Vermeil 🧪

Your Personal, Low-Cost, and Extensible AI Assistant

🚀 About Vermeil

Vermeil is a lightweight, powerful, and cost-effective AI assistant designed to run efficiently on your local machine. At its core is Vermeil, a customized Large Language Model (LLM) based on Microsoft's phi3:mini, fine-tuned for high-quality, responsive interaction.

The project's main goal is to make advanced, personalized AI accessible to everyone without the high costs of proprietary cloud services. With a focus on modularity, Vermeil can be extended using the Model Context Protocol (MCP), allowing you to integrate your own data and tools seamlessly.

✨ Key Features

• 🧠 Custom LLM: Powered by a fine-tuned phi3:mini model for optimal performance and quality.
• 💸 Low-Cost: Engineered to minimize resource consumption, making it perfect for local deployment.
• 🔧 Extensible Architecture: Easily add new capabilities and context using the Model Context Protocol (MCP).
• 🐍 Python-Powered: Built with Python 3.10 for modern, robust performance.
• 🔜 Cross-Platform Vision: Currently available for desktop, with mobile and IoT versions on the horizon!

🚧 Project Status: Under Active Development

Vermeil is currently under heavy development. Features are being added and improved continuously. While the core functionality is operational, you may encounter bugs. We appreciate your patience and welcome contributions to help us build a stable and feature-rich application.

🛠️ Getting Started

Follow these steps to get Vermeil running on your local machine.

Prerequisites

• Ollama: Vermeil uses Ollama to run the local LLM.
• Python 3.10: This project is built and tested with Python 3.10. It is highly recommended.
• Git: To clone the repository.

Step 1: Install and Set Up Ollama

1. Install Ollama:

   • For macOS & Linux:

     curl -fsSL https://ollama.com/install.sh | sh

   • For Windows: Download and run the installer from the Ollama website.

2. Download the Base Model: After installing Ollama, open your terminal and pull the phi3:mini model. This is the foundation for Vermeil.

   ollama pull phi3:mini

3. Create the Custom 'vermeil' Model: Vermeil is designed to work with a custom model entry. In the root of this project, you will find a Modelfile (click to see example). Use it to create the vermeil model by running:

   ollama create vermeil -f Modelfile

   (You can skip this step)

Step 2: Configure and Install the Project

1. Clone the repository:

   git clone https://github.com/Matrixxboy/vermeil.git
   cd vermeil

2. Set Model Configuration: Open the configuration file located at core_project/core/Model/model_config.py. Ensure the following variables are set to point to your local Ollama instance and the custom model:

   OLLAMA_URL = "http://localhost:11434/api/generate"
   #you can use default model : phi3:mini
   OLLAMA_MODEL = "vermeil" #or "phi3:mini"

3. Create a Virtual Environment & Install Dependencies:

   python3.10 -m venv venv
   source venv/bin/activate  # On Windows, use `venv\Scripts\activate`
   pip install -r requirements.txt

Step 3: Run the Application

Once Ollama is running and the project is configured, launch the application:

python main.py

🧩 Extending with Model Context Protocol (MCP)

One of Vermeil's most powerful features is its extensibility.
MCP allows you to inject custom context, tools, or data into the model before it processes a prompt.
This enables you to tailor Vermeil's knowledge and capabilities to your specific needs.

🔧 How to add your own context

1. Navigate to the core_project/MCP/ directory.
2. Add your custom protocol files or modules in this directory.
3. The application will automatically detect and load them, making them available to the LLM.

This system is designed to be simple yet powerful, allowing for deep customization without altering the core codebase.

---

🗺️ Roadmap

We have an exciting future planned for Vermeil! Here's what we're working on:

• Core AI Assistant (Desktop)
• Fully-featured Mobile Application
• IoT Device Integration for smart home commands
• Advanced Tool Integration via MCP
• Community-driven MCP library
• Further performance and cost-optimization research

---

🤝 Contributing

Contributions are what make the open-source community such an amazing place to learn, inspire, and create.
Any contributions you make are greatly appreciated.

If you have a suggestion that would make this better:

• Fork the Project

• Create your Feature Branch

  git checkout -b feature/AmazingFeature

• Commit your Changes

  git commit -m 'Add some AmazingFeature'

• Push to the Branch

  git push origin feature/AmazingFeature

Open a Pull Request

---

📄 License

Distributed under the MIT License. See LICENSE for more information.

Stay tuned for more updates! 🚀