Hybrid Search Sparse Dense

Operator builds sparse_vector as a Python dict {token:weight} and upserts to a hybrid-capable index.

Per docs, sparse_vector is {indices:[int...], values:[float...]} with matching lengths — both lists same order. The dict form must be converted to two parallel arrays. Validate before upsert; mismatched lengths or non-int indices are rejected.

Hybrid response carries a single 'score' field per match. Operator threshold-filters at score >= 0.5.

Hybrid score is a fused metric (alpha-blended) and not directly comparable to pure-dense cosine scores across alpha values. Calibrate thresholds per (index, alpha) on holdout. Do not share thresholds when alpha changes.

Operator created the hybrid-capable index with metric=cosine.

Per docs, sparse-dense hybrid requires metric=dotproduct (the only metric that admits sparse fusion in Pinecone's published formulation). Create the hybrid index with metric=dotproduct; using cosine breaks hybrid semantics. Confirm with describe_index.