AI编程终端工具

uv tool install --python 3.13 git+https://github.com/elblah/dt-aicoder
export OPENAI_API_KEY="your-api-key"
uvx aicoder

aicoder
> Help me refactor this Python function
[PLAN] AI: I'll analyze the function and suggest improvements...

> /plan toggle
*** Planning mode enabled (read-only)
[PLAN] > ls -la
Planning mode: Read-only tools only

> Edit the auth.py file to add JWT validation
└─ AI wants to call: edit_file
   File: src/auth.py
   Changes:
   - Add validate_jwt_token function
   - Update authenticate method
   
Approve? [a]llow once [s]ession [d]eny

export OPENAI_BASE_URL="http://localhost:4999"
export OPENAI_API_KEY="your-api-key"
export OPENAI_MODEL="gpt-5"           # Default: gpt-5-nano
export DEBUG=1                        # Enable debug mode
export YOLO_MODE=1                    # Bypass approvals

// mcp_tools.json
{
  "tools": [
    {
      "name": "deploy_app",
      "command": "kubectl",
      "args": ["apply", "-f", "manifest.yaml"],
      "description": "Deploy application to Kubernetes"
    }
  ]
}

> /theme ocean
✅ Applied theme: ocean

git clone https://github.com/elblah/dt-aicoder
cd dt-aicoder
uv tool install .
aicoder

# Comprehensive test suite
./run-tests.sh

aicoder/
├── core/           # Application core
├── tool_manager/   # Tool execution & approval
├── plugin_system/  # Extensible plugin framework  
├── streaming/      # Real-time response handling
└── themes/         # Visual customization

# Example: Custom notification plugin
class NotifyPlugin:
    def on_ai_response(self, message):
        # Send desktop notifications for AI responses
        subprocess.run(['notify-send', 'AI Response', message[:100]])
    
    def on_command_complete(self, command, result):
        # Notify when long-running commands finish
        if command.duration > 60:
            subprocess.run(['notify-send', 'Command Complete', command.name])