Skip to content

CHANGELOG.md

Latest commit

 

History

History
57 lines (51 loc) · 3.15 KB

CHANGELOG.md

File metadata and controls

57 lines (51 loc) · 3.15 KB

Changelog

[Unreleased]

Added

  • brain_expand MCP tool — expand a chunk_id with N surrounding chunks for full context retrieval
  • Groq enrichment backend support (GROQ_API_KEY, BRAINLAYER_GROQ_URL, BRAINLAYER_GROQ_MODEL)
  • [style] optional extra: ChromaDB vector store as alternative backend
  • faiss-cpu added to [brain] optional extra for fast ANN search
  • Deferred embedding in brain_store — async embedding pipeline for faster writes (#76)
  • Eval suite + entity injection in UserPromptSubmit hook (#72)
  • C6 eval coverage expansion — 18 new test cases for search scenarios (#75)
  • C7 failed query mining script — automated discovery of low-score searches (#74)
  • C1+C2 lifecycle hardening — stale MCP process cleanup + WAL checkpoint on start (#73)
  • Post-RRF importance and recency reranking (US-002) — boosts high-importance and recent chunks
  • FTS5 expansion to index summary, tags, resolved_query fields (US-001)
  • KG entity quality — validation, prompts, and cleanup pipeline (#69)
  • KG rebuild pipeline with audit fixes (119 entities, 153K entity-chunk links) (#67)
  • Groq rate limiter for enrichment backend (#68)

Fixed

  • brain_graph.py: replaced correlated subquery with literal 'claude_code' — eliminates N+1 subquery on large DBs (PR #62)
  • Removed tree-sitter>=0.21.0 from core dependencies
  • brainlayer serve docs: removed non-existent --http flag
  • brain_expand for manual chunk IDs — fixed lookup path (#78)
  • Consolidated DB paths — single canonical ~/.local/share/brainlayer/brainlayer.db across all scripts (#77)
  • brain_search now returns chunk_id in results (#66)
  • DB lock resilience — retry + queue for MCP writes and reads (#65)
  • format parameter renamed to avoid Python builtin shadowing (BUG-005)
  • Entity merge in cleanup script — UNIQUE constraint violation (#a125d68)
  • Search validation hardened with backfill coverage (a108aff)

Changed

  • MCP tool count updated from 7 to 8 (brain_expand added)
  • Brain graph clustering documented as Leiden + UMAP (was incorrectly listed as HDBSCAN + UMAP)
  • Test count: 698 → 929 (after eval suite expansion and search hardening)
  • [kg] optional extra documented in README and Optional Extras section
  • Search: hybrid RRF reranking now incorporates importance and recency signals
  • Embeddings: deferred from synchronous to async pipeline

[1.0.0] - 2026-02-19

Added

  • Initial open-source release as BrainLayer (formerly Zikaron)
  • Semantic search across AI conversation history (sqlite-vec + bge-large-en-v1.5)
  • 10-field LLM enrichment pipeline (Ollama / MLX backends)
  • Brain graph visualization (HDBSCAN clustering + UMAP 3D layout)
  • MCP server with 7 tools (+ 14 backward-compatible aliases) for Claude Code, Zed, Cursor
  • Interactive setup wizard (brainlayer init)
  • Centralized artifact storage (~/.local/share/brainlayer/storage/)
  • Multi-source indexing: Claude Code, WhatsApp, YouTube, Markdown, Claude Desktop
  • Communication style analysis pipeline
  • Obsidian vault export
  • FastAPI daemon with 25+ HTTP endpoints
  • GitHub Actions CI/CD with PyPI publishing
  • PII sanitization pipeline for safe cloud processing
  • Source-aware enrichment thresholds