README-Spaces.md

GainEnergy/OGAI-24B API

This repository contains the code to deploy the GainEnergy/OGAI-24B model as an API service on Hugging Face Spaces.

Features

• Full REST API with FastAPI
• Streaming support
• Configurable generation parameters
• Docker support for easy deployment
• Chat-formatted responses following the message role format (system, user, assistant)

Deployment on Hugging Face Spaces

1. Create a new Space on Hugging Face:

   • Go to https://huggingface.co/spaces
   • Click "Create new Space"
   • Choose a name for your Space
   • Select "Docker" as Space SDK
   • Select "GPU" as Hardware (A10G or better recommended)

2. Upload the following files to your Space:

   • app.py
   • requirements.txt
   • Dockerfile
   • .gitattributes (optional, for handling large files)

3. The Space will automatically build and deploy the Docker image.

Environment Variables

You can configure the following environment variables in your Hugging Face Space:

• MODEL_ID: The model ID to load (default: "GainEnergy/OGAI-24B")
• DEFAULT_MAX_LENGTH: Default maximum length for generation (default: 2048)
• DEFAULT_TEMPERATURE: Default temperature for generation (default: 0.7)

API Usage

Generate Text

Endpoint: POST /generate

Request Body:

{
  "messages": [
    {
      "role": "system",
      "content": "You are an assistant specialized in oil and gas engineering."
    },
    {
      "role": "user",
      "content": "Explain the principles of reservoir simulation."
    }
  ],
  "temperature": 0.7,
  "max_tokens": 2048,
  "top_p": 0.95,
  "top_k": 50,
  "stream": false
}

Response:

{
  "generated_text": "Reservoir simulation is a computational method used to predict the flow of fluids (oil, gas, and water) through porous media over time..."
}

Stream Generated Text

Endpoint: POST /generate_stream

This endpoint supports Server-Sent Events (SSE) for streaming responses. Set "stream": true in your request body.

Local Development

1. Clone the repository:

git clone https://huggingface.co/spaces/your-username/your-space-name
cd your-space-name

2. Install dependencies:

pip install -r requirements.txt

3. Run the application:

python app.py

4. The API will be available at http://localhost:7860

License

This project is licensed under the Apache 2.0 License - see the original model card for details.