Osmosis-MCP-4B-demo

# Osmosis-MCP-4B ## [Blog Post](https://osmosis.ai/blog/applying-rl-mcp) ## Prerequisites * Python 3.x * uvx * Access to a model server (e.g., running locally via vLLM/lm studio) * API Keys for: * Brave Search * Google Maps * AccuWeather ## Setup & Installation 0. **Install LM studio [here](https://lmstudio.ai) and use it to run the model from a http endpoint** 1. find and download model in discover <img width="944" alt="image" src="https://github.com/user-attachments/assets/7a379cc3-fe3f-4e4c-bc0f-9f17f753c849" /> 2. select model to load and load an osmosis model <img width="430" alt="image" src="https://github.com/user-attachments/assets/3d32833d-dcae-4f91-bb3f-47dd23e0570c" /> 3. start the web server with model loaded <img width="1130" alt="image" src="https://github.com/user-attachments/assets/51d4fd79-75dc-4808-8dd7-3b9f8c43ff77" /> 2. **Clone the repository (if applicable):** ```bash git clone https://github.com/Gulp-AI/Osmosis-MCP-4B-demo cd Osmosis-MCP-4B-demo ``` 3. **Install dependencies:** It's recommended to use a virtual environment. ```bash python -m venv .venv source .venv/bin/activate # On Windows: .venv\Scripts\activate pip install -r requirements.txt ``` 4. **Set up Environment Variables:** Create a `.env` file in the root of the project directory and add your API keys and model server configuration: ```env BRAVE_API_KEY="your_brave_api_key" GOOGLE_MAPS_API_KEY="your_google_maps_api_key" ACCUWEATHER_API_KEY="your_accuweather_api_key" APP_STYLE="gui" # or "tui" ``` if an api key is not provided, the tool will not be loaded. If `APP_STYLE` is omitted, the application defaults to GUI mode. The `main.py` script currently configures the model server URL directly. Ensure `http://localhost:1234/v1` is the correct endpoint for your Qwen model server (this is what lm studio uses). 5. **Serve local model:** Use a tool like lm studio to provide a usable endpoint. ## Environment Variables The application uses the following environment variables (loaded from a `.env` file): * `BRAVE_API_KEY`: Your API key for Brave Search. * `GOOGLE_MAPS_API_KEY`: Your API key for Google Maps. * `ACCUWEATHER_API_KEY`: Your API key for AccuWeather. * `APP_STYLE`: 'tui' or 'gui' ## Available Tools (MCP Servers) The agent is configured to use the following tools via MCP: * **Time:** Provides current time information. * **Brave Search:** Enables web search capabilities. * Requires: `BRAVE_API_KEY` * **Fetch:** Fetches content from URLs. * **Google Maps:** Provides location-based services. * Requires: `GOOGLE_MAPS_API_KEY` * **Weather:** Provides weather forecasts. * Requires: `ACCUWEATHER_API_KEY` * **Code Interpreter:** A built-in tool for executing Python code snippets. These servers need to be running and accessible for the agent to utilize their respective functionalities. The `main.py` script provides the commands to start these MCP servers. ## How to Run Graphical User Interface (GUI): This mode launches a web-based interface for interacting with the agent. This is the default mode. ```bash python main.py ``` The web UI will be accessible at the address provided by the `WebUI` component upon startup (on `http://localhost:7860`). ---