deepset
Haystack
Production-ready pipelines for search and RAG.
Haystack is an open-source framework from deepset for building search systems, RAG, and agent pipelines. It centers on explicit, typed Pipelines composed of swappable components — retrievers, rankers, generators, and tools.
Install
bash
pip install haystack-aiQuickstart
A minimal example to verify your setup.
python
from haystack import Pipeline
from haystack.components.generators import OpenAIGenerator
from haystack.components.builders import PromptBuilder
pipe = Pipeline()
pipe.add_component("prompt", PromptBuilder(template="Answer: {{ q }}"))
pipe.add_component("llm", OpenAIGenerator(model="gpt-4o-mini"))
pipe.connect("prompt.prompt", "llm.prompt")
print(pipe.run({"prompt": {"q": "What is Haystack?"}}))Core concepts
Pipelines
Typed DAGs of components with validated inputs and outputs — easy to test, serialize to YAML, and deploy.
Components
Retrievers, rankers, embedders, generators, and tools share a small interface. Swap implementations without rewiring.
Document stores
First-class integrations with Elasticsearch, OpenSearch, Weaviate, Qdrant, pgvector, and more.
Agents & tools
Tool-calling agents built on the same Pipeline primitives, with branching and looping support.
Common use cases
- ›Enterprise search over millions of documents
- ›Hybrid keyword + vector RAG
- ›Question-answering on technical docs
- ›Customer support copilots