Hugging Face Import Guide

Use this flow to pull GGUF-based models from Hugging Face into your selected runtime host.

What this importer expects

  • A model repository in username/repo format

  • At least one .gguf file in that repository

You can enter either a plain username/repo path or a Hugging Face URL; the app normalizes supported formats.

Open the Hugging Face importer

  1. Open Marketplace.

  2. Select Hugging Face.

  3. Enter the model path.

Select host

Set Host in the modal:

  • Local to pull into local runtime

  • A configured remote host to pull into remote runtime

Choose quantization and pull

After you enter a valid model path:

  1. The app checks Hugging Face and finds available GGUF quantizations.

  2. Select one quantization from the dropdown.

  3. Click Pull Model.

The app then starts model pull for an ID like hf.co/username/repo:QUANT on your selected host.

After pull completes

  1. Open Chat.

  2. Select the same host used during pull.

  3. Choose the pulled model and send a test prompt.

Troubleshooting

  • Format: username/repo message: fix the model ID format.

  • Model not found: confirm the repo exists and is public/accessible.

  • Quantization list does not appear: repo may not contain GGUF files.

  • Pull fails: verify host connectivity and available disk space.