Haystack

framework active open-source

An open-source framework by deepset for building production-ready LLM applications, with strong focus on RAG and search. Haystack uses a pipeline abstraction where components are connected declaratively, emphasizing production readiness and scalability.

Implements

Concepts this tool claims to implement:

  • Core pipeline components for retrieval, generation, and RAG. Strong document store abstractions and retriever implementations.

  • Retriever primary

    Multiple retriever types including BM25, embedding, and hybrid. Reranking components and document store integrations.

  • Workflow primary

    Pipeline abstraction with declarative component connections. YAML pipeline definitions for reproducibility.

  • Hybrid Search secondary

    Built-in support for combining sparse (BM25) and dense (embedding) retrieval with fusion methods.

Integration Surfaces

  • Python SDK
  • REST API
  • deepset Cloud

Details

Vendor
deepset
License
Apache-2.0
Runs On
local, cloud
Used By
human, system

Notes

Haystack is one of the oldest LLM application frameworks, predating LangChain. It's particularly strong for enterprise search and QA use cases, with production-focused features like pipeline serialization and monitoring.