Hybrid Search (English)
English summary of xb/doc/ai_application/HYBRID_SEARCH.md. It covers combining scalar filters with vector ranking.
Patterns
- SQL first – filter candidates with SQL, then vector search by IDs.
- Vector first – retrieve top-N vectors, then join with SQL for enrichment.
- Mixed – run both, merge results with business scoring.
Implementation tips
- Use
Cond blocks to wrap optional filters.
- Keep vector limits small when chaining to SQL (avoid large
IN lists).
- Monitor total latency; break down SQL vs vector cost.
References
doc/en/VECTOR_GUIDE.mddoc/en/QDRANT_GUIDE.md