https://github.com/yonasvalentin/codecontext-ai
🔒 Privacy-first AI models for code documentation. Your code stays on YOUR machine. Forever.
https://github.com/yonasvalentin/codecontext-ai
ai documentation local-ai machine-learning ollama open-source privacy privacy-first
Last synced: 2 months ago
JSON representation
🔒 Privacy-first AI models for code documentation. Your code stays on YOUR machine. Forever.
- Host: GitHub
- URL: https://github.com/yonasvalentin/codecontext-ai
- Owner: YonasValentin
- License: mit
- Created: 2025-07-30T17:48:03.000Z (2 months ago)
- Default Branch: main
- Last Pushed: 2025-07-30T17:55:02.000Z (2 months ago)
- Last Synced: 2025-07-30T20:35:10.512Z (2 months ago)
- Topics: ai, documentation, local-ai, machine-learning, ollama, open-source, privacy, privacy-first
- Language: Python
- Size: 64.5 KB
- Stars: 1
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Changelog: CHANGELOG.md
- Contributing: CONTRIBUTING.md
- License: LICENSE
Awesome Lists containing this project
README
# CodeContext AIâ„¢
Privacy-first AI models for automated code documentation generation. Train and run specialized documentation models locally without external dependencies.
## Features
- **Local Processing**: All inference runs locally with no external API calls
- **Specialized Models**: Fine-tuned models for README, API documentation, and changelog generation
- **Privacy Focused**: No data transmission or telemetry collection
- **Production Ready**: Docker support, comprehensive testing, and CI/CD integration
## Installation
### Quick Setup (Recommended)
```bash
# Automated setup with virtual environment
chmod +x setup_environment.sh
./setup_environment.sh
source venv/bin/activate
```
### Manual Setup
```bash
# Create virtual environment
python3 -m venv venv
source venv/bin/activate
# Install dependencies
pip install torch transformers pyyaml
pip install -r requirements.txt
pip install -e .
```
## Quick Start
```bash
# Run advisory analysis demo
python demo_advisory.py
# Train advisory model
make train MODEL=advisory
# Analyze code with advisory system
python -m codecontext_ai.guidance_cli analyze myfile.py --type refactor
# Scan directory for issues
python -m codecontext_ai.guidance_cli scan ./src --type security
```
## Architecture
### Models
| Model | Purpose | Base | Training |
|-------|---------|------|----------|
| codecontext-readme-7b | Project documentation | CodeLlama-7B | QLoRA fine-tuning |
| codecontext-api-7b | API documentation | CodeLlama-7B | QLoRA fine-tuning |
| codecontext-changelog-7b | Release notes | CodeLlama-7B | QLoRA fine-tuning |
### Training Pipeline
- **Base Model**: CodeLlama-7B with 4-bit quantization
- **Fine-tuning**: Parameter Efficient Fine-Tuning (PEFT) with LoRA
- **Optimization**: GGUF format for efficient local inference
- **Evaluation**: Multi-metric assessment (BLEU, ROUGE-L, semantic similarity)
## Development
### Setup
```bash
# Development environment
make install-dev
# Prepare training data
make prepare-data
# Run tests
make test
# Type checking and linting
make lint && make typecheck
```
### Training
```bash
# Train specific model
make train MODEL=readme
# Train all models
make train-all
# Convert to GGUF format
make convert-gguf MODEL=models/codecontext-readme-7b
```
### Evaluation
```bash
# Evaluate model performance
make evaluate MODEL=models/codecontext-readme-7b.gguf
# Run comprehensive benchmarks
make benchmark
```
## Configuration
Training configurations are stored in `configs/`:
- `readme.yaml`: README model parameters
- `api.yaml`: API documentation model parameters
- `changelog.yaml`: Changelog model parameters
## Privacy
- No external API calls during training or inference
- All processing occurs locally
- Open source codebase for transparency
- No data collection or telemetry
## Performance
Benchmark results on RTX 4090:
- Inference speed: 30-40 tokens/second
- Memory usage: 4-8GB RAM
- Model size: ~4GB per specialized model
## Contributing
1. Fork the repository
2. Create feature branch
3. Add tests for new functionality
4. Run test suite: `make test`
5. Submit pull request
See [CONTRIBUTING.md](CONTRIBUTING.md) for detailed guidelines.
## License
MIT License - see [LICENSE](LICENSE) for details.
## Support
- Issues: [GitHub Issues](https://github.com/YonasValentin/codecontext-ai/issues)
- Discussions: [GitHub Discussions](https://github.com/YonasValentin/codecontext-ai/discussions)