Feature Description
Allow users to connect Hugging Face Chat to models and inference endpoints belonging to organizations they have access to.
For example, if a user belongs to a private Hugging Face organization, they should be able to:
- Select their organization from Hugging Face Chat.
- Access models hosted privately within that organization.
- Connect to Hugging Face Inference Endpoints deployed under that organization.
- Use those private models/endpoints directly in Hugging Face Chat, alongside the existing public models.
This would be particularly useful for teams running private, fine-tuned, or organization-specific models where inference needs to remain within their organization's infrastructure.
Screenshots (if relevant)
Not applicable.
Implementation idea
The model selector could provide an option to select/connect an organization. After authentication, Hugging Face Chat could retrieve the organizations and resources accessible to the user and expose their private models and Inference Endpoints as selectable models.
For Inference Endpoints, the endpoint could be treated as a model/provider in the same way as the existing supported inference providers.
Ideally, authentication and authorization would rely on the user's existing Hugging Face permissions, so users only see private resources they are authorized to access.
Happy to help implement this if there is interest in the feature.
Feature Description
Allow users to connect Hugging Face Chat to models and inference endpoints belonging to organizations they have access to.
For example, if a user belongs to a private Hugging Face organization, they should be able to:
This would be particularly useful for teams running private, fine-tuned, or organization-specific models where inference needs to remain within their organization's infrastructure.
Screenshots (if relevant)
Not applicable.
Implementation idea
The model selector could provide an option to select/connect an organization. After authentication, Hugging Face Chat could retrieve the organizations and resources accessible to the user and expose their private models and Inference Endpoints as selectable models.
For Inference Endpoints, the endpoint could be treated as a model/provider in the same way as the existing supported inference providers.
Ideally, authentication and authorization would rely on the user's existing Hugging Face permissions, so users only see private resources they are authorized to access.
Happy to help implement this if there is interest in the feature.