Bee Reviewer 🐝

Bee Reviewer is a Python CLI tool that generates narrated, interactive code reviews. It uses a "Context-Aware" architecture where your AI agent (Claude, Gemini, etc.) directly generates review recipes as it works, ensuring the narration captures the full intent and design decisions of every task.

🌟 Features

• Context-Aware: Review recipes are generated inside the AI session, capturing planning and reasoning that raw git diffs often miss.
• Narrated Presentation: A cinematic web-based slideshow where your code is explained via high-quality text-to-speech.
• Turnkey Viewer: A self-contained web viewer with syntax highlighting, auto-play narration, and full playback controls.
• Cross-Agent Support: Works with Gemini CLI, Claude Code, Aider, and any other AI agent via BEE-REVIEWER.md.
• Zero Token Overhead: No secondary AI calls required for reviews.

🚀 Installation

Install via pip:

pip install bee-reviewer

🛠️ Getting Started

1. Initialize your project

Run this command in the root of your project:

bee-reviewer init

This will:

• Print the Bee Reviewer banner.
• Create a .bee-review/ directory to store your reviews.
• Detect your agent (CLAUDE.md, GEMINI.md) or create a fallback BEE-REVIEWER.md.
• Inject instructions telling the AI to generate a recipe.json in .bee-review/ after every completed task.

2. Work with your AI Agent

Continue working with your AI agent as usual. After every human prompt you provide, the agent will silently create a new review file in the .bee-review/ folder.

3. Launch a Review

When you're ready to watch your progress:

bee-reviewer review

• Select a review from the chronological menu.
• A local web server starts, and your browser opens to a narrated slideshow of your changes!

📖 Manual Usage & Advanced Tips

Custom Agent Launch

You can initialize and immediately launch your AI agent in one command:

bee-reviewer init --launch gemini

💡 Pro Tips for Better Reviews

• Baseline Commits: For the best review experience, run git commit -m "baseline" before starting a major task. This gives the AI a clean reference point for its narration.
• Large File Handling: The AI instructions include a 1,000-line limit per file. If a file is larger, the AI is instructed to include only the most relevant 1,000 lines to keep your sessions fast and lightweight.
• Recipe Management: All review recipes are stored as JSON files in the .bee-review/ directory. You can safely delete old files you no longer need, or share specific .json files with teammates.
• Shell Completion: bee-reviewer supports tab-completion. To enable it for your shell, run:

  bee-reviewer --install-completion
  

🛑 Stopping the Server

Press Ctrl+C in your terminal to stop the review server. This safely returns you to your shell.

📄 License

MIT