← Graph

Hybrid Search

concept 2 connections

Common retrieval strategy for LLM-powered applications: mix classic word-based search with vector-embedding search (and whatever else works) rather than relying on either alone. Hasiński recommends it for 2025 because there's still no good universal solution for LLM retrieval and the field churns constantly. Also suggests considering graph databases — especially LLM-generated ones — as part of the mix.

category
pattern
about
Hybrid Search concept
Recommends mixing keyword and vector search for LLM apps.
recommends
Hybrid Search concept
Recommends combining classic and vector search plus considering graph DBs.

Provenance

Read by
2 extractions