Haystack Core Integrations

This repository contains integrations to extend the capabilities of Haystack. The code in this repo is maintained by deepset, see each integration's README file for details around installation, usage and support.

Quick start

You will need hatch to work on or create new integrations, open this link and follow the install instructions for your operating system and platform.

All the integrations are self contained, so the first step before working on one is to cd into the proper folder. For example, to run the tests suite for the Chroma document store, from the root of the repo:

$ cd integrations/chroma
$ hatch run test:all

Hatch will take care of setting up an isolated Python environment and run the tests.

Please check out our Contribution Guidelines for all the details.

Inventory

Package	Type	PyPi Package	Status	Coverage (unit tests)	Coverage (combined)
aimlapi-haystack	Generator
alloydb-haystack	Document Store
amazon-bedrock-haystack	Embedder, Generator, Ranker, Downloader
amazon-sagemaker-haystack	Generator
anthropic-haystack	Generator
arcadedb-haystack	Document Store
astra-haystack	Document Store
azure-ai-search-haystack	Document Store
azure-doc-intelligence-haystack	Converter
brave-haystack	Web Search
chonkie-haystack	Preprocessor
chroma-haystack	Document Store
cohere-haystack	Embedder, Generator, Ranker
cometapi-haystack	Generator
deepeval-haystack	Evaluator
docling-haystack	Converter
dspy-haystack	Generator
e2b-haystack	Tool
elasticsearch-haystack	Document Store
faiss-haystack	Document Store
falkordb-haystack	Document Store
fastembed-haystack	Embedder, Ranker
firecrawl-haystack	Fetcher, Web Search
github-haystack	Connector
google-ai-haystack	Generator		Archived - use google-genai-haystack instead
google-genai-haystack	Embedder, Generator
google-vertex-haystack	Embedder, Generator		Archived - use google-genai-haystack instead
hanlp-haystack	Preprocessor
jina-haystack	Connector, Embedder, Ranker
kreuzberg-haystack	Converter
langfuse-haystack	Tracer
lara-haystack	Translator
libreoffice-haystack	Converter
llama-cpp-haystack	Generator
llama-stack-haystack	Generator
markitdown-haystack	Converter
mcp-haystack	Tool
meta-llama-haystack	Generator
mistral-haystack	Converter, Embedder, Generator
mongodb-atlas-haystack	Document Store
nvidia-haystack	Embedder, Generator, Ranker
ollama-haystack	Embedder, Generator
openrouter-haystack	Generator
opensearch-haystack	Document Store
optimum-haystack	Embedder
paddleocr-haystack	Converter
pinecone-haystack	Document Store
pgvector-haystack	Document Store
presidio-haystack	Preprocessor
pyversity-haystack	Ranker
qdrant-haystack	Document Store
ragas-haystack	Evaluator
snowflake-haystack	Retriever
sqlalchemy-haystack	Retriever
stackit-haystack	Embedder, Generator
supabase-haystack	Document Store
tavily-haystack	Websearch
togetherai-haystack	Generator
unstructured-fileconverter-haystack	File converter
valkey-haystack	Document Store
vllm-haystack	Embedder, Generator, Ranker
watsonx-haystack	Embedder, Generator
weave-haystack	Tracer
weaviate-haystack	Document Store
instructor-embedders-haystack	Embedder		Staged

Releasing

To release a new version of an integration to PyPI tag the commit with the right version number and push the tag to GitHub. The GitHub Actions workflow will take care of the rest.

1. Tag the commit with the right version number

   The tag needs to have the following format:

   git tag integrations/<INTEGRATION_FOLDER_NAME>-<version>
   

   For example, if we want to release version 1.0.99 of the google-vertex-haystack integration we'd have to push the tag:

   git tag integrations/google_vertex-v1.0.99
   

2. Push the tag to GitHub

   git push --tags origin
   

3. Wait for the CI to do its magic

   If the release is successful, the HaystackBot will push a commit on main to update the changelog.