Hybrid Search
AIHybrid search combines full-text search (keyword-based) with vector search (semantic). Keyword search excels at exact matches (e.g., IDs, proper nouns); vector search excels at meaning-based matches. Hybrid approaches merge or re-rank results from both for broader coverage.
Use Cases
- Mixed queries — "Invoice #12345 total" benefits from keyword (ID) and semantic (total) search
- Precision + recall — Keyword narrows; vector broadens; combining can improve both
- Domain-specific terms — Technical terms may match better with keywords; paraphrases with vectors
DocLD focuses on vector search via Pinecone. Metadata filters can approximate keyword scoping. Full hybrid search may be available in future releases.
Related Concepts
Hybrid search combines full-text search and vector search. Reranking can reorder mixed results. Top-k and RAG use retrieved chunks from search.