Model Management on Paddler Dashboard (Out-of-the-Box Experience) #162
Description
This feature aims to improve the out-of-the-box experience of the Paddler Dashboard by offering a broader selection of AI models available directly from the dashboard. Users should be able to fully manage the model lifecycle without manual setup or CLI usage.
The dashboard should act as a centralized model marketplace and manager, providing visibility, control, and status tracking for all supported models.
Functional Requirements
1. Model Catalog (Available Models)
The dashboard must display a list of available models, including:
- Model name
- Version
- Disk size
Short description / intended use (chat, coding, embeddings, lightweight, etc.)
Current status:
- Not Installed
- Downloading
- Installed
- Failed
Models must be visible even before installation to encourage exploration and quick setup.
2. Download Model
Users can initiate a model download directly from the dashboard.
The UI should display:
- Download progress (%)
- Download speed (optional)
- Estimated time remaining (optional)
Downloads must be:
- Asynchronous
- Resilient to page refreshes (state is maintained server-side)
3. Cancel Download
While a model is in the Downloading state, users must be able to cancel the download.
On cancellation, the system must:
- Terminate the active download process
- Clean up partial or corrupted files (or mark them as invalid)
- Revert the model status to Not Installed
4. Uninstall Model
For installed models, users can uninstall them via the dashboard.
Uninstalling must:
- Remove model files from disk
- Free occupied storage space
The system must check:
- Whether the model is currently in use
- If in use, display a warning or block the uninstall action
5. Reinstall Model
Reinstall must be supported for:
- Already installed models
- Models in a Failed state
Reinstall behavior:
- Performs an uninstall followed by a fresh download
- Resets all local caches and artifacts related to the model