NEWVectors or files. Pick a path.Start →

    Query Expansion

    Expand search queries with synonyms, related terms, and reformulations using LLM

    Why do anything?

    Users search with limited vocabulary. Exact query matching misses relevant documents with different terminology.

    Why now?

    LLMs can generate semantically equivalent query variations to improve recall.

    Why this feature?

    LLM expands queries with synonyms, related concepts, and alternative phrasings. Supports multiple expansion strategies.

    How It Works

    Query expansion uses LLM to generate semantically related query variations.

    1

    Query Analysis

    Analyze original query intent

    2

    Expansion Generation

    LLM generates synonyms and variations

    3

    Query Combination

    Combine expanded queries for search

    Why This Approach

    LLM understands semantic relationships. Expansion improves recall without sacrificing precision.

    Integration

    retriever = client.retrievers.create(stages=[{"type": "query_expand", "max_expansions": 5}])