Haystack
By deepset
Haystack is an open-source Python framework by deepset for building production-grade search and question-answering systems, including retrieval-augmented generation (RAG) pipelines, using large language models and vector search.
Definition
Haystack is an open-source Python framework by deepset for building production-grade search and question-answering systems, including retrieval-augmented generation (RAG) pipelines, using large language models and vector search.
Overview
Haystack lets developers compose modular "pipelines" out of components such as document stores, retrievers, rankers, and large language models (LLMs) to build applications like semantic search, document Q&A, and conversational agents. It was created by the Berlin-based startup deepset and has become a popular alternative to LangChain for teams that want tighter control over retrieval quality and evaluation. At its core, Haystack implements retrieval-augmented generation (RAG): a user query is converted into an embedding, matched against documents stored in a vector database such as Elasticsearch, OpenSearch, or a dedicated vector store, and the retrieved passages are passed as context to an LLM to generate a grounded answer. Pipelines can be extended with custom components, agents, and tool-calling logic for more complex, multi-step workflows. Haystack is widely used in enterprise settings where accuracy and traceability of AI-generated answers matter, such as internal knowledge bases, customer support search, and compliance-heavy document review. It integrates with model providers including OpenAI and Hugging Face, and is often taught alongside courses like Retrieval-Augmented Generation.
Key Features
- Modular pipeline architecture combining retrievers, rankers, and generators
- Native support for retrieval-augmented generation (RAG) and semantic search
- Integrations with major vector databases and document stores
- Built-in evaluation tools for measuring retrieval and answer quality
- Support for agents and multi-step tool-calling workflows
- Compatibility with open-source and commercial LLM providers
- REST API and UI tooling for deploying search applications
- Active open-source community maintained by deepset