Merge semantic and keyword retrieval with configurable alpha and RRF fusion in Python.
npx skills add https://github.com/juanre/llmemory --skill hybrid-search-juanre
Please help me install this Agent Skill. Skill: hybrid-search Source: https://github.com/juanre/llmemory/tree/main/skills/hybrid-search Command: npx skills add https://github.com/juanre/llmemory --skill hybrid-search-juanre
Hybrid search solves the challenge of delivering high-quality results by merging semantic vector similarity with exact keyword matching.
Install llmemory and dependencies, then initialize a memory instance and execute a hybrid search with a sample query to observe blended results.
High-intent search queries and answers about installing and using this skill.
Hybrid search applies Reciprocal Rank Fusion (RRF) to merge vector similarity scores with BM25-like keyword matching results. This fusion approach ensures that both semantic intent and exact lexical terms are accurately reflected in the final unified ranking.
You can tune the balance between semantic and lexical search results using a configurable alpha parameter. Adjusting this alpha weight allows you to prioritize vector similarity or exact keyword matching depending on your specific query type and data requirements.
Use hybrid search instead of standard vector search when your queries require exact keyword matching for specific terms like product codes or model numbers alongside semantic intent. It delivers significantly improved relevance for catalogs, documentation, and general search tasks with mixed query patterns.
Hybrid search supports pluggable vector and text search backends in Python using the llmemory library. This architecture allows you to configure parallel retrieval pipelines without being locked into a single specific vector database dependency.
To execute a hybrid search query in Python, install the llmemory library and its dependencies, initialize a memory instance, and run a sample query. The system will execute parallel vector and text retrieval, applying fusion to output blended results.
leonardoteodorooImplement hybrid search combining vector similarity with keyword matching for RAG systems.
zilliztechCombine BM25 keyword matching with vector semantic similarity in Milvus.
NotysotyDesign hybrid retrieval pipelines combining dense vector search with BM25 keyword matching.
timescaleCombine BM25 keyword search with semantic vector search in PostgreSQL.
