# 🚀 GitHub Release Upload for Quantized Models
This directory contains scripts to automatically upload the quantized FastAI models to GitHub releases with device authentication.
## 📋 Prerequisites
### 1. GitHub CLI Installation
**Windows:**
```powershell
winget install GitHub.cli
```
**macOS:**
```bash
brew install gh
```
**Linux:**
```bash
# Ubuntu/Debian
curl -fsSL https://cli.github.com/packages/githubcli-archive-keyring.gpg | sudo dd of=/usr/share/keyrings/githubcli-archive-keyring.gpg
echo "deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/githubcli-archive-keyring.gpg] https://cli.github.com/packages stable main" | sudo tee /etc/apt/sources.list.d/github-cli.list > /dev/null
sudo apt update && sudo apt install gh
# Other distributions: https://cli.github.com/manual/installation
```
### 2. Model Preparation
Ensure models are prepared for release:
```bash
uv run python scripts/prepare_model_release.py
```
This creates:
- `dist/fastai_intent_classifier_quantized.pkl` (63.0 MB) - Primary model
- `dist/fastai_intent_classifier.pkl` (121.1 MB) - Fallback model
## 🚀 Upload Methods
### Method 1: Python Script (Recommended)
```bash
uv run python scripts/upload_to_github.py
```
**Features:**
- Interactive wizard interface
- Device authentication flow
- Progress tracking
- Automatic release notes generation
- Error handling and validation
### Method 2: PowerShell Script (Windows)
```powershell
.\scripts\upload_to_github.ps1
```
**Features:**
- Windows-optimized interface
- Automatic model preparation
- Colorized output
- Error handling
### Method 3: Batch Script (Windows)
```cmd
scripts\upload_to_github.bat
```
**Features:**
- Simple double-click execution
- Basic error handling
- Virtual environment activation
## 🔐 Authentication Process
The script will guide you through GitHub authentication:
1. **Check existing auth:** Verifies if you're already logged in
2. **Device flow:** Opens browser for GitHub login if needed
3. **Verification:** Confirms authentication success
**No tokens or passwords needed!** Uses secure device authentication.
## 📤 Upload Process
### Interactive Flow
1. **Pre-checks:** Validates GitHub CLI, authentication, and model files
2. **Release info:** Prompts for tag, title, and prerelease status
3. **Notes generation:** Creates comprehensive release notes automatically
4. **Confirmation:** Shows preview before upload
5. **Creation:** Creates GitHub release
6. **Upload:** Uploads both model files with progress tracking
### Example Session
```
🚀 GITHUB RELEASE UPLOAD WIZARD
============================================================
✅ GitHub CLI found: gh version 2.40.1
✅ Already authenticated with GitHub CLI
📁 Checking model files...
✅ fastai_intent_classifier_quantized.pkl (63.0 MB)
✅ fastai_intent_classifier.pkl (121.1 MB)
📊 Total upload size: 184.1 MB
📋 Existing releases:
• v1.0.0
📝 Release Information:
Last release: v1.0.0
Enter release tag (e.g., v1.1.0): v1.1.0
Enter release title (default: 'IFS Cloud MCP Server v1.1.0'):
Is this a prerelease? (y/N): n
📄 Generated release notes preview:
----------------------------------------
# IFS Cloud MCP Server v1.1.0
## 🚀 Quantized Model Release
This release features the new **quantized FastAI intent classifier**...
----------------------------------------
Proceed with release creation and upload? (Y/n): y
🚀 Creating release v1.1.0...
✅ Release v1.1.0 created successfully!
📤 Uploading assets to release v1.1.0...
Uploading fastai_intent_classifier_quantized.pkl...
✅ fastai_intent_classifier_quantized.pkl uploaded successfully
Uploading fastai_intent_classifier.pkl...
✅ fastai_intent_classifier.pkl uploaded successfully
🎉 SUCCESS! Release v1.1.0 created and assets uploaded!
🔗 View release: https://github.com/graknol/ifs-cloud-core-mcp-server/releases/tag/v1.1.0
```
## 📋 Generated Release Notes
The script automatically generates comprehensive release notes including:
- **Performance metrics:** Size reduction, speed improvement
- **Usage instructions:** Download commands and configuration
- **Migration notes:** Upgrade process for existing users
- **Technical details:** Quantization benefits and compatibility
## 🔧 Configuration
### Repository Settings
Default: `graknol/ifs-cloud-core-mcp-server`
To change repository:
```python
uploader = GitHubReleaseUploader("your-username/your-repo")
```
### Model Files
Default paths:
- `dist/fastai_intent_classifier_quantized.pkl` - Primary model
- `dist/fastai_intent_classifier.pkl` - Fallback model
## ⚠️ Troubleshooting
### GitHub CLI Not Found
```bash
# Check installation
gh --version
# Reinstall if needed
winget install GitHub.cli # Windows
brew install gh # macOS
```
### Authentication Issues
```bash
# Check auth status
gh auth status
# Re-authenticate
gh auth login --web
```
### Model Files Missing
```bash
# Prepare models
uv run python scripts/prepare_model_release.py
# Check files exist
ls dist/
```
### Upload Failures
- Check internet connection
- Verify repository permissions
- Ensure sufficient GitHub storage quota
- Try smaller uploads if network issues
## 🎯 Post-Upload Steps
After successful upload:
1. **Update model downloader:**
```python
# In src/ifs_cloud_mcp_server/model_downloader.py
DEFAULT_TAG = "v1.1.0" # Update to new release
```
2. **Test download:**
```bash
# Test quantized model download
uv run python -m ifs_cloud_mcp_server.model_downloader --tag v1.1.0
```
3. **Update documentation:**
- README.md
- Installation instructions
- Version references
## 📊 Release Benefits
**For Users:**
- 48% smaller downloads
- 22% faster inference
- Same search accuracy
- Automatic fallback
**For Deployment:**
- Faster CI/CD pipelines
- Reduced bandwidth costs
- Better user experience
- Easier distribution
---
_The upload scripts handle all the complexity - just run and follow the interactive prompts!_ 🚀
