benfinklea-ComfyUI-ModelSync/README.md

# ComfyUI-ModelSync

**Model Auto-Update & Multi-GPU Sync for ComfyUI**

A custom node extension that provides:
- **Model Version Tracking** - Detect when HuggingFace/CivitAI models have updates
- **Auto-Update System** - Configurable automatic model downloads
- **Multi-GPU Sync** - Sync models to secondary machines filtered by VRAM capacity
- **Metadata Management** - Track installed models with hashes, versions, sources

## Features

### Model Version Tracking
- Automatically detect updates for models from HuggingFace and CivitAI
- Compare local files with remote versions using SHA256 hashes
- Cache version information to minimize API calls

### Auto-Update System
- Background scheduler for periodic update checks
- Configurable update intervals
- Automatic downloads with hash verification
- Backup old versions before updating

### Multi-GPU Sync
- Sync models to secondary machines using rsync over SSH
- Filter models by VRAM requirements
- Generate standalone sync scripts for manual execution
- Configurable per-target settings

### Web UI
- Integrated panel in ComfyUI interface
- View all tracked models and their status
- Configure sync targets and settings
- Trigger manual updates and syncs

## Installation

### Using ComfyUI Manager (Recommended)

Search for "ModelSync" in ComfyUI Manager and install.

### Manual Installation

1. Clone this repository into your ComfyUI custom_nodes directory:
   ```bash
   cd ComfyUI/custom_nodes
   git clone https://github.com/benfinklea/ComfyUI-ModelSync.git
   ```

2. Install dependencies:
   ```bash
   cd ComfyUI-ModelSync
   pip install -r requirements.txt
   ```

3. Restart ComfyUI

## Usage

### Web Interface

1. Click the "Model Sync" button in the ComfyUI menu
2. **Models Tab**: View and manage tracked models
   - Click "Scan" to discover models in your models directory
   - Click "Check Updates" to query HuggingFace/CivitAI for updates
   - Enable auto-update per model using the checkbox
   - Click "Download" to update individual models

3. **Sync Targets Tab**: Configure machines to sync to
   - Add targets with SSH host and models path
   - Set VRAM limits to filter which models sync
   - Click "Sync Now" to sync to a specific target

4. **Settings Tab**: Configure global options
   - Enable/disable auto-update
   - Set check interval
   - Toggle download verification and backups

### API Endpoints

| Endpoint | Method | Description |
|----------|--------|-------------|
| `/modelsync/models` | GET | List all tracked models |
| `/modelsync/models/scan` | POST | Scan filesystem for models |
| `/modelsync/models/{id}` | GET | Get model details |
| `/modelsync/models/{id}` | PATCH | Update model settings |
| `/modelsync/check` | POST | Check models for updates |
| `/modelsync/download/{id}` | POST | Download model update |
| `/modelsync/config` | GET/POST | Get/set configuration |
| `/modelsync/targets` | GET/POST | List/add sync targets |
| `/modelsync/targets/{id}` | DELETE | Remove sync target |
| `/modelsync/sync/{id}` | POST | Sync to specific target |
| `/modelsync/sync` | POST | Sync to all targets |
| `/modelsync/status` | GET | Get overall status |

### Setting Model Sources

To enable update checking for a model, set its source URL:

```bash
curl -X PATCH http://localhost:8188/modelsync/models/1 \
  -H "Content-Type: application/json" \
  -d '{"source_url": "https://huggingface.co/stabilityai/sdxl-vae/resolve/main/sdxl_vae.safetensors"}'
```

Or use the optional workflow node "Set Model Source".

### Configuring Sync Targets

Add a sync target:

```bash
curl -X POST http://localhost:8188/modelsync/targets \
  -H "Content-Type: application/json" \
  -d '{
    "name": "secondary-gpu",
    "ssh_host": "user@192.168.1.100",
    "model_path": "/home/user/ComfyUI/models",
    "max_vram_mb": 16000
  }'
```

### VRAM Filtering

Models are filtered based on estimated VRAM requirements:

| Pattern | Estimated VRAM |
|---------|----------------|
| wan, hunyuan, mochi | 48 GB |
| flux-dev | 24 GB |
| flux-schnell | 12 GB |
| sdxl | 12 GB |
| sd15, sd1.5 | 6 GB |
| lora | 2 GB |

Models larger than the target's `max_vram_mb` will be skipped during sync.

## Configuration

Default configuration values:

| Key | Default | Description |
|-----|---------|-------------|
| `auto_update_enabled` | false | Enable automatic updates |
| `check_interval_hours` | 24 | Hours between update checks |
| `cache_ttl_hours` | 6 | Hours to cache version info |
| `verify_downloads` | true | Verify file hash after download |
| `backup_before_update` | true | Create .bak before replacing |
| `max_concurrent_downloads` | 2 | Max simultaneous downloads |

## Optional Workflow Nodes

Three utility nodes are included for workflow integration:

- **Model Info (ModelSync)** - Output model metadata (source URL, VRAM, update status)
- **Check Model Update (ModelSync)** - Check if a specific model has updates
- **Set Model Source (ModelSync)** - Associate a model with its source URL

## Database

ModelSync uses SQLite for metadata storage. The database is stored at:
```
ComfyUI/custom_nodes/ComfyUI-ModelSync/modelsync.db
```

## Requirements

- Python 3.10+
- ComfyUI
- `huggingface_hub>=0.20.0`
- `aiohttp>=3.9.0`
- SSH and rsync (for multi-GPU sync)

## License

MIT License - see LICENSE file for details.

## Contributing

Contributions are welcome! Please open an issue or pull request on GitHub.

## Support

For issues and feature requests, please use the [GitHub issue tracker](https://github.com/benfinklea/ComfyUI-ModelSync/issues).