Generate clean, consistent PRs from commits - powered by AI and your git history.

📖 Full Documentation

Features ✨

• Multi-provider AI: Groq, Cerebras, Google Gemini, and OpenAI-compatible APIs
• Smart commit filtering: Excludes noise commits
• PR template support: Use your existing templates
• GitHub CLI integration: Generate gh pr create commands
• Multilingual: 12+ languages supported
• Custom context: Guide AI generation style

Installation 📦

npm install -g lazypr

Requires Node.js >= 22. See installation guide →

Quick start ⚡

1. Configure interactively (recommended for first-time setup):

   lazypr config

   This opens a guided menu where you can pick provider + model together and enter your API key with masked input (never echoed to terminal or shell history).

   Or use scriptable commands for CI/dotfiles:

   lazypr config set GROQ_API_KEY=<your-key>

2. Generate a PR:

   lazypr              # compares against main
   lzp                 # short alias
   lazypr develop      # compare against different branch

Complete quick start guide → | Usage examples →

Configuration ⚙️

lazypr supports two configuration flows that work alongside each other:

Interactive (recommended for setup)

lazypr config          # opens a Clack-powered interactive menu

The interactive menu lets you:

• Pick provider + model together with a combined selector (Groq, Cerebras, Google, OpenAI, or custom/local)
• Enter API keys with masked input — keys are never echoed to the terminal or shell history
• Configure locale, default branch, context, commit filtering, custom labels, retries, and timeout
• View the current config with secrets masked

Scriptable (CI and dotfiles)

lazypr config set KEY=VALUE    # Set configuration
lazypr config get KEY          # Get configuration
lazypr config remove KEY       # Remove configuration key
lazypr config list             # Show all settings

Examples:

# Set API keys (key is visible in shell history — use interactive flow to avoid this)
lazypr config set GROQ_API_KEY=<key>

# Switch provider and model
lazypr config set PROVIDER=google
lazypr config set MODEL=gemini-3.5-flash

# Use a local/custom model with OpenAI-compatible API
lazypr config set PROVIDER=openai
lazypr config set OPENAI_BASE_URL=http://localhost:11434/v1
lazypr config set MODEL=llama3.2

# View config
lazypr config list

Common settings:

• PROVIDER - AI provider (groq, cerebras, google, or openai)
• GROQ_API_KEY / CEREBRAS_API_KEY / GOOGLE_GENERATIVE_AI_API_KEY / OPENAI_API_KEY - API keys
• LOCALE - Output language (en, es, pt, fr, etc.)
• MODEL - AI model to use (any non-empty string; custom/local model IDs supported)
• FILTER_COMMITS - Smart commit filtering (true or false)
• CONTEXT - Custom guidance for AI generation
• OPENAI_BASE_URL - Custom endpoint for OpenAI-compatible APIs (Ollama, LM Studio, etc.)

View all settings → | Configuration examples →

Advanced Features

• Smart Commit Filtering - Automatically excludes merge commits, dependency updates, and formatting changes
• PR Templates - Use your existing .github templates
• GitHub CLI Integration - Generate gh pr create commands
• Multilingual Support - 12+ languages available
• Context Guidance - Customize AI generation style
• GitHub Actions - Automate PR generation in CI/CD
• OpenAI-Compatible APIs - Use local providers (Ollama, LM Studio) or third-party services

CLI Reference

lazypr [target] [options]

Options:
  -t, --template [name]      Use a PR template
  -l, --locale <language>    Output language
  -c, --context <text>       Custom AI guidance
  --gh                       Generate gh pr create command
  --no-filter                Disable smart filtering
  -u, --usage                Show token usage
  -h, --help                 Display help

View complete CLI reference →

Development

bun install      # Install dependencies
bun test         # Run tests
bun run build    # Build for production

Uses Bun for development, Node.js >= 22 for runtime.

Contributing 🤝

Issues and PRs are welcome. Open one on the repository’s Issues page.

License 📄

MIT © Raul Carini. See the LICENSE file for details.