PreReview

AI-powered code review before you commit. Uses the GitHub Copilot SDK to analyze your staged changes and provide actionable suggestions.

Features

• 🔍 AI-Powered Review - Uses GitHub Copilot SDK (Claude, GPT-5, Gemini, Grok) to review your code
• 🎯 Interactive Workflow - Fix, skip, or review suggestions one by one
• 🪝 Git Hook Integration - Automatically runs before each commit
• ⚙️ Configurable - Customize models, strictness, and ignore patterns
• 🎨 Beautiful Terminal UI - Clean, colorful output with lipgloss

Prerequisites

• GitHub Copilot subscription (Free, Pro, Business, or Enterprise)

• Copilot CLI installed:

  brew install copilot-cli

• Logged in to GitHub:

  copilot auth login

Installation

Using Go

go install github.com/emilushi/prereview@latest

From Source

git clone https://github.com/emilushi/prereview.git
cd prereview
go build -o prereview .
sudo mv prereview /usr/local/bin/

Quick Start

1. Review staged changes:

   git add .
   prereview

2. Install as pre-commit hook:

   prereview install

3. Create config file:

   prereview config init

Usage

Commands

Command	Description
prereview	Review staged changes interactively
prereview review	Same as above
prereview doctor	Check dependencies and setup
prereview install	Install as git pre-commit hook
prereview uninstall	Remove git pre-commit hook
prereview config init	Create default config file
prereview config list	Show current configuration
prereview config set <key> <value>	Set a config value
prereview config get <key>	Get a config value

Flags

Flag	Description
--model	AI model to use (claude, gpt-4, gemini, grok)
--strict	Require all issues to be fixed before committing
--tolerance	Review tolerance: strict, moderate, relaxed (default: moderate)
--force	Force commit even with unresolved suggestions
--verbose	Show detailed output
--config	Path to config file

Tolerance Levels

PreReview supports three tolerance levels to reduce false positives:

Level	Description
strict	Reports all potential issues including style nitpicks
moderate	Reports bugs, security issues, and significant quality concerns (default)
relaxed	Only reports definite bugs and critical security issues

# Use relaxed mode for less noise
prereview --tolerance relaxed

# Or set in config
prereview config set tolerance relaxed

Confidence Levels

Each suggestion includes a confidence level:

• high - Definite issue, should be fixed (>95% confident)
• medium - Likely an issue but could be intentional (70-95% confident)
• low - Possible issue, may be false positive (<70% confident)

Only high-confidence errors block commits by default. Low-confidence suggestions are shown but don't prevent you from committing.

Interactive Review

When reviewing, you have these options for each suggestion:

• [f]ix - Apply the suggested fix (if available)
• [s]kip - Skip this suggestion
• [v]iew - View the diff for context
• [q]uit - Abort the review

After reviewing all suggestions:

• [y]es - Proceed with commit
• [n]o - Abort
• [r]e-review - Review again (after making manual fixes)

Configuration

Create a .prereviewrc.yaml in your project root or ~/.prereviewrc.yaml for global settings:

# AI model to use
model: gpt-4

# Require all issues to be fixed
strict: false

# Review tolerance: strict, moderate, relaxed
# - strict: Report all potential issues
# - moderate: Report bugs and significant issues (default)
# - relaxed: Only report definite bugs and security issues
tolerance: moderate

# What severity level blocks commits: errors, warnings, all, none
# Default: errors (only high-confidence errors block)
block_on: errors

# Show detailed output
verbose: false

# Files to ignore
ignore_patterns:
  - "*.min.js"
  - "vendor/*"
  - "node_modules/*"

# Max file size to review (bytes)
max_file_size: 100000

# Additional coding standard files to detect (beyond auto-detected ones)
# These are file paths relative to repo root
coding_standards:
  - ".custom-lint.json"
  - "config/phpcs-custom.xml"

# Project-specific hints for the AI reviewer
# Use this to provide context that reduces false positives
project_hints:
  - "Factory methods may return different subtypes based on input"
  - "Data is sanitized at input time, not output time"
  - "We use dependency injection throughout the codebase"

Reducing False Positives

If you're experiencing too many false positives:

1. Use relaxed tolerance:

   prereview config set tolerance relaxed

2. Add project-specific context:

   project_hints:
     - "This project uses the Repository pattern"
     - "getEntity() returns different entity types based on the ID prefix"
     - "All user input is sanitized before storage"

3. Change what blocks commits:

   block_on: errors  # Only block on high-confidence errors

4. Force commit when needed:

   git commit  # Will prompt: "Proceed despite issues? [y/N]"
   # Or bypass completely:
   prereview --force
   git commit --no-verify

Authentication

PreReview uses the GitHub Copilot CLI for authentication. The CLI handles all authentication automatically using your GitHub Copilot subscription.

Setup

1. Install Copilot CLI:

   brew install copilot-cli

2. Login to GitHub:

   copilot auth login

3. Verify authentication:

   copilot auth status

That's it! PreReview will use your Copilot CLI credentials automatically.

Examples

Basic Review

$ git add src/utils.js
$ prereview

🔍 Reviewing 1 changed file(s)...

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
📄 src/utils.js [1/2]
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  Line 45

  ⚠ Missing null check

  Accessing property on potentially null value.

  Suggested fix:
  if (user?.email) { ... }

  best-practice

  [f]ix | [s]kip | [v]iew diff | [q]uit: f

  ✓ Applied fix

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Summary
  ✓ 1 fixed
  ⏭ 0 skipped
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Proceed with commit? [y]es | [n]o | [r]e-review: y

✓ Review complete: 1 fixed, 0 skipped

Strict Mode

prereview --strict

In strict mode, you cannot proceed with commit if any suggestions are skipped.

Development

Requirements

• Go 1.21+
• GitHub Copilot license

Building

go build -o prereview .

Testing

go test ./...

License

MIT License - see LICENSE for details.

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Run tests
5. Submit a pull request

Acknowledgments

• GitHub Copilot for AI capabilities
• Cobra for CLI framework
• Viper for configuration
• Lipgloss for terminal styling