Powered by Qwen-2.5-7B (Foundry Local) + Claude Sonnet 4.5 (GitHub Copilot SDK)
GenGitHubRepoPPT is an innovative hybrid AI solution that combines the power of local AI models with cloud-based AI agents to automatically generate professional PowerPoint presentations from GitHub repository README files.
This project leverages a two-stage AI pipeline:
- Content Generation Stage: Uses Qwen-2.5-7B (a 7-billion parameter language model) running locally via Microsoft Foundry Local to analyze GitHub README files and generate comprehensive PPT outlines and content
- Presentation Creation Stage: Utilizes GitHub Copilot SDK to transform the generated outline into a fully-formatted, professional PowerPoint presentation
- โ Local Privacy: Sensitive repository analysis happens locally with Qwen-2.5-7B CPU inference
- โ Professional Quality: GitHub Copilot SDK ensures polished, well-designed presentations
- โ Fast Generation: Complete PPT creation in approximately 5 minutes
- โ Multi-language Support: Supports English, Simplified Chinese, and Traditional Chinese
- โ Cost-Effective: Leverages free local models with minimal cloud API usage
- README Fetching: User provides GitHub repository URL โ System fetches README.md content
- Outline Generation: README content โ Foundry Local (Qwen-2.5-7B) โ Structured PPT outline
- PPT Creation: Outline โ GitHub Copilot SDK โ Professional PowerPoint file
- Download: User receives ready-to-use PPTX presentation
Foundry Local is a local AI runtime that enables running large language models on your own hardware with OpenAI-compatible APIs.
Features used in this project:
- Local deployment of Qwen-2.5-7B model
- CPU-based inference for accessibility
- OpenAI-compatible chat completion API
- No external API calls for content analysis
Why Foundry Local?
- Privacy-first approach for repository analysis
- No cloud dependencies for content generation
- Free and open-source
- Supports various quantized models for CPU inference
GitHub Copilot SDK provides programmatic access to GitHub Copilot's AI capabilities, enabling developers to build AI-powered applications.
Features used in this project:
- Integration with Claude Sonnet 4.5 model
- Custom skill system for PPT generation
- Streaming responses for real-time feedback
- Professional document creation capabilities
Why GitHub Copilot SDK?
- Access to state-of-the-art models (Claude Sonnet 4.5)
- Specialized skills for complex tasks
- High-quality output formatting
- Seamless integration with development workflows
- Python 3.10 or higher
- Microsoft Foundry Local running with Qwen-2.5-7B model
- GitHub Copilot SDK installed and configured
- Git (for cloning the repository)
# Clone the repository
git clone https://github.com/kinfey/GenGitHubRepoPPT.git
cd GenGitHubRepoPPT
# Navigate to app directory
cd app
# Install Python dependencies
pip install -r requirements.txt
# Install GitHub Copilot SDK (if not already installed)
# Follow instructions at: https://github.com/github/copilot-sdkCreate or verify your .env file in the app directory:
FOUNDRYLOCAL_ENDPOINT=Your Foundry Service Endpoint such as http://localhost:51498/v1
FOUNDRYLOCAL_API_KEY=nokey
FOUNDRYLOCAL_MODEL_DEPLOYMENT_NAME=qwen2.5-7b-instruct-generic-cpu:4# Make sure Foundry Local is running with Qwen-2.5-7B model
# Then start the Flask server
python app.pyVisit http://localhost:5001 in your browser.
GenGitHubRepoPPT/
โโโ README.md # This file
โโโ app/
โ โโโ app.py # Flask backend server
โ โโโ requirements.txt # Python dependencies
โ โโโ README.md # App-specific documentation
โ โโโ .env # Environment configuration
โ โโโ templates/
โ โ โโโ index.html # Main web interface
โ โโโ static/
โ โ โโโ css/
โ โ โ โโโ style.css # Application styles
โ โ โโโ js/
โ โ โโโ app.js # Main application logic
โ โ โโโ i18n.js # Internationalization
โ โโโ ppt/ # Generated PPT output directory
โ โโโ generated_ppts/ # Alternative output directory
โ โโโ .copilot_skills/ # Copilot SDK custom skills
โ โโโ ppt/
โ โโโ SKILL.md # PPT generation skill definition
- Open the web application at
http://localhost:5001 - Enter a valid GitHub repository URL (e.g.,
https://github.com/microsoft/foundry-local) - Click "Fetch README"
- The system generates a structured PPT outline using Qwen-2.5-7B
- Review the outline content
- Select your preferred language (English, ็ฎไฝไธญๆ, or ็น้ซไธญๆ)
- Click "Regenerate" if you want a different outline
- Click "Generate PPT" when satisfied
- Wait for the PPT generation (typically 3-5 minutes)
- Download your professional PowerPoint presentation
- Review and customize as needed
| Endpoint | Method | Description |
|---|---|---|
/ |
GET | Main web interface |
/api/fetch-readme |
POST | Fetch README from GitHub repository |
/api/generate-outline |
POST | Generate PPT outline using Foundry Local |
/api/generate-ppt |
POST | Create PPT file using GitHub Copilot SDK |
/api/download-ppt/<session_id> |
GET | Download generated PowerPoint file |
- Microsoft Foundry Local GitHub - Local AI runtime documentation
- GitHub Copilot SDK - Copilot SDK documentation and examples
- Qwen Models - Qwen 2.5 model family information
- Qwen-2.5-7B-Instruct - Model card on Hugging Face
- Claude Sonnet 4.5 - Anthropic's Claude model
- Flask Documentation - Python web framework
- OpenAI API Reference - API compatibility reference
- Edge AI For Beginners - Related technologies
- Microsoft Foundry Local - Related technologies
Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.
This project is licensed under the MIT License - see the LICENSE file for details.
- Microsoft for Foundry Local framework
- GitHub for Copilot SDK
- Alibaba Cloud for the Qwen model family
- Anthropic for Claude AI models
Made with โค๏ธ by kinfey
โญ Star this repo if you find it helpful!