feat(connectors): add MongoDB connector (#305) (#310)

* refactor(connectors): split KtxDialect into core and KtxSqlDialect Separate the dialect contract into a driver-agnostic core (display/ref formatting and type mapping) and a SQL-only extension (query generators). The catalog and entity-details paths resolve the core dialect for any snapshot driver, so it must stay free of SQL generation; this is the prerequisite refactor for adding non-SQL primary sources. - KtxDialect keeps type, formatDisplayRef, parseDisplayRef, columnDisplayTablePartCount, mapDataType, mapToDimensionType - KtxSqlDialect extends it with quoteIdentifier, formatTableName, and the query/sample/statistics generators; the 7 SQL dialects implement it - add getSqlDialectForDriver for SQL drivers; the 7 connectors and the relationship-benchmark harness consume it - thread the relationship pipeline (profiling/validation/composite/ discovery) as KtxSqlDialect | null so a non-SQL source skips coverage SQL and its candidates stay in review; local-enrichment builds the SQL dialect only when the connector advertises readOnlySql Pure extraction: no behavior change for the existing 7 drivers. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(connectors): add MongoDB connector for issue #305 Add a read-only MongoDB connector that treats a database as a primary context source: collections map to tables and inferred top-level fields to columns. MongoDB is the first non-SQL source (readOnlySql: false), so ktx sql and metric compilation do not apply, but its collections flow through ingest, descriptions, and relationship discovery. - schema-inference: infer a flat column schema from the most recent sample_size documents (by _id desc, or order_by for non-ObjectId keys). Union BSON types per field, mark multi-type fields mixed (string), keep sub-documents/arrays as a single opaque json column, derive nullability from presence, treat _id as the primary key - connector: KtxMongoDbScanConnector behind an injectable client seam; strictly read-only (find/listCollections/estimatedDocumentCount only), no executeReadOnly; resolves env:/file: via resolveKtxConfigReference - core-only KtxMongoDbDialect and a live-database introspection adapter - wire the mongodb driver: driver union, dialect registry, driver registration (scopeConfigKey databases), mongodbConnectionSchema, connection-drivers, normalizeDriver, the live-database route, and the ktx setup picker. ktx sql is refused by the read-only SQL capability gate - tests: schema inference, connector snapshot via a fake client, dialect, driver-schema parsing, and the ktx sql rejection Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * docs(integrations): document the MongoDB primary source Add a MongoDB section to the primary-sources reference: connection config (url, databases, enabled_tables, sample_size, order_by), mongodb+srv/TLS/ Atlas notes, the schema-inference explainer, a features matrix, and the non-SQL caveat. Update the frontmatter and connection field reference. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * fix(connectors): address review blockers on the MongoDB connector - introspect: skip estimatedDocumentCount for views. The count command is rejected on a MongoDB view (CommandNotSupportedOnView), so counting a view aborted introspect for the whole connection; compute estimatedRows only for real collections, as ClickHouse does. - sl: refuse a semantic-layer query against a non-SQL connection instead of defaulting it to the Postgres dialect. compileLocalSlQuery (the shared CLI + MCP path) now rejects a driver with no SQL dialect via the new isSqlQueryableDriver authority, keeping MongoDB context-only per issue #305. - tests: cover input.tableScope and the empty-scope skip for the Mongo connector (the scan layer does not post-filter), the view no-count path, and the ktx sl query refusal for a mongodb connection. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * polish(mongodb): compute sampled nullCount and document sampling caveats Address the non-blocking review notes: - sampleColumn now counts null/absent values over the sampled window instead of returning nullCount: null, since the documents are already in hand - warn that a custom order_by must be indexed (an unindexed sort hits MongoDB's in-memory sort limit on large collections) in the connection schema and docs - note that sampled values for nested fields are stringified, not faithfully serialized, so the json opacity is deliberate Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * docs(examples): add a MongoDB connector example A manual, container-backed example mirroring examples/postgres-historic: - docker-compose.yml + init/seed.js seed a representative dataset (nested documents, arrays, a Decimal128, a mixed-type field, a nullable field, an ObjectId reference, and a view) on first container start - scripts/smoke.sh + introspect-smoke.mjs assert the connector's inferred schema with no LLM credentials — the same introspection entry point ktx ingest's database-schema stage uses, including the view-no-count path - README.md documents the smoke and a full keyless ktx ingest run (claude-code LLM + managed sentence-transformers embeddings) Works with Docker Compose or podman compose. Verified end to end. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * chore: ignore examples/** in knip to fix dead-code false positives The MongoDB connector example files (examples/mongodb/init/seed.js and examples/mongodb/scripts/introspect-smoke.mjs) are used at runtime but were flagged as unused by knip. Add examples/** to the ignore array, matching the existing .context/** entry. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0114qQV8fJ5a5ME3XbMVRzbL * fix(mongodb): refuse non-SQL connections before SQL analysis `ktx sql` and the MCP sql_execution tool resolved a SQL-analysis dialect (falling back to Postgres for a non-SQL driver) and ran read-only validation before the connector capability gate refused the connection. For a MongoDB connection that spun up the parser/daemon and produced Postgres parser diagnostics instead of a clean non-SQL refusal. Route both entry points through a shared assertSqlQueryableConnection guard before dialect selection, mirroring compileLocalSlQuery. The federated duckdb path has no driver and is exempted at each call site. Add CLI and MCP regression tests asserting validation/connector work never starts for a MongoDB connection. * fix(mongodb): pass CI gates (dialect boundary, secrets, setup test) Three latent failures in the connector surfaced once CI ran on the branch: - connector.ts imported the concrete KtxMongoDbDialect, which the connector dialect-import boundary forbids. Route it through getDialectForDriver('mongodb') and widen inferKtxMongoCollectionColumns to the base KtxDialect (it only uses mapDataType/mapToDimensionType). - detect-secrets flagged a test ObjectId hex and the mongodb+srv example URL; annotate both with allowlist pragmas. - the "shows every supported database" setup test omitted the new MongoDB option. --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: Luca Martial <48870843+luca-martial@users.noreply.github.com> Co-authored-by: Luca Martial <lucamrtl@gmail.com> Co-authored-by: Andrey Avtomonov <andreybavt@gmail.com>
2026-07-01 08:59:39 +02:00 · 2026-06-29 15:17:56 +02:00 · 2026-06-29 15:17:56 +02:00 · 2afab61417
commit 2afab61417
parent 4f084186f1
59 changed files with 1971 additions and 129 deletions
--- a/examples/mongodb/README.md
+++ b/examples/mongodb/README.md
@ -0,0 +1,127 @@
+# MongoDB Connector Example
+
+A manual, self-contained example for the **ktx** MongoDB connector. It starts a
+local MongoDB, seeds a representative dataset, and exercises the connector both
+as a fast no-LLM introspection smoke and as a full `ktx ingest` run.
+
+MongoDB is a **context-only** primary source: collections become tables and
+inferred top-level fields become columns, but `ktx sql` and semantic-layer
+metric compilation do not apply. See
+[`docs-site/content/docs/integrations/primary-sources.mdx`](../../docs-site/content/docs/integrations/primary-sources.mdx).
+
+## Prerequisites
+
+- Docker with Compose v2, or Podman with `podman compose`
+- Node and pnpm matching the **ktx** workspace
+- The built CLI: `pnpm --filter @kaelio/ktx run build`
+- For the full ingest only: `uv` on `PATH` and a usable local Claude Code
+  session (the keyless `claude-code` LLM backend)
+
+## What the seed contains
+
+[`init/seed.js`](init/seed.js) creates the `app` database with:
+
+- `users` — `_id` (ObjectId), scalar fields, a nested `address`, an array
+  `tags`, a `Decimal128` `balance`, a `ref` field that holds more than one type
+  (inferred `mixed`), and an `age` field absent from one document (nullable)
+- `orders` — an ObjectId `user_id` reference for relationship discovery
+- `active_users` — a **view** (to confirm introspection never runs a count
+  command on a view)
+
+MongoDB applies the script once on first container start. Apply it by hand with:
+
+```bash
+mongosh "mongodb://localhost:27117" < examples/mongodb/init/seed.js
+```
+
+## Smoke (no LLM credentials)
+
+From the **ktx** repository root:
+
+```bash
+examples/mongodb/scripts/smoke.sh
+```
+
+It starts MongoDB on `127.0.0.1:27117`, seeds it, and asserts the connector's
+inferred schema (collections → tables, nested → `json`, `mixed`, nullability,
+`_id` primary key, and a view introspected with `estimatedRows: null`). This
+drives the same entry point `ktx ingest`'s "database schema" stage uses, without
+needing an LLM or embeddings.
+
+Podman:
+
+```bash
+KTX_MONGODB_COMPOSE="podman compose" examples/mongodb/scripts/smoke.sh
+```
+
+Set `KTX_MONGODB_KEEP=1` to leave the container running after the script exits.
+
+## Full `ktx ingest`
+
+The public database-ingest path requires a configured model and embeddings.
+This runs entirely locally with the keyless `claude-code` LLM backend and the
+**ktx**-managed `sentence-transformers` embedding daemon — no API keys.
+
+Start MongoDB and create a project:
+
+```bash
+docker compose -f examples/mongodb/docker-compose.yml up -d --wait   # or: podman compose
+node packages/cli/dist/bin.js admin init /tmp/ktx-mongodb-example
+```
+
+Add the connection and a keyless enrichment stack to
+`/tmp/ktx-mongodb-example/ktx.yaml`:
+
+```yaml
+connections:
+  mongo-prod:
+    driver: mongodb
+    url: mongodb://localhost:27117/app
+    databases:
+      - app
+llm:
+  provider:
+    backend: claude-code
+  models:
+    default: sonnet
+scan:
+  enrichment:
+    mode: llm
+    embeddings:
+      backend: sentence-transformers
+      model: all-MiniLM-L6-v2
+      dimensions: 384
+      sentenceTransformers:
+        base_url: ""
+```
+
+Test the connection and ingest:
+
+```bash
+node packages/cli/dist/bin.js connection test mongo-prod --project-dir /tmp/ktx-mongodb-example
+node packages/cli/dist/bin.js ingest mongo-prod --project-dir /tmp/ktx-mongodb-example --yes --plain
+```
+
+The first ingest starts the **ktx** embedding daemon and downloads the
+`all-MiniLM-L6-v2` model. Expected final state: `Database schema: done`.
+
+Inspect the result:
+
+- `raw-sources/mongo-prod/live-database/<run>/tables/*.json` — one per
+  collection, including the `active_users` view with `estimatedRows: null`
+- `raw-sources/mongo-prod/live-database/<run>/enrichment/relationships.json` —
+  inferred relationships sit in `review` (a non-SQL source has no read-only SQL
+  coverage validation), with `accepted: []`
+- `semantic-layer/mongo-prod/_schema/app.yaml` — the schema with per-column AI
+  descriptions
+
+`ktx sql -c mongo-prod "SELECT 1"` is refused by the read-only SQL capability
+gate, and `ktx sl query -c mongo-prod ...` is refused because MongoDB is not a
+SQL source.
+
+## Cleanup
+
+```bash
+docker compose -f examples/mongodb/docker-compose.yml down -v   # or: podman compose
+rm -rf /tmp/ktx-mongodb-example
+```
--- a/examples/mongodb/docker-compose.yml
+++ b/examples/mongodb/docker-compose.yml
@ -0,0 +1,14 @@
+services:
+  mongodb:
+    image: mongo:7
+    ports:
+      # Non-default host port so the example does not clash with a local MongoDB.
+      - "27117:27017"
+    healthcheck:
+      test: ["CMD-SHELL", "mongosh --quiet --eval \"db.runCommand({ ping: 1 }).ok\" | grep -q 1"]
+      interval: 2s
+      timeout: 5s
+      retries: 30
+    volumes:
+      # MongoDB runs *.js here once, on first start, against an empty data dir.
+      - ./init:/docker-entrypoint-initdb.d:ro
--- a/examples/mongodb/init/seed.js
+++ b/examples/mongodb/init/seed.js
@ -0,0 +1,64 @@
+// Seed a representative MongoDB dataset for the ktx connector example.
+//
+// MongoDB runs this once on first container start (it is mounted into
+// /docker-entrypoint-initdb.d). It can also be applied by hand:
+//   mongosh "mongodb://localhost:27117" < examples/mongodb/init/seed.js
+//
+// The shapes here exercise the connector's schema inference end to end:
+// scalar BSON types, a nested sub-document, an array, Decimal128, dates, a
+// field with more than one type (-> "mixed"), an absent field (-> nullable),
+// an ObjectId reference for relationship discovery, and a view (to confirm
+// introspection never runs a count command on a view).
+const app = db.getSiblingDB('app');
+
+app.users.drop();
+app.orders.drop();
+
+app.users.insertMany([
+  {
+    email: 'ada@example.com',
+    age: 31,
+    active: true,
+    created: new Date('2026-01-04T10:00:00Z'),
+    balance: NumberDecimal('120.50'),
+    address: { city: 'NY', zip: '10001' },
+    tags: ['admin', 'early-access'],
+    ref: 'abc',
+  },
+  {
+    email: 'grace@example.com',
+    active: false,
+    created: new Date('2026-02-11T08:30:00Z'),
+    balance: NumberDecimal('0.00'),
+    address: { city: 'SF', zip: '94016' },
+    tags: [],
+    ref: 42, // a second type for this field -> inferred "mixed"
+    // age intentionally absent -> inferred nullable
+  },
+  {
+    email: 'linus@example.com',
+    age: 27,
+    active: true,
+    created: new Date('2026-03-01T12:00:00Z'),
+    balance: NumberDecimal('9.99'),
+    address: { city: 'Austin', zip: '73301' },
+    tags: ['beta'],
+    ref: null,
+  },
+]);
+
+const userIds = app.users.find({}, { _id: 1 }).toArray().map((u) => u._id);
+
+app.orders.insertMany([
+  { user_id: userIds[0], total: 120.5, status: 'paid', placed: new Date('2026-03-02T09:00:00Z') },
+  { user_id: userIds[0], total: 9.99, status: 'pending', placed: new Date('2026-03-05T14:00:00Z') },
+  { user_id: userIds[1], total: 50.25, status: 'paid', placed: new Date('2026-03-06T16:00:00Z') },
+]);
+
+// A view, to confirm introspection does not issue a count command on it
+// (MongoDB rejects count on a view with CommandNotSupportedOnView).
+app.createView('active_users', 'users', [{ $match: { active: true } }]);
+
+print('users: ' + app.users.countDocuments());
+print('orders: ' + app.orders.countDocuments());
+print('collections: ' + app.getCollectionNames().join(', '));
--- a/examples/mongodb/scripts/introspect-smoke.mjs
+++ b/examples/mongodb/scripts/introspect-smoke.mjs
@ -0,0 +1,53 @@
+// Deterministic, no-LLM smoke for the MongoDB connector. Drives the same
+// introspection entry point ktx ingest's "database schema" stage uses, against
+// the seeded example database, and asserts the inferred schema.
+//
+// Usage: node introspect-smoke.mjs [mongoUrl]
+import { fileURLToPath } from 'node:url';
+import { dirname, resolve } from 'node:path';
+
+const here = dirname(fileURLToPath(import.meta.url));
+const ktxRoot = resolve(here, '../../..');
+const connectorUrl = `file://${resolve(
+  ktxRoot,
+  'packages/cli/dist/connectors/mongodb/live-database-introspection.js',
+)}`;
+
+const mongoUrl = process.argv[2] ?? 'mongodb://localhost:27117/app';
+
+const { createMongoDbLiveDatabaseIntrospection } = await import(connectorUrl);
+
+function assert(condition, message) {
+  if (!condition) {
+    throw new Error(`assertion failed: ${message}`);
+  }
+}
+
+const port = createMongoDbLiveDatabaseIntrospection({
+  connections: { 'mongo-example': { driver: 'mongodb', url: mongoUrl, databases: ['app'] } },
+});
+
+const snapshot = await port.extractSchema('mongo-example');
+const tables = new Map(snapshot.tables.map((table) => [table.name, table]));
+
+assert(snapshot.driver === 'mongodb', 'snapshot driver is mongodb');
+assert(['orders', 'users'].every((name) => tables.has(name)), 'users and orders collections introspected');
+
+const users = tables.get('users');
+const columns = new Map(users.columns.map((column) => [column.name, column]));
+assert(columns.get('_id')?.primaryKey === true && columns.get('_id')?.nullable === false, '_id is the non-null primary key');
+assert(columns.get('age')?.nullable === true, 'age is nullable (absent in one document)');
+assert(columns.get('email')?.nullable === false, 'email is non-nullable (present in every document)');
+assert(columns.get('address')?.normalizedType === 'json', 'nested address maps to opaque json');
+assert(columns.get('tags')?.normalizedType === 'json', 'array tags maps to opaque json');
+assert(columns.get('ref')?.nativeType === 'mixed', 'ref with two types is inferred as mixed');
+
+const view = tables.get('active_users');
+assert(view?.kind === 'view', 'active_users is a view');
+assert(view?.estimatedRows === null, 'a view is introspected without a count (estimatedRows null)');
+
+console.log(`OK: introspected ${snapshot.tables.length} collections from ${mongoUrl}`);
+for (const table of snapshot.tables) {
+  console.log(`  - ${table.db}.${table.name} (${table.kind}, ${table.columns.length} columns)`);
+}
+process.exit(0);
--- a/examples/mongodb/scripts/smoke.sh
+++ b/examples/mongodb/scripts/smoke.sh
@ -0,0 +1,37 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+# Manual smoke for the MongoDB connector: start MongoDB, seed it, and assert the
+# connector's schema introspection (the deterministic, no-LLM half of ktx ingest's
+# "database schema" stage). The full enrichment ingest is documented in README.md.
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+EXAMPLE_DIR="$(cd "$SCRIPT_DIR/.." && pwd)"
+KTX_ROOT="$(cd "$EXAMPLE_DIR/../.." && pwd)"
+COMPOSE_FILE="$EXAMPLE_DIR/docker-compose.yml"
+CONNECTOR="$KTX_ROOT/packages/cli/dist/connectors/mongodb/live-database-introspection.js"
+MONGO_URL="${KTX_MONGODB_URL:-mongodb://localhost:27117/app}"
+
+# Compose engine: docker by default, override for podman:
+#   KTX_MONGODB_COMPOSE="podman compose" examples/mongodb/scripts/smoke.sh
+COMPOSE="${KTX_MONGODB_COMPOSE:-docker compose}"
+
+cleanup() {
+  if [[ "${KTX_MONGODB_KEEP:-0}" != "1" ]]; then
+    $COMPOSE -f "$COMPOSE_FILE" down -v >/dev/null 2>&1 || true
+  fi
+}
+trap cleanup EXIT
+
+if [[ ! -f "$CONNECTOR" ]]; then
+  echo "Build the CLI first: pnpm --filter @kaelio/ktx run build" >&2
+  exit 1
+fi
+
+echo "Starting MongoDB and seeding (${COMPOSE})…"
+$COMPOSE -f "$COMPOSE_FILE" up -d --wait
+
+echo "Asserting connector introspection against ${MONGO_URL}…"
+node "$SCRIPT_DIR/introspect-smoke.mjs" "$MONGO_URL"
+
+echo "Smoke passed."