Haystack
🆕 Faster agents with parallel tool execution and guardrails & moderation for safer apps. See what's new in Haystack 2.15 🌟

The Production-Ready
Open Source AI Framework

Stars |

Highly
customizable

Don’t just use Haystack, build on top of it.

The flexible components and pipelines architecture allows you to build around your own specifications and use-cases. Whether you’re building a simple retrieval-augmented generation (RAG) app or a complex agentic pipeline with many moving parts.

Build with leading LLM providers and AI tools

Have the freedom of choice.

Thanks to our partnerships with leading LLM providers, vector databases, and AI tools such as OpenAI, Anthropic, Mistral, Weaviate, Pinecone and so many more

Production is where it gets real

Get your application in front of the world.

Haystack is built with production in mind. Pipelines are fully serializable and perfect for K8s native workflows. Logging and monitoring integrations give you the transparency you need. Deployment guides walk you through full-scale deployments on all clouds and on-prem.

Build Haystack apps faster with deepset Studio

  • Drag, drop, and construct Haystack pipelines
  • Bring your own files or connect your database
  • Deploy on deepset or export pipelines locally
  • Test, debug, and share your prototype
  • Free and open to everyone
Learn more
Start building

Learn how to extend Haystack with deepset AI Platform for faster building, easier iteration and instant deployment.

People in our community work for:

AWS
Nvidia
IBM
Intel
Comcast

Haystack Use Cases

  • Multimodal
    AI

    Architect a next generation AI app around all modalities, not just text. Haystack can do tasks like image generation, image captioning and audio transcription too.

  • Conversational
    AI

    All of our generators provide a standardized chat interface so that you can focus on building the perfect bot for your users.

  • Content
    Generation

    The flexibility and composability of Haystack’s prompt flow is unparalleled. Leverage our Jinja-2 templates and build a content generation engine that exactly matches your workflow.

  • Agentic
    Pipelines

    All our LLM generators come with a standard function-calling interface so that your LLM can leverage tools to achieve more. Our pipeline architecture provides branching and looping to support complex agent workflows.

  • Advanced
    RAG

    Build highly performant RAG pipelines with a multitude of retrieval and generation strategies. From hybrid retrieval to self-correction loops, Haystack has got you covered.

Join the community

Join our Discord

Our community on Discord is for everyone interested in NLP, using Haystack or even just getting started!

GET STARTED

Upcoming Haystack Events

Haystack Commmunity Talks

Breaking Down DeepSeek-R1

Evaluating AI with Haystack

Adding Tools to Agentic Pipelines & Other Experimental Features

  • Breaking Down DeepSeek-R1

  • Evaluating AI with Haystack

  • Adding Tools to Agentic Pipelines & Other Experimental Features