Haystack
Modular AI orchestration framework for production RAG and agentic pipelines
View on GitHubWhat it solves
Building production RAG and agent pipelines where every stage is independently testable and replaceable: without rewriting the plumbing every time you swap a model or retriever.
Overview
Haystack, maintained by deepset, is a component-graph framework for building production AI pipelines. Every stage (document loading, chunking, embedding, retrieval, reranking, generation) is a typed component that you wire into a directed graph. Components are independently testable and swappable; the same pipeline can run a different retriever or LLM without touching the surrounding code. Supports tool-calling agents in addition to RAG.
Key facts
- Language
- Python
- License
- Apache 2.0
- Maturity
- Stable
- Maintainer
- deepset-ai
- Reviewed
- 2026-06-11
Where to start
The generators directory shows how Haystack wraps different LLM providers behind a common interface, understanding this pattern explains how the whole component system works.
haystack/components/generators/Technologies
More in this domain
LangGraph
Stateful multi-agent orchestration for LLM applications
The state management problem in multi-agent systems: how you build agents that pass context across turns without losing the execution thread.
AutoGen
Multi-agent conversation framework from Microsoft Research
How to coordinate multiple specialized agents toward a shared goal without writing low-level orchestration logic by hand.
CrewAI
Role-based multi-agent orchestration for collaborative task execution
The boilerplate problem in multi-agent setup: defining roles, delegation rules, and task routing without writing a custom orchestration graph.