aixion

A powerful GitHub Action that enables seamless integration with various AI providers, allowing you to enhance your workflows with artificial intelligence capabilities. Supporting multiple providers including OpenAI, Anthropic, Groq, Mistral, and more.

Features

• Multi-Provider Support:

  • OpenAI
  • Anthropic (Claude)
  • Groq
  • Mistral
  • Cohere
  • DeepInfra
  • Fireworks
  • Google
  • Together AI
  • XAI

• Flexible Configuration:

  • Custom API endpoints
  • Configurable model parameters
  • Adjustable response settings
  • Header customization

• Easy Integration:

  • Simple workflow setup
  • Comprehensive output handling
  • Support for both chat and completion modes

Installation

1. Add the action to your workflow file (e.g., .github/workflows/ai.yml)
2. Configure your AI provider credentials as repository secrets
3. Customize the action parameters as needed

Basic Usage

Create a workflow file in your repository:

name: Basic AI Prompt
on:
  push:
    branches: [main]

jobs:
  generate:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - name: Generate AI Text
        uses: 0xjord4n/aixion@v1.2.1
        with:
          config: >
            {
              "provider": "openai",
              "provider_options": {
                "api_key": "${{ secrets.OPENAI_API_KEY }}"
              },
              "prompt": "Your prompt here",
              "model": "gpt-4"
            }

Input Methods

aixion supports both direct text input and file-based input for prompts and messages:

File-Based Input

You can now reference files for your prompts and system messages:

config: >
  {
    "provider": "openai",
    "provider_options": {
      "api_key": "${{ secrets.OPENAI_API_KEY }}"
    },
    "prompt_file": ".github/prompts/analysis.txt",
    "system_file": ".github/prompts/system.txt",
    "model": "gpt-4"
  }

For messages array, you can mix file-based and direct content:

config: >
  {
    "provider": "openai",
    "provider_options": {
      "api_key": "${{ secrets.OPENAI_API_KEY }}"
    },
    "messages": [
      {
        "role": "system",
        "content_file": ".github/prompts/system.txt"
      },
      {
        "role": "user",
        "content_file": ".github/prompts/user_query.txt"
      },
      {
        "role": "assistant",
        "content": "I understand your question. Let me help..."
      }
    ],
    "model": "gpt-4"
  }

Note: When using file-based inputs:

• Use prompt_file instead of prompt to read from a file
• Use system_file instead of system to read from a file
• In messages array, use content_file instead of content to read from a file
• File paths are relative to your repository root
• Files must exist and be readable
• You can mix direct content and file-based content in the messages array

Prompting Methods

aixion supports three different ways to interact with AI models:

1. Simple Prompt

Best for straightforward, single-turn interactions:

config: >
  {
    "provider": "openai",
    "provider_options": {
      "api_key": "${{ secrets.OPENAI_API_KEY }}"
    },
    "prompt": "What is the capital of France?",
    "model": "gpt-4"
  }

2. System + Prompt

Useful when you need to set specific behavior or context:

config: >
  {
    "provider": "openai",
    "provider_options": {
      "api_key": "${{ secrets.OPENAI_API_KEY }}"
    },
    "system": "You are a helpful programming assistant.",
    "prompt": "How do I write a hello world in Python?",
    "model": "gpt-4"
  }

3. Messages Array

Perfect for multi-turn conversations or complex interactions:

config: >
  {
    "provider": "openai",
    "provider_options": {
      "api_key": "${{ secrets.OPENAI_API_KEY }}"
    },
    "messages": [
      {
        "role": "system",
        "content": "You are a helpful programming assistant."
      },
      {
        "role": "user",
        "content": "What is a variable?"
      },
      {
        "role": "assistant",
        "content": "A variable is a container for storing data values."
      },
      {
        "role": "user",
        "content": "Can you show an example?"
      }
    ],
    "model": "gpt-4"
  }

Note: You should use only one of these methods per request. The precedence order is:

1. messages (if present, others are ignored)
2. system + prompt (if no messages)
3. prompt alone (if no messages or system)

Configuration Options

Provider Options

• api_key: Your provider's API key
• base_url: Custom API endpoint (optional)
• headers: Additional headers (optional)

Model Parameters

• temperature: Control randomness (0.0-2.0)
• max_tokens: Maximum response length
• top_p: Nucleus sampling parameter
• frequency_penalty: Repetition control
• presence_penalty: Topic diversity control
• stop: Array of sequences where the API will stop generating further tokens

Miscellaneous Options

• save_path: Path where the response will be saved (optional)

Outputs

The action provides the following outputs:

• text: Generated response
• usage: Token usage statistics
• finishReason: Completion status

Development

To contribute to this project:

1. Clone the repository:

   git clone https://github.com/0xjord4n/aixion.git

2. Install dependencies
3. Make your changes
4. Submit a pull request

Support

For issues, feature requests, or questions:

• Open an issue
• Submit a pull request

License

This project is licensed under the MIT License - see the LICENSE file for details.