NEWVectors or files. Pick a path.Start →

    RAG Context Builder

    Format search results as context for LLM generation with citation tracking

    Why do anything?

    RAG applications need properly formatted context from search results. Manual formatting is error-prone.

    Why now?

    RAG is the dominant pattern for grounded LLM applications.

    Why this feature?

    Format search results into LLM context with source citations, relevance ordering, and token management.

    How It Works

    RAG preparation formats search results for LLM consumption.

    1

    Content Selection

    Select relevant content from documents

    2

    Citation Assignment

    Assign citation identifiers

    3

    Context Formatting

    Format into context string with citations

    4

    Token Management

    Truncate to fit token budget

    Why This Approach

    Standardized RAG preparation ensures consistent context formatting. Citation tracking enables attribution.

    Integration

    retriever = client.retrievers.create(stages=[{"type": "rag_prepare", "max_tokens": 4000}])