DeepCode 🐽 - AI Coding Agent Based on LangChain

English | 中文

An intelligent AI-powered coding agent CLI (based on LangChain) that helps you write, analyze, and understand code with natural language commands.

Features

• AI-Powered Code Assistant: Interact with your codebase using natural language
• Human-in-the-Loop Approval: Review and approve tool executions before they run
• Tabbed Question Interface: Beautiful TUI for multi-step user interactions with:
  • Tabbed navigation between questions
  • Progress indicator ("Step X of Y")
  • Arrow key navigation (←→ for tabs, ↑↓ for options)
  • Preview panel for options with rich markdown/code preview
  • Dark mode styling with blue highlights
  • Submit validation (all questions must be answered)
  • Inline custom input for "Other" options (supports English/Chinese)
• Rich Terminal UI: Beautiful output with markdown rendering and syntax highlighting
• Multi-Model Support: Works with OpenAI-compatible APIs (OpenAI, DeepSeek, etc.)
• Extensible Middleware System: Add custom behaviors via middleware
• Async Architecture: Built on asyncio for responsive interactions

Installation

Prerequisites

• Python 3.11 or higher
• uv (recommended package manager)

Install with uv

# Clone the repository
git clone https://github.com/77z-zhou/DeepCode.git
cd deepcode

# Install dependencies
uv sync

# Or install in development mode
uv pip install -e .

Install with pip

pip install -e .

Quick Start

# Run deepcode
deepcode

# Or with Python
python -m deepcode

First run will prompt you to configure:

• API provider (OpenAI, DeepSeek, etc.)
• API key
• Model name
• Workspace directory

Usage

Interactive Mode

Start the CLI and chat with the agent:

deepcode

Example interactions:

You: Read the main.py file and summarize what it does

You: Add error handling to the parse_config function

You: Create a test for the User class

You: What files use the Database connection?

Built-in Commands

• /help - Show help information
• /clear - Clear the screen
• /model - Change the AI model
• /compact - Compact conversation context
• /cost - Show API usage costs
• /exit - Exit the program
• /init - Reinitialize configuration

Configuration

Configuration is stored in your project directory. The first run will guide you through setup:

# Sample configuration structure
{
    "model": "gpt-4",           # Model identifier
    "api_key": "sk-...",        # API key
    "base_url": null,           # Custom API base URL (optional)
    "workspace": ".",           # Working directory
    "recursion_limit": 50,      # Agent recursion limit
    "skills_dir": ".claude/skills"  # Custom skills directory
}

Project Structure

deepcode/
├── deepcode/
│   ├── __init__.py
│   ├── __main__.py           # Entry point
│   ├── cli/                  # CLI interface
│   │   ├── app.py            # Main CLI application
│   │   ├── commands/         # Built-in commands
│   │   └── utils.py          # TUI components (tabbed interface)
│   ├── core/
│   │   ├── agent.py          # Main agent logic
│   │   ├── model.py          # Model factory
│   │   └── middleware/       # Agent middleware
│   ├── config/               # Configuration management
│   └── bus/                  # Message bus for async communication
├── tests/                    # Test suite
├── pyproject.toml           # Project metadata
└── README.md

Development

Running Tests

# Run all tests
pytest

# Run with coverage
pytest --cov=deepcode

Code Style

This project follows standard Python conventions with type hints.

Key Components

• Tabbed Question Interface (deepcode/cli/utils.py):

  • _present_user_questions() - Main TUI entry point
  • Full prompt_toolkit implementation with custom key bindings
  • Rich markdown preview with syntax highlighting

• Agent (deepcode/core/agent.py):

  • LangGraph-based agent with LangChain integration
  • Streaming responses with progress callbacks
  • HITL (Human-in-the-Loop) support

• Message Bus (deepcode/bus/):

  • Async pub/sub for inbound/outbound messages
  • Coordinates between CLI and agent

Dependencies

• langchain - Agent framework
• langgraph - Agent orchestration
• prompt_toolkit - Terminal UI framework
• rich - Rich text and formatting
• typer - CLI framework
• pydantic - Data validation
• loguru - Logging
• questionary - Interactive prompts

License

MIT License - see LICENSE file for details

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Acknowledgments

Built with:

• LangChain - AI agent framework
• prompt_toolkit - Terminal UI
• Rich - Terminal formatting