Self-Driving Codebase

A GitHub repository that drives itself using AI agents powered by Gemini CLI and GitHub Actions. This project implements a system of autonomous agents that can triage issues, review code, update documentation, diagnose CI failures, and even implement features from issue descriptions.

Features

• Issue Triage - Automatically classifies and prioritizes new issues
• Code Review - Reviews pull requests with AI-powered analysis
• CI Doctor - Diagnoses CI/CD failures and suggests fixes
• Documentation Sync - Keeps README and docs updated with code changes
• Self-Coding Agent - Implements features from issue descriptions
• Continuous Improvement - Weekly analysis for code simplification opportunities

Quick Start

1. Set Up API Key

Get your Gemini API key from Google AI Studio:

1. Go to Google AI Studio
2. Click "Create API Key"
3. Copy the key

2. Add Secret to GitHub

1. Go to your repository Settings → Secrets and variables → Actions
2. Click New repository secret
3. Name: GEMINI_API_KEY
4. Value: Paste your API key
5. Click Add secret

3. Enable Actions

The workflows are already configured. Just push to your repository and they will activate.

Available Commands

Comment on any issue or PR to trigger agents:

Command	Description	Example
/implement	Implement a feature from issue	@gemini-cli /implement Add dark mode support
/fix	Fix a bug from issue	@gemini-cli /fix Users can't log in with Safari
/review	Review a pull request	@gemini-cli /review
/triage	Classify an issue	@gemini-cli /triage
/doc	Generate documentation	@gemini-cli /doc Update API docs
/debug	Diagnose CI failure	@gemini-cli /debug Build is failing
/explain	Explain code	@gemini-cli /explain this function
/help	Show available commands	@gemini-cli /help

How It Works

You create an issue
        │
        ▼
┌───────────────────┐
│  GitHub Actions   │
│    triggers       │
└───────────────────┘
        │
        ▼
┌───────────────────┐
│  Gemini CLI       │
│  processes request│
└───────────────────┘
        │
        ▼
┌───────────────────┐
│  Creates PR/      │
│  Comments/Issues  │
└───────────────────┘

Free Tier Limits

Gemini API (Free)

• Requests: 10/minute, 250/day
• Tokens: 250,000/minute
• Context: 1 million tokens

GitHub Actions (Free)

• Minutes: 2,000/month
• Concurrency: 20 jobs
• Storage: 500 MB

Workflows

Dispatcher (gemini-dispatch.yml)

Central routing for all agent commands. Runs on issue/PR comments.

Issue Triage (scheduled-triage.yml)

Daily analysis of open issues. Runs at 8 AM UTC.

Auto PR Review (auto-pr-review.yml)

Automatic code review on PR creation.

CI Doctor (ci-doctor.yml)

Diagnoses failed CI runs. Also runs nightly health checks.

Documentation Sync (documentation-sync.yml)

Updates docs when code changes. Runs on PR merge.

Self-Coding Agent (self-coding-agent.yml)

Implements features from issues. Triggered by /implement command.

Continuous Improvement (continuous-improvement.yml)

Weekly code analysis. Runs every Monday at 3 AM UTC.

Configuration

GEMINI.md

Edit this file to customize agent behavior for your codebase.

Workflow Settings

In GitHub Actions, you can set these variables:

• GEMINI_DEBUG - Enable debug logging (set to true)

Customization

Adding New Commands

Edit gemini-dispatch.yml to add new commands:

# In the dispatch job
if echo "$COMMENT" | grep -q "/your-command"; then
  echo "command=your-command" >> $GITHUB_OUTPUT
fi

Custom Agents

Create new workflow files following the existing patterns.

Security

• API keys stored as GitHub secrets
• Agents run with minimal permissions
• All operations logged
• No external service access without configuration

Troubleshooting

Agent not responding?

1. Check the Actions tab for errors
2. Verify GEMINI_API_KEY is set
3. Check for rate limit messages

Rate limit errors?

• Reduce workflow frequency
• Use simpler prompts
• Consider batching operations

Poor results?

• Be more specific in descriptions
• Add examples to issues
• Update GEMINI.md with better context

Contributing

Contributions welcome! Areas for improvement:

• Better prompts
• New agent capabilities
• Bug fixes
• Documentation improvements

License

MIT

Resources

• Gemini CLI Documentation
• GitHub Actions
• Google AI Studio

Test