feat(embeddings): storage layer — state.db schema + vector search + query cache#38

Merged

plind-junior merged 18 commits into

vouchdev:mainfrom

dripsmvcp:feat/semantic-search-phase-2-storage

May 21, 2026

Commits on May 20, 2026

docs(spec): semantic search as primary retrieval backend
dripsmvcp
committed
docs(plan): semantic search implementation plan
dripsmvcp
committed
feat(embeddings): add optional-deps extras and pytest markers
dripsmvcp
committed
feat(embeddings): create package skeleton
dripsmvcp
committed
feat(embeddings): Embedder ABC, registry, content_hash, MockEmbedder
dripsmvcp
committed
feat(embeddings): sentence-transformers all-mpnet-base-v2 default adapter
dripsmvcp
committed
feat(embeddings): sentence-transformers MiniLM-L6 alternative adapter
dripsmvcp
committed
feat(embeddings): fastembed BGE alternative (no-torch) adapter
dripsmvcp
committed
fix(ci): satisfy ruff SIM105 + add mypy overrides for optional deps
dripsmvcp
committed
fix(ci): skip embeddings test suite when numpy isn't installed
dripsmvcp
committed
fix(bundle): skip Pydantic validation for opaque source content files
dripsmvcp
committed
fix(embeddings): MockEmbedder uses uint32 scaling to avoid NaN/Inf
dripsmvcp
committed
feat(embeddings): state.db schema for embedding storage + put/get helpers
dripsmvcp
committed
feat(embeddings): NumPy brute-force cosine search over embedding_index
dripsmvcp
committed
feat(embeddings): sqlite-vec ANN path with NumPy fallback
dripsmvcp
committed
feat(embeddings): query embedding LRU cache
dripsmvcp
committed
style: fix ruff E402/SIM105/E702/I001 violations from Phase 2 storage tasks
dripsmvcp
committed
fix(embeddings): use outer-query for WHERE on cosine alias in search_embedding
dripsmvcp
committed