Code Reviewer

Context-aware AI-powered code review for TypeScript/Node.js. Uses static analysis, codebase indexing, and AI (Ollama, Claude, Gemini) to produce actionable review reports.

Requirements

• Node.js 20+
• Ollama (optional) for free local analysis
• API keys for Claude or Gemini if using cloud providers

Installing Ollama

Ollama runs large language models locally. Install it before using code-reviewer setup:

• macOS — Homebrew: brew install ollama (or GUI: brew install --cask ollama-app) — or install script: curl -fsSL https://ollama.com/install.sh | sh
• Linux — Snap: sudo snap install ollama — or install script: curl -fsSL https://ollama.com/install.sh | sh
• Windows — winget: winget install Ollama.Ollama — or download OllamaSetup.exe

After installation, start the server (it may auto-start on some systems):

ollama serve

Then run code-reviewer setup to pull a recommended model and generate your config.

Quick Start

# Install globally
npm install -g code-reviewer

# Setup Ollama (one-time)
code-reviewer setup

# Index your codebase
code-reviewer index

# Run your first review
code-reviewer analyze src/

Commands

| Command | Description | |---------|-------------| | analyze <paths...> | Analyze files or directories. Output formats: text (default), json, markdown. | | index | Index codebase for similar-pattern lookup. Use --force to rebuild, --incremental to update changed files only. Use -v, --verbose for detailed logs. | | setup | Interactive setup for Ollama: check install, pull a model, write config. | | models | List installed Ollama models. | | cache | Manage analysis cache. Use --stats or --clear. |

Analyze options

• -f, --format <type> — Output format: text, json, markdown
• --provider <provider> — AI provider: ollama, claude, gemini
• --model <model> — Override model (e.g. codellama:13b, gemini-2.0-flash)
• -v, --verbose — Show detailed progress (config, file resolution, parsing, cache hits, AI calls)

Configuration

Create .codereviewerrc.json in your project root:

{
  "ai": {
    "provider": "ollama",
    "ollama": { "host": "http://localhost:11434", "model": "codellama:13b" }
  },
  "rules": {
    "performance": true,
    "readability": true,
    "maintainability": true
  },
  "ignore": ["node_modules", "dist", "coverage"]
}

Environment variables

• AI_PROVIDER — Override provider (ollama, claude, gemini)
• OLLAMA_HOST — Ollama server URL
• GEMINI_API_KEY — For Gemini provider
• ANTHROPIC_API_KEY — For Claude provider

Providers

| Provider | Cost | Notes | |----------|------|-------| | Ollama | Free | Local models. Run ollama serve and pull a model. | | Gemini | Free tier | Set GEMINI_API_KEY. | | Claude | Paid | Set ANTHROPIC_API_KEY. |

CI/CD

Example GitHub Actions workflows are in .github/workflows/examples/:

• code-review-ollama.yml — Self-hosted runner with Ollama
• code-review-gemini.yml — Cloud runner, uses GEMINI_API_KEY secret
• code-review-claude.yml — Cloud runner, uses ANTHROPIC_API_KEY secret

Copy one into .github/workflows/ and configure secrets as needed.

Demo

Try the demo scenarios with intentional issues. See demo/README.md for the full list.

code-reviewer analyze demo/order-service.ts      # sequential awaits, retry logic
code-reviewer analyze demo/api-sequential-fetches.ts  # parallelizable API calls
code-reviewer analyze demo/  # all scenarios

License

MIT