AI Code Review Agent · GitHub Action

AI‑powered, auto‑stack code reviews for your pull requests. The action detects the tech stack (React/Next, Vue/Nuxt, Laravel/WordPress, Django/Flask, …), runs the LLM you choose, returns a structured JSON with score & findings and can fail CI when quality drops below your threshold.

---

✨ Features

Icon	Feature	Description
🤖	Multi‑provider LLM	OpenAI GPT‑4/4o, Anthropic Claude 3, Google Gemini‑Pro or any Ollama/Llama endpoint.
🔍	Automatic stack detection	Tailors the prompt to React/Next, Vue/Nuxt, Laravel/WordPress (PHP) or Django/Flask (Python).
📊	Structured report	Numeric score, recommendation plus an array of findings in JSON.
🎯	Fully configurable	Glob filters, min‑score gate, PR‑comment toggle.
🚀	Plug‑and‑play	Single uses: line – nothing to deploy.

---

Quick Start (OpenAI)

name: AI Code Review

on:
  pull_request:
    types: [opened, synchronize]

jobs:
  review:
    runs-on: ubuntu-latest
    steps:
      - uses: samushi/code-review-action@v1
        with:
          provider: openai                   # openai | anthropic | gemini | ollama
          github-token: ${{ secrets.GITHUB_TOKEN }}
          openai-api-key: ${{ secrets.OPENAI_API_KEY }}
          # optional
          min-score-threshold: 7
          fail-on-low-score: false

Secrets: add the API key required by your provider (OPENAI_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY). GITHUB_TOKEN is provided automatically.

---

📥 Inputs

Name	Required	Default	Description
provider	No	openai	openai · anthropic · gemini · ollama
github-token	Yes	${{ github.token }}	Token with repo scope to post PR comments.
openai-api-key	if provider = openai	–	OpenAI secret key.
anthropic-api-key	if provider = anthropic	–	Anthropic Claude secret.
gemini-api-key	if provider = gemini	–	Google Generative AI key.
ollama-base-url	if provider = ollama	http://localhost:11434	Base URL of your Ollama server.
ai-model	No	provider default	e.g. gpt-4o-mini, claude-3-opus-20240229, gemini-pro, llama3.
stack	No	auto	Force a stack (react, vue, laravel, django, …) or leave blank for auto‑detect.
file-patterns	No	**/*.{js,jsx,ts,tsx,php,py}	Comma‑separated glob patterns to include.
exclude-patterns	No	**/{node_modules,dist}/**	Patterns to ignore.
min-score-threshold	No	7	Fail CI if score < threshold & fail-on-low-score=true.
post-comment	No	true	Post the AI report as a PR comment.
fail-on-low-score	No	false	Mark job failed when quality is low.

---

📤 Outputs

Output	Description
review-score	Numeric score (1‑10).
recommendation	POSITIVE, NEEDS_CHANGES, or NEGATIVE.
issues-found	Total findings reported by the AI.
review-summary	One‑paragraph summary.

---

Provider Examples

Laravel (OpenAI)

with:
  provider: openai
  openai-api-key: ${{ secrets.OPENAI_API_KEY }}
  stack: laravel
  file-patterns: '**/*.php,**/*.blade.php'

Django (Claude 3 Sonnet)

with:
  provider: anthropic
  anthropic-api-key: ${{ secrets.ANTHROPIC_API_KEY }}
  stack: django
  ai-model: claude-3-sonnet-20240229

Vue (Gemini‑Pro)

with:
  provider: gemini
  gemini-api-key: ${{ secrets.GEMINI_API_KEY }}
  stack: vue

Local Llama 3 via Ollama (auto‑detect)

with:
  provider: ollama
  ollama-base-url: http://localhost:11434
  ai-model: llama3
  post-comment: false   # dry‑run

---

Roadmap

• Inline annotations via GitHub Checks API
• SARIF export for security scanners
• ESLint / PHPStan / flake8 integration

PRs are welcome – open an issue or contribute! Licensed under MIT.