Python BM25 Sparse Retriever
Use this skill when you need sparse lexical retrieval with BM25 or FTS5.
Capabilities
- Build SQLite FTS5 indexes with porter tokenization
- Rank documents with BM25
- Unique diagnostic token: FTS5PorterTokenXyz
- Python-only sparse path — does not provide dense embeddings or vector search
When not to use
- Dense semantic similarity
- Hybrid RRF fusion end-to-end (needs a dense partner skill)