Skip to content

Releases: ZanzyTHEbar/mcp-memory-libsql-go

v0.7.0

11 Oct 23:12

Choose a tag to compare

0.7.0 (2025-10-11)

🍕 Features

  • release: configure @semantic-release/github to add releases at the bottom (99b7056)

🐛 Bug Fixes

  • release: add @semantic-release/github to publish GitHub Releases (db01680)

v0.6.0

11 Oct 23:00

Choose a tag to compare

Full Changelog: v0.5.2...v0.6.0

v0.1.0

15 Aug 16:24

Choose a tag to compare

0.1.0 (2025-08-15)

🍕 Features

  • search: add BM25 ranking for FTS results with env-tunable k1/b and safe fallbacks (134339e)
  • add Docker support and metrics instrumentation (db54259)
  • docker: add docker-compose support for multi-profile setups; enhance README with detailed Docker usage instructions; introduce integration testing capabilities with a dedicated tester tool (4b46586)
  • search: add FTS5-backed text search with automatic LIKE fallback; detect FTS5 at startup and sync via triggers; fix README dims typo to 768 (17c5b27)
  • database: add GetRelations convenience wrapper; all tests green (1d14e45)
  • search: add Hybrid Search strategy (RRF fusion of text + vector); env toggle HYBRID_SEARCH with tunable weights (0af0748)
  • add jsonschema annotations to argument types for improved documentation (9ed4378)
  • docker: add multi-stage Dockerfiles for SSE and STDIO transports; update Makefile to include docker build target; enhance README with tool summary adjustments (ea9cf3f)
  • add OpenAI and Ollama embeddings providers (cd64197)
  • add pagination support to search functions in database and server (a1d93b4)
  • metrics: add tool result size histogram with sampling; keep low label cardinality\n\n- New Recorder.ObserveToolResultSize(tool, size) with sampling via METRICS_RESULT_SAMPLE_N\n- Prometheus histogram tool_result_size with generic buckets\n- Keeps cardinality low (labels: tool only)\n\nRefs: #3 (6e0987a)
  • add update and health check tools to MCP server (863e6cc)
  • embeddings: add VoyageAI provider and dimension adapter; wire provider and deps (417aa3c)
  • graph: add Walk and ShortestPath functions for graph traversal; implement corresponding argument types and server handlers (cad80a7)
  • entrypoint: allow MODE to override CMD; fallback to CMD when MODE unset (ffc012c)
  • docker: consolidate Dockerfile for multi-transport builds; remove separate Dockerfiles for SSE and STDIO (6c6d194)
  • tests: enhance integration tester with additional delete operations and increase step capacity (5d233f1)
  • enhance MCP server functionality with new tools and improved project context handling (2ffa1ba)
  • enhance MCP tool handlers with input/output schemas and annotations (cf027de)
  • enhance vector extraction and add test for entity creation (e07ba40)
  • implement DatabaseManager for entity and relation management (69fc23c)
  • database: implement GetNeighbors function to fetch 1-hop neighbors; add NeighborsArgs type for input parameters; enhance metrics with pool stats and statement cache tracking (a232560)
  • implement main application logic for MCP Memory LibSQL server (d441beb)
  • implement multiple embeddings providers and enhance DBManager (475ffeb)
  • embeddings: implement providers for OpenAI, Ollama, Gemini, Vertex AI, and LocalAI; wire env-based selection and docs; batch-generate embeddings when missing; enforce EMBEDDING_DIMS match (555ab06)
  • implement statement caching in DBManager for improved query performance (73181e2)
  • entrypoint: improve model loading and timeout handling (91d3a94)
  • db: introduce soft cascade via trigger on entities delete; simplify DeleteEntity\n\n- Add trg_entities_delete_cascade to remove observations/relations on entity delete\n- Replace manual multi-table delete with single DELETE relying on trigger\n\nRefs: #7 (fd79d3b)
  • metrics: record result sizes across tool handlers with sampled histogram\n\n- Observe counts of entities/relations/updates per tool\n- Uses ObserveToolResultSize (sampling via METRICS_RESULT_SAMPLE_N)\n\nRefs: #3 (bc119bb)
  • db: refine capability probing to be per-project DB handle (vector_top_k, fts5)\n\n- Track caps per project with capsByProject map\n- Adjust search codepaths to consult/update per-project flags\n- Avoid global probe; initialize on getDB(project)\n\nRefs: #5 (f3b528d)
  • multi-project: require projectName; add per-project token auth with optional env toggles and auto-init; enforce across tools (9206f65)
  • search: robust FTS queries + prefix handling; fallback improvements\n\n- FTS tokenizer (unicode61) with tokenchars=:-_@./ and prefix index\n- buildFTSMatchExpr to support Task:* (entity_name/content) and quoted phrases\n- downgrade to LIKE on parse/table errors without disabling FTS globally\n- normalize nil slices in read_graph/search_nodes/open_nodes for schema compliance\n- prompts: document query syntax/tips; README: add query rules & examples (85467eb)
  • graph): add neighbors/walk/shortest_path tools with DB implementations and tests; feat(metrics: stmt-cache/pool gauges + periodic emission; docs: update tools table and metrics section; bench: add search benches (e83617e)
  • observability: structured error logs per tool with low-cardinality key=value format\n\n- Add logToolError helper and log on failures across handlers\n- Keeps logs machine-parseable without new deps\n\nRefs: #3 (21982fd)
  • db: transactional bulk deletes with chunking; rely on cascade trigger\n\n- DeleteEntities now wraps deletes in a single tx, uses IN-chunking (<=500)\n- DeleteObservations uses tx + chunking for ids/contents\n- Leverages trg_entities_delete_cascade for cleanup\n\nRefs: #8 (8a52e82)

🐛 Bug Fixes

  • install: check for existing binary before installation and remove it if found (7d5d088)
  • db: close cached prepared statements on shutdown to prevent descriptor leaks\n\n- Close all cached *sql.Stmt per project before closing DBs\n- Clear stmtCache buckets\n- Leave behavior unchanged otherwise\n\nRefs: #2 (633710f)
  • makefile: correct conditional check in docker-test target for proper string compa...
Read more