Skip to main content

​
Configuration

CodeWolf is designed to be flexible and self-hosted.
All configuration is handled through environment variables and can be extended over time.

​
Environment Variables

All configuration is managed via the .env file in the root directory.

​
GitHub Configuration

GITHUB_APP_ID=
GITHUB_PRIVATE_KEY_PATH=

​
LLM Configuration

HF_TOKEN=
HF_MODEL=

​
LLM providers

CodeWolf follows a BYOK (Bring Your Own Key) approach. Currently supported:
  • Hugging Face (via app/llm/huggingface.js)
More providers will be added soon.

​
How it works

  • CodeWolf sends structured prompts to the configured model
  • The model returns analysis (bugs, security issues, suggestions)
  • Output is normalized and posted to the PR

​
Future Support

Planned support includes:
  • OpenAI
  • Anthropic
  • Gemini
  • Local models running with Ollama, etc.

​
Model Selection

Choosing the right model affects:
  • Review quality
  • Speed
  • Cost

​
Recommendations

  • Use larger models for deeper analysis
  • Use faster models for quicker feedback cycles

​
Future customization

Planned enhancements:
  • Custom prompts
  • Rule-based review guidelines
  • Team-specific coding standards

​
File Filtering

Before sending data to the LLM, CodeWolf filters out:
  • Large files
  • Generated code
  • Unsupported formats
This improves signal quality and reduces unnecessary processing.

​
Extending Configuration

The system is designed to be modular. You can extend:
  • Add new LLM providers in app/llm/
  • Modify review logic in app/core/reviewEngine.js
  • Adjust filtering rules based on your needs
Configuration in CodeWolf is intentionally simple today, with flexibility to evolve as the system grows.