Agentic AI Chat

A Cursor-like AI development environment with advanced agentic capabilities, on a Web interface.

WARNING: the project is mainly for demo purpose only, not for production use. It has severe security issues that essentially allows any user to access your system.

Quick Start

1. Activate the virtual environment:

   # On Windows
   venv\Scripts\activate
   
   # On macOS/Linux
   source venv/bin/activate

2. Configure your environment:

   • Copy .env.example to .env if you haven't already
   • Add your API keys in .env (optional)

3. Run the server:

   # Basic single-worker mode (default)
   python claude-tooling/run.py
   
   # Multi-worker mode for better concurrency. 
   # WARNING: It still has some issues. Ideally we would need an external data source to share data between workers.
   python claude-tooling/run.py --workers 4
   
   # Production mode with Gunicorn (Linux/Mac only)
   # WARNING: It still has some issues. Ideally we would need an external data source to share data between workers.
   python claude-tooling/run.py --use-gunicorn --workers 9

Features

• Advanced AI Assistant configured via .cursorrules
• Auto Tool Execution for seamless AI interaction
• File Management tools for reading and writing files
• Command Execution capabilities for terminal operations
• Web Tools for searching and extracting web content
• Conversation Root Directories - Each conversation is automatically assigned a unique directory in runs/<timestamp> for better organization of generated files
• Server Concurrency - Support for multiple workers to handle concurrent requests for better performance
• Enhanced UI for Tool Calls - Collapsible tool calls with clear visualization of inputs and results, automatic expansion when results arrive, and visual indicators for completed tool executions
• Modular Frontend Architecture - Well-structured frontend code with clear separation of concerns for better maintainability and extensibility
• Reliable Tool Completion Detection - Improved tracking of tool execution status and accurate auto-execution indicator management

Frontend Architecture

The frontend code follows a modular architecture with clear separation of concerns:

• state.js - Centralized state management using a singleton pattern
• api.js - API communication layer for backend interaction
• tools.js - Tool call handling and auto-execution management
• ui.js - User interface rendering and DOM operations
• events.js - Event handling and user interaction processing
• config.js - Configuration constants and settings
• filePreview.js - File preview functionality for generated files
• utils.js - Utility functions for common operations
• main.js - Application initialization and coordination

See frontend/README.md for detailed documentation of the frontend architecture.

Testing

The project includes a comprehensive test suite to ensure functionality and assist with code refactoring:

# Run all tests with the test runner script
python claude-tooling/scripts/run_tests.py

# Run specific test groups
python claude-tooling/scripts/run_tests.py --api
python claude-tooling/scripts/run_tests.py --auto-execute
python claude-tooling/scripts/run_tests.py --modularity

# Run tests with verbose output and coverage reports
python claude-tooling/scripts/run_tests.py --verbose --coverage

See tests/README.md for detailed information about the testing strategy and how to use tests for refactoring.

Dependencies

• Activate the Python virtual environment in ./venv
• Install dependencies: pip install -r claude-tooling/requirements.txt
• For web tools: playwright install chromium

CI/CD

GitHub Actions automatically run tests on pull requests and merges to main branches. View workflow details in .github/workflows/ci.yml.

License

MIT License