heare-developer 0.3.4

A command-line coding agent.

Heare Developer CLI

Heare Developer CLI is a powerful and interactive coding assistant that leverages Anthropic's Claude AI models to help developers with various tasks. It provides a sandbox environment where you can perform file operations, execute bash commands, and interact with an AI assistant for coding-related queries and tasks.

Key Features

1. Advanced AI Models: Access to Claude 3 models (Opus, Sonnet, Sonnet-3.5, Haiku) for varied needs and performance levels
2. Intelligent File Handling: Smart file mention system using @ syntax for referencing files
3. Sandbox Environment: Configurable sandbox modes for controlled file operations
4. Tool Integration: Built-in tools for file and system operations
5. Command Auto-completion: Intelligent command and path completion
6. Multi-line Input Support: Easy handling of multi-line code and text input
7. Token Usage Tracking: Real-time monitoring of token usage and associated costs
8. Rate Limiting Protection: Built-in rate limit handling with exponential backoff
9. Permission Management: Granular control over file and system operations
10. Rich Command History: Searchable command history with auto-suggestions

Installation

1. Clone the repository:

   git clone https://github.com/clusterfudge/heare-developer.git
   cd heare-developer
   

2. Install the required dependencies:

   pip install -r requirements.txt
   

3. Set up your Anthropic API key:

   export ANTHROPIC_API_KEY=your_api_key_here
   

   Or create a .env file containing:

   ANTHROPIC_API_KEY=your_api_key_here
   

Usage

Basic usage:

python -m heare.developer.cli [sandbox_path]

Command Line Options

• sandbox_path: Path to initialize the sandbox (default: current directory)
• --model: Choose the Claude AI model (default: sonnet-3.5)
  • Available options: opus, sonnet, sonnet-3.5, haiku
• --summary-cache: Specify path for summary cache (default: ~/.cache/heare.summary_cache)
• --sandbox-mode: Set sandbox mode for file operations
  • Options: REMEMBER_PER_RESOURCE, FORGET_IMMEDIATELY, REMEMBER_FOREVER
• --prompt: Provide initial prompt (prefix with @ to read from file)

Interactive Features

1. File References: Use @ syntax to reference files in your messages:

   > Can you explain the code in @src/main.py?
   

2. Multi-line Input:

   > {
   Here's my multi-line
   input that can include
   code or text
   }
   

3. Command Auto-completion:

   • Press Tab to auto-complete commands and file paths
   • @ followed by partial path triggers file path completion

4. Available Commands:

   • /quit or /exit: Exit the CLI
   • /restart: Clear chat history and start over
   • Various tool-specific commands (shown on startup)

Sandbox Modes

• REMEMBER_PER_RESOURCE: Remember permissions per resource (default)
• FORGET_IMMEDIATELY: Ask for permission each time
• REMEMBER_FOREVER: Remember all permissions

Development

The project follows a modular architecture:

• heare/developer/: Core CLI and developer tools
• heare/pm/: Project management functionality (WIP)
• tests/: Test suite

To contribute:

1. Fork the repository
2. Create a feature branch
3. Add tests for new functionality
4. Submit a pull request

License

[Insert your chosen license here]

Acknowledgements

This project uses:

• Anthropic's Claude AI models
• Rich for terminal UI
• Prompt Toolkit for command line interface
• Various other open source packages (see requirements.txt)