Filter retrieved context down to sentences relevant to a query.

Runs sentence segmentation, then a local cross-encoder (cross-encoder/ms-marco-MiniLM-L-6-v2 by default) to score every sentence's relevance to the query, keeps sentences scoring at or above threshold (0-1, default 0.5), optionally trims to max_tokens, and returns the filtered context along with real measured metrics.

Return per-sentence cross-encoder relevance scores for a context.

Segments context into sentences and scores each one's relevance to query with the local cross-encoder, without applying any threshold. Useful for inspecting why Uma would keep or drop a given sentence.

Return cumulative Uma filtering statistics for this server process.