New integration proposal: HPC-AI

[lioZ129]

Summary and motivation

I would like to propose adding an official HPC-AI integration to haystack-core-integrations.

HPC-AI provides an OpenAI-compatible chat completions API, so it fits well with Haystack’s existing pattern for provider-specific OpenAI-compatible integrations. A dedicated integration would give users a clearer setup, provider-specific docs, and tested behavior instead of relying on a generic OpenAI configuration.

This integration would mainly support users who want to use HPC-AI-hosted models in Haystack pipelines and agents, especially:

• minimax/minimax-m2.5
• moonshotai/kimi-k2.5

Detailed design

The integration would be added under integrations/hpc_ai and released as hpc-ai-haystack.

The initial scope would be intentionally small:

• add a single public component: HPCAIChatGenerator
• do not add a non-chat generator in the first version

HPCAIChatGenerator would reuse Haystack’s existing OpenAIChatGenerator behavior with HPC-AI-specific defaults:

• api_key=Secret.from_env_var("HPC_AI_API_KEY")
• api_base_url=os.getenv("HPC_AI_BASE_URL", "https://api.hpc-ai.com/inference/v1")
• model="minimax/minimax-m2.5"

The integration would declare these officially supported models:

• SUPPORTED_MODELS = ["minimax/minimax-m2.5", "moonshotai/kimi-k2.5"]

Because the API is OpenAI-compatible, the integration would reuse Haystack’s existing support for:

• chat completions
• streaming callbacks
• tool calling
• structured outputs
• sync and async usage

If accepted, the implementation would also include the usual repo wiring for a new integration:

• tests and CI workflow
• integration README and examples
• labeler update
• README inventory update
• companion docs in haystack
• integration page in haystack-integrations

To keep the scope easy to review and maintain, the first version would be chat-only and limited to the two models above.

Checklist

If the request is accepted, ensure the following checklist is complete before closing this issue.

Tasks

• The code is documented with docstrings and was merged in the main branch
• Docs are published at https://docs.haystack.deepset.ai/
• There is a Github workflow running the tests for the integration nightly and at every PR
• A new label named like integration:hpc-ai has been added to the list of labels for this repository
• The labeler.yml file has been updated
• The package has been released on PyPI
• An integration tile with a usage example has been added to https://github.com/deepset-ai/haystack-integrations
• The integration has been listed in the Inventory section of this repo README
• The feature was announced through social media

[lioZ129]

3 relative PRs (closed but can reopen): #3109 deepset-ai/haystack#11048 deepset-ai/haystack-integrations#438