UltraDev

A comprehensive GitHub Action that provides intelligent code reviews using multiple AI providers with framework-specific expertise.

🚀 Features

• Multi-AI Provider Support: Choose from OpenAI, Claude, Gemini, or DeepSeek
• Framework-Specific Reviews: Specialized prompts for Laravel, Vue/Nuxt, and React/Next.js
• Smart PR Analysis: Reviews only added lines with contextual understanding
• Confidence Scoring: Risk assessment and confidence levels for each review
• Professional Comments: Evidence-based feedback following industry best practices

📋 Required Inputs

Core Configuration

inputs:
  ai_provider:
    description: 'AI provider to use for code review'
    required: true
    # Options: 'openai', 'claude', 'gemini', 'deepseek'
  
  framework:
    description: 'Framework for specialized code review'
    required: true
    # Options: 'laravel', 'vue', 'nuxt', 'react', 'nextjs'
  
  github_token:
    description: 'GitHub token for PR access'
    required: true

Provider-Specific API Keys

Provide the API key corresponding to your chosen ai_provider:

  # For OpenAI
  openai_api_key:
    description: 'OpenAI API key'
    required: false
  openai_assistant_id:
    description: 'OpenAI Assistant ID'
    required: true # Required when ai_provider is 'openai'
  
  # For Claude
  claude_api_key:
    description: 'Anthropic Claude API key'
    required: false
  
  # For Gemini
  gemini_api_key:
    description: 'Google Gemini API key'
    required: false
  
  # For DeepSeek
  deepseek_api_key:
    description: 'DeepSeek API key'
    required: false

🔧 Usage Examples

OpenAI + Laravel

name: Code Review
on:
  pull_request:
    types: [opened, synchronize]

jobs:
  code-review:
    runs-on: ubuntu-latest
    steps:
      - name: AI Code Review
        uses: updivision/ultra-dev@v1
        with:
          ai_provider: 'openai'
          framework: 'laravel'
          github_token: ${{ secrets.GITHUB_TOKEN }}
          openai_api_key: ${{ secrets.OPENAI_API_KEY }}
          openai_assistant_id: ${{ secrets.OPENAI_ASSISTANT_ID }}

Claude + React

name: Code Review
on:
  pull_request:
    types: [opened, synchronize]

jobs:
  code-review:
    runs-on: ubuntu-latest
    steps:
      - name: AI Code Review
        uses: updivision/ultra-dev@v1
        with:
          ai_provider: 'claude'
          framework: 'react'
          github_token: ${{ secrets.GITHUB_TOKEN }}
          claude_api_key: ${{ secrets.CLAUDE_API_KEY }}

Gemini + Vue/Nuxt

name: Code Review
on:
  pull_request:
    types: [opened, synchronize]

jobs:
  code-review:
    runs-on: ubuntu-latest
    steps:
      - name: AI Code Review
        uses: updivision/ultra-dev@v1
        with:
          ai_provider: 'gemini'
          framework: 'nuxt'
          github_token: ${{ secrets.GITHUB_TOKEN }}
          gemini_api_key: ${{ secrets.GEMINI_API_KEY }}

DeepSeek + Next.js

name: Code Review
on:
  pull_request:
    types: [opened, synchronize]

jobs:
  code-review:
    runs-on: ubuntu-latest
    steps:
      - name: AI Code Review
        uses: updivision/ultra-dev@v1
        with:
          ai_provider: 'deepseek'
          framework: 'nextjs'
          github_token: ${{ secrets.GITHUB_TOKEN }}
          deepseek_api_key: ${{ secrets.DEEPSEEK_API_KEY }}

📌 Action Versioning

How to Reference This Action

When using this action in your workflows, you can choose between different versioning strategies:

Option 1: Use Latest Major Version (Recommended)

uses: updivision/ultra-dev@v1

• Automatically gets: Latest features and bug fixes within v1.x.x
• Best for: Most users who want convenience and latest improvements
• Updates: Automatically receives compatible updates

Option 2: Pin to Specific Version

uses: updivision/[email protected]

• Guarantees: Exact same behavior every time
• Best for: Production environments requiring strict reproducibility
• Updates: Manual - you control when to update

Why This Pattern?

Semantic versioning encourages precise tags like v1.0.0, v1.0.1, etc., that never change once published.

But you might want a convenience tag like v1 that always points to the latest v1.x.x release.

This lets you pick either:

• A fixed version (e.g., v1.0.1) to ensure stability, or
• A moving version (e.g., v1) to get the latest compatible update without changing your workflow

Setup Instructions

1. Get API Keys

OpenAI

1. Visit OpenAI Platform
2. Create an API key in your dashboard
3. Create a custom Assistant and note the Assistant ID

Claude (Anthropic)

1. Visit Anthropic Console
2. Generate an API key
3. Ensure you have access to Claude-3 models

Gemini (Google)

1. Visit Google AI Studio
2. Create a new API key
3. Enable the Generative AI API

DeepSeek

1. Visit DeepSeek Platform
2. Create an account and generate API key
3. Access DeepSeek Coder models

2. Configure GitHub Secrets

Add your API keys as repository secrets:

1. Go to your repository → Settings → Secrets and variables → Actions
2. Add the following secrets:
   • OPENAI_API_KEY (if using OpenAI)
   • OPENAI_ASSISTANT_ID (required for OpenAI)
   • CLAUDE_API_KEY (if using Claude)
   • GEMINI_API_KEY (if using Gemini)
   • DEEPSEEK_API_KEY (if using DeepSeek)

3. GitHub Token Permissions

Important: You must configure your repository to grant the necessary permissions to the GITHUB_TOKEN.

Setting Up Workflow Permissions

1. Go to your repository → Settings → Actions → General
2. Scroll down to Workflow permissions
3. Select "Read and write permissions"
4. Click Save

The action requires these specific permissions:

• contents: read - To access PR files
• pull-requests: write - To post review comments
• issues: write - To update PR status

Without proper permissions, the action will fail to post review comments on your pull requests.

🤝 Contributing

Adding New AI Providers

1. Create a new provider class in src/providers/
2. Implement the base provider interface
3. Add configuration in action.yml
4. Update documentation

Adding New Frameworks

1. Create framework-specific prompts in src/prompts/
2. Add framework validation in src/config/settings.py
3. Update documentation and examples

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• Framework-specific best practices derived from official documentation
• AI provider integrations following their respective best practices
• Community feedback and contributions