What is haiku-rag and what does it do?

Haiku-rag is an opinionated RAG (Retrieval-Augmented Generation) framework that combines LanceDB for vector storage, Pydantic AI for agentic reasoning, and Docling for document parsing. It simplifies building AI systems that retrieve and reason over document collections.

How do I install and run haiku-rag?

Install via `uvx haiku-rag@0.44.0`. The server runs over stdio transport, making it compatible with MCP clients. Check the project repository for configuration and startup examples.

What document types can haiku-rag process?

Docling handles multiple document formats including PDFs, Word documents, and other structured formats. Refer to Docling's documentation for the complete list of supported file types.

Can I use haiku-rag with my own LLM or API keys?

Haiku-rag integrates Pydantic AI, which supports multiple LLM providers. Check the repository for environment variable configuration and authentication setup for your chosen provider.

Is haiku-rag suitable for production RAG systems?

Haiku-rag provides an opinionated, batteries-included approach to RAG. For production use, review the project's stability guarantees, scalability considerations with LanceDB, and whether the included defaults match your performance and cost requirements.

MCP ServerSTDIOOfficialv0.44.0

Ggozad Haiku Rag MCP Server

Build agentic RAG systems with LanceDB vector storage, Pydantic AI agents, and Docling document parsing. For developers integrating retrieval-augmented generation into AI applications.

io.github.ggozad/haiku-rag

Try MCP with AI models View source

Hosted URL

Local install

Transport

STDIO

Auth

No auth required

What the Ggozad Haiku Rag MCP server does

How models use it and what it is built for.

Build agentic RAG systems with LanceDB vector storage, Pydantic AI agents, and Docling document parsing. For developers integrating retrieval-augmented generation into AI applications.

Connect to Ggozad Haiku Rag

Local install — runs as a subprocess.

uvx haiku-rag==0.44.0

Resources

Where to find authoritative docs and source for Ggozad Haiku Rag.

Source repository

Example prompts for Ggozad Haiku Rag

Paste any of these into Agent Studio after connecting Ggozad Haiku Rag.

Index a PDF document and query it with semantic search
Set up a multi-document RAG pipeline with LanceDB
Configure Pydantic AI agents to reason over retrieved context
What document formats does Docling support in this RAG setup

Ggozad Haiku Rag MCP server — FAQ

Common questions about connecting and running Ggozad Haiku Rag.

What is haiku-rag and what does it do?
Haiku-rag is an opinionated RAG (Retrieval-Augmented Generation) framework that combines LanceDB for vector storage, Pydantic AI for agentic reasoning, and Docling for document parsing. It simplifies building AI systems that retrieve and reason over document collections.
How do I install and run haiku-rag?
Install via `uvx haiku-rag@0.44.0`. The server runs over stdio transport, making it compatible with MCP clients. Check the project repository for configuration and startup examples.
What document types can haiku-rag process?
Docling handles multiple document formats including PDFs, Word documents, and other structured formats. Refer to Docling's documentation for the complete list of supported file types.
Can I use haiku-rag with my own LLM or API keys?
Haiku-rag integrates Pydantic AI, which supports multiple LLM providers. Check the repository for environment variable configuration and authentication setup for your chosen provider.
Is haiku-rag suitable for production RAG systems?
Haiku-rag provides an opinionated, batteries-included approach to RAG. For production use, review the project's stability guarantees, scalability considerations with LanceDB, and whether the included defaults match your performance and cost requirements.

Skip the local setup — run MCP in your browser

MCP Playground runs 10,000+ hosted MCP servers — GitHub, Linear, Notion, Stripe, Sentry and more — across Claude, GPT, Gemini, DeepSeek and 60+ AI models. Compare model answers side-by-side, save agent presets, share runs. Zero install.

Open Agent Studio