Haystack is an end-to-end framework that enables you to build powerful and production-ready pipelines for different search use cases. The Docker image comes with a web service configured to serve Haystack's rest_api to ease pipeline deployments in containerized environments.

To start the Docker container binding the TCP port 8000 locally, run:

docker run -p 8000:8000 deepset/haystack

If you need the container to access other services available in the host, run:

docker run -p 8000:8000 --network="host" deepset/haystack

The Docker image comes in six variants:

• haystack:gpu-<version> contains Haystack dependencies as well as what's needed to run the REST API and UI. It comes with the CUDA runtime and is capable of running on GPUs.
• haystack:cpu-remote-inference-<version> is a slimmed down version of the CPU image with the REST API and UI. It is specifically designed for PromptNode inferencing using remotely hosted models, such as Hugging Face Inference, OpenAI, Cohere, Anthropic, and similar.
• haystack:cpu-<version> contains Haystack dependencies as well as what's needed to run the REST API and UI. It has no support for GPU so must be run on CPU.
• haystack:base-gpu-<version> only contains the Haystack dependencies. It comes with the CUDA runtime and can run on GPUs.
• haystack:base-cpu-remote-inference-<version> is a slimmed down version of the CPU image, specifically designed for PromptNode inferencing using remotely hosted models, such as Hugging Face Inference, OpenAI, Cohere, Anthropic, and similar.
• haystack:base-cpu-<version> only contains the Haystack dependencies. It has no support for GPU so must be run on CPU.

Images are built with BuildKit and we use bake to orchestrate the process. You can build a specific image by running:

docker buildx bake gpu

You can override any variable defined in the docker-bake.hcl file and build custom images, for example if you want to use a branch from the Haystack repo, run:

HAYSTACK_VERSION=mybranch_or_tag BASE_IMAGE_TAG_SUFFIX=latest docker buildx bake gpu --no-cache

Haystack images support multiple architectures. But depending on your operating system and Docker environment, you might not be able to build all of them locally.

You may encounter the following error when trying to build the image:

multiple platforms feature is currently not supported for docker driver. Please switch to a different driver
(eg. “docker buildx create --use”)

To get around this, you need to override the platform option and limit local builds to the same architecture as your computer's. For example, on an Apple M1 you can limit the builds to ARM only by invoking bake like this:

docker buildx bake base-cpu --set "*.platform=linux/arm64"

View license information for the software contained in this image.

As with all Docker images, these likely also contain other software which may be under other licenses (such as Bash, etc from the base distribution, along with any direct or indirect dependencies of the primary software being contained).

As for any pre-built image usage, it is the image user's responsibility to ensure that any use of this image complies with any relevant licenses for all software contained within.