Spaces:
Paused
Paused
| # PRIX CLI - Local AI Code Review | |
| Run AI-powered code reviews locally before pushing to GitHub. | |
| ## Installation | |
| ```bash | |
| npm install | |
| npm run build | |
| npm link | |
| ``` | |
| ## Quick Start | |
| ```bash | |
| # Review staged files | |
| prix-cli --staged | |
| # Review changes since last commit | |
| prix-cli --diff HEAD~1 | |
| # Review specific files | |
| prix-cli --files src/app.ts,lib/util.ts | |
| ``` | |
| ## Pre-commit Hook | |
| Install the pre-commit hook for automatic reviews: | |
| ```bash | |
| npm run install-hooks | |
| ``` | |
| Or manually: | |
| ```bash | |
| ./scripts/install-hooks.sh | |
| ``` | |
| ## Configuration | |
| Environment variables: | |
| | Variable | Default | Description | | |
| | --------------------- | -------------------------------- | -------------------------------------------- | | |
| | `GROQ_API_KEY` | - | **Required.** Groq API key | | |
| | `AI_API_KEY` | - | Alternative API key | | |
| | `API_BASE_URL` | `https://api.groq.com/openai/v1` | API base URL | | |
| | `PRIX_MIN_SEVERITY` | `minor` | Minimum severity (critical/major/minor/info) | | |
| | `PRIX_MIN_CONFIDENCE` | `60` | Minimum confidence 0-100 | | |
| | `PRIX_DISABLE` | `false` | Set to `true` to skip reviews | | |
| ## Options | |
| ``` | |
| --staged Review staged files (git diff --cached) | |
| --diff <ref> Review files changed since given git ref | |
| --files <paths> Review specific files (comma-separated) | |
| --model <name> LLM model to use (default: llama-3.1-8b-instant) | |
| --min-confidence <n> Minimum confidence threshold (0-100) | |
| --min-severity <lvl> Minimum severity to report (critical|major|minor|info) | |
| --disable-review Skip code review, only generate summary | |
| --disable-notes Skip release notes generation | |
| --max-files <n> Maximum files to review (0 = unlimited) | |
| --path-filters <p> Path filters (comma-separated, e.g., "src/**,lib/**") | |
| --system-message <m> Additional system message | |
| --language <lang> Output language (default: en-US) | |
| --output <fmt> Output format: console (default) or json | |
| --help, -h Show help message | |
| ``` | |
| ## Exit Codes | |
| | Code | Meaning | | |
| | ---- | ------------------------------------- | | |
| | `0` | Success, no critical issues | | |
| | `1` | Critical issues found (blocked in CI) | | |
| ## Examples | |
| ### Basic Pre-commit Review | |
| ```bash | |
| prix-cli --staged | |
| ``` | |
| ### Only Critical Issues | |
| ```bash | |
| prix-cli --staged --min-severity critical --min-confidence 80 | |
| ``` | |
| ### JSON Output for CI | |
| ```bash | |
| prix-cli --staged --output json | |
| ``` | |
| ### Review Specific Paths | |
| ```bash | |
| prix-cli --diff HEAD --path-filters "src/**,tests/**" | |
| ``` | |
| ## Workflow | |
| 1. **Write code** and stage files with `git add` | |
| 2. **Run review** with `prix-cli --staged` | |
| 3. **Fix issues** or override with `git commit --no-verify` | |
| 4. **Push** when satisfied with the review | |
| ## GitHub Integration | |
| The CLI works alongside the GitHub App: | |
| - **Pre-commit (local)**: Catch obvious issues early, fast feedback | |
| - **PR Review (GitHub App)**: Full context, semantic search, detailed analysis | |
| Both use the same underlying AI models and confidence calibration. | |