v0.4.0

What's Changed

• chore: Bump GitHub Actions to Node.js 24 by @5queezer in #83
• feat: Add /release skill for PyPI publishing by @5queezer in #82
• chore: Remove unused infra/ and tools/ scaffolding by @5queezer in #84
• fix: detect embedding dimension mismatch on startup by @5queezer in #85
• fix: make distill seed execute instead of printing prompt by @5queezer in #86
• feat: add distill reembed CLI to rebuild vectors on model change by @5queezer in #87
• feat: retention purge, lineage tracking, temporal search, retrieval benchmarks by @5queezer in #88
• chore: bump version to 0.4.0 by @5queezer in #94

Full Changelog: v0.3.0...v0.4.0

v0.3.0

What's Changed

• feat: auto-observe pipeline — replace remember/confirm with hook-based capture by @5queezer in #80
• chore: bump version to 0.3.0 by @5queezer in #81

Full Changelog: v0.2.0...v0.3.0

v0.2.0

What's Changed

• fix: include skills/ directory in wheel distribution by @5queezer in #50
• Add emoji to Distill README title by @5queezer in #51
• feat: proactive remember with confirmed param by @5queezer in #56
• fix: privacy hardening — close bypass, add PII detection, harden prompt by @5queezer in #60
• feat: contradiction detection on remember (#58) by @5queezer in #61
• feat: multi-level memory with search ranking (#57) by @5queezer in #62
• feat: stale memory identification and cleanup tool (#59) by @5queezer in #63
• feat: distill export command with JSON/CSV output (#17) by @5queezer in #64
• docs: update MkDocs for Phase 1-3 features by @5queezer in #66
• feat: AWS Bedrock and Azure OpenAI embedding adapters (#19, #20) by @5queezer in #65
• feat: add Gemini provider and decouple storage/embedding/distillation config by @5queezer in #67
• fix: lower MIN_SEARCH_SCORE so search returns results (#76) by @5queezer in #77
• chore: bump version to 0.2.0 by @5queezer in #78

Full Changelog: v0.1.2...v0.2.0

v0.1.2

Fix PyPI README: replace Mermaid code fence with pre-rendered SVG, use absolute image URLs.

v0.1.0 — First PyPI release

Privacy-first team memory MCP server — first public release.

Install

pip install distill-mcp

Highlights

• Local LLM distillation via Ollama (raw text never leaves your device)
• SQLite + LanceDB local storage
• PostgreSQL + pgvector team backend (optional)
• GCP Vertex AI embeddings (optional)
• Git-based identity with Postgres RLS
• 8 MCP tools: remember, search, get, update, forget, confirm, list, search_memory

Documentation

https://5queezer.github.io/distill/