Changelog

0.2.1

• Fix entity extraction for hyphenated tech terms
• Improve consolidation dedup performance with numpy matmul optimization (144ms → 16ms)
• Add recall_cache_ttl and recall_cache_size configuration options

0.2.0

• 4-channel hybrid retrieval (semantic + BM25 + entity + temporal)
• Reciprocal Rank Fusion for result merging
• Cross-encoder reranking with ms-marco-MiniLM-L-6-v2
• Background consolidation scheduler with write-threshold triggers
• LLM-powered promotion from raw → insight
• OpenAI and Anthropic SDK wrappers
• FastAPI router with 17 endpoints
• Memory tools for agent self-management (5 tools)
• Temporal versioning with supersession chains
• Multi-tenant scoping with rate limiting and quotas

0.1.0

• Initial release
• Basic write/recall/list/delete operations
• PostgreSQL + pgvector storage
• Local ONNX embeddings (all-MiniLM-L6-v2)
• Importance decay and raw chunk expiry