Haystack

Haystack by deepset is a framework for building production-grade NLP and RAG pipelines with a composable, pipeline-based architecture. Each component (retriever, reader, generator, embedder) is a modular building block that can be swapped and configured independently.

Key use cases include:

• Building enterprise search systems that understand natural language queries
• Question answering over large document collections
• RAG pipelines with fine-grained control over retrieval strategies
• Document processing — converting, cleaning, splitting, and embedding documents
• Hybrid search combining keyword (BM25) and semantic retrieval
• Building conversational search interfaces with follow-up questions

Haystack is used by enterprise teams who need production-grade RAG with thorough testing, type safety, and fine-grained component control. Its pipeline architecture makes it easy to benchmark different configurations and swap components.

Haystack 2.x provides a clean API redesign with strong typing and a growing ecosystem of integrations.

1. Install Haystack:

   # Core + OpenAI
   pip install haystack-ai
   
   # With specific integrations
   pip install haystack-ai elasticsearch-haystack

2. Set your API key:

   export OPENAI_API_KEY='sk-...'

3. Build a simple RAG pipeline:

   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: {{question}}'
   ))
   pipe.add_component('llm', OpenAIGenerator(model='gpt-4o'))
   pipe.connect('prompt', 'llm')
   result = pipe.run({'prompt': {'question': 'What is RAG?'}})

4. Explore haystack.deepset.ai/tutorials for full RAG and search pipelines.

Pricing: Haystack is free and open source (Apache 2.0). deepset Cloud offers managed deployment and enterprise features. You pay only for LLM and infrastructure costs.