Recommended releases

Download Released
1.x-1.0.0-alpha1 (13 KB) Oct 27th, 2026 Notes

Hugging Face provider for the Backdrop CMS AI module.

Adds open-source models hosted on Hugging Face (https://huggingface.co/) to the
providers the ai module can route to, through the OpenAI-compatible Inference
Router or a Dedicated Inference Endpoint.

Supported operations

Operation Supported Notes
Chat Yes Streaming supported. JSON mode and JSON schema responses supported.
Completions Yes Sent as a single chat message; uses the chat endpoint.
Tool calling Yes Models the router catalog marks supports_tools on a live provider.
Thinking Manual The catalog has no reasoning flag; assign models on the Model capabilities page.
Embeddings Yes Router: the hf-inference feature-extraction pipeline. Dedicated endpoints: /v1/embeddings.
Vision Yes Models the router catalog lists with image input.
Image generation No
Moderation No
Speech-to-text No

Nothing is hardcoded. Chat models and their tool/vision flags come from the
router catalog (/v1/models, which is public). Embedding models are the
most-downloaded feature-extraction models that hf-inference serves, from the
Hub API. Custom models are offered for chat and embeddings. Adjust any of this
on the Model capabilities page.

Settings

The provider settings at admin/config/ai/settings add:

  • Base URL / Inference Endpoint — defaults to the serverless router
    https://router.huggingface.co/v1. Set it to a Dedicated Endpoint URL
    (https://xxxx.endpoints.huggingface.cloud/v1) to use your own deployment.
  • Custom / Additional Models — one model ID per line (or alias=model_id),
    added to the models discovered from the catalog.

Installation

Issues

Bugs and feature requests should be reported in the Issue Queue.

Current Maintainer

Justin Keiser

Credits

  • Created for Backdrop CMS by Justin Keiser.

  • Developed with AI assistance.

License

This project is GPL v2 software. See the LICENSE.txt file in this directory for complete text.