server.rs:212 was documenting v2.0.4's state (23 tools) but the
handle_tools_list test two functions below asserts 24 tools since
v2.0.5 (when `suppress` landed). Align the comment with the assertion
and point at the assertion as the source of truth so future tool
additions can't introduce comment drift without also tripping a test
failure.
Both parameters were advertised in the MCP tool schemas since v1.7+ but
were silently ignored at runtime — a schema-contract violation. Any
caller that set them got unfiltered results with no error or warning,
which is the worst possible failure mode for a public tool surface.
changelog.rs
- Parse `start` / `end` as ISO-8601/RFC-3339 timestamps; return an
explicit error on malformed input (previously: silent drop).
- In system-wide mode, over-fetch 4× limit when a time window is set,
then apply an inclusive [start, end] filter in Rust before the
sort+truncate. SQL-level filtering is a v2.1+ optimisation.
- Response JSON gains a `filter` field echoing the applied bounds so
callers can confirm the window was honored.
- Per-memory mode still ignores the window (semantically meaningless
when scoped to one memory's transition history).
intention_unified.rs
- `execute_check`: when `include_snoozed=true`, fold snoozed
intentions back into the check pool so their time/context triggers
can wake them when a matching condition appears. Previously
snoozed intentions were invisible to check regardless of the arg.
- Deduplicates defensively via a HashSet on intention.id in case
storage ever returns overlap.
Tests: 9 changelog + 37 intention_unified tests continue to pass.
Full vestige-mcp lib suite 419 passing, 0 failures.
All four PredictiveMemory calls (predict_needed_memories, get_proactive_suggestions,
get_top_interests, prediction_accuracy) return Result, and all four were being
swallowed by `.unwrap_or_default()` / `.unwrap_or(0.0)` — every lock-poisoning or
internal error produced a response indistinguishable from a genuine cold-start
"I have no predictions yet." Callers (dashboard, Claude Code, Cursor) had no way
to tell "the system is broken" from "there genuinely isn't anything to predict."
Now each call uses `unwrap_or_else` to (a) `tracing::warn!` the error with its
source channel for observability, (b) flip a local `degraded` flag. The JSON
response gains a new `predict_degraded: bool` field. Empty + degraded=false =
cold start (expected). Empty + degraded=true = something went wrong, check logs.
6 existing predict tests pass (return shape unchanged on success path).
Both tables were added speculatively in V4 and never received a single
INSERT or SELECT in the codebase. `knowledge_edges` has an elaborate
bi-temporal schema (valid_from, valid_until, confidence) and was marked
DEPRECATED in the same V4 migration that created it — the real edge
table is `memory_connections` (V3). `compressed_memories` was a tiered-
compression feature (compression_ratio, semantic_fidelity, model_used)
but `advanced/compression.rs` operates entirely in-memory and never
touches it.
Both tables are verified single-file references (only migrations.rs).
A grep across crates/ shows zero row reads or writes. Safe to drop
without behaviour change; frees schema space for future migrations.
The MCP `instructions` field is injected into every connecting client's
system prompt on Initialize — Claude Code, Cursor, Zed, Windsurf, and
anyone else running Vestige. That reach demands a default that serves
every user, not the project maintainer's personal workflow.
Default ("minimal", 3 sentences) tells the client how to use Vestige
and how to respond to explicit feedback signals. It is safe for every
audience: competitive coders, hobbyists saving recipes, Rails devs
saving bug fixes, enterprise deployments under system-prompt review.
The full composition mandate — Composing / Never-composed /
Recommendation shape + FSRS-trust blocking phrase + origin case study
— is load-bearing for decision-adjacent work but misfires on trivial
retrievals ("what's my favorite color"). Opt in on your own machine
with `VESTIGE_SYSTEM_PROMPT_MODE=full`; four hundred strangers do not
inherit one maintainer's trauma scar on every session.
Extracted into build_instructions() so the branch is a single env-var
check on Initialize, not a compile-time switch. main.rs --help output
advertises the new variable alongside VESTIGE_DASHBOARD_ENABLED.
Two unrelated upstream issues were stopping two of our four release targets.
Root-caused and fixed both so v2.0.5 ships on 100% of supported platforms.
Windows MSVC (new regression in v2.0.5):
`usearch 2.24.0` introduced a `memory_mapping_allocator_gt` template
that references the POSIX `MAP_FAILED` macro from <sys/mman.h>, which
doesn't exist on MSVC. Confirmed upstream as unum-cloud/usearch#746
(open). The bump from 2.23.0 happened during the v2.0.5 Cargo.lock
refresh. Pinned `usearch = "=2.23.0"` in crates/vestige-core/Cargo.toml
with a comment linking the upstream issue. Unpin when the fix lands.
Intel Mac (latent bug exposed by the ci workaround):
Root cause was feature-propagation, not the release workflow.
crates/vestige-mcp/Cargo.toml hardcoded
`features = ["bundled-sqlite", "embeddings", "vector-search"]` on its
vestige-core dep, which forcibly enabled vestige-core's embeddings
feature regardless of whether vestige-mcp's own `embeddings` feature
flag was set. With `--no-default-features` at the top level (the old
Intel Mac ci workaround), vestige-mcp's feature flags turned off but
vestige-core's embeddings stayed on through the hardcoded list. That
pulled in fastembed -> ort-sys, but without any of the flags that
select ort-sys' backend binaries, so the ort-sys build script failed
with "does not provide prebuilt binaries for the target
x86_64-apple-darwin with feature set (no features)".
Fix:
- Drop `embeddings` and `vector-search` from the hardcoded features
list in crates/vestige-mcp/Cargo.toml. Leave only `bundled-sqlite`
as an always-on base feature. The existing
`embeddings = ["vestige-core/embeddings"]` /
`vector-search = ["vestige-core/vector-search"]` flag declarations
now actually gate those features as intended.
- Bump the vestige-core dep version ref 2.0.4 -> 2.0.5 (was stale).
- Drop `cargo_flags: "--no-default-features"` from the Intel Mac
target in .github/workflows/release.yml. The original reason for
that workaround was sidestepping the same ort-sys issue, but with
the feature-propagation bug fixed, Intel Mac now builds with full
default features the same way aarch64-darwin does on the same
macos-14 runner.
Verification:
- `cargo tree -p vestige-mcp --no-default-features -i fastembed`
-> "did not match any packages" (fastembed truly absent now)
- `cargo tree -p vestige-mcp --no-default-features -i ort-sys` -> same
- `cargo build --release -p vestige-mcp` -> clean, 1m 21s, usearch 2.23.0
Same v2.0.5 tag. Rust source code identical to 8178beb. Re-triggering
the release workflow via workflow_dispatch will rebuild all four
platforms and upload to the existing v2.0.5 release page.
First AI memory system to model forgetting as a neuroscience-grounded
PROCESS rather than passive decay. Adds the `suppress` MCP tool (#24),
Rac1 cascade worker, migration V10, and dashboard forgetting indicators.
Based on:
- Anderson, Hanslmayr & Quaegebeur (2025), Nat Rev Neurosci — right
lateral PFC as the domain-general inhibitory controller; SIF
compounds with each stopping attempt.
- Cervantes-Sandoval et al. (2020), Front Cell Neurosci PMC7477079 —
Rac1 GTPase as the active synaptic destabilization mechanism.
What's new:
* `suppress` MCP tool — each call compounds `suppression_count` and
subtracts a `0.15 × count` penalty (saturating at 80%) from
retrieval scores during hybrid search. Distinct from delete
(removes) and demote (one-shot).
* Rac1 cascade worker — background sweep piggybacks the 6h
consolidation loop, walks `memory_connections` edges from
recently-suppressed seeds, applies attenuated FSRS decay to
co-activated neighbors. You don't just forget Jake — you fade
the café, the roommate, the birthday.
* 24h labile window — reversible via `suppress({id, reverse: true})`
within 24 hours. Matches Nader reconsolidation semantics.
* Migration V10 — additive-only (`suppression_count`, `suppressed_at`
+ partial indices). All v2.0.x DBs upgrade seamlessly on first launch.
* Dashboard: `ForgettingIndicator.svelte` pulses when suppressions
are active. 3D graph nodes dim to 20% opacity when suppressed.
New WebSocket events: `MemorySuppressed`, `MemoryUnsuppressed`,
`Rac1CascadeSwept`. Heartbeat carries `suppressed_count`.
* Search pipeline: SIF penalty inserted into the accessibility stage
so it stacks on top of passive FSRS decay.
* Tool count bumped 23 → 24. Cognitive modules 29 → 30.
Memories persist — they are INHIBITED, not erased. `memory.get(id)`
returns full content through any number of suppressions. The 24h
labile window is a grace period for regret.
Also fixes issue #31 (dashboard graph view buggy) as a companion UI
bug discovered during the v2.0.5 audit cycle:
* Root cause: node glow `SpriteMaterial` had no `map`, so
`THREE.Sprite` rendered as a solid-coloured 1×1 plane. Additive
blending + `UnrealBloomPass(0.8, 0.4, 0.85)` amplified the square
edges into hard-edged glowing cubes.
* Fix: shared 128×128 radial-gradient `CanvasTexture` singleton used
as the sprite map. Retuned bloom to `(0.55, 0.6, 0.2)`. Halved fog
density (0.008 → 0.0035). Edges bumped from dark navy `0x4a4a7a`
to brand violet `0x8b5cf6` with higher opacity. Added explicit
`scene.background` and a 2000-point starfield for depth.
* 21 regression tests added in `ui-fixes.test.ts` locking every
invariant in (shared texture singleton, depthWrite:false, scale
×6, bloom magic numbers via source regex, starfield presence).
Tests: 1,284 Rust (+47) + 171 Vitest (+21) = 1,455 total, 0 failed
Clippy: clean across all targets, zero warnings
Release binary: 22.6MB, `cargo build --release -p vestige-mcp` green
Versions: workspace aligned at 2.0.5 across all 6 crates/packages
Closes#31
CI failed on macOS + Ubuntu with clippy::collapsible_if on the
else-if branch of the exclude_types filter. Collapse the inner
`if` into the `let Some && ...` guard. Semantics preserved — the
includes branch is left as-is to keep include/exclude mutually
exclusive behavior.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Exact keyword matches (e.g. unique proper nouns like "Valladares") were
buried by semantic scoring in hybrid search. With a small limit, the
exact match could be missing entirely from results.
Adds a dedicated Stage 0 keyword-only pass (keyword_weight=1.0,
semantic_weight=0.0) before the main hybrid search. Results with strong
keyword scores (>= 0.8) are collected and merged into the hybrid results
via dedup. In the reranker stage, these keyword-priority results bypass
the cross-encoder entirely and receive a 2x score boost to survive
downstream pipeline stages (temporal, FSRS, utility, competition).
Bug inherited from Vestige 2.0.1.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Type filtering (include_types/exclude_types) was applied post-fetch after
the database LIMIT, which could return zero results when all top-N
results were of the filtered type. This pushes type filters into the SQL
WHERE clause in keyword_search_with_scores() so they apply before the
limit. Semantic results still get post-fetch filtering as a safety net
since the vector index cannot filter by type.
Also adds hybrid_search_filtered() as the new primary method, with the
original hybrid_search() delegating to it with no filters for backward
compatibility. The MCP search tool now exposes include_types and
exclude_types parameters.
Includes 5 new test cases covering include, exclude, precedence,
empty results, and backward compatibility.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Dashboard now disabled by default to reduce memory footprint
- Enable with VESTIGE_DASHBOARD_ENABLED=true
- Update --help text to document new env var
The Stage 8 `recommended` selector and the evidence sort both rank by
FSRS-6 trust only, discarding the `combined_score` signal that the
upstream hybrid_search + cross-encoder reranker just computed. Confidence
is then derived from `recommended.trust + evidence_count`, neither of
which moves with the query — so any query against the same corpus
returns the same primary memory and the same confidence score.
Empirical reproduction (15 deep_reference probes against an 11-memory
corpus, 9 with a unique correct answer + 6 with no relevant memories):
- Distinct primary memories returned : 1 / 15
- Confidence values returned : 1 distinct (0.82 for all)
- Ground-truth accuracy on specific queries : 1 / 9 (11.1%)
The single hit is coincidental: the always-returned memory happened to
be the correct answer for one query. Random guessing across the 11-memory
corpus would be ~9% baseline, so the tool is performing at random.
Fix
---
Replace trust-only ranking at three sites with a 50/50 composite of
combined_score (query relevance) and FSRS-6 trust:
let composite = |s: &ScoredMemory| s.combined_score as f64 * 0.5 + s.trust * 0.5;
Used in:
- cross_reference.rs:573 — `recommended` max_by
- cross_reference.rs:589 — `non_superseded` evidence sort_by
- cross_reference.rs:622 — `base_confidence` formula
The 50/50 weighting is a design choice — see PR body for the knob to
tweak if a different blend is preferred. The pre-existing updated_at
tiebreaker is preserved.
Tests
-----
Two regression tests, both verified to FAIL on `main` and PASS with the
fix via negative control (temporarily set the composite weights to
1.0 trust + 0.0 relevance and confirmed both tests fail again):
- test_recommended_uses_query_relevance_not_just_trust
Two-memory corpus, ingested in order so the off-topic memory wins
the trust tiebreaker. Query targets the on-topic memory. The fix
ensures `recommended` is the on-topic one.
- test_confidence_varies_with_query_relevance
Single-memory corpus. Identical execute() calls with a relevant
query and an irrelevant query. The fix ensures the relevant
query produces higher confidence.
Full crate suite: 410 / 410 passing (was 408 + 2 new).
Out of scope
------------
While running the live MCP probes I observed two further inconsistencies
in `cross_reference.rs` that I cannot reproduce in cargo test (the
synthetic test environment with mock embeddings does not trigger the
required combined_score > 0.2 floor condition):
- The `effective_sim` floor at line 551 fabricates contradictions
between memories with no real topical overlap when one contains a
CORRECTION_SIGNALS keyword.
- The Stage 5 `contradictions` field (strict) and the Stage 7
`pair_relations` feeding the reasoning text (loose, post-floor)
disagree, producing responses where `reasoning` claims N
contradictions while `contradictions` is empty and `status` is
"resolved".
I have empirical data for both from live MCP usage but no reproducible
cargo test, so they are intentionally not addressed in this PR. Happy to
file them as a separate issue with the raw probe data if useful.
Collapsed nested if statements into single conditions using
let-chains (if a && let Ok(b) = ...). Fixes CI clippy failures
on both macOS and Ubuntu.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The public JSON schema in schema() declares `in_minutes` and `file_pattern`
in snake_case, but TriggerSpec uses `#[serde(rename_all = "camelCase")]`
which makes serde expect `inMinutes` / `filePattern`. Snake_case inputs are
silently dropped to None, so time-based intentions with `in_minutes` never
fire (triggerAt becomes null) and file_pattern-only context intentions
never match.
Added `#[serde(alias = ...)]` so both naming conventions deserialize
correctly — purely additive, existing camelCase callers unaffected.
Two regression tests added, verified to FAIL without the aliases
(negative control confirmed the snake_case duration test sees
`triggerAt: null` and the file_pattern test sees an empty `triggered`
array). Both pass with the fix. Full crate suite: 408/408 passing.
Related to #25 (Bug #8 was half-fixed — check-side re-derivation works,
but the set-side was still dropping the value before it could be persisted).
- Removed vestige-agent and vestige-agent-py from workspace members
(ARC-AGI-3 code, not part of Vestige release — caused CI failure)
- Improved deep_reference reasoning chain: fuller output with arrows on
supersession reasoning, longer primary finding preview, fallback message
when no relations found, boosted relation detection for search results
with high combined_score
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Handle 'vestige/memory://' and 'vestige/codebase://' URIs by stripping the
provider prefix before scheme matching. This fixes compatibility with
MCP clients like OpenCode that prepend the provider name to resource URIs.
Fixes#19
When memories are created, promoted, deleted, or dreamed via MCP tools,
the 3D graph now shows spectacular live animations:
- Rainbow particle burst + elastic scale-up on MemoryCreated
- Ripple wave cascading to nearby nodes
- Green pulse + node growth on MemoryPromoted
- Implosion + dissolution on MemoryDeleted
- Edge growth animation on ConnectionDiscovered
- Purple cascade on DreamStarted/DreamProgress/DreamCompleted
- FIFO eviction at 50 live nodes to guard performance
Also: graph center defaults to most-connected node, legacy HTML
redirects to SvelteKit dashboard, CSS height chain fix in layout.
Testing: 150 unit tests (vitest), 11 e2e tests (Playwright with
MCP Streamable HTTP client), 22 proof screenshots.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds a second transport layer alongside stdio — Streamable HTTP on port
3928. Enables Claude.ai, remote clients, and web integrations to connect
to Vestige over HTTP with per-session McpServer instances.
- POST /mcp (JSON-RPC) + DELETE /mcp (session cleanup)
- Bearer token auth with constant-time comparison (subtle crate)
- Auto-generated UUID v4 token persisted with 0o600 permissions
- Per-session McpServer instances with 30-min idle reaper
- 100 max sessions, 50 concurrency limit, 256KB body limit
- --http-port flag + VESTIGE_HTTP_PORT / VESTIGE_HTTP_BIND env vars
- Module exports moved from binary to lib.rs for reusability
- vestige CLI gains `serve` subcommand via shared lib
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
ort-sys v2.0.0-rc.11 has no prebuilt ONNX Runtime binaries for
x86_64-apple-darwin, and vestige-mcp requires embeddings to compile.
- Remove x86_64-apple-darwin from CI release matrix (discontinued 2020)
- Fix vestige-mcp Cargo.toml: add default-features=false to vestige-core dep
- Extract sanitize_fts5_query to always-available fts.rs module
- Gate embeddings-only imports in storage/sqlite.rs behind #[cfg]
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Dashboard v2.1 "Nuclear" upgrade:
- Dark glassmorphism UI system (4-tier glass utilities, ambient orbs, nav glow)
- Graph3D decomposed from 806-line monolith into 10 focused modules
- Custom GLSL shaders (nebula FBM background, chromatic aberration, film grain, vignette)
- Enhanced dream mode with smooth 2s lerped transitions and aurora cycling
- Cognitive pipeline visualizer (7-stage search cascade animation)
- Temporal playback slider (scrub through memory evolution over time)
- Bioluminescent color palette for node types and events
Fix flaky CI test on macOS:
- vector::tests::test_add_and_search used near-identical test vectors (additive phase shift)
- Changed to multiplicative frequency so each seed produces a distinct vector
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
explore_connections and memory_graph returned empty results because
in-memory cognitive modules were never loaded from the database.
Connections were persisting to SQLite correctly (795 in production)
but the query path only checked empty ActivationNetwork.
- Add CognitiveEngine::hydrate() to load connections at startup
- Add storage fallback in explore_connections associations
- Hydrate live engine after dream persists new connections
- Add error logging for save_connection failures
- Add 7 integration tests for the full round-trip
Closes#14
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Triggers 2 (force after 6h stale) and 3 (mini-consolidation after 2h)
fired immediately on fresh schedulers even when user was active, because
they didn't check activity state. Added MIN_BRIEF_IDLE_MINS (5 min)
guard so both triggers require a brief idle period before firing.
Fixes test_consolidation_idle_trigger in CI.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
3 new MCP tools (16 → 19 total):
- importance_score: 4-channel neuroscience importance scoring (novelty/arousal/reward/attention)
- session_checkpoint: batch smart_ingest up to 20 items with PE Gating
- find_duplicates: cosine similarity clustering with union-find for dedup
CLI: vestige ingest command for memory ingestion via command line
Core: made get_node_embedding public, added get_all_embeddings for dedup scanning
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add web dashboard (axum) on port 3927 with memory browser, search, and
system stats. New MCP tools: memory_timeline, memory_changelog,
health_check, consolidate, stats, backup, export, gc. Search now supports
detail_level (brief/summary/full) to control token usage. Add backup_to()
and get_recent_state_transitions() to storage layer. Bump to v1.2.0.
P0 fixes:
- Add `vestige backup <path>` — full DB copy with WAL checkpoint flush
- Add `vestige export --format json|jsonl [--tags] [--since] <path>` —
paginated memory export with tag/date filtering
- Add `vestige gc --min-retention 0.1 [--max-age-days] [--dry-run] [--yes]`
— bulk cleanup of stale memories with safety prompts
- Fix apply_decay() scaling: batched pagination (500 rows/batch) with
explicit transactions instead of loading all nodes into memory
- Fix hidden MCP resources: memory://insights and memory://consolidation-log
now listed in resources/list (were implemented but undiscoverable)
P1 fixes:
- Add auto-consolidation on server startup: FSRS-6 decay runs in background
after 2s delay, only if last consolidation was >6 hours ago
- Add encryption at rest via SQLCipher feature flag: use --features encryption
with VESTIGE_ENCRYPTION_KEY env var (bundled-sqlite and encryption are
mutually exclusive)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Route ingest tool through smart_ingest (Prediction Error Gating) to
prevent duplicate memories when content is similar to existing entries
- Fix Intel Mac release build: use macos-13 runner for x86_64-apple-darwin
(macos-latest is now ARM64, causing silent cross-compile failures)
- Sync npm package version to 1.1.2 (was 1.0.0 in package.json, 1.1.0
in postinstall.js BINARY_VERSION)
- Add vestige-restore to npm makeExecutable list
- Remove abandoned packages/core/ TypeScript package (pre-Rust implementation
referencing FSRS-5, chromadb, ollama — 32K lines of dead code)
- Sync workspace Cargo.toml version to 1.1.2
Closes#5
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Previously, fastembed created .fastembed_cache in the current working
directory, polluting project folders with symlinks.
Now uses platform-appropriate cache directories:
- macOS: ~/Library/Caches/com.vestige.core/fastembed
- Linux: ~/.cache/vestige/fastembed
- Windows: %LOCALAPPDATA%\vestige\cache\fastembed
Can still be overridden with FASTEMBED_CACHE_PATH env var.
Fixes user feedback about .fastembed_cache appearing in random folders.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add #![allow(dead_code)] to deprecated tool modules (kept for
backwards compatibility but not exposed in MCP tool list)
- Mark unused functions with #[allow(dead_code)] annotations
- Fix unused variable warnings (prefix with _)
- Apply clippy auto-fixes for redundant closures and derives
- Fix test to account for protocol version negotiation
- Reorganize tools/mod.rs to clarify active vs deprecated tools
Security review: LOW RISK - no critical vulnerabilities found
Dead code review: deprecated tools properly annotated
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Fix silent errors in stdio.rs: clients now receive fallback error
responses instead of hanging when JSON serialization fails
- Fix UTF-8 panics in keyword.rs: use char-aware slicing instead of
byte offsets for query sanitization and term highlighting
- Fix UTF-8 panics in prospective_memory.rs: replace hard-coded byte
offsets with char-aware slicing for natural language parsing
- Fix UTF-8 panics in git.rs: convert byte positions to char positions
before slicing commit messages
- Fix feature flag bug in vestige-mcp: add proper [features] section
to forward embeddings and vector-search features from vestige-core,
enabling the #[cfg(feature = "embeddings")] initialization code
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
When embeddings fail to initialize in the MCP server context (e.g., due
to working directory issues), the error was silently swallowed and
smart_ingest would fall back to regular ingest without explanation.
Changes:
- Add init_embeddings() method to Storage for explicit initialization
- Initialize embeddings at MCP server startup with error logging
- Add check_ready() method to EmbeddingService for error access
- Log warning when is_ready() returns false
Now users will see clear error messages like:
"Failed to initialize embedding service: ..."
"Hint: Check FASTEMBED_CACHE_PATH or ensure ~/.fastembed_cache exists"
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Claude Desktop requests protocol version 2025-06-18 but Vestige was
responding with 2025-11-25, causing Claude Desktop to disconnect.
Now the server negotiates: if the client requests an older version,
use it. This maintains backward compatibility with older clients.
Fixes: Claude Desktop "Server transport closed unexpectedly" error
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
