A local AI model web interface using Hugging Face Transformers.

open-LocalModel provides a web-based chat interface for running various Hugging Face AI models locally. It supports both GPU and CPU execution with model quantization options.

• Multiple AI model support (DeepSeek, Qwen, Llama, Mixtral)
• Web-based chat interface
• Automatic chat saving and loading
• Model parameter configuration
• Hardware optimization (GPU/CPU with quantization)
• Model management (install, delete, switch)

• DeepSeek R1 series (1.5B, 8B, 14B)
• Qwen series (2.5-7B-Instruct, 3-8B-Base)
• Mixtral (Nous-Hermes-2-Mixtral-8x7B-DPO)
• open-neo Kyro series

• Python 3.8+
• 8GB+ RAM (varies by model)
• NVIDIA GPU recommended (CUDA support)

1. Clone the repository:

git clone https://github.com/llaa33219/open-LockalModel.git
cd open-LocalModel

2. Install dependencies:

pip install -r requirements.txt

3. Run the application:

python Run.py

4. Open your browser and navigate to http://localhost:5000

1. Install a model using the "Install Model" button
2. Select the installed model from the dropdown
3. Start chatting in the input field
4. Adjust model parameters in the settings panel

open-LocalModel/
├── AI_model.py          # AI model management
├── Chat.py              # Chat functionality
├── Web.py               # Flask web server
├── Run.py               # Main entry point
├── index.html           # Web interface
├── models/              # Downloaded models storage
├── Chat/                # Chat history storage
└── Setting-ai/          # Settings storage

Model settings are stored in models/models_info.json. Each model can have individual configurations for temperature, top-p, quantization, and other parameters.

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