OVOS OpenAI Skill

Description

The ovos-skill-openai project is an OpenVoiceOS fallback skill designed to forward user requests to the OpenAI GPT models. It maintains conversational context using the OpenAI Conversations API. This allows for a more coherent and engaging user experience.

Features

• Fallback to GPT models for queries
• Context management through Conversations API
• Customizable through settings.json

Prerequisites

• OpenVoiceOS installed
• Python 3.x

Installation

To install the ovos-skill-openai skill, run the following command:

pip3 install git+https://github.com/jaredcobb/ovos-skill-openai.git

Configuration

The settings.json configuration file may contain the following properties. Only api_key is required:

• api_key: Your OpenAI API key [REQUIRED]
• model: The OpenAI GPT model you wish to use (default is "gpt-3.5-turbo")
• system_prompt: The system prompt for initiating conversations with the GPT model. This is how you give the model some personality. We have a basic default.
• wait_timeout: The maximum amount of time OVOS should wait while speaking dialog to you. Some ChatGPT responses may take a while to finish.
• audio_files: An array of audio files (one or many) to shuffle and play while we wait for the OpenAI response. The gpt-4 model is particularly slow at times. This gives the user some feedback that it's processing. Must be an absolute path to a file on the device. Audio files should be less than three seconds long.

Sample JSON Config

{
  "api_key": "your-api-key-here",
  "model": "gpt-3.5-turbo",
  "system_prompt": "You are a helpful voice assistant with a friendly tone and fun sense of humor",
  "wait_timeout": 90,
  "audio_files": [
    "/home/ovos/.config/files/audio_files_example_01.mp3",
    "/home/ovos/.config/files/audio_files_example_02.mp3",
    "/home/ovos/.config/files/audio_files_example_03.mp3",
    "/home/ovos/.config/files/audio_files_example_04.mp3"
  ]
}

Usage

Once installed and configured, the skill will automatically forward queries it receives as fallbacks to the configured GPT model.

License

This project is licensed under the Apache License. See the LICENSE file for details.

Contributing

For contributing guidelines, please refer to the CONTRIBUTING.md file.

Support

If you encounter any issues, please open an issue on GitHub.