Hugging Face source
Which Hub APIs are used
BotConnector reads public Hub endpoints:
| Data | Endpoint |
|---|---|
| Model search/filter | GET https://huggingface.co/api/models?filter=gguf&search=… |
| Model metadata | GET https://huggingface.co/api/models/<repo> |
| File tree with sizes | GET https://huggingface.co/api/models/<repo>/tree/main |
| File downloads | https://huggingface.co/<repo>/resolve/main/<file> |
These are the same documented endpoints described in the Hub API reference. No personal token is needed for public models.
How the website uses them
This website never calls the Hub from the browser. Requests go to /api/hf/search and /api/hf/model/<repo> on this site's server, which:
- caches Hub responses for ~10 minutes,
- strips any upstream fields not needed for display,
- escapes all model-provided metadata before rendering.
How the desktop app uses them
The desktop app performs the same reads from your machine and stores a snapshot of metadata with the downloaded model, so installed models remain inspectable offline. See Offline usage.
What is NOT read
- No authentication tokens are bundled or sent from the website.
- Private repositories are never listed; only public GGUF repos appear.
- README text is not rendered as HTML on the website — only structured metadata.
Gated models
Gated repositories appear with a Gated badge. Downloading them requires accepting the license on huggingface.co with your account; BotConnector does not bypass gating. See Gated models.