Shunsuke Hayashi 25d358f96e feat: Add config system, agent mode, and session management

- Add comprehensive Config system with TOML support (~/.miyabi/config.toml)
- Add Agent module for autonomous task execution with tool approval
- Add Session management for conversation persistence
- Extend CLI with new commands: init, sessions, agent, version
- Add CLI flags: --model, --max-tokens, --thinking, --config, --session
- Fix all clippy warnings (16 → 0)
- Improve code quality with collapsible pattern matching

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

2025-11-22 22:22:11 +09:00

3.2 KiB

Raw Blame History

miyabi-cli-standalone

Autonomous development powered by Miyabi - AI-driven development framework.

Getting Started

Prerequisites

# Set environment variables
cp .env.example .env
# Edit .env and add your tokens

Installation

npm install

Development

npm run dev          # Run development server
npm run build        # Build project
npm test             # Run tests
npm run typecheck    # Check types
npm run lint         # Lint code

Project Structure

miyabi-cli-standalone/
├── src/              # Source code
│   └── index.ts     # Entry point
├── tests/           # Test files
│   └── example.test.ts
├── .claude/         # AI agent configuration
│   ├── agents/      # Agent definitions
│   └── commands/    # Custom commands
├── .github/
│   ├── workflows/   # CI/CD automation
│   └── labels.yml   # Label system (53 labels)
├── CLAUDE.md        # AI context file
└── package.json

Miyabi Framework

This project uses 7 autonomous AI agents:

CoordinatorAgent - Task planning & orchestration
IssueAgent - Automatic issue analysis & labeling
CodeGenAgent - AI-powered code generation
ReviewAgent - Code quality validation (80+ score)
PRAgent - Automatic PR creation
DeploymentAgent - CI/CD deployment automation
TestAgent - Test execution & coverage

Workflow

Create Issue: Describe what you want to build
Agents Work: AI agents analyze, implement, test
Review PR: Check generated pull request
Merge: Automatic deployment

Label System

Issues transition through states automatically:

📥 state:pending - Waiting for agent assignment
🔍 state:analyzing - Being analyzed
🏗️ state:implementing - Code being written
👀 state:reviewing - Under review
✅ state:done - Completed & merged

Commands

# Check project status
npx miyabi status

# Watch for changes (real-time)
npx miyabi status --watch

# Create new issue
gh issue create --title "Add feature" --body "Description"

Configuration

Environment Variables

Required variables (see .env.example):

GITHUB_TOKEN - GitHub personal access token
ANTHROPIC_API_KEY - Claude API key (optional for local development)
REPOSITORY - Format: owner/repo

GitHub Actions

Workflows are pre-configured in .github/workflows/:

CI/CD pipeline
Automated testing
Deployment automation
Agent execution triggers

Note: Set repository secrets at: https://github.com/ShunsukeHayashi/miyabi-cli-standalone/settings/secrets/actions

Required secrets:

GITHUB_TOKEN (auto-provided by GitHub Actions)
ANTHROPIC_API_KEY (add manually for agent execution)

Documentation

Miyabi Framework: https://github.com/ShunsukeHayashi/Miyabi
NPM Package: https://www.npmjs.com/package/miyabi
Label System: See .github/labels.yml
Agent Operations: See CLAUDE.md

Support

Issues: https://github.com/ShunsukeHayashi/Miyabi/issues
Discord: [Coming soon]

License

MIT

✨ Generated by Miyabi

3.2 KiB Raw Blame History