capybara-vibe 0.1.2

Async-first AI-powered CLI coding assistant implementing the ReAct agent pattern

Capybara Vibe

Async-first AI-powered CLI coding assistant implementing the ReAct agent pattern.

Capybara Vibe is a powerful command-line interface designed to assist developers by leveraging Large Language Models (LLMs) to understand, plan, and execute coding tasks directly within your terminal. It supports interactive chat, one-off command execution, and session management, all built on a robust async architecture.

Features

• ReAct Agent Pattern: Implements sophisticated reasoning and acting loops to solve complex problems.
• Interactive Chat: Natural language conversation with context awareness.
• Multi-Mode Operation:
  • standard: Balanced autonomy and safety.
  • safe: Restricted permissions, asking for confirmation before critical actions.
  • plan: Focuses on planning without execution permissions.
  • auto: High autonomy for trusted tasks.
• Session Management: Save, list, and resume conversation sessions.
• Model Flexibility: Easy switching between different LLM providers and models.
• Configuration UI: Built-in Web UI for easy configuration management.
• MCP Support: Model Context Protocol (MCP) enabled for enhanced context integration.

Installation

Ensure you have Python 3.10 or higher installed.

pip install capybara-vibe

Getting Started

1. Configuration

Before using Capybara, you need to initialize the configuration. You can do this via an interactive Web UI or through the CLI.

Web UI (Recommended):

capybara init

This will open a configuration page in your browser where you can set up your LLM providers (e.g., OpenAI, Anthropic) and other settings.

CLI Only:

capybara init --cli

2. Basic Usage

Interactive Chat: Start a new chat session to brainstorm or ask questions.

capybara chat

You can also start with an initial message:

capybara chat "Refactor the authentication middleware in src/"

Single Command Run: Execute a specific task and exit.

capybara run "Write a unit test for the user login function"

3. Usage Options

Select Model: Specify a model for a specific command:

capybara chat --model gpt-4-turbo

Set Default Model: View available models and set a global default:

capybara model

Or set it directly:

capybara model gpt-4-turbo

Modes: Run in different modes depending on your needs:

capybara chat --mode safe   # Asks for permission before file edits
capybara chat --mode plan   # Only generates plans, no file edits

4. Session Management

List Sessions: See your recent activity.

capybara sessions

Resume a Session: Pick up where you left off using the Session ID.

capybara resume <session_id>

Development

To contribute to Capybara Vibe Coding:

1. Clone the repository.
2. Install dependencies:

   pip install -e ".[dev]"
   

3. Run tests:

   pytest
   

License

MIT