haiku.rag

Haiku RAG combines the power of LanceDB, Pydantic AI, and Docling to provide a hybrid search experience that integrates vector and full-text search capabilities. It supports complex question answering with citations and allows for multi-agent workflows that can plan, search, evaluate, and synthesize information effectively. The server also features a conversational interface for multi-turn interactions, ensuring a seamless user experience. In addition to its robust search and question answering functionalities, Haiku RAG offers tools for visual grounding, document structure awareness, and file monitoring to keep data up-to-date. With local-first architecture and support for various cloud providers, it ensures flexibility and ease of use. The MCP server exposes its capabilities for integration with AI assistants, making it a versatile tool for developers and researchers alike.

{
  "mcpServers": {
    "haiku-rag": {
      "args": [
        "haiku-rag"
      ],
      "command": "uvx"
    }
  }
}