AIOS: give a LLM full access to your terminal for the sake of science.

This is basically https://github.com/VictorTaelin/AI-scripts/blob/main/chatsh.mjs implemented in Python, with additional features like image analysis, configuration files.

Example usage:

• Logs of adding streaming to Gemini on tiny-ai-client

output-cropped-3x.mp4

Installation

1. Clone this repository.
2. Run poetry install in the project directory.
3. Alternatively, install via pip: pip install pi-ai-os

Usage

aios [options] [initial message]

Options:

• -m, --model: Specify the AI model to use (default: claude-3-5-sonnet-20240620)
• -c, --config: Path to the configuration file
• -a, --assistant: Use an assistant model for context gathering, hardcoded to Claude Haiku for speed and cost
• -v, --verbose: Increase output verbosity

Example:

export ANTHROPIC_API_KEY=your_api_key_here
aios -m claude-3-5-sonnet-20240620 "List all files in the current directory"

Features

Image Analysis

To include an image in your query, use the /image command followed by the path to the image:

/image path/to/your/image.jpg Describe this image

Configuration File

AIOS can use a configuration file to provide context about your project. By default, it looks for a file named pi_ai_os.yaml in the current directory. You can specify a different config file using the -c option.

Example configuration file:

allowed_extensions:
  - .txt
  - .md
  - .py
ignored_dirs:
  - .git
  - node_modules

Context-Aware Assistance

Use the -a or --assistant flag to enable context-aware assistance. This feature uses an assistant model to analyze the project structure and select relevant files for providing context to the main AI model.

Clearing Context

Type /clear or /c during your interaction with AIOS to clear the conversation context.

Verbose Mode

Use the -v or --verbose flag to increase output verbosity, which can be helpful for debugging or understanding the AI's decision-making process.