Every model, tool and connector an agent can be given.

Cross-encoder reranker that reorders retrieved chunks by true relevance before they reach the model's context.

Retrieval-tuned embedding model for English corpora where recall on long technical documents matters.

Splits documents on structure rather than character count, so a passage retrieved is a passage that makes sense.

Vector similarity and keyword matching run together, then merge — so an exact product code is found as reliably as a paraphrase.