From e94e7d124a4d1b862cb0e05b75b9551ce7ab1f3c Mon Sep 17 00:00:00 2001 From: Ragnor Comerford Date: Mon, 1 Jun 2026 13:11:36 +0200 Subject: [PATCH 01/11] fix(bootstrap): pin RustFS to beta.3 + allow insecure default creds (#136) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit `local-rustfs-bootstrap.sh` defaulted RUSTFS_IMAGE to the floating `rustfs/rustfs:latest`, which resolved to 1.0.0-beta.4 (2026-05-21). beta.4 added a credentials-policy check that refuses to start when the access/secret keys are values it treats as "default" (rustfsadmin/rustfsadmin, the script's defaults) — so a fresh bootstrap broke at RustFS startup. Pin the default to 1.0.0-beta.3 to match CI (.github/workflows/ci.yml) and the v0.5.0 release notes, and additionally pass RUSTFS_ALLOW_INSECURE_DEFAULT_CREDENTIALS=true so the script stays forward-compatible if RUSTFS_IMAGE is overridden to beta.4+. Co-authored-by: Ragnor Comerford --- scripts/local-rustfs-bootstrap.sh | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) diff --git a/scripts/local-rustfs-bootstrap.sh b/scripts/local-rustfs-bootstrap.sh index 29427de..c4fdcbe 100755 --- a/scripts/local-rustfs-bootstrap.sh +++ b/scripts/local-rustfs-bootstrap.sh @@ -6,7 +6,14 @@ SOURCE_REF="${SOURCE_REF:-main}" RELEASE_CHANNEL="${RELEASE_CHANNEL:-edge}" WORKDIR="${WORKDIR:-$PWD/.omnigraph-rustfs-demo}" RUSTFS_CONTAINER_NAME="${RUSTFS_CONTAINER_NAME:-omnigraph-rustfs-demo}" -RUSTFS_IMAGE="${RUSTFS_IMAGE:-rustfs/rustfs:latest}" +# Pinned to 1.0.0-beta.3 (2026-05-14) — the last known-good tag, matching CI +# (.github/workflows/ci.yml). `rustfs/rustfs:latest` (1.0.0-beta.4, 2026-05-21) +# added a credentials-policy check that refuses to start when the access/secret +# keys are values it considers "default" (rustfsadmin/rustfsadmin here). This +# script still works on beta.4+ because it passes +# RUSTFS_ALLOW_INSECURE_DEFAULT_CREDENTIALS=true below — so overriding +# RUSTFS_IMAGE to a newer tag is safe. +RUSTFS_IMAGE="${RUSTFS_IMAGE:-rustfs/rustfs:1.0.0-beta.3}" RUSTFS_DATA_DIR="${RUSTFS_DATA_DIR:-$WORKDIR/rustfs-data}" BUCKET="${BUCKET:-omnigraph-local}" PREFIX="${PREFIX:-repos/context}" @@ -265,6 +272,7 @@ start_rustfs() { -v "$RUSTFS_DATA_DIR:/data" \ -e RUSTFS_ACCESS_KEY="$AWS_ACCESS_KEY_ID" \ -e RUSTFS_SECRET_KEY="$AWS_SECRET_ACCESS_KEY" \ + -e RUSTFS_ALLOW_INSECURE_DEFAULT_CREDENTIALS=true \ "$RUSTFS_IMAGE" \ /data >/dev/null } From 353c0c876ab11c0ed4c087fcb32b894ec184df6d Mon Sep 17 00:00:00 2001 From: Ragnor Comerford Date: Mon, 1 Jun 2026 13:28:38 +0200 Subject: [PATCH 02/11] fix(branch): make branch delete correct under partial failure (#137) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * test(lance): pin force_delete_branch surface guard Pin the Lance 6.0.1 force_delete_branch behavior the branch-delete single-authority redesign relies on: plain delete_branch errors on a missing ref, force_delete_branch removes an existing forked branch, and the local-store quirk where force_delete on a fully-absent branch still errors (worked around by the upcoming TableStore::force_delete_branch). Re-pin the docs/dev/lance.md alignment stanza (9 guards; 4 runtime). * feat(storage): add force branch-delete to TableStore + CommitGraph Add TableStore::force_delete_branch and CommitGraph::force_delete_branch (idempotent: tolerate an already-absent branch via Lance RefNotFound / NotFound), plus CommitGraph::list_branches for the cleanup reconciler to diff against the manifest authority. RefConflict (referencing descendants) is still surfaced. Unused until the branch-delete rewire. * test(maintenance): red — cleanup reconciles orphaned branch forks Forge a Lance branch on the Person table that the manifest never references (a zombie fork from an incomplete prior delete) and assert cleanup reclaims it while leaving main intact. Fails today: cleanup does not yet reconcile orphaned forks. Goes green with the next commit. * fix(maintenance): reconcile orphaned branch forks in cleanup Add reconcile_orphaned_branches: force_delete_branch every per-table and commit-graph Lance branch absent from the manifest branch set (the authority), children-before-parents. Folded into cleanup_all_tables, runs before version GC. Idempotent and authority-derived; no-ops once nothing is orphaned, and would harmlessly find nothing if a future Lance atomic multi-dataset branch op prevented orphans. Adds TableStore::list_branches and exposes graph_commits_uri(pub crate). Turns the maintenance red test green. * test(failpoints): red — branch_delete partial failure converges Add the branch_delete.before_table_cleanup failpoint hook (inert without the feature) and a regression test: a cleanup-step failure after the manifest authority flip must leave branch_delete returning Ok, the branch gone, the orphan stranded, then reclaimed by cleanup, and the name reusable. Fails today: cleanup_deleted_branch_tables propagates the error as a hard failure. Goes green with the next commit. * fix(branch): best-effort fork reclaim after the manifest flip Make branch_delete treat per-table forks and the commit-graph branch as derived state reclaimed best-effort with force_delete_branch after the manifest authority flip. A reclaim failure (transient error, or the branch_delete.before_table_cleanup failpoint) is logged via tracing::warn and swallowed: the branch is already gone and the cleanup reconciler converges the orphan. cleanup_deleted_branch_tables no longer returns an error or blocks the call. Turns the partial-failure recovery test green. * test(failpoints): red — recreate over orphaned fork is actionable After a partial-failure delete leaves a fork orphaned, recreating the branch name and writing to the previously-forked table before cleanup runs currently surfaces the opaque ExpectedVersionMismatch ("stale view ... expected manifest table version N"). Assert instead a clear error pointing the user at cleanup. Goes green with the next commit. * fix(branch): actionable orphan-collision error in fork_branch_from_state When a fork's create_branch collides with an existing target ref, reuse it only if its head matches source_version (a legitimate concurrent first-write). A version mismatch means a zombie fork from an incomplete prior delete: return a manifest_conflict pointing the user at `omnigraph cleanup`, instead of the opaque ExpectedVersionMismatch. Turns the recreate-over-orphan red test green. * docs(invariants): single-authority branch-lifecycle + Lance forward-compat Record branch delete in the Current Truth Matrix: manifest is the single authority flipped atomically first, per-table forks + commit-graph branch are derived state reclaimed best-effort with the cleanup reconciler as backstop, and reusing a name whose reclaim failed surfaces an actionable error. Note the reconciler is authority-derived and degrades to a no-op under a future Lance atomic multi-dataset branch op, the same shape as invariant 7. * test(failpoints): red — cleanup isolates a single-table failure Add the cleanup.table_gc failpoint hook (inert without the feature) and an error: Option field on TableCleanupStats (mechanical, always None for now). Regression test: a one-shot version-GC failure for one table must not abort the whole cleanup — assert cleanup still succeeds, surfaces the failure per-table in stats, and the independent reconcile pass still reclaimed an orphan. Fails today: the version-GC collect aborts on the first table error. Goes green with the next commit. * fix(maintenance): fault-isolate cleanup per table Make the cleanup sweep do as much as it can and converge on re-run instead of aborting wholesale on one table's transient error (invariant 13). The version-GC loop now records a per-table failure on its stats row (error: Some) and logs it rather than collecting into a Result that aborts; reconcile_orphaned_branches isolates per-table and commit-graph failures into BranchReconcileStats.failures. The CLI reports any failed tables and tells the user to rerun cleanup. Addresses the Devin review finding. Turns the single-table-failure test green. * test(failpoints): red — branch_create heals commit-graph zombie + is atomic Add the branch_delete.before_commit_graph_reclaim failpoint hook and two regression tests: (a) recreating a name whose delete left a commit-graph zombie must succeed (today it dies on Lance's internal Clone error), and (b) branch_create must roll back the manifest branch when the derived commit-graph branch fails (today it leaves the manifest branch created while returning Err). Both fail now; green with the next commit. The existing branch_create_failpoint_triggers test still passes. * fix(branch): make branch_create atomic + heal commit-graph zombie branch_create now flips the manifest authority first, then creates the derived commit-graph branch in create_commit_graph_branch, force-dropping any orphaned commit-graph ref left by an incomplete prior delete (the manifest branch is fresh, so a same-named commit-graph branch is provably a zombie). If commit-graph creation fails, the manifest branch is rolled back so the name never half-exists. Addresses the Codex review finding. Turns the two branch_create red tests green; existing tests unaffected. * test(failpoints): red — fork collision misclassifies live concurrent fork Add the fork.before_classify failpoint hook and a concurrency test: when a concurrent first-write legitimately wins the fork race, the loser must get a retryable refresh-and-retry, not the misleading run-cleanup orphan error. Today the version-comparison misclassifies the live fork as an orphan (the Cursor finding). Goes green with the next commit. * fix(branch): manifest-arbitrated fork-collision classification Classify a fork collision by the manifest authority instead of comparing Lance branch versions. Before forking, open_owned_dataset_for_branch_write re-reads the live manifest: if the table is already forked on the active branch, a concurrent first-write won and the loser gets a retryable refresh-and-retry (not a misleading orphan error). fork_branch_from_state no longer guesses from versions — a create collision past that check is an orphan, so it returns the actionable cleanup error. Addresses the Cursor finding; turns the live-concurrent-fork test green, zombie path unchanged. * test(failpoints): close branch-lifecycle test gaps Three coverage additions for the branch-delete work (behavior already correct; these lock it in and catch regressions): - cleanup_isolates_reconcile_failure: inject a force-delete failure into the reconcile loop (new cleanup.reconcile_fork hook) and assert the sweep continues + converges on re-run. Directly covers the reconcile loop the Devin finding was about (previously only version-GC was). - cleanup_reclaims_orphaned_commit_graph_branch: forge a commit-graph orphan via the delete reclaim failpoint and assert cleanup's reconcile_commit_graph_orphans drops it (previously untested). - fork_collision_with_live_concurrent_fork_is_retryable: replace the fixed 300ms sleep with a deterministic readiness signal (cfg_callback + compare_exchange atomics) so the two-writer ordering can't flake. Full failpoints suite 31/0. --- crates/omnigraph-cli/src/main.rs | 15 +- crates/omnigraph/src/db/commit_graph.rs | 33 +- crates/omnigraph/src/db/graph_coordinator.rs | 70 ++- crates/omnigraph/src/db/omnigraph.rs | 38 +- crates/omnigraph/src/db/omnigraph/optimize.rs | 242 +++++++++- .../omnigraph/src/db/omnigraph/table_ops.rs | 16 + crates/omnigraph/src/table_store.rs | 66 ++- crates/omnigraph/tests/failpoints.rs | 446 ++++++++++++++++++ .../omnigraph/tests/lance_surface_guards.rs | 48 ++ crates/omnigraph/tests/maintenance.rs | 69 +++ docs/dev/invariants.md | 8 + docs/dev/lance.md | 3 +- docs/user/branches-commits.md | 6 +- docs/user/maintenance.md | 8 +- 14 files changed, 1004 insertions(+), 64 deletions(-) diff --git a/crates/omnigraph-cli/src/main.rs b/crates/omnigraph-cli/src/main.rs index b7e3041..d98c302 100644 --- a/crates/omnigraph-cli/src/main.rs +++ b/crates/omnigraph-cli/src/main.rs @@ -2699,20 +2699,33 @@ async fn main() -> Result<()> { "table_key": s.table_key, "bytes_removed": s.bytes_removed, "old_versions_removed": s.old_versions_removed, + "error": s.error, })).collect::>(), }); print_json(&value)?; } else { let total_bytes: u64 = stats.iter().map(|s| s.bytes_removed).sum(); let total_versions: u64 = stats.iter().map(|s| s.old_versions_removed).sum(); + let failed: Vec<&str> = stats + .iter() + .filter(|s| s.error.is_some()) + .map(|s| s.table_key.as_str()) + .collect(); println!( "cleanup {} ({}) — removed {} versions ({} bytes) across {} tables", uri, policy_desc, total_versions, total_bytes, - stats.len() + stats.len() - failed.len() ); + if !failed.is_empty() { + println!( + " {} table(s) failed and will be retried on the next cleanup: {}", + failed.len(), + failed.join(", ") + ); + } } } Command::Graphs { command } => match command { diff --git a/crates/omnigraph/src/db/commit_graph.rs b/crates/omnigraph/src/db/commit_graph.rs index 565bd69..9531a64 100644 --- a/crates/omnigraph/src/db/commit_graph.rs +++ b/crates/omnigraph/src/db/commit_graph.rs @@ -169,6 +169,37 @@ impl CommitGraph { self.refresh().await } + /// Idempotently drop the commit-graph branch `name`, tolerating an + /// already-absent branch (see [`TableStore::force_delete_branch`] for the + /// same semantics). Used by the best-effort reclaim in `branch_delete` and + /// the `cleanup` orphan reconciler. `RefConflict` (referencing descendants) + /// is still surfaced. + pub async fn force_delete_branch(&mut self, name: &str) -> Result<()> { + let mut ds = Dataset::open(&graph_commits_uri(&self.root_uri)) + .await + .map_err(|e| OmniError::Lance(e.to_string()))?; + match ds.force_delete_branch(name).await { + Ok(()) => {} + Err(lance::Error::RefNotFound { .. }) | Err(lance::Error::NotFound { .. }) => {} + Err(e) => return Err(OmniError::Lance(e.to_string())), + } + self.refresh().await + } + + /// List the named branches present on the commit-graph dataset. The + /// `cleanup` reconciler diffs this against the manifest branch set to find + /// orphaned commit-graph branches to reclaim. + pub async fn list_branches(&self) -> Result> { + let ds = Dataset::open(&graph_commits_uri(&self.root_uri)) + .await + .map_err(|e| OmniError::Lance(e.to_string()))?; + let branches = ds + .list_branches() + .await + .map_err(|e| OmniError::Lance(e.to_string()))?; + Ok(branches.into_keys().collect()) + } + pub async fn append_commit( &mut self, manifest_branch: Option<&str>, @@ -345,7 +376,7 @@ impl CommitGraph { } } -fn graph_commits_uri(root_uri: &str) -> String { +pub(crate) fn graph_commits_uri(root_uri: &str) -> String { format!("{}/{}", root_uri.trim_end_matches('/'), GRAPH_COMMITS_DIR) } diff --git a/crates/omnigraph/src/db/graph_coordinator.rs b/crates/omnigraph/src/db/graph_coordinator.rs index a721036..dfe2767 100644 --- a/crates/omnigraph/src/db/graph_coordinator.rs +++ b/crates/omnigraph/src/db/graph_coordinator.rs @@ -211,14 +211,47 @@ impl GraphCoordinator { let branch = normalize_branch_name(name)? .ok_or_else(|| OmniError::manifest("cannot create branch 'main'".to_string()))?; self.ensure_commit_graph_initialized().await?; + + // Manifest authority flip first. self.manifest.create_branch(&branch).await?; - failpoints::maybe_fail("branch_create.after_manifest_branch_create")?; - if let Some(commit_graph) = &mut self.commit_graph { - commit_graph.create_branch(&branch).await?; + + // Derived commit-graph branch. If anything after the authority flip + // fails, roll back the manifest branch so the branch never half-exists + // (a manifest branch with no commit-graph branch breaks the next write). + if let Err(err) = self.create_commit_graph_branch(&branch).await { + if let Err(rollback_err) = self.manifest.delete_branch(&branch).await { + tracing::warn!( + target: "omnigraph::branch_create", + branch = %branch, + error = %rollback_err, + "rollback of manifest branch failed after commit-graph create failure", + ); + } + return Err(err); } Ok(()) } + /// Create the derived commit-graph branch for `branch`, healing a zombie ref + /// left by an incomplete prior delete. The manifest branch was just created + /// fresh, so any existing commit-graph branch with this name is provably + /// orphaned and is force-dropped before recreating. + async fn create_commit_graph_branch(&mut self, branch: &str) -> Result<()> { + failpoints::maybe_fail("branch_create.after_manifest_branch_create")?; + let Some(commit_graph) = &mut self.commit_graph else { + return Ok(()); + }; + if commit_graph + .list_branches() + .await? + .iter() + .any(|existing| existing == branch) + { + commit_graph.force_delete_branch(branch).await?; + } + commit_graph.create_branch(branch).await + } + pub async fn branch_delete(&mut self, name: &str) -> Result<()> { let branch = normalize_branch_name(name)? .ok_or_else(|| OmniError::manifest("cannot delete branch 'main'".to_string()))?; @@ -229,20 +262,43 @@ impl GraphCoordinator { ))); } + // Manifest authority flip — the single atomic op that makes the branch + // cease to exist. Must succeed; everything after is derived state + // reclaimed best-effort. self.manifest.delete_branch(&branch).await?; + // Commit-graph branch is derived state. Reclaim best-effort with the + // idempotent force variant: a failure here (or a missing dataset) is + // reconciled by `cleanup` and must not fail the delete after the + // authority already flipped. + if let Err(err) = self.reclaim_commit_graph_branch(&branch).await { + tracing::warn!( + target: "omnigraph::branch_delete::cleanup", + branch = %branch, + error = %err, + "best-effort commit-graph branch reclaim failed; cleanup will reconcile", + ); + } + + Ok(()) + } + + /// Best-effort, idempotent reclaim of the commit-graph branch `branch`. + /// Tolerates an absent commit-graph dataset (a graph that never committed). + async fn reclaim_commit_graph_branch(&mut self, branch: &str) -> Result<()> { + failpoints::maybe_fail("branch_delete.before_commit_graph_reclaim")?; if let Some(commit_graph) = &mut self.commit_graph { - commit_graph.delete_branch(&branch).await?; + commit_graph.force_delete_branch(branch).await } else if self .storage .exists(&graph_commits_uri(self.root_uri())) .await? { let mut commit_graph = CommitGraph::open(self.root_uri()).await?; - commit_graph.delete_branch(&branch).await?; + commit_graph.force_delete_branch(branch).await + } else { + Ok(()) } - - Ok(()) } pub async fn snapshot_at_version(&self, version: u64) -> Result { diff --git a/crates/omnigraph/src/db/omnigraph.rs b/crates/omnigraph/src/db/omnigraph.rs index 5c92ac3..eb58623 100644 --- a/crates/omnigraph/src/db/omnigraph.rs +++ b/crates/omnigraph/src/db/omnigraph.rs @@ -1058,11 +1058,14 @@ impl Omnigraph { Ok(()) } - async fn cleanup_deleted_branch_tables( - &self, - branch: &str, - owned_tables: &[(String, String)], - ) -> Result<()> { + /// Best-effort reclaim of the per-table Lance forks a just-deleted branch + /// owned. Runs AFTER the manifest authority flip, so the branch is already + /// gone and these forks are unreachable orphans. A failure here (transient + /// object-store error, the `branch_delete.before_table_cleanup` failpoint) + /// is logged and swallowed: the `cleanup` reconciler is the guaranteed + /// backstop that converges any leftover orphan. Uses `force_delete_branch` + /// so a partially-reclaimed retry is idempotent. + async fn cleanup_deleted_branch_tables(&self, branch: &str, owned_tables: &[(String, String)]) { let mut seen_paths = HashSet::new(); let mut cleanup_targets = owned_tables .iter() @@ -1073,15 +1076,21 @@ impl Omnigraph { for (table_key, table_path) in cleanup_targets { let dataset_uri = self.table_store.dataset_uri(&table_path); - if let Err(err) = self.table_store.delete_branch(&dataset_uri, branch).await { - return Err(OmniError::manifest_internal(format!( - "branch '{}' was deleted but cleanup failed for {}: {}", - branch, table_key, err - ))); + let outcome = match crate::failpoints::maybe_fail("branch_delete.before_table_cleanup") + { + Ok(()) => self.table_store.force_delete_branch(&dataset_uri, branch).await, + Err(injected) => Err(injected), + }; + if let Err(err) = outcome { + tracing::warn!( + target: "omnigraph::branch_delete::cleanup", + branch = %branch, + table = %table_key, + error = %err, + "best-effort fork reclaim failed; cleanup will reconcile the orphan", + ); } } - - Ok(()) } async fn delete_branch_storage_only(&self, branch: &str) -> Result<()> { @@ -1105,9 +1114,12 @@ impl Omnigraph { .map(|entry| (entry.table_key.clone(), entry.table_path.clone())) .collect::>(); + // Authority flip (+ best-effort commit-graph reclaim) — must succeed. self.coordinator.write().await.branch_delete(branch).await?; + // Best-effort per-table fork reclaim; cleanup reconciles any leftover. self.cleanup_deleted_branch_tables(branch, &owned_tables) - .await + .await; + Ok(()) } pub(crate) fn normalize_branch_name(branch: &str) -> Result> { diff --git a/crates/omnigraph/src/db/omnigraph/optimize.rs b/crates/omnigraph/src/db/omnigraph/optimize.rs index e158dc7..c703836 100644 --- a/crates/omnigraph/src/db/omnigraph/optimize.rs +++ b/crates/omnigraph/src/db/omnigraph/optimize.rs @@ -64,12 +64,15 @@ pub struct TableOptimizeStats { pub committed: bool, } -/// Per-table outcome of `cleanup_all_tables`. +/// Per-table outcome of `cleanup_all_tables`. `error` is `Some` when this +/// table's version GC failed; cleanup is fault-isolated per table, so a single +/// table's failure is recorded here rather than aborting the whole sweep. #[derive(Debug, Clone)] pub struct TableCleanupStats { pub table_key: String, pub bytes_removed: u64, pub old_versions_removed: u64, + pub error: Option, } /// Run Lance `compact_files` on every node + edge table on `main`. @@ -138,6 +141,26 @@ pub async fn cleanup_all_tables( db.ensure_schema_state_valid().await?; db.ensure_schema_apply_idle("cleanup").await?; + // Reclaim orphaned branch forks (from an incomplete prior `branch_delete`) + // before version GC. Authority-derived and idempotent; the eager + // best-effort reclaim in `branch_delete` covers the common case, this is + // the guaranteed backstop. Logged for observability. + let reconciled = reconcile_orphaned_branches(db).await?; + if !reconciled.reclaimed.is_empty() { + tracing::info!( + count = reconciled.reclaimed.len(), + reclaimed = ?reconciled.reclaimed, + "cleanup reconciled orphaned branch forks" + ); + } + if !reconciled.failures.is_empty() { + tracing::warn!( + count = reconciled.failures.len(), + failures = ?reconciled.failures, + "cleanup could not reconcile some orphaned forks; will retry next cleanup" + ); + } + let before_timestamp = options.older_than.map(|d| Utc::now() - d); let keep_versions = options.keep_versions; @@ -160,36 +183,205 @@ pub async fn cleanup_all_tables( let concurrency = maint_concurrency().min(table_tasks.len()).max(1); let table_store = &db.table_store; - let results: Vec> = futures::stream::iter(table_tasks.into_iter()) + // Fault-isolated per table: a single table's GC failure is recorded on its + // stats row (`error: Some`) and logged, never aborting the healthy tables. + // cleanup is the convergence backstop, so it must do as much as it can and + // converge on re-run rather than fail wholesale (invariant 13). + let results: Vec = futures::stream::iter(table_tasks.into_iter()) .map(|(table_key, full_path)| async move { - let ds = table_store - .open_dataset_head_for_write(&table_key, &full_path, None) - .await?; - let before_version = keep_versions - .map(|n| ds.version().version.saturating_sub(n as u64)) - .filter(|v| *v > 0); - let policy = CleanupPolicy { - before_timestamp, - before_version, - delete_unverified: false, - error_if_tagged_old_versions: false, - clean_referenced_branches: false, - delete_rate_limit: None, - }; - let removed: RemovalStats = lance::dataset::cleanup::cleanup_old_versions(&ds, policy) - .await - .map_err(|e| OmniError::Lance(e.to_string()))?; - Ok(TableCleanupStats { - table_key, - bytes_removed: removed.bytes_removed, - old_versions_removed: removed.old_versions, - }) + let outcome: Result = async { + crate::failpoints::maybe_fail("cleanup.table_gc")?; + let ds = table_store + .open_dataset_head_for_write(&table_key, &full_path, None) + .await?; + let before_version = keep_versions + .map(|n| ds.version().version.saturating_sub(n as u64)) + .filter(|v| *v > 0); + let policy = CleanupPolicy { + before_timestamp, + before_version, + delete_unverified: false, + error_if_tagged_old_versions: false, + clean_referenced_branches: false, + delete_rate_limit: None, + }; + lance::dataset::cleanup::cleanup_old_versions(&ds, policy) + .await + .map_err(|e| OmniError::Lance(e.to_string())) + } + .await; + match outcome { + Ok(removed) => TableCleanupStats { + table_key, + bytes_removed: removed.bytes_removed, + old_versions_removed: removed.old_versions, + error: None, + }, + Err(err) => { + tracing::warn!( + target: "omnigraph::cleanup", + table = %table_key, + error = %err, + "version GC failed for table; other tables unaffected", + ); + TableCleanupStats { + table_key, + bytes_removed: 0, + old_versions_removed: 0, + error: Some(err.to_string()), + } + } + } }) .buffer_unordered(concurrency) .collect() .await; - results.into_iter().collect() + Ok(results) +} + +/// Outcome of [`reconcile_orphaned_branches`]: the `(owner, branch)` pairs +/// reclaimed and the `(owner, error)` pairs that failed, where `owner` is a +/// table key (e.g. `node:Person`) or `"_graph_commits"`. Per-owner failures are +/// isolated and recorded here, not propagated — the next reconcile converges. +#[derive(Debug, Clone, Default)] +pub struct BranchReconcileStats { + pub reclaimed: Vec<(String, String)>, + pub failures: Vec<(String, String)>, +} + +/// Drop every per-table and commit-graph Lance branch that the manifest no +/// longer references. +/// +/// Orphaned forks arise when a `branch_delete` flips the manifest authority +/// (atomic) but a downstream best-effort reclaim does not complete. They are +/// unreachable through any snapshot — no manifest entry can name them — yet +/// they pin their `tree/{branch}/` storage and can block reusing the branch +/// name. This is the guaranteed convergence backstop: it is idempotent and +/// derived purely from the manifest authority, so it no-ops once everything is +/// reconciled, and it would harmlessly find nothing if a future Lance atomic +/// multi-dataset branch op prevented orphans from forming. +/// +/// The keep-set is the full (unfiltered) manifest branch list, so system +/// branches' forks are never reclaimed; `main`/default is not a named Lance +/// branch and so is never a candidate. Referencing children are dropped before +/// parents (Lance refuses to delete a referenced parent) by ordering longest +/// branch names first. +pub async fn reconcile_orphaned_branches(db: &Omnigraph) -> Result { + use std::collections::HashSet; + + let keep: HashSet = db + .coordinator + .read() + .await + .all_branches() + .await? + .into_iter() + .collect(); + + let resolved = db.resolved_branch_target(None).await?; + let snapshot = resolved.snapshot; + let table_targets: Vec<(String, String)> = all_table_keys(&db.catalog()) + .into_iter() + .filter_map(|table_key| { + let entry = snapshot.entry(&table_key)?; + let full_path = format!("{}/{}", db.root_uri, entry.table_path); + Some((table_key, full_path)) + }) + .collect(); + + let mut stats = BranchReconcileStats::default(); + + // Per-table fault isolation: one table's transient failure is recorded and + // logged, never aborting the rest of the sweep. + for (table_key, full_path) in table_targets { + let listed = match db.table_store.list_branches(&full_path).await { + Ok(listed) => listed, + Err(err) => { + tracing::warn!( + target: "omnigraph::cleanup", + table = %table_key, + error = %err, + "listing branches failed during reconcile; skipping table", + ); + stats.failures.push((table_key.clone(), err.to_string())); + continue; + } + }; + for branch in orphan_branches(listed, &keep) { + let outcome = match crate::failpoints::maybe_fail("cleanup.reconcile_fork") { + Ok(()) => db.table_store.force_delete_branch(&full_path, &branch).await, + Err(injected) => Err(injected), + }; + match outcome { + Ok(()) => stats.reclaimed.push((table_key.clone(), branch)), + Err(err) => { + tracing::warn!( + target: "omnigraph::cleanup", + table = %table_key, + branch = %branch, + error = %err, + "reclaiming orphaned fork failed; will retry next cleanup", + ); + stats.failures.push((table_key.clone(), err.to_string())); + } + } + } + } + + // Commit-graph orphans (best-effort: the dataset may not exist on a graph + // that has never committed; any failure is isolated and retried next time). + if let Err(err) = reconcile_commit_graph_orphans(db, &keep, &mut stats).await { + tracing::warn!( + target: "omnigraph::cleanup", + error = %err, + "commit-graph orphan reconcile failed; will retry next cleanup", + ); + stats.failures.push(("_graph_commits".to_string(), err.to_string())); + } + + Ok(stats) +} + +/// Commit-graph half of [`reconcile_orphaned_branches`], split out so its +/// errors can be isolated. Returns `Ok` when the commit-graph dataset is absent. +async fn reconcile_commit_graph_orphans( + db: &Omnigraph, + keep: &std::collections::HashSet, + stats: &mut BranchReconcileStats, +) -> Result<()> { + let commits_uri = crate::db::commit_graph::graph_commits_uri(db.root_uri()); + if !db.storage_adapter().exists(&commits_uri).await? { + return Ok(()); + } + let mut commit_graph = crate::db::commit_graph::CommitGraph::open(db.root_uri()).await?; + for branch in orphan_branches(commit_graph.list_branches().await?, keep) { + match commit_graph.force_delete_branch(&branch).await { + Ok(()) => stats.reclaimed.push(("_graph_commits".to_string(), branch)), + Err(err) => { + tracing::warn!( + target: "omnigraph::cleanup", + branch = %branch, + error = %err, + "reclaiming orphaned commit-graph branch failed; will retry next cleanup", + ); + stats.failures.push(("_graph_commits".to_string(), err.to_string())); + } + } + } + Ok(()) +} + +/// Filter `present` Lance branches down to those absent from the manifest +/// `keep` set, ordered children-before-parents (longest name first) so Lance's +/// referenced-parent `RefConflict` cannot block reclamation. +fn orphan_branches(present: Vec, keep: &std::collections::HashSet) -> Vec { + let mut orphans: Vec = present + .into_iter() + .filter(|branch| !keep.contains(branch)) + .collect(); + orphans.sort_by(|a, b| b.len().cmp(&a.len()).then_with(|| a.cmp(b))); + orphans } fn all_table_keys(catalog: &omnigraph_compiler::catalog::Catalog) -> Vec { diff --git a/crates/omnigraph/src/db/omnigraph/table_ops.rs b/crates/omnigraph/src/db/omnigraph/table_ops.rs index 0e89c45..3ed9c43 100644 --- a/crates/omnigraph/src/db/omnigraph/table_ops.rs +++ b/crates/omnigraph/src/db/omnigraph/table_ops.rs @@ -483,6 +483,22 @@ pub(super) async fn open_owned_dataset_for_branch_write( Ok((ds, Some(active_branch.to_string()))) } source_branch => { + crate::failpoints::maybe_fail("fork.before_classify")?; + // Authority check before forking: re-read the live manifest. If this + // table is already forked on active_branch, a concurrent first-write + // won the race and our snapshot is stale — that is a retryable + // conflict, not an orphan. (A zombie fork is never in the manifest, + // so this only fires for a live concurrent fork.) + let live = db.snapshot_for_branch(Some(active_branch)).await?; + if let Some(entry) = live.entry(table_key) { + if entry.table_branch.as_deref() == Some(active_branch) { + return Err(OmniError::manifest_expected_version_mismatch( + table_key, + entry_version, + entry.table_version, + )); + } + } fork_dataset_from_entry_state( db, table_key, diff --git a/crates/omnigraph/src/table_store.rs b/crates/omnigraph/src/table_store.rs index 46b15b0..10123b0 100644 --- a/crates/omnigraph/src/table_store.rs +++ b/crates/omnigraph/src/table_store.rs @@ -177,6 +177,45 @@ impl TableStore { .map_err(|e| OmniError::Lance(e.to_string())) } + /// List the named Lance branches present on the dataset at `dataset_uri`. + /// The `cleanup` orphan reconciler diffs this against the manifest branch + /// set to find orphaned per-table forks. `main`/default is not a named + /// branch and never appears here. + pub async fn list_branches(&self, dataset_uri: &str) -> Result> { + let ds = Dataset::open(dataset_uri) + .await + .map_err(|e| OmniError::Lance(e.to_string()))?; + let branches = ds + .list_branches() + .await + .map_err(|e| OmniError::Lance(e.to_string()))?; + Ok(branches.into_keys().collect()) + } + + /// Idempotently drop `branch` from the dataset at `dataset_uri`. + /// + /// Unlike [`delete_branch`](Self::delete_branch), this tolerates an + /// already-absent branch — both a missing contents ref (Lance's + /// `force_delete_branch` handles that) and a missing `tree/{branch}/` + /// directory (the local-store `NotFound` quirk pinned by + /// `lance_surface_guards::force_delete_branch_semantics`). Safe to call on a + /// possibly-orphaned or already-reclaimed fork. + /// + /// A branch that still has referencing descendants (`RefConflict`) is NOT + /// tolerated: that is a real ordering error and surfaces as `OmniError::Lance`. + /// Used by the eager best-effort reclaim in `cleanup_deleted_branch_tables` + /// and the `cleanup` orphan reconciler. + pub async fn force_delete_branch(&self, dataset_uri: &str, branch: &str) -> Result<()> { + let mut ds = Dataset::open(dataset_uri) + .await + .map_err(|e| OmniError::Lance(e.to_string()))?; + match ds.force_delete_branch(branch).await { + Ok(()) => Ok(()), + Err(lance::Error::RefNotFound { .. }) | Err(lance::Error::NotFound { .. }) => Ok(()), + Err(e) => Err(OmniError::Lance(e.to_string())), + } + } + pub async fn open_dataset_at_state( &self, table_path: &str, @@ -243,21 +282,24 @@ impl TableStore { .map_err(|e| OmniError::Lance(e.to_string()))?; self.ensure_expected_version(&source_ds, table_key, source_version)?; - match source_ds + if source_ds .create_branch(target_branch, source_version, None) .await + .is_err() { - Ok(_) => {} - Err(create_err) => match self - .open_dataset_head(dataset_uri, Some(target_branch)) - .await - { - Ok(ds) => { - self.ensure_expected_version(&ds, table_key, source_version)?; - return Ok(ds); - } - Err(_) => return Err(OmniError::Lance(create_err.to_string())), - }, + // The target branch ref already exists. The caller + // (`open_owned_dataset_for_branch_write`) re-reads the live manifest + // before forking and returns a retryable error when a concurrent + // writer legitimately holds the fork, so reaching here means the + // manifest does NOT reference this fork: it is an orphan from an + // incomplete prior `branch_delete`. Surface the actionable cleanup + // error rather than guessing from Lance branch versions. + return Err(OmniError::manifest_conflict(format!( + "branch '{}' has orphaned table state for '{}' from an incomplete \ + prior delete; run `omnigraph cleanup` to reclaim it before reusing \ + this branch name", + target_branch, table_key + ))); } let ds = self diff --git a/crates/omnigraph/tests/failpoints.rs b/crates/omnigraph/tests/failpoints.rs index 5ea71c5..149c63a 100644 --- a/crates/omnigraph/tests/failpoints.rs +++ b/crates/omnigraph/tests/failpoints.rs @@ -41,6 +41,452 @@ async fn branch_create_failpoint_triggers() { ); } +// Branch delete flips the manifest authority first, then reclaims the per-table +// forks best-effort. A failure during that reclaim (here, the +// `branch_delete.before_table_cleanup` failpoint, standing in for a transient +// object-store error) must NOT fail the call: the branch is already gone, and +// `cleanup` reconciles the stranded fork. The branch name is reusable after. +#[tokio::test] +async fn branch_delete_partial_failure_converges_via_cleanup() { + let _scenario = FailScenario::setup(); + let dir = tempfile::tempdir().unwrap(); + let uri = dir.path().to_str().unwrap().to_string(); + let mut main = helpers::init_and_load(&dir).await; + + main.branch_create("feature").await.unwrap(); + let mut feature = Omnigraph::open(&uri).await.unwrap(); + helpers::mutate_branch( + &mut feature, + "feature", + MUTATION_QUERIES, + "insert_person", + &mixed_params(&[("$name", "Eve")], &[("$age", 22)]), + ) + .await + .unwrap(); + drop(feature); + + let person_uri = node_table_uri(&uri, "Person"); + { + let ds = lance::Dataset::open(&person_uri).await.unwrap(); + assert!( + ds.list_branches().await.unwrap().contains_key("feature"), + "precondition: the owned table fork exists before delete" + ); + } + + // Inject a failure during per-table cleanup, AFTER the manifest authority + // flip. branch_delete must still succeed (best-effort reclaim). + { + let _fp = ScopedFailPoint::new("branch_delete.before_table_cleanup", "return"); + main.branch_delete("feature").await.expect( + "branch_delete is best-effort after the manifest flip: a cleanup-step \ + failure must not fail the call", + ); + } + + // Authority flipped: the branch is gone. + assert_eq!(main.branch_list().await.unwrap(), vec!["main".to_string()]); + + // The eager reclaim failed, so the orphan is stranded until cleanup. + { + let ds = lance::Dataset::open(&person_uri).await.unwrap(); + assert!( + ds.list_branches().await.unwrap().contains_key("feature"), + "failed eager reclaim should leave the orphan for cleanup to reconcile" + ); + } + + // cleanup converges: the orphan is reclaimed. + main.cleanup(omnigraph::db::CleanupPolicyOptions { + keep_versions: Some(1), + older_than: None, + }) + .await + .unwrap(); + { + let ds = lance::Dataset::open(&person_uri).await.unwrap(); + assert!( + !ds.list_branches().await.unwrap().contains_key("feature"), + "cleanup should reconcile the orphaned fork away" + ); + } + + // The name is reusable after cleanup reclaims the orphan. + main.branch_create("feature").await.unwrap(); + let mut feature2 = Omnigraph::open(&uri).await.unwrap(); + helpers::mutate_branch( + &mut feature2, + "feature", + MUTATION_QUERIES, + "insert_person", + &mixed_params(&[("$name", "Frank")], &[("$age", 41)]), + ) + .await + .unwrap(); +} + +// Reusing a branch name whose delete left an orphaned fork (before `cleanup` +// reconciles it) must fail with a clear, actionable error pointing at +// `cleanup`, not the opaque `ExpectedVersionMismatch` that leaks from the fork +// path. The recreate itself succeeds; the first write to the previously-forked +// table is where the stale orphan collides. +#[tokio::test] +async fn recreate_over_orphaned_fork_before_cleanup_is_actionable() { + let _scenario = FailScenario::setup(); + let dir = tempfile::tempdir().unwrap(); + let uri = dir.path().to_str().unwrap().to_string(); + let mut main = helpers::init_and_load(&dir).await; + + main.branch_create("feature").await.unwrap(); + let mut feature = Omnigraph::open(&uri).await.unwrap(); + helpers::mutate_branch( + &mut feature, + "feature", + MUTATION_QUERIES, + "insert_person", + &mixed_params(&[("$name", "Eve")], &[("$age", 22)]), + ) + .await + .unwrap(); + drop(feature); + + // Partial delete: leaves the Person fork orphaned (cleanup not yet run). + { + let _fp = ScopedFailPoint::new("branch_delete.before_table_cleanup", "return"); + main.branch_delete("feature").await.unwrap(); + } + + // Recreate the name and write to the previously-forked table WITHOUT a + // cleanup in between. + main.branch_create("feature").await.unwrap(); + let mut feature2 = Omnigraph::open(&uri).await.unwrap(); + let err = helpers::mutate_branch( + &mut feature2, + "feature", + MUTATION_QUERIES, + "insert_person", + &mixed_params(&[("$name", "Frank")], &[("$age", 41)]), + ) + .await + .expect_err("write should collide with the stale orphaned fork"); + + let msg = err.to_string(); + assert!( + msg.contains("cleanup") + && (msg.contains("orphan") || msg.contains("incomplete prior delete")), + "expected an actionable orphaned-fork error pointing at cleanup, got: {msg}" + ); + assert!( + !msg.contains("expected manifest table version"), + "should not surface the opaque ExpectedVersionMismatch, got: {msg}" + ); +} + +// cleanup is the guaranteed convergence backstop, so one table's transient +// failure must not abort the whole sweep. Inject a one-shot version-GC failure +// for a single table and assert: cleanup still succeeds, the failure is +// surfaced per-table in the returned stats, and the independent reconcile pass +// still reclaimed an orphan. +#[tokio::test] +async fn cleanup_isolates_single_table_failure() { + let _scenario = FailScenario::setup(); + let dir = tempfile::tempdir().unwrap(); + let uri = dir.path().to_str().unwrap().to_string(); + let mut db = helpers::init_and_load(&dir).await; + + // Forge an orphaned fork on the Person table (a reconcile target). + let person_uri = node_table_uri(&uri, "Person"); + { + let mut ds = lance::Dataset::open(&person_uri).await.unwrap(); + let base = ds.version().version; + ds.create_branch("ghost", base, None).await.unwrap(); + } + + // One table's version GC fails once; the sweep must isolate it. + let _fp = ScopedFailPoint::new("cleanup.table_gc", "1*return"); + let stats = db + .cleanup(omnigraph::db::CleanupPolicyOptions { + keep_versions: Some(1), + older_than: None, + }) + .await + .expect("a single table's GC failure must not abort cleanup"); + + let errored = stats.iter().filter(|s| s.error.is_some()).count(); + assert_eq!( + errored, 1, + "exactly one table's GC failure should be surfaced in stats, got {errored}" + ); + assert!( + stats.len() >= 4, + "every node+edge table should still appear in the stats" + ); + + // The reconcile pass is independent of the GC failure, so the orphan is gone. + { + let ds = lance::Dataset::open(&person_uri).await.unwrap(); + assert!( + !ds.list_branches().await.unwrap().contains_key("ghost"), + "reconcile should reclaim the orphan despite the GC failure" + ); + } +} + +// Companion to the version-GC isolation test, exercising the OTHER cleanup +// loop: a force-delete failure while reconciling one orphaned fork must be +// isolated (logged, not propagated) so the sweep continues, and a later +// cleanup converges. This is the loop the Devin finding was about. +#[tokio::test] +async fn cleanup_isolates_reconcile_failure() { + let _scenario = FailScenario::setup(); + let dir = tempfile::tempdir().unwrap(); + let uri = dir.path().to_str().unwrap().to_string(); + let mut db = helpers::init_and_load(&dir).await; + + // Forge an orphaned fork the reconcile pass will try to reclaim. + let person_uri = node_table_uri(&uri, "Person"); + { + let mut ds = lance::Dataset::open(&person_uri).await.unwrap(); + let base = ds.version().version; + ds.create_branch("ghost", base, None).await.unwrap(); + } + + // Inject a one-shot failure into the reconcile force-delete. The sweep must + // not abort. + { + let _fp = ScopedFailPoint::new("cleanup.reconcile_fork", "1*return"); + db.cleanup(omnigraph::db::CleanupPolicyOptions { + keep_versions: Some(1), + older_than: None, + }) + .await + .expect("a reconcile force-delete failure must not abort cleanup"); + } + // The blocked orphan is still present (the failure was isolated, not retried). + { + let ds = lance::Dataset::open(&person_uri).await.unwrap(); + assert!( + ds.list_branches().await.unwrap().contains_key("ghost"), + "the orphan whose reclaim was injected-to-fail should remain" + ); + } + // A second cleanup with no injected failure converges. + db.cleanup(omnigraph::db::CleanupPolicyOptions { + keep_versions: Some(1), + older_than: None, + }) + .await + .unwrap(); + { + let ds = lance::Dataset::open(&person_uri).await.unwrap(); + assert!( + !ds.list_branches().await.unwrap().contains_key("ghost"), + "the second cleanup should reconcile the orphan" + ); + } +} + +// The cleanup reconciler must reclaim orphaned commit-graph branches, not just +// per-table forks. A delete whose best-effort commit-graph reclaim fails leaves +// a commit-graph orphan; the next cleanup must drop it. +#[tokio::test] +async fn cleanup_reclaims_orphaned_commit_graph_branch() { + let _scenario = FailScenario::setup(); + let dir = tempfile::tempdir().unwrap(); + let uri = dir.path().to_str().unwrap().to_string(); + let mut db = helpers::init_and_load(&dir).await; + + db.branch_create("feature").await.unwrap(); + // Delete, failing the commit-graph reclaim → commit-graph "feature" orphan + // (manifest branch gone, commit-graph branch left behind). + { + let _fp = ScopedFailPoint::new("branch_delete.before_commit_graph_reclaim", "return"); + db.branch_delete("feature").await.unwrap(); + } + + let commits_uri = format!("{}/_graph_commits.lance", uri.trim_end_matches('/')); + { + let ds = lance::Dataset::open(&commits_uri).await.unwrap(); + assert!( + ds.list_branches().await.unwrap().contains_key("feature"), + "precondition: the commit-graph branch should be orphaned after the failed reclaim" + ); + } + + db.cleanup(omnigraph::db::CleanupPolicyOptions { + keep_versions: Some(1), + older_than: None, + }) + .await + .unwrap(); + + { + let ds = lance::Dataset::open(&commits_uri).await.unwrap(); + assert!( + !ds.list_branches().await.unwrap().contains_key("feature"), + "cleanup should reclaim the orphaned commit-graph branch" + ); + } +} + +// A branch_delete whose best-effort commit-graph reclaim fails leaves a +// commit-graph "zombie" branch. Recreating that name must heal the zombie and +// succeed (branch_create force-deletes a stale commit-graph ref since the +// manifest branch is created fresh), instead of dying on the leftover ref. +#[tokio::test] +async fn branch_create_recreates_over_commit_graph_zombie() { + let _scenario = FailScenario::setup(); + let dir = tempfile::tempdir().unwrap(); + let db = Omnigraph::init(dir.path().to_str().unwrap(), helpers::TEST_SCHEMA) + .await + .unwrap(); + + db.branch_create("feature").await.unwrap(); + { + // Fail the best-effort commit-graph reclaim → commit-graph "feature" + // zombie survives the delete (manifest authority still flips). + let _fp = ScopedFailPoint::new("branch_delete.before_commit_graph_reclaim", "return"); + db.branch_delete("feature").await.unwrap(); + } + assert_eq!(db.branch_list().await.unwrap(), vec!["main".to_string()]); + + db.branch_create("feature") + .await + .expect("branch_create should heal the zombie commit-graph branch and succeed"); + assert!( + db.branch_list() + .await + .unwrap() + .contains(&"feature".to_string()) + ); +} + +// branch_create is authority-then-derived: if the derived commit-graph branch +// cannot be created, the manifest branch (the authority) must be rolled back so +// the branch does not half-exist. The existing failpoint fires right after the +// manifest create, standing in for any post-authority failure. +#[tokio::test] +async fn branch_create_rolls_back_manifest_on_commit_graph_failure() { + let _scenario = FailScenario::setup(); + let dir = tempfile::tempdir().unwrap(); + let db = Omnigraph::init(dir.path().to_str().unwrap(), helpers::TEST_SCHEMA) + .await + .unwrap(); + + let err = { + let _fp = ScopedFailPoint::new("branch_create.after_manifest_branch_create", "return"); + db.branch_create("feature").await.unwrap_err() + }; + assert!( + !db.branch_list() + .await + .unwrap() + .contains(&"feature".to_string()), + "branch_create must roll back the manifest branch when the derived \ + commit-graph branch fails, got error: {err}" + ); +} + +// A fork collision must be classified by the manifest authority, not by Lance +// branch versions. When a concurrent first-write legitimately wins the fork +// race, the loser sees a version mismatch — but that is a stale snapshot, not +// an orphan, so it must be a retryable "refresh and retry", never a misleading +// "run cleanup". +// +// Ordering is made deterministic (no sleeps) via a callback at the fork point: +// `compare_exchange` lets only the FIRST arrival (writer A) record readiness and +// block until released; later arrivals (writer B) fall through. The test waits +// on the readiness flag, lets B win and commit the fork, then releases A. +static FORK_A_AT_POINT: std::sync::atomic::AtomicBool = std::sync::atomic::AtomicBool::new(false); +static FORK_RELEASE_A: std::sync::atomic::AtomicBool = std::sync::atomic::AtomicBool::new(false); + +#[tokio::test(flavor = "multi_thread")] +async fn fork_collision_with_live_concurrent_fork_is_retryable() { + use std::sync::atomic::Ordering::SeqCst; + + let _scenario = FailScenario::setup(); + FORK_A_AT_POINT.store(false, SeqCst); + FORK_RELEASE_A.store(false, SeqCst); + + let dir = tempfile::tempdir().unwrap(); + let uri = dir.path().to_str().unwrap().to_string(); + let main = helpers::init_and_load(&dir).await; + main.branch_create("feature").await.unwrap(); + + // First arrival (A) records readiness and blocks until released; the rest + // (B) fall through immediately. Bounded spin so a mistake can't hang forever. + fail::cfg_callback("fork.before_classify", || { + if FORK_A_AT_POINT + .compare_exchange(false, true, SeqCst, SeqCst) + .is_ok() + { + for _ in 0..2000 { + if FORK_RELEASE_A.load(SeqCst) { + break; + } + std::thread::sleep(std::time::Duration::from_millis(5)); + } + } + }) + .unwrap(); + + let uri_a = uri.clone(); + let writer_a = tokio::spawn(async move { + let mut a = Omnigraph::open(&uri_a).await.unwrap(); + helpers::mutate_branch( + &mut a, + "feature", + MUTATION_QUERIES, + "insert_person", + &mixed_params(&[("$name", "Eve")], &[("$age", 22)]), + ) + .await + }); + + // Wait (bounded) until A is parked at the fork point. + for _ in 0..600 { + if FORK_A_AT_POINT.load(SeqCst) { + break; + } + tokio::time::sleep(std::time::Duration::from_millis(5)).await; + } + assert!( + FORK_A_AT_POINT.load(SeqCst), + "writer A never reached the fork point" + ); + + // B wins the fork and commits it. + let mut b = Omnigraph::open(&uri).await.unwrap(); + helpers::mutate_branch( + &mut b, + "feature", + MUTATION_QUERIES, + "insert_person", + &mixed_params(&[("$name", "Frank")], &[("$age", 41)]), + ) + .await + .unwrap(); + + // Release A; it resumes, re-reads the manifest, and sees the fork is live. + FORK_RELEASE_A.store(true, SeqCst); + let err = writer_a + .await + .unwrap() + .expect_err("A's stale-snapshot fork should be a retryable conflict"); + fail::remove("fork.before_classify"); + + let msg = err.to_string(); + assert!( + !msg.contains("cleanup"), + "a live concurrent fork must not be misclassified as an orphan, got: {msg}" + ); + assert!( + msg.contains("refresh and retry") || msg.contains("expected manifest table version"), + "expected a retryable stale-view error, got: {msg}" + ); +} + #[tokio::test(flavor = "multi_thread")] async fn graph_publish_failpoint_triggers_before_commit_append() { let _scenario = FailScenario::setup(); diff --git a/crates/omnigraph/tests/lance_surface_guards.rs b/crates/omnigraph/tests/lance_surface_guards.rs index b65a808..ed1f22e 100644 --- a/crates/omnigraph/tests/lance_surface_guards.rs +++ b/crates/omnigraph/tests/lance_surface_guards.rs @@ -242,3 +242,51 @@ async fn _compile_delete_result_field_shape() -> lance::Result<()> { let _num_deleted: u64 = result.num_deleted_rows; Ok(()) } + +// --- Guard 9: force_delete_branch semantics -------------------------------- +// +// The branch-delete reconciler (`db/omnigraph/optimize.rs::reconcile_orphaned_branches`) +// and the eager best-effort reclaim in `cleanup_deleted_branch_tables` call +// `force_delete_branch` to drop orphaned branch refs. The single-authority +// design relies on three facts pinned here: +// 1. plain `delete_branch` errors on a missing ref (so the design uses the +// force variant instead); +// 2. `force_delete_branch` removes an existing (forked) branch — the orphan +// case, where a `tree/{branch}/` exists; +// 3. `force_delete_branch` on a *fully-absent* branch (no tree dir) still +// errors on the local store, because `remove_dir_all`'s NotFound is not +// caught for Lance's native error variant. `TableStore::force_delete_branch` +// wraps this to be fully idempotent. Pin the raw quirk so a future Lance +// fix (which would let us simplify the wrapper) is noticed. + +#[tokio::test] +async fn force_delete_branch_semantics() { + let dir = tempfile::tempdir().unwrap(); + let uri = dir.path().join("guard9.lance"); + let uri = uri.to_str().unwrap(); + let mut ds = fresh_dataset(uri).await; + + // (1) Plain delete of a never-created branch errors (RefNotFound). + assert!( + ds.delete_branch("nope").await.is_err(), + "Dataset::delete_branch on a missing ref should error; if this is now \ + Ok, the reconciler could drop the force variant." + ); + + // (2) force_delete_branch removes an existing (forked) branch. + let base = ds.version().version; + ds.create_branch("feature", base, None).await.unwrap(); + ds.force_delete_branch("feature").await.unwrap(); + assert!( + !ds.list_branches().await.unwrap().contains_key("feature"), + "force_delete_branch should remove an existing branch ref" + ); + + // (3) Quirk: force_delete on a fully-absent branch errors on the local + // store (worked around by TableStore::force_delete_branch). + assert!( + ds.force_delete_branch("never").await.is_err(), + "force_delete_branch on a fully-absent branch no longer errors — \ + TableStore::force_delete_branch's NotFound tolerance can be simplified." + ); +} diff --git a/crates/omnigraph/tests/maintenance.rs b/crates/omnigraph/tests/maintenance.rs index 3c6ab30..722bdc4 100644 --- a/crates/omnigraph/tests/maintenance.rs +++ b/crates/omnigraph/tests/maintenance.rs @@ -7,11 +7,24 @@ mod helpers; use std::time::Duration; +use lance::Dataset; use omnigraph::db::{CleanupPolicyOptions, Omnigraph}; use omnigraph::loader::{LoadMode, load_jsonl}; use helpers::{TEST_DATA, TEST_SCHEMA, count_rows, init_and_load}; +/// Filesystem URI of a node sub-table, mirroring the engine's layout +/// (FNV-1a of the type name under `nodes/`). Matches the helper in +/// `failpoints.rs`; used to inspect/forge Lance branches directly in tests. +fn node_table_uri(root: &str, type_name: &str) -> String { + let mut hash: u64 = 0xcbf2_9ce4_8422_2325; + for &b in type_name.as_bytes() { + hash ^= b as u64; + hash = hash.wrapping_mul(0x100_0000_01b3); + } + format!("{}/nodes/{hash:016x}", root.trim_end_matches('/')) +} + #[tokio::test] async fn optimize_on_empty_graph_returns_stats_per_table_with_no_changes() { let dir = tempfile::tempdir().unwrap(); @@ -158,3 +171,59 @@ async fn cleanup_then_optimize_preserves_rows_and_table_remains_writable() { .unwrap(); assert_eq!(count_rows(&db, "node:Person").await, people_before); } + +#[tokio::test] +async fn cleanup_reconciles_orphaned_branch_forks() { + // An incomplete prior `branch_delete` can leave a per-table Lance branch + // that the manifest no longer references (a "zombie" fork). It is + // unreachable through any snapshot but pins its `tree/{branch}/` storage. + // `cleanup` must reconcile it away: drop every Lance branch absent from the + // manifest authority, without touching `main`. + let dir = tempfile::tempdir().unwrap(); + let uri = dir.path().to_str().unwrap().to_string(); + let mut db = init_and_load(&dir).await; + + let people_before = count_rows(&db, "node:Person").await; + assert!(people_before > 0, "fixture should seed Person rows"); + + // Forge an orphaned fork the manifest never knew about. + let person_uri = node_table_uri(&uri, "Person"); + { + let mut ds = Dataset::open(&person_uri).await.unwrap(); + let base = ds.version().version; + ds.create_branch("ghost", base, None).await.unwrap(); + assert!( + ds.list_branches().await.unwrap().contains_key("ghost"), + "precondition: orphaned fork staged" + ); + } + + db.cleanup(CleanupPolicyOptions { + keep_versions: Some(1), + older_than: None, + }) + .await + .unwrap(); + + // Orphan reclaimed; main untouched. + { + let ds = Dataset::open(&person_uri).await.unwrap(); + assert!( + !ds.list_branches().await.unwrap().contains_key("ghost"), + "cleanup should reconcile the orphaned 'ghost' fork away" + ); + } + assert_eq!( + count_rows(&db, "node:Person").await, + people_before, + "cleanup must not disturb main while reconciling orphans" + ); + + // Idempotent: a second cleanup with the orphan already gone is a no-op. + db.cleanup(CleanupPolicyOptions { + keep_versions: Some(1), + older_than: None, + }) + .await + .unwrap(); +} diff --git a/docs/dev/invariants.md b/docs/dev/invariants.md index 70477d4..0cf295c 100644 --- a/docs/dev/invariants.md +++ b/docs/dev/invariants.md @@ -99,6 +99,7 @@ Use it this way: | Multi-table commit | Manifest CAS plus recovery sidecars; not a single Lance primitive | [writes.md](writes.md), [architecture.md](architecture.md) | | Constructive mutations | In-memory `MutationStaging`, one end-of-query table commit per touched table, then one manifest publish | [writes.md](writes.md), [execution.md](execution.md) | | Deletes | Inline-commit residual; delete-only queries allowed, mixed insert/update/delete rejected by D2 | [query-language.md](../user/query-language.md), [writes.md](writes.md) | +| Branch delete | Manifest is the single authority, flipped atomically first; per-table forks + commit-graph branch are derived state, reclaimed best-effort (`force_delete_branch`) with the `cleanup` reconciler as the guaranteed backstop. Reusing a name whose reclaim failed before `cleanup` surfaces an actionable error | [branches-commits.md](../user/branches-commits.md), [maintenance.md](../user/maintenance.md) | | Schema validation | Type checks, required fields, defaults, edge endpoint checks, and edge cardinality are enforced on write paths | [schema-language.md](../user/schema-language.md), [execution.md](execution.md) | | Unique constraints | Intra-batch and write-path checks exist; full cross-version uniqueness is still a gap | [schema-language.md](../user/schema-language.md) | | Storage trait | `TableStorage` exists as the sealed staged-write surface; full call-site migration and capability/stat surfaces are incomplete | [writes.md](writes.md), [architecture.md](architecture.md) | @@ -107,6 +108,13 @@ Use it this way: | Auth | Bearer token hashing and server-side actor resolution are implemented at the HTTP boundary | [server.md](../user/server.md), [policy.md](../user/policy.md) | | Tests | Tempdir-backed Lance tests are the current substrate; there is no `MemStorage` test backend | [testing.md](testing.md) | +The branch-delete reconciler is authority-derived: it reclaims orphaned forks +today and degrades to a no-op if Lance ships an atomic multi-dataset branch +operation, so the design composes with that future rather than blocking it. This +is the same shape as invariant 7 (indexes are derived state); prefer it over a +recovery-sidecar-style approach for any new multi-dataset metadata operation, +since the sidecar would be scaffolding to remove once the substrate closes the gap. + ## Known Gaps Do not hide these behind invariant wording. Either move them forward or keep diff --git a/docs/dev/lance.md b/docs/dev/lance.md index ef83f2c..100da6f 100644 --- a/docs/dev/lance.md +++ b/docs/dev/lance.md @@ -175,7 +175,8 @@ Migration from Lance 4.0.0 → 6.0.1 landed in this cycle (DataFusion 52 → 53, - **Lance #6658 closed** (2026-05-14) but `DeleteBuilder::execute_uncommitted` did **not** ship in v6.0.1 — binary search across the release stream shows it first appears in `v7.0.0-beta.10` (the closing commits landed on main but didn't backport to the 6.x line). Tracked as MR-A: migrate `delete_where` to staged, retire the parse-time D2 mutation rule, extend recovery sidecar coverage. **Gated on the Lance v7.x bump**, not this PR. v7.0.0-rc.1 dropped 2026-05-21. - **Lance #6666 still open** (`build_index_metadata_from_segments` public): vector-index two-phase blocked; inline `create_vector_index` residual retained. - **Lance #6877 still open** (`MergeInsertBuilder` dup-rowid): PR #109's `SourceDedupeBehavior::FirstSeen` + `check_batch_unique_by_keys` precondition stay load-bearing. +- **`Dataset::force_delete_branch`** (`branches().delete(name, force=true)`, dataset.rs:524) tolerates a missing branch-*contents* ref (vs plain `delete_branch`'s `RefNotFound`), but on the local store still errors `NotFound` if the branch `tree/` directory is fully absent (`remove_dir_all`'s NotFound is not caught for Lance's native error variant, refs.rs:526-549). Both variants still refuse a branch with referencing descendants (`RefConflict`). `TableStore::force_delete_branch` wraps this to be fully idempotent (tolerates already-absent). The single-authority branch-delete redesign uses it for orphan reclamation (eager best-effort reclaim + cleanup reconciler). Pinned by `lance_surface_guards.rs::force_delete_branch_semantics`. Branch delete is "flip the ref atomically, then `remove_dir_all(tree/{branch})`"; branch-exclusive data lives under `tree/{branch}/` so a drop reclaims it immediately without touching `main`. -Surface guards added: `crates/omnigraph/tests/lance_surface_guards.rs` (8 named guards; 3 runtime + 5 compile-only). Future Lance bumps re-run this file first as the smoke check. Two additional guards from the original plan deferred to follow-up (`manifest_cas_returns_row_level_contention_variant` needs full publisher-race harness; `table_version_metadata_byte_compatible_with_v4` needs `pub(crate)` reach extension). +Surface guards added: `crates/omnigraph/tests/lance_surface_guards.rs` (9 named guards; 4 runtime + 5 compile-only). Future Lance bumps re-run this file first as the smoke check. Two additional guards from the original plan deferred to follow-up (`manifest_cas_returns_row_level_contention_variant` needs full publisher-race harness; `table_version_metadata_byte_compatible_with_v4` needs `pub(crate)` reach extension). Bump this date stanza on the next alignment pass. diff --git a/docs/user/branches-commits.md b/docs/user/branches-commits.md index de6c653..c1894f9 100644 --- a/docs/user/branches-commits.md +++ b/docs/user/branches-commits.md @@ -8,10 +8,10 @@ Lance supports branching at the dataset level: a branch is a named lineage of ve OmniGraph builds *graph branches* on top by branching every sub-table coherently: -- `branch_create(name)` / `branch_create_from(target, name)` — disallowed name `main`; fails if branch exists; ensures the schema-apply lock is idle. +- `branch_create(name)` / `branch_create_from(target, name)` — disallowed name `main`; fails if branch exists; ensures the schema-apply lock is idle. Atomic and authority-first like `branch_delete`: it flips the `__manifest` branch (authority), then creates the derived commit-graph branch, force-dropping any orphaned commit-graph ref left by an incomplete prior delete (the manifest branch is fresh, so a same-named commit-graph branch is provably a zombie). If commit-graph creation fails, the manifest branch is rolled back so the name never half-exists. - `branch_list()` — returns public branches, **filters internal** `__run__…` and `__schema_apply_lock__` prefixes. -- `branch_delete(name)` — refuses if there are descendants or active runs on the branch; cleans up owned per-branch fragments. -- **Lazy forking**: a branch only forks a sub-table when that sub-table is first mutated on it. Pure-read branches share fragments with their source. +- `branch_delete(name)` — refuses if there are descendants or active runs on the branch. The manifest is the single authority for branch existence: deletion flips the `__manifest` branch ref first (one atomic op), after which the branch is gone from every snapshot. The owned per-table forks and the commit-graph branch are derived state, reclaimed best-effort with `force_delete_branch` after the flip. A failure during that reclaim (transient object-store error) does not fail the call or block the authority flip; the leftover forks are unreachable orphans that the [`cleanup`](maintenance.md) reconciler converges. One consequence: if a delete's best-effort reclaim fails, reusing that branch name before the next `cleanup` surfaces a clear error pointing at `cleanup` (the stale fork would otherwise collide on first write). +- **Lazy forking**: a branch only forks a sub-table when that sub-table is first mutated on it. Pure-read branches share fragments with their source. A fork collision is classified by the manifest authority, not by Lance branch versions: if the live manifest already records the fork on the active branch, a concurrent first-write won and the caller gets a retryable "refresh and retry"; if the manifest does not, a physical branch there is an orphan and the caller is pointed at `cleanup`. - `sync_branch(branch)` — re-binds the in-memory handle to the latest head of the branch. ## L2 — Commit graph (`db/commit_graph.rs`) diff --git a/docs/user/maintenance.md b/docs/user/maintenance.md index 08ae8da..9839ea1 100644 --- a/docs/user/maintenance.md +++ b/docs/user/maintenance.md @@ -14,9 +14,15 @@ - Lance `cleanup_old_versions()` per table. - Removes manifests (and their unique fragments) older than the retention policy. - `CleanupPolicyOptions { keep_versions: Option, older_than: Option }` — at least one is required. -- Returns `[TableCleanupStats { table_key, bytes_removed, old_versions_removed }]`. +- Returns `[TableCleanupStats { table_key, bytes_removed, old_versions_removed, error }]`. +- **Fault-isolated per table.** A single table's transient failure (version GC or + orphan reclaim) is recorded on that table's stats row (`error: Some(..)`, logged + via `tracing`) and never aborts the healthy tables — cleanup is the convergence + backstop, so it does as much as it can and converges on re-run. The CLI reports + any failed tables; rerun `cleanup` to retry them. - CLI guards with `--confirm`; without it, prints a preview line. - **Recovery floor:** `--keep < 3` may garbage-collect Lance versions that the open-time recovery sweep needs as a rollback target (the sweep restores to the branch's manifest-pinned table version, which is HEAD-1 in the typical Phase B → Phase C drift case). Default `--keep 10` is safe. +- **Orphaned-branch reconciliation:** before the version GC, cleanup runs `reconcile_orphaned_branches`, which `force_delete_branch`es any per-table or commit-graph Lance branch absent from the manifest branch list. These orphans arise when a `branch_delete` flips the manifest authority but a downstream best-effort reclaim does not complete (see [branches-commits.md](branches-commits.md)). The reconciler is authority-derived and idempotent (it no-ops once nothing is orphaned), runs regardless of the `keep_versions` / `older_than` values (those gate version GC only), and never reclaims `main` or system-branch forks. Reclaimed forks are logged via `tracing::info`. ## Tombstones From fab105bcce97ce6eb543d83c723fa3c4befb8bb5 Mon Sep 17 00:00:00 2001 From: Andrew Altshuler Date: Mon, 1 Jun 2026 12:56:21 +0100 Subject: [PATCH 03/11] fix(release): generate audit-clean Homebrew formula (#134) The generated formula failed `brew audit --strict` with 5 problems: `version` declared after `license`, and `url`/`sha256` placed directly inside `on_macos`/`on_linux` (forbidden by FormulaAudit/ComponentsOrder). Order `version` before `license`, hoist `head`/`livecheck` above the platform blocks, and nest `url`/`sha256` in `on_arm`/`on_intel`. Add a `brew audit --strict --online` gate to the release workflow so a malformed formula can never be published again. Verified clean against v0.6.0. Co-authored-by: Claude Opus 4.8 (1M context) --- .github/workflows/release.yml | 11 +++++++++++ scripts/update-homebrew-formula.sh | 29 ++++++++++++++++------------- 2 files changed, 27 insertions(+), 13 deletions(-) diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml index 48ab38c..3a66ff2 100644 --- a/.github/workflows/release.yml +++ b/.github/workflows/release.yml @@ -121,6 +121,17 @@ jobs: run: | ./scripts/update-homebrew-formula.sh "${GITHUB_REF_NAME}" homebrew-tap/Formula/omnigraph.rb + - name: Audit generated formula + if: env.HOMEBREW_TAP_SKIP != '1' + run: | + # Audit the checked-out tap by name (brew audit rejects bare paths + # and needs tap context). Symlink the checkout into Homebrew's Taps + # tree so `modernrelay/tap/omnigraph` resolves to it. + tap_dir="$(brew --repository)/Library/Taps/modernrelay/homebrew-tap" + mkdir -p "$(dirname "$tap_dir")" + ln -sfn "$PWD/homebrew-tap" "$tap_dir" + brew audit --strict --online modernrelay/tap/omnigraph + - name: Commit and push formula update if: env.HOMEBREW_TAP_SKIP != '1' working-directory: homebrew-tap diff --git a/scripts/update-homebrew-formula.sh b/scripts/update-homebrew-formula.sh index 90a5dea..f2f0df9 100755 --- a/scripts/update-homebrew-formula.sh +++ b/scripts/update-homebrew-formula.sh @@ -64,20 +64,8 @@ cat >"$FORMULA_PATH" < Date: Mon, 1 Jun 2026 22:50:31 +0200 Subject: [PATCH 04/11] Stored-query registry foundation + config/CLI RFC-002 (#128) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * MR-969: add stored-query registry config surface Introduce the `queries:` block in omnigraph.yaml — an inline `name -> entry` map of stored queries, per-graph (`graphs..queries`) and top-level for single-graph mode, mirroring how `policy` is wired in both modes. Each entry points at a `.gq` file and carries optional MCP exposure settings (`expose`, `tool_name`), defaulting to not-exposed. Additive: absent `queries:` leaves current behavior unchanged. - QueryEntry { file, mcp: McpSettings { expose, tool_name } } - `queries` field on TargetConfig + OmnigraphConfig (serde default) - query_entries() / target_query_entries() accessors - resolve_query_file() — base_dir-relative `.gq` path resolution - round-trip + absent-block tests Co-Authored-By: Claude Opus 4.8 (1M context) * Add stored-query registry loader and GraphHandle wiring Add a `queries` module: QueryRegistry loads each declared `.gq` entry, parses it, and selects the query whose symbol matches the manifest key, asserting the two agree (key == `query ` symbol). Identity is the query name; a key/symbol mismatch is a load-time error. Errors are collected, not fail-fast, so a bad registry surfaces every broken entry at once. Schema type-checking is deliberately left to a separate pass so the loader stays callable without an open engine. Thread an `Option>` through GraphHandle alongside the per-graph policy; the URI-canonicalizing clone propagates it. Production openers default to None for now — the boot path loads and attaches the registry in a later change. - QueryRegistry::{from_specs, load, lookup, iter}; StoredQuery::is_mutation - GraphHandle.queries field, propagated on canonical clone - registry unit tests: identity match/mismatch, multi-query selection, per-entry parse errors, error collection, mutation classification Co-Authored-By: Claude Opus 4.8 (1M context) * docs: add RFC-002 config & CLI architecture Layered config (user-global ~/.config/omnigraph/ + per-project), a unifying `target` abstraction resolving to (locus, graph, sub-state, credential) with embedded-URI XOR remote-server loci, multi-server × multi-graph client targeting, credentials by-reference, and the file-naming decision: project and server config are one artifact (`omnigraph.yaml`); the only differently-named file is the user-global `config.yaml`, split by scope not role. Includes the 12-factor bind portability rule (prefer --bind/OMNIGRAPH_BIND over a committed server.bind) and the defined-locally / invoked-remotely model for stored queries. Derived from first principles working backwards from what the engine enables; validated against kube/Helix/git/compose. Linked from docs/dev/index.md. Proposed; phased rollout for the MR-973/974/981 family. Co-Authored-By: Claude Opus 4.8 (1M context) * Add check() to validate stored queries against the live schema A pure check(registry, catalog) that type-checks every stored query via the same typecheck_query_decl the engine runs for inline queries — no parallel implementation. Failures are collected, not fail-fast, so an operator sees every broken query (e.g. a type/property a migration renamed or removed) in one pass. Breakages are fatal (the boot path will refuse to start); warnings are advisory. Pure over (registry, catalog) so it is callable both at boot (engine catalog) and offline from the CLI without an open engine. Advisory lint: an mcp.expose:true query that declares a Vector(N) parameter warns — an LLM cannot supply a raw embedding vector; such a query should take a String parameter and embed server-side. Warns rather than rejects, since service-to-service callers may pass vectors. - CheckReport { breakages, warnings }; has_breakages / is_clean - tests: valid query, unknown type, unknown property, collect-not-fail-fast, vector-param-exposed warns, unexposed silent Co-Authored-By: Claude Opus 4.8 (1M context) * Drop internal plan-label refs from stored-query config comments Doc comments referenced sequencing labels ("C2") that mean nothing to a reader; reword to describe the behavior directly. Comment-only. Co-Authored-By: Claude Opus 4.8 (1M context) * docs: reconcile aliases with the role model in RFC-002 Place the existing client-only `aliases:` block in the client/server role split: aliases are client-role (CLI, embedded, ungated) and may live in both user-global and project config; `queries:` is server-role (deployment manifest only). They overlap as "name -> .gq"; `queries:` is the superset, and the end-state subsumes aliases (definition -> queries, target/branch/format -> client invocation context, positional args -> CLI sugar). v1 keeps aliases unchanged. Co-Authored-By: Claude Opus 4.8 (1M context) * docs: make RFC-002 config global-first, project-optional The global user config is the primary, self-sufficient default; the CLI works from any directory with no project file (the kubectl/aws/gh posture), a deliberate flip from today's project-anchored behavior. The project omnigraph.yaml becomes an optional repo-scoped override and the deployment manifest. Uniform schema, both layers optional; global can hold any section including a personal server's graphs/queries. Additive: project still overrides global; the flip adds a fallback layer below the project file rather than removing it. Co-Authored-By: Claude Opus 4.8 (1M context) * docs: justify XDG ~/.config/omnigraph over legacy ~/.omnigraph in RFC-002 Make the rationale explicit: XDG-first because OmniGraph is a client that will cache remote catalogs and keep session state alongside secrets, and XDG separates config / cache / state into distinct dirs (clear cache without touching creds; backups skip cache) whereas a single ~/.omnigraph/ mixes them. Honor ~/.omnigraph/ as a fallback for the peer-group (aws/kube/docker/helix) expectation. Add XDG_CACHE_HOME / XDG_STATE_HOME to the override precedence. Co-Authored-By: Claude Opus 4.8 (1M context) * docs: build RFC-002 credentials on the existing env-file mechanism OmniGraph already has credentials-by-reference: bearer_token_env names the env var, and auth.env_file is a git-ignored dotenv the CLI auto-loads (real env vars win), resolved via resolve_remote_bearer_token. The RFC's proposed credentials.yaml + token_env were redundant parallel inventions. Reconcile: reuse bearer_token_env (extend to servers.) and auth.env_file (add a global ~/.config/omnigraph/.env layered under the project .env.omni); OS keychain is an additive future resolver. No new credentials.yaml. Updated summary, non-goals, background, file-naming, credentials, example, login, migration, rollout. Co-Authored-By: Claude Opus 4.8 (1M context) * docs: use single ~/.omnigraph dir (Helix-style), not XDG, in RFC-002 Reverse the earlier XDG-first call. The prior argument rested on a false dichotomy (single-dir => mixed config/cache/state); in fact the peer tools (aws, kube, helix) achieve separation via SUBDIRECTORIES inside one ~/.tool/ dir (~/.aws/sso/cache/, ~/.kube/cache/), getting cache hygiene AND one discoverable place. So everything goes under ~/.omnigraph/: config.yaml, credentials (dotenv, 0600), cache/, state/. Lower cognitive load, matches what DB/cloud-CLI users expect, matches Helix. OMNIGRAPH_HOME overrides; $XDG_CONFIG_HOME optionally honored but ~/.omnigraph/ is canonical. Updated all paths, the rationale paragraph, the file-naming table (added a cache/state row), and env precedence. Co-Authored-By: Claude Opus 4.8 (1M context) * docs: reconcile RFC-002 with shipped/planned CLI tickets Align with reality found in existing tickets: - Noun is graph/graphs, not target/targets (MR-603 done renamed the config key targets->graphs, flag --graph). Use graphs:/--graph; an entry is embedded (uri) XOR remote (server + remote graph name). - ~/.omnigraph/ confirmed by MR-581 (og template pull, done) which already quick-starts templates there. - Templates already exist (MR-581/MR-531) — not invented here. - The init family is already specced (init, quickstart MR-973, serve MR-970, prune MR-972, mcp install MR-974, agent-mode MR-981); this RFC only adds the user route (~/.omnigraph/config.yaml + login). - aliases: -> operations: planned (MR-839). - bearer_token_env gap tracked in MR-971. - query lint/check already exist (MR-639) — registry validator must not collide with the singular `query check`. Add a Reconciliation section; fix the canonical example to graphs:/--graph. Also: merge semantics refined (deep-merge settings, replace named entries, replace lists, config view --resolved --show-origin). Co-Authored-By: Claude Opus 4.8 (1M context) * docs: correct stale-ticket claims and fold init/bootstrap design into RFC-002 Verify against code, not ticket statuses (MR-581 is marked done but is stale/unbuilt): no ~/.omnigraph usage, no template/serve/quickstart/ prune/login commands exist; config still uses aliases: (no operations:). So ~/.omnigraph/ stands on peer-convention merits alone, and templates are a design question, not a foothold. Add §7.5: the three-tier init model (user route = login + ~/.omnigraph/config.yaml; thin project init; fat quickstart + templates) with first-principles positions (split init/login, in-place refuse-if-exists, interactive vs --auto/agent-mode, --template flag, secrets-on-scaffold gitignore rule). This RFC owns only the user route; the rest are sibling tickets (MR-973/970/972/974/981). Co-Authored-By: Claude Opus 4.8 (1M context) * docs: breadboard + slice Shape A in RFC-002 Add the implementation breadboard (places P1-P5, affordances N1-N14 with NEW markers, mermaid) and five vertical slices for the selected config/ CLI/init shape: V1 global layer + merge engine + config view; V2 remote graphs + HTTP-client path + credential resolution; V3 omnigraph login; V4 init-hardening + quickstart + templates (rides MR-970); V5 agent-mode (MR-981). Rollout reordered to the slice sequence; spikes X1-X4 gate their owning slice. V1-V2 close the substantive client->server gap. Co-Authored-By: Claude Opus 4.8 (1M context) * Add InvokeQuery Cedar action (coarse, graph-scoped) A per-graph, branch-scoped action that gates invoking a server-side stored query by name. Coarse for now: an `invoke_query` allow rule permits any stored query on the graph; a future, additive refinement adds an optional per-query-name scope without changing rules written against the coarse action. Enforcement is at the HTTP boundary; the engine `_as` writers still enforce read/change per the query body, so a stored mutation is double-gated (invoke_query to reach the tool, change for the write). No call site yet — the invocation handler wires it in a later change (same pattern as Admin/GraphList added ahead of consumers). - variant + as_str/resource_kind(Graph)/FromStr/uses_branch_scope - Cedar schema: invoke_query appliesTo Graph - tests: per-graph allow/deny, branch-scope accepted Co-Authored-By: Claude Opus 4.8 (1M context) * Load and type-check stored queries at server boot, refusing breakage At startup the server now loads each graph's stored-query registry, type-checks every query against that graph's live schema, and refuses to boot if any query references a type/property the schema doesn't have (same posture as bad policy YAML) — so schema drift surfaces at the deploy boundary, not silently at invocation. Non-blocking warnings are logged. The validated registry is attached to the GraphHandle (the two production sites previously held `queries: None`). Loading (parse + key==symbol identity) happens at settings-build time where the config is in scope; the schema type-check happens after each engine opens (single mode in `open_single_with_queries`, multi mode in `open_single_graph`). `open_with_bearer_tokens_and_policy` delegates with an empty registry so its 18 test callers are unchanged; the public `new_*` constructors are unchanged (only the private build path threads the registry). - ServerConfigMode::Single / GraphStartupConfig carry the loaded registry - boot tests: valid registry boots; type-broken query refuses boot + names it Co-Authored-By: Claude Opus 4.8 (1M context) * Add `omnigraph queries validate` and `queries list` CLI `queries validate` type-checks the stored-query registry against the live schema offline — it opens the selected graph, runs the same check() the server runs at boot, prints breakages/warnings (human or --json), and exits non-zero on any breakage — so an operator can catch a query broken by a schema change without restarting the server. `queries list` prints each registered query's name, MCP exposure, and typed params. Named `validate` (not `check`) to avoid overlap with the existing `omnigraph lint` — `query check`/`query lint` are already deprecated argv-shims to `lint`. Registry entries resolve like the server: a named graph uses its per-graph `queries:`; otherwise the top-level one. - Queries subcommand group; reuses QueryRegistry::load + check from omnigraph-server; local-only (needs the schema), mirrors lint - tests: clean registry exits 0, broken query exits non-zero + names it, list shows the query and its typed params Co-Authored-By: Claude Opus 4.8 (1M context) * Route registry selection through one shared query_entries_for The "which queries: block applies for graph X" rule existed twice — the server boot path and the CLI's registry_entries — and had already drifted: the CLI carried an unreachable unwrap_or_else fallback the server lacked. Add OmnigraphConfig::query_entries_for(graph: Option<&str>) as the single definition (named graph -> its per-graph block; otherwise top-level) and route all three sites through it: server single mode, server multi-graph loop, and the CLI. The CLI's dead fallback arm is deleted; CLI and server now resolve identically by construction. No behavior change. Extends the config round-trip test to pin the selector, including the unknown-name -> top-level fallback the deleted CLI arm covered. * Funnel registry validation through one validate_and_attach gate The check -> refuse-on-breakage -> log-warnings -> empty->None block was copy-pasted across both open paths (single mode and the multi-graph per-graph open), differing only by the graph label. A third opener could attach a registry that was never schema-checked. Extract validate_and_attach(queries, catalog, label) -> Option> as the single gate both paths call, so attaching an unchecked registry is no longer expressible. The catalog handle is an owned Arc, so calling it before the multi-mode policy match (which rebinds db) is borrow-clean. No behavior change. Adds a direct unit test of the helper (empty / clean / breakage incl. the graph label in the message) — covering the multi-graph path's logic, which previously had no boot-refusal coverage. * Resolve param types structurally in the MCP vector lint The exposed-query advisory detected vector params with type_name.starts_with("Vector(") — a second copy of the compiler's own ScalarType::from_str_name vector parsing that could drift from it. Key the lint off PropType::from_param_type_name + ScalarType::Vector(_) instead, the one canonical resolver the type system already uses. Any future param-suppliability lint now reads the structured type rather than scanning the surface string. Behavior-preserving: the grammar forbids list-of-vector params (list_type = "[" base_type "]", and base_type excludes Vector), so the only input where the structured and string checks could differ is unparseable. Adds a guard test that an exposed String param does not false-trigger the warning. * Refuse duplicate MCP tool names across exposed stored queries The effective MCP tool name (explicit tool_name, else the query name) is a second identity namespace beside the registry key, but nothing enforced it unique — two exposed queries could claim one catalog key, and each consumer re-derived the name ad hoc. Add StoredQuery::effective_tool_name() as the one definition, and a load-time uniqueness pass in from_specs over exposed queries: a collision is a collected LoadError naming the loser and the winner. Scoped to exposed queries (unexposed have no MCP tool); deterministic over the BTreeMap so the first-declared wins and the error order is stable. New (rare) refusal: a config with colliding exposed tool names now fails `omnigraph queries validate` offline and refuses server boot, the same posture as a malformed registry. Release-note-worthy. Test-first: duplicate_exposed_tool_name_is_a_load_error (red before the pass, green after) + a CLI offline test; the unexposed sibling pins the exposed-only scope; effective_tool_name asserts folded into the load test. * docs: document the queries registry, CLI, and invoke_query action The stored-query surface shipped without user docs. Add it, per the same-PR maintenance contract: - policy.md: invoke_query as per-graph action #10 (branch-scoped), with the double-gating note; renumber graph_list; add it to the branch_scope list. - cli-reference.md: the `queries validate | list` command, and the `queries:` config block (per-graph + top-level) with mcp.expose/tool_name and the tool-name uniqueness rule. - server.md: boot-time stored-query type-check (refuse on breakage), noting invocation over HTTP/MCP is not yet exposed. * Add POST /queries/{name} stored-query invocation handler Invoke a curated server-side stored query by name: source + name come from the per-graph queries: registry, the client sends only runtime inputs (params, branch, snapshot). Gated by the invoke_query Cedar action at the boundary; the handler delegates to the existing run_query/run_mutate, whose inner Read/Change enforce still runs — so a stored mutation is double-gated (invoke_query to reach the tool, change for the write). - InvokeStoredQueryRequest + an untagged InvokeStoredQueryResponse { Read(ReadOutput), Change(ChangeOutput) } → one Json<_> return type and a oneOf 200 schema (a correct contract, not a wrong-but-simple one). - Route lives in per_graph_protected → single-mode /queries/{name} and multi-mode /graphs/{id}/queries/{name} for free. - Deny == unknown: an invoke_query denial and a missing query both return the same 404, so the catalog can't be probed by an unauthorized caller. - OpenAPI regenerated; tests cover read, mutation double-gate (403 vs 200), bad-param 400, and the identical-404 deny path. Completes the MR-969 V1 invocation slice (registry + /queries/{name} + invoke_query). * docs: stored-query invocation endpoint; flip the not-yet-exposed caveat Now that POST /queries/{name} ships (C7), document it: add the endpoint to server.md's inventory + an invocation section (body, untagged read/mutate envelope, invoke_query gate, double-gated mutations, deny == 404), and flip the startup note that said invocation was not yet exposed. In policy.md, replace "no invocation call site yet" on the invoke_query action with a pointer to the endpoint. * Scope the stored-query 404-hiding claim to non-invoke_query callers Review found the deny==404 catalog-hiding was overstated as a contract: it holds only at the outer invoke_query gate. A caller that HOLDS invoke_query but lacks read/change gets the inner gate's 403 for an existing query vs 404 for an unknown one — so existence is visible to grant-holders by design (the intended double-gate). The handler docstring, OpenAPI 404 description, and server.md all claimed the 404 was airtight against any denied actor. Correct the wording in all three (no behavior change) and add the missing symmetric test (invoke_query but no read -> 403 for an existing query, 404 for unknown) so the actual contract is pinned. Also document that in default-deny mode (tokens, no policy) every invocation 404s until an invoke_query rule is configured. Nits: the from_specs collision comment said "first declared wins" but it is lexicographically-first by name (BTreeMap); the effective_tool_name docstring overclaimed the CLI display routes through it (it resolves the rule on its own output DTO). * Default mcp.expose to true (the manifest entry is the opt-in) expose controls MCP-catalog membership only — it is not an authorization gate (invocation is gated by invoke_query regardless). So requiring a per-query mcp.expose: true was friction with no safety benefit: a non-exposed query is still HTTP-invocable by name. Flip the default so declaring a query in the manifest exposes it to the agent tool catalog by default; expose: false is the escape hatch for service-only queries. Both the absent-mcp path (Default impl) and the present-but-no-expose path (serde default fn) now yield true. Doc comments + cli-reference updated; the config round-trip test asserts the new default. * Add GET /queries stored-query catalog endpoint List a graph's mcp.expose stored queries as a typed tool catalog so a client (the MCP server) can register them as tools without fetching .gq source. Each entry carries name, MCP tool_name, description/instruction, a read/mutate flag, and decomposed typed params (kind enum: string|bool|int| bigint|float|date|datetime|blob|vector|list, plus item_kind for lists and vector_dim) — so the consumer builds an input schema with a closed match and never re-parses omnigraph type spelling. I64/U64 are bigint (string on the wire): a JSON number loses precision past 2^53 and the engine already accepts decimal strings. Read-gated (works in default-deny; the catalog is graph-wide, authorized against main). NOT Cedar-filtered per query yet — a reader can list a query whose invoke_query they lack (documented gap until per-query authz lands); invocation stays invoke_query-gated + deny==404. - api: QueriesCatalogOutput / QueryCatalogEntry / ParamDescriptor / ParamKind + query_catalog_entry (reuses PropType::from_param_type_name; scalar_kind is exhaustive, so a new ScalarType is a compile error here until catalogued). - GET /queries route in per_graph_protected (→ /graphs/{id}/queries in multi mode); OpenAPI regenerated; path allowlists updated. - Tests: projection unit (every kind, list, vector, nullable, mutation, empty) + handler (exposed-only filter, read-gate probe-oracle, empty registry). * docs: GET /queries stored-query catalog endpoint Document the catalog: the endpoint table row (GET /queries, read-gated), a catalog section (typed-param kind enum, bigint/date/datetime/blob-as-string, graph-wide/branch-independent, mcp.expose default true, the read-gated probe-oracle gap), and flip the startup note now that the catalog ships. * Collect file-I/O and parse errors in QueryRegistry::load in one pass load() early-returned on any unreadable .gq file, masking parse / identity / tool-name-collision errors in the OTHER (readable) files — so an operator fixed the missing file, restarted, and only then saw the next broken query. Now it collects I/O errors but still runs from_specs on the readable specs and returns the union, so every broken entry surfaces at once (matching the collected-errors contract the rest of the registry already follows). Safe: from_specs' tool-name collision check runs over loaded queries only, so dropping an I/O-failed entry can only under-report a collision, never invent one. I/O errors are ordered first (BTreeMap key order), then spec errors. Adds a load-level test (tempdir: a valid, a missing, and a parse-broken .gq) asserting all three surface in one Err — confirmed red before the fix. * Make invoke_query graph-scoped (one branch authority) invoke_query gates reaching the curated stored-query surface — a graph-level capability. Per-branch/snapshot access is already enforced by the inner read/change gate in run_query/run_mutate (authorized against the resolved branch), so branch-scoping the outer gate was redundant AND wrong for snapshot reads (it defaulted to main). Drop the branch dimension: remove InvokeQuery from uses_branch_scope (it joins admin as graph-scoped) and authorize the boundary gate with branch: None. Lossless: an actor confined to branch X by their read/change rules can still only invoke a stored query that touches X. A rule that sets branch_scope on invoke_query is now rejected by validate() — write invoke_query in its own rule. Ripple (atomic): restructure the server invoke fixture so invoke_query sits in its own branch_scope-free rule; invert invoke_query_is_branch_scoped -> invoke_query_rejects_branch_scope; the per-graph authorize test uses branch: None; docs (policy.md, server.md, the InvokeQuery doc). No wire/OpenAPI change. * Resolve graph config by identity, not server mode Which policy/queries block applies for a graph was decided three different, mode-dependent ways: single-mode boot used top-level even for a named graph; multi-mode used per-graph (and silently ignored a top-level queries block); the CLI used per-graph for a named target. So `queries validate --target prod` could check a different registry than the single-mode server loaded, and a named graph's per-graph policy/queries were silently shadowed. Make config a function of graph IDENTITY: a graph served by NAME (--target/server.graph, a graphs: entry) uses its own graphs..{policy, queries}; a bare URI is anonymous and uses top-level. One rule, applied by single-mode boot, multi-mode boot, and the CLI — so they can't diverge and the CLI predicts the server exactly. No silent ignore: serving a named graph while a top-level policy/queries block is populated now refuses boot, naming the block (the multi-mode top-level-policy bail, extended to queries and to single-mode-named). The CLI's `queries validate` derives the schema URI and the registry from ONE selection, and a positional URI forces anonymous (ignoring cli.graph) so the two can't come from different graphs. BREAKING (released behavior): single mode by name (--target/server.graph) with top-level policy/queries previously used top-level; it now uses the per-graph block and refuses boot if top-level is also populated. Bare-URI single mode is unchanged. Loud, with migration text pointing at graphs.. - config: resolve_policy_file_for (policy sibling of query_entries_for, no top-level fallback) + populated_top_level_blocks for the coherence check. - characterization tests (single-mode named -> per-graph; named + top-level -> bail; multi-mode top-level queries -> bail; CLI positional-URI -> top-level). - docs: policy.md, server.md, cli-reference.md. * docs: RFC-002 credentials keyed by server name (keychain/profile/env) Reworks the RFC's credentials model: secrets are keyed by server name — OS keychain `omnigraph:` (preferred) -> a `[]` profile in `~/.omnigraph/credentials` -> `OMNIGRAPH_TOKEN[_]` env (CI), the AWS/gh/kube model. `servers.` is endpoint-only by default but may carry an explicit, secret-free `auth: { token: { env|file|command|keychain } }` source. The shipped `bearer_token_env` + `.env.omni` dotenv remain a legacy compat path; no `credentials.yaml`. * docs: RFC-002 — typed graph locator (storage/server/graph_id), not a uri string Add §1.1: the resolved graph address is a typed GraphLocator (Embedded{storage} | Remote{server, graph_id}), not a flat uri: String. Diagnoses the string model's cost in the code today (~16 is_remote_uri forks, TargetConfig can't express multi-server x multi-graph, the CLI bails on remote, the ts SDK models baseUrl+graphId separately) and settles the YAML naming so the key names the locus: - storage: (embedded) — shipped uri: is a deprecated alias - server: + graph_id: (remote) — graph_id defaults to the entry key - storage xor server, reject both/neither (no silent ambiguity) Kills the graphs:/graph: collision and the uri:-might-be-a-server ambiguity. Updates the §1/§8 examples and the entry-shape notes to the new naming. * Test: queries list must reject an unknown --target queries list opens no graph URI, so unknown-graph validation does not ride along on resolve_target_uri the way it does for every other command. The new test reproduces the gap: with an unknown --target the command currently exits 0 and prints the (empty) top-level registry instead of erroring like the URI-resolving commands do. Fails against current code; the fix follows. * Validate the graph selection in queries list Graph-existence validation was a side effect of URI resolution: every URI-resolving command rejects an unknown --target via resolve_target_uri, but queries list opens no URI, so query_entries_for(Some(unknown)) silently fell back to the top-level registry and showed the wrong (or empty) catalog. Make membership a property of the selection: add the fallible resolve_graph_selection alongside the infallible query_entries_for (a known name passes through, an unknown name errors with the same message as resolve_target_uri, None stays anonymous), and validate the selection in execute_queries_list. query_entries_for is unchanged — server boot's bare-URI path still needs its None -> top-level arm. * Surface policy-engine errors from stored-query invoke The invoke handler mapped every authorize_request failure to 404 ('stored query not found'), which collapsed the authorization decision (deny -> 403) together with operational failures (no actor -> 401, Cedar evaluation error -> 500). A real policy-engine 500 was hidden as a missing query. Separate the two concerns instead of sniffing the masked status. Extract authorize() returning an Authz { Allowed, Denied(msg) } decision and reserve Err for operational failures only; authorize_request becomes a thin wrapper that maps Denied -> 403, so the 16 deny-as-403 callers are unchanged. The invoke handler now matches the decision directly: a denial stays 404 (deny == missing, so the catalog can't be probed without the grant), while a 401/500 propagates with its true status. 500 is now a reachable outcome on POST /queries/{name}; document it in the endpoint responses and regenerate openapi.json. * Extract the named-graph/top-level coherence rule into one helper The rule 'a named graph uses its own graphs. block, so a populated top-level block is a config error' lived inline in single-mode server boot. Extract it to OmnigraphConfig::ensure_top_level_blocks_honored so the same definition can be shared by the CLI selection gate (next commit) and the two can't drift. Boot calls the helper; the message is reworded context-neutral (drops 'serving') so it reads correctly from both boot and the CLI. Behavior-preserving: multi-graph mode keeps its own unconditional check, and single_mode_named_graph_rejects_top_level_blocks still passes. * Test: queries validate/list must reject a named graph with a top-level block Server boot refuses a config where a graph is selected by name yet a top-level queries:/policy.file block is populated (the block would be silently ignored). The CLI's queries validate/list resolve the same named selection but skip that coherence check, so they give a false green / list the per-graph block. The new test reproduces it: validate prints OK and list succeeds where boot would refuse. Fails against current code; the fix follows. * Enforce top-level coherence in the single CLI selection gate queries validate validated graph membership only as a side effect of URI resolution and queries list only via resolve_graph_selection's membership check; neither applied the named-graph/top-level coherence rule server boot enforces, so both gave a false green on a config boot refuses. Fold ensure_top_level_blocks_honored into resolve_graph_selection so it is the single gate that returns only valid + server-coherent selections, and route resolve_selected_graph (queries validate) through it; queries list already calls the gate. A named graph with a populated top-level block now errors in both commands, matching boot. A positional URI stays anonymous (top-level honored), so queries_validate_positional_uri_ignores_default_graph is unaffected. * docs: RFC-003 — MCP server surface for omnigraph-server Detailed MCP-transport design for the stored-query/MCP work, building on the shipped #128 registry. Corrects the draft against the branch head: the coarse invoke_query gate + 404 denial-masking are already wired (server_invoke_query), so per-query invoke_query scope (PolicyRequest has no query-name dimension yet) is the real prerequisite; positions the doc as superseding rfc-001's MCP transport (/mcp/tools+/mcp/invoke) and reconciles the shipped mcp.expose YAML form and the schema-introspection non-goal; grounds the parity surface in the actual omnigraph-ts package (13 tools with read/change ids, 2 resources). * docs(config): clarify graph config boundaries * fix(config): enforce graph-scoped policies and query validation * fix(cli): require graph selection for scoped query registries * fix(server): preserve named graph id in single mode policy * fix(cli): share graph identity for policy resolution * test(cli): cover policy tooling server graph selection * fix(cli): honor server graph for policy tooling --------- Co-authored-by: Claude Opus 4.8 (1M context) --- crates/omnigraph-cli/src/main.rs | 679 +++++++++++++++-- crates/omnigraph-cli/tests/cli.rs | 292 ++++++++ crates/omnigraph-cli/tests/system_local.rs | 182 ++++- crates/omnigraph-cli/tests/system_remote.rs | 4 +- crates/omnigraph-policy/src/lib.rs | 95 ++- crates/omnigraph-server/src/api.rs | 159 ++++ crates/omnigraph-server/src/config.rs | 408 ++++++++++- crates/omnigraph-server/src/lib.rs | 620 +++++++++++++++- crates/omnigraph-server/src/queries.rs | 688 ++++++++++++++++++ crates/omnigraph-server/src/registry.rs | 12 + crates/omnigraph-server/tests/openapi.rs | 33 + crates/omnigraph-server/tests/server.rs | 637 +++++++++++++++- crates/omnigraph/src/db/omnigraph.rs | 37 +- .../src/db/omnigraph/schema_apply.rs | 129 ++-- docs/dev/index.md | 2 + docs/dev/rfc-002-config-cli-architecture.md | 590 +++++++++++++++ docs/dev/rfc-003-mcp-server-surface.md | 270 +++++++ docs/user/cli-reference.md | 12 +- docs/user/policy.md | 20 +- docs/user/server.md | 27 +- openapi.json | 319 ++++++++ 21 files changed, 5017 insertions(+), 198 deletions(-) create mode 100644 crates/omnigraph-server/src/queries.rs create mode 100644 docs/dev/rfc-002-config-cli-architecture.md create mode 100644 docs/dev/rfc-003-mcp-server-surface.md diff --git a/crates/omnigraph-cli/src/main.rs b/crates/omnigraph-cli/src/main.rs index d98c302..879f070 100644 --- a/crates/omnigraph-cli/src/main.rs +++ b/crates/omnigraph-cli/src/main.rs @@ -9,6 +9,7 @@ use clap::{Arg, ArgAction, Args, CommandFactory, FromArgMatches, Parser, Subcomm use color_eyre::eyre::{Result, bail}; use omnigraph::db::{Omnigraph, ReadTarget, SnapshotId}; use omnigraph::loader::LoadMode; +use omnigraph::storage::normalize_root_uri; use omnigraph_compiler::query::parser::parse_query; use omnigraph_compiler::schema::parser::parse_schema; use omnigraph_compiler::{ @@ -24,9 +25,10 @@ use omnigraph_server::api::{ SnapshotTableOutput, commit_output, ingest_output, read_output, schema_apply_output, snapshot_payload, }; +use omnigraph_server::queries::{QueryRegistry, check, format_check_breakages}; use omnigraph_server::{ AliasCommand, OmnigraphConfig, PolicyAction, PolicyDecision, PolicyEngine, PolicyRequest, - PolicyTestConfig, ReadOutputFormat, load_config, + PolicyTestConfig, ReadOutputFormat, graph_resource_id_for_selection, load_config, }; use reqwest::Method; use reqwest::header::AUTHORIZATION; @@ -153,6 +155,11 @@ enum Command { #[arg(long)] json: bool, }, + /// Operate on the server-side stored-query registry (`queries:`). + Queries { + #[command(subcommand)] + command: QueriesCommand, + }, /// Show graph snapshot Snapshot { /// Graph URI @@ -502,6 +509,35 @@ enum PolicyCommand { }, } +#[derive(Debug, Subcommand)] +enum QueriesCommand { + /// Type-check the stored-query registry against the live schema. + /// + /// Distinct from `omnigraph lint` (which lints one `.gq` file): + /// this validates the whole `queries:` registry — opening the graph + /// to read its schema and confirming every stored query still + /// type-checks. Exits non-zero on any breakage. + Validate { + /// Graph URI + uri: Option, + #[arg(long)] + target: Option, + #[arg(long)] + config: Option, + #[arg(long)] + json: bool, + }, + /// List the registered stored queries (name, MCP exposure, params). + List { + #[arg(long)] + target: Option, + #[arg(long)] + config: Option, + #[arg(long)] + json: bool, + }, +} + #[derive(Debug, Args, Clone)] struct ParamsArgs { #[arg(long, conflicts_with = "params_file")] @@ -743,25 +779,66 @@ fn load_cli_config(config_path: Option<&PathBuf>) -> Result { Ok(config) } -fn resolve_policy_engine(config: &OmnigraphConfig) -> Result { - let policy_file = config - .resolve_policy_file() - .ok_or_else(|| color_eyre::eyre::eyre!("policy.file must be set in omnigraph.yaml"))?; - PolicyEngine::load_graph(&policy_file, &policy_graph_id(config)) +#[derive(Debug, Clone)] +struct ResolvedCliGraph { + uri: String, + selected: Option, + graph_id: String, + policy_file: Option, + is_remote: bool, } -/// Open a local-URI graph and, when `policy.file` is configured in -/// `omnigraph.yaml`, install the resolved `PolicyEngine` on the engine -/// handle so every direct-engine write goes through -/// `Omnigraph::enforce(...)` (MR-722). Without a configured policy this -/// is identical to a bare `Omnigraph::open`. -/// -/// Returns owned `Omnigraph`; chained on top of `Omnigraph::open(...)`'s -/// existing future to keep call sites narrow. -async fn open_local_db_with_policy(uri: &str, config: &OmnigraphConfig) -> Result { - let db = Omnigraph::open(uri).await?; - if config.resolve_policy_file().is_some() { - let engine = Arc::new(resolve_policy_engine(config)?); +impl ResolvedCliGraph { + fn selected(&self) -> Option<&str> { + self.selected.as_deref() + } +} + +struct ResolvedPolicyContext { + policy_file: PathBuf, + graph_id: String, +} + +fn resolve_policy_context(config: &OmnigraphConfig) -> Result { + let selected = config.resolve_policy_tooling_graph_selection()?; + let policy_file = config + .resolve_policy_file_for(selected) + .ok_or_else(|| { + color_eyre::eyre::eyre!( + "policy.file or graphs..policy.file must be set in omnigraph.yaml" + ) + })?; + let graph_id = match selected { + Some(name) => graph_resource_id_for_selection(Some(name), ""), + None => graph_resource_id_for_selection(None, "default"), + }; + Ok(ResolvedPolicyContext { + policy_file, + graph_id, + }) +} + +fn resolve_policy_engine(context: &ResolvedPolicyContext) -> Result { + PolicyEngine::load_graph(&context.policy_file, &context.graph_id) +} + +fn resolve_policy_engine_for_graph(graph: &ResolvedCliGraph) -> Result { + let policy_file = graph.policy_file.as_ref().ok_or_else(|| { + color_eyre::eyre::eyre!( + "policy.file or graphs..policy.file must be set in omnigraph.yaml" + ) + })?; + PolicyEngine::load_graph(policy_file, &graph.graph_id) +} + +/// Open a local graph and install the policy resolved for the same graph +/// identity that produced the URI. A named graph uses +/// `graphs..policy.file`; an explicit positional URI is anonymous and +/// uses the legacy top-level `policy.file`. +async fn open_local_db_with_policy(graph: &ResolvedCliGraph) -> Result { + let db = Omnigraph::open(&graph.uri).await?; + if graph.policy_file.is_some() { + let engine = Arc::new(resolve_policy_engine_for_graph(graph)?); Ok(db.with_policy(engine as Arc)) } else { Ok(db) @@ -778,22 +855,16 @@ fn resolve_cli_actor<'a>(cli_as: Option<&'a str>, config: &'a OmnigraphConfig) - cli_as.or(config.cli.actor.as_deref()) } -fn resolve_policy_tests_path(config: &OmnigraphConfig) -> Result { - config.resolve_policy_tests_file().ok_or_else(|| { - color_eyre::eyre::eyre!( - "policy.tests.yaml requires policy.file to be set in omnigraph.yaml" - ) - }) +fn resolve_policy_tests_path(context: &ResolvedPolicyContext) -> PathBuf { + context.policy_file.with_file_name("policy.tests.yaml") } -fn policy_graph_id(config: &OmnigraphConfig) -> String { - if let Some(name) = &config.project.name { - return name.clone(); +fn normalize_policy_graph_uri(uri: &str) -> Result { + if is_remote_uri(uri) { + Ok(uri.trim_end_matches('/').to_string()) + } else { + Ok(normalize_root_uri(uri)?) } - config - .resolve_target_uri(None, None, config.server_graph_name()) - .or_else(|_| config.resolve_target_uri(None, None, config.cli_graph_name())) - .unwrap_or_else(|_| "default".to_string()) } fn resolve_remote_bearer_token( @@ -877,6 +948,47 @@ fn resolve_uri( config.resolve_target_uri(cli_uri, cli_target, config.cli_graph_name()) } +fn resolve_cli_graph( + config: &OmnigraphConfig, + cli_uri: Option, + cli_target: Option<&str>, +) -> Result { + let selected = if cli_uri.is_some() { + None + } else { + cli_target + .map(str::to_string) + .or_else(|| config.cli_graph_name().map(str::to_string)) + }; + config.resolve_graph_selection(selected.as_deref())?; + let uri = resolve_uri(config, cli_uri, cli_target)?; + let normalized_uri = normalize_policy_graph_uri(&uri)?; + let graph_id = graph_resource_id_for_selection(selected.as_deref(), &normalized_uri); + Ok(ResolvedCliGraph { + graph_id, + is_remote: is_remote_uri(&uri), + policy_file: config.resolve_policy_file_for(selected.as_deref()), + selected, + uri, + }) +} + +fn resolve_local_graph( + config: &OmnigraphConfig, + cli_uri: Option, + cli_target: Option<&str>, + operation: &str, +) -> Result { + let graph = resolve_cli_graph(config, cli_uri, cli_target)?; + if graph.is_remote { + bail!( + "{} is only supported against local graph URIs in this milestone", + operation + ); + } + Ok(graph) +} + /// Parse a Go-style compact duration: `7d`, `24h`, `30m`, `90s`, or a plain /// integer as seconds. Used by the `cleanup --older-than` flag. fn parse_duration_arg(s: &str) -> Result { @@ -915,14 +1027,7 @@ fn resolve_local_uri( cli_target: Option<&str>, operation: &str, ) -> Result { - let uri = resolve_uri(config, cli_uri, cli_target)?; - if is_remote_uri(&uri) { - bail!( - "{} is only supported against local graph URIs in this milestone", - operation - ); - } - Ok(uri) + Ok(resolve_local_graph(config, cli_uri, cli_target, operation)?.uri) } fn resolve_branch( @@ -1609,6 +1714,248 @@ async fn execute_query_lint( )) } +#[derive(serde::Serialize)] +struct QueriesIssue { + query: String, + message: String, +} + +#[derive(serde::Serialize)] +struct QueriesValidateOutput { + ok: bool, + breakages: Vec, + warnings: Vec, +} + +#[derive(serde::Serialize)] +struct QueriesParam { + name: String, + #[serde(rename = "type")] + type_name: String, + nullable: bool, +} + +#[derive(serde::Serialize)] +struct QueriesListItem { + name: String, + mcp_expose: bool, + tool_name: Option, + mutation: bool, + params: Vec, +} + +#[derive(serde::Serialize)] +struct QueriesListOutput { + queries: Vec, +} + +/// Resolve the selected graph to `(local URI, registry selection)` from one +/// precedence, so a command's schema and its stored-query registry can never +/// come from different graphs. A **positional URI is anonymous** (top-level +/// registry, ignoring the configured default graph); otherwise `--target` +/// or the configured `cli.graph` names the graph (its per-graph block). +/// Mirrors the server's single-mode identity rule. +fn resolve_selected_graph( + config: &OmnigraphConfig, + cli_uri: Option, + cli_target: Option<&str>, + operation: &str, +) -> Result<(String, Option)> { + let graph = resolve_local_graph(config, cli_uri, cli_target, operation)?; + Ok((graph.uri, graph.selected)) +} + +/// Load the stored-query registry for an already-resolved graph selection +/// (`None` = anonymous → top-level; `Some(name)` = that graph's block). +fn load_registry_or_report( + config: &OmnigraphConfig, + selected: Option<&str>, +) -> Result { + QueryRegistry::load(config, config.query_entries_for(selected)).map_err(|errors| { + color_eyre::eyre::eyre!( + "stored-query registry failed to load:\n {}", + errors + .iter() + .map(|e| e.to_string()) + .collect::>() + .join("\n ") + ) + }) +} + +fn graph_query_registry_names(config: &OmnigraphConfig) -> Vec<&str> { + config + .graphs + .iter() + .filter_map(|(name, graph)| (!graph.queries.is_empty()).then_some(name.as_str())) + .collect() +} + +fn resolve_registry_selection_for_list( + config: &OmnigraphConfig, + target: Option<&str>, +) -> Result> { + let selected = target + .map(str::to_string) + .or_else(|| config.cli_graph_name().map(str::to_string)); + if let Some(name) = selected.as_deref() { + config.resolve_graph_selection(Some(name))?; + return Ok(selected); + } + + if !config.query_entries().is_empty() { + return Ok(None); + } + + let graph_names = graph_query_registry_names(config); + if graph_names.is_empty() { + return Ok(None); + } + + bail!( + "stored-query registries are configured for graph{} {} but no graph was selected. Pass `--target {}` or set `cli.graph`.", + if graph_names.len() == 1 { "" } else { "s" }, + graph_names.join(", "), + graph_names[0], + ) +} + +fn validate_registry_for_catalog( + registry: &QueryRegistry, + catalog: &omnigraph_compiler::catalog::Catalog, + label: &str, +) -> omnigraph::error::Result<()> { + let report = check(registry, catalog); + if report.has_breakages() { + return Err(omnigraph::error::OmniError::manifest( + format_check_breakages(label, &report), + )); + } + Ok(()) +} + +async fn execute_queries_validate( + uri: Option, + target: Option, + config_path: Option<&PathBuf>, + json: bool, +) -> Result<()> { + let config = load_cli_config(config_path)?; + // One selection drives both the schema URI and the registry, so a + // positional URI and a `--target` can't validate different graphs. + let (uri, selected) = + resolve_selected_graph(&config, uri, target.as_deref(), "queries validate")?; + let registry = load_registry_or_report(&config, selected.as_deref())?; + let db = Omnigraph::open(&uri).await?; + let report = check(®istry, &db.catalog()); + + let output = QueriesValidateOutput { + ok: !report.has_breakages(), + breakages: report + .breakages + .iter() + .map(|b| QueriesIssue { + query: b.query.clone(), + message: b.message.clone(), + }) + .collect(), + warnings: report + .warnings + .iter() + .map(|w| QueriesIssue { + query: w.query.clone(), + message: w.message.clone(), + }) + .collect(), + }; + + if json { + print_json(&output)?; + } else { + if output.breakages.is_empty() { + println!( + "OK {} stored quer{} type-check against the schema", + registry.len(), + if registry.len() == 1 { "y" } else { "ies" } + ); + } + for issue in &output.breakages { + println!("ERROR query '{}': {}", issue.query, issue.message); + } + for issue in &output.warnings { + println!("WARN query '{}': {}", issue.query, issue.message); + } + } + + if report.has_breakages() { + io::stdout().flush()?; + std::process::exit(1); + } + Ok(()) +} + +fn execute_queries_list( + target: Option, + config_path: Option<&PathBuf>, + json: bool, +) -> Result<()> { + let config = load_cli_config(config_path)?; + let selected = resolve_registry_selection_for_list(&config, target.as_deref())?; + let registry = load_registry_or_report(&config, selected.as_deref())?; + + let output = QueriesListOutput { + queries: registry + .iter() + .map(|q| QueriesListItem { + name: q.name.clone(), + mcp_expose: q.expose, + tool_name: q.tool_name.clone(), + mutation: q.is_mutation(), + params: q + .decl + .params + .iter() + .map(|p| QueriesParam { + name: p.name.clone(), + type_name: p.type_name.clone(), + nullable: p.nullable, + }) + .collect(), + }) + .collect(), + }; + + if json { + print_json(&output)?; + } else if output.queries.is_empty() { + println!("(no stored queries registered)"); + } else { + for q in &output.queries { + let kind = if q.mutation { "mutation" } else { "read" }; + let params = q + .params + .iter() + .map(|p| { + format!( + "${}: {}{}", + p.name, + p.type_name, + if p.nullable { "?" } else { "" } + ) + }) + .collect::>() + .join(", "); + let mcp = if q.mcp_expose { + format!(" [mcp: {}]", q.tool_name.as_deref().unwrap_or(&q.name)) + } else { + String::new() + }; + println!("{kind} {}({params}){mcp}", q.name); + } + } + Ok(()) +} + async fn execute_read( uri: &str, query_source: &str, @@ -1655,7 +2002,7 @@ async fn execute_read_remote( } async fn execute_change( - uri: &str, + graph: &ResolvedCliGraph, query_source: &str, query_name: Option<&str>, branch: &str, @@ -1665,7 +2012,7 @@ async fn execute_change( ) -> Result { let (selected_name, query_params) = select_named_query(query_source, query_name)?; let params = query_params_from_json(&query_params, params_json)?; - let db = open_local_db_with_policy(uri, config).await?; + let db = open_local_db_with_policy(graph).await?; let actor = resolve_cli_actor(cli_as_actor, config); let result = db .mutate_as(branch, query_source, &selected_name, ¶ms, actor) @@ -1893,9 +2240,10 @@ async fn main() -> Result<()> { json, } => { let config = load_cli_config(config.as_ref())?; - let uri = resolve_local_uri(&config, uri, target.as_deref(), "load")?; + let graph = resolve_local_graph(&config, uri, target.as_deref(), "load")?; + let uri = graph.uri.clone(); let branch = resolve_branch(&config, branch, None, "main"); - let db = open_local_db_with_policy(&uri, &config).await?; + let db = open_local_db_with_policy(&graph).await?; let actor = resolve_cli_actor(cli.as_actor.as_deref(), &config); let result = db .load_file_as(&branch, &data.to_string_lossy(), mode.into(), actor) @@ -1936,10 +2284,11 @@ async fn main() -> Result<()> { let config = load_cli_config(config.as_ref())?; let bearer_token = resolve_remote_bearer_token(&config, uri.as_deref(), target.as_deref())?; - let uri = resolve_uri(&config, uri, target.as_deref())?; + let graph = resolve_cli_graph(&config, uri, target.as_deref())?; + let uri = graph.uri.clone(); let branch = resolve_branch(&config, branch, None, "main"); let from = resolve_branch(&config, from, None, "main"); - let payload = if is_remote_uri(&uri) { + let payload = if graph.is_remote { let data = fs::read_to_string(&data)?; remote_json::( &http_client, @@ -1955,7 +2304,7 @@ async fn main() -> Result<()> { ) .await? } else { - let db = open_local_db_with_policy(&uri, &config).await?; + let db = open_local_db_with_policy(&graph).await?; let actor = resolve_cli_actor(cli.as_actor.as_deref(), &config); let result = db .ingest_file_as( @@ -1986,9 +2335,10 @@ async fn main() -> Result<()> { let config = load_cli_config(config.as_ref())?; let bearer_token = resolve_remote_bearer_token(&config, uri.as_deref(), target.as_deref())?; - let uri = resolve_uri(&config, uri, target.as_deref())?; + let graph = resolve_cli_graph(&config, uri, target.as_deref())?; + let uri = graph.uri.clone(); let from = resolve_branch(&config, from, None, "main"); - let payload = if is_remote_uri(&uri) { + let payload = if graph.is_remote { remote_json::( &http_client, Method::POST, @@ -2001,7 +2351,7 @@ async fn main() -> Result<()> { ) .await? } else { - let db = open_local_db_with_policy(&uri, &config).await?; + let db = open_local_db_with_policy(&graph).await?; let actor = resolve_cli_actor(cli.as_actor.as_deref(), &config); db.branch_create_from_as(ReadTarget::branch(&from), &name, actor) .await?; @@ -2027,8 +2377,9 @@ async fn main() -> Result<()> { let config = load_cli_config(config.as_ref())?; let bearer_token = resolve_remote_bearer_token(&config, uri.as_deref(), target.as_deref())?; - let uri = resolve_uri(&config, uri, target.as_deref())?; - let payload = if is_remote_uri(&uri) { + let graph = resolve_cli_graph(&config, uri, target.as_deref())?; + let uri = graph.uri.clone(); + let payload = if graph.is_remote { remote_json::( &http_client, Method::GET, @@ -2061,8 +2412,9 @@ async fn main() -> Result<()> { let config = load_cli_config(config.as_ref())?; let bearer_token = resolve_remote_bearer_token(&config, uri.as_deref(), target.as_deref())?; - let uri = resolve_uri(&config, uri, target.as_deref())?; - let payload = if is_remote_uri(&uri) { + let graph = resolve_cli_graph(&config, uri, target.as_deref())?; + let uri = graph.uri.clone(); + let payload = if graph.is_remote { remote_json::( &http_client, Method::DELETE, @@ -2072,7 +2424,7 @@ async fn main() -> Result<()> { ) .await? } else { - let db = open_local_db_with_policy(&uri, &config).await?; + let db = open_local_db_with_policy(&graph).await?; let actor = resolve_cli_actor(cli.as_actor.as_deref(), &config); db.branch_delete_as(&name, actor).await?; BranchDeleteOutput { @@ -2098,9 +2450,10 @@ async fn main() -> Result<()> { let config = load_cli_config(config.as_ref())?; let bearer_token = resolve_remote_bearer_token(&config, uri.as_deref(), target.as_deref())?; - let uri = resolve_uri(&config, uri, target.as_deref())?; + let graph = resolve_cli_graph(&config, uri, target.as_deref())?; + let uri = graph.uri.clone(); let into = resolve_branch(&config, into, None, "main"); - let payload = if is_remote_uri(&uri) { + let payload = if graph.is_remote { remote_json::( &http_client, Method::POST, @@ -2113,7 +2466,7 @@ async fn main() -> Result<()> { ) .await? } else { - let db = open_local_db_with_policy(&uri, &config).await?; + let db = open_local_db_with_policy(&graph).await?; let actor = resolve_cli_actor(cli.as_actor.as_deref(), &config); let outcome = db.branch_merge_as(&source, &into, actor).await?; BranchMergeOutput { @@ -2248,9 +2601,10 @@ async fn main() -> Result<()> { let config = load_cli_config(config.as_ref())?; let bearer_token = resolve_remote_bearer_token(&config, uri.as_deref(), target.as_deref())?; - let uri = resolve_uri(&config, uri, target.as_deref())?; + let graph = resolve_cli_graph(&config, uri, target.as_deref())?; + let uri = graph.uri.clone(); let schema_source = fs::read_to_string(&schema)?; - let output = if is_remote_uri(&uri) { + let output = if graph.is_remote { // MR-694 PR B: SchemaApplyRequest gained an // allow_data_loss field so Hard-mode drops are no // longer CLI-only. The previous bail is gone; the @@ -2268,13 +2622,22 @@ async fn main() -> Result<()> { ) .await? } else { - let db = open_local_db_with_policy(&uri, &config).await?; + let db = open_local_db_with_policy(&graph).await?; let actor = resolve_cli_actor(cli.as_actor.as_deref(), &config); + let registry = load_registry_or_report(&config, graph.selected())?; + let registry = (!registry.is_empty()).then_some(registry); + let label = graph.selected().unwrap_or(&uri).to_string(); let result = db - .apply_schema_as( + .apply_schema_as_with_catalog_check( &schema_source, omnigraph::db::SchemaApplyOptions { allow_data_loss }, actor, + |catalog| { + if let Some(registry) = registry.as_ref() { + validate_registry_for_catalog(registry, catalog, &label)?; + } + Ok(()) + }, ) .await?; schema_apply_output(&uri, result) @@ -2331,6 +2694,23 @@ async fn main() -> Result<()> { .await?; finish_query_lint(&output, json)?; } + Command::Queries { command } => match command { + QueriesCommand::Validate { + uri, + target, + config, + json, + } => { + execute_queries_validate(uri, target, config.as_ref(), json).await?; + } + QueriesCommand::List { + target, + config, + json, + } => { + execute_queries_list(target, config.as_ref(), json)?; + } + }, Command::Snapshot { uri, target, @@ -2436,7 +2816,8 @@ async fn main() -> Result<()> { .as_deref() .or_else(|| alias_config.and_then(|alias| alias.graph.as_deref())); let bearer_token = resolve_remote_bearer_token(&config, uri.as_deref(), target_name)?; - let uri = resolve_uri(&config, uri, target_name)?; + let graph = resolve_cli_graph(&config, uri, target_name)?; + let uri = graph.uri.clone(); let query_source = resolve_query_source( &config, query.as_ref(), @@ -2458,7 +2839,7 @@ async fn main() -> Result<()> { alias_config.and_then(|alias| alias.branch.clone()), )?; let query_name = name.or_else(|| alias_config.and_then(|alias| alias.name.clone())); - let output = if is_remote_uri(&uri) { + let output = if graph.is_remote { execute_read_remote( &http_client, &uri, @@ -2521,7 +2902,8 @@ async fn main() -> Result<()> { .as_deref() .or_else(|| alias_config.and_then(|alias| alias.graph.as_deref())); let bearer_token = resolve_remote_bearer_token(&config, uri.as_deref(), target_name)?; - let uri = resolve_uri(&config, uri, target_name)?; + let graph = resolve_cli_graph(&config, uri, target_name)?; + let uri = graph.uri.clone(); let query_source = resolve_query_source( &config, query.as_ref(), @@ -2543,7 +2925,7 @@ async fn main() -> Result<()> { "main", ); let query_name = name.or_else(|| alias_config.and_then(|alias| alias.name.clone())); - let output = if is_remote_uri(&uri) { + let output = if graph.is_remote { execute_change_remote( &http_client, &uri, @@ -2556,7 +2938,7 @@ async fn main() -> Result<()> { .await? } else { execute_change( - &uri, + &graph, &query_source, query_name.as_deref(), &branch, @@ -2575,20 +2957,19 @@ async fn main() -> Result<()> { Command::Policy { command } => match command { PolicyCommand::Validate { config } => { let config = load_cli_config(config.as_ref())?; - let engine = resolve_policy_engine(&config)?; - let policy_file = config - .resolve_policy_file() - .expect("policy file should exist after resolve_policy_engine"); + let context = resolve_policy_context(&config)?; + let engine = resolve_policy_engine(&context)?; println!( "policy valid: {} [{} actors]", - policy_file.display(), + context.policy_file.display(), engine.known_actor_count() ); } PolicyCommand::Test { config } => { let config = load_cli_config(config.as_ref())?; - let engine = resolve_policy_engine(&config)?; - let tests_path = resolve_policy_tests_path(&config)?; + let context = resolve_policy_context(&config)?; + let engine = resolve_policy_engine(&context)?; + let tests_path = resolve_policy_tests_path(&context); let tests = PolicyTestConfig::load(&tests_path)?; engine.run_tests(&tests)?; println!("policy tests passed: {} cases", tests.cases.len()); @@ -2601,7 +2982,8 @@ async fn main() -> Result<()> { target_branch, } => { let config = load_cli_config(config.as_ref())?; - let engine = resolve_policy_engine(&config)?; + let context = resolve_policy_context(&config)?; + let engine = resolve_policy_engine(&context)?; let request = PolicyRequest { action, branch, @@ -2774,7 +3156,8 @@ mod tests { use super::{ DEFAULT_BEARER_TOKEN_ENV, apply_bearer_token, bearer_token_from_env_file, legacy_change_request_body, load_cli_config, load_env_file_into_process, - normalize_bearer_token, parse_env_assignment, resolve_remote_bearer_token, + normalize_bearer_token, parse_env_assignment, resolve_policy_context, + resolve_cli_graph, resolve_remote_bearer_token, }; use omnigraph_server::load_config; use reqwest::header::AUTHORIZATION; @@ -3034,4 +3417,150 @@ graphs: } } } + + #[test] + fn graph_identity_resolve_policy_context_named_cli_graph_uses_graph_key_not_project_name_or_uri() { + let temp = tempdir().unwrap(); + let config_path = temp.path().join("omnigraph.yaml"); + fs::write( + &config_path, + r#" +project: + name: misleading-project +graphs: + local: + uri: /tmp/local-policy-graph.omni + policy: + file: ./policy.yaml +cli: + graph: local +"#, + ) + .unwrap(); + + let config = load_config(Some(&config_path)).unwrap(); + let context = resolve_policy_context(&config).unwrap(); + assert_eq!(context.graph_id, "local"); + } + + #[test] + fn graph_identity_resolve_policy_context_server_graph_uses_graph_key_when_cli_graph_absent() { + let temp = tempdir().unwrap(); + let config_path = temp.path().join("omnigraph.yaml"); + fs::write( + &config_path, + r#" +project: + name: misleading-project +graphs: + local: + uri: /tmp/local-policy-graph.omni + policy: + file: ./server-policy.yaml +server: + graph: local +"#, + ) + .unwrap(); + + let config = load_config(Some(&config_path)).unwrap(); + let context = resolve_policy_context(&config).unwrap(); + assert_eq!(context.graph_id, "local"); + assert!(context.policy_file.ends_with("server-policy.yaml")); + } + + #[test] + fn graph_identity_resolve_policy_context_anonymous_uses_top_level_default_identity() { + let temp = tempdir().unwrap(); + let config_path = temp.path().join("omnigraph.yaml"); + fs::write( + &config_path, + r#" +project: + name: misleading-project +graphs: + local: + uri: /tmp/local-policy-graph.omni +policy: + file: ./top-policy.yaml +"#, + ) + .unwrap(); + + let config = load_config(Some(&config_path)).unwrap(); + let context = resolve_policy_context(&config).unwrap(); + assert_eq!(context.graph_id, "default"); + assert!(context.policy_file.ends_with("top-policy.yaml")); + } + + #[test] + fn graph_identity_resolve_cli_graph_named_target_uses_graph_key_not_project_name_or_uri() { + let temp = tempdir().unwrap(); + let config_path = temp.path().join("omnigraph.yaml"); + fs::write( + &config_path, + r#" +project: + name: misleading-project +graphs: + prod: + uri: s3://bucket/prod-graph/ + policy: + file: ./prod-policy.yaml +"#, + ) + .unwrap(); + + let config = load_config(Some(&config_path)).unwrap(); + let graph = resolve_cli_graph(&config, None, Some("prod")).unwrap(); + assert_eq!(graph.selected(), Some("prod")); + assert_eq!(graph.graph_id, "prod"); + assert_eq!(graph.uri, "s3://bucket/prod-graph/"); + } + + #[test] + fn graph_identity_resolve_cli_graph_positional_uri_uses_anonymous_normalized_uri() { + let temp = tempdir().unwrap(); + let config_path = temp.path().join("omnigraph.yaml"); + fs::write( + &config_path, + r#" +project: + name: misleading-project +graphs: + local: + uri: /tmp/configured-graph.omni + policy: + file: ./policy.yaml +cli: + graph: local +"#, + ) + .unwrap(); + + let config = load_config(Some(&config_path)).unwrap(); + let local_graph_path = temp.path().join("explicit-graph.omni"); + let local_graph = resolve_cli_graph( + &config, + Some(format!("file://{}", local_graph_path.display())), + None, + ) + .unwrap(); + assert_eq!(local_graph.selected(), None); + assert_eq!( + local_graph.graph_id, + local_graph_path.to_string_lossy().as_ref() + ); + assert_eq!(local_graph.policy_file, None); + + let s3_graph = resolve_cli_graph( + &config, + Some("s3://bucket/anonymous-graph/".to_string()), + None, + ) + .unwrap(); + assert_eq!(s3_graph.selected(), None); + assert_eq!(s3_graph.graph_id, "s3://bucket/anonymous-graph"); + assert_eq!(s3_graph.policy_file, None); + } } diff --git a/crates/omnigraph-cli/tests/cli.rs b/crates/omnigraph-cli/tests/cli.rs index 6e5de37..9682d9a 100644 --- a/crates/omnigraph-cli/tests/cli.rs +++ b/crates/omnigraph-cli/tests/cli.rs @@ -2376,3 +2376,295 @@ fn graphs_list_against_local_uri_errors_with_remote_only_message() { "expected 'remote multi-graph server URL' rejection in stderr; got:\n{stderr}" ); } + +fn queries_test_config(graph_uri: &str, entry: &str, gq_file: &str) -> String { + format!( + "graphs:\n local:\n uri: '{}'\n queries:\n {entry}:\n file: ./{gq_file}\n\ + cli:\n graph: local\npolicy: {{}}\n", + graph_uri.replace('\'', "''") + ) +} + +#[test] +fn queries_validate_exits_zero_on_clean_registry() { + let graph = SystemGraph::loaded(); + graph.write_query( + "find_person.gq", + "query find_person($name: String) { match { $p: Person { name: $name } } return { $p.age } }", + ); + let config = graph.write_config( + "omnigraph.yaml", + &queries_test_config(&graph.path().to_string_lossy(), "find_person", "find_person.gq"), + ); + let output = output_success(cli().arg("queries").arg("validate").arg("--config").arg(&config)); + let stdout = stdout_string(&output); + assert!(stdout.contains("OK"), "stdout:\n{stdout}"); +} + +#[test] +fn queries_validate_exits_nonzero_on_type_broken_query() { + let graph = SystemGraph::loaded(); + // `Widget` is not in the fixture schema. + graph.write_query("ghost.gq", "query ghost() { match { $w: Widget } return { $w.name } }"); + let config = graph.write_config( + "omnigraph.yaml", + &queries_test_config(&graph.path().to_string_lossy(), "ghost", "ghost.gq"), + ); + let output = output_failure(cli().arg("queries").arg("validate").arg("--config").arg(&config)); + let stdout = stdout_string(&output); + assert!( + stdout.contains("ghost"), + "validation should name the broken query; stdout:\n{stdout}" + ); +} + +#[test] +fn queries_list_prints_registered_query() { + let graph = SystemGraph::loaded(); + graph.write_query( + "find_person.gq", + "query find_person($name: String) { match { $p: Person { name: $name } } return { $p.age } }", + ); + // Exposed with an explicit tool name so the list shows the MCP suffix. + let config = graph.write_config( + "omnigraph.yaml", + &format!( + concat!( + "graphs:\n", + " local:\n", + " uri: '{}'\n", + " queries:\n", + " find_person:\n", + " file: ./find_person.gq\n", + " mcp: {{ expose: true, tool_name: lookup_person }}\n", + "cli:\n", + " graph: local\n", + "policy: {{}}\n", + ), + graph.path().to_string_lossy().replace('\'', "''") + ), + ); + let output = output_success(cli().arg("queries").arg("list").arg("--config").arg(&config)); + let stdout = stdout_string(&output); + assert!(stdout.contains("find_person"), "stdout:\n{stdout}"); + assert!( + stdout.contains("$name: String"), + "list should show typed params; stdout:\n{stdout}" + ); + assert!( + stdout.contains("[mcp: lookup_person]"), + "list should show the MCP tool name for exposed queries; stdout:\n{stdout}" + ); +} + +#[test] +fn queries_list_requires_graph_selection_for_per_graph_only_registries() { + let graph = SystemGraph::loaded(); + graph.write_query( + "find_person.gq", + "query find_person($name: String) { match { $p: Person { name: $name } } return { $p.age } }", + ); + let config = graph.write_config( + "omnigraph.yaml", + &format!( + concat!( + "graphs:\n", + " local:\n", + " uri: '{}'\n", + " queries:\n", + " find_person:\n", + " file: ./find_person.gq\n", + "policy: {{}}\n", + ), + graph.path().to_string_lossy().replace('\'', "''") + ), + ); + + let output = output_failure(cli().arg("queries").arg("list").arg("--config").arg(&config)); + let stderr = String::from_utf8_lossy(&output.stderr); + assert!( + stderr.contains("local") && stderr.contains("--target local"), + "error must name the graph and give a concrete selection hint; stderr:\n{stderr}" + ); +} + +#[test] +fn queries_list_without_graph_selection_lists_top_level_registry() { + let graph = SystemGraph::loaded(); + graph.write_query( + "top_find.gq", + "query top_find($name: String) { match { $p: Person { name: $name } } return { $p.age } }", + ); + let config = graph.write_config( + "omnigraph.yaml", + concat!( + "queries:\n", + " top_find:\n", + " file: ./top_find.gq\n", + "policy: {}\n", + ), + ); + + let output = output_success(cli().arg("queries").arg("list").arg("--config").arg(&config)); + let stdout = stdout_string(&output); + assert!(stdout.contains("top_find"), "stdout:\n{stdout}"); +} + +#[test] +fn queries_list_unknown_target_errors() { + // `queries list` opens no graph URI, so unknown-graph validation can't ride + // along on URI resolution the way it does for every other command. An + // unknown `--target` must still error (naming the graph) instead of + // silently falling back to the top-level registry and showing the wrong + // (or empty) catalog. + let graph = SystemGraph::loaded(); + graph.write_query( + "find_person.gq", + "query find_person($name: String) { match { $p: Person { name: $name } } return { $p.age } }", + ); + let config = graph.write_config( + "omnigraph.yaml", + &queries_test_config(&graph.path().to_string_lossy(), "find_person", "find_person.gq"), + ); + let output = output_failure( + cli() + .arg("queries") + .arg("list") + .arg("--target") + .arg("nonexistent") + .arg("--config") + .arg(&config), + ); + let stderr = String::from_utf8_lossy(&output.stderr); + assert!( + stderr.contains("nonexistent"), + "error must name the unknown graph; stderr:\n{stderr}" + ); +} + +#[test] +fn queries_commands_reject_named_graph_with_populated_top_level_block() { + // A named graph (here via `cli.graph`) uses its own `graphs.` block, + // so a populated top-level `queries:` block would be silently ignored — a + // config the server REFUSES to boot. `queries validate`/`list` must reject + // it too (matching boot) instead of validating/listing the per-graph block + // and giving a false green. + let graph = SystemGraph::loaded(); + graph.write_query( + "find_person.gq", + "query find_person($name: String) { match { $p: Person { name: $name } } return { $p.age } }", + ); + let config = graph.write_config( + "omnigraph.yaml", + &format!( + concat!( + "graphs:\n", + " local:\n", + " uri: '{}'\n", + " queries:\n", + " find_person:\n", + " file: ./find_person.gq\n", + "cli:\n", + " graph: local\n", + "queries:\n", // populated top-level block: the coherence violation + " legacy:\n", + " file: ./legacy.gq\n", + "policy: {{}}\n", + ), + graph.path().to_string_lossy().replace('\'', "''") + ), + ); + // Both resolve `local` from cli.graph (no positional URI), so both must + // error and name the graph + the ignored block — like server boot does. + for sub in ["validate", "list"] { + let output = output_failure(cli().arg("queries").arg(sub).arg("--config").arg(&config)); + let stderr = String::from_utf8_lossy(&output.stderr); + assert!( + stderr.contains("local") && stderr.contains("queries"), + "`queries {sub}` must reject a named graph with a populated top-level block; stderr:\n{stderr}" + ); + } +} + +#[test] +fn queries_validate_exits_nonzero_on_duplicate_tool_name() { + // Two exposed queries claiming one MCP tool name is a load-time + // collision — `queries validate` must fail (offline, before the engine + // opens) and name both queries plus the contested tool. + let graph = SystemGraph::loaded(); + graph.write_query("a.gq", "query a() { match { $p: Person } return { $p.name } }"); + graph.write_query("b.gq", "query b() { match { $p: Person } return { $p.name } }"); + let config = graph.write_config( + "omnigraph.yaml", + &format!( + concat!( + "graphs:\n", + " local:\n", + " uri: '{}'\n", + " queries:\n", + " a:\n", + " file: ./a.gq\n", + " mcp: {{ expose: true, tool_name: dup }}\n", + " b:\n", + " file: ./b.gq\n", + " mcp: {{ expose: true, tool_name: dup }}\n", + "cli:\n", + " graph: local\n", + "policy: {{}}\n", + ), + graph.path().to_string_lossy().replace('\'', "''") + ), + ); + let output = output_failure(cli().arg("queries").arg("validate").arg("--config").arg(&config)); + let stderr = String::from_utf8_lossy(&output.stderr); + assert!( + stderr.contains("dup") && stderr.contains("'a'") && stderr.contains("'b'"), + "duplicate tool name should be reported naming both queries; stderr:\n{stderr}" + ); +} + +#[test] +fn queries_validate_positional_uri_ignores_default_graph() { + // A positional URI is anonymous → the schema AND the registry both come + // from top-level, even when `cli.graph` names a graph whose per-graph + // queries would fail. Pins that the URI and registry can't diverge. + let graph = SystemGraph::loaded(); + graph.write_query( + "clean.gq", + "query clean($name: String) { match { $p: Person { name: $name } } return { $p.age } }", + ); + // `Widget` is not in the fixture schema — the default graph's per-graph + // query would break validate if it were (wrongly) selected. + graph.write_query("broken.gq", "query broken() { match { $w: Widget } return { $w.name } }"); + let config = graph.write_config( + "omnigraph.yaml", + concat!( + "cli:\n graph: prod\n", + "graphs:\n", + " prod:\n", + " uri: /nonexistent-prod.omni\n", + " queries:\n", + " broken:\n", + " file: ./broken.gq\n", + "queries:\n", + " clean:\n", + " file: ./clean.gq\n", + "policy: {}\n", + ), + ); + // Positional URI = the real loaded graph; selection is anonymous, so the + // CLEAN top-level registry validates (not prod's broken one). + let output = output_success( + cli() + .arg("queries") + .arg("validate") + .arg(graph.path()) + .arg("--config") + .arg(&config), + ); + let stdout = stdout_string(&output); + assert!( + stdout.contains("OK"), + "positional URI must validate the top-level registry, not the cli.graph default; stdout:\n{stdout}" + ); +} diff --git a/crates/omnigraph-cli/tests/system_local.rs b/crates/omnigraph-cli/tests/system_local.rs index 08f653d..4fc3e9a 100644 --- a/crates/omnigraph-cli/tests/system_local.rs +++ b/crates/omnigraph-cli/tests/system_local.rs @@ -74,14 +74,36 @@ project: graphs: local: uri: {} + policy: + file: ./policy.yaml cli: graph: local branch: main query: roots: - . -policy: - file: ./policy.yaml +", + yaml_string(&graph.path().to_string_lossy()) + ) +} + +fn local_policy_server_graph_config(graph: &SystemGraph) -> String { + format!( + "\ +project: + name: policy-e2e-local +graphs: + local: + uri: {} + policy: + file: ./policy.yaml +server: + graph: local +cli: + branch: main +query: + roots: + - . ", yaml_string(&graph.path().to_string_lossy()) ) @@ -1000,49 +1022,55 @@ query vector_search($q: String) { #[test] fn local_cli_policy_tooling_is_end_to_end() { // Sanity check for the read-only policy CLI surfaces. These don't - // mutate the graph — they just parse and evaluate the policy file — - // so they don't depend on PR #4's engine-side enforcement. + // mutate the graph; they parse and evaluate the effective policy for + // named graph selections, including per-graph policy files. let graph = SystemGraph::loaded(); let config = graph.write_config("omnigraph-policy.yaml", &local_policy_config(&graph)); + let server_graph_config = graph.write_config( + "omnigraph-policy-server.yaml", + &local_policy_server_graph_config(&graph), + ); graph.write_config("policy.yaml", POLICY_E2E_YAML); graph.write_config("policy.tests.yaml", POLICY_E2E_TESTS_YAML); - let validate = output_success( - cli() - .arg("policy") - .arg("validate") - .arg("--config") - .arg(&config), - ); - assert!(stdout_string(&validate).contains("policy valid:")); + for config in [&config, &server_graph_config] { + let validate = output_success( + cli() + .arg("policy") + .arg("validate") + .arg("--config") + .arg(config), + ); + assert!(stdout_string(&validate).contains("policy valid:")); - let tests = output_success(cli().arg("policy").arg("test").arg("--config").arg(&config)); - assert!(stdout_string(&tests).contains("policy tests passed: 2 cases")); + let tests = output_success(cli().arg("policy").arg("test").arg("--config").arg(config)); + assert!(stdout_string(&tests).contains("policy tests passed: 2 cases")); - let explain = output_success( - cli() - .arg("policy") - .arg("explain") - .arg("--config") - .arg(&config) - .arg("--actor") - .arg("act-bruno") - .arg("--action") - .arg("change") - .arg("--branch") - .arg("main"), - ); - let explain_stdout = stdout_string(&explain); - assert!(explain_stdout.contains("decision: deny")); - assert!(explain_stdout.contains("branch: main")); + let explain = output_success( + cli() + .arg("policy") + .arg("explain") + .arg("--config") + .arg(config) + .arg("--actor") + .arg("act-bruno") + .arg("--action") + .arg("change") + .arg("--branch") + .arg("main"), + ); + let explain_stdout = stdout_string(&explain); + assert!(explain_stdout.contains("decision: deny")); + assert!(explain_stdout.contains("branch: main")); + } } #[test] fn local_cli_change_enforces_engine_layer_policy() { - // Asserts MR-722 PR #4: when `policy.file` is configured in - // `omnigraph.yaml`, the CLI loads PolicyEngine into Omnigraph and - // every direct-engine write hits `enforce(action, scope, actor)` — - // identical to what the HTTP server gets, regardless of transport. + // Asserts MR-722 PR #4: when the selected graph has a configured + // policy file, the CLI loads PolicyEngine into Omnigraph and every + // direct-engine write hits `enforce(action, scope, actor)` — identical + // to what the HTTP server gets, regardless of transport. // // Three cases, each discriminating: // @@ -1135,6 +1163,32 @@ fn local_cli_change_enforces_engine_layer_policy() { assert_eq!(verify["rows"][0]["p.name"], "RagnorOnMain"); } +#[test] +fn local_cli_positional_uri_does_not_inherit_default_graph_policy() { + let graph = SystemGraph::loaded(); + let config = graph.write_config("omnigraph-policy.yaml", &local_policy_config(&graph)); + graph.write_config("policy.yaml", POLICY_E2E_YAML); + let mutation_file = insert_person_query(&graph, "system-local-policy-positional.gq"); + + let allowed = parse_stdout_json(&output_success( + cli() + .arg("--as") + .arg("act-bruno") + .arg("change") + .arg("--config") + .arg(&config) + .arg("--uri") + .arg(graph.path()) + .arg("--query") + .arg(&mutation_file) + .arg("--params") + .arg(r#"{"name":"PositionalUriBruno","age":4}"#) + .arg("--json"), + )); + assert_eq!(allowed["affected_nodes"], 1); + assert_eq!(allowed["actor_id"], "act-bruno"); +} + // ─── MR-722 PR A: CLI×writer matrix ─────────────────────────────────────── // // The change writer is covered above by `local_cli_change_enforces_engine_layer_policy`. @@ -1293,6 +1347,62 @@ fn local_cli_schema_apply_enforces_engine_layer_policy() { assert_eq!(allowed["applied"], true); } +#[test] +fn local_cli_schema_apply_rejects_stored_query_breakage_before_publish() { + let graph = SystemGraph::loaded(); + graph.write_query( + "stored-find-person.gq", + "query find_person($name: String) { match { $p: Person { name: $name } } return { $p.age } }", + ); + let config = graph.write_config( + "omnigraph-stored-query-schema.yaml", + &format!( + "\ +graphs: + local: + uri: {} + queries: + find_person: + file: ./stored-find-person.gq +cli: + graph: local + branch: main +query: + roots: + - . +policy: {{}} +", + yaml_string(&graph.path().to_string_lossy()) + ), + ); + let renamed_schema = std::fs::read_to_string(fixture("test.pg")) + .unwrap() + .replace("age: I32?", "years: I32? @rename_from(\"age\")"); + let schema_path = graph.write_file("stored-query-breaks.pg", &renamed_schema); + + let rejected = output_failure( + cli() + .arg("schema") + .arg("apply") + .arg("--config") + .arg(&config) + .arg("--schema") + .arg(&schema_path) + .arg("--json"), + ); + let stderr = String::from_utf8_lossy(&rejected.stderr); + assert!( + stderr.contains("find_person") && stderr.contains("schema check"), + "schema apply should reject the stored-query breakage before publish; stderr: {stderr}" + ); + + let schema = stdout_string(&output_success( + cli().arg("schema").arg("show").arg("--config").arg(&config), + )); + assert!(schema.contains("age: I32?")); + assert!(!schema.contains("years: I32?")); +} + #[test] fn local_cli_branch_create_enforces_engine_layer_policy() { let graph = SystemGraph::loaded(); @@ -1448,6 +1558,8 @@ project: graphs: local: uri: {} + policy: + file: ./policy.yaml cli: graph: local branch: main @@ -1455,8 +1567,6 @@ cli: query: roots: - . -policy: - file: ./policy.yaml ", yaml_string(&graph.path().to_string_lossy()), actor, diff --git a/crates/omnigraph-cli/tests/system_remote.rs b/crates/omnigraph-cli/tests/system_remote.rs index c86e32e..45bf502 100644 --- a/crates/omnigraph-cli/tests/system_remote.rs +++ b/crates/omnigraph-cli/tests/system_remote.rs @@ -60,10 +60,10 @@ project: graphs: local: uri: {} + policy: + file: ./policy.yaml server: graph: local -policy: - file: ./policy.yaml ", yaml_string(&graph.path().to_string_lossy()) ) diff --git a/crates/omnigraph-policy/src/lib.rs b/crates/omnigraph-policy/src/lib.rs index 6459fcd..cb59796 100644 --- a/crates/omnigraph-policy/src/lib.rs +++ b/crates/omnigraph-policy/src/lib.rs @@ -56,6 +56,21 @@ pub enum PolicyAction { /// from v0.6.0; operators add and remove graphs by editing /// `omnigraph.yaml` and restarting. GraphList, + /// Gates invoking a server-side stored query by name. Per-graph and + /// **graph-scoped** (no branch dimension, like `Admin`): the per-branch + /// access of the query body is enforced by the inner `Read`/`Change` + /// gate, so branch-scoping this outer gate would be redundant (and was + /// wrong for snapshot reads). A rule that sets `branch_scope` on + /// `invoke_query` is rejected by `validate()`. In this release it is + /// **coarse**: an `invoke_query` allow rule permits *any* stored query + /// on the graph (no per-query dimension yet); a future, additive + /// refinement adds an optional query-name scope. + /// + /// This gate sits at the HTTP boundary. The engine `_as` writers still + /// enforce `Read`/`Change` per the query body, so a stored *mutation* + /// is double-gated: `invoke_query` to reach the tool, plus `change` for + /// the write itself. + InvokeQuery, } impl PolicyAction { @@ -70,6 +85,7 @@ impl PolicyAction { Self::BranchMerge => "branch_merge", Self::Admin => "admin", Self::GraphList => "graph_list", + Self::InvokeQuery => "invoke_query", } } @@ -99,7 +115,8 @@ impl PolicyAction { | Self::BranchCreate | Self::BranchDelete | Self::BranchMerge - | Self::Admin => PolicyResourceKind::Graph, + | Self::Admin + | Self::InvokeQuery => PolicyResourceKind::Graph, } } } @@ -155,6 +172,7 @@ impl FromStr for PolicyAction { "branch_merge" => Ok(Self::BranchMerge), "admin" => Ok(Self::Admin), "graph_list" => Ok(Self::GraphList), + "invoke_query" => Ok(Self::InvokeQuery), other => bail!("unknown policy action '{other}'"), } } @@ -806,6 +824,7 @@ namespace Omnigraph { action "branch_delete" appliesTo { principal: Actor, resource: Graph, context: RequestContext }; action "branch_merge" appliesTo { principal: Actor, resource: Graph, context: RequestContext }; action "admin" appliesTo { principal: Actor, resource: Graph, context: RequestContext }; + action "invoke_query" appliesTo { principal: Actor, resource: Graph, context: RequestContext }; action "graph_list" appliesTo { principal: Actor, resource: Server, context: RequestContext }; } @@ -1264,6 +1283,80 @@ rules: assert!(!deny.allowed); } + #[test] + fn invoke_query_authorizes_per_graph() { + let policy: PolicyConfig = serde_yaml::from_str( + r#" +version: 1 +groups: + team: [act-alice] + others: [act-bruno] +rules: + - id: team-invoke-queries + allow: + actors: { group: team } + actions: [invoke_query] +"#, + ) + .unwrap(); + let engine = PolicyCompiler::compile(&policy, "graph").unwrap(); + + let allow = engine + .authorize( + "act-alice", + &PolicyRequest { + action: PolicyAction::InvokeQuery, + branch: None, + target_branch: None, + }, + ) + .unwrap(); + assert!(allow.allowed); + assert_eq!( + allow.matched_rule_id.as_deref(), + Some("team-invoke-queries") + ); + + // Actor outside the group → deny. + let deny = engine + .authorize( + "act-bruno", + &PolicyRequest { + action: PolicyAction::InvokeQuery, + branch: None, + target_branch: None, + }, + ) + .unwrap(); + assert!(!deny.allowed); + } + + #[test] + fn invoke_query_rejects_branch_scope() { + // invoke_query is graph-scoped (like admin) — per-branch access is + // enforced by the inner read/change gate — so a rule that puts a + // `branch_scope` qualifier on it is rejected at validate(). + let policy: PolicyConfig = serde_yaml::from_str( + r#" +version: 1 +groups: + team: [act-alice] +rules: + - id: team-invoke-any-branch + allow: + actors: { group: team } + actions: [invoke_query] + branch_scope: any +"#, + ) + .unwrap(); + let err = policy.validate().unwrap_err().to_string(); + assert!( + err.contains("branch_scope") && err.contains("invoke_query"), + "branch_scope on invoke_query must be rejected: {err}" + ); + } + #[test] fn server_scoped_rule_cannot_use_branch_scope() { let policy: PolicyConfig = serde_yaml::from_str( diff --git a/crates/omnigraph-server/src/api.rs b/crates/omnigraph-server/src/api.rs index 2c818ae..4a6024f 100644 --- a/crates/omnigraph-server/src/api.rs +++ b/crates/omnigraph-server/src/api.rs @@ -1,8 +1,11 @@ use omnigraph::db::{GraphCommit, MergeOutcome, ReadTarget, SchemaApplyResult, Snapshot}; use omnigraph::error::{MergeConflict, MergeConflictKind}; use omnigraph::loader::{IngestResult, LoadMode}; +use crate::queries::StoredQuery; use omnigraph_compiler::SchemaMigrationStep; +use omnigraph_compiler::query::ast::Param; use omnigraph_compiler::result::QueryResult; +use omnigraph_compiler::types::{PropType, ScalarType}; use serde::{Deserialize, Serialize}; use serde_json::Value; use utoipa::{IntoParams, ToSchema}; @@ -300,6 +303,162 @@ pub struct ChangeRequest { pub branch: Option, } +/// Body for `POST /queries/{name}` — invokes the server-side stored query +/// named in the path. The query source and name come from the registry, +/// never the body; only the runtime inputs are supplied here. +#[derive(Debug, Clone, Default, Serialize, Deserialize, ToSchema)] +pub struct InvokeStoredQueryRequest { + /// JSON object whose keys match the stored query's declared parameters. + #[serde(default)] + pub params: Option, + /// Branch to run against. Defaults to `main`; for a stored mutation the + /// write targets this branch. + #[serde(default)] + pub branch: Option, + /// Snapshot id to read from (read queries only — rejected for a stored + /// mutation). Mutually exclusive with `branch`. + #[serde(default)] + pub snapshot: Option, +} + +/// Response for `POST /queries/{name}`: the read envelope for a stored +/// read, or the mutation envelope for a stored mutation. Serialized +/// **untagged**, so the wire shape is exactly [`ReadOutput`] or +/// [`ChangeOutput`] — classification follows the stored query, not a +/// wrapper field. +#[derive(Debug, Serialize, ToSchema)] +#[serde(untagged)] +pub enum InvokeStoredQueryResponse { + Read(ReadOutput), + Change(ChangeOutput), +} + +/// The kind of a stored-query parameter, decomposed so a client (e.g. an +/// MCP server) can build a typed input schema with a closed `match` and +/// never re-parse omnigraph's type spelling. `bigint`/`date`/`datetime`/ +/// `blob` are carried as JSON strings on the wire: a 64-bit integer past +/// 2^53 loses precision as a JSON number, and Date/DateTime are ISO +/// strings, Blob a blob-URI string. +#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize, ToSchema)] +#[serde(rename_all = "snake_case")] +pub enum ParamKind { + String, + Bool, + Int, + #[serde(rename = "bigint")] + BigInt, + Float, + Date, + #[serde(rename = "datetime")] + DateTime, + Blob, + Vector, + List, +} + +/// One declared parameter of a stored query, projected for the catalog. +#[derive(Debug, Clone, Serialize, Deserialize, ToSchema)] +pub struct ParamDescriptor { + pub name: String, + pub kind: ParamKind, + /// Element kind when `kind == list` (always a scalar — the grammar + /// forbids lists of vectors or nested lists). + #[serde(skip_serializing_if = "Option::is_none")] + pub item_kind: Option, + /// Dimension when `kind == vector`. + #[serde(skip_serializing_if = "Option::is_none")] + pub vector_dim: Option, + /// `false` → the caller must supply it; `true` → optional. + pub nullable: bool, +} + +/// One entry in the stored-query catalog (`GET /queries`). +#[derive(Debug, Clone, Serialize, Deserialize, ToSchema)] +pub struct QueryCatalogEntry { + /// Registry key / invoke path segment (`POST /queries/{name}`). + pub name: String, + /// MCP tool id (the `tool_name` override, else `name`). + pub tool_name: String, + #[serde(skip_serializing_if = "Option::is_none")] + pub description: Option, + #[serde(skip_serializing_if = "Option::is_none")] + pub instruction: Option, + /// `true` for a stored mutation → an MCP read-only hint of `false`. + pub mutation: bool, + pub params: Vec, +} + +/// Response for `GET /queries`: the `mcp.expose` subset of a graph's +/// stored-query registry, each with typed parameters. +#[derive(Debug, Clone, Serialize, Deserialize, ToSchema)] +pub struct QueriesCatalogOutput { + pub queries: Vec, +} + +/// Total map from a resolved scalar to its catalog kind. Exhaustive on +/// purpose: a new `ScalarType` is a compile error here until catalogued. +fn scalar_kind(scalar: ScalarType) -> ParamKind { + match scalar { + ScalarType::String => ParamKind::String, + ScalarType::Bool => ParamKind::Bool, + ScalarType::I32 | ScalarType::U32 => ParamKind::Int, + ScalarType::I64 | ScalarType::U64 => ParamKind::BigInt, + ScalarType::F32 | ScalarType::F64 => ParamKind::Float, + ScalarType::Date => ParamKind::Date, + ScalarType::DateTime => ParamKind::DateTime, + ScalarType::Blob => ParamKind::Blob, + ScalarType::Vector(_) => ParamKind::Vector, + } +} + +fn param_descriptor(param: &Param) -> ParamDescriptor { + match PropType::from_param_type_name(¶m.type_name, param.nullable) { + Some(pt) if pt.list => ParamDescriptor { + name: param.name.clone(), + kind: ParamKind::List, + item_kind: Some(scalar_kind(pt.scalar)), + vector_dim: None, + nullable: param.nullable, + }, + Some(pt) => { + let (kind, vector_dim) = match pt.scalar { + ScalarType::Vector(dim) => (ParamKind::Vector, Some(dim)), + other => (scalar_kind(other), None), + }; + ParamDescriptor { + name: param.name.clone(), + kind, + item_kind: None, + vector_dim, + nullable: param.nullable, + } + } + // Unreachable for a parsed query (every declared param type is + // grammatical); fall back to an opaque string so the field is still + // usable rather than dropped. + None => ParamDescriptor { + name: param.name.clone(), + kind: ParamKind::String, + item_kind: None, + vector_dim: None, + nullable: param.nullable, + }, + } +} + +/// Project a loaded stored query into its catalog entry (typed params, +/// MCP tool name, read/mutate flag, description/instruction). +pub fn query_catalog_entry(query: &StoredQuery) -> QueryCatalogEntry { + QueryCatalogEntry { + name: query.name.clone(), + tool_name: query.effective_tool_name().to_string(), + description: query.decl.description.clone(), + instruction: query.decl.instruction.clone(), + mutation: query.is_mutation(), + params: query.decl.params.iter().map(param_descriptor).collect(), + } +} + #[derive(Debug, Clone, Default, Serialize, Deserialize, ToSchema)] pub struct SchemaApplyRequest { /// Project schema in `.pg` source form. The diff against the current diff --git a/crates/omnigraph-server/src/config.rs b/crates/omnigraph-server/src/config.rs index 87737d0..b308b72 100644 --- a/crates/omnigraph-server/src/config.rs +++ b/crates/omnigraph-server/src/config.rs @@ -9,6 +9,13 @@ use serde::{Deserialize, Serialize}; pub const DEFAULT_CONFIG_FILE: &str = "omnigraph.yaml"; +pub fn graph_resource_id_for_selection( + selected_graph: Option<&str>, + normalized_uri: &str, +) -> String { + selected_graph.unwrap_or(normalized_uri).to_string() +} + #[derive(Debug, Clone, Default, Serialize, Deserialize)] pub struct ProjectConfig { pub name: Option, @@ -24,6 +31,14 @@ pub struct TargetConfig { /// graph's HTTP-layer Cedar enforcement. #[serde(default)] pub policy: PolicySettings, + /// Per-graph stored-query registry: an inline `name -> entry` + /// map. Mirrors the per-graph `policy` shape — each + /// `graphs..queries` declares that graph's stored queries. Absent + /// (or empty) = no stored queries for the graph. v1 is inline-only; + /// an external `queries.yaml` manifest indirection is a deferred + /// convenience. + #[serde(default)] + pub queries: BTreeMap, } #[derive(Debug, Clone, Copy, Default, Eq, PartialEq, Serialize, Deserialize, ValueEnum)] @@ -90,6 +105,50 @@ pub struct PolicySettings { pub file: Option, } +/// One stored-query registry entry. The map **key** is the query's +/// identity — it must equal the `query ` symbol declared inside +/// the referenced `.gq` file (asserted when the registry loads). +/// Renaming the key (or the symbol) is a breaking change to callers, by +/// design. +#[derive(Debug, Clone, Serialize, Deserialize)] +pub struct QueryEntry { + /// Path to the `.gq` file (relative to the config's `base_dir`). The + /// file may declare several queries; the registry selects the one + /// whose symbol matches the map key. + pub file: String, + #[serde(default)] + pub mcp: McpSettings, +} + +/// MCP exposure for a stored query. A *deployment* concern (the same +/// `.gq` may be exposed in one graph and hidden in another), so it lives +/// in YAML rather than in the `.gq` source. **Default `expose: true`** — +/// declaring a query in the manifest *is* the opt-in, so it appears in the +/// MCP tool catalog (`GET /queries`) by default; set `expose: false` to +/// keep a query HTTP/service-callable but hidden from the agent tool list. +/// `expose` governs catalog membership only — it is **not** an +/// authorization gate (invocation is gated by `invoke_query`), so a hidden +/// query is still invocable by name with the right permission. +#[derive(Debug, Clone, Serialize, Deserialize)] +pub struct McpSettings { + #[serde(default = "mcp_expose_default")] + pub expose: bool, + pub tool_name: Option, +} + +fn mcp_expose_default() -> bool { + true +} + +impl Default for McpSettings { + fn default() -> Self { + Self { + expose: mcp_expose_default(), + tool_name: None, + } + } +} + #[derive(Debug, Clone, Copy, Eq, PartialEq, Serialize, Deserialize)] #[serde(rename_all = "snake_case")] pub enum AliasCommand { @@ -137,6 +196,12 @@ pub struct OmnigraphConfig { pub aliases: BTreeMap, #[serde(default)] pub policy: PolicySettings, + /// Top-level stored-query registry, used in single-graph + /// mode — mirrors how the top-level `policy` applies to the single + /// graph. In multi-graph mode this is unused; each graph's + /// `graphs..queries` applies instead. + #[serde(default)] + pub queries: BTreeMap, #[serde(skip)] base_dir: PathBuf, } @@ -152,6 +217,7 @@ impl Default for OmnigraphConfig { query: QueryDefaults::default(), aliases: BTreeMap::new(), policy: PolicySettings::default(), + queries: BTreeMap::new(), base_dir: PathBuf::new(), } } @@ -244,6 +310,124 @@ impl OmnigraphConfig { .map(|path| self.resolve_config_path(path)) } + /// The top-level stored-query registry entries (single-graph mode). + pub fn query_entries(&self) -> &BTreeMap { + &self.queries + } + + /// The per-graph stored-query registry entries for a named target + /// (multi-graph mode). Returns `None` if the target is unknown. + pub fn target_query_entries( + &self, + target_name: &str, + ) -> Option<&BTreeMap> { + self.graphs.get(target_name).map(|target| &target.queries) + } + + /// The stored-query registry entries that apply for a graph + /// selection — the single definition of "which `queries:` block + /// governs graph X", shared by server boot and the CLI so the two + /// can't drift. A named graph present in `graphs:` uses its + /// per-graph block; everything else (no selection, or a name that is + /// not a known graph, e.g. a bare URI) falls back to the top-level + /// block (single-graph mode). + pub fn query_entries_for(&self, graph: Option<&str>) -> &BTreeMap { + match graph { + Some(name) if self.graphs.contains_key(name) => &self.graphs[name].queries, + _ => &self.queries, + } + } + + /// The single CLI gate that turns a raw graph selection into a *validated* + /// one — the fallible counterpart to the infallible + /// [`OmnigraphConfig::query_entries_for`]. Both `queries` subcommands route + /// their selection through here so neither can skip a check the other (or + /// server boot) applies: + /// * a known name passes through, but only after the same coherence check + /// server boot enforces + /// ([`OmnigraphConfig::ensure_top_level_blocks_honored`]) — a named graph + /// with a populated top-level block is rejected; + /// * an unknown name errors with the **same** message + /// [`OmnigraphConfig::resolve_target_uri`] produces, so a command that + /// opens no URI rejects an unknown `--target` exactly like the + /// URI-resolving commands do; + /// * an anonymous selection (`None`, e.g. a bare URI) stays anonymous, + /// resolving to the top-level registry downstream (top-level honored). + pub fn resolve_graph_selection<'a>(&self, graph: Option<&'a str>) -> Result> { + match graph { + Some(name) if self.graphs.contains_key(name) => { + self.ensure_top_level_blocks_honored(Some(name))?; + Ok(Some(name)) + } + Some(name) => bail!("graph '{}' not found in {}", name, DEFAULT_CONFIG_FILE), + None => Ok(None), + } + } + + pub fn resolve_policy_tooling_graph_selection(&self) -> Result> { + self.resolve_graph_selection(self.cli_graph_name().or_else(|| self.server_graph_name())) + } + + /// The policy file that applies for a graph selection — the policy + /// sibling of [`OmnigraphConfig::query_entries_for`], so policy and + /// queries resolve by the same identity rule. A named graph in + /// `graphs:` uses its per-graph `policy.file` with **no** top-level + /// fallback (a named graph with no per-graph policy has no policy — + /// that keeps the boot-time coherence check meaningful); anything else + /// (no selection, or a bare URI) uses the top-level `policy.file`. + pub fn resolve_policy_file_for(&self, graph: Option<&str>) -> Option { + match graph { + Some(name) if self.graphs.contains_key(name) => self.resolve_target_policy_file(name), + _ => self.resolve_policy_file(), + } + } + + /// Names of any top-level config blocks (`policy.file`, `queries:`) + /// that are populated. Used by the boot-time coherence check: when a + /// **named** graph is served (single-mode by name, or multi-mode), + /// the top-level blocks are not honored, so a populated one is a + /// configuration error rather than a silent no-op. + pub fn populated_top_level_blocks(&self) -> Vec<&'static str> { + let mut blocks = Vec::new(); + if self.policy.file.is_some() { + blocks.push("policy.file"); + } + if !self.queries.is_empty() { + blocks.push("queries"); + } + blocks + } + + /// A named graph uses its own `graphs.` block, so a populated + /// top-level block would be silently ignored — a config error. The single + /// definition of that rule, shared by server boot and the CLI selection + /// gate ([`OmnigraphConfig::resolve_graph_selection`]) so the two can't + /// drift. An anonymous selection (`None`, e.g. a bare URI) legitimately + /// honors the top-level blocks, so it is never rejected here. + pub fn ensure_top_level_blocks_honored(&self, selected: Option<&str>) -> Result<()> { + if let Some(name) = selected { + let unhonored = self.populated_top_level_blocks(); + if !unhonored.is_empty() { + bail!( + "named graph '{name}' uses its own `graphs.{name}.…` block, but top-level {} \ + {} set and would be ignored. Move it to `graphs.{name}` (e.g. \ + `graphs.{name}.policy.file`, `graphs.{name}.queries`).", + unhonored.join(" and "), + if unhonored.len() == 1 { "is" } else { "are" }, + ); + } + } + Ok(()) + } + + /// Resolve a stored-query `.gq` file path (from a registry entry), + /// relative to the config's `base_dir`. Mirrors policy-file + /// resolution; the registry loader calls this to turn each entry's + /// `file:` value into an absolute path. + pub fn resolve_query_file(&self, value: &str) -> PathBuf { + self.resolve_config_path(value) + } + /// Resolve the server-level policy file path (used by management /// endpoints). Returns `None` if `server.policy.file` is not set. pub fn resolve_server_policy_file(&self) -> Option { @@ -387,7 +571,9 @@ mod tests { use tempfile::tempdir; - use super::{ReadOutputFormat, TableCellLayout, load_config_in}; + use super::{ + ReadOutputFormat, TableCellLayout, graph_resource_id_for_selection, load_config_in, + }; #[test] fn load_config_reads_yaml_defaults_from_current_dir() { @@ -451,6 +637,114 @@ policy: {} assert!(config.graphs.is_empty()); } + #[test] + fn graph_resource_id_for_selection_uses_name_or_anonymous_uri() { + assert_eq!( + graph_resource_id_for_selection(Some("local"), "/tmp/graph.omni"), + "local" + ); + assert_eq!( + graph_resource_id_for_selection(None, "/tmp/graph.omni"), + "/tmp/graph.omni" + ); + } + + #[test] + fn resolve_graph_selection_validates_membership_and_coherence() { + let temp = tempdir().unwrap(); + fs::write( + temp.path().join("omnigraph.yaml"), + "graphs:\n local:\n uri: ./demo.omni\n", + ) + .unwrap(); + let config = load_config_in(temp.path(), None).unwrap(); + + // A known graph passes through unchanged. + assert_eq!(config.resolve_graph_selection(Some("local")).unwrap(), Some("local")); + // An anonymous selection stays anonymous (→ top-level registry downstream). + assert_eq!(config.resolve_graph_selection(None).unwrap(), None); + // An unknown name errors, naming the graph (matching resolve_target_uri). + let err = config.resolve_graph_selection(Some("ghost")).unwrap_err().to_string(); + assert!( + err.contains("ghost") && err.contains("not found"), + "unknown graph must error naming it: {err}" + ); + + // Coherence: a named graph plus a populated top-level block is the + // config server boot refuses, so the gate rejects it too (shared rule + // via ensure_top_level_blocks_honored). An anonymous selection still + // passes — top-level is honored when no graph is named. + let temp2 = tempdir().unwrap(); + fs::write( + temp2.path().join("omnigraph.yaml"), + "graphs:\n local:\n uri: ./demo.omni\npolicy:\n file: ./top.yaml\n", + ) + .unwrap(); + let incoherent = load_config_in(temp2.path(), None).unwrap(); + let err = incoherent + .resolve_graph_selection(Some("local")) + .unwrap_err() + .to_string(); + assert!( + err.contains("local") && err.contains("policy.file"), + "named graph + populated top-level block must be rejected, naming both: {err}" + ); + assert_eq!( + incoherent.resolve_graph_selection(None).unwrap(), + None, + "anonymous selection still honors top-level" + ); + } + + #[test] + fn policy_tooling_graph_selection_prefers_cli_then_server_and_validates() { + let temp = tempdir().unwrap(); + fs::write( + temp.path().join("omnigraph.yaml"), + "graphs:\n local:\n uri: ./local.omni\n prod:\n uri: ./prod.omni\n\ + server:\n graph: local\ncli:\n graph: prod\n", + ) + .unwrap(); + let config = load_config_in(temp.path(), None).unwrap(); + assert_eq!( + config.resolve_policy_tooling_graph_selection().unwrap(), + Some("prod") + ); + + let temp = tempdir().unwrap(); + fs::write( + temp.path().join("omnigraph.yaml"), + "graphs:\n local:\n uri: ./local.omni\nserver:\n graph: local\n", + ) + .unwrap(); + let config = load_config_in(temp.path(), None).unwrap(); + assert_eq!( + config.resolve_policy_tooling_graph_selection().unwrap(), + Some("local") + ); + + let temp = tempdir().unwrap(); + fs::write(temp.path().join("omnigraph.yaml"), "policy: {}\n").unwrap(); + let config = load_config_in(temp.path(), None).unwrap(); + assert_eq!(config.resolve_policy_tooling_graph_selection().unwrap(), None); + + let temp = tempdir().unwrap(); + fs::write( + temp.path().join("omnigraph.yaml"), + "graphs:\n local:\n uri: ./local.omni\nserver:\n graph: ghost\n", + ) + .unwrap(); + let config = load_config_in(temp.path(), None).unwrap(); + let err = config + .resolve_policy_tooling_graph_selection() + .unwrap_err() + .to_string(); + assert!( + err.contains("ghost") && err.contains("not found"), + "unknown server.graph must use graph-selection validation: {err}" + ); + } + #[test] fn resolve_query_path_searches_config_roots() { let temp = tempdir().unwrap(); @@ -489,6 +783,118 @@ policy: {} assert_eq!(resolved, config_dir.join("local.gq")); } + #[test] + fn queries_block_round_trips_inline_and_per_graph() { + let temp = tempdir().unwrap(); + fs::write( + temp.path().join("omnigraph.yaml"), + r#" +graphs: + prod: + uri: s3://bucket/prod + queries: + find_user: + file: ./queries/find_user.gq + mcp: + expose: true + tool_name: lookup_user + internal_audit: + file: ./queries/audit.gq +queries: + single_mode_q: + file: ./q.gq +"#, + ) + .unwrap(); + + let config = load_config_in(temp.path(), None).unwrap(); + + // Per-graph registry (multi-graph mode). + let prod = config.target_query_entries("prod").unwrap(); + assert_eq!(prod.len(), 2); + let find_user = &prod["find_user"]; + assert_eq!(find_user.file, "./queries/find_user.gq"); + assert!(find_user.mcp.expose); + assert_eq!(find_user.mcp.tool_name.as_deref(), Some("lookup_user")); + // Default exposure is true (the manifest entry is the opt-in); tool_name absent. + let audit = &prod["internal_audit"]; + assert!(audit.mcp.expose); + assert!(audit.mcp.tool_name.is_none()); + + // Top-level registry (single-graph mode). + assert_eq!(config.query_entries().len(), 1); + + // The shared selector resolves the same blocks the server boot + // and the CLI use: a known graph → its per-graph block; no + // selection or an unknown name → the top-level block (the latter + // pins the behavior of the CLI's now-deleted fallback arm). + assert_eq!(config.query_entries_for(Some("prod")).len(), 2); + assert_eq!(config.query_entries_for(None).len(), 1); + assert_eq!(config.query_entries_for(Some("nonexistent")).len(), 1); + + // Path resolution joins against base_dir, like policy files. + assert_eq!( + config.resolve_query_file(&find_user.file), + temp.path().join("./queries/find_user.gq") + ); + } + + #[test] + fn resolve_policy_file_for_follows_identity() { + let temp = tempdir().unwrap(); + fs::write( + temp.path().join("omnigraph.yaml"), + "policy:\n file: ./top.yaml\ngraphs:\n prod:\n uri: s3://b/prod\n \ + policy:\n file: ./prod.yaml\n bare:\n uri: s3://b/bare\n", + ) + .unwrap(); + let config = load_config_in(temp.path(), None).unwrap(); + + // Named graph with its own policy → per-graph (not top-level). + assert!( + config + .resolve_policy_file_for(Some("prod")) + .unwrap() + .ends_with("prod.yaml") + ); + // Named graph with NO per-graph policy → None (no top-level fallback; + // load-bearing for the boot coherence check). + assert!(config.resolve_policy_file_for(Some("bare")).is_none()); + // Anonymous (bare URI) or an unknown name → top-level. + assert!( + config + .resolve_policy_file_for(None) + .unwrap() + .ends_with("top.yaml") + ); + assert!( + config + .resolve_policy_file_for(Some("nope")) + .unwrap() + .ends_with("top.yaml") + ); + } + + #[test] + fn queries_block_absent_yields_empty_registry() { + let temp = tempdir().unwrap(); + fs::write( + temp.path().join("omnigraph.yaml"), + "graphs:\n local:\n uri: ./demo.omni\n", + ) + .unwrap(); + + let config = load_config_in(temp.path(), None).unwrap(); + // Additive: no `queries:` anywhere → empty registries everywhere. + assert!(config.query_entries().is_empty()); + assert!( + config + .target_query_entries("local") + .unwrap() + .is_empty() + ); + } + #[test] fn policy_block_accepts_non_empty_mapping() { let temp = tempdir().unwrap(); diff --git a/crates/omnigraph-server/src/lib.rs b/crates/omnigraph-server/src/lib.rs index ad41f9d..60ebef3 100644 --- a/crates/omnigraph-server/src/lib.rs +++ b/crates/omnigraph-server/src/lib.rs @@ -4,6 +4,7 @@ pub mod config; pub mod graph_id; pub mod identity; pub mod policy; +pub mod queries; pub mod registry; pub mod workload; @@ -11,6 +12,8 @@ pub use graph_id::GraphId; pub use identity::{AuthSource, GraphKey, ResolvedActor, Scope, TenantId}; pub use registry::{GraphHandle, GraphRegistry, InsertError, RegistryLookup, RegistrySnapshot}; +use crate::queries::{QueryRegistry, check, format_check_breakages}; + use std::collections::{HashMap, HashSet}; use std::fs; use std::io; @@ -22,7 +25,8 @@ use api::{ BranchCreateOutput, BranchCreateRequest, BranchDeleteOutput, BranchListOutput, BranchMergeOutput, BranchMergeRequest, ChangeOutput, ChangeRequest, CommitListOutput, CommitListQuery, ErrorCode, ErrorOutput, ExportRequest, GraphInfo, GraphListResponse, - HealthOutput, IngestOutput, IngestRequest, QueryRequest, ReadOutput, ReadRequest, + HealthOutput, IngestOutput, IngestRequest, InvokeStoredQueryRequest, + InvokeStoredQueryResponse, QueriesCatalogOutput, QueryRequest, ReadOutput, ReadRequest, SchemaApplyOutput, SchemaApplyRequest, SchemaOutput, SnapshotQuery, ingest_output, schema_apply_output, snapshot_payload, }; @@ -40,12 +44,13 @@ use color_eyre::eyre::{Result, WrapErr, bail}; pub use config::{ AliasCommand, AliasConfig, CliDefaults, DEFAULT_CONFIG_FILE, OmnigraphConfig, PolicySettings, ProjectConfig, QueryDefaults, ReadOutputFormat, ServerDefaults, TableCellLayout, TargetConfig, - load_config, + graph_resource_id_for_selection, load_config, }; use futures::stream; use omnigraph::db::{Omnigraph, ReadTarget}; use omnigraph::error::{ManifestConflictDetails, ManifestErrorKind, OmniError}; use omnigraph::storage::normalize_root_uri; +use omnigraph_compiler::catalog::Catalog; use omnigraph_compiler::json_params_to_param_map; use omnigraph_compiler::query::parser::parse_query; use omnigraph_compiler::{JsonParamMode, ParamMap}; @@ -93,6 +98,8 @@ fn hash_bearer_token(token: &str) -> BearerTokenHash { server_export, #[allow(deprecated)] server_change, server_mutate, + server_list_queries, + server_invoke_query, server_schema_apply, server_schema_get, server_ingest, @@ -157,8 +164,16 @@ pub enum ServerConfigMode { /// set to a named target. Single { uri: String, + /// Cedar graph resource id for the single graph. A named selection + /// uses the graph name; an anonymous URI uses the normalized URI to + /// preserve legacy single-graph policy identity. + graph_id: String, /// Top-level `policy.file` (single-graph Cedar policy). policy_file: Option, + /// Top-level stored-query registry, loaded and identity-checked + /// at settings-build time; type-checked against the schema when + /// the engine opens. + queries: QueryRegistry, }, /// Multi-graph invocation — `--config omnigraph.yaml` with a /// non-empty `graphs:` map and no single-mode selector. @@ -185,6 +200,10 @@ pub struct GraphStartupConfig { pub graph_id: String, pub uri: String, pub policy_file: Option, + /// Per-graph stored-query registry, loaded and identity-checked at + /// settings-build time; type-checked against the schema when this + /// graph's engine opens. + pub queries: QueryRegistry, } /// Runtime routing for the server. Single mode = legacy @@ -285,7 +304,31 @@ impl AppState { ) -> Self { let bearer_tokens = hash_bearer_tokens(bearer_tokens); let per_graph_policy = policy_engine.map(Arc::new); - Self::build_single_mode(uri, db, bearer_tokens, per_graph_policy, Arc::new(workload)) + Self::build_single_mode(uri, db, bearer_tokens, per_graph_policy, Arc::new(workload), None) + } + + /// Like `new_single`, but attaches a pre-validated stored-query + /// registry. Private — the production single-mode boot path + /// (`open_single_with_queries`) is the only caller; every public + /// `new_*` constructor builds with no stored queries. + fn new_single_with_queries( + uri: String, + db: Omnigraph, + bearer_tokens: Vec<(String, String)>, + policy_engine: Option, + workload: workload::WorkloadController, + queries: Option>, + ) -> Self { + let bearer_tokens = hash_bearer_tokens(bearer_tokens); + let per_graph_policy = policy_engine.map(Arc::new); + Self::build_single_mode( + uri, + db, + bearer_tokens, + per_graph_policy, + Arc::new(workload), + queries, + ) } pub fn new(uri: String, db: Omnigraph) -> Self { @@ -377,6 +420,39 @@ impl AppState { uri: impl Into, bearer_tokens: Vec<(String, String)>, policy_file: Option<&PathBuf>, + ) -> Result { + Self::open_single_with_queries( + uri, + bearer_tokens, + policy_file, + QueryRegistry::default(), + ) + .await + } + + /// Single-mode boot with a stored-query registry: open the engine, + /// **type-check the registry against the live schema and refuse to + /// start on a breakage** (same posture as bad policy YAML), log + /// non-blocking warnings, then attach the registry to the handle. + /// With an empty registry the check is a no-op and no registry is + /// attached — that is the path `open_with_bearer_tokens_and_policy` + /// (no stored queries) takes. + pub async fn open_single_with_queries( + uri: impl Into, + bearer_tokens: Vec<(String, String)>, + policy_file: Option<&PathBuf>, + queries: QueryRegistry, + ) -> Result { + Self::open_single_with_queries_for_graph_id(uri, bearer_tokens, policy_file, queries, None) + .await + } + + async fn open_single_with_queries_for_graph_id( + uri: impl Into, + bearer_tokens: Vec<(String, String)>, + policy_file: Option<&PathBuf>, + queries: QueryRegistry, + graph_id: Option, ) -> Result { // The "policy requires tokens" invariant is enforced once by // `classify_server_runtime_state` in `serve()`, before either @@ -384,16 +460,24 @@ impl AppState { // time we get here, the (policy, no-tokens) combination has // already been rejected — no second bail needed. let uri = normalize_root_uri(&uri.into()).wrap_err("normalize graph URI")?; + let graph_id = graph_id.unwrap_or_else(|| uri.clone()); let db = Omnigraph::open(&uri).await?; + + // Validate the registry against the live schema and resolve it to + // an attachable handle (refuse boot on breakage). + let registry = validate_and_attach(queries, &db.catalog(), &graph_id)?; + let policy_engine = match policy_file { - Some(path) => Some(PolicyEngine::load_graph(path, &uri)?), + Some(path) => Some(PolicyEngine::load_graph(path, &graph_id)?), None => None, }; - Ok(Self::new_with_bearer_tokens_and_policy( + Ok(Self::new_single_with_queries( uri, db, bearer_tokens, policy_engine, + workload::WorkloadController::from_env(), + registry, )) } @@ -408,6 +492,7 @@ impl AppState { bearer_tokens: Arc<[(BearerTokenHash, Arc)]>, policy_engine: Option>, workload: Arc, + queries: Option>, ) -> Self { // Engine-layer policy gate (MR-722). With a per-graph policy // installed, every `_as` writer on `Omnigraph` calls into the @@ -436,6 +521,7 @@ impl AppState { uri, engine: Arc::new(db), policy: policy_engine, + queries, }); Self { routing: GraphRouting::Single { handle }, @@ -750,6 +836,58 @@ pub fn init_tracing() { let _ = tracing_subscriber::fmt().with_env_filter(filter).try_init(); } +/// Log each non-blocking advisory from a registry check report. +fn log_registry_warnings(label: &str, report: &queries::CheckReport) { + for warning in &report.warnings { + warn!(graph = label, query = %warning.query, "stored query: {}", warning.message); + } +} + +fn validate_registry_against_catalog( + registry: &QueryRegistry, + catalog: &Catalog, + label: &str, +) -> omnigraph::error::Result<()> { + let report = check(registry, catalog); + if report.has_breakages() { + return Err(OmniError::manifest(format_check_breakages(label, &report))); + } + log_registry_warnings(label, &report); + Ok(()) +} + +/// Validate a loaded stored-query registry against the live schema and +/// resolve it to an attachable handle. Refuses boot on any breakage +/// (same posture as bad policy YAML), logs the non-blocking warnings, +/// and collapses an empty registry to `None` (nothing attached). This is +/// the single gate every open path funnels through, so no opener can +/// attach a registry that has not been schema-checked. `label` names the +/// graph in messages. +fn validate_and_attach( + queries: QueryRegistry, + catalog: &Catalog, + label: &str, +) -> Result>> { + validate_registry_against_catalog(&queries, catalog, label) + .map_err(|err| color_eyre::eyre::eyre!(err.to_string()))?; + Ok(if queries.is_empty() { + None + } else { + Some(Arc::new(queries)) + }) +} + +/// Format every load error (parse / identity failure) into a multi-line +/// boot-abort message. +fn format_registry_load_errors(label: &str, errors: &[queries::LoadError]) -> String { + let joined = errors + .iter() + .map(|e| e.to_string()) + .collect::>() + .join("\n "); + format!("graph '{label}': stored-query registry failed to load:\n {joined}") +} + pub fn load_server_settings( config_path: Option<&PathBuf>, cli_uri: Option, @@ -799,15 +937,43 @@ pub fn load_server_settings( let uri = normalize_root_uri(&raw_uri).wrap_err_with(|| { format!("normalize single-graph URI '{raw_uri}' from server settings") })?; - let policy_file = config.resolve_policy_file(); - ServerConfigMode::Single { uri, policy_file } + // Config follows graph IDENTITY, not mode: a bare URI is anonymous + // (top-level config); a graph chosen by name uses its per-graph + // `graphs..{policy,queries}`. `resolve_target_uri` already + // errored on an unknown name, so a `Some(name)` here is a known graph. + let selected: Option<&str> = if has_cli_uri { + None + } else { + cli_target.as_deref().or_else(|| config.server_graph_name()) + }; + // A named selection must not leave a populated top-level block + // silently unused — refuse boot and point at the per-graph block. The + // same rule the CLI selection gate enforces, shared via one helper so + // the boot check and `omnigraph queries validate`/`list` can't drift. + config.ensure_top_level_blocks_honored(selected)?; + // Load + identity-check now (no engine needed); the schema + // type-check happens when the engine opens. + let policy_file = config.resolve_policy_file_for(selected); + let queries = QueryRegistry::load(&config, config.query_entries_for(selected)) + .map_err(|errs| color_eyre::eyre::eyre!(format_registry_load_errors(&uri, &errs)))?; + let graph_id = graph_resource_id_for_selection(selected, &uri); + ServerConfigMode::Single { + uri, + graph_id, + policy_file, + queries, + } } else if has_explicit_config && has_graphs_map { - if config.resolve_policy_file().is_some() { + // Multi mode: every graph uses its per-graph block; top-level + // policy/queries are never honored, so a populated one is an error. + let unhonored = config.populated_top_level_blocks(); + if !unhonored.is_empty() { bail!( - "top-level `policy.file` is single-graph/CLI-local policy only; \ - in multi-graph mode move per-graph rules to \ - `graphs..policy.file` and move `graph_list` rules to \ - `server.policy.file`." + "multi-graph mode: top-level {} {} not honored — each graph uses its own \ + `graphs..…` block. Move per-graph rules there (and any \ + `graph_list` policy to `server.policy.file`).", + unhonored.join(" and "), + if unhonored.len() == 1 { "is" } else { "are" }, ); } // Rule 4 → Multi mode. Build a startup config per graph. @@ -823,10 +989,17 @@ pub fn load_server_settings( let uri = normalize_root_uri(&raw_uri).wrap_err_with(|| { format!("normalize URI '{raw_uri}' for graph '{name}' in omnigraph.yaml") })?; + // Per-graph `queries:`, selected through the shared + // `query_entries_for` so server and CLI resolve identically. + // Load + identity-check now; the schema type-check happens + // when this graph's engine opens. + let queries = QueryRegistry::load(&config, config.query_entries_for(Some(name.as_str()))) + .map_err(|errs| color_eyre::eyre::eyre!(format_registry_load_errors(name, &errs)))?; graphs.push(GraphStartupConfig { graph_id: name.clone(), uri, policy_file: config.resolve_target_policy_file(name), + queries, }); } let config_path = config_path @@ -949,6 +1122,8 @@ pub fn build_app(state: AppState) -> Router { server_change })) .route("/mutate", post(server_mutate)) + .route("/queries", get(server_list_queries)) + .route("/queries/{name}", post(server_invoke_query)) .route("/schema", get(server_schema_get)) .route("/schema/apply", post(server_schema_apply)) .route( @@ -1046,10 +1221,28 @@ pub async fn serve(config: ServerConfig) -> Result<()> { let bind = config.bind.clone(); let state = match config.mode { - ServerConfigMode::Single { uri, policy_file } => { + ServerConfigMode::Single { + uri, + graph_id, + policy_file, + queries, + } => { let uri_for_log = uri.clone(); - info!(uri = %uri_for_log, bind = %bind, mode = "single", "serving omnigraph"); - AppState::open_with_bearer_tokens_and_policy(uri, tokens, policy_file.as_ref()).await? + info!( + uri = %uri_for_log, + graph_id = %graph_id, + bind = %bind, + mode = "single", + "serving omnigraph" + ); + AppState::open_single_with_queries_for_graph_id( + uri, + tokens, + policy_file.as_ref(), + queries, + Some(graph_id), + ) + .await? } ServerConfigMode::Multi { graphs, @@ -1131,6 +1324,12 @@ async fn open_single_graph(cfg: GraphStartupConfig) -> Result> .await .map_err(|err| color_eyre::eyre::eyre!("open graph '{}' at {}: {err}", graph_id, uri))?; + // Validate this graph's stored queries against the live schema and + // resolve them to an attachable handle (refuse boot on breakage). + // Done before the policy match rebinds `db`; the catalog handle is an + // owned `Arc`, so no borrow of `db` survives into the match. + let queries = validate_and_attach(cfg.queries, &db.catalog(), graph_id.as_str())?; + let (policy_arc, db) = match &cfg.policy_file { Some(path) => { let policy = PolicyEngine::load_graph(path, graph_id.as_str())?; @@ -1146,6 +1345,7 @@ async fn open_single_graph(cfg: GraphStartupConfig) -> Result> uri, engine: Arc::new(db), policy: policy_arc, + queries, })) } @@ -1479,7 +1679,21 @@ fn log_policy_decision(actor_id: &str, request: &PolicyRequest, decision: &Polic ); } -/// HTTP-layer Cedar policy gate. Two sources of the policy engine: +/// The allow/deny **decision** an authorization check produces, kept +/// separate from the operational failures (`Err`) that can occur while +/// computing it. [`authorize_request`] collapses `Denied` to a 403; a caller +/// that needs to remap a denial without also remapping operational failures +/// (the stored-query invoke handler hides a denial as a 404) matches on this +/// directly, so a real 401 (missing bearer) or 500 (policy-evaluation error) +/// keeps its true status instead of being masked as the denial's response. +enum Authz { + Allowed, + Denied(String), +} + +/// HTTP-layer Cedar policy gate, returning the allow/deny [`Authz`] decision +/// and reserving `Err` for operational failures (401 missing bearer, 500 +/// policy-evaluation error). Two sources of the policy engine: /// * Per-graph handler — passes `handle.policy.as_deref()` so the /// graph's Cedar rules govern read/change/branch_*/schema_apply. /// * Management handler — passes `state.server_policy.as_deref()` so @@ -1493,11 +1707,11 @@ fn log_policy_decision(actor_id: &str, request: &PolicyRequest, decision: &Polic /// dropped from the type), so handlers cannot smuggle it through the /// request. See `actor_id_resolves_from_bearer_token_ignoring_client_supplied_headers` /// at `tests/server.rs`. -fn authorize_request( +fn authorize( actor: Option<&ResolvedActor>, policy: Option<&PolicyEngine>, request: PolicyRequest, -) -> std::result::Result<(), ApiError> { +) -> std::result::Result { let Some(engine) = policy else { // No PolicyEngine installed. Three runtime states can reach this: // @@ -1524,21 +1738,23 @@ fn authorize_request( // operator's only path to enabling it is configuring an // explicit `server.policy.file` in omnigraph.yaml. if request.action.resource_kind() == PolicyResourceKind::Server { - return Err(ApiError::forbidden( + return Ok(Authz::Denied( "server-scoped actions require an explicit `server.policy.file` \ configured in omnigraph.yaml — the management surface is closed \ by default in every runtime state, including --unauthenticated, \ - so that server topology is never exposed without operator opt-in.", + so that server topology is never exposed without operator opt-in." + .to_string(), )); } if actor.is_some() && request.action != PolicyAction::Read { - return Err(ApiError::forbidden( + return Ok(Authz::Denied( "server runs in default-deny mode (bearer tokens configured but no \ policy file). Only `read` actions are permitted; configure \ - `policy.file` in omnigraph.yaml to enable other actions.", + `policy.file` in omnigraph.yaml to enable other actions." + .to_string(), )); } - return Ok(()); + return Ok(Authz::Allowed); }; let Some(actor) = actor else { return Err(ApiError::unauthorized("missing bearer token")); @@ -1560,9 +1776,26 @@ fn authorize_request( .map_err(|err| ApiError::internal(format!("policy: {err}")))?; log_policy_decision(actor_id, &request, &decision); if decision.allowed { - Ok(()) + Ok(Authz::Allowed) } else { - Err(ApiError::forbidden(decision.message)) + Ok(Authz::Denied(decision.message)) + } +} + +/// Thin wrapper over [`authorize`] for the handlers that treat any denial as a +/// 403: a denial becomes `ApiError::forbidden`, and operational failures +/// (401 missing bearer, 500 policy-evaluation error) propagate unchanged. The +/// stored-query invoke handler does **not** use this — it consumes the +/// [`Authz`] decision directly to hide a denial as a 404 while letting an +/// operational failure keep its true status. +fn authorize_request( + actor: Option<&ResolvedActor>, + policy: Option<&PolicyEngine>, + request: PolicyRequest, +) -> std::result::Result<(), ApiError> { + match authorize(actor, policy, request)? { + Authz::Allowed => Ok(()), + Authz::Denied(message) => Err(ApiError::forbidden(message)), } } @@ -2001,6 +2234,194 @@ async fn server_mutate( )) } +/// Path parameter for `POST /queries/{name}`. +#[derive(Deserialize)] +struct QueryNamePath { + name: String, +} + +fn parse_optional_invoke_body( + body: Bytes, +) -> std::result::Result { + if body.is_empty() { + return Ok(InvokeStoredQueryRequest::default()); + } + serde_json::from_slice::>(&body) + .map(|request| request.unwrap_or_default()) + .map_err(|err| { + ApiError::bad_request(format!("invalid stored-query invocation body: {err}")) + }) +} + +#[utoipa::path( + post, + path = "/queries/{name}", + tag = "queries", + operation_id = "invoke_query", + params(("name" = String, Path, description = "Stored query name (the registry key)")), + request_body = Option, + responses( + (status = 200, description = "Read envelope (ReadOutput) or mutation envelope (ChangeOutput), serialized untagged", body = InvokeStoredQueryResponse), + (status = 400, description = "Bad request (param type error; snapshot on a stored mutation)", body = ErrorOutput), + (status = 401, description = "Unauthorized", body = ErrorOutput), + (status = 403, description = "Forbidden (the inner `change` gate for a stored mutation)", body = ErrorOutput), + (status = 404, description = "Unknown stored query, or `invoke_query` denied — indistinguishable to a caller without the grant", body = ErrorOutput), + (status = 409, description = "Merge conflict", body = ErrorOutput), + (status = 429, description = "Per-actor admission cap exceeded; honor `Retry-After` header", body = ErrorOutput), + (status = 500, description = "Policy evaluation error (a denial is reported as 404, not 500)", body = ErrorOutput), + ), + security(("bearer_token" = [])), +)] +/// Invoke a curated, server-side stored query by name. +/// +/// The query source comes from the graph's `queries:` registry, not the +/// request body — callers send only runtime inputs (`params`, `branch`, +/// `snapshot`). Gated by the `invoke_query` Cedar action at the boundary; +/// a stored *mutation* additionally passes the engine's `change` gate +/// (double-gated). An actor **without** `invoke_query` cannot tell a denied +/// query from a missing one — both return the same 404, so the catalog +/// can't be probed without the grant. Once `invoke_query` is held, the +/// inner `read`/`change` gate may surface a 403 for an existing query the +/// actor can't run (the intended double-gate signal). +async fn server_invoke_query( + State(state): State, + Extension(handle): Extension>, + actor: Option>, + Path(QueryNamePath { name }): Path, + body: Bytes, +) -> std::result::Result, ApiError> { + let req = parse_optional_invoke_body(body)?; + // A caller without `invoke_query` can't tell a denial from a missing + // query: both 404 with this exact message, so the catalog can't be + // probed without the grant. (A caller that holds invoke_query may still + // see the inner gate's 403 for an existing query it can't run — intended.) + const NOT_FOUND: &str = "stored query not found"; + let actor_ref = actor.as_ref().map(|Extension(actor)| actor); + + // Boundary gate (authentication already ran in `require_bearer_auth`). + // A denial is hidden as 404 (deny == missing, so the catalog can't be + // probed without the grant), but operational failures (401 missing bearer, + // 500 policy-evaluation error) propagate with their true status via `?` + // rather than being masked as a missing query. + match authorize( + actor_ref, + handle.policy.as_deref(), + PolicyRequest { + action: PolicyAction::InvokeQuery, + // Graph-scoped: no branch dimension. The per-branch/snapshot + // access is enforced by the inner read/change gate in the + // runner, so the outer gate must not resolve a branch (doing so + // was wrong for snapshot reads). + branch: None, + target_branch: None, + }, + )? { + Authz::Allowed => {} + Authz::Denied(_) => return Err(ApiError::not_found(NOT_FOUND)), + } + + // Resolve against the per-graph registry (same 404 on a miss). + let stored = handle + .queries + .as_ref() + .and_then(|registry| registry.lookup(&name)) + .ok_or_else(|| ApiError::not_found(NOT_FOUND))?; + + // Detach what we need before `handle` moves into the runner — the + // registry borrow lives inside `handle`. + let source = Arc::clone(&stored.source); + let query_name = stored.name.clone(); + let is_mutation = stored.is_mutation(); + + info!( + graph = %handle.uri, + actor = ?actor_ref.map(|a| a.actor_id.as_ref()), + query = %query_name, + kind = if is_mutation { "mutate" } else { "read" }, + "stored query invoked" + ); + + if is_mutation { + if req.snapshot.is_some() { + return Err(ApiError::bad_request( + "stored mutation cannot target a snapshot", + )); + } + let branch = req.branch.unwrap_or_else(|| "main".to_string()); + let output = run_mutate( + state, + handle, + actor_ref, + &source, + Some(&query_name), + req.params.as_ref(), + branch, + ) + .await?; + Ok(Json(InvokeStoredQueryResponse::Change(output))) + } else { + let (selected, target, result) = run_query( + handle, + actor_ref, + &source, + Some(&query_name), + req.params.as_ref(), + req.branch, + req.snapshot, + true, + ) + .await?; + Ok(Json(InvokeStoredQueryResponse::Read(api::read_output( + selected, &target, result, + )))) + } +} + +#[utoipa::path( + get, + path = "/queries", + tag = "queries", + operation_id = "list_queries", + responses( + (status = 200, description = "Stored-query catalog (the mcp.expose subset, with typed params)", body = QueriesCatalogOutput), + (status = 401, description = "Unauthorized", body = ErrorOutput), + (status = 403, description = "Forbidden", body = ErrorOutput), + ), + security(("bearer_token" = [])), +)] +/// List the graph's exposed stored queries as a typed tool catalog. +/// +/// Returns the `mcp.expose == true` subset of the `queries:` registry, each +/// with its MCP tool name, read/mutate flag, description/instruction, and +/// typed parameters — enough for a client to register them as tools without +/// fetching `.gq` source. Read-gated; the catalog is graph-wide (branch +/// independent — `read` is authorized against `main`). **Not** Cedar-filtered +/// per query yet, so it can list a query whose `invoke_query` the caller +/// lacks (a known gap until per-query authorization lands). +async fn server_list_queries( + Extension(handle): Extension>, + actor: Option>, +) -> std::result::Result, ApiError> { + authorize_request( + actor.as_ref().map(|Extension(actor)| actor), + handle.policy.as_deref(), + PolicyRequest { + action: PolicyAction::Read, + branch: Some("main".to_string()), + target_branch: None, + }, + )?; + let queries = match handle.queries.as_ref() { + Some(registry) => registry + .iter() + .filter(|q| q.expose) + .map(api::query_catalog_entry) + .collect(), + None => Vec::new(), + }; + Ok(Json(QueriesCatalogOutput { queries })) +} + #[utoipa::path( get, path = "/schema", @@ -2088,18 +2509,26 @@ async fn server_schema_apply( .map_err(ApiError::from_workload_reject)?; let result = { let db = &handle.engine; + let registry = handle.queries.as_deref(); + let label = handle.key.graph_id.as_str().to_string(); // Engine-layer policy enforcement (MR-722): pass the resolved // actor through so apply_schema_as can call enforce() with the // authoritative identity. With a policy installed in AppState, // engine-side enforcement re-checks the same decision the // HTTP-layer authorize_request just made above. PR #3 collapses // the redundancy. - db.apply_schema_as( + db.apply_schema_as_with_catalog_check( &request.schema_source, omnigraph::db::SchemaApplyOptions { allow_data_loss: request.allow_data_loss, }, actor_id, + |catalog| { + if let Some(registry) = registry { + validate_registry_against_catalog(registry, catalog, &label)?; + } + Ok(()) + }, ) .await .map_err(ApiError::from_omni)? @@ -2658,12 +3087,133 @@ mod tests { use std::fs; use tempfile::tempdir; + /// `authorize` returns the allow/deny **decision** (`Authz`) and reserves + /// `Err` for operational failures, so the invoke handler can hide a denial + /// as 404 without also masking a 401/500. Pins each outcome. + #[test] + fn authorize_splits_decision_from_operational_error() { + use super::{Authz, PolicyAction, PolicyCompiler, PolicyConfig, PolicyRequest, ResolvedActor, authorize}; + use std::sync::Arc; + + fn req(action: PolicyAction) -> PolicyRequest { + PolicyRequest { action, branch: None, target_branch: None } + } + let actor = ResolvedActor::cluster_static(Arc::from("act-alice")); + + // --- No policy engine installed (open / default-deny modes) --- + // A server-scoped action is denied in every no-policy state. + assert!(matches!( + authorize(Some(&actor), None, req(PolicyAction::GraphList)).unwrap(), + Authz::Denied(_) + )); + // Authenticated actor + a non-read per-graph action → default-deny. + assert!(matches!( + authorize(Some(&actor), None, req(PolicyAction::Change)).unwrap(), + Authz::Denied(_) + )); + // `read` is the one per-graph action permitted without a policy. + assert!(matches!( + authorize(Some(&actor), None, req(PolicyAction::Read)).unwrap(), + Authz::Allowed + )); + // Open mode (no actor, no policy) → allowed. + assert!(matches!( + authorize(None, None, req(PolicyAction::Read)).unwrap(), + Authz::Allowed + )); + + // --- Policy engine installed --- + let policy: PolicyConfig = serde_yaml::from_str( + "version: 1\n\ + groups:\n team: [act-alice]\n\ + rules:\n - id: team-read\n allow:\n actors: { group: team }\n actions: [read]\n branch_scope: any\n", + ) + .unwrap(); + let engine = PolicyCompiler::compile(&policy, "graph").unwrap(); + + // A matched allow rule → Allowed. + assert!(matches!( + authorize( + Some(&actor), + Some(&engine), + PolicyRequest { action: PolicyAction::Read, branch: Some("main".to_string()), target_branch: None }, + ) + .unwrap(), + Authz::Allowed + )); + // Known actor, no matching allow rule → Denied, carrying the decision message. + match authorize( + Some(&actor), + Some(&engine), + PolicyRequest { action: PolicyAction::Change, branch: Some("main".to_string()), target_branch: None }, + ) + .unwrap() + { + Authz::Denied(message) => assert!(!message.is_empty(), "a deny carries its decision message"), + Authz::Allowed => panic!("change must be denied: only read is allowed"), + } + // Policy installed but no actor → operational failure (`Err`), NOT a + // decision. This is the split that keeps a 401/500 from being masked + // as the denial's response in the invoke handler. + assert!( + authorize(None, Some(&engine), req(PolicyAction::Read)).is_err(), + "a missing actor with a policy installed is an operational error, not a deny" + ); + } + #[test] fn hash_bearer_token_produces_32_byte_output() { let hash = hash_bearer_token("any-token"); assert_eq!(hash.len(), 32); } + /// The single gate both open paths funnel through: it refuses a + /// schema breakage (naming the graph label + query), attaches a clean + /// registry, and collapses an empty one to `None`. Pure over its args + /// (no engine), so it covers the multi-graph path's logic too — the + /// only per-path difference is the `label`, asserted here. + #[test] + fn validate_and_attach_gates_on_schema_and_collapses_empty() { + use crate::queries::{QueryRegistry, RegistrySpec}; + use omnigraph_compiler::catalog::build_catalog; + use omnigraph_compiler::schema::parser::parse_schema; + + let schema = parse_schema("node User {\nname: String\n}\n").unwrap(); + let catalog = build_catalog(&schema).unwrap(); + let spec = |name: &str, source: &str| RegistrySpec { + name: name.to_string(), + source: source.to_string(), + expose: false, + tool_name: None, + }; + + // Empty registry → nothing attached, no error. + let empty = + super::validate_and_attach(QueryRegistry::default(), &catalog, "g").unwrap(); + assert!(empty.is_none()); + + // A query that type-checks → attached. + let ok = QueryRegistry::from_specs(vec![spec( + "find_user", + "query find_user() { match { $u: User } return { $u.name } }", + )]) + .unwrap(); + assert!(super::validate_and_attach(ok, &catalog, "g").unwrap().is_some()); + + // A query referencing a type the schema lacks → boot refusal that + // names both the graph label and the offending query. + let broken = QueryRegistry::from_specs(vec![spec( + "ghost", + "query ghost() { match { $w: Widget } return { $w.name } }", + )]) + .unwrap(); + let err = super::validate_and_attach(broken, &catalog, "graph-x").unwrap_err(); + let msg = err.to_string(); + assert!(msg.contains("graph-x"), "labels the graph: {msg}"); + assert!(msg.contains("ghost"), "names the query: {msg}"); + assert!(msg.contains("schema check"), "mentions the schema check: {msg}"); + } + #[test] fn hash_bearer_token_is_deterministic() { assert_eq!( @@ -2707,7 +3257,10 @@ server: let settings = load_server_settings(Some(&config), None, None, None, false).unwrap(); match &settings.mode { - ServerConfigMode::Single { uri, .. } => assert_eq!(uri, "/tmp/demo.omni"), + ServerConfigMode::Single { uri, graph_id, .. } => { + assert_eq!(uri, "/tmp/demo.omni"); + assert_eq!(graph_id, "local"); + } ServerConfigMode::Multi { .. } => panic!("expected Single mode, got Multi"), } assert_eq!(settings.bind, "0.0.0.0:9090"); @@ -2739,7 +3292,10 @@ server: ) .unwrap(); match &settings.mode { - ServerConfigMode::Single { uri, .. } => assert_eq!(uri, "/tmp/override.omni"), + ServerConfigMode::Single { uri, graph_id, .. } => { + assert_eq!(uri, "/tmp/override.omni"); + assert_eq!(graph_id, "/tmp/override.omni"); + } ServerConfigMode::Multi { .. } => panic!("expected Single mode, got Multi"), } assert_eq!(settings.bind, "0.0.0.0:9999"); @@ -2768,7 +3324,10 @@ server: load_server_settings(Some(&config), None, Some("dev".to_string()), None, false) .unwrap(); match &settings.mode { - ServerConfigMode::Single { uri, .. } => assert_eq!(uri, "http://127.0.0.1:8080"), + ServerConfigMode::Single { uri, graph_id, .. } => { + assert_eq!(uri, "http://127.0.0.1:8080"); + assert_eq!(graph_id, "dev"); + } ServerConfigMode::Multi { .. } => panic!("expected Single mode, got Multi"), } } @@ -2848,6 +3407,7 @@ server: .to_string_lossy() .into_owned(), policy_file: None, + queries: crate::queries::QueryRegistry::default(), }], config_path: temp.path().join("omnigraph.yaml"), server_policy_file: Some(policy_path), @@ -2895,7 +3455,9 @@ server: .join("graph.omni") .to_string_lossy() .into_owned(), + graph_id: "default".to_string(), policy_file: None, + queries: crate::queries::QueryRegistry::default(), }, bind: "127.0.0.1:0".to_string(), allow_unauthenticated: false, diff --git a/crates/omnigraph-server/src/queries.rs b/crates/omnigraph-server/src/queries.rs new file mode 100644 index 0000000..bf131c8 --- /dev/null +++ b/crates/omnigraph-server/src/queries.rs @@ -0,0 +1,688 @@ +//! Stored-query registry. +//! +//! A server-side registry of named, parameter-typed `.gq` queries that +//! operators declare in `omnigraph.yaml` (per-graph, or top-level in +//! single mode) and the server loads at startup. Each entry is parsed +//! and its identity asserted here (`load`); type-checking against the +//! live schema happens separately (a `check` pass) so the loader stays +//! callable without an open engine (the CLI's offline `queries check`). +//! +//! Identity is the query **name**: the manifest key must equal the +//! `query ` symbol declared in the referenced `.gq` file. The two +//! are asserted equal at load — one name, two places that must agree. +//! Renaming either is a breaking change to callers, by design. + +use std::collections::BTreeMap; +use std::fs; +use std::sync::Arc; + +use omnigraph_compiler::catalog::Catalog; +use omnigraph_compiler::query::ast::QueryDecl; +use omnigraph_compiler::query::parser::parse_query; +use omnigraph_compiler::query::typecheck::typecheck_query_decl; +use omnigraph_compiler::types::{PropType, ScalarType}; + +use crate::config::{OmnigraphConfig, QueryEntry}; + +/// One loaded stored query. `source` is the full `.gq` file text — the +/// invocation handler hands it to `run_query` / `run_mutate` verbatim, +/// which reuse the same parse/IR/exec path as the inline routes (no +/// parallel implementation). +#[derive(Debug, Clone)] +pub struct StoredQuery { + /// Identity: manifest key == `query ` symbol. + pub name: String, + /// Full `.gq` source text the query was selected from. + pub source: Arc, + /// Parsed declaration (params, mutations, description, …). + pub decl: QueryDecl, + /// Whether this query is listed in the MCP tool catalog (`GET /queries`). + /// Default `true` (the manifest entry is the opt-in); `expose: false` + /// keeps it HTTP/service-callable but hidden from the agent tool list. + /// Catalog membership only — not an authorization gate. + pub expose: bool, + /// Optional MCP tool-name override; defaults to `name`. + pub tool_name: Option, +} + +impl StoredQuery { + /// `true` if the selected declaration contains insert/update/delete + /// statements — drives read-vs-mutate routing at invocation time. + pub fn is_mutation(&self) -> bool { + !self.decl.mutations.is_empty() + } + + /// The MCP tool name this query is catalogued under: the explicit + /// `tool_name` override, else the query `name`. The catalog key — + /// enforced unique across exposed queries at load. Server-side + /// consumers (the uniqueness check, the future catalog projection) read + /// this; the CLI `queries list` resolves the same rule on its own DTO. + pub fn effective_tool_name(&self) -> &str { + self.tool_name.as_deref().unwrap_or(&self.name) + } +} + +/// A loaded, identity-checked stored-query registry for one graph. +#[derive(Debug, Clone, Default)] +pub struct QueryRegistry { + by_name: BTreeMap, +} + +/// In-memory registry entry before file I/O. Used by [`QueryRegistry::load`] +/// (after reading each `.gq` from disk) and directly by tests. +#[derive(Debug, Clone)] +pub struct RegistrySpec { + pub name: String, + pub source: String, + pub expose: bool, + pub tool_name: Option, +} + +/// A single registry load failure. Collected (not fail-fast) so a bad +/// `omnigraph.yaml` surfaces every broken entry at once, matching the +/// bad-policy-YAML posture. +#[derive(Debug, Clone)] +pub struct LoadError { + /// The offending query name, when the failure is entry-scoped. + pub query: Option, + pub message: String, +} + +impl std::fmt::Display for LoadError { + fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + match &self.query { + Some(name) => write!(f, "stored query '{name}': {}", self.message), + None => write!(f, "stored query registry: {}", self.message), + } + } +} + +impl QueryRegistry { + /// Build a registry from in-memory specs: parse each source, select + /// the declaration whose symbol equals the manifest key, and assert + /// they agree. Collects every failure. No schema type-checking here + /// — that is [`check`]. + pub fn from_specs(specs: Vec) -> Result> { + let mut by_name = BTreeMap::new(); + let mut errors = Vec::new(); + + for spec in specs { + match parse_query(&spec.source) { + Ok(file) => { + match file.queries.into_iter().find(|q| q.name == spec.name) { + Some(decl) => { + by_name.insert( + spec.name.clone(), + StoredQuery { + name: spec.name, + source: Arc::from(spec.source), + decl, + expose: spec.expose, + tool_name: spec.tool_name, + }, + ); + } + None => errors.push(LoadError { + query: Some(spec.name.clone()), + message: format!( + "no `query {}` declaration found in its `.gq` file \ + (the registry key must match the query symbol)", + spec.name + ), + }), + } + } + Err(err) => errors.push(LoadError { + query: Some(spec.name), + message: format!("parse error: {err}"), + }), + } + } + + // Exposed queries are catalogued under their effective tool name; + // two claiming one name is an MCP-namespace collision. Refuse it at + // load (collected, not fail-fast), naming the loser and the winner. + // Iterating the `BTreeMap` makes the winner deterministic (the + // lexicographically-first query name; config is a map, so YAML + // declaration order isn't preserved anyway) and the error order + // stable. Scoped to a block so these borrows of `by_name` end + // before it is moved into `Self`. + { + let mut claimed: BTreeMap<&str, &str> = BTreeMap::new(); + for query in by_name.values().filter(|q| q.expose) { + let tool = query.effective_tool_name(); + if let Some(winner) = claimed.insert(tool, &query.name) { + errors.push(LoadError { + query: Some(query.name.clone()), + message: format!( + "MCP tool name '{tool}' already claimed by exposed query '{winner}'" + ), + }); + } + } + } + + if errors.is_empty() { + Ok(Self { by_name }) + } else { + Err(errors) + } + } + + /// Read each registry entry's `.gq` file from disk and build the + /// registry. `entries` is either the top-level `queries` map (single + /// mode) or a graph's `queries` map (multi mode); `config` resolves + /// each entry's relative `file:` path against `base_dir`. + pub fn load( + config: &OmnigraphConfig, + entries: &BTreeMap, + ) -> Result> { + let mut specs = Vec::with_capacity(entries.len()); + let mut errors = Vec::new(); + for (name, entry) in entries { + let path = config.resolve_query_file(&entry.file); + match fs::read_to_string(&path) { + Ok(source) => specs.push(RegistrySpec { + name: name.clone(), + source, + expose: entry.mcp.expose, + tool_name: entry.mcp.tool_name.clone(), + }), + Err(err) => errors.push(LoadError { + query: Some(name.clone()), + message: format!("cannot read '{}': {err}", path.display()), + }), + } + } + + // Parse/identity/uniqueness-check the readable specs even when some + // files failed to read, so every broken entry (I/O, parse, identity, + // tool-name collision) surfaces in one pass rather than one per + // restart. I/O errors come first (in `entries` key order), then the + // spec errors. A non-empty `errors` always fails the load. + match Self::from_specs(specs) { + Ok(registry) if errors.is_empty() => Ok(registry), + Ok(_) => Err(errors), + Err(spec_errors) => { + errors.extend(spec_errors); + Err(errors) + } + } + } + + pub fn lookup(&self, name: &str) -> Option<&StoredQuery> { + self.by_name.get(name) + } + + pub fn iter(&self) -> impl Iterator { + self.by_name.values() + } + + pub fn is_empty(&self) -> bool { + self.by_name.is_empty() + } + + pub fn len(&self) -> usize { + self.by_name.len() + } +} + +/// A stored query that fails to type-check against the live schema — +/// e.g. it references a node/edge type or property that was renamed or +/// removed by a migration. Breakages **block server boot** (same posture +/// as bad policy YAML), surfacing schema drift at the deploy boundary +/// rather than silently at invocation time. +#[derive(Debug, Clone)] +pub struct Breakage { + pub query: String, + pub message: String, +} + +/// A non-blocking advisory found during validation. Logged at boot; +/// never blocks startup. Currently: an MCP-exposed query that declares a +/// parameter an agent cannot realistically supply. +#[derive(Debug, Clone)] +pub struct Warning { + pub query: String, + pub message: String, +} + +/// Outcome of validating a registry against a schema. Breakages are +/// fatal (boot refuses); warnings are advisory. +#[derive(Debug, Clone, Default)] +pub struct CheckReport { + pub breakages: Vec, + pub warnings: Vec, +} + +impl CheckReport { + pub fn has_breakages(&self) -> bool { + !self.breakages.is_empty() + } + + pub fn is_clean(&self) -> bool { + self.breakages.is_empty() && self.warnings.is_empty() + } +} + +/// Validate a loaded registry against the live schema. +/// +/// Pure over `(registry, catalog)` — takes an already-parsed registry and +/// a catalog, so it is callable both at server boot (with the engine's +/// `catalog()`) and offline from the CLI (`omnigraph queries check`), +/// without coupling to server config or an open engine connection. +/// +/// Every query is type-checked via the same `typecheck_query_decl` the +/// engine runs for inline queries — no parallel implementation. Failures +/// are **collected, not fail-fast**, so an operator sees every broken +/// query in one pass. +/// +/// Advisory lint (warn, never block): an `mcp.expose: true` query that +/// declares a `Vector(N)` parameter. An LLM cannot supply a raw embedding +/// vector; such a query should take a `String` parameter and let the +/// engine embed it server-side at query time. Service-to-service callers +/// may legitimately pass vectors, so this warns rather than rejects. +pub fn check(registry: &QueryRegistry, catalog: &Catalog) -> CheckReport { + let mut report = CheckReport::default(); + for query in registry.iter() { + if let Err(err) = typecheck_query_decl(catalog, &query.decl) { + report.breakages.push(Breakage { + query: query.name.clone(), + message: err.to_string(), + }); + } + if query.expose { + for param in &query.decl.params { + // Resolve to the structured type via the compiler's own + // resolver rather than string-matching `Vector(` — one + // canonical definition of "is a vector", so this lint can't + // drift from how the parser/type system spells the type. + let is_vector = PropType::from_param_type_name(¶m.type_name, param.nullable) + .is_some_and(|pt| matches!(pt.scalar, ScalarType::Vector(_))); + if is_vector { + report.warnings.push(Warning { + query: query.name.clone(), + message: format!( + "MCP-exposed query declares a `{}` parameter `${}` that agents \ + cannot supply; use a `String` parameter for server-side embedding", + param.type_name, param.name + ), + }); + } + } + } + } + report +} + +/// Format every breakage in a registry check report into a multi-line +/// operator-facing message, naming each offending query. +pub fn format_check_breakages(label: &str, report: &CheckReport) -> String { + let joined = report + .breakages + .iter() + .map(|b| format!("query '{}': {}", b.query, b.message)) + .collect::>() + .join("\n "); + format!( + "graph '{label}': {} stored quer{} failed the schema check:\n {joined}", + report.breakages.len(), + if report.breakages.len() == 1 { + "y" + } else { + "ies" + } + ) +} + +#[cfg(test)] +mod tests { + use super::*; + + fn spec(name: &str, source: &str, expose: bool) -> RegistrySpec { + RegistrySpec { + name: name.to_string(), + source: source.to_string(), + expose, + tool_name: None, + } + } + + fn spec_tool(name: &str, source: &str, expose: bool, tool_name: &str) -> RegistrySpec { + RegistrySpec { + name: name.to_string(), + source: source.to_string(), + expose, + tool_name: Some(tool_name.to_string()), + } + } + + #[test] + fn key_equal_symbol_loads() { + let reg = QueryRegistry::from_specs(vec![spec( + "find_user", + "query find_user($id: String) { match { $u: User } return { $u.name } }", + true, + )]) + .unwrap(); + let q = reg.lookup("find_user").unwrap(); + assert_eq!(q.name, "find_user"); + assert!(q.expose); + assert_eq!(q.decl.params.len(), 1); + assert!(!q.is_mutation()); + // No override → the effective tool name is the query name. + assert_eq!(q.effective_tool_name(), "find_user"); + + // An explicit override is what the catalog keys on. + let with_tool = QueryRegistry::from_specs(vec![spec_tool( + "find_user", + "query find_user($id: String) { match { $u: User } return { $u.name } }", + true, + "lookup_user", + )]) + .unwrap(); + assert_eq!( + with_tool.lookup("find_user").unwrap().effective_tool_name(), + "lookup_user" + ); + } + + #[test] + fn key_mismatch_is_an_identity_error() { + let errors = QueryRegistry::from_specs(vec![spec( + "find_user", + // symbol is `lookup`, key is `find_user` — must be rejected. + "query lookup($id: String) { match { $u: User } return { $u.name } }", + false, + )]) + .unwrap_err(); + assert_eq!(errors.len(), 1); + assert_eq!(errors[0].query.as_deref(), Some("find_user")); + assert!(errors[0].message.contains("must match the query symbol")); + } + + #[test] + fn multi_query_file_selects_the_matching_symbol() { + let source = "query a($x: I64) { match { $u: User } return { $u.name } }\n\ + query b($y: String) { match { $u: User } return { $u.name } }"; + let reg = QueryRegistry::from_specs(vec![spec("b", source, false)]).unwrap(); + let q = reg.lookup("b").unwrap(); + assert_eq!(q.name, "b"); + assert_eq!(q.decl.params[0].name, "y"); + assert!(reg.lookup("a").is_none(), "only the selected symbol is registered"); + } + + #[test] + fn duplicate_exposed_tool_name_is_a_load_error() { + // Two MCP-exposed queries claiming one tool name is an ambiguity in + // the catalog key space — refused at load, naming both queries and + // the contested tool. + let errors = QueryRegistry::from_specs(vec![ + spec_tool("a", "query a() { match { $u: User } return { $u.name } }", true, "dup"), + spec_tool("b", "query b() { match { $u: User } return { $u.name } }", true, "dup"), + ]) + .unwrap_err(); + assert_eq!(errors.len(), 1); + let msg = errors[0].to_string(); + assert!(msg.contains("'dup'"), "names the contested tool: {msg}"); + assert!(msg.contains("'a'"), "names the winning query: {msg}"); + assert!(msg.contains("'b'"), "names the losing query: {msg}"); + } + + #[test] + fn duplicate_tool_name_among_unexposed_is_allowed() { + // Unexposed queries have no MCP tool, so a shared effective tool + // name is inert — must not error (pins the exposed-only scope). + let reg = QueryRegistry::from_specs(vec![ + spec_tool("a", "query a() { match { $u: User } return { $u.name } }", false, "dup"), + spec_tool("b", "query b() { match { $u: User } return { $u.name } }", false, "dup"), + ]) + .unwrap(); + assert_eq!(reg.len(), 2); + } + + #[test] + fn parse_error_surfaces_per_entry() { + let errors = + QueryRegistry::from_specs(vec![spec("broken", "query broken( {{ not valid", false)]) + .unwrap_err(); + assert_eq!(errors[0].query.as_deref(), Some("broken")); + assert!(errors[0].message.contains("parse error")); + } + + #[test] + fn errors_collect_rather_than_fail_fast() { + let errors = QueryRegistry::from_specs(vec![ + spec("good", "query good() { match { $u: User } return { $u.name } }", false), + spec("mismatch", "query other() { match { $u: User } return { $u.name } }", false), + spec("broken", "query broken(", false), + ]) + .unwrap_err(); + // `good` loads cleanly; only the mismatch and the parse error are + // reported, and both surface in one pass (not fail-fast). + assert_eq!(errors.len(), 2); + } + + #[test] + fn mutation_body_classifies_as_mutation() { + let reg = QueryRegistry::from_specs(vec![spec( + "add_user", + "query add_user($name: String) { insert User { name: $name } }", + false, + )]) + .unwrap(); + assert!(reg.lookup("add_user").unwrap().is_mutation()); + } + + // --- check(registry, catalog) --- + + use omnigraph_compiler::catalog::build_catalog; + use omnigraph_compiler::schema::parser::parse_schema; + + fn test_catalog() -> Catalog { + let schema = parse_schema( + r#" +node User { +name: String +age: I32? +embedding: Vector(4) +} +"#, + ) + .unwrap(); + build_catalog(&schema).unwrap() + } + + #[test] + fn check_passes_for_valid_query() { + let reg = QueryRegistry::from_specs(vec![spec( + "find_user", + "query find_user($name: String) { match { $u: User { name: $name } } return { $u.age } }", + false, + )]) + .unwrap(); + let report = check(®, &test_catalog()); + assert!(report.is_clean(), "unexpected: {:?}", report); + } + + #[test] + fn check_reports_unknown_type_as_breakage() { + let reg = QueryRegistry::from_specs(vec![spec( + "ghost", + // `Widget` is not in the schema. + "query ghost() { match { $w: Widget } return { $w.name } }", + false, + )]) + .unwrap(); + let report = check(®, &test_catalog()); + assert!(report.has_breakages()); + assert_eq!(report.breakages[0].query, "ghost"); + } + + #[test] + fn check_reports_unknown_property_as_breakage() { + let reg = QueryRegistry::from_specs(vec![spec( + "bad_prop", + // `User` exists but has no `nickname`. + "query bad_prop() { match { $u: User } return { $u.nickname } }", + false, + )]) + .unwrap(); + let report = check(®, &test_catalog()); + assert!(report.has_breakages()); + assert_eq!(report.breakages[0].query, "bad_prop"); + } + + #[test] + fn check_collects_every_breakage_not_fail_fast() { + let reg = QueryRegistry::from_specs(vec![ + spec("a", "query a() { match { $w: Widget } return { $w.x } }", false), + spec("b", "query b() { match { $g: Gadget } return { $g.y } }", false), + spec( + "ok", + "query ok() { match { $u: User } return { $u.name } }", + false, + ), + ]) + .unwrap(); + let report = check(®, &test_catalog()); + assert_eq!(report.breakages.len(), 2, "both bad queries reported: {:?}", report); + } + + #[test] + fn vector_param_on_exposed_query_warns() { + let reg = QueryRegistry::from_specs(vec![spec( + "vec_search", + "query vec_search($q: Vector(4)) { match { $u: User } return { $u.name } \ + order { nearest($u.embedding, $q) } limit 3 }", + true, // mcp.expose + )]) + .unwrap(); + let report = check(®, &test_catalog()); + assert!(!report.has_breakages(), "valid query: {:?}", report); + assert_eq!(report.warnings.len(), 1); + assert_eq!(report.warnings[0].query, "vec_search"); + } + + #[test] + fn vector_param_on_unexposed_query_is_silent() { + let reg = QueryRegistry::from_specs(vec![spec( + "vec_search", + "query vec_search($q: Vector(4)) { match { $u: User } return { $u.name } \ + order { nearest($u.embedding, $q) } limit 3 }", + false, // not exposed — vector param is fine for service-to-service callers + )]) + .unwrap(); + let report = check(®, &test_catalog()); + assert!(report.is_clean(), "unexpected: {:?}", report); + } + + #[test] + fn non_vector_param_on_exposed_query_does_not_warn() { + // The recommended `String` alternative on an exposed query does not + // resolve to a Vector, so the embedding advisory stays silent. Guards + // the structured type check against a false positive (and pins that + // only `Vector(_)` triggers the warning). + let reg = QueryRegistry::from_specs(vec![spec( + "search", + "query search($name: String) { match { $u: User { name: $name } } return { $u.name } }", + true, + )]) + .unwrap(); + let report = check(®, &test_catalog()); + assert!(report.is_clean(), "no breakage or warning expected: {:?}", report); + } + + // --- catalog projection (api::query_catalog_entry) --- + + #[test] + fn catalog_entry_projects_every_param_kind() { + use crate::api::{self, ParamKind}; + let reg = QueryRegistry::from_specs(vec![spec_tool( + "all_types", + "query all_types($s: String, $i: I32, $big: I64, $u: U64, $f: F64, $b: Bool, \ + $d: Date, $dt: DateTime, $blob: Blob, $opt: String?, $list: [I32], $vec: Vector(4)) \ + { match { $x: User } return { $x.name } }", + true, + "all", + )]) + .unwrap(); + let entry = api::query_catalog_entry(reg.lookup("all_types").unwrap()); + assert_eq!(entry.name, "all_types"); + assert_eq!(entry.tool_name, "all"); + assert!(!entry.mutation); + + let by: std::collections::HashMap<_, _> = + entry.params.iter().map(|p| (p.name.as_str(), p)).collect(); + assert_eq!(by["s"].kind, ParamKind::String); + assert_eq!(by["i"].kind, ParamKind::Int); + assert_eq!(by["big"].kind, ParamKind::BigInt, "I64 → bigint (string on the wire)"); + assert_eq!(by["u"].kind, ParamKind::BigInt, "U64 → bigint"); + assert_eq!(by["f"].kind, ParamKind::Float); + assert_eq!(by["b"].kind, ParamKind::Bool); + assert_eq!(by["d"].kind, ParamKind::Date); + assert_eq!(by["dt"].kind, ParamKind::DateTime); + assert_eq!(by["blob"].kind, ParamKind::Blob); + assert!(!by["s"].nullable); + assert!(by["opt"].nullable, "String? → nullable"); + assert_eq!(by["list"].kind, ParamKind::List); + assert_eq!(by["list"].item_kind, Some(ParamKind::Int), "[I32] → list of int"); + assert_eq!(by["vec"].kind, ParamKind::Vector); + assert_eq!(by["vec"].vector_dim, Some(4)); + } + + #[test] + fn catalog_entry_flags_mutation_and_empty_params() { + use crate::api; + let reg = QueryRegistry::from_specs(vec![spec( + "add_user", + "query add_user($name: String) { insert User { name: $name } }", + true, + )]) + .unwrap(); + let entry = api::query_catalog_entry(reg.lookup("add_user").unwrap()); + assert!(entry.mutation, "insert body → mutation flag"); + + let reg2 = QueryRegistry::from_specs(vec![spec( + "no_params", + "query no_params() { match { $u: User } return { $u.name } }", + true, + )]) + .unwrap(); + let entry2 = api::query_catalog_entry(reg2.lookup("no_params").unwrap()); + assert!(entry2.params.is_empty(), "no declared params → empty list"); + } + + // --- load() error collection (file I/O + parse in one pass) --- + + #[test] + fn load_collects_io_and_parse_errors_in_one_pass() { + use crate::config::load_config; + let temp = tempfile::tempdir().unwrap(); + std::fs::write( + temp.path().join("good.gq"), + "query good() { match { $u: User } return { $u.name } }", + ) + .unwrap(); + std::fs::write(temp.path().join("broken.gq"), "query broken( {{ not valid").unwrap(); + // `missing.gq` is deliberately not written (an I/O failure). + std::fs::write( + temp.path().join("omnigraph.yaml"), + "queries:\n good:\n file: ./good.gq\n \ + missing:\n file: ./missing.gq\n broken:\n file: ./broken.gq\n", + ) + .unwrap(); + let config = load_config(Some(&temp.path().join("omnigraph.yaml"))).unwrap(); + + let errors = QueryRegistry::load(&config, config.query_entries()).unwrap_err(); + let joined = errors.iter().map(|e| e.to_string()).collect::>().join("\n"); + // Both the missing file AND the parse error surface in one pass — + // the I/O failure must not mask the parse failure. + assert!(joined.contains("missing"), "I/O error must surface: {joined}"); + assert!( + joined.contains("broken") && joined.contains("parse error"), + "the parse error in a readable file must surface in the same pass: {joined}" + ); + assert!(!joined.contains("'good'"), "the valid entry is not an error: {joined}"); + } +} diff --git a/crates/omnigraph-server/src/registry.rs b/crates/omnigraph-server/src/registry.rs index 5897ad1..54115e4 100644 --- a/crates/omnigraph-server/src/registry.rs +++ b/crates/omnigraph-server/src/registry.rs @@ -29,6 +29,7 @@ use tokio::sync::Mutex; use crate::identity::GraphKey; use crate::policy::PolicyEngine; +use crate::queries::QueryRegistry; /// Open handle for a single graph in the registry. Cheap to clone (`Arc`-wrapped /// engine + policy). Cluster-mode handlers extract this via @@ -47,6 +48,11 @@ pub struct GraphHandle { /// `_as` writers"; the HTTP-layer `require_bearer_auth` middleware still /// runs regardless. pub policy: Option>, + /// Per-graph stored-query registry, loaded and validated at + /// startup. `None` means the operator declared no stored queries for + /// this graph — `POST /queries/{name}` then 404s. Mirrors the + /// optional `policy` shape. + pub queries: Option>, } /// Immutable snapshot of the registry's current state. Replaced atomically @@ -245,6 +251,7 @@ fn canonicalize_handle_uri( uri: canonical_uri.clone(), engine: Arc::clone(&handle.engine), policy: handle.policy.clone(), + queries: handle.queries.clone(), }); Ok((canonical_uri, canonical_handle)) } @@ -276,6 +283,7 @@ mod tests { uri: graph_uri, engine: Arc::new(engine), policy: None, + queries: None, }) } @@ -340,12 +348,14 @@ mod tests { uri: shared_uri.clone(), engine: Arc::clone(&engine), policy: None, + queries: None, }); let h2 = Arc::new(GraphHandle { key: GraphKey::cluster(GraphId::try_from("beta").unwrap()), uri: shared_uri, engine, policy: None, + queries: None, }); let registry = GraphRegistry::new(); @@ -411,12 +421,14 @@ mod tests { uri: shared_uri.clone(), engine: Arc::clone(&engine), policy: None, + queries: None, }); let h2 = Arc::new(GraphHandle { key: GraphKey::cluster(GraphId::try_from("beta").unwrap()), uri: shared_uri, engine, policy: None, + queries: None, }); let err = match GraphRegistry::from_handles(vec![h1, h2]) { Ok(_) => panic!("expected DuplicateUri, got Ok"), diff --git a/crates/omnigraph-server/tests/openapi.rs b/crates/omnigraph-server/tests/openapi.rs index a2542db..3d13e74 100644 --- a/crates/omnigraph-server/tests/openapi.rs +++ b/crates/omnigraph-server/tests/openapi.rs @@ -168,6 +168,8 @@ const EXPECTED_PATHS: &[&str] = &[ "/export", "/change", "/mutate", + "/queries", + "/queries/{name}", "/schema", "/schema/apply", "/ingest", @@ -701,6 +703,8 @@ fn protected_endpoints_reference_bearer_token_security() { ("/read", "post"), ("/change", "post"), ("/schema/apply", "post"), + ("/queries", "get"), + ("/queries/{name}", "post"), ("/ingest", "post"), ("/export", "post"), ("/snapshot", "get"), @@ -913,6 +917,34 @@ fn post_endpoints_have_request_body() { } } +#[test] +fn invoke_stored_query_request_body_is_optional() { + let doc = openapi_json(); + let request_body = &doc["paths"]["/queries/{name}"]["post"]["requestBody"]; + assert!( + request_body.is_object(), + "POST /queries/{{name}} should document its optional request body" + ); + assert_eq!( + request_body["required"].as_bool().unwrap_or(false), + false, + "stored-query invocation body should be optional" + ); + let schema = &request_body["content"]["application/json"]["schema"]; + let ref_path = schema["$ref"] + .as_str() + .or_else(|| { + schema["oneOf"] + .as_array() + .and_then(|schemas| schemas.iter().find_map(|schema| schema["$ref"].as_str())) + }) + .unwrap(); + assert!( + ref_path.contains("InvokeStoredQueryRequest"), + "POST /queries/{{name}} requestBody should reference InvokeStoredQueryRequest, got {ref_path}" + ); +} + // --------------------------------------------------------------------------- // Serialization round-trip test // --------------------------------------------------------------------------- @@ -1117,6 +1149,7 @@ async fn app_for_multi_mode(graph_ids: &[&str]) -> (Vec, Rout uri: graph_uri, engine: Arc::new(engine), policy: None, + queries: None, })); dirs.push(dir); } diff --git a/crates/omnigraph-server/tests/server.rs b/crates/omnigraph-server/tests/server.rs index 3ace80e..4a49a14 100644 --- a/crates/omnigraph-server/tests/server.rs +++ b/crates/omnigraph-server/tests/server.rs @@ -8,7 +8,7 @@ use axum::body::{Body, to_bytes}; use axum::http::header::AUTHORIZATION; use axum::http::{Method, Request, StatusCode}; use lance::index::DatasetIndexExt; -use omnigraph::db::{Omnigraph, ReadTarget, SchemaApplyOptions}; +use omnigraph::db::{Omnigraph, ReadTarget}; use omnigraph::error::OmniError; use omnigraph::loader::{LoadMode, load_jsonl}; use omnigraph_policy::{PolicyChecker, PolicyEngine}; @@ -16,6 +16,7 @@ use omnigraph_server::api::{ BranchCreateRequest, BranchMergeRequest, ChangeRequest, ErrorOutput, ExportRequest, IngestRequest, QueryRequest, ReadRequest, SchemaApplyRequest, SchemaOutput, }; +use omnigraph_server::queries::{QueryRegistry, RegistrySpec}; use omnigraph_server::{AppState, build_app}; use serde_json::{Value, json}; use serial_test::serial; @@ -141,6 +142,469 @@ fn graph_path(root: &Path) -> PathBuf { root.join("server.omni") } +fn stored_query_registry(specs: &[(&str, &str, bool)]) -> QueryRegistry { + QueryRegistry::from_specs( + specs + .iter() + .map(|(name, source, expose)| RegistrySpec { + name: name.to_string(), + source: source.to_string(), + expose: *expose, + tool_name: None, + }) + .collect(), + ) + .expect("specs parse and key==symbol") +} + +#[tokio::test] +async fn server_boots_with_a_valid_stored_query_registry() { + // A stored query that type-checks against the fixture schema + // (`Person { name, age }`) must let the server boot. + let temp = init_loaded_graph().await; + let graph = graph_path(temp.path()); + let registry = stored_query_registry(&[( + "find_person", + "query find_person($name: String) { match { $p: Person { name: $name } } return { $p.age } }", + false, + )]); + let state = AppState::open_single_with_queries( + graph.to_string_lossy().to_string(), + vec![], + None, + registry, + ) + .await; + assert!(state.is_ok(), "valid registry should boot: {:?}", state.err()); +} + +#[tokio::test] +async fn server_refuses_boot_on_type_broken_stored_query() { + // A stored query referencing a type not in the schema (`Widget`) + // must abort boot, naming the offending query. + let temp = init_loaded_graph().await; + let graph = graph_path(temp.path()); + let registry = stored_query_registry(&[( + "ghost", + "query ghost() { match { $w: Widget } return { $w.name } }", + false, + )]); + let result = AppState::open_single_with_queries( + graph.to_string_lossy().to_string(), + vec![], + None, + registry, + ) + .await; + // `AppState` is not `Debug`, so match rather than `expect_err`. + let err = match result { + Ok(_) => panic!("type-broken stored query must refuse boot"), + Err(err) => err, + }; + let msg = err.to_string(); + assert!(msg.contains("ghost"), "error should name the broken query: {msg}"); + assert!( + msg.contains("schema check"), + "error should mention the schema check: {msg}" + ); +} + +/// Build a single-mode app with a stored-query registry plus a bearer→actor +/// pairing and a policy, so invoke tests exercise the `invoke_query` +/// boundary gate and the inner read/change gates together. +async fn app_with_stored_queries( + specs: &[(&str, &str, bool)], + tokens: &[(&str, &str)], + policy: &str, +) -> (tempfile::TempDir, Router) { + let temp = init_loaded_graph().await; + let graph = graph_path(temp.path()); + let policy_path = temp.path().join("policy.yaml"); + fs::write(&policy_path, policy).unwrap(); + let registry = stored_query_registry(specs); + let state = AppState::open_single_with_queries( + graph.to_string_lossy().to_string(), + tokens + .iter() + .map(|(actor, token)| ((*actor).to_string(), (*token).to_string())) + .collect(), + Some(&policy_path), + registry, + ) + .await + .unwrap(); + (temp, build_app(state)) +} + +/// - `act-invoke`: invoke_query + read (stored reads, not mutations) +/// - `act-full`: invoke_query + read + change (stored mutations) +/// - `act-noinvoke`: read only, no invoke_query (boundary-denied) +/// - `act-invokeonly`: invoke_query only, no read (clears the boundary, inner read denies) +const INVOKE_POLICY_YAML: &str = r#" +version: 1 +groups: + invokers: ["act-invoke"] + full: ["act-full"] + readers: ["act-noinvoke"] + invoke_only: ["act-invokeonly"] +protected_branches: [main] +rules: + # invoke_query is graph-scoped — its own rules, no branch_scope. + - id: invokers-can-invoke + allow: + actors: { group: invokers } + actions: [invoke_query] + - id: full-can-invoke + allow: + actors: { group: full } + actions: [invoke_query] + - id: invoke-only-can-invoke + allow: + actors: { group: invoke_only } + actions: [invoke_query] + # read / change are branch-scoped. + - id: invokers-can-read + allow: + actors: { group: invokers } + actions: [read] + branch_scope: any + - id: full-can-read-change + allow: + actors: { group: full } + actions: [read, change] + branch_scope: any + - id: readers-can-read + allow: + actors: { group: readers } + actions: [read] + branch_scope: any +"#; + +const STORED_QUERY_SCHEMA_APPLY_POLICY_YAML: &str = r#" +version: 1 +groups: + admins: [act-ragnor] +protected_branches: [main] +rules: + - id: admins-can-invoke + allow: + actors: { group: admins } + actions: [invoke_query] + - id: admins-can-read + allow: + actors: { group: admins } + actions: [read] + branch_scope: any + - id: admins-can-schema-apply + allow: + actors: { group: admins } + actions: [schema_apply] + target_branch_scope: protected +"#; + +const FIND_PERSON_GQ: &str = + "query find_person($name: String) { match { $p: Person { name: $name } } return { $p.age } }"; + +fn invoke_request(name: &str, token: &str, body: Value) -> Request { + Request::builder() + .uri(format!("/queries/{name}")) + .method(Method::POST) + .header("content-type", "application/json") + .header("authorization", format!("Bearer {token}")) + .body(Body::from(serde_json::to_vec(&body).unwrap())) + .unwrap() +} + +fn invoke_request_bytes( + name: &str, + token: &str, + body: impl Into, + content_type: Option<&str>, +) -> Request { + let mut builder = Request::builder() + .uri(format!("/queries/{name}")) + .method(Method::POST) + .header("authorization", format!("Bearer {token}")); + if let Some(content_type) = content_type { + builder = builder.header("content-type", content_type); + } + builder.body(body.into()).unwrap() +} + +#[tokio::test(flavor = "multi_thread")] +async fn invoke_stored_read_returns_rows() { + let (_temp, app) = app_with_stored_queries( + &[("find_person", FIND_PERSON_GQ, false)], + &[("act-invoke", "t-invoke")], + INVOKE_POLICY_YAML, + ) + .await; + let (status, body) = json_response( + &app, + invoke_request("find_person", "t-invoke", json!({ "params": { "name": "Alice" } })), + ) + .await; + assert_eq!(status, StatusCode::OK, "body: {body}"); + assert_eq!(body["query_name"], "find_person"); + assert_eq!(body["row_count"], 1, "Alice is in the fixture; body: {body}"); + assert!(body["rows"].is_array(), "read envelope shape; body: {body}"); +} + +#[tokio::test(flavor = "multi_thread")] +async fn invoke_stored_read_accepts_absent_or_empty_body() { + let no_param_query = "query list_people() { match { $p: Person } return { $p.name } }"; + let (_temp, app) = app_with_stored_queries( + &[("list_people", no_param_query, false)], + &[("act-invoke", "t-invoke")], + INVOKE_POLICY_YAML, + ) + .await; + + let (status, body) = json_response( + &app, + invoke_request_bytes("list_people", "t-invoke", Body::empty(), None), + ) + .await; + assert_eq!(status, StatusCode::OK, "body: {body}"); + assert_eq!(body["query_name"], "list_people"); + + let (status, body) = json_response( + &app, + invoke_request_bytes( + "list_people", + "t-invoke", + Body::empty(), + Some("application/json"), + ), + ) + .await; + assert_eq!(status, StatusCode::OK, "body: {body}"); + + let (status, body) = json_response( + &app, + invoke_request_bytes( + "list_people", + "t-invoke", + Body::from("{}"), + Some("application/json"), + ), + ) + .await; + assert_eq!(status, StatusCode::OK, "body: {body}"); + + let (status, body) = json_response( + &app, + invoke_request_bytes( + "list_people", + "t-invoke", + Body::from("{"), + Some("application/json"), + ), + ) + .await; + assert_eq!(status, StatusCode::BAD_REQUEST, "body: {body}"); + assert!( + body["error"] + .as_str() + .unwrap_or_default() + .contains("invalid stored-query invocation body"), + "malformed JSON should be rejected as bad request; body: {body}" + ); +} + +#[tokio::test(flavor = "multi_thread")] +async fn invoke_stored_mutation_double_gates_on_change() { + let specs: &[(&str, &str, bool)] = &[( + "add_person", + "query add_person($name: String) { insert Person { name: $name } }", + false, + )]; + let (_temp, app) = app_with_stored_queries( + specs, + &[("act-invoke", "t-invoke"), ("act-full", "t-full")], + INVOKE_POLICY_YAML, + ) + .await; + + // Has invoke_query but NOT change → the inner change gate denies (403). + let (status, body) = json_response( + &app, + invoke_request("add_person", "t-invoke", json!({ "params": { "name": "Eve" } })), + ) + .await; + assert_eq!( + status, + StatusCode::FORBIDDEN, + "invoke_query without change must 403; body: {body}" + ); + + // Has invoke_query + change → applied. + let (status, body) = json_response( + &app, + invoke_request("add_person", "t-full", json!({ "params": { "name": "Eve" } })), + ) + .await; + assert_eq!(status, StatusCode::OK, "body: {body}"); + assert_eq!(body["affected_nodes"], 1, "body: {body}"); +} + +#[tokio::test(flavor = "multi_thread")] +async fn invoke_stored_query_bad_param_is_400() { + let (_temp, app) = app_with_stored_queries( + &[("find_person", FIND_PERSON_GQ, false)], + &[("act-invoke", "t-invoke")], + INVOKE_POLICY_YAML, + ) + .await; + // `name` is declared String; pass a number. + let (status, body) = json_response( + &app, + invoke_request("find_person", "t-invoke", json!({ "params": { "name": 123 } })), + ) + .await; + assert_eq!(status, StatusCode::BAD_REQUEST, "body: {body}"); + assert!( + body["error"].as_str().unwrap_or_default().contains("name"), + "400 should name the offending param; body: {body}" + ); +} + +#[tokio::test(flavor = "multi_thread")] +async fn invoke_unknown_query_and_denied_actor_return_identical_404() { + let (_temp, app) = app_with_stored_queries( + &[("find_person", FIND_PERSON_GQ, false)], + &[("act-invoke", "t-invoke"), ("act-noinvoke", "t-noinvoke")], + INVOKE_POLICY_YAML, + ) + .await; + + // Authorized actor, unknown query name → 404. + let (unknown_status, unknown_body) = + json_response(&app, invoke_request("does_not_exist", "t-invoke", json!({}))).await; + // Denied actor (no invoke_query), real query name → 404. + let (denied_status, denied_body) = json_response( + &app, + invoke_request("find_person", "t-noinvoke", json!({ "params": { "name": "Alice" } })), + ) + .await; + + assert_eq!(unknown_status, StatusCode::NOT_FOUND); + assert_eq!(denied_status, StatusCode::NOT_FOUND); + assert_eq!( + unknown_body, denied_body, + "deny must be byte-identical to a missing query (no catalog probing)" + ); +} + +#[tokio::test(flavor = "multi_thread")] +async fn invoke_query_holder_without_read_sees_403_not_404() { + // The 404-hiding is for callers WITHOUT invoke_query. An actor that + // HOLDS invoke_query but lacks `read` clears the boundary gate, then the + // inner read gate denies → 403 for an EXISTING read query, vs 404 for an + // unknown one. Existence is visible to grant-holders by design (the + // documented double-gate); this pins that actual contract. + let (_temp, app) = app_with_stored_queries( + &[("find_person", FIND_PERSON_GQ, false)], + &[("act-invokeonly", "t-invokeonly")], + INVOKE_POLICY_YAML, + ) + .await; + let (exists_status, _) = json_response( + &app, + invoke_request("find_person", "t-invokeonly", json!({ "params": { "name": "Alice" } })), + ) + .await; + let (absent_status, _) = + json_response(&app, invoke_request("does_not_exist", "t-invokeonly", json!({}))).await; + assert_eq!( + exists_status, + StatusCode::FORBIDDEN, + "an existing read query the holder can't read → inner-gate 403" + ); + assert_eq!(absent_status, StatusCode::NOT_FOUND, "unknown query still 404s"); +} + +fn get_request(uri: &str, token: &str) -> Request { + Request::builder() + .uri(uri) + .method(Method::GET) + .header("authorization", format!("Bearer {token}")) + .body(Body::empty()) + .unwrap() +} + +#[tokio::test(flavor = "multi_thread")] +async fn list_queries_returns_only_exposed_with_typed_params() { + let (_temp, app) = app_with_stored_queries( + &[ + ("find_person", FIND_PERSON_GQ, true), + ( + "add_person", + "query add_person($name: String) { insert Person { name: $name } }", + true, + ), + ("hidden", "query hidden() { match { $p: Person } return { $p.name } }", false), + ], + &[("act-invoke", "t-invoke")], + INVOKE_POLICY_YAML, + ) + .await; + let (status, body) = json_response(&app, get_request("/queries", "t-invoke")).await; + assert_eq!(status, StatusCode::OK, "body: {body}"); + + let entries = body["queries"].as_array().unwrap(); + let names: Vec<&str> = entries.iter().map(|q| q["name"].as_str().unwrap()).collect(); + assert!( + names.contains(&"find_person") && names.contains(&"add_person"), + "exposed queries listed: {names:?}" + ); + assert!(!names.contains(&"hidden"), "non-exposed query hidden from the catalog: {names:?}"); + + let fp = entries.iter().find(|q| q["name"] == "find_person").unwrap(); + assert_eq!(fp["mutation"], false); + assert_eq!(fp["tool_name"], "find_person"); + assert_eq!(fp["params"][0]["name"], "name"); + assert_eq!(fp["params"][0]["kind"], "string"); + let ap = entries.iter().find(|q| q["name"] == "add_person").unwrap(); + assert_eq!(ap["mutation"], true, "stored insert → mutation"); +} + +#[tokio::test(flavor = "multi_thread")] +async fn list_queries_is_read_gated_so_a_non_invoker_can_list() { + // The catalog is read-gated (not invoke_query-gated), so a reader who + // lacks invoke_query still enumerates the exposed queries — the + // documented probe-oracle gap until per-query Cedar filtering lands. + let (_temp, app) = app_with_stored_queries( + &[("find_person", FIND_PERSON_GQ, true)], + &[("act-noinvoke", "t-noinvoke")], + INVOKE_POLICY_YAML, + ) + .await; + let (status, body) = json_response(&app, get_request("/queries", "t-noinvoke")).await; + assert_eq!(status, StatusCode::OK, "read-gated catalog; body: {body}"); + let names: Vec<&str> = body["queries"] + .as_array() + .unwrap() + .iter() + .map(|q| q["name"].as_str().unwrap()) + .collect(); + assert!( + names.contains(&"find_person"), + "a reader lists the catalog despite lacking invoke_query: {names:?}" + ); +} + +#[tokio::test(flavor = "multi_thread")] +async fn list_queries_is_empty_when_no_registry() { + let (_temp, app) = app_for_loaded_graph_with_auth("demo-token").await; + let (status, body) = json_response(&app, get_request("/queries", "demo-token")).await; + assert_eq!(status, StatusCode::OK, "body: {body}"); + assert!( + body["queries"].as_array().unwrap().is_empty(), + "no stored-query registry → empty catalog" + ); +} + fn drifted_test_schema() -> String { fs::read_to_string(fixture("test.pg")) .unwrap() @@ -423,6 +887,83 @@ async fn schema_apply_route_updates_graph_for_authorized_admin() { ); } +#[tokio::test(flavor = "multi_thread")] +async fn schema_apply_route_rejects_stored_query_breakage_before_publish() { + let (temp, app) = app_with_stored_queries( + &[("find_person", FIND_PERSON_GQ, true)], + &[("act-ragnor", "admin-token")], + STORED_QUERY_SCHEMA_APPLY_POLICY_YAML, + ) + .await; + + let request = Request::builder() + .method(Method::POST) + .uri("/schema/apply") + .header("content-type", "application/json") + .header("authorization", "Bearer admin-token") + .body(Body::from( + serde_json::to_vec(&SchemaApplyRequest { + schema_source: renamed_age_schema(), + ..Default::default() + }) + .unwrap(), + )) + .unwrap(); + let (status, payload) = json_response(&app, request).await; + assert_eq!(status, StatusCode::BAD_REQUEST, "body: {payload}"); + let message = payload["error"].as_str().unwrap_or_default(); + assert!( + message.contains("find_person") && message.contains("schema check"), + "registry breakage should name the stored query; body: {payload}" + ); + + let reopened = Omnigraph::open(graph_path(temp.path()).to_str().unwrap()) + .await + .unwrap(); + let person = &reopened.catalog().node_types["Person"]; + assert!(person.properties.contains_key("age")); + assert!(!person.properties.contains_key("years")); + + let (invoke_status, invoke_body) = json_response( + &app, + invoke_request( + "find_person", + "admin-token", + json!({ "params": { "name": "Alice" } }), + ), + ) + .await; + assert_eq!(invoke_status, StatusCode::OK, "body: {invoke_body}"); + assert_eq!(invoke_body["row_count"], 1); +} + +#[tokio::test(flavor = "multi_thread")] +async fn schema_apply_route_noop_keeps_valid_stored_query_registry() { + let (_temp, app) = app_with_stored_queries( + &[("find_person", FIND_PERSON_GQ, true)], + &[("act-ragnor", "admin-token")], + STORED_QUERY_SCHEMA_APPLY_POLICY_YAML, + ) + .await; + + let request = Request::builder() + .method(Method::POST) + .uri("/schema/apply") + .header("content-type", "application/json") + .header("authorization", "Bearer admin-token") + .body(Body::from( + serde_json::to_vec(&SchemaApplyRequest { + schema_source: fs::read_to_string(fixture("test.pg")).unwrap(), + ..Default::default() + }) + .unwrap(), + )) + .unwrap(); + let (status, payload) = json_response(&app, request).await; + assert_eq!(status, StatusCode::OK, "body: {payload}"); + assert_eq!(payload["applied"], false); +} + #[tokio::test] async fn schema_apply_route_requires_schema_apply_policy_permission() { let (_temp, app) = app_for_graph_with_auth_tokens_and_policy( @@ -4690,6 +5231,7 @@ mod multi_graph_startup { uri: graph_uri, engine: Arc::new(engine), policy: None, + queries: None, })); dirs.push(dir); } @@ -4985,12 +5527,14 @@ graphs: uri: graph_uri.clone(), engine: Arc::clone(&engine), policy: None, + queries: None, }); let beta = Arc::new(GraphHandle { key: GraphKey::cluster(GraphId::try_from("beta").unwrap()), uri: format!("file://{graph_uri}/"), engine, policy: None, + queries: None, }); match GraphRegistry::from_handles(vec![alpha, beta]) { @@ -5016,6 +5560,7 @@ graphs: uri: format!("file://{graph_uri}/"), engine: Arc::new(engine), policy: None, + queries: None, }); let registry = GraphRegistry::from_handles(vec![handle]).unwrap(); @@ -5138,11 +5683,11 @@ graphs: let err = load_server_settings(Some(&config_path), None, None, None, true).unwrap_err(); let msg = err.to_string(); assert!( - msg.contains("top-level `policy.file` is single-graph/CLI-local policy only"), - "expected single-graph policy guidance, got: {msg}" + msg.contains("top-level") && msg.contains("policy.file") && msg.contains("not honored"), + "expected top-level-not-honored guidance, got: {msg}" ); assert!( - msg.contains("graphs..policy.file"), + msg.contains("graphs."), "expected per-graph migration guidance, got: {msg}" ); assert!( @@ -5151,6 +5696,88 @@ graphs: ); } + #[test] + fn mode_inference_multi_rejects_top_level_queries() { + // Symmetric to the policy guard: a top-level `queries:` block in + // multi-graph mode is not honored (each graph uses its own), so it + // is a loud error rather than a silent no-op. + let temp = tempfile::tempdir().unwrap(); + let config_path = temp.path().join("omnigraph.yaml"); + fs::write( + &config_path, + "queries:\n q:\n file: ./q.gq\ngraphs:\n alpha:\n uri: /tmp/alpha.omni\n", + ) + .unwrap(); + let err = load_server_settings(Some(&config_path), None, None, None, true).unwrap_err(); + let msg = err.to_string(); + assert!( + msg.contains("queries") && msg.contains("not honored"), + "top-level queries must be rejected in multi-graph mode: {msg}" + ); + } + + #[test] + fn single_mode_named_graph_rejects_top_level_blocks() { + // Serving a graph by name (`--target`/`server.graph`) uses its + // per-graph block; a populated top-level block would be silently + // shadowed, so boot refuses and names the per-graph location. + let temp = tempfile::tempdir().unwrap(); + let config_path = temp.path().join("omnigraph.yaml"); + fs::write( + &config_path, + "policy:\n file: ./top.yaml\ngraphs:\n prod:\n uri: /tmp/prod.omni\n", + ) + .unwrap(); + let err = + load_server_settings(Some(&config_path), None, Some("prod".to_string()), None, true) + .unwrap_err(); + let msg = err.to_string(); + assert!( + msg.contains("prod") && msg.contains("policy.file") && msg.contains("graphs.prod"), + "named single-mode + top-level policy must refuse, naming the graph: {msg}" + ); + } + + #[test] + fn single_mode_named_graph_uses_per_graph_policy_and_queries() { + // The identity rule: `--target prod` attaches `graphs.prod`'s own + // policy + queries, not the top-level ones (which are absent here). + let temp = tempfile::tempdir().unwrap(); + fs::write( + temp.path().join("prod.gq"), + "query pq() { match { $u: User } return { $u.name } }", + ) + .unwrap(); + let config_path = temp.path().join("omnigraph.yaml"); + fs::write( + &config_path, + "graphs:\n prod:\n uri: /tmp/prod.omni\n policy:\n file: ./prod-policy.yaml\n \ + queries:\n pq:\n file: ./prod.gq\n", + ) + .unwrap(); + let settings = + load_server_settings(Some(&config_path), None, Some("prod".to_string()), None, true) + .unwrap(); + match settings.mode { + ServerConfigMode::Single { + graph_id, + policy_file, + queries, + .. + } => { + assert_eq!(graph_id, "prod", "named single-mode keeps graph identity"); + assert!( + policy_file + .as_ref() + .is_some_and(|p| p.ends_with("prod-policy.yaml")), + "per-graph policy attached: {policy_file:?}" + ); + assert!(queries.lookup("pq").is_some(), "per-graph query attached"); + } + other => panic!("expected Single mode, got {other:?}"), + } + } + #[test] fn mode_inference_normalizes_multi_graph_uris() { let temp = tempfile::tempdir().unwrap(); @@ -5383,6 +6010,7 @@ graphs: uri: graph_uri, engine: Arc::new(engine), policy: None, + queries: None, }); let tokens = vec![("act-andrew".to_string(), "secret-token".to_string())]; let workload = omnigraph_server::workload::WorkloadController::from_env(); @@ -5450,6 +6078,7 @@ graphs: uri: graph_uri, engine: Arc::new(engine), policy: None, + queries: None, })); } diff --git a/crates/omnigraph/src/db/omnigraph.rs b/crates/omnigraph/src/db/omnigraph.rs index eb58623..9d1403d 100644 --- a/crates/omnigraph/src/db/omnigraph.rs +++ b/crates/omnigraph/src/db/omnigraph.rs @@ -67,6 +67,12 @@ pub struct SchemaApplyResult { pub steps: Vec, } +#[derive(Debug, Clone)] +pub struct SchemaApplyPreview { + pub plan: SchemaMigrationPlan, + pub catalog: Catalog, +} + /// Top-level handle to an Omnigraph database. /// /// An Omnigraph is a Lance-native graph database with git-style branching. @@ -493,6 +499,14 @@ impl Omnigraph { schema_apply::plan_schema(self, desired_schema_source, options).await } + pub async fn preview_schema_apply_with_options( + &self, + desired_schema_source: &str, + options: SchemaApplyOptions, + ) -> Result { + schema_apply::preview_schema_apply(self, desired_schema_source, options).await + } + pub async fn apply_schema(&self, desired_schema_source: &str) -> Result { self.apply_schema_as(desired_schema_source, SchemaApplyOptions::default(), None) .await @@ -523,7 +537,28 @@ impl Omnigraph { options: SchemaApplyOptions, actor: Option<&str>, ) -> Result { - schema_apply::apply_schema(self, desired_schema_source, options, actor).await + self.apply_schema_as_with_catalog_check(desired_schema_source, options, actor, |_| Ok(())) + .await + } + + pub async fn apply_schema_as_with_catalog_check( + &self, + desired_schema_source: &str, + options: SchemaApplyOptions, + actor: Option<&str>, + validate_catalog: F, + ) -> Result + where + F: FnOnce(&Catalog) -> Result<()>, + { + schema_apply::apply_schema( + self, + desired_schema_source, + options, + actor, + validate_catalog, + ) + .await } pub(crate) async fn ensure_schema_apply_idle(&self, operation: &str) -> Result<()> { diff --git a/crates/omnigraph/src/db/omnigraph/schema_apply.rs b/crates/omnigraph/src/db/omnigraph/schema_apply.rs index 0dcf0f9..35fe161 100644 --- a/crates/omnigraph/src/db/omnigraph/schema_apply.rs +++ b/crates/omnigraph/src/db/omnigraph/schema_apply.rs @@ -48,50 +48,17 @@ pub(super) async fn plan_schema( Ok(plan) } -pub(super) async fn apply_schema( - db: &Omnigraph, - desired_schema_source: &str, - options: SchemaApplyOptions, - actor: Option<&str>, -) -> Result { - // Engine-layer policy gate (MR-722 chassis core). - // - // Fires BEFORE acquiring the schema-apply lock or doing any other - // work. When no PolicyChecker is installed this is a no-op and - // the apply path behaves exactly as it did before MR-722. When - // a PolicyChecker IS installed and the actor is None, this is a - // hard error — see Omnigraph::enforce's docstring for the - // forget-the-actor-footgun reasoning. - // - // Scope is TargetBranch("main") to match the HTTP-layer convention - // for SchemaApply: branch=None, target_branch=Some("main"). Cedar - // policies in the wild use `target_branch_scope: protected` to - // gate schema applies, so the engine-layer call has to set the - // target_branch shape that activates that predicate. Wrong scope - // here = silent policy mismatch with HTTP. See - // `omnigraph_policy::ResourceScope::to_branch_pair` for the mapping. - db.enforce( - omnigraph_policy::PolicyAction::SchemaApply, - &omnigraph_policy::ResourceScope::TargetBranch("main".to_string()), - actor, - )?; - - acquire_schema_apply_lock(db).await?; - let result = apply_schema_with_lock(db, desired_schema_source, options).await; - let release_result = release_schema_apply_lock(db).await; - match (result, release_result) { - (Ok(result), Ok(())) => Ok(result), - (Ok(_), Err(err)) => Err(err), - (Err(err), Ok(())) => Err(err), - (Err(err), Err(_)) => Err(err), - } +struct PlannedSchemaApply { + plan: SchemaMigrationPlan, + desired_ir: SchemaIR, + desired_catalog: Catalog, } -pub(super) async fn apply_schema_with_lock( +async fn plan_schema_for_apply( db: &Omnigraph, desired_schema_source: &str, options: SchemaApplyOptions, -) -> Result { +) -> Result { db.ensure_schema_state_valid().await?; let branches = db.coordinator.read().await.all_branches().await?; // Skip `main` and internal system branches. The schema-apply lock branch @@ -123,6 +90,87 @@ pub(super) async fn apply_schema_with_lock( .unwrap_or_else(|| "unsupported schema migration plan".to_string()); return Err(OmniError::manifest(message)); } + + let mut desired_catalog = build_catalog_from_ir(&desired_ir)?; + fixup_blob_schemas(&mut desired_catalog); + Ok(PlannedSchemaApply { + plan, + desired_ir, + desired_catalog, + }) +} + +pub(super) async fn preview_schema_apply( + db: &Omnigraph, + desired_schema_source: &str, + options: SchemaApplyOptions, +) -> Result { + let planned = plan_schema_for_apply(db, desired_schema_source, options).await?; + Ok(SchemaApplyPreview { + plan: planned.plan, + catalog: planned.desired_catalog, + }) +} + +pub(super) async fn apply_schema( + db: &Omnigraph, + desired_schema_source: &str, + options: SchemaApplyOptions, + actor: Option<&str>, + validate_catalog: F, +) -> Result +where + F: FnOnce(&Catalog) -> Result<()>, +{ + // Engine-layer policy gate (MR-722 chassis core). + // + // Fires BEFORE acquiring the schema-apply lock or doing any other + // work. When no PolicyChecker is installed this is a no-op and + // the apply path behaves exactly as it did before MR-722. When + // a PolicyChecker IS installed and the actor is None, this is a + // hard error — see Omnigraph::enforce's docstring for the + // forget-the-actor-footgun reasoning. + // + // Scope is TargetBranch("main") to match the HTTP-layer convention + // for SchemaApply: branch=None, target_branch=Some("main"). Cedar + // policies in the wild use `target_branch_scope: protected` to + // gate schema applies, so the engine-layer call has to set the + // target_branch shape that activates that predicate. Wrong scope + // here = silent policy mismatch with HTTP. See + // `omnigraph_policy::ResourceScope::to_branch_pair` for the mapping. + db.enforce( + omnigraph_policy::PolicyAction::SchemaApply, + &omnigraph_policy::ResourceScope::TargetBranch("main".to_string()), + actor, + )?; + + acquire_schema_apply_lock(db).await?; + let result = apply_schema_with_lock(db, desired_schema_source, options, validate_catalog).await; + let release_result = release_schema_apply_lock(db).await; + match (result, release_result) { + (Ok(result), Ok(())) => Ok(result), + (Ok(_), Err(err)) => Err(err), + (Err(err), Ok(())) => Err(err), + (Err(err), Err(_)) => Err(err), + } +} + +pub(super) async fn apply_schema_with_lock( + db: &Omnigraph, + desired_schema_source: &str, + options: SchemaApplyOptions, + validate_catalog: F, +) -> Result +where + F: FnOnce(&Catalog) -> Result<()>, +{ + let planned = plan_schema_for_apply(db, desired_schema_source, options).await?; + validate_catalog(&planned.desired_catalog)?; + let PlannedSchemaApply { + plan, + desired_ir, + desired_catalog, + } = planned; if plan.steps.is_empty() { return Ok(SchemaApplyResult { supported: true, @@ -132,9 +180,6 @@ pub(super) async fn apply_schema_with_lock( }); } - let mut desired_catalog = build_catalog_from_ir(&desired_ir)?; - fixup_blob_schemas(&mut desired_catalog); - let snapshot = db.snapshot().await; let base_manifest_version = snapshot.version(); let mut added_tables = BTreeSet::new(); diff --git a/docs/dev/index.md b/docs/dev/index.md index d9ba5e5..600c969 100644 --- a/docs/dev/index.md +++ b/docs/dev/index.md @@ -59,6 +59,8 @@ Working documents for in-flight feature work. Removed when the work lands. |---|---| | Schema-lint chassis v1 (MR-694) — `--allow-data-loss`, soft/hard drops | [schema-lint-v1-plan.md](schema-lint-v1-plan.md) | | Inline + stored queries, request/response envelope, MCP (MR-656 / MR-976 / MR-969) | [rfc-001-queries-envelope-mcp.md](rfc-001-queries-envelope-mcp.md) | +| Config & CLI architecture — layered config, client targeting, file naming (MR-973 / MR-974 / MR-981) | [rfc-002-config-cli-architecture.md](rfc-002-config-cli-architecture.md) | +| MCP server surface — full tool parity, stored queries, modular auth (MR-969 / MR-956 / MR-974) | [rfc-003-mcp-server-surface.md](rfc-003-mcp-server-surface.md) | ## Boundary diff --git a/docs/dev/rfc-002-config-cli-architecture.md b/docs/dev/rfc-002-config-cli-architecture.md new file mode 100644 index 0000000..0a8e573 --- /dev/null +++ b/docs/dev/rfc-002-config-cli-architecture.md @@ -0,0 +1,590 @@ +# RFC: Config & CLI Architecture — Layered Config, Client Targeting, File Naming + +**Status:** Proposed +**Date:** 2026-05-30 +**Tickets:** MR-668 (multi-graph server, shipped — the dependency this builds on), MR-969 (stored queries + MCP — supplies the in-repo agent tool surface), MR-973 (quickstart / onboarding), MR-974 (agent setup surface), MR-981 (agent-friendly CLI hardening) +**Target release:** v0.8.x (tentative; phased — see Rollout) + +## Summary + +OmniGraph today has a single config file, `omnigraph.yaml`, read both by the CLI (operating the embedded engine) and by `omnigraph-server` (hosting graphs). There is **no client-side configuration that targets a *running server*** — to talk to a deployed `omnigraph-server` you drop to `curl` or the `omnigraph-ts` client. This is the one real gap in an otherwise coherent design (storage-URI addressing, multi-graph routing, per-graph policy). + +This RFC defines the config and CLI architecture that closes that gap, derived from first principles — *working backwards from what OmniGraph uniquely enables* rather than copying kubeconfig / `helix.toml`. The result: + +1. A **global-first layered config** — user-global (`~/.omnigraph/`) is the **primary, self-sufficient default**; per-project (`./omnigraph.yaml`) is an *optional* override + deployment manifest. One uniform schema, both layers optional; the CLI works from any directory with **no project file** (the `kubectl`/`aws`/`gh` posture), unlike today's project-anchored behavior. +2. A single unifying noun — the **target** — that resolves a name to a concrete `(locus, graph, sub-state, credential)` tuple, where the locus is **embedded (storage URI) XOR remote (server endpoint)**. +3. A **multi-server × multi-graph** client model (OmniGraph hosts N graphs per server and there are M servers — unlike Helix's one-cluster-one-graph). +4. **Credentials by reference, keyed by server name** (the AWS/gh/kube model) — OS keychain `omnigraph:` (preferred) → a `[]` profile in `~/.omnigraph/credentials` → `OMNIGRAPH_TOKEN[_]` env (CI). `servers.` is endpoint-only by default but may carry an explicit, secret-free `auth: { token: { env|file|command|keychain } }` source; no `credentials.yaml`; the shipped `bearer_token_env` + dotenv stay as a legacy compat path. Every committed/GitOps'd surface stays secret-free. +5. A **file-naming** decision: project and server config are **the same artifact, same name** (`omnigraph.yaml`); the only differently-named file is the user-global `config.yaml`, justified by **scope, not role**. + +The design optimizes jointly for **DX** (one command surface across embedded and remote; clone-and-go) and **AX** (agent experience: one flat resolved context, secrets structurally unreachable, branch-pinned reproducible reads, and a GitOps'd capability surface). + +## Reconciliation with shipped / planned CLI work + +Verified **against the code**, not ticket statuses (which are unreliable — e.g. MR-581 is marked done but is stale and unbuilt). Findings and the corrections they force: + +- **Noun is `graph`/`graphs`, NOT `target`/`targets`.** The config key is `graphs:` in `config.rs` and the flag is `--graph`. **This RFC uses `graphs:`/`--graph` throughout**; the unifying noun is a **`graphs:` entry** that is *embedded* (`storage:`, formerly `uri:`) XOR *remote* (`server:` + `graph_id:` defaulting to the entry key) — a typed locator (§1.1). Read any lingering `targets:`/`--target` below as `graphs:`/`--graph`. +- **`~/.omnigraph/` stands on its own merits** (Helix/aws/kube peer convention), **not** on precedent — there is **no `~/.omnigraph/` usage in the code** today. (MR-581 / MR-531 templates-into-`~/.omnigraph/` are *stale tickets, unbuilt*.) +- **Templates do not exist** in the code (no `template` command). The template mechanism is a *design question for this RFC / the init family*, not an existing foothold. +- **What actually exists in the CLI** (verified): `init, query(read), mutate(change), load, ingest, branch, schema, lint, snapshot, export, commit, policy, optimize, cleanup, graphs`. **Not built:** `serve, quickstart, template, prune, login`. `omnigraph init` exists (with `scaffold_config_if_missing`, `main.rs:1415`); the rest of the "init family" (`quickstart` MR-973, `serve` MR-970, `prune`/`init --force` MR-972/975, `mcp install`/skills MR-974, agent-mode MR-981) are **unbuilt tickets**, some stale. +- **Config still uses `aliases:`** (no `operations:` in code; MR-839 unbuilt). §6's reconciliation talks about `aliases:` as-is, noting `operations:` is a *proposed* rename. +- **`bearer_token_env` exists** (per-graph, `config.rs`); MR-971 flags a CLI-parity / server-side gap. The per-`servers.` extension lands on top of that. +- **A top-level `omnigraph lint` command exists** (verified). A stored-query *registry* validator must pick a verb that doesn't read as a competing lint/check. + +## Motivation + +Three problems, in priority order: + +- **No client→server targeting config.** The moment an operator stands up `omnigraph-server` — for bearer auth + Cedar at a network boundary + admission control + multi-graph routing — the CLI can't address it. `curl` is the fallback. There is no named, switchable, credential-carrying way to say "run this against `prod` on the team server." +- **Multi-server × multi-graph has no first-class expression.** OmniGraph genuinely runs N graphs per server across M servers. The same graph is **multi-homed** — `s3://b/prod` may be `prod` on server A, `production` on server B, and opened directly by the CLI. Today's flat `graphs:` map (name→storage-URI) can't express "graph `production` on server `prod-eu`." +- **Solo-first and embedded-first are unserved by the remote story.** A solo developer with no projects should define everything in `~`. A developer iterating locally (embedded, no server) and then pointing at staging (remote) should change *one word*, not learn a second command surface. + +MR-668 shipped the server side (multiple graphs per server). MR-969 ships the in-repo agent tool surface (stored queries / MCP). This RFC supplies the **client and config layer** that lets humans and agents target that surface coherently — the foundation under MR-973 / MR-974 / MR-981. + +## Non-Goals + +- **A control plane / dashboard for config.** Operators edit files and (for servers) restart. No runtime config-mutation API. Matches the MR-668 / MR-969 operational model. +- **Hot reload.** Restart-only for server-side config, matching MR-668 and MR-969. +- **Embedding secrets in any config file.** Credentials are by-reference; the git-ignored `auth.env_file` dotenv (or, later, the OS keychain) holds tokens. Never a committable `*.yaml`. +- **Renaming the project manifest by role.** No `omnigraph.server.yaml` / `omnigraph.client.yaml`. Role lives in sections, not filenames (see Design §3). +- **Dropping embedded mode.** Embedded-first is load-bearing for the file-naming decision; this RFC assumes it stays. +- **Cross-graph / cross-server tool listing in MCP.** Clients loop over per-graph catalogs (a MR-969 non-goal, restated). + +## Background + +OmniGraph runs on Lance 6.x: typed nodes/edges in per-type Lance datasets, atomic multi-table commits via a `__manifest` table, branchable and time-travelable. The CLI (`omnigraph`) operates the **embedded engine** directly against a storage URI — no HTTP client in its runtime dependencies. `omnigraph-server` (Axum) is a *separate* HTTP front-end over the same engine, with bearer auth + per-graph Cedar (MR-668). The two read the same `omnigraph.yaml` but never connect to each other. + +OmniGraph **already has a credentials-by-reference mechanism**, which this RFC builds on rather than replacing: `TargetConfig.bearer_token_env` names the env var holding a graph's bearer token, and `auth.env_file` points at a git-ignored dotenv (`.env.omni`) that the CLI auto-loads into the process (`load_env_file_into_process`) with real-env-vars-win precedence; `resolve_remote_bearer_token` resolves a token via env var then dotenv named lookup. `.env.omni` is already in `.gitignore`. + +The six **irreducible enablers** that drive the design (referenced as E1–E6 below): + +| # | Enabler | Consequence | +|---|---|---| +| E1 | A graph is a **self-contained storage URI**; the substrate (object store + manifest CAS) is the source of truth — no server required to read/write. | A graph is addressable **directly (embedded)**, not only via a server. | +| E2 | A server hosts **many graphs**; **many servers** exist. | The remote address space is **`{server} × {graph_id}`**. | +| E3 | The same graph is **multi-homed** under different per-locus names. | **Name ≠ identity.** Resolution is mandatory. | +| E4 | **Branch / commit / snapshot** are first-class addressable sub-state. | An address is *graph @ branch/snapshot*, not just graph. | +| E5 | Enforcement is **two-layered**: engine-layer Cedar (`_as` writers, works embedded) + HTTP-boundary bearer+Cedar (server only). | *How* you reach a graph determines *which* enforcement applies. | +| E6 | **Stored queries / MCP tools are a per-graph registry defined in the project config** (MR-969). | The **agent tool surface is version-controlled in the repo**. | + +Competitors collapse dimensions OmniGraph keeps live: **Helix** fuses E2+E3 (one cluster = one graph); **namidb** fuses E1+E3 into the URI (`s3://b?ns=prod`) and serves one namespace per process. OmniGraph has all of E1–E6 at once, so its config resolves a richer space — but the richness is *earned* by capability. + +## Design + +### 1. The address space and the `target` abstraction + +Every OmniGraph address is a tuple: + +``` +(locus, graph, sub-state, credential) + locus = embedded(URI) XOR remote(server-endpoint) # E1, E2 + graph = a URI (embedded) | a graph_id on a server (remote) # E3 + sub-state = branch | snapshot # E4 + credential = cloud-storage creds (embedded) | bearer token (remote) # E5 +``` + +The config's only job is **name → this tuple**. Define one noun — a **target** — that resolves to either shape: + +```yaml +targets: + dev: # embedded — substrate-direct (E1) + storage: s3://team-bucket/dev.omni + branch: main # sub-state (E4) + staging: # remote — resolves a server by reference (E2/E3) + server: staging # → looked up in `servers` + graph_id: prod # the graph's id on that server (defaults to the entry key) + branch: review +``` + +`--target staging` resolves: project `targets.staging` → `{server: staging, graph_id: prod, branch: review}` → `servers.staging` → `{endpoint, token-by-ref}` → final `(remote(https://…), prod, review, $TOKEN)`. Embedded targets skip the server hop and use cloud-storage credentials. + +**Two concepts, not kubeconfig's three.** kube splits cluster / user / context; that 3-way split is its most-cursed UX. A target *bundles* server+graph+branch+defaults under one name; the **only** thing split out is `servers`, because endpoints+credentials are shared across many targets and are secret-bearing (different ownership and rate-of-change; see §2). Result: **2 nouns — `servers` and `targets`.** Embedded `targets` (`storage:`) subsume today's `graphs:` entries. + +### 1.1 The resolved address is a typed *locator*, not a `uri` string + +The shipped config models a graph as a single `uri: String`, and code branches on `is_remote_uri(uri)`. That conflates two structurally different addresses: an **embedded** graph is a *complete, self-contained* address — one storage URI = one graph, opened directly via the embedded engine; a **remote** graph is a *server endpoint + a `graph_id`* — one server hosts N graphs. A bare server URL **is not a graph**; it lacks the `graph_id`. The cost of the string model, in the code today: + +- the CLI re-decides "server or file?" via `is_remote_uri` at ~16 call sites; +- `TargetConfig` (one `uri` field) **cannot express** multi-server × multi-graph or a multi-homed graph (E2/E3) — "graph `production` on server `prod-eu`" has no representation; +- the CLI **bails on remote URIs** for most operations, precisely because the string can't carry the `graph_id`; +- the `omnigraph-ts` SDK had to model `baseUrl` **+** `graphId` *separately* (rewriting `/graphs/{graphId}/…`) — it invented the structure the string lacks. + +So the *resolved* address is a **typed locator**, not a string: + +```rust +enum GraphLocator { + Embedded { storage: StorageUri }, // file:// , s3:// — a complete graph + Remote { server: ServerId, graph_id: GraphId }, // which server + which graph (+ bearer creds) +} +``` + +A `graphs:` entry resolves into this **once**; downstream code dispatches on the variant (the breadboard's `GraphConn = Embedded(engine) | Remote(http)`) instead of re-sniffing a scheme at each call site. The `uri` string becomes an *input format* for the embedded variant, never the address itself. + +**YAML naming follows the locator — the *key* names the locus**, so neither the value's scheme nor a comment is load-bearing: + +| Locus | Key | Value | +|---|---|---| +| Embedded | **`storage:`** (shipped `uri:` is a deprecated alias) | a storage URI (`s3://…`, `file://…`) | +| Remote | **`server:`** | a name in `servers:` (its `endpoint` + creds resolve by name, §5) | +| Remote graph id | **`graph_id:`** | the id on that server — **defaults to the entry key**; set only when the local alias differs | + +An entry has `storage:` **xor** `server:` — the deserializer rejects *both* and *neither* (no silent ambiguity). This removes two prior confusions: `graphs:` (the map) vs `graph:` (the remote id), and `uri:`-might-be-a-server. + +```yaml +servers: + prod-eu: { endpoint: https://og-eu.internal:8080 } +graphs: + dev: { storage: s3://team-bucket/dev.omni } # embedded + production: { server: prod-eu } # remote — graph_id = "production" (the key) + staging: { server: prod-eu, graph_id: prod } # remote — alias ≠ server's id +``` + +### 1.2 Invalid configs are rejected by design + +The DX rule is: **a config field is either honored or rejected, never silently ignored**. The loader therefore has two phases: + +1. Parse YAML into a loose/raw shape that preserves origin (`base_dir`, layer, line/path when available). +2. Convert once into a typed, role-aware resolved config. Every command receives the resolved form, not the raw YAML structs. + +The typed graph shape is: + +```rust +enum GraphEntry { + Embedded(EmbeddedGraphEntry), + Remote(RemoteGraphEntry), +} + +struct EmbeddedGraphEntry { + storage: StorageUri, + branch: Option, + policy: Option, + queries: QueryRegistrySpec, +} + +struct RemoteGraphEntry { + server: ServerId, + graph_id: GraphId, + branch: Option, +} +``` + +That makes these rules structural rather than advisory: + +- A graph entry must specify **exactly one** locator: `storage:`/legacy `uri:` xor `server:`. +- `policy:` and `queries:` are valid only on `Embedded` graph entries, because they define the capability surface of a graph this process opens directly. A `Remote` graph entry points at a server; that server owns policy and stored-query definitions. +- `omnigraph-server` may serve only `Embedded` graph entries. A server manifest entry with `server:` is rejected: a server should not "host" a graph by proxying another server. +- A named graph uses its own graph entry. Top-level `policy:` / `queries:` are a legacy anonymous-bare-URI compatibility path only; if a named graph is selected while top-level blocks would be ignored, config validation errors with a migration hint. +- A client-defined remote graph discovers stored queries from the server (`GET /queries`) and invokes them (`POST /queries/{name}`); it does not define `queries:` locally for that remote graph. + +Examples that must fail fast: + +```yaml +graphs: + prod: + storage: s3://team-bucket/prod.omni + server: prod-us # invalid: storage xor server +``` + +```yaml +graphs: + prod: + server: prod-us + graph_id: production + policy: { file: ./policies/prod.yaml } # invalid: remote graph policy lives on the server + queries: + find_user: { file: ./queries/find_user.gq } # invalid: remote graph queries are discovered +``` + +`omnigraph config view --resolved --show-origin` is the user-facing debugger for this boundary: it shows the final `Embedded` or `Remote` graph and where every honored field came from. Fields that cannot be honored never make it into the resolved view; they fail validation first. + +### 2. Layered config — global-first, uniform schema, project-optional + +**Posture: global-first, project-optional.** OmniGraph's CLI is primarily a *client* (it operates against graphs and servers, embedded or remote), so it sits on the **global-first** side of the CLI-config axis — like `kubectl` / `aws` / `gh` / `docker`, and unlike *project-first* tools (`git` / `cargo` / `terraform`) whose primary config is per-repo. The **global user config is the primary, self-sufficient default**; the project file is an *optional* repo-scoped override (and, when present, the deployment manifest). `omnigraph query --target prod` must work from **any directory with no project file**, exactly as `kubectl get pods --context prod` works from anywhere. *(This is a deliberate flip from today, where the CLI reads `./omnigraph.yaml` and does not even walk parent dirs — i.e. today it is project-anchored.)* + +**Rule: the two layers share ONE raw schema, and each is fully self-sufficient** (the git-layering mechanism — same schema at both levels; you never need a repo to have a working config). Do **not** specialize the file format by layer. Instead, run the same role-aware validation everywhere (§1.2): the global and project layers may both define graph locators, defaults, servers, and aliases, but fields that are meaningless for a resolved graph variant are rejected rather than ignored. For example, `queries:` is valid for an embedded graph this config opens directly; it is invalid on a remote graph entry because remote stored queries are server-owned and discovered. + +This makes the **zero-project case the default, not an edge case**: a solo user (or an agent) defines everything needed for client work in `~/.omnigraph/config.yaml` — servers, embedded + remote graph locators, defaults, aliases, and optionally personal embedded-graph query registries — and **never creates a project file**. A team adds `./omnigraph.yaml` only when it wants repo-scoped overrides or a committed, GitOps'd deployment manifest. Global-first does **not** forbid project files; it stops *requiring* them (the kubectl model: `~/.kube/config` is sufficient and default; per-project kubeconfigs are opt-in via `KUBECONFIG`). + +| Layer | Required? | Typical use | Path | +|---|---|---|---| +| Global | no | **the default** — solo/agent's entire config; shared servers+creds for teams; even a personal server's graphs/queries | `~/.omnigraph/config.yaml` | +| Project | no | **opt-in** — repo-scoped overrides + the committed deployment manifest (graphs, queries, policy) | `./omnigraph.yaml` | + +**Precedence (low → high):** built-in defaults < global < project < env vars < CLI flags. With no project file it collapses to **built-in < global < env < flags** — the common global-only path. + +**Merge semantics — "closest layer wins, at the smallest meaningful unit"** (the field consensus: git / kubeconfig / cargo / Helm / VS Code): +- **Settings objects** (`defaults`, `auth`, `server`) → **deep-merge per field**: a project sets `defaults.graph` and *inherits* the global `defaults.output_format`. (VS Code / cargo behavior.) +- **Named-resource maps** (`servers`, `graphs` / compat `targets`, `queries`, `aliases`) → **union by key; on a collision the higher layer's entry REPLACES the lower wholesale** — *no field-level deep-merge within an entry*. (kubeconfig: union contexts by name.) The footgun this avoids: global `servers.prod = {endpoint, policy}`, project `servers.prod = {endpoint: other}` — deep-merge would silently retain the old fields; replace makes the project's `prod` self-contained and predictable. +- **Lists/arrays** → **replace, never append** (Helm convention; appending is order-sensitive and surprising). +- **Scalars** → higher layer wins. +- **Relative paths carry their origin's base_dir.** A `queries:` entry's `.gq` path, or a `policy.file`, resolves against the directory of the layer it was *defined in* — global entries under `~/.omnigraph/`, project entries under the project dir. +- **Inspectable (non-negotiable):** `omnigraph config view --resolved --show-origin` prints each final value *and which layer set it* (the `git config --show-origin` / `kubectl config view` rule). A layered config without origin-tracing is a debugging trap. + +### 3. Roles, and the file-naming decision (same name for project = server) + +`omnigraph.yaml` carries two *roles* that diverge in prod and collapse on a laptop: + +- **Server role** (read by `omnigraph-server`): `graphs:` entries that are **embedded storage locators**, per-graph `policy.file`, **`queries:` — the stored-query/MCP registry lives here**, plus serving knobs. Remote graph locators are rejected in this role. +- **Client role** (read by the CLI/agent): `servers:`, embedded or remote `graphs:` locators, `defaults:`, `aliases:`. A remote graph locator points at server-owned capabilities; it cannot define local `policy:` or `queries:`. + +**Project config and server config are the same artifact, hence the same name.** The server *serves the project*: the file that says "these graphs exist, with these stored queries and this policy" is simultaneously the project manifest and the server's deploy config. Role is distinguished by which *sections* are populated, never by filename. Readers ignore sections that are not theirs (today's file already does this with `cli:` vs `server:`). + +**Why not kube's role-split.** Two coherent models exist: (A) one project file with role-sections (Helix `helix.toml` holds both `[local.dev]` and `[enterprise.production]`; compose; Cargo), and (B) deployment-manifest strictly separate from client config (kubectl — you never put a context in `deployment.yaml`). kube is the sharpest topological analog (multi-server × multi-graph, one client targeting many), so B has a real claim. The tiebreaker is **E1: OmniGraph is embedded-first.** In embedded mode the manifest's `graphs:` *is* the local target list — manifest and local-client-view are the same object, so splitting them (B) fights the grain and forces two files for local work. kube splits because it has **no** embedded mode (client always remote+global). So: take the half kube is right about — *remote* client targeting (`servers:`, endpoints, creds) is a separate concern in a separate **user-global** file (`config.yaml`, like `~/.kube/config`); reject the half it is wrong about for us — do **not** split the *project* layer by role. **The second name (`config.yaml`) is justified by scope (user-global), not role.** *(If OmniGraph ever dropped embedded mode and went pure-remote, model B's strict split would become cleanest.)* + +### 4. File naming + +Principles from the field: **one global dir** `~/.omnigraph/` (like `~/.aws`/`~/.kube`/`~/.helix`), with config/cache/state as **subdirectories** (separation without XDG's three-root scatter); **secrets keyed by server name in the OS keychain or a separate git-ignored profile file** (AWS/gh model, not a new `credentials.yaml`); **project-root manifest keeps the app-named file** (`Cargo.toml`, `package.json`); **`.yaml`, not `.yml`**; keep OmniGraph's established names. The genuinely *new* decisions are the **global** dir's existence and keyed-by-name resolution with an explicit `auth.token` override (MR-971); the shipped `bearer_token_env` + `auth.env_file` mechanism remains as legacy compat. + +| Artifact | Path / name | Why | +|---|---|---| +| Project = server config (one artifact) | `./omnigraph.yaml` | **Keep.** Root manifest like `Cargo.toml` / `compose.yaml` / `helix.toml`. Same name for both roles because it is one file. In prod the server's deploy repo and an app repo each have their own `omnigraph.yaml` — same name, different repos. | +| Global user config | `~/.omnigraph/config.yaml` | **One dir** (`~/.omnigraph/`, like `~/.aws`/`~/.kube`/`~/.helix`). Named `config.yaml` *not* `omnigraph.yaml` — the name signals scope (and `~/.aws/config`, `~/.kube/config`, `~/.helix/config` all do this). Holds the full schema so a solo user needs nothing else. | +| Credentials | OS keychain (`omnigraph:`, preferred) → `~/.omnigraph/credentials` profile file (`[]`, `0600`, git-ignored). **Keyed by server name**, inside the one dir. | **Key by name, AWS/gh model** — `~/.aws/credentials [profile]`, `~/.kube/config users:`, `~/.helix/credentials`. *Not* a `credentials.yaml`, and *not* a per-server hand-named env var; the secret lives under the server name (no indirection). Legacy `bearer_token_env` + `.env.omni` dotenv remain as a compat path. See §5. | +| Cache / state | `~/.omnigraph/cache/`, `~/.omnigraph/state/` | Subdirs of the one dir (like `~/.aws/sso/cache/`, `~/.kube/cache/`) — cache is `rm -rf`-safe and backup-excludable without scattering across XDG roots. | +| Cedar policy | `./policies/.yaml` + `.tests.yaml` | **Keep.** Referenced by `policy.file`. | +| Schema | `./*.pg` (e.g. `schema.pg`) | **Keep.** | +| Stored queries | `./queries/*.gq` | **Keep.** `.gq` sources referenced by the `queries:` registry. | + +**Global dir: `~/.omnigraph/` — one place, with subdirectories.** Everything OmniGraph keeps for a user lives under a single `~/.omnigraph/` directory, matching the peer group (`~/.aws`, `~/.kube`, `~/.docker`) and the direct competitor (`~/.helix`). This is what DB/cloud-CLI users expect and the lowest-cognitive-load shape. + +*Separation and "one place" are not in conflict* — the decisive realization. The peer tools get config/cache/state separation via **subdirectories inside the one dir**, not via XDG's three scattered roots: `~/.aws/sso/cache/`, `~/.kube/cache/`. So OmniGraph keeps `~/.omnigraph/config.yaml`, `~/.omnigraph/credentials`, `~/.omnigraph/cache/` (catalogs — `rm -rf`-safe, backup-excludable), `~/.omnigraph/state/` (session, logs) — getting cache hygiene **and** a single discoverable location, without the XDG scatter. An earlier draft argued XDG on a false dichotomy (it assumed single-dir ⇒ mixed); subdirs dissolve it. `~/.omnigraph/` is canonical and documented; `$XDG_CONFIG_HOME` may optionally be honored if a user has set it, but XDG is not part of the mental model. + +**Env / override precedence (the `KUBECONFIG` analog):** +- `OMNIGRAPH_CONFIG=/path` — explicit config file, highest precedence. +- `OMNIGRAPH_HOME=/path` → the global dir (default `~/.omnigraph/`); `$XDG_CONFIG_HOME` optionally honored if a user has set it, but `~/.omnigraph/` is canonical. +- Cache and state are subdirs of the one dir: `~/.omnigraph/cache/` (cached remote catalogs), `~/.omnigraph/state/` (session, logs). +- Per-server token resolution: an explicit `auth: { token: {...} }` source (env/file/command/keychain) wins if set; otherwise **keyed by the server name** — `OMNIGRAPH_TOKEN_` (or `OMNIGRAPH_TOKEN` for the active server) → OS keychain `omnigraph:` → the `[]` profile in `~/.omnigraph/credentials`; legacy `bearer_token_env` still honored. See §5. + +### 5. Credentials, connection tiers, and bind portability (12-factor) + +**Credentials are by-reference everywhere, never inlined — and keyed by the *server name*, not by a hand-invented env-var name.** This is the one place the design departs from simply reusing the shipped `bearer_token_env` mechanism, because that mechanism is sub-optimal for a multi-server client: it forces the operator to invent and coordinate an env-var name per server (three steps to add a server: pick a var, name it in config, set it in the store). The peer group (AWS profiles, `gh` hosts, kubeconfig users, docker auths) instead keys the secret **by the server's name** — no indirection. OmniGraph should match that. + +**Resolution for server `` (no config field required):** +1. **`OMNIGRAPH_TOKEN_`** env var (name-derived, upper-snake), else **`OMNIGRAPH_TOKEN`** for the active server — the CI/headless override (12-factor). +2. **OS keychain** entry `omnigraph:` — the preferred interactive store (no plaintext on disk); written by `omnigraph login `. +3. **`~/.omnigraph/credentials`** — an AWS-style profile file keyed by server name (mode `0600`, git-ignored), the fallback when no keychain: + ```ini + [prod-us] + token = … + [prod-eu] + token = … + ``` +So a `servers.` with no token field resolves by name — adding a server is one step (`omnigraph login `), and "multiple servers, multiple tokens" falls out for free. + +**But implicit must not be the *only* path — explicit sourcing is a first-class option** (the DX/AX lesson). Pure-convention is invisible (you must *know* `OMNIGRAPH_TOKEN_`), can't integrate with a secrets-manager's fixed var name, and can't do dynamic/short-lived tokens. So a server may declare an explicit `auth:` block — a **method-agnostic wrapper** (today only `token:` for bearer; `mtls:`/`oidc:` are the future siblings, so the credential model never has to be re-keyed) holding a tagged token *source*. Secrets are *still* never inlined (every source is a reference): + +```yaml +servers: + prod-us: + endpoint: https://og-us… + auth: { token: { env: OG_PROD_US_TOKEN } } # explicit env var — self-documenting (= legacy bearer_token_env) + prod-eu: + endpoint: https://og-eu… + auth: { token: { command: [vault, read, -field=token, secret/og] } } # dynamic / short-lived + edge: + endpoint: https://og-edge… + auth: { token: { file: /run/secrets/og-token } } # k8s/docker mounted secret + staging: + endpoint: https://og-staging… # no auth: → implicit chain (below) +``` + +| `auth.token:` source | when | DX/AX value | +|---|---|---| +| *(auth omitted)* | the common case | zero-config; `omnigraph login` populates keychain `omnigraph:` | +| `{ env: VAR }` | secrets-manager / CI injects a fixed var | **self-documenting** — config states the source; = the legacy `bearer_token_env` | +| `{ file: PATH }` | k8s/docker secret mounted as a file | no env plumbing | +| `{ command: [...] }` | Vault, cloud IAM, `gh auth token` | **dynamic tokens** — first-class exec, the capability pure-env/keychain can't give (kube `exec` / AWS `credential_process`) | +| `{ keychain: ENTRY }` | pin a non-default keychain entry | explicit override of the name-derived default | + +**Resolution per server:** if `auth.token:` is set, use that source (no fallthrough). Else the **implicit chain**: `OMNIGRAPH_TOKEN_` (or `OMNIGRAPH_TOKEN` for the active server) → keychain `omnigraph:` → `[]` in `~/.omnigraph/credentials` (`0600`, git-ignored). `omnigraph login ` writes/rotates only that server's secret; per-server precedence is independent; sharing is opt-in (same env var or source). The `command` source runs locally with the operator's own privileges and is defined only in operator-owned config (never server-supplied), so it adds no remote-execution surface. The `auth:` wrapper is method-agnostic so adding mTLS/OIDC later is a new sibling key, not a breaking re-key (Hyrum's Law: the field name is a contract once shipped). There is **no `credentials.yaml`** and **no inlined secret**. *Convention for the floor, explicit for control — and explicit is legible to agents and never inlines a secret.* + +**Back-compat.** The shipped per-graph `bearer_token_env` + `auth.env_file` dotenv (`resolve_remote_bearer_token`, real-env-wins) keeps working unchanged for existing single-server setups; `bearer_token_env` is just the legacy flat alias for `auth: { token: { env } }`. Resolution tries an explicit `auth.token:` (or legacy `bearer_token_env`) first, then the keyed-by-name chain — so nothing breaks, but the zero-config default is the no-boilerplate keyed-by-name path. (MR-971 — the `bearer_token_env` parity gap — is where this resolver work lands.) + +**Three connection tiers** (Supabase/Prisma teach the zero-config floor): +1. **Env vars** — `OMNIGRAPH_SERVER=https://…` + `OMNIGRAPH_TOKEN=…`: zero-config remote, no file (the `DATABASE_URL` floor). +2. **Global `config.yaml`** — named `servers:` + `graphs:` for multi-server setups (the AWS-profiles convenience). +3. **Project `omnigraph.yaml`** — project-pinned targets/graphs, committed. + +**Keep `omnigraph.yaml` a *portable* manifest (12-factor).** Deploy-specific runtime that varies per environment — the **bind host/port**, worker counts — should be supplied by **`--bind` / `OMNIGRAPH_BIND` (flags/env)**, *not* a committed `server.bind:` baked into the manifest. A manifest that hardcodes `0.0.0.0:8080` is not portable across deploys and leaks an environment detail into a version-controlled file. The same-named `omnigraph.yaml` stays portable across deploys precisely because the volatile, per-environment knobs live in env/flags (12-factor config), while the stable, portable definition (graphs, queries, policy) lives in the file. This is the one concrete lesson taken from kube's model-B without adopting its file split: portability via env/flags, not via a second file. + +### 6. Where stored queries live: defined locally, invoked remotely + +A stored query splits across two axes; do not conflate them: +- **Definition** (`.gq` source + `queries:` entry) lives next to the **embedded graph entry that owns it**. For a hosted remote graph, that is the **deployment manifest** read by `omnigraph-server`; for a personal embedded graph, it may be the user's own config. It never lives on a client-side `Remote` graph entry. +- **Discovery** ("what tools exist for me?") is fetched from the **server** (Cedar-filtered `GET /queries` / MCP catalog) at connect time. +- **Invocation** is **remote** (client → server, HTTP/MCP) — or **embedded** (the CLI opens the graph directly and reads the same manifest). + +For remote use, the client carries *pointers to servers*, not query definitions; it **discovers and invokes**, never defines. This is the **capability-as-code guarantee for agents**: an agent can only invoke tools the server's *committed, reviewed* config exposes — it **cannot define a new tool at runtime**. Definition is structurally outside the agent's reach. + +`queries:` (graph-capability registry, Cedar-gated when served remotely, MCP-visible when exposed) and `aliases:` (client CLI shortcut) overlap — both can name `.gq`-backed operations. This RFC keeps them siblings (the MR-969 decision); the clean long-term is **one registry, two invocation surfaces** (embedded + remote), with `aliases:` subsumed. Out of scope here. + +#### Reconciling `aliases:` with the role model + +`aliases:` is the pre-MR-969, **client-role, embedded-only, ungated** ancestor of `queries:`. An alias bundles `command` (read/change), `query` (`.gq` path), `name` (symbol), `args` (positional param names), and `graph`/`branch`/`format` defaults; the CLI runs it embedded. The server never reads it. So: + +- **Role:** `aliases:` is **client-role** (CLI behavior) → it may live in **both** the user-global `config.yaml` and the project manifest, layered. `queries:` is **graph-capability role** → it lives only on an `Embedded` graph entry, and for remote server graphs that means the server deployment manifest. *Who opens the graph determines where query definitions can live.* +- **Difference:** `aliases:` = embedded invocation, no gating, explicit `command`, bundles client defaults + positional args. `queries:` = remote (+future embedded), Cedar + `mcp.expose`, **infers** read/mutate, bundles only MCP settings. +- **Convergence:** decompose an alias — *definition* (name→.gq+symbol) → `queries:` (the superset: typed, validated, gated, multi-surface, no redundant `command`); *target/branch/format* → client invocation context (`--target`/`--branch`/`--format` or `defaults:`), not baked per-query; *positional `args`* → thin CLI sugar or dropped (agents/services use named JSON params). End-state: one `queries:` registry + the client config model subsumes `aliases:`. +- **Validation:** a file-backed alias (`query: ./foo.gq`) may target only an embedded graph. A remote graph shortcut must be explicit that it invokes a server-owned stored query, e.g. `invoke: find_user`, so the client cannot smuggle a new `.gq` definition into a remote capability surface. +- **v1:** keep `aliases:` unchanged. Footgun worth a load-time warn: an alias and a query with the same name in one manifest are different namespaces invoked differently (`--alias X` vs `POST /queries/X`). + +```yaml +aliases: + local_owner: + command: query + query: ./queries/owner.gq + name: owner + graph: dev # valid only if `dev` resolves Embedded + + remote_owner: + invoke: find_user + graph: prod # valid only if `prod` resolves Remote; source lives on the server + args: [name] +``` + +### 7. CLI surface + +- `omnigraph login ` — interactive auth; stores the token keyed by server name in the OS keychain (`omnigraph:`) or the `[]` profile of `~/.omnigraph/credentials` (0600). The `gh auth login` analog. +- `omnigraph use ` — set the active graph (writes the appropriate layer). The `kubectl config use-context` analog. +- `omnigraph config view [--resolved] [--show-origin] []` — print the merged config and, with `--resolved`, the final tuple **plus the origin layer of every field** (the `git config --show-origin` / `kubectl config view` analog). Resolution is never a mystery. +- All existing verbs (`query`, `mutate`, `load`, `schema`, `branch`, …) gain `--graph `; resolution decides embedded vs remote transparently. + +### 7.5 Init, login, and bootstrap — three tiers (folds in the Q2 design) + +Scaffolding splits into three tiers by *scope* and *fatness*, mirroring the field (supabase `init` vs `login`; HelixDB thin `init` vs fat `chef`). Most of this lives in sibling tickets; this RFC owns only the **user route**. + +| Tier | Command | Scope | What it does | Model | Status | +|---|---|---|---|---|---| +| **User route** | `omnigraph login []` | user (`~/.omnigraph/`) | auth + write `~/.omnigraph/config.yaml` / `credentials`; first-run global setup | gh / supabase `login` | **this RFC** (unbuilt) | +| **Thin project init** | `omnigraph init` | project, in-place | create graph + `scaffold_config_if_missing` (`omnigraph.yaml` + minimal `.pg`/`.gq`); refuse-if-exists or `--force` | `cargo init`, `prisma init` | exists; `--force` purge = MR-975 | +| **Fat bootstrap** | `omnigraph quickstart [--template ] [--auto]` | project, possibly new-dir | scaffold + seed data + `serve start` + agent prompt file | HelixDB `chef`, `create-next-app` | MR-973 (unbuilt) | + +**Design positions** (first-principles, since none of the fat tier is built): +- **Split `init` (project) from `login` (user)** — never one command writing to both `$HOME` and the project (the supabase line, not the dbt line). `init`=project scaffold; `login`=user credential + global config. +- **`init` is in-place + refuse-if-exists** (cargo/prisma/terraform default): don't clobber; adopt existing files; require `--force` to overwrite (and `--force` purges Lance state per MR-975). +- **Interactive for humans, `--auto`/agent-mode for automation** (npm `-y`, create-* `--CI`, MR-981 `--machine`). In `OMNIGRAPH_AGENT_MODE` any prompt → fail with a repair hint. +- **Templates are a `--template ` flag on the fat tier** (create-vite model), with the *content* (schema + queries + seed) coming from a template source. Mechanism is a design question (bundled-in vs `og template pull` from a repo vs `npm create-*`-style delegation) — **not** an existing foothold (MR-581 stale). Lean: a small set of bundled templates first (generic `Person→Knows`, plus promote `omnigraph-intel-bootstrap`), `--template ` later. +- **`init`/`quickstart` can scaffold the `graphs:` map with one or more entries**; "init with specific graphs" = the scaffolded `graphs:` block (embedded `storage:` locally; the agent/operator adds remote `server:` entries via `login` + editing). +- **Secrets-on-scaffold rule** (prisma/dbt/supabase all do this): anything that writes a token also keeps it out of VCS. `login` prefers the OS keychain (no file); the `~/.omnigraph/credentials` profile fallback is `0600` and git-ignored, and any project-local `.env`-shaped file gets a `.gitignore` entry. + +### 8. Concrete shape + +**Global** `~/.omnigraph/config.yaml` (per-user, secret-free): +```yaml +servers: # endpoint only — token is keyed by the server name + prod-us: { endpoint: https://og-us.internal:8080 } + prod-eu: { endpoint: https://og-eu.internal:8080 } + staging: { endpoint: https://og-staging.internal:8080 } +graphs: + personal: { storage: ~/graphs/personal.omni } +defaults: + graph: personal +aliases: + my_people: + command: query + query: ~/queries/people.gq + name: list_people + graph: personal +``` + +**Project client** `./omnigraph.yaml` (committed, secret-free, portable — no `server.bind`). Note the shipped noun is `graphs:` (MR-603); an entry is embedded (`storage:`) XOR remote (`server:` + `graph_id:`, §1.1): +```yaml +graphs: + dev: { storage: s3://team-bucket/dev.omni, branch: main } # embedded + staging: { server: staging, graph_id: prod, branch: review } # remote → graph `prod` on server `staging` + prod-us: { server: prod-us, graph_id: production } + prod-eu: { server: prod-eu, graph_id: production } # multi-homed: same graph, another server +defaults: { graph: dev, output_format: table } +aliases: + owner: + command: query + query: ./queries/owner.gq + name: owner + args: [name] + graph: dev +``` +Select with `--graph ` (shipped flag, MR-603). + +**Server deployment** `./omnigraph.yaml` (committed in the deploy repo, read by `omnigraph-server`). Every served graph is an embedded storage locator; server-owned policy and stored-query definitions live here: +```yaml +graphs: + production: + storage: s3://team-bucket/prod.omni + policy: + file: ./policies/prod.yaml + queries: + find_user: + file: ./queries/find_user.gq + mcp: { expose: true, tool_name: lookup_user } + +server: + policy: + file: ./policies/server.yaml +``` + +**Credentials** are keyed by server name — `omnigraph login prod-us` writes the OS keychain entry `omnigraph:prod-us` (or a `[prod-us]` profile in `~/.omnigraph/credentials`, 0600, git-ignored); `OMNIGRAPH_TOKEN_PROD_US` overrides for CI. No token fields in any config file; no committable secrets. + +## DX + +1. **One command surface, two loci.** `query --graph dev` (embedded) and `--graph staging` (remote) are the same command; only resolution differs. Change one word, not a mental model. +2. **Clone-and-go.** Project config names servers+graphs; teammate runs `omnigraph login staging` once and every target resolves. The git + `gh auth login` model. +3. **Multi-server × multi-graph is the default.** Remote graph entries reference `server` by name; `servers` is a global named map; graphs are per-server. `prod-us` and `prod-eu` both serving `production` is two graph entries — Helix cannot express this. +4. **Solo-first.** Everything in `~`, no project required. +5. **Laptop-to-fleet on one schema.** Local = one `omnigraph.yaml` (both roles); prod = role-split across repos. No second format to learn. + +## AX (agent experience) + +1. **One flat resolved context, never a config to navigate.** target→server→endpoint→token resolves *before* the agent sees anything. The agent reasons about tools, not topology (the LLM-safe-surface principle extended to config). +2. **Secrets are structurally outside the agent's reach.** The repo it operates in has no tokens; they are in the global layer / keychain, outside its view. An agent *cannot* exfiltrate a prod token from project config because it is not there. +3. **Branch/snapshot-pinned contexts** (E4) — hand an agent a `branch: review` / `--snapshot v42` target and its reads are reproducible and cannot see uncommitted main-line state. No kubeconfig analog. +4. **The agent's capabilities are a GitOps'd artifact** (E6) — which graphs exist, which stored-query tools it may call, and which Cedar rules gate them are all in the version-controlled server config. Powers change only via a reviewed PR, deployed by restart. Infrastructure-as-code for what the AI can do. +5. **Config + policy compose.** Config = "where am I pointed + which token"; Cedar = "what may I do there." Orthogonal; no enforcement logic leaks into config. + +## GitOps — three surfaces, secrets in none + +| Surface | Repo | Contents | Deploy | Secrets | +|---|---|---|---|---| +| Server deployment config | infra/deploy repo | `graphs:`, policy, **`queries:` + `.gq` files** | commit → CI → **server restart** (no hot reload) | none — by-reference | +| Project client config | app repo | `graphs:` → embedded storage or remote server+graph | committed, read by CLI/agent | none | +| Global user config | **not GitOps'd** — machine-local `~` | `servers:` + creds-by-ref | `omnigraph login` writes it | refs only (like `~/.kube/config`) | + +## Comparison + +| Property | kubeconfig | Helix | git | compose | **OmniGraph (this RFC)** | +|---|---|---|---|---|---| +| Named remote endpoints + creds-by-ref | ✅ | ✅ | partial | partial | ✅ (global `servers`) | +| Global + project layering, uniform schema | ✗ | ✗ | ✅ | ✗ | ✅ | +| Embedded OR remote under one name | ✗ | ✗ | n/a | ✗ | ✅ (E1) | +| Multi-server × multi-graph | ✅ | ✗ | n/a | n/a | ✅ (E2) | +| Branch/snapshot in the address | ✗ | ✗ | partial | ✗ | ✅ (E4) | +| Agent tool surface in the repo | ✗ | ✗ (separate bundle) | n/a | n/a | ✅ (E6) | +| Project manifest renamed by role | — | no | — | no | **no** | +| Concept count | 3 | 1 | 2 | 1 | **2 (servers/targets)** | + +## Migration / backwards compatibility + +- **Additive.** Today's `omnigraph.yaml` (`graphs:`, `cli:`, `server:`, `aliases:`, `policy:`) keeps working unchanged. `graphs:` entries are equivalent to embedded `targets:` with a `storage:` (shipped `uri:` is a deprecated alias); both resolve. +- **`targets:` is new** and optional. `servers:` is new and optional. Absent → today's behavior. +- **Global `~/.omnigraph/config.yaml` is new.** Absent → only project + env + flags, exactly as now. Its addition is the **global-first posture flip**: today the CLI is project-anchored (reads `./omnigraph.yaml`, no parent walk); the global config becomes the new primary discovery path so the CLI works with no project file. Existing project-only workflows are unchanged (project still overrides global); the flip is additive — it adds a fallback layer below the project file, it does not remove the project file. +- **`graphs:` → `targets:` is an evolution, not a break.** Both can coexist; `targets:` is the superset (adds remote + branch pinning). A future cleanup may alias `graphs:` to embedded `targets:`. +- **`server.bind` stays supported** but documentation steers operators to `--bind` / `OMNIGRAPH_BIND` for portability; no removal. +- **Credentials: keyed-by-name is new; `bearer_token_env` is the compat path.** The primary design (keychain / `[]` profile / `OMNIGRAPH_TOKEN_`) is new resolver work (lands on MR-971). The shipped `bearer_token_env` + `auth.env_file` dotenv (`resolve_remote_bearer_token`) is **unchanged and still honored** — existing single-server dotenv setups keep working, and the resolver honors an explicit `auth: { token: {...} }` source (env/file/command/keychain) with `bearer_token_env` as its flat legacy alias. No `credentials.yaml`. +- **Validation tightens invalid mixes, not valid legacy use.** Top-level `policy:` / `queries:` remain only for anonymous bare-URI compatibility. Named graphs use per-entry fields. Remote graph entries with local `policy:` / `queries:` and server manifests with `server:` graph locators are rejected because there is no correct way to honor those fields. + +## Open questions + +- **`graphs:` vs `targets:` naming churn.** Do we rename `graphs:` → `targets:` (with a deprecation alias) or keep `graphs:` for embedded and add `targets:` for remote? Leaning: keep both, document `targets:` as the superset. +- **Keychain integration scope.** Keychain is now the *primary* credential store (§5), so this is on the critical path, not optional: macOS Keychain first (matches operator practice) with the `0600` `[]` profile file as fallback; Linux Secret Service / `pass` later. Open: which keyring crate, and the exact `OMNIGRAPH_TOKEN_` name-derivation (upper-snake, non-alnum → `_`). +- **Project-local `servers:`.** Allowed (e.g. a localhost dev server), merged with global. Confirm creds stay by-reference even for project-local servers (yes). +- **`aliases:` ⇄ `queries:` convergence.** Out of scope here; tracked separately. One registry with embedded + remote invocation surfaces is the target end state. +- **Single-file `KUBECONFIG`-style list.** Do we support `OMNIGRAPH_CONFIG` pointing at multiple files (colon-joined), or a single file only? Start single; revisit if demand appears. + +## Implementation — breadboard + slices (Shape A) + +Shaped via requirements + a fit check (Shape A — global-first layered config + unified `graphs:` entry + three-tier init — selected over a project-first minimal option and a Helix-clone). This section breadboards A and slices it. **Bold** = NEW. + +### Places + +| # | Place | What | +|---|---|---| +| P1 | Disk | `~/.omnigraph/{config.yaml, credentials, cache/, state/}` + project `omnigraph.yaml` + `.env.omni` | +| P2 | Config resolution | runs on every command: load layers → merge → resolve `--graph` | +| P3 | Command execution | embedded engine OR remote HTTP client | +| P4 | Remote `omnigraph-server` | existing HTTP surface (`/query`, `/mutate`, `/queries/{name}`) | +| P5 | Scaffold | `login` / `init` / `quickstart` | + +### Affordances + +| # | Place | Affordance | NEW? | Wires | +|---|---|---|---|---| +| U1 | P1 | `~/.omnigraph/config.yaml` (operator edits) | **N** | → N1 | +| U2 | P1 | project `./omnigraph.yaml` | — | → N1 | +| U3 | P1 | `~/.omnigraph/credentials` / `.env.omni` dotenv (secrets, git-ignored) | — | → N4 | +| U4 | P3 | `omnigraph --graph ` (any command) | — | → N14 | +| U5 | P5 | `omnigraph login []` | **N** | → N11 | +| U6 | P5 | `omnigraph init` / `quickstart [--template]` | partly | → N12 / N13 | +| U7 | P2 | `omnigraph config view --resolved --show-origin` | **N** | → N10 | +| N1 | P2 | `load_layered_config()` — global (N3) + project (cwd), serde each | **N** | → N2 | +| N2 | P2 | **merge engine** — deep-merge settings; replace named-resource entries; replace lists; **retain provenance** and raw field origins | **N⚠️** | → N5, → S_merged | +| N3 | P2 | global-dir resolver — `OMNIGRAPH_HOME` else `~/.omnigraph/` | **N** | → N1 | +| N4 | P2 | `load_env_file_into_process` — dotenv, real-env-wins (existing) | — | → N9 | +| N5 | P2 | `resolve_graph(name, merged)` → typed `Embedded`/`Remote` locator; rejects invalid role/field combinations before execution | **N⚠️** | → N6 | +| N6 | P3 | `GraphConn` — `Embedded(engine)` \| `Remote(http)` dispatch | **N⚠️** | → N7, → N8 | +| N7 | P3 | embedded path — `Omnigraph::open(uri)` (existing) | — | → engine | +| N8 | P3 | **HTTP-client path** — POST `/query`/`/mutate`/`/queries/{name}` | **N⚠️** | → P4, → N9 | +| N9 | P2 | `resolve_bearer_token(server)` — explicit `auth.token` source if set, else **keyed by name**: `OMNIGRAPH_TOKEN_`/`OMNIGRAPH_TOKEN` → keychain `omnigraph:` → `[]` profile; legacy `bearer_token_env`/dotenv (MR-971) | **N⚠️** | → N8 | +| N10 | P2 | `config view` handler — merged + per-field origin (needs N2 provenance) | **N** | → U7 | +| N11 | P5 | `login` handler — interactive auth → write `config.yaml` + `credentials` (0600) + `.gitignore` | **N⚠️** | → S_global | +| N12 | P5 | `init` handler — `scaffold_config_if_missing` + create graph; refuse-if-exists/`--force` purge (MR-975) | partly | → S_project | +| N13 | P5 | `quickstart` handler — scaffold + `--template` + seed + `serve start` + agent prompt (MR-973; needs serve MR-970) | **N⚠️** | → S_project | +| N14 | P3 | agent-mode wrapper — `--machine`/`OMNIGRAPH_AGENT_MODE`: JSON, structured errors, never-prompt, typed exit codes (MR-981) | **N⚠️** | → N1 | +| S_global | P1 | `~/.omnigraph/config.yaml` + `credentials` | **N** | read by N1/N9 | +| S_project | P1 | `./omnigraph.yaml` + `.env.omni` | — | read by N1/N4 | +| S_merged | P2 | in-memory resolved config (per command, with provenance) | **N** | read by N5/N10 | +| S_cache | P1 | `~/.omnigraph/cache/` (remote catalogs) | **N** | read by N8 | + +```mermaid +flowchart TB + subgraph P1["P1: Disk"] + U1["U1: ~/.omnigraph/config.yaml"] + U2["U2: ./omnigraph.yaml"] + U3["U3: credentials dotenv"] + end + subgraph P2["P2: Config resolution"] + N3["N3: global-dir (OMNIGRAPH_HOME)"] + N1["N1: load_layered_config"] + N2["N2: merge engine (+provenance)"] + N4["N4: dotenv loader"] + N5["N5: resolve_graph(--graph)"] + N9["N9: resolve_bearer_token"] + N10["N10: config view"] + end + subgraph P3["P3: Command execution"] + U4["U4: omnigraph --graph"] + N14["N14: agent-mode wrapper"] + N6["N6: GraphConn embedded|remote"] + N7["N7: embedded Omnigraph::open"] + N8["N8: HTTP-client POST"] + end + subgraph P5["P5: Scaffold"] + U5["U5: login"]; U6["U6: init/quickstart"] + N11["N11: login handler"]; N12["N12: init"]; N13["N13: quickstart"] + end + P4["P4: remote omnigraph-server"] + U1-->N1; U2-->N1; N3-->N1; N1-->N2-->N5-->N6 + U3-->N4-->N9-->N8 + U4-->N14-->N1 + N6-->N7; N6-->N8-->P4 + N2-->N10-->U7["U7: config view --resolved"] + U5-->N11; U6-->N12; U6-->N13 + classDef ui fill:#ffb6c1,stroke:#d87093,color:#000 + classDef n fill:#d3d3d3,stroke:#808080,color:#000 + class U1,U2,U3,U4,U5,U6,U7 ui + class N1,N2,N3,N4,N5,N6,N7,N8,N9,N10,N11,N12,N13,N14 n +``` + +### Slices (vertical, each demo-able) + +| # | Slice | Parts/affordances | Demo | +|---|---|---|---| +| **V1** | **Global layer + merge + `config view`** | A1–A4 · N1,N2,N3,N10 · U1,U7,S_global,S_merged | Put config in `~/.omnigraph/`, run `omnigraph config view --resolved --show-origin` from any dir → merged result with per-field origin; existing embedded commands work global-first with no project file | +| **V2** | **Remote graphs + HTTP client + creds** | A5–A7 · N5,N6,N8,N9 · S_cache | Define a `server:` graph entry; `omnigraph query --graph prod` hits the remote server (`curl`-free); embedded `--graph dev` still local | +| **V3** | **`omnigraph login`** | A8 · N11,U5 | `omnigraph login prod` writes `~/.omnigraph/credentials` (0600) + `.gitignore`; V2 remote query now works with no manual env | +| **V4** | **Thin-init hardening + quickstart + templates** | A9 · N12,N13,U6 (needs serve MR-970) | `omnigraph quickstart --template person-knows` scaffolds + seeds + serves; `init --force` purges (MR-975) | +| **V5** | **Agent-mode** | A10 · N14,U4 (MR-981) | `OMNIGRAPH_AGENT_MODE=1 omnigraph query …` → JSON + structured errors + typed exit codes; never-prompt | + +V1 is the foundation (global-first + merge + view). V2 closes the substantive client→server gap. V3 is credential ergonomics. V4/V5 ride sibling tickets (MR-970/973/981). MR-969 (stored queries) ships independently and is reached by N8's `/queries/{name}` once V2 lands. + +## Rollout + +The slices above are the rollout order: **V1 (global layer + merge) → V2 (remote graphs + HTTP client) → V3 (login) → V4 (quickstart/templates, on MR-970) → V5 (agent-mode, MR-981).** V1–V2 close the substantive gap (global-first config + `curl`-free server access); V3–V5 are ergonomics that ride sibling tickets. Evaluate after V2 against early-adopter and agent-onboarding (MR-973 / MR-974) signal. The spikes (X1 HTTP-client, X2 merge engine, X3 resolver+provenance, X4 login) resolve before their owning slice. + +## Prior art + +- kubeconfig (clusters / users / contexts; `KUBECONFIG`; `kubectl config view`) +- Helix CLI v2 (`helix.toml` local+enterprise instance blocks; `~/.helix/config`; `~/.helix/credentials`) +- AWS CLI (`~/.aws/config` + `~/.aws/credentials` split; named profiles; `credential_process`) +- git (`~/.gitconfig` + `.git/config`; `--show-origin`) +- Cargo (`Cargo.toml` manifest + `~/.cargo/config.toml`) +- Supabase / Prisma (one project manifest; connection via `DATABASE_URL` env) +- 12-factor app (config that varies by deploy lives in the environment) diff --git a/docs/dev/rfc-003-mcp-server-surface.md b/docs/dev/rfc-003-mcp-server-surface.md new file mode 100644 index 0000000..32fbce5 --- /dev/null +++ b/docs/dev/rfc-003-mcp-server-surface.md @@ -0,0 +1,270 @@ +# RFC: MCP Server Surface for `omnigraph-server` — Full Tool Parity, Stored Queries, Modular Auth + +**Status:** Proposed +**Date:** 2026-06-01 +**Tickets:** MR-969 (stored queries + MCP exposure — the surface this completes), MR-956 (federated auth / WorkOS OAuth — the auth substrate this consumes), MR-971 (per-server credential resolver), MR-974 (agent setup surface — the installer that wires this), MR-668 (multi-graph server — shipped, the routing this builds on) +**Builds on:** [omnigraph#128](https://github.com/ModernRelay/omnigraph/pull/128) (`ragnorc/stored-queries-mcp`) — the shipped stored-query registry, `GET /queries`, `POST /queries/{name}`, and the coarse `invoke_query` gate. +**Supersedes:** the MCP-transport portion of [rfc-001-queries-envelope-mcp.md](rfc-001-queries-envelope-mcp.md) (`/mcp/tools` + `/mcp/invoke`). See [Relationship to RFC-001](#relationship-to-rfc-001). +**Target release:** v0.8.x (phased — see Rollout) + +## Summary + +Add a first-class **MCP (Model Context Protocol) server surface to `omnigraph-server`**, exposed over **Streamable HTTP**, that projects the server's operations as MCP tools and resources for LLM clients (Claude Code/Desktop/web, Cursor, etc.). Two populations of tools share one projection path: + +1. **Built-in operational tools** — parity with the existing `@modernrelay/omnigraph-mcp` stdio package's **13 tools** (`health`, `snapshot`, `read`, `schema_get`, `branches_list`, `commits_list`, `commits_get`, `change`, `ingest`, `branches_create`, `branches_delete`, `branches_merge`, `schema_apply`) and its **2 resources** (`omnigraph://schema`, `omnigraph://branches`), plus a new server-scoped `graphs_list` tool and an `omnigraph://graphs` resource (multi-graph mode). +2. **Dynamic stored-query tools** — one MCP tool per `mcp.expose: true` entry in the `queries:` registry (MR-969 / #128), with parameters typed from the `.gq` declaration via the shipped `query_catalog_entry` / `param_descriptor` projection. + +Every tool is **authorized by the server's existing Cedar policy engine**. The MCP layer never implements its own authentication: it consumes an **already-resolved `ResolvedActor`** from the server's bearer middleware (`require_bearer_auth` today; the `TokenVerifier` seam when MR-956 lands), so the **same MCP endpoint serves on-prem (static or customer-OIDC tokens) and our cloud (WorkOS OAuth) by configuration only**. Cloud OAuth is an additive layer (RFC 9728 protected-resource metadata) that slots in with zero MCP changes. + +The end-state collapses two diverging tool implementations into one: the in-server MCP is the canonical, Cedar-gated, remotely-reachable surface; the stdio package becomes a thin stdio↔HTTP proxy (local on-ramp) over it. + +> **Key caveat, stated up front (see §5.9 below):** the headline "a token scoped via Cedar to a *specific set* of stored queries" requires **per-query `invoke_query` scope**, which is *designed* (rfc-001) but **not yet implemented** — the shipped action is coarse (any stored query on the graph, or none). Per-actor Cedar curation works today for *built-in vs ad-hoc vs admin* tools and for *stored-vs-ad-hoc*; sub-selecting individual stored queries per actor is gated on a prerequisite (PR 0b). Until then, stored-query curation is graph-level (registry membership + `mcp.expose`). + +## Relationship to RFC-001 + +[rfc-001-queries-envelope-mcp.md](rfc-001-queries-envelope-mcp.md) (MR-656 / MR-976 / MR-969) is the parent design for stored queries + the response envelope + MCP. This RFC is the **detailed MCP-transport design** that #128 left for a follow-up, and it **revises rfc-001 in three places where the shipped code or the MCP wire protocol diverged from rfc-001's sketch**: + +1. **Transport shape.** rfc-001 sketched `GET /mcp/tools` + `POST /mcp/invoke` (a bespoke REST pair). **That is not the MCP wire protocol — real MCP clients cannot connect to it.** This RFC implements actual MCP JSON-RPC over Streamable HTTP and reuses `query_catalog_entry` as a *projection source*, not a parallel surface. (rfc-001's own Open Question already leaned toward Streamable HTTP.) +2. **Exposure config.** rfc-001 specified inline `.gq` pragmas (`@mcp(expose=…)`, default `expose=false`). **#128 shipped a different mechanism:** YAML `queries..mcp.expose` in `omnigraph.yaml`, **default `true`** (declaring a query in the manifest *is* the opt-in). This RFC builds on the shipped YAML form; the `.gq`-pragma design in rfc-001 is superseded for exposure. +3. **Schema introspection.** rfc-001 lists "Schema introspection through MCP" as a **non-goal** ("agents see types through declared return shapes"). This RFC **revises that**: the operational-parity tools include `schema_get` and `omnigraph://schema` — *because the shipped stdio package already exposes both*. The non-goal is achieved by *policy*, not omission: `schema_get`/`omnigraph://schema` are Cedar-gated by `Read`, and the recommended locked-down agent policy denies `Read`, so a curated agent still never sees the schema. (rfc-001's intent is preserved; the mechanism moves from "don't build it" to "build it, gate it.") + +Everything else in rfc-001 (two-paths-one-engine, per-query `invoke_query` *as the intended scope*, the response envelope, multi-graph per-graph endpoints) this RFC consumes unchanged. + +> **Numbering note:** the `TokenVerifier`/WorkOS auth design is referred to in code (`crates/omnigraph-server/src/identity.rs`) as "RFC 0001," which is a *different* document from this repo's `docs/dev/rfc-001-queries-envelope-mcp.md`. To avoid the collision this RFC cites the auth substrate as **MR-956** throughout, never "RFC 0001." + +## Reconciliation with shipped code (verified against `ragnorc/stored-queries-mcp` HEAD) + +Verified against `crates/omnigraph-server/src/{lib.rs,api.rs}` and `crates/omnigraph-policy/src/lib.rs` at the current branch head (not the #128 PR body, and not `api.rs` alone): + +- ✅ `GET /queries` returns the `mcp.expose == true` subset as `QueriesCatalogOutput { queries: [QueryCatalogEntry] }`, each with typed `ParamDescriptor`s, `tool_name`, `description`, `instruction`, and a `mutation` flag. **MCP-ready projection, but exposed as bespoke REST/JSON — not the MCP wire protocol.** +- ✅ `POST /queries/{name}` route exists (`server_invoke_query`, `lib.rs`). +- ✅ `query_catalog_entry()` / `param_descriptor()` with an exhaustive `ScalarType → ParamKind` map (a new scalar is a compile error). +- ✅ `InvokeQuery` Cedar action defined in `omnigraph-policy`. +- ✅ **`InvokeQuery` IS enforced** at `POST /queries/{name}`: `server_invoke_query` calls `authorize(PolicyAction::InvokeQuery)` and **masks a denial to a 404 identical to "unknown query"** so the catalog isn't probeable (the denial-masking the previous draft of this RFC reported as missing is shipped — it lives in `lib.rs`, not `api.rs`). The stored-mutation path is already double-gated: `InvokeQuery` outer, then `Change` inside `run_mutate`. +- ✅ **Reuse path exists:** `run_query` / `run_mutate` are already decoupled from their HTTP request bodies and take registry-supplied `(source, name, params, branch/snapshot)`. MCP `tools/call` for both stored and ad-hoc tools delegates to these — no new business logic. +- ❌ **Per-query (`invoke_query[name]`) scope is NOT implemented.** `PolicyRequest` carries only `{action, branch, target_branch}` — **no query-name dimension** — and the action is documented coarse ("permits *any* stored query on the graph"). rfc-001 *designed* per-name scope; it is unbuilt. This RFC's per-query Cedar filtering (§5.4) and recommended agent policy (§5.9) depend on it → tracked as **PR 0b**. +- ❌ No MCP protocol surface (`initialize`/`tools/list`/`tools/call`, JSON-RPC, transport). +- ❌ No `TokenVerifier` trait yet — `require_bearer_auth` resolves a `ResolvedActor` inline (static-hash). The trait/`OidcJwtVerifier` are MR-956 (draft). The MCP layer's only requirement — *consume `ResolvedActor`* — is satisfiable today. + +Stack (verified `Cargo.toml`): Axum + utoipa (OpenAPI) + `omnigraph-policy` (Cedar) + `futures` + `tokio`. **No MCP crate present.** `edition = "2024"`. + +## Motivation + +- **One curated, safe, remotely-reachable tool surface.** MR-969's thesis: hand an LLM a token Cedar-scoped to a set of tools and it sees exactly those typed tools — cannot construct ad-hoc queries it isn't permitted, cannot read the schema it isn't permitted, cannot reach other graphs. Today the only MCP is the stdio package: local-only, full surface, ungated. +- **Parity, so the in-server MCP can be the single implementation.** Operators/agents already depend on the operational tools. Supporting them server-side behind one Cedar gate lets the stdio package degrade to a proxy and removes two diverging tool sets. +- **On-prem and cloud from one endpoint.** A managed cloud (WorkOS OAuth) and an on-prem/air-gapped deploy (static or customer-OIDC tokens) must serve the same MCP without forks or MCP-specific auth. +- **Foundation for the agent on-ramp (MR-974).** `omnigraph mcp install --agent ` needs a decided transport + a stable endpoint. + +## Goals + +- Project built-in tools + stored queries as MCP tools through **one** registry abstraction. +- `tools/list` and the callable set are **identical for argument-independent authorization**, both driven by Cedar (see §5.4 for the branch-scoped caveat). +- The MCP layer is **auth-method-agnostic**: it consumes `ResolvedActor`, never a raw token, never branches on how auth happened. +- The same endpoint works on-prem (static/OIDC) and cloud (WorkOS OAuth), switched by config; cloud OAuth is additive (RFC 9728). +- No new business logic: MCP tools delegate to the same `run_query`/`run_mutate`/branch/schema functions the HTTP routes call. +- Behaviour-neutral when unused: no MCP traffic = no change. + +## Non-Goals + +- **Building/hosting an OAuth authorization server.** The server is a Resource Server; WorkOS AuthKit+Connect is the AS (MR-956). The MCP endpoint validates tokens, never issues them, never holds client secrets. +- **OAuth/WorkOS implementation itself** — MR-956's work. This RFC leaves a clean RFC-9728 hook and consumes `ResolvedActor`. +- **MCP prompts, elicitation, `tools/list_changed`, resource subscriptions, server-initiated messages.** None needed → enables a stateless POST-only transport (§5.6). +- **stdio transport inside the server.** stdio stays in the TS package (now a proxy). +- **Cross-graph tool listing.** Per-graph catalogs only (MR-969 + RFC-002 non-goal). +- **Hot reload of the query registry.** Restart-only (MR-969). + +## Background + +`omnigraph-server` (Axum) already implements every operation this RFC exposes as an authenticated HTTP route; each authorizes via a `PolicyAction` against the Cedar policy for a server-resolved actor and calls into the engine. The existing stdio MCP package is a *client* of these routes (it owns no business logic). MR-956 will introduce a `TokenVerifier` trait (`StaticHashTokenVerifier` today inline, `OidcJwtVerifier` for OIDC/WorkOS) producing the `ResolvedActor { actor_id, tenant_id: Option, scopes: Vec, source }` that already exists in `identity.rs` and is consumed by Cedar — token *validation* is offline (cached JWKS), so on-prem/air-gapped has no request-path dependency on the cloud. + +## Design + +### 5.1 One tool model: a `McpTool` trait, two populators + +Both built-in and stored-query tools implement one trait so `tools/list` / `tools/call` never special-case: + +```rust +trait McpTool: Send + Sync { + fn name(&self) -> &str; // MCP tool id (stable) + fn title(&self) -> Option<&str>; + fn description(&self) -> &str; + fn input_schema(&self) -> serde_json::Value; // JSON Schema (draft 2020-12) + fn annotations(&self) -> ToolAnnotations; // readOnlyHint / destructiveHint / idempotentHint + /// The Cedar request(s) this call requires, given parsed args. Used BOTH at + /// list-time (dry-run filter, default args) and call-time (enforce, real args). + fn authorization(&self, args: &ToolArgs) -> Vec; + async fn call(&self, ctx: &GraphCtx, args: ToolArgs) -> Result; +} +``` + +- **Built-ins**: ~14 static impls, each delegating to the *same* function its HTTP route calls (`run_query`, `run_mutate`, branch ops, `apply_schema_as`, …). `input_schema` authored once (or derived from each route's existing `utoipa`/`ToSchema` DTO). +- **Stored queries**: generated `McpTool` instances, one per `mcp.expose` entry; `input_schema` from `param_descriptor` (§5.3); `authorization` → `InvokeQuery` (coarse today; `InvokeQuery{name}` after PR 0b) then the inner `Read`/`Change`. + +`ToolRegistry` for a graph = the static built-ins + the dynamic stored-query tools resolved from that graph's `GraphHandle` registry. + +### 5.2 Tool catalog (parity) and Cedar mapping + +Each built-in **reuses the exact `PolicyAction` its HTTP route already enforces** — verified against the handlers in `lib.rs`, not invented: + +| MCP tool | Scope | Read/Mutate | Cedar action (verified from route) | +|---|---|---|---| +| `health` | server | read | none (liveness/version) | +| `graphs_list` *(new)* | server | read | `GraphList` | +| `snapshot` | graph | read | `Read` | +| `schema_get` | graph | read | `Read` | +| `branches_list` | graph | read | `Read` | +| `commits_list`, `commits_get` | graph | read | `Read` | +| `read` (ad-hoc `.gq`) / `query` *(alias)* | graph | read | `Read` | +| `change` (ad-hoc `.gq`) / `mutate` *(alias)* | graph | mutate | `Change` | +| `ingest` (NDJSON) | graph | mutate | `Change` (+ `BranchCreate` when forking a new branch) | +| `branches_create` | graph | mutate | `BranchCreate` | +| `branches_delete` | graph | mutate | `BranchDelete` | +| `branches_merge` | graph | mutate | `BranchMerge` | +| `schema_apply` (`allow_data_loss`) | graph | mutate | `SchemaApply` | +| **stored query** (`find_user`, …) | graph | inferred | `InvokeQuery` (coarse; `InvokeQuery{name}` after PR 0b) + inner `Read`/`Change` | + +There is **no `Ingest` and no separate `snapshot`/`Export` action** — `ingest` enforces `Change`, `snapshot` enforces `Read`. (`Export` exists but maps to the `/export` route, which this RFC does not expose as a tool.) + +**Tool id parity vs. canonicalization.** The shipped stdio package uses tool ids **`read`/`change`** (and calls the deprecated `/read`,`/change` routes). The server HTTP surface canonicalized to `/query`,`/mutate` with `/read`,`/change` deprecated (MR-656). To keep existing package clients working *and* align with the server, the MCP exposes **`query`/`mutate` as canonical with `read`/`change` retained as deprecated-but-live aliases** (both dispatch to the same handler). Open Q7 asks whether to drop the aliases later. + +Resources (§5.5): `omnigraph://schema`, `omnigraph://branches` (parity), plus `omnigraph://graphs` *(new)* — each gated by the same action as its list/get route (`Read`, `Read`, `GraphList`). + +### 5.3 `ParamDescriptor → JSON Schema` (stored-query tools) + +| `ParamKind` | JSON Schema | Notes | +|---|---|---| +| String | `{"type":"string"}` | | +| Bool | `{"type":"boolean"}` | | +| Int (i32/u32) | `{"type":"integer"}` | | +| BigInt (i64/u64) | `{"type":"string","pattern":"^-?\\d+$"}` | JSON numbers lose precision >2⁵³ → string (matches the shipped `api.rs` rationale). (Open Q1) | +| Float (f32/f64) | `{"type":"number"}` | | +| Date | `{"type":"string","format":"date"}` | | +| DateTime | `{"type":"string","format":"date-time"}` | | +| Blob | `{"type":"string","contentEncoding":"base64"}` | | +| Vector | `{"type":"array","items":{"type":"number"},"minItems":dim,"maxItems":dim}` | uses `vector_dim` | +| List | `{"type":"array","items":}` | scalar items only (grammar guarantees) | + +`nullable == false` → param is in `required`. Annotations: `mutation` → `{readOnlyHint:false, destructiveHint:true}`; else `{readOnlyHint:true}`. `description` → tool description; `instruction` → appended to description (or `_meta`). (The shipped `check()` already warns when an `mcp.expose` query declares a `Vector` param an LLM can't supply.) + +For built-in tools the schema is hand-authored from the route DTO; e.g. `query` → `{source: string, branch?: string, params?: object}`; `schema_apply` → `{schema: string, allow_data_loss?: boolean}`; `ingest` → `{ndjson: string, mode?: "merge"|"append"|"overwrite", branch?: string}`. + +### 5.4 `tools/list` (Cedar-filtered) and `tools/call` (dispatch + masking) + +- **`tools/list`**: build the `ToolRegistry`; for each tool evaluate `authorization(default_args)` against the actor's Cedar policy; **emit only tools that authorize**. Authz decisions memoized per request. Stored-query tools additionally require `mcp.expose: true`. + - **Exactness caveat (R7 is conditional):** the listed set equals the callable set **only for tools whose authorization is argument-independent** (`health`, `graphs_list`, `snapshot`, `schema_get`, `branches_list`, `commits_*`, ad-hoc `query`/`mutate`, and stored queries under the *coarse* action). For **branch-scoped tools** (`branches_create`/`merge` with `target_branch_scope`, and any branch-scoped `Read`/`Change` rule), list-time uses `default_args` (e.g. branch `main`) and cannot know the real target, so the listed set is a *best-effort approximation* of callability — a call may still be denied (or, rarely, a hidden tool would have been allowed). `tools/call` is always the authoritative gate. The contract is: **list never shows a tool the actor can't ever call; for branch-scoped tools it may show one the actor can call only on some branches.** +- **`tools/call`**: resolve `name` → `McpTool` (masked-404 if unknown *or* `mcp.expose:false`); parse+validate args against `input_schema`; enforce `authorization(args)` (mutations stay double-gated: `InvokeQuery` then `Change`); on success `call`. **Denial masking** lives in one place (the dispatcher): an authz denial is returned identically to "unknown tool" (§5.10), reusing the same deny≡missing principle already shipped at `POST /queries/{name}`. + +### 5.5 Resources + +Advertise `resources` capability (`subscribe:false, listChanged:false`). `resources/list` → the URIs the actor may read; `resources/read` → schema `.pg` text / branches JSON / (multi-graph) graphs JSON, each gated by the corresponding action (`Read`, `Read`, `GraphList`). A locked-down agent denied `Read` simply never sees `omnigraph://schema` or `omnigraph://branches` — this is how rfc-001's "agents don't introspect schema" intent is met *by policy* (§Relationship-to-RFC-001). + +### 5.6 Transport: Streamable HTTP, stateless, POST-only + +- **Streamable HTTP** (MCP's current standard; we're already an HTTP server). One endpoint per scope (§5.7). +- Because the server emits **no** server-initiated messages, implement the **minimal conformant** shape: client `POST`s JSON-RPC, server replies `application/json`. **No SSE channel, no `Mcp-Session-Id`, stateless** — each request authenticated independently via the bearer middleware. Honour the `MCP-Protocol-Version` header. SSE/sessions can be added later if subscriptions land. +- **JSON-RPC methods:** `initialize` (advertise `{tools:{listChanged:false}, resources:{listChanged:false, subscribe:false}}` + serverInfo/version), `notifications/initialized` (no-op ack), `ping`, `tools/list`, `tools/call`, `resources/list`, `resources/read`. `prompts/list` returns empty if probed. +- **Library decision (Open Q2):** spike `rmcp` (official Rust MCP SDK) for conformance + Streamable-HTTP/Axum on edition 2024; **fall back to a hand-rolled ~150 LOC JSON-RPC-over-POST** (only the methods above) on friction. Given the tiny surface, hand-roll is an acceptable default. + +### 5.7 Endpoint routing (server- vs graph-scoped) + +- **Single-graph mode:** `POST /mcp` — graph tools + server tools (`health`, `graphs_list`). +- **Multi-graph mode (MR-668):** `POST /graphs/{graph_id}/mcp` — graph-scoped tools for that graph; plus a server-level `POST /mcp` exposing only server-scoped tools (`health`, `graphs_list`). A per-graph endpoint never lists another graph's tools (isolation, tested). Mirrors the shipped `/graphs/{graph_id}/…` cluster routing. (Open Q5: confirm naming + whether server tools also appear on the per-graph endpoint.) + +### 5.8 Modular / decoupled auth (the cross-cutting requirement) + +**Invariant (load-bearing, satisfiable today):** the MCP handler receives an **already-resolved `ResolvedActor`** and **branches on nothing** about how the token was verified. No token parsing, no method check, no OAuth inside the MCP module. Today that actor comes from `require_bearer_auth`; when MR-956 lands it comes from a `TokenVerifier` — the MCP code is identical either way. + +``` +request → [auth middleware: ResolvedActor] → [MCP route] → Cedar → McpTool +``` + +**Server side — auth is config, not code:** + +| Deployment | Verifier | MCP change | +|---|---|---| +| On-prem, static bearer | `require_bearer_auth` / `StaticHashTokenVerifier` | none | +| On-prem, customer IdP | `OidcJwtVerifier` → customer issuer (MR-956) | none | +| Our cloud | `OidcJwtVerifier` → WorkOS, `tenant_id = Some(org_id)` (MR-956) | none | + +Token validation is offline (cached JWKS) — on-prem/air-gapped keeps working with no request-path cloud dependency. The MCP endpoint never terminates OAuth and never holds a client secret (Resource Server only). + +**Cloud client negotiation — additive, no MCP changes:** when MR-956 lands, the server publishes RFC 9728 `/.well-known/oauth-protected-resource` and returns `WWW-Authenticate: Bearer ..., resource_metadata="..."` on 401. A compliant MCP client (Claude) then auto-negotiates: static bearer to an on-prem endpoint; on a cloud 401 it discovers the WorkOS AS and runs OAuth/PKCE itself — **same endpoint URL, zero client-side branching.** This RFC only requires that MCP routes flow through the standard 401 path so that hook can be added later without touching MCP. + +**Multi-user identity pass-through (cloud):** the *caller's* token (a WorkOS JWT, audience-bound per-tenant) must reach the server so Cedar enforces per-user/per-tenant policy — never a shared service token. The MCP endpoint validates it offline and maps `org_id → tenant_id`. This is why the **remote path is the in-server HTTP MCP that Claude connects to directly** (its token flows through), not a stdio bridge impersonating a user. + +**Client-side credential acquisition (CLI/SDK/proxy) — pluggable `CredentialSource`** (RFC-002 §5, MR-971), keyed by server name, so OAuth is a future *sibling key*, not a re-key: + +```yaml +servers: + onprem: { endpoint: https://og.internal:8080, auth: { token: { env: OG_TOKEN } } } + edge: { endpoint: https://og-edge, auth: { token: { command: [vault, read, -field=token, secret/og] } } } + cloud: { endpoint: https://api.omnigraph.cloud, auth: { oauth: { issuer: workos } } } # future sibling +``` + +Implicit chain when `auth:` omitted: `OMNIGRAPH_TOKEN_` → keychain `omnigraph:` → `[]` in `~/.omnigraph/credentials`; legacy `bearer_token_env` honoured. Secrets never inlined. + +### 5.9 Safety model — Cedar is the gate, default-deny is the floor + +With ad-hoc `query`/`mutate`/`schema_apply` present as tools, the **only** thing protecting an untrusted agent is the Cedar policy. Therefore: + +- **Default-deny when tokens are configured** (MR-723, shipped) is the floor — an actor with no grants sees an empty tool list. +- **What works today (coarse action):** a policy can hide all ad-hoc tools and admin tools per-actor (`deny Read, Change, SchemaApply, Branch*`) while allowing stored queries (`allow InvokeQuery`). That already reproduces "can't run ad-hoc, can't read schema, can only call stored queries" — the agent sees *every* exposed stored query plus nothing else. +- **What needs PR 0b (per-query scope):** selecting *which* stored queries an actor may call (`allow InvokeQuery [find_user, list_orders]`, deny the rest). The shipped `invoke_query` is coarse (all stored queries or none). Until PR 0b adds a query-name dimension to `PolicyRequest` + the Cedar schema (rfc-001's intended design), per-actor sub-selection of stored queries is **not expressible**; curation is graph-level (which `.gq` files are registered + `mcp.expose`). +- `schema_apply`, `branches_delete`, ad-hoc `mutate` require an explicit admin-tier grant; never in a default agent policy. +- (Open Q3) Optional `mcp.allow_adhoc` server switch defaulting **off** for the ad-hoc `query`/`mutate` tools — defence-in-depth independent of Cedar, and independent of PR 0b. + +### 5.10 Result shaping and error mapping + +- **Success:** `tools/call` returns `content: [{type:"text", text:}]` where `` is the route's existing output envelope (read rows / mutation summary, i.e. `ReadOutput` / `ChangeOutput`). (Open Q4: also emit `structuredContent` + `outputSchema` — defer; text-JSON for v1.) +- **Tool execution error** (bad params after schema validation, engine error): result with `isError:true` + a text content block. +- **Authorization denial / unknown tool / `mcp.expose:false`:** a single JSON-RPC error (`-32602`, message `"unknown tool"`) — identical for all three so policy isn't probeable (same principle as the shipped `POST /queries/{name}` 404 masking). +- **Auth failure** (bad/absent bearer): HTTP 401 from the middleware *before* MCP — carries `WWW-Authenticate` (the RFC 9728 hook), never masked as a tool error. (This is exactly the path the shipped `authorize`/`authorize_request` split preserves: operational failures keep their status; only *denials* are masked.) + +## Relationship to the `@modernrelay/omnigraph-mcp` stdio package + +Verified surface of the package (`omnigraph-ts`, pkg version `0.3.0`, `@modelcontextprotocol/sdk@^1.29.0`, **stdio only**): **13 tools** (`health`, `snapshot`, `read`, `schema_get`, `branches_list`, `commits_list`, `commits_get`, `change`, `ingest`, `branches_create`, `branches_delete`, `branches_merge`, `schema_apply`) and **2 resources** (`omnigraph://schema`, `omnigraph://branches`). It is a thin client over the SDK → HTTP routes and **forwards the caller's bearer verbatim** (no inspection). + +Once parity lands, **collapse to one implementation**: the in-server MCP is canonical (Cedar-gated, remote-capable, the path that becomes a Claude-web connector via MR-956). The stdio package degrades to a **thin stdio↔HTTP proxy** forwarding JSON-RPC (and the incoming `Authorization`) to `/mcp` — staying the local on-ramp for Claude Code/Desktop while sharing one tool set, one Cedar gate. Transition: keep the current independent stdio package on its `0.3.x`/`0.6.x` line; ship proxy mode in a later TS minor once the server endpoint is GA. (Note: the package is currently several minors behind the server — its vendored `spec/openapi.json` predates the stored-query routes — so it needs the standard re-sync regardless of MCP work.) + +## Testing + +- **Protocol conformance:** `initialize` handshake + advertised capabilities; `tools/list` shape; `tools/call` happy path; JSON-RPC error envelopes (`-32601` unknown method, `-32602` invalid params / unknown tool); `resources/list` + `resources/read`. +- **Cedar filtering (coarse, today):** an actor with `allow InvokeQuery` + `deny Read/Change` sees *all* exposed stored queries but **not** `query`/`mutate`/`schema_get`; `tools/call query` returns masked "unknown tool"; an admin sees the full catalog. +- **Cedar filtering (per-query, gated on PR 0b):** actor scoped to `InvokeQuery [find_user]` sees *only* `find_user`; `tools/call list_orders` masks. **This test ships with PR 0b**, not PR 1 — it cannot pass against the coarse action. +- **Parity per built-in:** each tool round-trips against the same expectations as its HTTP route (reuse route tests); `read`/`change` aliases dispatch identically to `query`/`mutate`. +- **Double-gating:** a stored mutation requires both `InvokeQuery` and `Change`; `schema_apply` requires `SchemaApply`. +- **`mcp.expose:false`:** absent from `GET /queries` and MCP `tools/list`; still service-callable by name through `POST /queries/{name}` when the actor has `invoke_query`, but not MCP-callable. +- **Schema generation:** table-driven over every `ParamKind` incl. nullable / list / vector(dim). +- **Branch-scoped list approximation:** assert the documented R7 caveat — a branch-scoped policy lists `branches_create`, and `tools/call` is the authoritative gate (a denied target still 403s/masks). +- **Multi-graph isolation:** `/graphs/a/mcp` never lists graph `b`'s tools; server `/mcp` exposes only server tools. +- **Auth decoupling:** the MCP suite is green under the current `require_bearer_auth` and under a mock OIDC `ResolvedActor` source — proving verifier-agnosticism. A 401 carries `WWW-Authenticate`. +- **OpenAPI:** the JSON-RPC endpoint is not REST — document only the envelope in utoipa (or exclude); keep `openapi.json` drift test green (`OMNIGRAPH_UPDATE_OPENAPI=1` to regenerate on intentional change). +- **Cross-repo smoke (optional):** point `@modelcontextprotocol/sdk` (TS) at the HTTP endpoint in an `omnigraph-ts` integration test. + +## Rollout — phased by risk + +- **PR 0a — extract the reusable invoke path (small).** The coarse `invoke_query` gate + 404 denial-masking are **already shipped** in `server_invoke_query`. Extract the read/mutate dispatch into `invoke_stored_query(handle, name, params, branch/snapshot, actor)` so MCP `tools/call` and the HTTP route share one path. No behaviour change. *(Replaces the previous draft's "PR 0 — wire the gate", which was already done.)* +- **PR 0b — per-query `invoke_query` scope (the safety prerequisite).** Add a query-name dimension to `PolicyRequest` + the Cedar schema (rfc-001's intended design), wire it at `POST /queries/{name}` and in the stored-query `McpTool::authorization`. Independently useful (the `allow InvokeQuery [find_user]` policy). **Gates the per-query Cedar-filtering test and §5.9's recommended agent policy.** +- **PR 1 — MCP transport + read-only parity + stored-query reads.** Endpoint(s), `initialize`/`tools/list`/`tools/call`/`resources/*`, the `McpTool` registry, Cedar-filtered listing, the read-only built-ins (`health`, `graphs_list`, `snapshot`, `read`/`query`, `schema_get`, `branches_list`, `commits_*`) + resources + stored-query *reads*. All auth-agnostic. +- **PR 2 — mutating parity + stored-query mutations.** `change`/`mutate`, `ingest`, `branches_create/delete/merge`, `schema_apply`, stored-query mutations + the `mcp.allow_adhoc` switch. +- **PR 3 — docs + agent on-ramp hook.** `docs/user/server.md` MCP section (incl. the recommended agent policy + the coarse-vs-per-query caveat), `openapi.json` sync, the `omnigraph mcp install` config target (MR-974), and the downstream `omnigraph-ts` re-sync/proxy follow-up. +- **Later (separate, MR-956):** RFC 9728 protected-resource metadata + WorkOS — slots in with zero MCP changes. +- **Later (TS minor):** stdio package → proxy mode. + +## Migration / backwards compatibility + +- **Additive.** No `queries:` and no MCP traffic → today's behaviour unchanged. New endpoints are new routes. +- **Cedar default-deny** (when tokens configured) means MCP exposes nothing until an actor is granted — safe by default. +- The stdio package keeps working unchanged; proxy mode is opt-in later. +- `openapi.json` only gains the documented MCP envelope; existing REST routes untouched. + +## Open Questions + +1. **BigInt/u64 as JSON string** (recommended, precision-safe) vs number. +2. **`rmcp` vs hand-rolled** JSON-RPC (spike `rmcp` on edition 2024; default to hand-roll on friction). +3. **Default-off `mcp.allow_adhoc`** for ad-hoc `query`/`mutate` (recommended) vs always-on + Cedar-only. +4. **`structuredContent` + `outputSchema`** now vs text-JSON v1 (recommend v1 text-JSON). +5. **Endpoint paths:** `/mcp` + `/graphs/{id}/mcp` — confirm naming and whether server-scoped tools also appear on the per-graph endpoint. +6. **Stateless POST-only** confirmed (no near-term server-initiated messages) — revisit only if subscriptions land. +7. **Legacy alias tools** (`read`/`change`): keep for client compat (the shipped package uses them), or drop and rely on `query`/`mutate`? +8. **PR 0b shape:** per-query scope as a Cedar *resource* (`StoredQuery::"find_user"`) vs a `query_name` *context attribute* + policy condition — affects how `allow InvokeQuery [list]` is authored. diff --git a/docs/user/cli-reference.md b/docs/user/cli-reference.md index 0326e64..019e4ad 100644 --- a/docs/user/cli-reference.md +++ b/docs/user/cli-reference.md @@ -20,10 +20,11 @@ A reference for the `omnigraph` binary's command surface and `omnigraph.yaml` sc | `run list \| show \| publish \| abort` | transactional run ops | | `schema plan \| apply \| show (alias: get)` | migrations | | `lint` (alias: `check`) | offline / graph-backed query validation. Replaces `query lint` / `query check`, which are kept as deprecated argv-level shims that print a one-line warning and rewrite to `omnigraph lint` | +| `queries validate \| list` | operate on the server-side stored-query registry (the `queries:` block). `validate` type-checks every stored query against the live schema offline (opens the selected graph; exits non-zero on any breakage), catching schema drift without restarting the server; `list` prints the selected registry's query names, MCP exposure, and typed params. For per-graph registries, pass `--target ` or set `cli.graph`; with no graph selection, `list` shows only top-level `queries:`. Distinct from `lint`, which validates a single `.gq` file | | `optimize` | non-destructive Lance compaction | | `cleanup --keep N --older-than 7d --confirm` | destructive version GC | | `embed` | offline JSONL embedding pipeline | -| `policy validate \| test \| explain` | Cedar tooling | +| `policy validate \| test \| explain` | Cedar tooling. Selects `cli.graph`, else `server.graph`, else top-level `policy.file` | | `version` / `-v` | print `omnigraph 0.3.x` | ## `omnigraph.yaml` schema @@ -34,6 +35,13 @@ graphs: : uri: bearer_token_env: + queries: # per-graph stored-query registry (server-role; multi-graph mode) + : # key MUST equal the `query ` symbol inside the .gq + file: # relative to this config's directory + mcp: + expose: true # default true: listed in the MCP catalog (GET /queries); set false to hide (still HTTP-callable) + tool_name: # optional MCP tool-name override (defaults to ; + # must be unique across exposed queries) server: graph: bind: @@ -59,6 +67,8 @@ aliases: graph: branch: format: +queries: # top-level registry — applies only to a bare-URI (anonymous) graph; a graph served by name uses its `graphs..queries`. Mirrors top-level `policy`. + : { file: } # mcp.expose defaults to true policy: file: ./policy.yaml ``` diff --git a/docs/user/policy.md b/docs/user/policy.md index 749d3be..ec0d214 100644 --- a/docs/user/policy.md +++ b/docs/user/policy.md @@ -14,10 +14,11 @@ Per-graph actions (bind to `Omnigraph::Graph::""`): 6. `branch_delete` 7. `branch_merge` 8. `admin` — reserved for policy-management surfaces (hot reload, audit log, approvals). No call site today; see MR-724 for the reservation rationale. +9. `invoke_query` — gates invoking a server-side stored query (the `queries:` registry). Graph-scoped (like `admin`) — per-branch access is enforced by the inner `read` / `change` gate, so a rule that sets `branch_scope` on `invoke_query` is rejected. Coarse in this release: an `invoke_query` allow rule permits any stored query on the graph; a future, additive refinement adds an optional per-query-name scope without changing rules written against the coarse action. Enforced at `POST /queries/{name}` (see [server](server.md)). A stored *mutation* is double-gated: `invoke_query` to reach the tool, plus `change` for the write itself (the engine `_as` writers still enforce per the query body). Server-scoped action (v0.6.0+; binds to `Omnigraph::Server::"root"`): -9. `graph_list` — `GET /graphs` registry enumeration (multi-graph mode) +10. `graph_list` — `GET /graphs` registry enumeration (multi-graph mode) Server-scoped actions cannot use `branch_scope` or `target_branch_scope` — they operate on the registry, not on a graph's branches. A rule cannot mix server-scoped and per-graph actions; split into separate rules. (Runtime `graph_create` / `graph_delete` are reserved but not shipped in v0.6.0; operators add/remove graphs by editing `omnigraph.yaml` and restarting.) @@ -46,10 +47,15 @@ graphs: # no per-graph policy → no engine-layer Cedar enforcement on beta ``` -Top-level `policy.file` is single-graph / CLI-local policy only. Multi-graph -server startup rejects it because applying one graph policy to every configured -graph is ambiguous. Move per-graph rules to `graphs..policy.file` and -move `graph_list` rules to `server.policy.file`. +**Config follows graph identity, not server mode.** A graph served by **name** +(`--target ` or `server.graph`) uses its own `graphs..policy.file`, +exactly as in multi-graph mode. Top-level `policy.file` applies only to an +**anonymous** graph — one served by a bare `` with no `graphs:` entry. +Serving a **named** graph (single- or multi-graph mode) while top-level +`policy.file` (or `queries:`) is populated **refuses boot**, naming the block, +since the top-level value would otherwise be silently shadowed by the per-graph +block. Move per-graph rules to `graphs..policy.file` and `graph_list` +rules to `server.policy.file`. Each graph's HTTP request flows through its own per-graph policy. The management endpoint (`GET /graphs`) flows through the server-level policy. When `server.policy.file` is unset, `GET /graphs` is denied in every runtime state, including `--unauthenticated`; with bearer tokens configured, it returns 403 after admission control because `graph_list` is not a `read`-equivalent action. The operator must explicitly authorize via `server-policy.yaml` to expose `/graphs`. @@ -92,6 +98,10 @@ bearer token. ## CLI +Policy tooling resolves its graph like server single-mode policy: `cli.graph` +wins, otherwise `server.graph` is used, otherwise the top-level `policy.file` +is validated/tested/explained as the anonymous policy. + - `omnigraph policy validate` — parse + count actors, exit 1 on parse error. - `omnigraph policy test` — run cases in `policy.tests.yaml`, exit 1 on any expectation mismatch. - `omnigraph policy explain --actor … --action … [--branch …] [--target-branch …]` — show decision and matched rule. diff --git a/docs/user/server.md b/docs/user/server.md index 6f55e16..67b5afe 100644 --- a/docs/user/server.md +++ b/docs/user/server.md @@ -6,7 +6,9 @@ Axum 0.8 + tokio + utoipa-generated OpenAPI. **Two modes** (v0.6.0+): single-gra ### Single-graph mode (legacy) -`omnigraph-server ` or `omnigraph-server --target --config omnigraph.yaml`. Routes are flat — `/snapshot`, `/read`, `/branches`, etc. Behavior unchanged from v0.6.0. +`omnigraph-server ` or `omnigraph-server --target --config omnigraph.yaml`. Routes are flat — `/snapshot`, `/read`, `/branches`, etc. + +**Config follows graph identity.** A bare `` is an *anonymous* graph and uses the **top-level** `policy.file` / `queries:`. A graph chosen by **name** (`--target` / `server.graph`) uses its own `graphs..{policy.file, queries}` — the same block multi-graph mode uses. ⚠️ *Changed from v0.6.0, which always used top-level config in single mode: a named-graph config that puts `policy`/`queries` at top-level now **refuses boot** and points you at `graphs..…` (move the block there). Bare-`` single mode is unchanged.* ### Multi-graph mode (v0.6.0+) @@ -20,6 +22,10 @@ Mode inference (four-rule matrix): 4. `--config` + non-empty `graphs:` + no single-mode selector → **multi** 5. otherwise → error with migration hint +### Stored-query validation at startup + +If a graph declares a `queries:` registry (see [cli-reference](cli-reference.md)), the server **loads and type-checks every stored query against that graph's live schema at startup** and **refuses to boot** if any query references a type or property the schema lacks — the same fail-loud posture as a malformed policy file, so schema drift surfaces at the deploy boundary rather than at invocation. Two MCP-exposed queries claiming the same tool name is likewise a boot error. Non-blocking advisories (e.g. an MCP-exposed query with a vector parameter an agent cannot supply) are logged. Validate offline before deploying with `omnigraph queries validate`. Discover the exposed queries as a typed tool catalog with `GET /queries`, and invoke one over HTTP with `POST /queries/{name}` (both below). + ## Endpoint inventory Per-graph endpoints — same body shape across modes; URLs differ: @@ -34,6 +40,8 @@ Per-graph endpoints — same body shape across modes; URLs differ: | POST | `/export` | `/graphs/{id}/export` | bearer + `export` | NDJSON stream | `server_export` | | POST | `/mutate` | `/graphs/{id}/mutate` | bearer + `change` | mutation (canonical; `query`/`name`; accepts legacy `query_source`/`query_name` as serde aliases) | `server_mutate` | | POST | `/change` | `/graphs/{id}/change` | bearer + `change` | **deprecated** alias of `/mutate` (carries `Deprecation: true` + `Link: ; rel="successor-version"`) | `server_change` | +| GET | `/queries` | `/graphs/{id}/queries` | bearer + `read` | list the `mcp.expose` stored queries as a typed tool catalog | `server_list_queries` | +| POST | `/queries/{name}` | `/graphs/{id}/queries/{name}` | bearer + `invoke_query` (+ `change` for a stored mutation) | invoke a named query from the `queries:` registry; deny == 404 | `server_invoke_query` | | GET | `/schema` | `/graphs/{id}/schema` | bearer + `read` | get current `.pg` source | `server_schema_get` | | POST | `/schema/apply` | `/graphs/{id}/schema/apply` | bearer + `schema_apply` (target=`main`) | migrate | `server_schema_apply` | | POST | `/ingest` | `/graphs/{id}/ingest` | bearer + `branch_create` (if new) + `change` | bulk load | `server_ingest` (32 MB body limit) | @@ -50,6 +58,23 @@ Server-level management endpoints (v0.6.0+): |---|---|---|---|---| | GET | `/graphs` | bearer + `graph_list` on `Server::"root"` | list registered graphs | `server_graphs_list` (405 in single mode) | +### Stored-query catalog (`GET /queries`) + +List the graph's **`mcp.expose`** stored queries as a typed tool catalog — enough for a client (e.g. an MCP server) to register each as a tool without fetching `.gq` source. Each entry: `{ name, tool_name, description, instruction, mutation, params }`, where each param is `{ name, kind, item_kind?, vector_dim?, nullable }`. `kind` is one of `string | bool | int | bigint | float | date | datetime | blob | vector | list` (decomposed so a consumer maps it with a closed `switch`, never re-parsing GQ type spelling). `bigint` (I64/U64), `date`, `datetime`, and `blob` are carried as JSON **strings** — a 64-bit integer loses precision as a JSON number, dates are ISO strings, and a blob is a URI string. + +- **Read-gated** (works in default-deny mode). The catalog is **graph-wide** (branch-independent; `read` is authorized against `main`). +- **`mcp.expose` defaults to `true`** — declaring a query in `queries:` lists it; set `mcp: { expose: false }` to keep it HTTP/service-callable but hidden from the catalog. +- **Not Cedar-filtered per query (yet).** A caller with `read` but not `invoke_query` can *list* a query they can't *invoke* (which would 404). Closing that gap is future per-query authorization; for now the catalog is a discovery surface and `invoke_query` remains the invocation gate. + +### Stored-query invocation (`POST /queries/{name}`) + +Invoke a curated, server-side stored query by **name** — the source comes from the graph's `queries:` registry, so the client never sends `.gq`. The request body itself is optional; omit it for no-param queries, or send `{ "params": { … }, "branch": "main", "snapshot": null }`, where every field is optional and `params` keys match the query's declared parameters. The response is the **read envelope** (`ReadOutput`) for a stored read or the **mutation envelope** (`ChangeOutput`) for a stored mutation — serialized untagged, so the wire shape is identical to `/query` / `/mutate`. + +- **Gate:** `invoke_query` (per-graph, graph-scoped) at the boundary. A stored *mutation* is **double-gated** — it also passes the engine's `change` gate, so an actor with `invoke_query` but not `change` gets `403`. +- **Deny == unknown, for callers without `invoke_query`:** for a caller lacking the grant, an `invoke_query` denial and an unknown query name return the **same `404`** (identical body), so the catalog can't be probed. A caller that *holds* `invoke_query` may still get the inner gate's `403` for an existing query it can't `read`/`change` (the double-gate, above) — so existence is visible to grant-holders by design. +- **Requires an explicit policy grant when auth is on.** In default-deny mode (bearer tokens but no `policy.file`), only `read` is permitted, so *every* `/queries/{name}` call returns `404` until an `invoke_query` rule is configured. +- A stored mutation cannot target a `snapshot` (`400`); a parameter type error is a structured `400` naming the parameter. + ## Adding and removing graphs (multi mode) Runtime add/remove via API is **not** exposed in v0.6.0 — neither diff --git a/openapi.json b/openapi.json index d1fa337..08d39c4 100644 --- a/openapi.json +++ b/openapi.json @@ -829,6 +829,177 @@ ] } }, + "/queries": { + "get": { + "tags": [ + "queries" + ], + "summary": "List the graph's exposed stored queries as a typed tool catalog.", + "description": "Returns the `mcp.expose == true` subset of the `queries:` registry, each\nwith its MCP tool name, read/mutate flag, description/instruction, and\ntyped parameters — enough for a client to register them as tools without\nfetching `.gq` source. Read-gated; the catalog is graph-wide (branch\nindependent — `read` is authorized against `main`). **Not** Cedar-filtered\nper query yet, so it can list a query whose `invoke_query` the caller\nlacks (a known gap until per-query authorization lands).", + "operationId": "list_queries", + "responses": { + "200": { + "description": "Stored-query catalog (the mcp.expose subset, with typed params)", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/QueriesCatalogOutput" + } + } + } + }, + "401": { + "description": "Unauthorized", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/ErrorOutput" + } + } + } + }, + "403": { + "description": "Forbidden", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/ErrorOutput" + } + } + } + } + }, + "security": [ + { + "bearer_token": [] + } + ] + } + }, + "/queries/{name}": { + "post": { + "tags": [ + "queries" + ], + "summary": "Invoke a curated, server-side stored query by name.", + "description": "The query source comes from the graph's `queries:` registry, not the\nrequest body — callers send only runtime inputs (`params`, `branch`,\n`snapshot`). Gated by the `invoke_query` Cedar action at the boundary;\na stored *mutation* additionally passes the engine's `change` gate\n(double-gated). An actor **without** `invoke_query` cannot tell a denied\nquery from a missing one — both return the same 404, so the catalog\ncan't be probed without the grant. Once `invoke_query` is held, the\ninner `read`/`change` gate may surface a 403 for an existing query the\nactor can't run (the intended double-gate signal).", + "operationId": "invoke_query", + "parameters": [ + { + "name": "name", + "in": "path", + "description": "Stored query name (the registry key)", + "required": true, + "schema": { + "type": "string" + } + } + ], + "requestBody": { + "content": { + "application/json": { + "schema": { + "oneOf": [ + { + "type": "null" + }, + { + "$ref": "#/components/schemas/InvokeStoredQueryRequest" + } + ] + } + } + } + }, + "responses": { + "200": { + "description": "Read envelope (ReadOutput) or mutation envelope (ChangeOutput), serialized untagged", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/InvokeStoredQueryResponse" + } + } + } + }, + "400": { + "description": "Bad request (param type error; snapshot on a stored mutation)", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/ErrorOutput" + } + } + } + }, + "401": { + "description": "Unauthorized", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/ErrorOutput" + } + } + } + }, + "403": { + "description": "Forbidden (the inner `change` gate for a stored mutation)", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/ErrorOutput" + } + } + } + }, + "404": { + "description": "Unknown stored query, or `invoke_query` denied — indistinguishable to a caller without the grant", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/ErrorOutput" + } + } + } + }, + "409": { + "description": "Merge conflict", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/ErrorOutput" + } + } + } + }, + "429": { + "description": "Per-actor admission cap exceeded; honor `Retry-After` header", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/ErrorOutput" + } + } + } + }, + "500": { + "description": "Policy evaluation error (a denial is reported as 404, not 500)", + "content": { + "application/json": { + "schema": { + "$ref": "#/components/schemas/ErrorOutput" + } + } + } + } + }, + "security": [ + { + "bearer_token": [] + } + ] + } + }, "/query": { "post": { "tags": [ @@ -1628,6 +1799,40 @@ } } }, + "InvokeStoredQueryRequest": { + "type": "object", + "description": "Body for `POST /queries/{name}` — invokes the server-side stored query\nnamed in the path. The query source and name come from the registry,\nnever the body; only the runtime inputs are supplied here.", + "properties": { + "branch": { + "type": [ + "string", + "null" + ], + "description": "Branch to run against. Defaults to `main`; for a stored mutation the\nwrite targets this branch." + }, + "params": { + "description": "JSON object whose keys match the stored query's declared parameters." + }, + "snapshot": { + "type": [ + "string", + "null" + ], + "description": "Snapshot id to read from (read queries only — rejected for a stored\nmutation). Mutually exclusive with `branch`." + } + } + }, + "InvokeStoredQueryResponse": { + "oneOf": [ + { + "$ref": "#/components/schemas/ReadOutput" + }, + { + "$ref": "#/components/schemas/ChangeOutput" + } + ], + "description": "Response for `POST /queries/{name}`: the read envelope for a stored\nread, or the mutation envelope for a stored mutation. Serialized\n**untagged**, so the wire shape is exactly [`ReadOutput`] or\n[`ChangeOutput`] — classification follows the stored query, not a\nwrapper field." + }, "LoadMode": { "type": "string", "description": "Shadow enum for documenting [`LoadMode`] in the OpenAPI schema.", @@ -1698,6 +1903,120 @@ } } }, + "ParamDescriptor": { + "type": "object", + "description": "One declared parameter of a stored query, projected for the catalog.", + "required": [ + "name", + "kind", + "nullable" + ], + "properties": { + "item_kind": { + "oneOf": [ + { + "type": "null" + }, + { + "$ref": "#/components/schemas/ParamKind", + "description": "Element kind when `kind == list` (always a scalar — the grammar\nforbids lists of vectors or nested lists)." + } + ] + }, + "kind": { + "$ref": "#/components/schemas/ParamKind" + }, + "name": { + "type": "string" + }, + "nullable": { + "type": "boolean", + "description": "`false` → the caller must supply it; `true` → optional." + }, + "vector_dim": { + "type": [ + "integer", + "null" + ], + "format": "int32", + "description": "Dimension when `kind == vector`.", + "minimum": 0 + } + } + }, + "ParamKind": { + "type": "string", + "description": "The kind of a stored-query parameter, decomposed so a client (e.g. an\nMCP server) can build a typed input schema with a closed `match` and\nnever re-parse omnigraph's type spelling. `bigint`/`date`/`datetime`/\n`blob` are carried as JSON strings on the wire: a 64-bit integer past\n2^53 loses precision as a JSON number, and Date/DateTime are ISO\nstrings, Blob a blob-URI string.", + "enum": [ + "string", + "bool", + "int", + "bigint", + "float", + "date", + "datetime", + "blob", + "vector", + "list" + ] + }, + "QueriesCatalogOutput": { + "type": "object", + "description": "Response for `GET /queries`: the `mcp.expose` subset of a graph's\nstored-query registry, each with typed parameters.", + "required": [ + "queries" + ], + "properties": { + "queries": { + "type": "array", + "items": { + "$ref": "#/components/schemas/QueryCatalogEntry" + } + } + } + }, + "QueryCatalogEntry": { + "type": "object", + "description": "One entry in the stored-query catalog (`GET /queries`).", + "required": [ + "name", + "tool_name", + "mutation", + "params" + ], + "properties": { + "description": { + "type": [ + "string", + "null" + ] + }, + "instruction": { + "type": [ + "string", + "null" + ] + }, + "mutation": { + "type": "boolean", + "description": "`true` for a stored mutation → an MCP read-only hint of `false`." + }, + "name": { + "type": "string", + "description": "Registry key / invoke path segment (`POST /queries/{name}`)." + }, + "params": { + "type": "array", + "items": { + "$ref": "#/components/schemas/ParamDescriptor" + } + }, + "tool_name": { + "type": "string", + "description": "MCP tool id (the `tool_name` override, else `name`)." + } + } + }, "QueryRequest": { "type": "object", "description": "Inline read-query request for `POST /query`.\n\nFriendlier-named alternative to [`ReadRequest`] for ad-hoc reads and\nAI-agent integration. Mutations are rejected with 400 — use `POST\n/mutate` (or its deprecated alias `POST /change`) for write queries.\nField names are deliberately short (`query`, `name`) to match the GQ\nkeyword and the CLI `-e` flag.", From d54bccb9401551415d426c45714c20caee8ebe86 Mon Sep 17 00:00:00 2001 From: Ragnor Comerford Date: Tue, 2 Jun 2026 17:12:00 +0200 Subject: [PATCH 05/11] fix(optimize): skip blob-bearing tables to avoid Lance compaction crash (#138) * test(optimize): pin Lance blob-column compaction failure as a surface guard Lance compact_files mis-decodes blob-v2 columns under its forced BlobHandling::AllBinary read ("more fields in the schema than provided column indices"), failing even a pristine uniform-V2_2 multi-fragment blob table; reads use descriptor handling and are unaffected. Guard 10 reproduces this and is self-retiring: it turns red on the Lance bump that fixes the bug, forcing LANCE_SUPPORTS_BLOB_COMPACTION to flip. * fix(optimize): skip blob-bearing tables instead of crashing compaction omnigraph optimize aborted the whole sweep when any node/edge table had a Blob property: Lance compact_files cannot decode blob-v2 columns under AllBinary (the column-index error pinned by the surface guard). Skip blob-bearing tables behind a LANCE_SUPPORTS_BLOB_COMPACTION gate and report them via TableOptimizeStats.skipped / SkipReason (surfaced in the CLI and a tracing::warn) instead of erroring, which also isolates the failure so the other tables still compact. Reads/writes are unaffected; only fragment/space reclamation on blob tables is deferred until the upstream Lance fix. Adds a maintenance.rs regression test (validated red with the column-index symptom before the fix, green after), a concise v0.6.1 release note, and updates docs (maintenance, cli-reference, AGENTS capability matrix, invariants Known Gaps, lance.md audit, constants). * refactor(optimize): make TableOptimizeStats and SkipReason non_exhaustive Both are returned result types, never built by callers, so #[non_exhaustive] makes this the last field/variant addition that can break downstream literal construction and keeps future ones non-breaking (review feedback on the public-field addition). The v0.6.1 Compatibility Notes call out the source-level change. Also drops the now-stale "RED today / GREEN after the fix lands" narration in the optimize_skips_blob_table_and_reports_skip test (historical regression context now that the fix is in this branch), and folds in the expanded v0.6.1 release note. * chore(release): bump workspace to v0.6.1 Coherent version bump to accompany the v0.6.1 release note: all five crate manifests + path-dependency constraints, Cargo.lock, the AGENTS.md surveyed-version line, and openapi.json info.version move 0.6.0 -> 0.6.1. Matches the established release pattern (#118 landed the v0.6.0 note + bump together) and resolves the Codex/Devin review flag that a v0.6.1 note without a bump leaves CARGO_PKG_VERSION reporting 0.6.0 and mixed package versions. --- AGENTS.md | 4 +- Cargo.lock | 10 +- crates/omnigraph-cli/Cargo.toml | 10 +- crates/omnigraph-cli/src/main.rs | 9 +- crates/omnigraph-compiler/Cargo.toml | 2 +- crates/omnigraph-policy/Cargo.toml | 2 +- crates/omnigraph-server/Cargo.toml | 8 +- crates/omnigraph/Cargo.toml | 8 +- crates/omnigraph/src/db/mod.rs | 2 +- crates/omnigraph/src/db/omnigraph.rs | 2 +- crates/omnigraph/src/db/omnigraph/optimize.rs | 129 ++++++++++++++++-- .../omnigraph/tests/lance_surface_guards.rs | 85 ++++++++++++ crates/omnigraph/tests/maintenance.rs | 93 ++++++++++++- docs/dev/invariants.md | 9 ++ docs/dev/lance.md | 3 +- docs/releases/v0.6.1.md | 26 ++++ docs/user/cli-reference.md | 2 +- docs/user/constants.md | 1 + docs/user/maintenance.md | 3 +- openapi.json | 2 +- 20 files changed, 363 insertions(+), 47 deletions(-) create mode 100644 docs/releases/v0.6.1.md diff --git a/AGENTS.md b/AGENTS.md index 68de6b8..b876749 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -16,7 +16,7 @@ Tools that support `@`-imports (Claude Code) auto-include all three files via th `CLAUDE.md` is a symlink to this file — there is exactly one source of truth. Edit `AGENTS.md`. -**Version surveyed:** 0.6.0 +**Version surveyed:** 0.6.1 **Workspace crates:** `omnigraph-compiler`, `omnigraph` (engine), `omnigraph-policy`, `omnigraph-cli`, `omnigraph-server` **Storage substrate:** Lance 6.x (columnar, versioned, branchable) **License:** MIT @@ -237,7 +237,7 @@ omnigraph policy explain --actor act-alice --action change --branch main | Per-dataset versioning + time travel | ✅ | `snapshot_at_version`, `entity_at`, snapshot-pinned reads across many tables | | Per-dataset branches | ✅ | **Graph-level** branches (atomic across all sub-tables), lazy fork, system branch filtering | | Atomic single-dataset commits | ✅ | **Multi-table publish via three layers**, NOT a single Lance primitive: (1) per-table Lance `commit_staged` for the data write, (2) `__manifest` row-level CAS via `ManifestBatchPublisher` for cross-table ordering, (3) the open-time recovery sweep for the residual gap between (1) and (2). All three layers ship; the four migrated writers (`MutationStaging::finalize`, `schema_apply`, `branch_merge`, `ensure_indices`) write a `__recovery/{ulid}.json` sidecar before Phase B and delete it after Phase C. The next `Omnigraph::open` (gated on `OpenMode::ReadWrite`) runs the sweep in `db/manifest/recovery.rs`: classify, decide all-or-nothing per sidecar, roll forward via single `ManifestBatchPublisher::publish` or roll back via `Dataset::restore`, and record an audit row in `_graph_commit_recoveries.lance` (queryable via `omnigraph commit list --filter actor=omnigraph:recovery`). Continuous in-process recovery (no restart needed between Phase B failure and recovery) is the goal of a future background reconciler. Engine writes route through a sealed `TableStorage` trait exposing `stage_*` + `commit_staged` as the canonical staged-write surface; documented inline-commit residuals (`delete_where`, `create_vector_index`, plus legacy `append_batch` / `merge_insert_batches` / `overwrite_batch` / `create_*_index`) remain on the trait until upstream Lance ships a public two-phase API ([#6658](https://github.com/lance-format/lance/issues/6658), [#6666](https://github.com/lance-format/lance/issues/6666)) and the migration of every call site completes. | -| Compaction (`compact_files`) | ✅ | `omnigraph optimize` orchestrates over all node/edge tables, bounded concurrency | +| Compaction (`compact_files`) | ✅ | `omnigraph optimize` orchestrates over all node/edge tables, bounded concurrency; **skips blob-bearing tables** (reported via `TableOptimizeStats.skipped`, not silent), gated on `LANCE_SUPPORTS_BLOB_COMPACTION` until the upstream blob-v2 compaction-decode bug is fixed (see [docs/dev/invariants.md](docs/dev/invariants.md) Known Gaps) | | Cleanup (`cleanup_old_versions`) | ✅ | `omnigraph cleanup` with `--keep` / `--older-than` policy | | BTREE / inverted (FTS) / vector indexes | ✅ | `ensure_indices` builds them on every relevant column; idempotent; lazy across branches | | `merge_insert` upsert | ✅ | `LoadMode::Merge`, mutation `update`/`insert`/`delete` lowering | diff --git a/Cargo.lock b/Cargo.lock index a3d6d62..3223b9c 100644 --- a/Cargo.lock +++ b/Cargo.lock @@ -4543,7 +4543,7 @@ dependencies = [ [[package]] name = "omnigraph-cli" -version = "0.6.0" +version = "0.6.1" dependencies = [ "assert_cmd", "clap", @@ -4565,7 +4565,7 @@ dependencies = [ [[package]] name = "omnigraph-compiler" -version = "0.6.0" +version = "0.6.1" dependencies = [ "ahash", "arrow-array", @@ -4586,7 +4586,7 @@ dependencies = [ [[package]] name = "omnigraph-engine" -version = "0.6.0" +version = "0.6.1" dependencies = [ "arc-swap", "arrow-array", @@ -4627,7 +4627,7 @@ dependencies = [ [[package]] name = "omnigraph-policy" -version = "0.6.0" +version = "0.6.1" dependencies = [ "cedar-policy", "clap", @@ -4640,7 +4640,7 @@ dependencies = [ [[package]] name = "omnigraph-server" -version = "0.6.0" +version = "0.6.1" dependencies = [ "arc-swap", "async-trait", diff --git a/crates/omnigraph-cli/Cargo.toml b/crates/omnigraph-cli/Cargo.toml index 0d35ed8..641068e 100644 --- a/crates/omnigraph-cli/Cargo.toml +++ b/crates/omnigraph-cli/Cargo.toml @@ -1,6 +1,6 @@ [package] name = "omnigraph-cli" -version = "0.6.0" +version = "0.6.1" edition = "2024" description = "CLI for the Omnigraph graph database." license = "MIT" @@ -13,10 +13,10 @@ name = "omnigraph" path = "src/main.rs" [dependencies] -omnigraph = { package = "omnigraph-engine", path = "../omnigraph", version = "0.6.0" } -omnigraph-compiler = { path = "../omnigraph-compiler", version = "0.6.0" } -omnigraph-policy = { path = "../omnigraph-policy", version = "0.6.0" } -omnigraph-server = { path = "../omnigraph-server", version = "0.6.0" } +omnigraph = { package = "omnigraph-engine", path = "../omnigraph", version = "0.6.1" } +omnigraph-compiler = { path = "../omnigraph-compiler", version = "0.6.1" } +omnigraph-policy = { path = "../omnigraph-policy", version = "0.6.1" } +omnigraph-server = { path = "../omnigraph-server", version = "0.6.1" } clap = { workspace = true } color-eyre = { workspace = true } serde = { workspace = true } diff --git a/crates/omnigraph-cli/src/main.rs b/crates/omnigraph-cli/src/main.rs index 879f070..29b55c4 100644 --- a/crates/omnigraph-cli/src/main.rs +++ b/crates/omnigraph-cli/src/main.rs @@ -3011,18 +3011,19 @@ async fn main() -> Result<()> { "fragments_removed": s.fragments_removed, "fragments_added": s.fragments_added, "committed": s.committed, + "skipped": s.skipped.map(|r| r.as_str()), })).collect::>(), }); print_json(&value)?; } else { println!("optimize {} — {} tables", uri, stats.len()); for s in &stats { - if s.committed { + if let Some(reason) = s.skipped { + println!(" {:<40} skipped ({reason})", s.table_key); + } else if s.committed { println!( " {:<40} frags {} → {} ✓", - s.table_key, - s.fragments_removed + s.fragments_added - s.fragments_added, - s.fragments_added + s.table_key, s.fragments_removed, s.fragments_added ); } else { println!(" {:<40} no-op", s.table_key); diff --git a/crates/omnigraph-compiler/Cargo.toml b/crates/omnigraph-compiler/Cargo.toml index 229b862..545db83 100644 --- a/crates/omnigraph-compiler/Cargo.toml +++ b/crates/omnigraph-compiler/Cargo.toml @@ -1,6 +1,6 @@ [package] name = "omnigraph-compiler" -version = "0.6.0" +version = "0.6.1" edition = "2024" description = "Schema/query compiler for Omnigraph. Zero Lance dependency." license = "MIT" diff --git a/crates/omnigraph-policy/Cargo.toml b/crates/omnigraph-policy/Cargo.toml index dacda35..3d14fc5 100644 --- a/crates/omnigraph-policy/Cargo.toml +++ b/crates/omnigraph-policy/Cargo.toml @@ -1,6 +1,6 @@ [package] name = "omnigraph-policy" -version = "0.6.0" +version = "0.6.1" edition = "2024" description = "Policy / authorization layer for Omnigraph — Cedar-backed PolicyEngine, PolicyChecker trait, ResourceScope enum." license = "MIT" diff --git a/crates/omnigraph-server/Cargo.toml b/crates/omnigraph-server/Cargo.toml index e9a0e46..5994aa1 100644 --- a/crates/omnigraph-server/Cargo.toml +++ b/crates/omnigraph-server/Cargo.toml @@ -1,6 +1,6 @@ [package] name = "omnigraph-server" -version = "0.6.0" +version = "0.6.1" edition = "2024" description = "HTTP server for the Omnigraph graph database." license = "MIT" @@ -19,9 +19,9 @@ default = [] aws = ["dep:aws-config", "dep:aws-sdk-secretsmanager"] [dependencies] -omnigraph = { package = "omnigraph-engine", path = "../omnigraph", version = "0.6.0" } -omnigraph-compiler = { path = "../omnigraph-compiler", version = "0.6.0" } -omnigraph-policy = { path = "../omnigraph-policy", version = "0.6.0" } +omnigraph = { package = "omnigraph-engine", path = "../omnigraph", version = "0.6.1" } +omnigraph-compiler = { path = "../omnigraph-compiler", version = "0.6.1" } +omnigraph-policy = { path = "../omnigraph-policy", version = "0.6.1" } axum = { workspace = true } clap = { workspace = true } color-eyre = { workspace = true } diff --git a/crates/omnigraph/Cargo.toml b/crates/omnigraph/Cargo.toml index 1fa3436..70f51d8 100644 --- a/crates/omnigraph/Cargo.toml +++ b/crates/omnigraph/Cargo.toml @@ -1,6 +1,6 @@ [package] name = "omnigraph-engine" -version = "0.6.0" +version = "0.6.1" edition = "2024" description = "Runtime engine for the Omnigraph graph database." license = "MIT" @@ -16,8 +16,8 @@ default = [] failpoints = ["dep:fail", "fail/failpoints"] [dependencies] -omnigraph-compiler = { path = "../omnigraph-compiler", version = "0.6.0" } -omnigraph-policy = { path = "../omnigraph-policy", version = "0.6.0" } +omnigraph-compiler = { path = "../omnigraph-compiler", version = "0.6.1" } +omnigraph-policy = { path = "../omnigraph-policy", version = "0.6.1" } lance = { workspace = true } lance-datafusion = { workspace = true } datafusion = { workspace = true } @@ -51,7 +51,7 @@ chrono = { workspace = true } arc-swap = { workspace = true } [dev-dependencies] -omnigraph-compiler = { path = "../omnigraph-compiler", version = "0.6.0" } +omnigraph-compiler = { path = "../omnigraph-compiler", version = "0.6.1" } tokio = { workspace = true } lance-namespace-impls = { workspace = true } serial_test = "3" diff --git a/crates/omnigraph/src/db/mod.rs b/crates/omnigraph/src/db/mod.rs index d0b292f..8702f88 100644 --- a/crates/omnigraph/src/db/mod.rs +++ b/crates/omnigraph/src/db/mod.rs @@ -13,7 +13,7 @@ pub use manifest::{Snapshot, SubTableEntry, SubTableUpdate}; pub(crate) use omnigraph::ensure_public_branch_ref; pub use omnigraph::{ CleanupPolicyOptions, InitOptions, MergeOutcome, Omnigraph, OpenMode, SchemaApplyOptions, - SchemaApplyResult, TableCleanupStats, TableOptimizeStats, + SchemaApplyResult, SkipReason, TableCleanupStats, TableOptimizeStats, }; pub(crate) use run_registry::is_internal_run_branch; diff --git a/crates/omnigraph/src/db/omnigraph.rs b/crates/omnigraph/src/db/omnigraph.rs index 9d1403d..7b8a3f6 100644 --- a/crates/omnigraph/src/db/omnigraph.rs +++ b/crates/omnigraph/src/db/omnigraph.rs @@ -33,7 +33,7 @@ mod optimize; mod schema_apply; mod table_ops; -pub use optimize::{CleanupPolicyOptions, TableCleanupStats, TableOptimizeStats}; +pub use optimize::{CleanupPolicyOptions, SkipReason, TableCleanupStats, TableOptimizeStats}; pub use schema_apply::SchemaApplyOptions; use super::commit_graph::GraphCommit; diff --git a/crates/omnigraph/src/db/omnigraph/optimize.rs b/crates/omnigraph/src/db/omnigraph/optimize.rs index c703836..fff3f54 100644 --- a/crates/omnigraph/src/db/omnigraph/optimize.rs +++ b/crates/omnigraph/src/db/omnigraph/optimize.rs @@ -40,6 +40,20 @@ fn maint_concurrency() -> usize { .unwrap_or(DEFAULT_MAINT_CONCURRENCY) } +/// Whether the installed Lance can compact a dataset that contains blob +/// columns. `false` today: Lance `compact_files` forces +/// `BlobHandling::AllBinary` on the read side, and the blob-v2 struct decoder +/// mis-counts columns ("there were more fields in the schema than provided +/// column indices"), failing even a pristine uniform-V2_2 multi-fragment blob +/// table. Reads are unaffected (queries use descriptor handling). +/// +/// While `false`, [`optimize_all_tables`] skips blob-bearing tables and reports +/// [`SkipReason::BlobColumnsUnsupportedByLance`] instead of aborting the whole +/// sweep. Flip to `true` once the upstream Lance fix ships — the +/// `lance_surface_guards.rs::compact_files_still_fails_on_blob_columns` guard +/// turns red on that bump and forces this flip. Tracked in `docs/dev/lance.md`. +const LANCE_SUPPORTS_BLOB_COMPACTION: bool = false; + /// Retention knobs for [`cleanup_all_tables`]. At least one must be set or /// nothing is cleaned. If both are set, Lance applies them as AND (a manifest /// is kept if it satisfies either — i.e. only manifests older than BOTH the @@ -52,8 +66,45 @@ pub struct CleanupPolicyOptions { pub older_than: Option, } -/// Per-table outcome of `optimize_all_tables`. +/// Why `optimize` did not compact a table. Typed so callers branch on the +/// reason rather than sniffing a string. One variant today, gated by +/// [`LANCE_SUPPORTS_BLOB_COMPACTION`]. +#[derive(Debug, Clone, Copy, PartialEq, Eq)] +#[non_exhaustive] +pub enum SkipReason { + /// The table has one or more `Blob` columns. Lance `compact_files` forces + /// `BlobHandling::AllBinary`, which mis-decodes blob-v2 columns; see + /// [`LANCE_SUPPORTS_BLOB_COMPACTION`] and `docs/dev/lance.md`. + BlobColumnsUnsupportedByLance, +} + +impl SkipReason { + /// Stable machine-readable token for serialized output (e.g. CLI `--json`). + /// Once emitted this is part of the output contract — keep it stable. + pub fn as_str(&self) -> &'static str { + match self { + SkipReason::BlobColumnsUnsupportedByLance => "blob_columns_unsupported_by_lance", + } + } +} + +impl std::fmt::Display for SkipReason { + /// Human-readable reason for CLI and log output. + fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { + let msg = match self { + SkipReason::BlobColumnsUnsupportedByLance => { + "blob columns — Lance compaction unsupported" + } + }; + f.write_str(msg) + } +} + +/// Per-table outcome of `optimize_all_tables`. This is a returned result type, +/// not built by callers, so it is `#[non_exhaustive]`: future fields stay +/// non-breaking and downstream code reads fields rather than constructing it. #[derive(Debug, Clone)] +#[non_exhaustive] pub struct TableOptimizeStats { pub table_key: String, /// Number of source fragments that were rewritten by Lance. @@ -62,6 +113,33 @@ pub struct TableOptimizeStats { pub fragments_added: usize, /// Did this table get a new Lance manifest version from the compaction? pub committed: bool, + /// `Some(reason)` if this table was deliberately not compacted. When set, + /// `fragments_removed == 0`, `fragments_added == 0`, and `!committed`. + pub skipped: Option, +} + +impl TableOptimizeStats { + /// Stat for a table that Lance actually compacted. + fn compacted(table_key: String, metrics: &CompactionMetrics, committed: bool) -> Self { + Self { + table_key, + fragments_removed: metrics.fragments_removed, + fragments_added: metrics.fragments_added, + committed, + skipped: None, + } + } + + /// Stat for a table that was deliberately skipped (compaction not attempted). + fn skipped(table_key: String, reason: SkipReason) -> Self { + Self { + table_key, + fragments_removed: 0, + fragments_added: 0, + committed: false, + skipped: Some(reason), + } + } } /// Per-table outcome of `cleanup_all_tables`. `error` is `Some` when this @@ -84,14 +162,21 @@ pub async fn optimize_all_tables(db: &Omnigraph) -> Result = all_table_keys(&db.catalog()) - .into_iter() - .filter_map(|table_key| { - let entry = snapshot.entry(&table_key)?; + // Compute per-table state (path + whether it has blob columns) up front, in + // a scope that drops the catalog handle before the async stream starts. + let table_tasks: Vec<(String, String, bool)> = { + let catalog = db.catalog(); + let mut tasks = Vec::new(); + for table_key in all_table_keys(&catalog) { + let Some(entry) = snapshot.entry(&table_key) else { + continue; + }; let full_path = format!("{}/{}", db.root_uri, entry.table_path); - Some((table_key, full_path)) - }) - .collect(); + let has_blob = !blob_properties_for_table_key(&catalog, &table_key)?.is_empty(); + tasks.push((table_key, full_path, has_blob)); + } + tasks + }; if table_tasks.is_empty() { return Ok(Vec::new()); @@ -101,7 +186,24 @@ pub async fn optimize_all_tables(db: &Omnigraph) -> Result> = futures::stream::iter(table_tasks.into_iter()) - .map(|(table_key, full_path)| async move { + .map(|(table_key, full_path, has_blob)| async move { + // Lance `compact_files` mis-decodes blob-v2 columns under the forced + // `BlobHandling::AllBinary` read (see LANCE_SUPPORTS_BLOB_COMPACTION). + // Skip blob-bearing tables and report it rather than aborting the + // whole sweep — the other tables still compact. + if has_blob && !LANCE_SUPPORTS_BLOB_COMPACTION { + tracing::warn!( + target: "omnigraph::optimize", + table = %table_key, + "skipping compaction: table has blob columns the current Lance \ + cannot rewrite (blob-v2 AllBinary decode bug); other tables \ + unaffected — rerun after the Lance fix", + ); + return Ok(TableOptimizeStats::skipped( + table_key, + SkipReason::BlobColumnsUnsupportedByLance, + )); + } let mut ds = table_store .open_dataset_head_for_write(&table_key, &full_path, None) .await?; @@ -111,12 +213,11 @@ pub async fn optimize_all_tables(db: &Omnigraph) -> Result RecordBatch { + let ids: Vec = (start..start + n).map(|i| format!("n{i}")).collect(); + let data = + LargeBinaryArray::from_iter_values((start..start + n).map(|i| format!("blob{i}"))); + let blob_uri = StringArray::from(vec![None::<&str>; n as usize]); + let DataType::Struct(fields) = lance::blob::blob_field("content", true).data_type().clone() + else { + unreachable!("blob_field is always a Struct"); + }; + let content = StructArray::new( + fields, + vec![Arc::new(data) as _, Arc::new(blob_uri) as _], + None, + ); + let schema = Arc::new(Schema::new(vec![ + Field::new("id", DataType::Utf8, false), + lance::blob::blob_field("content", true), + ])); + RecordBatch::try_new( + schema, + vec![Arc::new(StringArray::from(ids)) as _, Arc::new(content) as _], + ) + .unwrap() + } + + async fn write(uri: &str, batch: RecordBatch, mode: WriteMode) { + let schema = batch.schema(); + let reader = RecordBatchIterator::new(vec![Ok(batch)], schema); + // Blob v2 requires file version >= 2.2; without the pin the *write* + // would fail with a different error, masking the guard's intent. + let params = WriteParams { + mode, + enable_stable_row_ids: true, + data_storage_version: Some(LanceFileVersion::V2_2), + ..Default::default() + }; + Dataset::write(reader, uri, Some(params)).await.unwrap(); + } + + let dir = tempfile::tempdir().unwrap(); + let uri = dir.path().join("guard10-blob.lance"); + let uri = uri.to_str().unwrap(); + + // Uniform V2_2, two fragments → forces compaction to actually rewrite. + write(uri, blob_batch(0, 2), WriteMode::Create).await; + write(uri, blob_batch(100, 2), WriteMode::Append).await; + + let mut ds = Dataset::open(uri).await.unwrap(); + assert!( + ds.get_fragments().len() >= 2, + "guard needs a multi-fragment table to trigger a real compaction rewrite" + ); + + let result = compact_files(&mut ds, CompactionOptions::default(), None).await; + let err = result.expect_err( + "compact_files unexpectedly SUCCEEDED on a blob table — the Lance blob-v2 \ + compaction bug is fixed. Flip LANCE_SUPPORTS_BLOB_COMPACTION to true in \ + db/omnigraph/optimize.rs, remove the blob-skip branch, and re-pin docs/dev/lance.md.", + ); + assert!( + err.to_string() + .contains("more fields in the schema than provided column indices"), + "blob compaction failed with an unexpected error (Lance internals may have \ + shifted): {err}" + ); +} diff --git a/crates/omnigraph/tests/maintenance.rs b/crates/omnigraph/tests/maintenance.rs index 722bdc4..3e61677 100644 --- a/crates/omnigraph/tests/maintenance.rs +++ b/crates/omnigraph/tests/maintenance.rs @@ -8,7 +8,7 @@ mod helpers; use std::time::Duration; use lance::Dataset; -use omnigraph::db::{CleanupPolicyOptions, Omnigraph}; +use omnigraph::db::{CleanupPolicyOptions, Omnigraph, SkipReason}; use omnigraph::loader::{LoadMode, load_jsonl}; use helpers::{TEST_DATA, TEST_SCHEMA, count_rows, init_and_load}; @@ -72,6 +72,97 @@ async fn optimize_after_load_then_again_is_idempotent() { } } +// Regression: `optimize` must not crash on a graph that has a `Blob` table. +// +// Lance `compact_files` forces `BlobHandling::AllBinary`, which mis-decodes +// blob-v2 columns ("more fields in the schema than provided column indices"), +// failing even a pristine uniform-V2_2 multi-fragment blob table. `optimize` +// must skip blob-bearing tables (and report the skip) rather than aborting the +// whole sweep. +// +// Before the skip fix, `optimize()` returned that Lance error here and aborted +// the whole sweep; it now skips the blob table (`doc.skipped == Some(..)`) +// while the sibling non-blob `Tag` table still compacts. The skip is gated by +// `LANCE_SUPPORTS_BLOB_COMPACTION`; the surface guard +// `compact_files_still_fails_on_blob_columns` flags when the upstream Lance fix +// makes the skip (and this test's blob arm) removable. +#[tokio::test] +async fn optimize_skips_blob_table_and_reports_skip() { + let dir = tempfile::tempdir().unwrap(); + let uri = dir.path().to_str().unwrap(); + // One Blob node type (`Doc`) + one plain node type (`Tag`): proves the blob + // table is skipped while a non-blob table in the same sweep still compacts. + let schema = "\ +node Doc {\n slug: String @key\n content: Blob\n}\n\ +node Tag {\n slug: String @key\n}\n"; + let mut db = Omnigraph::init(uri, schema).await.unwrap(); + + // Multi-fragment blob table: Overwrite creates fragment 1; each Merge of + // new keys appends another. A >=2-fragment blob table is exactly what + // crashes `compact_files` today (single fragment would no-op and not crash). + load_jsonl( + &mut db, + "{\"type\":\"Doc\",\"data\":{\"slug\":\"d1\",\"content\":\"base64:aGVsbG8x\"}}\n{\"type\":\"Doc\",\"data\":{\"slug\":\"d2\",\"content\":\"base64:aGVsbG8y\"}}", + LoadMode::Overwrite, + ) + .await + .unwrap(); + load_jsonl( + &mut db, + "{\"type\":\"Doc\",\"data\":{\"slug\":\"d3\",\"content\":\"base64:aGVsbG8z\"}}", + LoadMode::Merge, + ) + .await + .unwrap(); + load_jsonl( + &mut db, + "{\"type\":\"Doc\",\"data\":{\"slug\":\"d4\",\"content\":\"base64:aGVsbG80\"}}", + LoadMode::Merge, + ) + .await + .unwrap(); + // Plain table, also multi-fragment so it has something to compact. + load_jsonl( + &mut db, + "{\"type\":\"Tag\",\"data\":{\"slug\":\"t1\"}}\n{\"type\":\"Tag\",\"data\":{\"slug\":\"t2\"}}", + LoadMode::Merge, + ) + .await + .unwrap(); + load_jsonl( + &mut db, + "{\"type\":\"Tag\",\"data\":{\"slug\":\"t3\"}}", + LoadMode::Merge, + ) + .await + .unwrap(); + + let stats = db + .optimize() + .await + .expect("optimize must not crash on a graph with a Blob table"); + + let doc = stats + .iter() + .find(|s| s.table_key == "node:Doc") + .expect("Doc stat present"); + let tag = stats + .iter() + .find(|s| s.table_key == "node:Tag") + .expect("Tag stat present"); + // The blob table is skipped (and reported), not compacted. + assert_eq!( + doc.skipped, + Some(SkipReason::BlobColumnsUnsupportedByLance), + "blob table must be reported as skipped", + ); + assert!(!doc.committed, "skipped blob table is not compacted"); + assert_eq!(doc.fragments_removed, 0); + assert_eq!(doc.fragments_added, 0); + // The plain (non-blob) table is unaffected by the skip. + assert_eq!(tag.skipped, None, "non-blob table must not be skipped"); +} + #[tokio::test] async fn cleanup_without_any_policy_option_errors() { let dir = tempfile::tempdir().unwrap(); diff --git a/docs/dev/invariants.md b/docs/dev/invariants.md index 0cf295c..5ee4f17 100644 --- a/docs/dev/invariants.md +++ b/docs/dev/invariants.md @@ -130,6 +130,15 @@ them explicit. - **Deletes and vector indexes:** `delete_where` and vector index creation still advance Lance HEAD inline because the required public Lance APIs are missing. Keep D2 and recovery coverage in place until those residuals are removed. +- **Blob-column compaction:** Lance `compact_files` mis-decodes blob-v2 columns + under its forced `BlobHandling::AllBinary` read ("more fields in the schema + than provided column indices"), so `optimize` skips any table with a `Blob` + property — reporting `SkipReason::BlobColumnsUnsupportedByLance` (loud, not a + silent drop) behind the `LANCE_SUPPORTS_BLOB_COMPACTION` gate. Reads and writes + are unaffected; only space/fragment reclamation on blob tables is deferred. + Remove the skip when the upstream Lance fix lands — the + `lance_surface_guards.rs::compact_files_still_fails_on_blob_columns` guard + turns red on that bump to force it. - **Planner capability/stat surfaces:** cost-aware planning, complete capability advertisement, and explain-with-cost are roadmap. Do not describe them as implemented. diff --git a/docs/dev/lance.md b/docs/dev/lance.md index 100da6f..9d2b990 100644 --- a/docs/dev/lance.md +++ b/docs/dev/lance.md @@ -176,7 +176,8 @@ Migration from Lance 4.0.0 → 6.0.1 landed in this cycle (DataFusion 52 → 53, - **Lance #6666 still open** (`build_index_metadata_from_segments` public): vector-index two-phase blocked; inline `create_vector_index` residual retained. - **Lance #6877 still open** (`MergeInsertBuilder` dup-rowid): PR #109's `SourceDedupeBehavior::FirstSeen` + `check_batch_unique_by_keys` precondition stay load-bearing. - **`Dataset::force_delete_branch`** (`branches().delete(name, force=true)`, dataset.rs:524) tolerates a missing branch-*contents* ref (vs plain `delete_branch`'s `RefNotFound`), but on the local store still errors `NotFound` if the branch `tree/` directory is fully absent (`remove_dir_all`'s NotFound is not caught for Lance's native error variant, refs.rs:526-549). Both variants still refuse a branch with referencing descendants (`RefConflict`). `TableStore::force_delete_branch` wraps this to be fully idempotent (tolerates already-absent). The single-authority branch-delete redesign uses it for orphan reclamation (eager best-effort reclaim + cleanup reconciler). Pinned by `lance_surface_guards.rs::force_delete_branch_semantics`. Branch delete is "flip the ref atomically, then `remove_dir_all(tree/{branch})`"; branch-exclusive data lives under `tree/{branch}/` so a drop reclaims it immediately without touching `main`. +- **Lance blob-v2 `compact_files` bug** (no public issue found as of 2026-06): `compact_files` disables binary-copy for blob datasets and forces `BlobHandling::AllBinary` on the read side; the v2.1+ structural decoder then mis-counts column infos for the blob-v2 struct and fails with `Invalid user input: there were more fields in the schema than provided column indices / infos` (`lance-encoding/src/decoder.rs::ColumnInfoIter::expect_next`). This fails even a pristine uniform-V2_2 multi-fragment blob table; vector/list/scalar/ragged columns and mixed file versions all compact fine. Reads/queries use descriptor handling (`BlobHandling::default()`) and are unaffected. `optimize` skips blob-bearing tables behind `LANCE_SUPPORTS_BLOB_COMPACTION = false` (`db/omnigraph/optimize.rs`), reporting `SkipReason::BlobColumnsUnsupportedByLance`. Pinned by `lance_surface_guards.rs::compact_files_still_fails_on_blob_columns`, which turns red when the bug is fixed → flip the gate, remove the skip branch + the `maintenance.rs::optimize_skips_blob_table_and_reports_skip` skip assertions. -Surface guards added: `crates/omnigraph/tests/lance_surface_guards.rs` (9 named guards; 4 runtime + 5 compile-only). Future Lance bumps re-run this file first as the smoke check. Two additional guards from the original plan deferred to follow-up (`manifest_cas_returns_row_level_contention_variant` needs full publisher-race harness; `table_version_metadata_byte_compatible_with_v4` needs `pub(crate)` reach extension). +Surface guards added: `crates/omnigraph/tests/lance_surface_guards.rs` (10 named guards; 5 runtime + 5 compile-only). Future Lance bumps re-run this file first as the smoke check. Two additional guards from the original plan deferred to follow-up (`manifest_cas_returns_row_level_contention_variant` needs full publisher-race harness; `table_version_metadata_byte_compatible_with_v4` needs `pub(crate)` reach extension). Bump this date stanza on the next alignment pass. diff --git a/docs/releases/v0.6.1.md b/docs/releases/v0.6.1.md new file mode 100644 index 0000000..aafe1af --- /dev/null +++ b/docs/releases/v0.6.1.md @@ -0,0 +1,26 @@ +# Omnigraph v0.6.1 + +v0.6.1 focuses on operational polish after v0.6.0: stored-query registries, safer branch cleanup, more complete release artifacts, and a Lance blob-compaction workaround. + +## Highlights + +- **Stored-query registries.** `omnigraph.yaml` can declare curated `queries:` blocks per graph. Servers load and type-check them at startup, `omnigraph queries validate` checks them offline, `omnigraph queries list` shows exposed queries and typed params, `GET /queries` exposes a typed catalog, and `POST /queries/{name}` invokes a stored query without accepting ad hoc `.gq` source from the client. +- **Stored-query policy gate.** New Cedar action `invoke_query` gates the stored-query invocation surface. Stored mutations are double-gated: `invoke_query` to reach the stored query and `change` for the actual write. +- **Safer branch deletion.** `branch_delete` now treats the manifest as the authority, flips branch visibility atomically, and reclaims per-table/commit-graph forks as derived state. If best-effort reclaim is interrupted, `cleanup` reconciles orphaned forks; reusing a branch name before cleanup reports an actionable error. +- **Blob-safe optimize.** `omnigraph optimize` skips tables with `Blob` properties instead of failing the whole sweep on Lance's blob-v2 compaction decode bug. Skips are visible in human output, `--json` as `skipped`, `TableOptimizeStats.skipped`, and logs; non-blob tables still compact normally. +- **Deployment improvements.** The container entrypoint now composes `OMNIGRAPH_TARGET_URI` with `OMNIGRAPH_CONFIG`, so operators can keep the graph URI in env while loading policy/query config from a mounted file. The local RustFS bootstrap pins RustFS beta.3 and allows the current insecure local-dev default credentials. +- **Windows release support.** Tagged and edge releases now publish Windows x86_64 archives containing `omnigraph.exe` and `omnigraph-server.exe`, with a PowerShell installer and Windows install docs. +- **Release tooling.** Homebrew formula generation was tightened to produce audit-clean formulas. + +## Compatibility Notes + +- A graph selected by name (`--target` or `server.graph`) now uses `graphs..policy` and `graphs..queries`. Top-level `policy` / `queries` blocks are only for anonymous bare-URI single-graph mode; using them with a named graph now fails loudly with migration guidance. +- `mcp.expose` defaults to `true` for stored-query registry entries. Set `mcp: { expose: false }` for service-only queries that should not appear in the catalog. +- `invoke_query` is graph-scoped, not branch-scoped. Branch/snapshot access remains enforced by the inner `read` / `change` gate. +- Blob tables are not compacted until the upstream Lance fix lands, so fragment count and deleted-row space on blob tables are not reclaimed by `optimize`. Reads, writes, and query results are unaffected; no on-disk migration is required. +- `TableOptimizeStats` is now `#[non_exhaustive]` and gains a `skipped: Option` field (so does the new `SkipReason` enum). This is a source-level change only for downstream code that built this returned result struct by literal — rare, since it is produced by `optimize` and consumed by reading its fields; field access is unaffected, and `#[non_exhaustive]` keeps future additions non-breaking. + +## Docs And Cleanup + +- Public docs were updated for stored queries, policy, server routes, deployment, Windows installation, branch deletion, maintenance, and the `runs` docs rename to `writes`. +- README copy and release documentation were refreshed; older release notes had small typo/wording fixes. diff --git a/docs/user/cli-reference.md b/docs/user/cli-reference.md index 019e4ad..8263919 100644 --- a/docs/user/cli-reference.md +++ b/docs/user/cli-reference.md @@ -21,7 +21,7 @@ A reference for the `omnigraph` binary's command surface and `omnigraph.yaml` sc | `schema plan \| apply \| show (alias: get)` | migrations | | `lint` (alias: `check`) | offline / graph-backed query validation. Replaces `query lint` / `query check`, which are kept as deprecated argv-level shims that print a one-line warning and rewrite to `omnigraph lint` | | `queries validate \| list` | operate on the server-side stored-query registry (the `queries:` block). `validate` type-checks every stored query against the live schema offline (opens the selected graph; exits non-zero on any breakage), catching schema drift without restarting the server; `list` prints the selected registry's query names, MCP exposure, and typed params. For per-graph registries, pass `--target ` or set `cli.graph`; with no graph selection, `list` shows only top-level `queries:`. Distinct from `lint`, which validates a single `.gq` file | -| `optimize` | non-destructive Lance compaction | +| `optimize` | non-destructive Lance compaction (skips tables with `Blob` columns; `--json` reports a `skipped` field) | | `cleanup --keep N --older-than 7d --confirm` | destructive version GC | | `embed` | offline JSONL embedding pipeline | | `policy validate \| test \| explain` | Cedar tooling. Selects `cli.graph`, else `server.graph`, else top-level `policy.file` | diff --git a/docs/user/constants.md b/docs/user/constants.md index 527aaea..8f13555 100644 --- a/docs/user/constants.md +++ b/docs/user/constants.md @@ -11,6 +11,7 @@ | Internal manifest schema version | `INTERNAL_MANIFEST_SCHEMA_VERSION = 2` | `db/manifest/migrations.rs` | | Merge stage batch | `MERGE_STAGE_BATCH_ROWS = 8192` | `exec/merge.rs` | | Maintenance concurrency | `OMNIGRAPH_MAINTENANCE_CONCURRENCY=8` | `db/omnigraph/optimize.rs` | +| Lance blob compaction support | `LANCE_SUPPORTS_BLOB_COMPACTION = false` | `db/omnigraph/optimize.rs` | | Graph index cache size | `8` (LRU) | `runtime_cache.rs` | | Default body limit | `1 MB` | `omnigraph-server/lib.rs` | | Ingest body limit | `32 MB` | `omnigraph-server/lib.rs` | diff --git a/docs/user/maintenance.md b/docs/user/maintenance.md index 9839ea1..3628fa0 100644 --- a/docs/user/maintenance.md +++ b/docs/user/maintenance.md @@ -7,7 +7,8 @@ - Lance `compact_files()` on every node + edge table on `main`. - Rewrites small fragments into fewer large ones; old fragments remain reachable via older manifests. - Bounded by `OMNIGRAPH_MAINTENANCE_CONCURRENCY` (default 8). -- Returns `[TableOptimizeStats { table_key, fragments_removed, fragments_added, committed }]`. +- Returns `[TableOptimizeStats { table_key, fragments_removed, fragments_added, committed, skipped }]`. +- **Blob tables are skipped.** A table that declares any `Blob` property is not compacted: it is reported with `skipped: Some(BlobColumnsUnsupportedByLance)` (and logged via `tracing::warn`) instead of compacted, and the rest of the sweep proceeds normally. The current Lance `compact_files` mis-decodes blob-v2 columns under its forced `BlobHandling::AllBinary` read; **reads and writes are unaffected** — only compaction is. This is gated by `LANCE_SUPPORTS_BLOB_COMPACTION` (`db/omnigraph/optimize.rs`) and removed when the upstream Lance fix lands (see [docs/dev/lance.md](../dev/lance.md)). Consequence: fragment count and deleted-row space on blob tables are not reclaimed until then; query results are never affected. ## `cleanup_all_tables(db, options)` — destructive diff --git a/openapi.json b/openapi.json index 08d39c4..aced64d 100644 --- a/openapi.json +++ b/openapi.json @@ -7,7 +7,7 @@ "name": "MIT", "identifier": "MIT" }, - "version": "0.6.0" + "version": "0.6.1" }, "paths": { "/branches": { From 96dbe9dec00b41b68907708d7535437677d3fde7 Mon Sep 17 00:00:00 2001 From: Andrew Altshuler Date: Sat, 6 Jun 2026 00:44:48 +0300 Subject: [PATCH 06/11] fix(release): make Homebrew audit non-blocking + set up brew on runner (#140) The v0.6.1 Release shipped binaries but the Homebrew tap update job died at the audit step (brew not on the ubuntu runner; exit 127), skipping the formula push so the tap stayed at 0.6.0. - Install Homebrew via Homebrew/actions/setup-homebrew so brew is available. - Make both the setup and audit steps continue-on-error: they are best-effort diagnostics (the formula is correct by construction via update-homebrew-formula.sh), so neither can skip the actual tap publish. - Drop --online from brew audit for deterministic, network-independent linting. --- .github/workflows/release.yml | 18 ++++++++++++++++-- 1 file changed, 16 insertions(+), 2 deletions(-) diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml index 3a66ff2..a265c40 100644 --- a/.github/workflows/release.yml +++ b/.github/workflows/release.yml @@ -121,16 +121,30 @@ jobs: run: | ./scripts/update-homebrew-formula.sh "${GITHUB_REF_NAME}" homebrew-tap/Formula/omnigraph.rb + # Diagnostic only: brew is not on PATH on the ubuntu runner by default, so + # set it up explicitly. Both this setup and the audit below are best-effort + # canaries, not gates — continue-on-error on each keeps a failed/flaky brew + # (the action is pinned to a moving @master ref) from skipping the actual + # tap publish below. The formula is correct by construction + # (update-homebrew-formula.sh), so brew tooling must never block the push. + - name: Set up Homebrew + if: env.HOMEBREW_TAP_SKIP != '1' + continue-on-error: true + uses: Homebrew/actions/setup-homebrew@master + - name: Audit generated formula if: env.HOMEBREW_TAP_SKIP != '1' + continue-on-error: true run: | # Audit the checked-out tap by name (brew audit rejects bare paths # and needs tap context). Symlink the checkout into Homebrew's Taps - # tree so `modernrelay/tap/omnigraph` resolves to it. + # tree so `modernrelay/tap/omnigraph` resolves to it. Offline audit + # (no --online) keeps it deterministic; it still catches the + # ComponentsOrder/structure class of problems. tap_dir="$(brew --repository)/Library/Taps/modernrelay/homebrew-tap" mkdir -p "$(dirname "$tap_dir")" ln -sfn "$PWD/homebrew-tap" "$tap_dir" - brew audit --strict --online modernrelay/tap/omnigraph + brew audit --strict modernrelay/tap/omnigraph - name: Commit and push formula update if: env.HOMEBREW_TAP_SKIP != '1' From c7365bf8efd4500d6af16b00eec34d4c2202ca2b Mon Sep 17 00:00:00 2001 From: Andrew Altshuler Date: Sat, 6 Jun 2026 18:09:47 +0300 Subject: [PATCH 07/11] ci(codeowners): un-trap required checks, auto-render, generate owner tables (#142) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The CODEOWNERS required checks blocked every PR — the real root cause was a name mismatch, compounded by a path filter: - branch-protection.json required the contexts `CODEOWNERS / drift` and `CODEOWNERS / noedit` (the GitHub UI "workflow / job-id" display form), but the jobs report check-run names from their `name:` fields — "CODEOWNERS matches source" / "CODEOWNERS not hand-edited". The required contexts therefore never matched any reported check and sat permanently pending. - The workflow was also path-filtered to CODEOWNERS files, so it didn't even run for most PRs. Net effect: with both required checks unsatisfiable, every PR could only land via admin override (e.g. #140). Fixes: - A: drop the `paths:` filter so the workflow runs on every PR and both required contexts always report. - name fix: point branch-protection.json at the actual job names verbatim, and add a doc note that the contexts must equal the job `name:` values. - B: the `drift` job now re-renders and, on same-repo PRs, auto-commits the regenerated artifacts back to the branch (mirrors the openapi.json job in ci.yml); forks / manual runs strict-check instead. Contributors no longer run the script by hand. - D: render-codeowners.py also generates a "who owns what" path->owners + roles table spliced into docs/dev/codeowners.md between markers, so the human-readable view never drifts. Idempotent; CODEOWNERS output unchanged. - docs: correct the stale `enforce_admins: true` line (JSON and live are false). NOTE: the branch-protection.json change only takes effect after an admin runs `./scripts/apply-branch-protection.sh` (deliberate manual step, per docs/dev/branch-protection.md). Until then `main` still requires the old mismatched contexts, so this PR itself needs an admin-override merge — the last one that should be necessary. Co-authored-by: Claude Opus 4.8 (1M context) --- .github/branch-protection.json | 4 +- .github/scripts/render-codeowners.py | 81 ++++++++++++++++++++++++++-- .github/workflows/codeowners.yml | 72 ++++++++++++++++++++----- docs/dev/branch-protection.md | 4 +- docs/dev/codeowners.md | 39 ++++++++++---- 5 files changed, 168 insertions(+), 32 deletions(-) diff --git a/.github/branch-protection.json b/.github/branch-protection.json index 61b7d33..7ca46b9 100644 --- a/.github/branch-protection.json +++ b/.github/branch-protection.json @@ -7,8 +7,8 @@ "Check AGENTS.md Links", "Test Workspace", "Test omnigraph-server --features aws", - "CODEOWNERS / drift", - "CODEOWNERS / noedit" + "CODEOWNERS matches source", + "CODEOWNERS not hand-edited" ] }, "enforce_admins": false, diff --git a/.github/scripts/render-codeowners.py b/.github/scripts/render-codeowners.py index f243d0c..5e96545 100755 --- a/.github/scripts/render-codeowners.py +++ b/.github/scripts/render-codeowners.py @@ -1,10 +1,14 @@ #!/usr/bin/env python3 -"""Render .github/CODEOWNERS from .github/codeowners-roles.yml. +"""Render .github/CODEOWNERS and the ownership tables in +docs/dev/codeowners.md from .github/codeowners-roles.yml. -The yml is the source of truth — editing CODEOWNERS directly is -rejected by CI (see .github/workflows/codeowners.yml). This script -expands the role-based yml into the flat path→owners format GitHub -expects. +The yml is the source of truth. This script expands the role-based yml +into (1) the flat path→owners format GitHub expects in +`.github/CODEOWNERS`, and (2) the "who owns what" markdown tables spliced +between the generated-region markers in `docs/dev/codeowners.md`. Both are +derived artifacts; CI re-renders them on every PR (see +.github/workflows/codeowners.yml) and auto-commits the result on same-repo +PRs, so the source of truth and the human-readable view never drift. Usage: python3 .github/scripts/render-codeowners.py @@ -16,6 +20,7 @@ Exits non-zero on: one owner; otherwise CODEOWNERS would assign nobody and GitHub would silently fall back to "no required reviewer", which defeats the purpose). + - Missing generated-region markers in docs/dev/codeowners.md. """ from __future__ import annotations @@ -34,6 +39,13 @@ except ImportError: REPO_ROOT = Path(__file__).resolve().parents[2] SOURCE = REPO_ROOT / ".github" / "codeowners-roles.yml" OUTPUT = REPO_ROOT / ".github" / "CODEOWNERS" +DOCS = REPO_ROOT / "docs" / "dev" / "codeowners.md" + +# The "who owns what" tables in docs/dev/codeowners.md are spliced between +# these markers so the human-readable view never drifts from the source of +# truth. Edit codeowners-roles.yml and re-render — never the table by hand. +DOCS_BEGIN = "" +DOCS_END = "" BANNER = """\ # AUTOGENERATED from .github/codeowners-roles.yml. Do not edit by hand. @@ -75,6 +87,62 @@ def owners_for(role_names: list[str], roles: dict) -> list[str]: return seen +def _oneline(text: str) -> str: + """Collapse a folded/multi-line YAML description into one cell of text.""" + return " ".join((text or "").split()) + + +def ownership_tables(spec: dict, roles: dict) -> str: + """Render the human-readable "who owns what" markdown — a path→owners + table (the operative view at PR time, in last-match-wins order with the + catch-all first) plus a role→members table. Spliced into the docs between + the markers so it is always current with the source of truth.""" + out: list[str] = [] + + out.append("**Path → owners** (GitHub applies *last match wins*; the `*` " + "catch-all is listed first and is overridden by the specific " + "patterns below it):") + out.append("") + out.append("| Path | Owners | Role(s) |") + out.append("|---|---|---|") + if "default" in spec: + owners = " ".join(owners_for(spec["default"], roles)) + out.append(f"| `*` | {owners} | {', '.join(spec['default'])} |") + for pattern, role_names in (spec.get("paths") or {}).items(): + owners = " ".join(owners_for(role_names, roles)) + out.append(f"| `{pattern}` | {owners} | {', '.join(role_names)} |") + out.append("") + + out.append("**Roles**:") + out.append("") + out.append("| Role | Members | Description |") + out.append("|---|---|---|") + for name, role in roles.items(): + members = " ".join(f"@{m}" for m in (role.get("members") or [])) + out.append(f"| `{name}` | {members} | {_oneline(role.get('description', ''))} |") + out.append("") + + return "\n".join(out) + + +def splice_docs(table_md: str) -> None: + """Replace the region between DOCS_BEGIN/DOCS_END in the docs file with the + freshly generated tables, leaving surrounding prose untouched.""" + if not DOCS.exists(): + sys.exit(f"error: docs file not found: {DOCS}") + text = DOCS.read_text() + if DOCS_BEGIN not in text or DOCS_END not in text: + sys.exit( + f"error: ownership markers not found in {DOCS.relative_to(REPO_ROOT)}. " + f"Add the lines:\n {DOCS_BEGIN}\n {DOCS_END}\n" + f"around the generated table region." + ) + head, rest = text.split(DOCS_BEGIN, 1) + _, tail = rest.split(DOCS_END, 1) + new = f"{head}{DOCS_BEGIN}\n\n{table_md}\n{DOCS_END}{tail}" + DOCS.write_text(new) + + def main() -> int: if not SOURCE.exists(): sys.exit(f"error: source file not found: {SOURCE}") @@ -127,6 +195,9 @@ def main() -> int: OUTPUT.write_text(rendered) print(f"wrote {OUTPUT.relative_to(REPO_ROOT)}") + + splice_docs(ownership_tables(spec, roles)) + print(f"updated {DOCS.relative_to(REPO_ROOT)}") return 0 diff --git a/.github/workflows/codeowners.yml b/.github/workflows/codeowners.yml index 19d5835..75b3515 100644 --- a/.github/workflows/codeowners.yml +++ b/.github/workflows/codeowners.yml @@ -1,19 +1,24 @@ name: CODEOWNERS +# Runs on EVERY pull request (no paths filter). The two jobs below are +# required status checks on `main`; a path-filtered required check never +# reports for PRs outside the filter and leaves them permanently "pending" +# (the trap that forced admin-override merges). Always-run + cheap +# short-circuit is what keeps them honest. on: pull_request: - paths: - - '.github/codeowners-roles.yml' - - '.github/CODEOWNERS' - - '.github/scripts/render-codeowners.py' - - '.github/workflows/codeowners.yml' workflow_dispatch: -# Read-only; we never push from this workflow. +# `drift` auto-commits the regenerated artifacts back to same-repo PR +# branches, so it needs write access. permissions: - contents: read + contents: write jobs: + # NOTE: the job `name:` values below ("CODEOWNERS matches source" / + # "CODEOWNERS not hand-edited") ARE the status-check contexts that + # .github/branch-protection.json must list verbatim. Renaming a job here + # is a branch-protection change — update the JSON and re-apply. drift: name: CODEOWNERS matches source runs-on: ubuntu-latest @@ -28,19 +33,56 @@ jobs: - name: Install PyYAML run: pip install pyyaml - - name: Re-render CODEOWNERS + - name: Re-render CODEOWNERS + ownership docs run: python3 .github/scripts/render-codeowners.py - - name: Reject drift + # Same-repo PR: push the regenerated artifacts back so contributors + # never have to run the script locally. Mirrors the openapi.json + # auto-commit in ci.yml (separate shallow clone of the head branch so + # the pushed commit carries only the regenerated files). + - name: Commit regenerated artifacts to PR branch + if: | + github.event_name == 'pull_request' && + github.event.pull_request.head.repo.full_name == github.repository + env: + GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} run: | - if ! git diff --quiet .github/CODEOWNERS; then - echo "::error::.github/CODEOWNERS is out of sync with .github/codeowners-roles.yml." - echo "::error::Run \`python3 .github/scripts/render-codeowners.py\` locally and commit the result." + if git diff --quiet -- .github/CODEOWNERS docs/dev/codeowners.md; then + echo "CODEOWNERS and ownership docs already in sync." + exit 0 + fi + tmp=$(mktemp -d) + git clone --depth 1 --branch "${{ github.head_ref }}" \ + "https://x-access-token:${GITHUB_TOKEN}@github.com/${{ github.repository }}.git" \ + "$tmp" + cp .github/CODEOWNERS "$tmp/.github/CODEOWNERS" + cp docs/dev/codeowners.md "$tmp/docs/dev/codeowners.md" + cd "$tmp" + if git diff --quiet -- .github/CODEOWNERS docs/dev/codeowners.md; then + echo "Head branch already matches; nothing to push." + exit 0 + fi + git config user.name "github-actions[bot]" + git config user.email "41898282+github-actions[bot]@users.noreply.github.com" + git add .github/CODEOWNERS docs/dev/codeowners.md + git commit -m "chore: regenerate CODEOWNERS + ownership docs" + git push + + # Fork PR / workflow_dispatch: cannot push back, so enforce drift + # strictly. The contributor runs the script and commits the result. + - name: Verify in sync (forks / manual runs) + if: | + !(github.event_name == 'pull_request' && + github.event.pull_request.head.repo.full_name == github.repository) + run: | + if ! git diff --quiet -- .github/CODEOWNERS docs/dev/codeowners.md; then + echo "::error::Generated CODEOWNERS / ownership docs are out of sync with .github/codeowners-roles.yml." + echo "::error::Run \`python3 .github/scripts/render-codeowners.py\` and commit the result." echo "--- diff ---" - git --no-pager diff .github/CODEOWNERS + git --no-pager diff -- .github/CODEOWNERS docs/dev/codeowners.md exit 1 fi - echo "CODEOWNERS is in sync with its source." + echo "Generated artifacts are in sync with their source." noedit: name: CODEOWNERS not hand-edited @@ -52,6 +94,8 @@ jobs: fetch-depth: 0 - name: Reject hand-edits to generated file + # Only meaningful for PRs (needs a base to diff against). + if: github.event_name == 'pull_request' run: | base="origin/${{ github.base_ref }}" git fetch origin "${{ github.base_ref }}" --quiet diff --git a/docs/dev/branch-protection.md b/docs/dev/branch-protection.md index 9b2fa78..2b6cc37 100644 --- a/docs/dev/branch-protection.md +++ b/docs/dev/branch-protection.md @@ -8,7 +8,7 @@ This page explains what the policy says and how to change it. | Setting | Value | Why | |---|---|---| -| **Required status checks (strict)** | `Classify Changes`, `Check AGENTS.md Links`, `Test Workspace`, `Test omnigraph-server --features aws`, `CODEOWNERS / drift`, `CODEOWNERS / noedit` | Every PR must pass workspace tests, AGENTS.md link integrity, and the CODEOWNERS hygiene checks. `strict: true` requires the branch to be up-to-date with `main` before merge. | +| **Required status checks (strict)** | `Classify Changes`, `Check AGENTS.md Links`, `Test Workspace`, `Test omnigraph-server --features aws`, `CODEOWNERS matches source`, `CODEOWNERS not hand-edited` | Every PR must pass workspace tests, AGENTS.md link integrity, and the CODEOWNERS hygiene checks. The two CODEOWNERS contexts must equal the job `name:` values in `.github/workflows/codeowners.yml` **verbatim** — a context naming a job that never reports (the old `CODEOWNERS / drift` used the job *id*, and the job was path-filtered) leaves every PR permanently pending and forces admin overrides. `strict: true` requires the branch to be up-to-date with `main` before merge. | | **Required approving reviews** | `1` | At least one reviewer. With a 2-person team, going higher would block all merges when one person is unavailable. | | **Require code-owner reviews** | `true` | The reviewer must be a code owner per `.github/CODEOWNERS`. This is what makes the codeowners chassis enforced. | | **Dismiss stale reviews on new commits** | `true` | A push after approval invalidates the prior review. Prevents the "approve, then sneak in unreviewed changes" pattern. | @@ -16,7 +16,7 @@ This page explains what the policy says and how to change it. | **Disallow force pushes** | `true` | No history rewrites on `main`. | | **Disallow branch deletions** | `true` | `main` cannot be deleted. | | **Required conversation resolution** | `true` | All review comment threads must be resolved before merge. | -| **Enforce on admins** | `true` | Even repository admins go through the gates. The point is no bypasses. | +| **Enforce on admins** | `false` | Admins can override the gates (`enforce_admins: false` in the JSON). This is the intended escape hatch for the 2-person team; tightening to `true` is tracked under hardening below. | | **Required signed commits** | not yet | Not enabled. Would lock out maintainers until everyone enrolls GPG/SSH commit signing. Tracked as a follow-up. | ## How to apply diff --git a/docs/dev/codeowners.md b/docs/dev/codeowners.md index 9a7fb50..14bba0b 100644 --- a/docs/dev/codeowners.md +++ b/docs/dev/codeowners.md @@ -4,24 +4,45 @@ This setup gives every role change a reviewable PR and a permanent in-repository audit trail (`git log .github/codeowners-roles.yml`). -## Current roles +## Who owns what -| Role | Members | Scope | +The tables below are **generated** from `.github/codeowners-roles.yml` by `.github/scripts/render-codeowners.py` (the same render that produces `.github/CODEOWNERS`). They are the always-current "who owns what at this commit" view — don't edit them by hand; edit the yml and re-render. + + + +**Path → owners** (GitHub applies *last match wins*; the `*` catch-all is listed first and is overridden by the specific patterns below it): + +| Path | Owners | Role(s) | |---|---|---| -| `engineering` | `@ragnorc` | All code under `crates/**`, repository infrastructure, default for unmapped paths | -| `docs` | `@ragnorc` | `docs/**`, README.md, AGENTS.md, CLAUDE.md, SECURITY.md | +| `*` | @ragnorc | engineering | +| `crates/**` | @ragnorc | engineering | +| `docs/**` | @ragnorc | docs | +| `README.md` | @ragnorc | docs | +| `AGENTS.md` | @ragnorc | docs | +| `CLAUDE.md` | @ragnorc | docs | +| `SECURITY.md` | @ragnorc | docs | -GitHub treats multiple owners in a CODEOWNERS line as **"any one of them satisfies the review requirement"**. To require N distinct approvers on a specific path, layer a CI check on top (not currently configured). +**Roles**: + +| Role | Members | Description | +|---|---|---| +| `engineering` | @ragnorc | All production code under crates/**. Engine, CLI, server, compiler. | +| `docs` | @ragnorc | Documentation under docs/**, plus repo-level docs (README.md, AGENTS.md, CLAUDE.md symlink, SECURITY.md). | + + + +GitHub treats multiple owners on a CODEOWNERS line as **"any one of them satisfies the review requirement"**. To require N distinct approvers on a specific path, layer a CI check on top (not currently configured). ## How to change role membership or path mappings 1. Edit `.github/codeowners-roles.yml`. -2. Run `python3 .github/scripts/render-codeowners.py` (requires PyYAML; `pip install pyyaml`). -3. Commit both files in the same PR. +2. Open a PR. **CI re-renders for you**: the `CODEOWNERS` workflow regenerates `.github/CODEOWNERS` and the ownership tables above and auto-commits them back to your PR branch on same-repository PRs — you don't have to run the script locally (though you can: `python3 .github/scripts/render-codeowners.py`, requires PyYAML). + +On a fork (where CI can't push back), the workflow instead fails with the diff so you can run the script and commit it yourself. CI fails the PR if: -- `CODEOWNERS` was edited without a corresponding yml change, or -- The yml was changed but the rendered `CODEOWNERS` doesn't match. +- a fork PR left a generated artifact out of sync, or +- `CODEOWNERS` was edited without a corresponding yml change (the `CODEOWNERS not hand-edited` check). ## How to add a new role From 343f1f17ed8e86032aef6d9a466a778a9c39b6bd Mon Sep 17 00:00:00 2001 From: Andrew Altshuler Date: Sat, 6 Jun 2026 23:58:08 +0300 Subject: [PATCH 08/11] governance: external contribution model (issues/discussions/RFCs/PRs) (#143) Formalize the public contribution surface. Maintainers keep a separate internal process and are exempt from the intake gates; everyone stays bound by review, CODEOWNERS, and branch protection. Model: - Issues = problem reports only (bug form + config.yml redirects ideas to Discussions and disables blank issues). - Discussions = ideas + RFC incubation. - RFCs = anyone (incl. external) authors docs/rfcs/NNNN-*.md; a maintainer merging it is acceptance. Distinct from the maintainer-internal docs/dev/rfc-00N-* track. - PRs = link an `accepted` issue or accepted RFC, or use the trivial fast-lane (typos/docs/deps). Enforced softly to start (template + review). Adds GOVERNANCE.md, rewrites CONTRIBUTING.md, adds docs/rfcs/ (README + template), .github issue/PR/discussion templates. Wires docs/rfcs/ into the doc-link checker (excluded like releases; linked from docs/dev/index.md). Co-authored-by: Claude Opus 4.8 (1M context) --- .github/DISCUSSION_TEMPLATE/rfc.yml | 34 +++++++++ .github/ISSUE_TEMPLATE/bug_report.yml | 55 +++++++++++++ .github/ISSUE_TEMPLATE/config.yml | 13 ++++ .github/PULL_REQUEST_TEMPLATE.md | 29 +++++++ CONTRIBUTING.md | 38 +++++++-- GOVERNANCE.md | 106 ++++++++++++++++++++++++++ docs/dev/index.md | 12 +++ docs/rfcs/0000-template.md | 54 +++++++++++++ docs/rfcs/README.md | 66 ++++++++++++++++ scripts/check-agents-md.sh | 7 +- 10 files changed, 406 insertions(+), 8 deletions(-) create mode 100644 .github/DISCUSSION_TEMPLATE/rfc.yml create mode 100644 .github/ISSUE_TEMPLATE/bug_report.yml create mode 100644 .github/ISSUE_TEMPLATE/config.yml create mode 100644 .github/PULL_REQUEST_TEMPLATE.md create mode 100644 GOVERNANCE.md create mode 100644 docs/rfcs/0000-template.md create mode 100644 docs/rfcs/README.md diff --git a/.github/DISCUSSION_TEMPLATE/rfc.yml b/.github/DISCUSSION_TEMPLATE/rfc.yml new file mode 100644 index 0000000..2a63525 --- /dev/null +++ b/.github/DISCUSSION_TEMPLATE/rfc.yml @@ -0,0 +1,34 @@ +labels: ["rfc"] +body: + - type: markdown + attributes: + value: | + Use this to **incubate an RFC** — socialize a design and reach rough + consensus before writing the formal document. When it's ready, graduate + it into a pull request that adds `docs/rfcs/NNNN-title.md` + (see [docs/rfcs/README.md](../blob/main/docs/rfcs/README.md)); a + maintainer merging that PR is acceptance. + + For a plain feature request or open-ended idea, use the **Ideas** + category instead. For bugs, open an [Issue](../../issues/new/choose). + - type: textarea + id: problem + attributes: + label: Problem / motivation + description: What needs solving, and why is it worth the long-run cost? + validations: + required: true + - type: textarea + id: sketch + attributes: + label: Proposed direction (sketch) + description: A rough shape of the design. Detail comes later in the RFC document. + validations: + required: true + - type: textarea + id: invariants + attributes: + label: Invariants touched + description: Which items in docs/dev/invariants.md does this affect or risk? Any deny-list brush? + validations: + required: false diff --git a/.github/ISSUE_TEMPLATE/bug_report.yml b/.github/ISSUE_TEMPLATE/bug_report.yml new file mode 100644 index 0000000..8e19465 --- /dev/null +++ b/.github/ISSUE_TEMPLATE/bug_report.yml @@ -0,0 +1,55 @@ +name: Bug report +description: Report a reproducible problem or wrong behavior in OmniGraph. +title: "bug: " +labels: ["bug", "needs-triage"] +body: + - type: markdown + attributes: + value: | + Issues are for **reporting problems** — concrete, reproducible bugs. + For ideas, feature requests, or questions, please use + [Discussions](../../discussions) instead. + For a security vulnerability, follow [SECURITY.md](../../blob/main/SECURITY.md) — do **not** file it here. + + A maintainer will triage this; once labelled **`accepted`** it's open for a pull request + (see [GOVERNANCE.md](../../blob/main/GOVERNANCE.md)). + - type: textarea + id: what-happened + attributes: + label: What happened + description: What went wrong, and what you expected instead. + validations: + required: true + - type: textarea + id: repro + attributes: + label: Steps to reproduce + description: Minimal steps, commands, schema/query, or a failing snippet. + placeholder: | + 1. omnigraph init ... + 2. omnigraph ... + 3. observed: ... / expected: ... + validations: + required: true + - type: input + id: version + attributes: + label: Version + description: Output of `omnigraph --version` (or the engine/crate version) and how you installed it. + validations: + required: true + - type: input + id: environment + attributes: + label: Environment + description: OS, architecture, and storage backend (local FS / S3 / RustFS / MinIO). + validations: + required: false + - type: textarea + id: logs + attributes: + label: Logs / output + description: Relevant error text or logs. Will be rendered as code. + render: shell + validations: + required: false diff --git a/.github/ISSUE_TEMPLATE/config.yml b/.github/ISSUE_TEMPLATE/config.yml new file mode 100644 index 0000000..50720b8 --- /dev/null +++ b/.github/ISSUE_TEMPLATE/config.yml @@ -0,0 +1,13 @@ +# Issues are for problem reports only. Disable blank issues so everything is +# routed: bugs through the form, everything else to Discussions / SECURITY.md. +blank_issues_enabled: false +contact_links: + - name: 💡 Idea, feature request, or RFC + url: https://github.com/ModernRelay/omnigraph/discussions + about: Propose features and designs in Discussions. RFCs graduate from there into a docs/rfcs/ pull request. + - name: ❓ Question or help + url: https://github.com/ModernRelay/omnigraph/discussions + about: Ask in Discussions — questions are not tracked as Issues. + - name: 🔒 Security vulnerability + url: https://github.com/ModernRelay/omnigraph/blob/main/SECURITY.md + about: Report security issues privately per SECURITY.md — never as a public Issue. diff --git a/.github/PULL_REQUEST_TEMPLATE.md b/.github/PULL_REQUEST_TEMPLATE.md new file mode 100644 index 0000000..2a548c7 --- /dev/null +++ b/.github/PULL_REQUEST_TEMPLATE.md @@ -0,0 +1,29 @@ + + +## What & why + + + +## Backing issue / RFC + + + +- [ ] Fixes an **accepted** issue: Closes # +- [ ] Implements / is an **accepted** RFC: +- [ ] **Trivial fast-lane** (typo / docs / dependency bump / comment / one-line CI) — no issue/RFC required + +## Checklist + +- [ ] Change is focused (one logical change) +- [ ] Tests added/updated for behavior changes (or N/A) +- [ ] Public docs updated if user-facing surface changed (or N/A) +- [ ] Reviewed against [docs/dev/invariants.md](../blob/main/docs/dev/invariants.md) — no Hard Invariant weakened, no deny-list item hit (or justified) + +## Notes for reviewers + + diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md index 8d9c687..2d77ef0 100644 --- a/CONTRIBUTING.md +++ b/CONTRIBUTING.md @@ -1,10 +1,29 @@ # Contributing -Small bug fixes and documentation improvements are welcome directly through pull -requests. +Thanks for your interest in OmniGraph. This page is the practical how-to; the +rules and decision authority behind it live in [GOVERNANCE.md](GOVERNANCE.md). -For larger changes, please open an issue or design discussion first so the -proposed direction is clear before implementation starts. +## Start in the right place + +| I want to… | Go to | Notes | +|---|---|---| +| **Report a bug** or wrong behavior | **[Open an Issue](../../issues/new/choose)** | Concrete and reproducible. A maintainer triages it; once labelled **`accepted`** it's open for a PR. | +| **Suggest a feature / share an idea / ask** | **[Start a Discussion](../../discussions)** | Ideas and questions live here, not in Issues. | +| **Propose a design / RFC** | **An RFC pull request** | Anyone can author one — see [docs/rfcs/README.md](docs/rfcs/README.md). A maintainer merging it is acceptance. | +| **Fix something / implement a change** | **A pull request** | Must link an `accepted` issue or an accepted RFC — unless it's trivial (below). | +| **Report a security vulnerability** | **[SECURITY.md](SECURITY.md)** | Do **not** open a public Issue. | + +### When can I just open a PR? +The **trivial fast-lane** — open directly, no prior issue/RFC needed: typo and +wording fixes, doc corrections, dependency bumps, comment fixes, obvious +one-line CI tweaks. Anything more substantial needs a backing `accepted` issue +or accepted RFC first, so the *why* is agreed before the *how* is reviewed. A PR +that turns out to be non-trivial will be redirected — that's about process, not +the merit of the change. + +> **Maintainers (ModernRelay team)** follow a separate internal process and are +> not bound by the intake rules above. Everyone is bound by review, CODEOWNERS, +> branch protection, and CI. ## Development @@ -49,6 +68,11 @@ CI runs both. ## Pull Requests -- keep changes focused -- include tests for behavior changes when practical -- update public docs when the user-facing surface changes +- **Link the backing issue or RFC** (`Closes #123`, or reference the RFC) — or + mark the PR as trivial per the fast-lane. +- Keep changes focused; one logical change per PR. +- Include tests for behavior changes when practical. +- Update public docs when the user-facing surface changes. + +New to the codebase? Read [AGENTS.md](AGENTS.md) — the architecture map and the +always-on invariants every change is reviewed against. diff --git a/GOVERNANCE.md b/GOVERNANCE.md new file mode 100644 index 0000000..5878f1f --- /dev/null +++ b/GOVERNANCE.md @@ -0,0 +1,106 @@ +# Governance + +This document describes how **external contributions** to OmniGraph are +proposed, accepted, and merged. It exists so an outside contributor can answer, +without asking: *where does my report/idea/change go, who decides, and what has +to happen before code lands?* + +> **Scope.** This governs the public contribution surface — Issues, +> Discussions, RFCs, and pull requests from people outside the ModernRelay +> team. **Maintainers operate under a separate internal process** and are not +> bound by the intake gates below. Everyone, maintainer or not, is still bound +> by the universal gates: branch protection on `main` and CODEOWNERS review +> (see [docs/dev/branch-protection.md](docs/dev/branch-protection.md) and +> [docs/dev/codeowners.md](docs/dev/codeowners.md)). + +## Roles + +| Role | Who | Authority | +|---|---|---| +| **Maintainer** | The code owners in [`.github/CODEOWNERS`](.github/CODEOWNERS) (generated from [`.github/codeowners-roles.yml`](.github/codeowners-roles.yml)) | Validate issues, accept/reject RFCs, review and merge PRs, set direction. Final decision authority. | +| **Contributor** | Anyone else | Report problems (Issues), propose ideas (Discussions), author RFCs, and open pull requests. | + +Decision authority rests with the maintainers. CODEOWNERS is the single source +of truth for who that is; this document does not duplicate the list. + +## The three channels + +Each channel has one job. Using the right one is the first thing we ask of a +contribution. + +| Channel | Purpose | Not for | +|---|---|---| +| **[Issues](../../issues)** | **Report a problem** — a bug, a regression, a documented behavior that's wrong. Something concrete and reproducible. | Feature requests, ideas, questions, or design proposals (→ Discussions). | +| **[Discussions](../../discussions)** | **Propose and explore** — new ideas, feature requests, questions, and the incubation of RFCs. | Bug reports (→ Issues). | +| **Pull requests** | **Land a sanctioned change** — a fix for a *validated* issue, an *accepted* RFC, or a trivial change (see fast-lane). | Substantive change with no backing issue/RFC — it will be redirected. | + +## How a change becomes mergeable + +``` + ┌─────────── bug ───────────┐ ┌──────── idea / feature ────────┐ + ▼ │ ▼ │ + Issue (problem report) │ Discussion (idea / RFC incubation) │ + │ │ │ │ + maintainer triage │ rough consensus │ + │ │ │ graduate │ + ▼ │ ▼ │ + label: accepted ──────────┐ │ RFC PR (docs/rfcs/NNNN-*.md) │ + │ │ │ │ │ + │ │ │ maintainer review │ + ▼ ▼ │ ▼ │ + Pull request ◀──────────┴──────────│── merged == accepted │ + (links the issue or the accepted RFC) ◀───────┘ (implementation PRs reference it) │ + │ + review + CODEOWNERS + branch protection + ▼ + merged +``` + +### Issues → validated +A new issue starts unlabeled. A maintainer triages it and, if it's a real, +in-scope problem, applies the **`accepted`** label. **Only `accepted` issues are +open for a contributor PR.** This prevents the "I fixed an issue you hadn't +agreed was a problem" rejection. Want to fix something? Get the issue accepted +first, or pick one already labelled `accepted` / `help wanted`. + +### Discussions → RFCs → accepted +Ideas and feature requests start in **Discussions**. Anyone — including external +contributors — may then **author an RFC** by opening a pull request that adds +`docs/rfcs/NNNN-title.md` (see [docs/rfcs/README.md](docs/rfcs/README.md)). The +RFC is reviewed as code; **a maintainer merging it is the act of acceptance** +(it becomes the durable decision record). Implementation PRs then reference the +accepted RFC. + +Authoring an RFC is open to everyone; **accepting one is a maintainer +decision.** Maintainers may also decline an RFC, with rationale, by closing it. + +### Pull requests → sanctioned +A contributor PR must do one of: +1. link a maintainer-**`accepted`** issue it fixes, or +2. be (or reference) an **accepted RFC**, or +3. qualify for the **trivial fast-lane**. + +**Trivial fast-lane** — these may be opened directly, no prior issue/RFC: +typo and wording fixes, documentation corrections, dependency bumps, comment +fixes, and obviously-correct one-line CI tweaks. When in doubt, open an Issue or +Discussion first; a PR that turns out to be non-trivial will be asked to. + +A substantive PR with no backing issue/RFC will be closed with a pointer to the +right channel — not as a judgment of the idea, but to keep design discussion +where it's reviewable. + +## What maintainers do *not* gate +Maintainers' own changes do not pass through the intake gates above — the team +runs a separate internal process. The universal gates (review, CODEOWNERS, +branch protection, CI) apply to everyone. Enforcement of the intake rules is, to +start, **by convention and review** (PR template + labels); an automated check +keyed to author association may be added later if volume warrants. + +## Code of conduct & security +- Conduct: [CODE_OF_CONDUCT.md](CODE_OF_CONDUCT.md). +- Security issues are **not** public Issues — see [SECURITY.md](SECURITY.md). + +## Changing this document +Governance changes the same way code does: a pull request, reviewed by +maintainers. This file describes the external surface; the internal maintainer +process is intentionally out of scope here. diff --git a/docs/dev/index.md b/docs/dev/index.md index 600c969..1e41342 100644 --- a/docs/dev/index.md +++ b/docs/dev/index.md @@ -51,6 +51,18 @@ constraints. User-facing behavior should still be documented through | Install and deployment packaging | [install.md](../user/install.md), [deployment.md](../user/deployment.md) | | Release history | [releases/](../releases/) | +## Contribution & Governance + +| Area | Read | +|---|---| +| How to contribute (external) | [CONTRIBUTING.md](../../CONTRIBUTING.md) | +| Governance model, roles, decision authority | [GOVERNANCE.md](../../GOVERNANCE.md) | +| Public contribution RFC track | [rfcs/](../rfcs/) | + +The `docs/rfcs/` track is the **public, externally-authorable** RFC process. The +maintainer/internal RFCs below (`rfc-00N-*.md`) are a separate, team-owned +track; don't conflate the two. + ## Active Implementation Plans Working documents for in-flight feature work. Removed when the work lands. diff --git a/docs/rfcs/0000-template.md b/docs/rfcs/0000-template.md new file mode 100644 index 0000000..48f4bda --- /dev/null +++ b/docs/rfcs/0000-template.md @@ -0,0 +1,54 @@ +# RFC NNNN: + +| | | +|---|---| +| **Status** | Proposed | +| **Author(s)** | <your name / handle> | +| **Discussion** | <link to the originating Discussion, if any> | +| **Implementation** | <issue/PR links, filled in as work lands> | + +> Status is maintained by maintainers: `Proposed` while the PR is open, +> `Accepted` on merge, `Declined` on close, `Superseded by NNNN` later. + +## Summary + +One paragraph: what this changes, in plain terms. + +## Motivation + +What problem does this solve, and why is it worth the ongoing cost? Tie it to a +concrete need (a Discussion, a recurring issue, a user request). Per the +project's first principle, argue the *long-run liability*, not just the +short-term convenience. + +## Guide-level explanation + +Explain the change as you'd teach it to a user or contributor: new commands, +syntax, API shapes, behavior. Examples first. + +## Reference-level design + +The precise design: data structures, IR/AST/planner changes, storage/format +impact, migration path, error behavior. Enough that a reviewer can find the +holes. + +## Invariants & deny-list check + +Which Hard Invariants in [../dev/invariants.md](../dev/invariants.md) does this +touch? Does it brush against any deny-list item — and if so, why is this the +justified exception? State explicitly that no invariant is weakened, or which +Known Gap moves. + +## Drawbacks & alternatives + +What does this cost, what did you reject, and why. "Do nothing" is a valid +alternative to weigh. + +## Reversibility + +Is this reversible? On-disk/wire/format and substrate choices are near-permanent +and demand more evidence; a CLI flag or doc is cheap to undo. Say which this is. + +## Unresolved questions + +What's deliberately left open for review to settle. diff --git a/docs/rfcs/README.md b/docs/rfcs/README.md new file mode 100644 index 0000000..99cdd76 --- /dev/null +++ b/docs/rfcs/README.md @@ -0,0 +1,66 @@ +# RFCs + +Substantial changes to OmniGraph — new user-facing surface, format or protocol +changes, anything irreversible or cross-cutting — go through a lightweight RFC +so the design is agreed *as reviewable code* before implementation starts. This +is the public RFC track, open to **anyone, including external contributors**. + +This complements the always-on review bar in +[../dev/invariants.md](../dev/invariants.md): the invariants say *what every +change must respect*; an RFC says *why this particular change is worth making and +how*. + +> **Two tracks, don't conflate them.** This `docs/rfcs/` directory is the +> **public contribution** track (anyone authors; maintainers accept). The +> maintainer-internal RFCs under `docs/dev/rfc-00N-*.md` are a separate, +> team-owned track for in-flight internal work. If you're an outside +> contributor, you're in the right place here. + +## When you need one + +- **RFC required:** new query/schema/CLI/HTTP surface; on-disk or wire-format + changes; a new substrate dependency; anything the deny-list in + [../dev/invariants.md](../dev/invariants.md) flags; anything irreversible + ("reversibility shapes evidence demand"). +- **RFC not required:** bug fixes for an `accepted` issue, and the trivial + fast-lane (typos, docs, deps) — see [../../CONTRIBUTING.md](../../CONTRIBUTING.md). + +If you're unsure, start a [Discussion](../../../discussions); a maintainer will +tell you whether it needs an RFC. + +## Lifecycle + +``` +Discussion (incubate, get rough consensus) + │ graduate + ▼ +RFC pull request → adds docs/rfcs/NNNN-title.md (Status: Proposed) + │ +maintainer review ──▶ changes requested / declined (PR closed, with rationale) + │ + ▼ +merged == Accepted (the merged file is the durable decision record) + │ + ▼ +Implementation PR(s) reference the accepted RFC +``` + +- **Author:** anyone. **Acceptance:** a maintainer decision, performed by + merging the RFC PR. Declining is closing it with rationale. +- The merged RFC *is* the accepted record — there is no separate sign-off step. +- Later reversals don't edit history: supersede with a new RFC that links back + and flip the old one's `Status` to `Superseded`. + +## Numbering & naming + +- File: `docs/rfcs/NNNN-kebab-title.md`, where `NNNN` is the next free + zero-padded integer (`0001`, `0002`, …). `0000-template.md` is reserved. +- Pick the number when you open the PR; if it collides with another in-flight + RFC, the second to merge bumps theirs. + +## Status values + +`Proposed` (open PR) · `Accepted` (merged) · `Declined` (closed) · +`Superseded by NNNN` · `Implemented` (set once the work lands, optional). + +Copy [0000-template.md](0000-template.md) to start. diff --git a/scripts/check-agents-md.sh b/scripts/check-agents-md.sh index abc6469..02a177a 100755 --- a/scripts/check-agents-md.sh +++ b/scripts/check-agents-md.sh @@ -34,10 +34,15 @@ PY canonical=() while IFS= read -r line; do canonical+=("$line") -done < <(find docs -type f -name '*.md' ! -path 'docs/releases/*' ! -path 'docs/internal/*' | sort) +done < <(find docs -type f -name '*.md' ! -path 'docs/releases/*' ! -path 'docs/internal/*' ! -path 'docs/rfcs/*' | sort) if [[ -d docs/releases ]]; then canonical+=("docs/releases/") fi +# RFCs are a growing collection (like releases): represent the directory, not +# every per-RFC file. The dir must be linked from an audience index. +if [[ -d docs/rfcs ]]; then + canonical+=("docs/rfcs/") +fi linked=() for index_file in "${index_files[@]}"; do From fd8e078a77fcce8be31b3ec3c18614427555b6fe Mon Sep 17 00:00:00 2001 From: Andrew Altshuler <andrew@collectivelab.io> Date: Sun, 7 Jun 2026 18:05:01 +0300 Subject: [PATCH 09/11] ci(codeowners): add aaltshuler to engineering role (#147) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Restores aaltshuler as an `engineering` code-owner (removed in #142), so `crates/**` and repo-infra PRs have a second reviewer besides the sole owner ragnorc — unblocking review of author-ragnorc PRs (e.g. #132) that ragnorc cannot self-approve. Edited the source of truth (.github/codeowners-roles.yml) and re-rendered .github/CODEOWNERS + the docs/dev/codeowners.md tables via .github/scripts/render-codeowners.py, per the documented flow. Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --- .github/CODEOWNERS | 4 ++-- .github/codeowners-roles.yml | 1 + docs/dev/codeowners.md | 6 +++--- 3 files changed, 6 insertions(+), 5 deletions(-) diff --git a/.github/CODEOWNERS b/.github/CODEOWNERS index d4ecfa5..e937724 100644 --- a/.github/CODEOWNERS +++ b/.github/CODEOWNERS @@ -8,9 +8,9 @@ # CI fails if this file drifts from its source, and rejects PRs that # edit this file directly without also editing the yml. -* @ragnorc +* @ragnorc @aaltshuler -crates/** @ragnorc +crates/** @ragnorc @aaltshuler docs/** @ragnorc README.md @ragnorc AGENTS.md @ragnorc diff --git a/.github/codeowners-roles.yml b/.github/codeowners-roles.yml index c5e36a9..ce4014d 100644 --- a/.github/codeowners-roles.yml +++ b/.github/codeowners-roles.yml @@ -22,6 +22,7 @@ roles: compiler. members: - ragnorc + - aaltshuler docs: description: > diff --git a/docs/dev/codeowners.md b/docs/dev/codeowners.md index 14bba0b..50c4dc7 100644 --- a/docs/dev/codeowners.md +++ b/docs/dev/codeowners.md @@ -14,8 +14,8 @@ The tables below are **generated** from `.github/codeowners-roles.yml` by `.gith | Path | Owners | Role(s) | |---|---|---| -| `*` | @ragnorc | engineering | -| `crates/**` | @ragnorc | engineering | +| `*` | @ragnorc @aaltshuler | engineering | +| `crates/**` | @ragnorc @aaltshuler | engineering | | `docs/**` | @ragnorc | docs | | `README.md` | @ragnorc | docs | | `AGENTS.md` | @ragnorc | docs | @@ -26,7 +26,7 @@ The tables below are **generated** from `.github/codeowners-roles.yml` by `.gith | Role | Members | Description | |---|---|---| -| `engineering` | @ragnorc | All production code under crates/**. Engine, CLI, server, compiler. | +| `engineering` | @ragnorc @aaltshuler | All production code under crates/**. Engine, CLI, server, compiler. | | `docs` | @ragnorc | Documentation under docs/**, plus repo-level docs (README.md, AGENTS.md, CLAUDE.md symlink, SECURITY.md). | <!-- END GENERATED OWNERSHIP --> From 54842808dbd981e61e0a4be2cf987fc0a52b2584 Mon Sep 17 00:00:00 2001 From: Ragnor Comerford <ragnor.comerford@gmail.com> Date: Sun, 7 Jun 2026 17:33:14 +0200 Subject: [PATCH 10/11] feat(engine): sweep & remove legacy __run__ branch guard (MR-770) (#132) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * feat(engine): sweep legacy __run__ branches via v2→v3 manifest migration Pre-v0.4.0 graphs can carry stale `__run__<id>` staging branches on the `__manifest` dataset, left by the Run state machine removed in MR-771. Lance's `list_branches` still enumerates them, so they leak into `branch_list()` and count as blocking branches at schema-apply time. Add a one-time `migrate_v2_to_v3` arm to the internal-schema dispatcher: on the first read-write open it enumerates `__manifest` branches, deletes every `__run__*` ref, and bumps the stamp to 3. Idempotent under retry (re-enumerates fresh each run). The `"__run__"` prefix is inlined so the migration does not depend on the run_registry guard that MR-770 removes next. This is the prerequisite sweep; the guard removal follows in the next commit. * refactor(engine): remove the legacy __run__ branch guard (MR-770) With the v2→v3 migration sweeping stale `__run__*` branches off `__manifest` on first read-write open, the defense-in-depth `is_internal_run_branch` guard is no longer needed. - delete `db/run_registry.rs`; drop the module + re-export from `db/mod.rs` - collapse `is_internal_system_branch` to the schema-apply-lock check only - `ensure_public_branch_ref`: drop the run-ref rejection; `__run__*` is now an ordinary branch name - `branch_merge`: reject `is_internal_system_branch` (was run-only) so the schema-apply lock is rejected consistently with create/delete — a small, deliberate tightening - update the inline schema-apply test + the writes integration tests (`public_branch_apis_reject_internal_run_refs` → `public_branch_apis_reject_internal_system_refs`, which also asserts `__run__*` now creates successfully) - docs: flip the "pending production sweep / defense-in-depth" notes to "auto-swept by the v2→v3 migration"; document the read-only-open limitation Known residual: the inert `_graph_runs.lance` / `_graph_run_actors.lance` bytes remain until a `StorageAdapter::delete_prefix` primitive lands. * fix(engine): run __run__ sweep at Omnigraph::open, not only on publish Review (PR #132) caught a regression: removing __run__ from `is_internal_system_branch` exposed legacy `__run__*` branches to the schema-apply blocking-branch checks (schema_apply.rs:104 and :778) and to `branch_list()`, but the v2→v3 sweep ran only inside the publisher's `load_publish_state`. On a pre-v0.4.0 graph whose first write is a schema apply, the blocking-branch check fires before any publish, so apply failed with "found non-main branches: __run__…". The same lazy timing also created a reverse hazard: a user-created `__run__*` branch on a still-v2 graph could be deleted by the first publish's sweep. Fix: run the internal-schema migration in `Omnigraph::open(ReadWrite)` (new `manifest::migrate_on_open`), before the coordinator reads branch state. The sweep now lands before any branch-observing code, and a graph is stamped v3 at open — so the one-time sweep can never catch a legitimately-created branch. Both checks and `branch_list` see the swept graph; correct by construction for every write path. Accepted residual: a read-only open of an unmigrated legacy graph still lists `__run__*` (read-only opens must not write, so they can't sweep). Documented. Regression test `legacy_run_branch_is_swept_on_open_and_does_not_block_schema_apply` confirmed RED before the fix (panicked on the branch_list leak assertion) and GREEN after. Also updates the stale schema_apply.rs comment, the writes.md "Migration code" section, and adds the v3 row to storage.md's migration table. * test(engine): sweep multiple legacy __run__ branches; doc nit Strengthen the v2→v3 migration test to synthesize three `__run__*` branches (a real legacy graph accumulates one per run) so the migration's delete loop is exercised on a single reused dataset handle, not just a single branch. Confirms multi-branch deletion is safe. Also drop a stale "active runs" reference from the branch_delete doc line. * fix(engine): force-delete in __run__ sweep for concurrency safety `migrate_v2_to_v3` ran `Dataset::delete_branch` (= `branches().delete(.., false)`), which errors "BranchContents not found" if the branch is already gone. Since the sweep now runs in `Omnigraph::open(ReadWrite)`, two processes opening the same legacy v2 graph concurrently would race: one wins each delete, the other's open fails. The migration only claimed idempotency under *sequential* retry. Switch to `Dataset::force_delete_branch` (= `delete(.., true)`), Lance's documented path for cleaning up zombie branches, which tolerates an already-absent branch. The sweep is now idempotent under concurrent runners and robust to partial/zombie state. Found in self-review; no behavior change for the common single-open path. * docs(release): note MR-770 __run__ cleanup in v0.6.1 * docs(branches): reconcile branch cleanup semantics --- crates/omnigraph/src/db/manifest.rs | 16 ++++ .../omnigraph/src/db/manifest/migrations.rs | 55 ++++++++++++- crates/omnigraph/src/db/manifest/tests.rs | 74 +++++++++++++++++ crates/omnigraph/src/db/mod.rs | 7 +- crates/omnigraph/src/db/omnigraph.rs | 81 +++++++++++++++---- .../src/db/omnigraph/schema_apply.rs | 10 +-- crates/omnigraph/src/db/run_registry.rs | 16 ---- crates/omnigraph/src/exec/merge.rs | 4 +- crates/omnigraph/src/exec/mod.rs | 2 +- crates/omnigraph/tests/writes.rs | 33 ++++---- docs/dev/writes.md | 18 +++-- docs/releases/v0.6.1.md | 2 + docs/user/audit.md | 2 +- docs/user/branches-commits.md | 10 +-- docs/user/constants.md | 6 +- docs/user/storage.md | 5 +- 16 files changed, 269 insertions(+), 72 deletions(-) delete mode 100644 crates/omnigraph/src/db/run_registry.rs diff --git a/crates/omnigraph/src/db/manifest.rs b/crates/omnigraph/src/db/manifest.rs index 7fcf7de..3b2886f 100644 --- a/crates/omnigraph/src/db/manifest.rs +++ b/crates/omnigraph/src/db/manifest.rs @@ -48,6 +48,22 @@ const OBJECT_TYPE_TABLE_VERSION: &str = "table_version"; const OBJECT_TYPE_TABLE_TOMBSTONE: &str = "table_tombstone"; const TABLE_VERSION_MANAGEMENT_KEY: &str = "table_version_management"; +/// Apply pending internal-schema migrations against `__manifest` on the +/// open-for-write path, independent of a publish. +/// +/// `Omnigraph::open(ReadWrite)` calls this before the coordinator reads branch +/// state, so branch-observing code (`branch_list`, the schema-apply +/// blocking-branch checks) sees the post-migration graph. In particular the +/// v2→v3 step sweeps legacy `__run__*` staging branches off `__manifest` +/// (MR-770); running it here closes the window where those branches would +/// otherwise block schema apply before the first publish runs the migration. +/// +/// Idempotent: a no-op stamp read when the on-disk version already matches. +pub(crate) async fn migrate_on_open(root_uri: &str) -> Result<()> { + let mut dataset = open_manifest_dataset(root_uri, None).await?; + migrations::migrate_internal_schema(&mut dataset).await +} + /// Immutable point-in-time view of the database. /// /// Cheap to create (no storage I/O). All reads within a query go through one diff --git a/crates/omnigraph/src/db/manifest/migrations.rs b/crates/omnigraph/src/db/manifest/migrations.rs index bbb7995..e2801fe 100644 --- a/crates/omnigraph/src/db/manifest/migrations.rs +++ b/crates/omnigraph/src/db/manifest/migrations.rs @@ -46,7 +46,11 @@ use crate::error::{OmniError, Result}; /// - v2 — `__manifest.object_id` carries the unenforced-PK annotation, /// engaging Lance's bloom-filter conflict resolver at commit time. Added /// alongside `expected_table_versions` OCC on `ManifestBatchPublisher::publish`. -pub(super) const INTERNAL_MANIFEST_SCHEMA_VERSION: u32 = 2; +/// - v3 — one-time sweep of legacy `__run__<id>` staging branches left on the +/// `__manifest` dataset by the pre-v0.4.0 Run state machine (removed in +/// MR-771). Once swept, the `is_internal_run_branch` defense-in-depth guard +/// is no longer needed (MR-770). +pub(super) const INTERNAL_MANIFEST_SCHEMA_VERSION: u32 = 3; const INTERNAL_SCHEMA_VERSION_KEY: &str = "omnigraph:internal_schema_version"; const OBJECT_ID_PK_KEY: &str = "lance-schema:unenforced-primary-key"; @@ -89,6 +93,10 @@ pub(super) async fn migrate_internal_schema(dataset: &mut Dataset) -> Result<()> migrate_v1_to_v2(dataset).await?; current = 2; } + 2 => { + migrate_v2_to_v3(dataset).await?; + current = 3; + } other => { return Err(OmniError::manifest_internal(format!( "no internal-schema migration registered for v{} → v{}", @@ -122,6 +130,51 @@ async fn migrate_v1_to_v2(dataset: &mut Dataset) -> Result<()> { set_stamp(dataset, 2).await } +/// v2 → v3: sweep legacy `__run__<id>` staging branches off the `__manifest` +/// dataset, then bump the stamp. +/// +/// The pre-v0.4.0 Run state machine (removed in MR-771) created graph-level +/// staging branches named `__run__<ulid>` on `__manifest`. MR-771 stopped +/// creating them but left any pre-existing ones in place; Lance's +/// `list_branches` still enumerates them, so they leak into `branch_list()` +/// and count as blocking branches at schema-apply time. This one-time sweep +/// removes them so the `is_internal_run_branch` guard can retire (MR-770). +/// +/// The `"__run__"` prefix is inlined here on purpose: this migration must keep +/// working after the `run_registry` module (the guard) is deleted, so it does +/// not depend on it. +/// +/// Idempotent under both sequential retry and concurrent runners: each run +/// re-enumerates `list_branches` fresh, and `force_delete_branch` tolerates a +/// branch that is already gone — so a crash before the stamp bump, or a second +/// process opening the same legacy graph at the same time, never errors out. +async fn migrate_v2_to_v3(dataset: &mut Dataset) -> Result<()> { + const LEGACY_RUN_BRANCH_PREFIX: &str = "__run__"; + let branches = dataset + .list_branches() + .await + .map_err(|e| OmniError::Lance(e.to_string()))?; + let run_branches: Vec<String> = branches + .into_keys() + .filter(|name| { + name.trim_start_matches('/') + .starts_with(LEGACY_RUN_BRANCH_PREFIX) + }) + .collect(); + for name in run_branches { + // `force_delete_branch` deletes even when the `BranchContents` is + // already gone. Plain `delete_branch` errors "BranchContents not + // found", which would fail a second concurrent open (or a retry that + // raced another runner) after the first one swept the branch. Force is + // exactly Lance's documented path for cleaning up zombie branches. + dataset + .force_delete_branch(&name) + .await + .map_err(|e| OmniError::Lance(e.to_string()))?; + } + set_stamp(dataset, 3).await +} + async fn set_stamp(dataset: &mut Dataset, version: u32) -> Result<()> { dataset .update_schema_metadata([(INTERNAL_SCHEMA_VERSION_KEY.to_string(), version.to_string())]) diff --git a/crates/omnigraph/src/db/manifest/tests.rs b/crates/omnigraph/src/db/manifest/tests.rs index effa0b5..885a2a8 100644 --- a/crates/omnigraph/src/db/manifest/tests.rs +++ b/crates/omnigraph/src/db/manifest/tests.rs @@ -1461,6 +1461,80 @@ async fn test_publish_migrates_pre_stamp_manifest_to_current_version() { assert!(reopened.snapshot().entry("node:Person").is_some()); } +#[tokio::test] +async fn test_v2_to_v3_sweeps_legacy_run_branches_on_write_open() { + let dir = tempfile::tempdir().unwrap(); + let uri = dir.path().to_str().unwrap(); + let catalog = build_test_catalog(); + let mut mc = ManifestCoordinator::init(uri, &catalog).await.unwrap(); + + // Synthesize a pre-MR-770 graph: several stale `__run__` staging branches + // left on `__manifest` (a real legacy graph accumulates one per run), plus + // a real user branch that must survive the sweep. Multiple run branches + // exercise the migration's delete loop on a single reused dataset handle. + mc.create_branch("__run__01J9LEGACY").await.unwrap(); + mc.create_branch("__run__01J9SECOND").await.unwrap(); + mc.create_branch("__run__01J9THIRD").await.unwrap(); + mc.create_branch("feature").await.unwrap(); + let before = mc.list_branches().await.unwrap(); + assert_eq!( + before.iter().filter(|b| b.starts_with("__run__")).count(), + 3, + "precondition: three legacy run branches exist on __manifest; got {before:?}", + ); + + // Rewind the internal-schema stamp to v2 so the next write-open runs the + // v2 → v3 sweep arm (init stamps at the current version, which is past it). + { + let mut ds = open_manifest_dataset(uri, None).await.unwrap(); + ds.update_schema_metadata([( + "omnigraph:internal_schema_version".to_string(), + Some("2".to_string()), + )]) + .await + .unwrap(); + let post = open_manifest_dataset(uri, None).await.unwrap(); + assert_eq!(super::migrations::read_stamp(&post), 2, "stamp rewound to v2"); + } + + // A no-op publish forces the open-for-write path, which runs the migration. + let mut expected = HashMap::new(); + expected.insert("node:Person".to_string(), 1); + GraphNamespacePublisher::new(uri, None) + .publish(&[], &expected) + .await + .unwrap(); + + // Stamp advanced to current; the legacy run branch is physically gone from + // `__manifest` (checked via the raw, unfiltered manifest list — not the + // guard-filtered `branch_list`), and the real branch + `main` survive. + let post = open_manifest_dataset(uri, None).await.unwrap(); + assert_eq!( + super::migrations::read_stamp(&post), + super::migrations::INTERNAL_MANIFEST_SCHEMA_VERSION, + ); + let reopened = ManifestCoordinator::open(uri).await.unwrap(); + let after = reopened.list_branches().await.unwrap(); + assert!( + !after.iter().any(|b| b.starts_with("__run__")), + "legacy run branch must be swept; got {after:?}", + ); + assert!(after.iter().any(|b| b == "feature"), "user branch must survive"); + assert!(after.iter().any(|b| b == "main"), "main must survive"); + + // Idempotent: a second write-open finds the stamp at current and does not + // re-run the sweep or error. + GraphNamespacePublisher::new(uri, None) + .publish(&[], &expected) + .await + .unwrap(); + let final_ds = open_manifest_dataset(uri, None).await.unwrap(); + assert_eq!( + super::migrations::read_stamp(&final_ds), + super::migrations::INTERNAL_MANIFEST_SCHEMA_VERSION, + ); +} + #[tokio::test] async fn test_publish_rejects_manifest_stamped_at_future_version() { let dir = tempfile::tempdir().unwrap(); diff --git a/crates/omnigraph/src/db/mod.rs b/crates/omnigraph/src/db/mod.rs index 8702f88..13e1c74 100644 --- a/crates/omnigraph/src/db/mod.rs +++ b/crates/omnigraph/src/db/mod.rs @@ -3,7 +3,6 @@ pub mod graph_coordinator; pub mod manifest; mod omnigraph; mod recovery_audit; -mod run_registry; mod schema_state; pub(crate) mod write_queue; @@ -15,7 +14,6 @@ pub use omnigraph::{ CleanupPolicyOptions, InitOptions, MergeOutcome, Omnigraph, OpenMode, SchemaApplyOptions, SchemaApplyResult, SkipReason, TableCleanupStats, TableOptimizeStats, }; -pub(crate) use run_registry::is_internal_run_branch; pub(crate) const SCHEMA_APPLY_LOCK_BRANCH: &str = "__schema_apply_lock__"; @@ -69,5 +67,8 @@ pub(crate) fn is_schema_apply_lock_branch(name: &str) -> bool { } pub(crate) fn is_internal_system_branch(name: &str) -> bool { - is_internal_run_branch(name) || is_schema_apply_lock_branch(name) + // Legacy `__run__*` staging branches (Run state machine, removed MR-771) + // are swept off `__manifest` by the v2→v3 internal-schema migration, so the + // only internal branch the engine still creates is the schema-apply lock. + is_schema_apply_lock_branch(name) } diff --git a/crates/omnigraph/src/db/omnigraph.rs b/crates/omnigraph/src/db/omnigraph.rs index 7b8a3f6..ba2b70e 100644 --- a/crates/omnigraph/src/db/omnigraph.rs +++ b/crates/omnigraph/src/db/omnigraph.rs @@ -346,6 +346,16 @@ impl Omnigraph { mode: OpenMode, ) -> Result<Self> { let root = normalize_root_uri(uri)?; + // Apply pending internal-schema migrations before the coordinator reads + // branch state, so `branch_list` and the schema-apply blocking-branch + // checks observe the post-migration graph — notably the v2→v3 sweep of + // legacy `__run__*` staging branches (MR-770). ReadWrite only: a + // read-only open must not trigger object-store writes, so a read-only + // open of an unmigrated legacy graph still lists `__run__*` until its + // first read-write open (an accepted, documented limitation). + if matches!(mode, OpenMode::ReadWrite) { + crate::db::manifest::migrate_on_open(&root).await?; + } // Open the coordinator first so the schema-staging recovery sweep can // compare its snapshot against any leftover staging files. let mut coordinator = GraphCoordinator::open(&root, Arc::clone(&storage)).await?; @@ -1491,12 +1501,6 @@ pub(crate) fn normalize_branch_name(branch: &str) -> Result<Option<String>> { } pub(crate) fn ensure_public_branch_ref(branch: &str, operation: &str) -> Result<()> { - if super::is_internal_run_branch(branch) { - return Err(OmniError::manifest(format!( - "{} does not allow internal run ref '{}'", - operation, branch - ))); - } if is_internal_system_branch(branch) { return Err(OmniError::manifest(format!( "{} does not allow internal system ref '{}'", @@ -1900,7 +1904,6 @@ fn json_value_from_array(array: &dyn Array, row: usize) -> Result<serde_json::Va #[cfg(test)] mod tests { use super::*; - use crate::db::is_internal_run_branch; use crate::db::manifest::ManifestCoordinator; use async_trait::async_trait; use serde_json::Value; @@ -2238,11 +2241,11 @@ edge WorksAt: Person -> Company #[tokio::test] async fn test_apply_schema_succeeds_after_load() { // Historical: schema apply used to be blocked by leftover - // `__run__` branches. A defense-in-depth filter now skips - // internal system branches, and run branches were made - // ephemeral on every terminal state — so in practice no - // `__run__` branch survives publish. The filter still guards - // the invariant. + // `__run__` branches. The Run state machine was removed in + // MR-771, so a fresh graph never creates a `__run__` branch; + // legacy ones are swept by the v2→v3 manifest migration. This + // asserts the invariant a current graph upholds: publish leaves + // no `__run__` branch behind, so schema apply proceeds. let dir = tempfile::tempdir().unwrap(); let uri = dir.path().to_str().unwrap(); let mut db = Omnigraph::init(uri, TEST_SCHEMA).await.unwrap(); @@ -2257,8 +2260,8 @@ edge WorksAt: Person -> Company let all_branches = db.coordinator.read().await.all_branches().await.unwrap(); assert!( - !all_branches.iter().any(|b| is_internal_run_branch(b)), - "run branch should be deleted after publish, got: {:?}", + !all_branches.iter().any(|b| b.starts_with("__run__")), + "no __run__ branch should exist after publish, got: {:?}", all_branches ); @@ -2270,6 +2273,56 @@ edge WorksAt: Person -> Company assert!(result.applied, "schema apply should have applied"); } + /// Regression (MR-770): a pre-v0.4.0 graph that still carries a stale + /// `__run__*` branch on `__manifest` must not block schema apply. The + /// v2→v3 sweep runs in `Omnigraph::open(ReadWrite)` — before the + /// schema-apply blocking-branch check — so apply succeeds with no + /// intervening publish. + /// + /// Confirmed to fail before the open-time migration landed: the reopened + /// graph still listed `__run__legacy`, and `apply_schema` returned + /// "found non-main branches: __run__legacy". + #[tokio::test] + async fn legacy_run_branch_is_swept_on_open_and_does_not_block_schema_apply() { + let dir = tempfile::tempdir().unwrap(); + let uri = dir.path().to_str().unwrap(); + let mut db = Omnigraph::init(uri, TEST_SCHEMA).await.unwrap(); + + // Synthesize a legacy graph: a stale `__run__` branch on `__manifest` + // plus the manifest stamp rewound to v2 (pre-sweep). + db.branch_create("__run__legacy").await.unwrap(); + drop(db); + { + let mut ds = lance::Dataset::open(&format!("{}/__manifest", uri)) + .await + .unwrap(); + ds.update_schema_metadata([( + "omnigraph:internal_schema_version".to_string(), + Some("2".to_string()), + )]) + .await + .unwrap(); + } + + // Reopen (ReadWrite): the open-time migration must sweep `__run__legacy` + // before any branch-observing code runs. + let db = Omnigraph::open(uri).await.unwrap(); + let branches = db.branch_list().await.unwrap(); + assert!( + !branches.iter().any(|b| b.starts_with("__run__")), + "open-time migration must sweep legacy __run__ branches; got {branches:?}", + ); + + // Schema apply must proceed with no intervening publish — the + // blocking-branch check no longer sees `__run__legacy`. + let desired = TEST_SCHEMA.replace( + " age: I32?\n}", + " age: I32?\n nickname: String?\n}", + ); + let result = db.apply_schema(&desired).await.unwrap(); + assert!(result.applied, "schema apply should have applied"); + } + #[tokio::test] async fn test_apply_schema_adds_index_for_existing_property() { let dir = tempfile::tempdir().unwrap(); diff --git a/crates/omnigraph/src/db/omnigraph/schema_apply.rs b/crates/omnigraph/src/db/omnigraph/schema_apply.rs index 35fe161..7cb3193 100644 --- a/crates/omnigraph/src/db/omnigraph/schema_apply.rs +++ b/crates/omnigraph/src/db/omnigraph/schema_apply.rs @@ -61,11 +61,11 @@ async fn plan_schema_for_apply( ) -> Result<PlannedSchemaApply> { db.ensure_schema_state_valid().await?; let branches = db.coordinator.read().await.all_branches().await?; - // Skip `main` and internal system branches. The schema-apply lock branch - // is excluded because it is the cluster-wide schema-apply serializer. - // `__run__*` branches are no longer created; the filter remains as - // defense-in-depth for legacy graphs with leftover staging branches. - // A future production sweep will let this guard go. + // Skip `main` and internal system branches (the schema-apply lock branch, + // the cluster-wide schema-apply serializer). Legacy `__run__*` staging + // branches were swept off `__manifest` by the v2→v3 migration that runs in + // `Omnigraph::open(ReadWrite)` before this check (MR-770), so they no + // longer appear here. let blocking_branches = branches .into_iter() .filter(|branch| branch != "main" && !is_internal_system_branch(branch)) diff --git a/crates/omnigraph/src/db/run_registry.rs b/crates/omnigraph/src/db/run_registry.rs deleted file mode 100644 index ee3d336..0000000 --- a/crates/omnigraph/src/db/run_registry.rs +++ /dev/null @@ -1,16 +0,0 @@ -// The Run state machine has been removed. Mutations now write directly -// to target tables and use the publisher's `expected_table_versions` -// CAS for cross-table OCC; `__run__<id>` staging branches and the -// `_graph_runs.lance` state machine no longer exist. -// -// What remains is the branch-name predicate, kept as a defense-in-depth -// guard against users naming a public branch `__run__*`. A future -// production sweep of legacy `_graph_runs.lance` rows and stale -// `__run__*` branches will let this predicate (and this file) go too. - -pub(crate) const INTERNAL_RUN_BRANCH_PREFIX: &str = "__run__"; - -pub(crate) fn is_internal_run_branch(name: &str) -> bool { - name.trim_start_matches('/') - .starts_with(INTERNAL_RUN_BRANCH_PREFIX) -} diff --git a/crates/omnigraph/src/exec/merge.rs b/crates/omnigraph/src/exec/merge.rs index 2e5f32e..eb6c4a3 100644 --- a/crates/omnigraph/src/exec/merge.rs +++ b/crates/omnigraph/src/exec/merge.rs @@ -1087,9 +1087,9 @@ impl Omnigraph { target: &str, actor_id: Option<&str>, ) -> Result<MergeOutcome> { - if is_internal_run_branch(source) || is_internal_run_branch(target) { + if is_internal_system_branch(source) || is_internal_system_branch(target) { return Err(OmniError::manifest(format!( - "branch_merge does not allow internal run refs ('{}' -> '{}')", + "branch_merge does not allow internal system refs ('{}' -> '{}')", source, target ))); } diff --git a/crates/omnigraph/src/exec/mod.rs b/crates/omnigraph/src/exec/mod.rs index 33a7e41..ce72d42 100644 --- a/crates/omnigraph/src/exec/mod.rs +++ b/crates/omnigraph/src/exec/mod.rs @@ -35,7 +35,7 @@ use time::format_description::well_known::Rfc3339; use crate::db::commit_graph::CommitGraph; use crate::db::manifest::ManifestCoordinator; -use crate::db::{MergeOutcome, Omnigraph, is_internal_run_branch}; +use crate::db::{MergeOutcome, Omnigraph, is_internal_system_branch}; use crate::db::{ReadTarget, Snapshot}; use crate::embedding::EmbeddingClient; use crate::error::{MergeConflict, MergeConflictKind, OmniError, Result}; diff --git a/crates/omnigraph/tests/writes.rs b/crates/omnigraph/tests/writes.rs index 13cb10f..0a309c9 100644 --- a/crates/omnigraph/tests/writes.rs +++ b/crates/omnigraph/tests/writes.rs @@ -371,11 +371,10 @@ async fn cancelled_mutation_future_leaves_no_state() { // Cancel-safety property: no graph-level run/staging state remains. // - // Note: `branch_list()` already filters `__run__*` via - // `is_internal_system_branch`, so a runtime "no `__run__` branches" check - // would be vacuous. The structural property that no `__run__` branches - // can ever be created is enforced by deletion of `begin_run` etc. in - // (verified by the build itself — those symbols no longer exist). + // No `__run__` branches can ever be created: the Run state machine + // (`begin_run` etc.) was deleted in MR-771 — verified by the build itself, + // those symbols no longer exist. Any legacy `__run__*` branch on an + // upgraded graph is swept by the v2→v3 manifest migration. // // (1) The branch list is unchanged: cancellation/completion cannot // synthesize new public branches. @@ -442,34 +441,40 @@ async fn repeated_loads_do_not_accumulate_branches() { assert_eq!(db.branch_list().await.unwrap(), vec!["main".to_string()]); } -/// User code must not be able to write to internal `__run__*` names. -/// The branch-name guard predicate is kept as defense-in-depth; it -/// will be removed once a future production sweep retires the legacy -/// branches. +/// After MR-770, `__run__*` is an ordinary branch name — the Run state machine +/// and its `is_internal_run_branch` guard are gone. The surviving internal-ref +/// guard still rejects the active `__schema_apply_lock__` branch on the public +/// create/merge APIs. #[tokio::test] -async fn public_branch_apis_reject_internal_run_refs() { +async fn public_branch_apis_reject_internal_system_refs() { let dir = tempfile::tempdir().unwrap(); let mut db = init_and_load(&dir).await; - let create_err = db.branch_create("__run__synthetic").await.unwrap_err(); + // `__run__*` is no longer reserved — creating it now succeeds. + db.branch_create("__run__formerly_reserved") + .await + .expect("__run__ prefix is a normal branch name post-MR-770"); + + // The schema-apply lock branch is still rejected on public branch APIs. + let create_err = db.branch_create("__schema_apply_lock__").await.unwrap_err(); let OmniError::Manifest(err) = create_err else { panic!("expected Manifest error"); }; assert!( - err.message.contains("internal run ref"), + err.message.contains("internal system ref"), "unexpected error: {}", err.message ); let merge_err = db - .branch_merge("__run__synthetic", "main") + .branch_merge("__schema_apply_lock__", "main") .await .unwrap_err(); let OmniError::Manifest(err) = merge_err else { panic!("expected Manifest error"); }; assert!( - err.message.contains("internal run refs"), + err.message.contains("internal system refs"), "unexpected error: {}", err.message ); diff --git a/docs/dev/writes.md b/docs/dev/writes.md index 974f7a6..8b692b4 100644 --- a/docs/dev/writes.md +++ b/docs/dev/writes.md @@ -14,8 +14,11 @@ publisher's row-level CAS on `__manifest` is the single fence. - No `RunRecord`, no `_graph_runs.lance`, no `_graph_run_actors.lance`. - No `omnigraph run *` CLI subcommands and no `/runs/*` HTTP endpoints. -- No `__run__<id>` staging branches. (Legacy on-disk artifacts from - pre-MR-771 repos are inert; MR-770 sweeps them in production.) +- No `__run__<id>` staging branches; `__run__*` is no longer a reserved + name. The branch-name guard was removed in MR-770, and any stale + `__run__*` branch on an upgraded graph is swept off `__manifest` by the + v2→v3 internal-schema migration on first read-write open. (The inert + `_graph_runs.lance` bytes remain until a `delete_prefix` primitive lands.) - Cancelled mutation futures leave **no graph-level state** — only orphaned Lance fragments, which the existing `omnigraph cleanup` pipe reclaims. @@ -245,9 +248,14 @@ list`. ## Migration code -`db/manifest/migrations.rs` does not change. Active deletion of -`_graph_runs.lance` belongs in MR-770 (the production sweep) — this PR -stops *creating* run state but does not destroy legacy bytes on disk. +`db/manifest/migrations.rs` carries the v2→v3 internal-schema step (MR-770): +a one-time sweep that deletes legacy `__run__*` staging branches off +`__manifest`. It runs in `Omnigraph::open(ReadWrite)` (via +`manifest::migrate_on_open`, before the coordinator reads branch state) and +again on the publisher's write path; both are idempotent once the stamp is at +v3. Deleting the inert `_graph_runs.lance` / `_graph_run_actors.lance` dataset +*bytes* is still deferred — it needs a `StorageAdapter::delete_prefix` +primitive — but those bytes are invisible to graph-level state. ## Mid-query partial failure: closed by MR-794 diff --git a/docs/releases/v0.6.1.md b/docs/releases/v0.6.1.md index aafe1af..0acc34b 100644 --- a/docs/releases/v0.6.1.md +++ b/docs/releases/v0.6.1.md @@ -7,6 +7,7 @@ v0.6.1 focuses on operational polish after v0.6.0: stored-query registries, safe - **Stored-query registries.** `omnigraph.yaml` can declare curated `queries:` blocks per graph. Servers load and type-check them at startup, `omnigraph queries validate` checks them offline, `omnigraph queries list` shows exposed queries and typed params, `GET /queries` exposes a typed catalog, and `POST /queries/{name}` invokes a stored query without accepting ad hoc `.gq` source from the client. - **Stored-query policy gate.** New Cedar action `invoke_query` gates the stored-query invocation surface. Stored mutations are double-gated: `invoke_query` to reach the stored query and `change` for the actual write. - **Safer branch deletion.** `branch_delete` now treats the manifest as the authority, flips branch visibility atomically, and reclaims per-table/commit-graph forks as derived state. If best-effort reclaim is interrupted, `cleanup` reconciles orphaned forks; reusing a branch name before cleanup reports an actionable error. +- **Legacy `__run__` cleanup (MR-770).** Removed the last functional remnant of the Run state machine (retired in v0.4.0): the `__run__` branch-name guard. A new v2→v3 `__manifest` internal-schema migration sweeps any stale `__run__*` staging branches on the first read-write open, so `__run__*` is no longer a reserved branch name. This closes the "unpromoted `__run__` branches block reads" condition behind the zombie-run cascade incident; the inert `_graph_runs.lance` row cleanup is tracked separately (it needs a `delete_prefix` primitive). - **Blob-safe optimize.** `omnigraph optimize` skips tables with `Blob` properties instead of failing the whole sweep on Lance's blob-v2 compaction decode bug. Skips are visible in human output, `--json` as `skipped`, `TableOptimizeStats.skipped`, and logs; non-blob tables still compact normally. - **Deployment improvements.** The container entrypoint now composes `OMNIGRAPH_TARGET_URI` with `OMNIGRAPH_CONFIG`, so operators can keep the graph URI in env while loading policy/query config from a mounted file. The local RustFS bootstrap pins RustFS beta.3 and allows the current insecure local-dev default credentials. - **Windows release support.** Tagged and edge releases now publish Windows x86_64 archives containing `omnigraph.exe` and `omnigraph-server.exe`, with a PowerShell installer and Windows install docs. @@ -17,6 +18,7 @@ v0.6.1 focuses on operational polish after v0.6.0: stored-query registries, safe - A graph selected by name (`--target` or `server.graph`) now uses `graphs.<name>.policy` and `graphs.<name>.queries`. Top-level `policy` / `queries` blocks are only for anonymous bare-URI single-graph mode; using them with a named graph now fails loudly with migration guidance. - `mcp.expose` defaults to `true` for stored-query registry entries. Set `mcp: { expose: false }` for service-only queries that should not appear in the catalog. - `invoke_query` is graph-scoped, not branch-scoped. Branch/snapshot access remains enforced by the inner `read` / `change` gate. +- **Legacy `__run__` migration.** Graphs created before v0.4.0 are migrated automatically on the first **read-write** open by a v0.6.1 binary (one-time `__manifest` stamp v2→v3 sweep of stale `__run__*` branches). No action required. Two caveats: (1) a graph opened **read-only** still lists any stale `__run__*` branch until its first read-write open, since the migration is write-path-only like all manifest migrations — long-lived read-only deployments should be opened read-write once after upgrading; (2) the inert `_graph_runs.lance` / `_graph_run_actors.lance` dataset bytes are left in place until a future `delete_prefix` primitive (they are invisible to graph-level state). - Blob tables are not compacted until the upstream Lance fix lands, so fragment count and deleted-row space on blob tables are not reclaimed by `optimize`. Reads, writes, and query results are unaffected; no on-disk migration is required. - `TableOptimizeStats` is now `#[non_exhaustive]` and gains a `skipped: Option<SkipReason>` field (so does the new `SkipReason` enum). This is a source-level change only for downstream code that built this returned result struct by literal — rare, since it is produced by `optimize` and consumed by reading its fields; field access is unaffected, and `#[non_exhaustive]` keeps future additions non-breaking. diff --git a/docs/user/audit.md b/docs/user/audit.md index e8abe5b..ab028ac 100644 --- a/docs/user/audit.md +++ b/docs/user/audit.md @@ -4,4 +4,4 @@ - `_as` variants of every write API let callers override the actor: `mutate_as`, `ingest_as`, `branch_merge_as`, `apply_schema_as`, etc. - Actor IDs are persisted on `GraphCommit.actor_id` with split storage in `_graph_commit_actors.lance` (the commit graph is split into `_graph_commits.lance` for the linkage and `_graph_commit_actors.lance` for the actor map). - HTTP server uses the bearer-token actor automatically; CLI uses the local user / explicit env (no implicit actor). -- Pre-v0.4.0 graphs also stored actor IDs on `RunRecord.actor_id` in `_graph_runs.lance` / `_graph_run_actors.lance`. The Run state machine was removed in MR-771; those files are inert post-v0.4.0 and reclaimed by MR-770's production sweep. +- Pre-v0.4.0 graphs also stored actor IDs on `RunRecord.actor_id` in `_graph_runs.lance` / `_graph_run_actors.lance`. The Run state machine was removed in MR-771; those files are inert post-v0.4.0. The v2→v3 manifest migration sweeps any stale `__run__*` branches on first write-open (MR-770); the inert dataset bytes remain until a `delete_prefix` primitive lands. diff --git a/docs/user/branches-commits.md b/docs/user/branches-commits.md index c1894f9..0565186 100644 --- a/docs/user/branches-commits.md +++ b/docs/user/branches-commits.md @@ -9,8 +9,8 @@ Lance supports branching at the dataset level: a branch is a named lineage of ve OmniGraph builds *graph branches* on top by branching every sub-table coherently: - `branch_create(name)` / `branch_create_from(target, name)` — disallowed name `main`; fails if branch exists; ensures the schema-apply lock is idle. Atomic and authority-first like `branch_delete`: it flips the `__manifest` branch (authority), then creates the derived commit-graph branch, force-dropping any orphaned commit-graph ref left by an incomplete prior delete (the manifest branch is fresh, so a same-named commit-graph branch is provably a zombie). If commit-graph creation fails, the manifest branch is rolled back so the name never half-exists. -- `branch_list()` — returns public branches, **filters internal** `__run__…` and `__schema_apply_lock__` prefixes. -- `branch_delete(name)` — refuses if there are descendants or active runs on the branch. The manifest is the single authority for branch existence: deletion flips the `__manifest` branch ref first (one atomic op), after which the branch is gone from every snapshot. The owned per-table forks and the commit-graph branch are derived state, reclaimed best-effort with `force_delete_branch` after the flip. A failure during that reclaim (transient object-store error) does not fail the call or block the authority flip; the leftover forks are unreachable orphans that the [`cleanup`](maintenance.md) reconciler converges. One consequence: if a delete's best-effort reclaim fails, reusing that branch name before the next `cleanup` surfaces a clear error pointing at `cleanup` (the stale fork would otherwise collide on first write). +- `branch_list()` — returns public branches, **filters the internal** `__schema_apply_lock__` branch. +- `branch_delete(name)` — refuses if there are descendants on the branch, or if it is the current branch. The manifest is the single authority for branch existence: deletion flips the `__manifest` branch ref first (one atomic op), after which the branch is gone from every snapshot. The owned per-table forks and the commit-graph branch are derived state, reclaimed best-effort with `force_delete_branch` after the flip. A failure during that reclaim (transient object-store error) does not fail the call or block the authority flip; the leftover forks are unreachable orphans that the [`cleanup`](maintenance.md) reconciler converges. One consequence: if a delete's best-effort reclaim fails, reusing that branch name before the next `cleanup` surfaces a clear error pointing at `cleanup` (the stale fork would otherwise collide on first write). - **Lazy forking**: a branch only forks a sub-table when that sub-table is first mutated on it. Pure-read branches share fragments with their source. A fork collision is classified by the manifest authority, not by Lance branch versions: if the live manifest already records the fork on the active branch, a concurrent first-write won and the caller gets a retryable "refresh and retry"; if the manifest does not, a physical branch there is an orphan and the caller is pointed at `cleanup`. - `sync_branch(branch)` — re-binds the in-memory handle to the latest head of the branch. @@ -51,10 +51,10 @@ Notes: ## L2 — Internal system branches -Filtered from `branch_list()` but visible to internals: +Internal or legacy branch refs: -- `__schema_apply_lock__` — serializes schema migrations. -- `__run__<run-id>` — legacy from the pre-v0.4.0 Run state machine (removed in MR-771). The branch-name guard predicate `is_internal_run_branch` is kept as defense-in-depth so users cannot create a branch matching the legacy prefix; the filter will be removed once production legacy branches are swept (MR-770). +- `__schema_apply_lock__` — serializes schema migrations; filtered from `branch_list()` but visible to internals. +- `__run__<run-id>` — legacy from the pre-v0.4.0 Run state machine (removed in MR-771). These are swept off `__manifest` on the first read-write open by the v2→v3 internal-schema migration (MR-770), and `__run__*` is no longer a reserved name. Known limitation: a pre-v0.4.0 graph opened **read-only** still surfaces any stale `__run__*` branch in `branch_list()` until its first read-write open (the migration is write-path-only, like all manifest migrations). ## L2 — Recovery audit trail diff --git a/docs/user/constants.md b/docs/user/constants.md index 8f13555..210155e 100644 --- a/docs/user/constants.md +++ b/docs/user/constants.md @@ -4,11 +4,11 @@ |---|---|---| | `MANIFEST_DIR` | `__manifest` | `db/manifest/layout.rs` | | Commit graph dir | `_graph_commits.lance` | `db/commit_graph.rs` | -| Run registry dir (legacy, removed MR-771) | `_graph_runs.lance` | inert post-v0.4.0; reclaimed by MR-770 | -| Run branch prefix (legacy, removed MR-771) | `__run__` | filtered by `is_internal_run_branch` defense-in-depth | +| Run registry dir (legacy, removed MR-771) | `_graph_runs.lance` | inert post-v0.4.0; bytes remain until a `delete_prefix` primitive lands | +| Run branch prefix (legacy, removed MR-771/MR-770) | `__run__` | swept off `__manifest` by the v2→v3 migration; no longer a reserved name | | Schema apply lock | `__schema_apply_lock__` | `db/mod.rs` | | Manifest publisher retry budget | `PUBLISHER_RETRY_BUDGET = 5` | `db/manifest/publisher.rs` | -| Internal manifest schema version | `INTERNAL_MANIFEST_SCHEMA_VERSION = 2` | `db/manifest/migrations.rs` | +| Internal manifest schema version | `INTERNAL_MANIFEST_SCHEMA_VERSION = 3` | `db/manifest/migrations.rs` | | Merge stage batch | `MERGE_STAGE_BATCH_ROWS = 8192` | `exec/merge.rs` | | Maintenance concurrency | `OMNIGRAPH_MAINTENANCE_CONCURRENCY=8` | `db/omnigraph/optimize.rs` | | Lance blob compaction support | `LANCE_SUPPORTS_BLOB_COMPACTION = false` | `db/omnigraph/optimize.rs` | diff --git a/docs/user/storage.md b/docs/user/storage.md index c22d4d6..d1c52b5 100644 --- a/docs/user/storage.md +++ b/docs/user/storage.md @@ -22,7 +22,7 @@ OmniGraph is **not** a single Lance dataset; it is a *graph* of datasets coordin - `edges/{fnv1a64-hex(edge_type_name)}` — one Lance dataset per edge type - `__manifest/` — the catalog of all sub-tables and their published versions - `_graph_commits.lance` / `_graph_commit_actors.lance` — the commit graph and its actor map - - (legacy `_graph_runs.lance` / `_graph_run_actors.lance` from pre-v0.4.0 graphs are inert; the run state machine was removed in MR-771 and these files are cleaned up via MR-770's production sweep) + - (legacy `_graph_runs.lance` / `_graph_run_actors.lance` from pre-v0.4.0 graphs are inert; the run state machine was removed in MR-771. The v2→v3 manifest migration sweeps stale `__run__*` branches on first write-open; the inert dataset bytes themselves remain until a `delete_prefix` storage primitive lands) - **Manifest row schema** (`object_id, object_type, location, metadata, base_objects, table_key, table_version, table_branch, row_count`): - `object_type` ∈ `table | table_version | table_tombstone` - `table_key` ∈ `node:<TypeName> | edge:<EdgeName>` @@ -47,6 +47,7 @@ Adding a new on-disk shape change is one constant bump (`INTERNAL_MANIFEST_SCHEM |---|---| | v1 (implicit, pre-stamp) | `__manifest.object_id` had no PK annotation; publisher had no row-level CAS protection. | | v2 | `__manifest.object_id` carries `lance-schema:unenforced-primary-key=true`; row-level CAS engaged. Stamped as `omnigraph:internal_schema_version=2`. | +| v3 | One-time sweep of legacy `__run__*` staging branches (pre-v0.4.0 Run state machine, removed MR-771) off `__manifest`. Runs at `Omnigraph::open(ReadWrite)` and on publish. Stamped as `omnigraph:internal_schema_version=3`. | ## On-disk layout @@ -91,7 +92,7 @@ flowchart TB - **Graph root** is one directory (or S3 prefix). Everything below is part of one OmniGraph graph. - **`__manifest/`** is a Lance dataset whose rows describe which sub-table version is published at which graph-branch. Reading a snapshot starts here. - **`nodes/`** and **`edges/`** are sibling directories holding one Lance dataset per declared type. Names are `fnv1a64-hex` of the type name to keep paths fixed-length and case-safe. -- **`_graph_commits.lance`** is an L2 dataset that records the graph-level commit DAG, with a paired `_graph_commit_actors.lance` for the actor map. (Pre-v0.4.0 graphs also have inert `_graph_runs.lance` / `_graph_run_actors.lance` from the removed Run state machine; MR-770 sweeps these in production.) +- **`_graph_commits.lance`** is an L2 dataset that records the graph-level commit DAG, with a paired `_graph_commit_actors.lance` for the actor map. (Pre-v0.4.0 graphs also have inert `_graph_runs.lance` / `_graph_run_actors.lance` from the removed Run state machine; the v2→v3 migration sweeps their stale `__run__*` branches, and the dataset bytes are reclaimed once `delete_prefix` lands.) - **`_graph_commit_recoveries.lance`** — one row per recovery sweep action. Joined to `_graph_commits.lance` by `graph_commit_id`; the linked commit row carries `actor_id=omnigraph:recovery`. Operators correlate recoveries with the original mutations they rolled forward / back via this join. See `crates/omnigraph/src/db/recovery_audit.rs`. - **`__recovery/{ulid}.json`** — transient sidecar files written by the four migrated writers (`MutationStaging::finalize`, `schema_apply`, `branch_merge`, `ensure_indices`) before Phase B begins, deleted after Phase C succeeds. A sidecar persisting after process exit means the writer crashed in the Phase B → Phase C window; the next `Omnigraph::open` recovery sweep processes it. Steady-state directory is empty. See `crates/omnigraph/src/db/manifest/recovery.rs`. - **`_refs/branches/{name}.json`** is graph-level branch metadata — pointers from a branch name to the manifest version it heads. From 4a66d6e071ce95eabe49cd496356fba0617901be Mon Sep 17 00:00:00 2001 From: Aaron Goh <aaronwgoh5@gmail.com> Date: Sun, 7 Jun 2026 20:37:37 +0200 Subject: [PATCH 11/11] fix(loader): accept multi-line (pretty-printed) JSON in load (#146) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The loader read input line-by-line (reader.lines() + serde_json::from_str per line), so any delta where a JSON object spanned multiple lines failed with 'invalid JSON on line 1: EOF while parsing an object'. Compact JSONL worked; pretty-printed JSON never did. Switch to a streaming value deserializer (Deserializer::from_reader().into_iter::<Value>()), which treats any whitespace (including newlines inside objects) as a separator — so both compact JSONL and pretty-printed JSON load. Error labels switch from line numbers to record numbers (line numbers are meaningless once objects span lines). Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: Ragnor Comerford <ragnor.comerford@gmail.com> --- crates/omnigraph/src/loader/mod.rs | 35 ++++++++++++++++-------------- 1 file changed, 19 insertions(+), 16 deletions(-) diff --git a/crates/omnigraph/src/loader/mod.rs b/crates/omnigraph/src/loader/mod.rs index 46a46e2..d5d74c0 100644 --- a/crates/omnigraph/src/loader/mod.rs +++ b/crates/omnigraph/src/loader/mod.rs @@ -288,21 +288,24 @@ async fn load_jsonl_reader<R: BufRead>( let mut node_rows: HashMap<String, Vec<JsonValue>> = HashMap::new(); let mut edge_rows: HashMap<String, Vec<(String, String, JsonValue)>> = HashMap::new(); - for (line_num, line) in reader.lines().enumerate() { - let line = line?; - let line = line.trim(); - if line.is_empty() { - continue; - } - let value: JsonValue = serde_json::from_str(line).map_err(|e| { - OmniError::manifest(format!("invalid JSON on line {}: {}", line_num + 1, e)) + // Parse a stream of JSON values. Accepts both compact JSONL (one object + // per line) and pretty-printed JSON where a single object spans multiple + // lines — serde's streaming deserializer treats any whitespace (including + // newlines) between top-level values as a separator. + for (idx, parsed) in serde_json::Deserializer::from_reader(reader) + .into_iter::<JsonValue>() + .enumerate() + { + let record_num = idx + 1; + let value: JsonValue = parsed.map_err(|e| { + OmniError::manifest(format!("invalid JSON at record {}: {}", record_num, e)) })?; if let Some(type_name) = value.get("type").and_then(|v| v.as_str()) { if !catalog.node_types.contains_key(type_name) { return Err(OmniError::manifest(format!( - "line {}: unknown node type '{}'", - line_num + 1, + "record {}: unknown node type '{}'", + record_num, type_name ))); } @@ -317,8 +320,8 @@ async fn load_jsonl_reader<R: BufRead>( } else if let Some(edge_name) = value.get("edge").and_then(|v| v.as_str()) { if catalog.lookup_edge_by_name(edge_name).is_none() { return Err(OmniError::manifest(format!( - "line {}: unknown edge type '{}'", - line_num + 1, + "record {}: unknown edge type '{}'", + record_num, edge_name ))); } @@ -326,14 +329,14 @@ async fn load_jsonl_reader<R: BufRead>( .get("from") .and_then(|v| v.as_str()) .ok_or_else(|| { - OmniError::manifest(format!("line {}: edge missing 'from'", line_num + 1)) + OmniError::manifest(format!("record {}: edge missing 'from'", record_num)) })? .to_string(); let to = value .get("to") .and_then(|v| v.as_str()) .ok_or_else(|| { - OmniError::manifest(format!("line {}: edge missing 'to'", line_num + 1)) + OmniError::manifest(format!("record {}: edge missing 'to'", record_num)) })? .to_string(); let data = value @@ -347,8 +350,8 @@ async fn load_jsonl_reader<R: BufRead>( .push((from, to, data)); } else { return Err(OmniError::manifest(format!( - "line {}: expected 'type' or 'edge' field", - line_num + 1 + "record {}: expected 'type' or 'edge' field", + record_num ))); } }