Skip to main content

Build AI Agents with Haystack + Needle

Haystack Logo

GitHub Repository

The Needle Haystack integration provides powerful RAG components for Haystack projects, including NeedleDocumentStore and NeedleEmbeddingRetriever. These components allow you to seamlessly integrate Needle's advanced search capabilities into your Haystack pipelines.

Installation​

Install the package via pip:

pip install needle-haystack-ai

API Keys​

To use the Needle Haystack integration, you'll need to set up your environment variables:

  • NEEDLE_API_KEY - Get your API key from Developer settings
  • OPENAI_API_KEY - Required if using OpenAI in your pipeline

Components​

NeedleDocumentStore​

The NeedleDocumentStore component provides a connection to your Needle collections. In Needle, document stores are called collections. For detailed information about collections, see our Collections documentation.

NeedleEmbeddingRetriever​

The NeedleEmbeddingRetriever component enables semantic search over your documents using Needle's advanced retrieval capabilities.

Example Pipeline​

Here's how to build a RAG pipeline using Needle components with Haystack:

from needle_haystack import NeedleDocumentStore, NeedleEmbeddingRetriever
from haystack import Pipeline
from haystack.components.generators import OpenAIGenerator
from haystack.components.builders import PromptBuilder

# Initialize Needle components
document_store = NeedleDocumentStore(collection_id="<your-collection-id>")
retriever = NeedleEmbeddingRetriever(document_store=document_store)

# Create prompt template
prompt_template = """
Given the following retrieved documents, generate a concise and informative answer to the query:

Query: {{query}}
Documents:
{% for doc in documents %}
    {{ doc.content }}
{% endfor %}

Answer:
"""

# Set up pipeline components
prompt_builder = PromptBuilder(template=prompt_template)
llm = OpenAIGenerator()

# Create and configure pipeline
pipeline = Pipeline()
pipeline.add_component("retriever", retriever)
pipeline.add_component("prompt_builder", prompt_builder)
pipeline.add_component("llm", llm)

# Connect components
pipeline.connect("retriever", "prompt_builder.documents")
pipeline.connect("prompt_builder", "llm")

# Run the pipeline
prompt = "What is the topic of the news?"
result = pipeline.run({
    "retriever": {"text": prompt},
    "prompt_builder": {"query": prompt}
})

# Get the response
print(result['llm']['replies'][0])

Support​

If you have questions or need assistance:

License​

The Needle Haystack integration is distributed under the terms of the MIT license.