ai-llm-search-retrieval

What problem does it solve?

Traditional keyword search often misses semantic meaning, while pure vector search can struggle with exact matches. This Skill provides operational patterns for building modern search systems with hybrid retrieval and advanced ranking for significant relevance gains.

Core Features & Use Cases

• Hybrid Search (Modern Standard): Combine BM25 lexical search with dense vector search and Reciprocal Rank Fusion (RRF) for optimal relevance across diverse query types.
• Scalable Indexing: Implement billion-scale HNSW-IF and multi-vector HNSW indexing for efficient approximate nearest neighbor search on massive datasets.
• Ranking Pipelines: Design multi-stage ranking architectures, including cross-encoder reranking and LLM-based query rewriting, to refine search results and improve quality.

Quick Start

Use the ai-llm-search-retrieval skill to design a hybrid search configuration for an e-commerce product catalog, combining BM25 and vector search.