007 AI code review Agent

A blazing-fast, audit-grade AI code reviewer powered by Bun, Vercel AI SDK, and @ai-sdk/google.

This agent analyzes Git diffs, streams intelligent feedback, and helps you maintain clean, scalable codebases—file by file.

---

🚀 Features

• 🔍 Git-aware code review: Automatically detects file changes and provides AI-powered suggestions.
• 🧠 Gemini 2.5 Flash: Leverages Google’s latest LLM for fast, insightful feedback.
• 🛠️ Custom tools: Includes getFileChangesInDirectoryTool for precise diff analysis.
• 📦 Bun-native: Lightning-fast startup, dependency resolution, and native .env support.
• 🧵 Real-time streaming: Delivers feedback directly to your terminal as it’s generated.

---

📦 Tech Stack

Tool	Purpose
Bun	Runtime & package manager
Vercel AI	Streaming AI orchestration
@ai-sdk/google	Gemini model integration
Zod	Schema validation for tool inputs
simple-git	Git diff parsing and file tracking
dotenv	Environment variable management

---

🛠️ Setup Guide

1. Clone the repository

Navigate to the agent directory:

cd 007---Setup/my-agent

2. Install Bun

Bun is a fast JavaScript runtime. To install it:

curl -fsSL https://bun.sh/install | bash

verify:

bun --version

💡 If you're using Linux, make sure ~/.bun/bin is added to your PATH.

---

3. Install dependencies

Run:

bun install

This installs all packages listed in package.json.

---

4.Configure your .env file**

Create a .env file in the workspace root (007---Setup/):

touch .env

Then add your API key and runtime settings:

# === API Keys ===
GOOGLE_GENERATIVE_AI_API_KEY=your-google-api-key-here

# === Agent Configuration ===
AGENT_NAME=007
AGENT_MODE=diff-summary
LOG_LEVEL=info

# === Git Integration ===
GIT_AUTHOR=Azuka
GIT_EMAIL=example@gmail.com

# === Runtime Settings ===
PORT=3000
TIMEOUT_MS=10000

🔐 Important: Never commit .env to version control. Add it to .gitignore.

---

5. Verify .env is loaded correctly

In index.ts, you can temporarily add:

console.log("Loaded API key:", process.env.GOOGLE_GENERATIVE_AI_API_KEY);

Run:

bun run my-agent/index.ts

If it prints your key, you're good to go.

---

6. Run the agent

Start the agent:

bun run my-agent/index.ts    

It will stream code review output based on your prompt.

---

🧠 How It Works

1. The agent uses simple-git to detect file changes in a specified directory.
2. Changes are analyzed by Gemini 2.5 Flash via streamText() from Vercel AI.
3. A custom system prompt ensures consistent tone, style, and focus for reviews.
4. Feedback is streamed in real-time to your terminal for immediate insights.

---

📁 Project Structure

007---Setup/                      # 🧠 Workspace root
├── README.md                     # Workspace-level documentation
├── bun.lock                      # Bun lockfile for reproducible installs
├── package.json                  # Workspace config (can declare workspaces)
├── setup.md                      # Contributor onboarding guide
│
├── my-agent/                     # 🧠 Modular agent logic
│   ├── index.ts                  # Agent entry point
│   ├── package.json              # Agent-specific dependencies and scripts
│   ├── tsconfig.json             # TypeScript config scoped to agent
│   ├── .env                      # API keys and runtime settings
│
│   ├── prompts.ts                # System prompt for reviewer behavior
│   ├── tools/                    # 🔧 Modular tools
│   │   ├── gitDiff.ts            # Git diff integration
│   │   ├── markdownWriter.ts     # Markdown formatting utility
│   │   ├── commitMessage.ts      # Commit message generator
│   │   └── checkDuplicates.ts    # Duplicate file detection
│
│   ├── test/                     # ✅ Agent tests
│   │   ├── agent.test.ts         # Prompt handling and edge cases
│   │   └── tools.test.ts         # Git tooling and utilities
│
│   └── setup.md                  # Agent-specific onboarding 

---

🧪 Example Usage

await codeReviewAgent(
  "Review the code changes in '../my-agent' directory, make your reviews and suggestions file by file"
);

This triggers a file-by-file review of changes in the specified directory, with suggestions streamed to your terminal.

---

🛡️ Security & Auditability

• File filtering: Excludes sensitive files like bun.lock and dist/.
• Audit-grade design: Built for reliable, reproducible code reviews.
• Extensible: Easily integrates with CI pipelines or commit history analysis.

---

✨ Roadmap

• 🔄 Support for staged vs. committed diff reviews
• 🧪 Test coverage analysis
• �� Summarization of lint and type errors
• 🧭 Contributor onboarding assistant

---

🧑‍💻 Author

Built by Azuka, a backend engineer passionate about automation, reproducibility, and scalable systems. Connect on X for updates or collaboration!

---

📄 License

MIT License — fork, remix, and deploy your own 007 AI Agent with freedom.

---

🙌 Contributing

Contributions are welcome! Please open an issue or pull request on GitHub. Let’s build better code review tools together.