Retour aux applications
Haiku RAG
par ggozad
Agentic RAG powered by LanceDB, Pydantic AI, and Docling with hybrid search and research agents.
0 étoiles
3 vues
Fonctionne dans:claude
Expose:ToolsResources
What it does
Haiku RAG is an opinionated agentic Retrieval-Augmented Generation (RAG) system that connects your AI assistant to local documents. It uses LanceDB for embedded storage and Docling for structure-aware document processing, enabling high-fidelity retrieval from PDFs and other formats.
Tools
search: Performs hybrid vector and full-text search with Reciprocal Rank Fusion.ask: Question answering with precise citations including page numbers and section headings.research: Multi-agent workflow to plan, search, evaluate, and synthesize complex information.rlm: Runs complex analytical tasks via sandboxed Python code execution for data aggregation and computation.add-src: Indexes new documents or URLs into the local knowledge base.
Installation
Install the package via pip:
pip install haiku.rag
Add to your claude_desktop_config.json:
{
"mcpServers": {
"haiku-rag": {
"command": "haiku-rag",
"args": ["serve", "--mcp", "--stdio"]
}
}
}
Supported hosts
- Claude Desktop
Installation rapide
pip install haiku.rag && haiku-rag serve --mcp --stdioInformations
- Tarification
- free
- Publié
- 4/20/2026
- étoiles
