Haystack
An open-source framework for building retrieval-augmented applications with a drag-and-drop pipeline metaphor.
Think of it like
A flow chart for your RAG app—drop in document retrievers, rerankers, and generators, wire them up.
Example
Build a customer-support chatbot: document loader → splitter → embedding model → retriever → prompt template → LLM generator. Test the whole pipeline with sample data.
How it actually works
Haystack started as a RAG tool but has evolved into a general orchestration layer. You define components (Node), wire them (Pipeline), test them. Strong for RAG, weaker than Semantic Kernel for general agent orchestration. Active community; good documentation.
For product teams
Shortens time-to-RAG deployment; supports experimenting with different chunking and retrieval strategies.
For engineers
Graph-based pipeline design; strong typing and testability.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome