Skip to content

bm25

diapason.tools.storage.bm25

BM25 memory backend — classic term-frequency retrieval.

BM25Memory

Bases: MemoryBackend

In-memory BM25 (Okapi) retrieval backend.

Uses the rank_bm25 library to score documents against a query using the classic BM25 probabilistic ranking function. All data lives in memory — there is no persistence across restarts.

clear()

Remove all stored documents — always via Rust backend.

count()

Return the number of stored documents — always via Rust backend.

delete(doc_id)

Delete a document by id — always via Rust backend.

retrieve(query, *, top_k=5, **kwargs)

Search for query and return the top-k results — always via Rust backend.

store(content, *, source='', metadata=None)

Persist content and return a unique document id.