AI Code Reviewer

A reusable GitHub Action that automatically reviews Pull Requests using AI models via OpenRouter. Model-agnostic - use Claude, GPT-4, Gemini, Llama, or any model available on OpenRouter.

Features

• Model Agnostic: Use any AI model available on OpenRouter
• Reusable: Host once, use across all your repositories
• Customizable: Configure system prompts, review focus areas, and more
• Automatic Reviews: Triggers on PR open, sync, and reopen

Quick Start

Using the Action in Your Repository

Create .github/workflows/ai-review.yml in your repository:

name: AI Code Review

on:
  pull_request:
    types: [opened, synchronize, reopened]

permissions:
  contents: read
  pull-requests: write

jobs:
  review:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: AI Code Review
        uses: YOUR_USERNAME/ai-code-reviewer@v1
        with:
          openrouter_api_key: ${{ secrets.OPENROUTER_API_KEY }}
          github_token: ${{ secrets.GITHUB_TOKEN }}

Replace YOUR_USERNAME/ai-code-reviewer with your published action's path.

Setup

1. Get an OpenRouter API Key

1. Go to OpenRouter
2. Sign up or log in
3. Navigate to API Keys
4. Create a new API key

2. Add the Secret to Your Repository

1. Go to your GitHub repository → Settings → Secrets and variables → Actions
2. Add a new repository secret:
   • Name: OPENROUTER_API_KEY
   • Value: Your OpenRouter API key

Configuration Options

Inputs

Input	Required	Default	Description
openrouter_api_key	Yes	-	Your OpenRouter API key
github_token	Yes	-	GitHub token for posting comments
model	No	anthropic/claude-3.5-sonnet	AI model to use
system_prompt	No	-	Override the entire system prompt
custom_prompt	No	-	Additional instructions appended to the prompt
review_focus	No	-	Comma-separated focus areas
temperature	No	0.3	Model temperature (0-1)
max_diff_size	No	100000	Max diff size in characters
post_comment	No	true	Whether to post review as PR comment

Outputs

Output	Description
review	The generated review content
model_used	The AI model that was used

Usage Examples

Basic Usage

- uses: YOUR_USERNAME/ai-code-reviewer@v1
  with:
    openrouter_api_key: ${{ secrets.OPENROUTER_API_KEY }}
    github_token: ${{ secrets.GITHUB_TOKEN }}

Using a Different Model

- uses: YOUR_USERNAME/ai-code-reviewer@v1
  with:
    openrouter_api_key: ${{ secrets.OPENROUTER_API_KEY }}
    github_token: ${{ secrets.GITHUB_TOKEN }}
    model: 'openai/gpt-4-turbo'

Custom System Prompt

- uses: YOUR_USERNAME/ai-code-reviewer@v1
  with:
    openrouter_api_key: ${{ secrets.OPENROUTER_API_KEY }}
    github_token: ${{ secrets.GITHUB_TOKEN }}
    system_prompt: |
      You are a senior security engineer reviewing code for vulnerabilities.
      Focus exclusively on security issues and rate their severity.
      Ignore style and formatting issues.

Focus on Specific Areas

- uses: YOUR_USERNAME/ai-code-reviewer@v1
  with:
    openrouter_api_key: ${{ secrets.OPENROUTER_API_KEY }}
    github_token: ${{ secrets.GITHUB_TOKEN }}
    review_focus: 'security,performance,bugs'

Available focus areas:

• security - Security vulnerabilities and unsafe practices
• performance - Performance bottlenecks and inefficiencies
• bugs - Logic errors, edge cases, runtime issues
• quality - Code smells and anti-patterns
• readability - Code clarity and maintainability
• testing - Test coverage and quality
• documentation - Missing or outdated docs
• architecture - Design patterns and modularity
• accessibility - UI accessibility issues
• error-handling - Error handling and logging

Additional Instructions

- uses: YOUR_USERNAME/ai-code-reviewer@v1
  with:
    openrouter_api_key: ${{ secrets.OPENROUTER_API_KEY }}
    github_token: ${{ secrets.GITHUB_TOKEN }}
    custom_prompt: |
      Pay special attention to:
      - SQL injection vulnerabilities
      - Proper input validation
      - Error messages that might leak sensitive info

Using Review Output in Subsequent Steps

- name: AI Code Review
  id: review
  uses: YOUR_USERNAME/ai-code-reviewer@v1
  with:
    openrouter_api_key: ${{ secrets.OPENROUTER_API_KEY }}
    github_token: ${{ secrets.GITHUB_TOKEN }}
    post_comment: 'false'

- name: Use Review Output
  run: |
    echo "Model used: ${{ steps.review.outputs.model_used }}"
    echo "Review: ${{ steps.review.outputs.review }}"

Available Models

Popular models on OpenRouter:

Provider	Model ID	Notes
Anthropic	anthropic/claude-3.5-sonnet	Default, excellent for code
Anthropic	anthropic/claude-3-opus	Most capable
OpenAI	openai/gpt-4-turbo	Fast and capable
OpenAI	openai/gpt-4o	Latest GPT-4
Google	google/gemini-pro-1.5	Good balance
Meta	meta-llama/llama-3.1-70b-instruct	Open source
Mistral	mistralai/mistral-large	European option
DeepSeek	deepseek/deepseek-coder	Code specialist

See the full list at OpenRouter Models.

---

Publishing This Action

To host and reuse this action across your projects:

Option 1: Public Repository (Recommended)

1. Create a new GitHub repository named ai-code-reviewer (or your preferred name)

2. Push this code to the repository:

   git init
   git add .
   git commit -m "Initial commit: AI Code Reviewer action"
   git remote add origin https://github.com/YOUR_USERNAME/ai-code-reviewer.git
   git push -u origin main

3. Create a release:

   git tag -a v1 -m "Release v1"
   git push origin v1

4. Use in any repository:

   uses: YOUR_USERNAME/ai-code-reviewer@v1

Option 2: GitHub Marketplace

1. Follow Option 1 steps first

2. Go to your repository → Releases → Draft a new release

3. Check "Publish this Action to the GitHub Marketplace"

4. Fill in the marketplace listing details

5. Publish the release

After publishing to the Marketplace, users can find and use your action more easily.

Option 3: Private Repository (GitHub Enterprise)

For private/internal use:

1. Create a private repository with the action code

2. Reference it using the full path:

   uses: YOUR_ORG/ai-code-reviewer@v1

3. Ensure the workflow has access to the private repository

---

Repository Structure

ai-code-reviewer/
├── action.yml              # Action definition
├── scripts/
│   └── review.js           # Main review script
├── .github/
│   └── workflows/
│       └── pr-code-review.yml  # Example workflow
└── README.md

Troubleshooting

Review not posting

1. Check that OPENROUTER_API_KEY is set correctly
2. Verify the workflow has pull-requests: write permission
3. Check the Actions log for error messages

API errors

1. Verify your API key is valid
2. Check your OpenRouter account has credits
3. Ensure the model ID is correct (check OpenRouter Models)

Diff too large

The script truncates diffs larger than 100KB by default. You can adjust this with max_diff_size.

License

MIT