privateGPT MCP Server

# PrivateGPT OpenAI-Compatible API Agent ## Description The PrivateGPT OpenAI-Compatible API Agent is a client that communicates with a private GPT server and provides an interface to openai-compatible libraries and tools. It comes in two variations. Variation 1 (*openai_compatible_api.p*) uses the PGPT API directly, Variation 2 (*openai_mcp_api.py*) uses communicates through the PGPT MCP Server. --- ## Prerequisites - Python 3.8 or higher - Access to the PrivateGPT server - For Variation 2, access to MCP server --- ## Setup 1. **Clone the repository:** ```bash git clone [https://github.com/pgpt-dev/MCP-Server-for-MAS-Developments.git](https://github.com/pgpt-dev/MCP-Server-for-MAS-Developments.git) cd MCP-Server-for-MAS-Developments ``` 2. **Optional: Create and activate a virtual environment:** ```bash python -m venv venv ``` - **Windows:** ```bash .\venv\Scripts\activate ``` - **Unix/MacOS:** ```bash source venv/bin/activate ``` 3. **Install dependencies:** ```bash pip install -r .\agents\OpenAI_Compatible_API_Agent\requirements.txt ``` 4. **Customise configuration file:** - 4.1 **OpenAI-Compatible API via PGPT API:** Copy the `pgpt_api_proxy.json.example` file to `pgpt_api_proxy.json` e.g. with `cp .\agents\OpenAI_Compatible_API_Agent\pgpt_api_proxy.json.example .\agents\OpenAI_Compatible_API_Agent\pgpt_api_proxy.json` Make sure that the `pgpt_api_proxy.json` is configured correctly and contains all necessary fields. The file should look like this: ```json { "base_url": "<Url to your private GPT instance>", "proxy_user": "<Proxy User>", "proxy_password": "<Proxy Password>", "access_header": "<Access Header>", "groups": ["<Groups to access>"], "language": "en", "use_public": true, "api_ip": "0.0.0.0", "api_port": 8001, "whitelist_keys": [] } ``` - 4.2 **OpenAI-Compatible API via MCP Server:** Copy the `pgpt_openai_api_mcp.json.example` file to `pgpt_openai_api_mcp.json` e.g. with `cp .\pgpt_openai_api_mcp.json.example .\pgpt_openai_api_mcp.json` Make sure that the `pgpt_openai_api_mcp.json` is configured correctly and contains all necessary fields. The file should look like this: ```json { "email": "<Your PGPT account>", "password": "<Your PGPT password>", "groups": ["<Groups to access>"], "language": "en", "api_ip": "0.0.0.0", "api_port": 8002, "whitelist_keys": [], "mcp_server": { "host": "172.24.123.123", "port": 5000 } } ``` 5. **Start the AI agents:** - 5.1 **Start the OpenAI-compatible API agent that uses the PGPT API directly:** ```bash python -m agents.OpenAI_Compatible_API_Agent.Python.openai_compatible_api ``` - 5.2 **Start the OpenAI-compatible API agent that uses the MCP server to communicate:** ```bash python -m agents.OpenAI_Compatible_API_Agent.Python.openai_mcp_api ``` - 5.3 **In order to run the OpenAI Compatible Agent with Docker:** Make sure Docker is installed on your System. ```bash cd agents\OpenAI_Compatible_API_Agent ``` To build and run Docker in Foreground: ```bash docker compose up --build ``` To build and run Docker in Background: ```bash docker compose up --build -d ``` 6. **Generate API key** Generate a key with the following command: ```bash python -m agents.OpenAI_Compatible_API_Agent.Python.generate_api_key --email "<PGPT account>" --password "<PGPT Password>" ``` This API key can be used for OpenAI compatible Clients. Make sure you either enter it to the config under whitelist_keys (see 4.1 and 4.2) or whitelist_keys is empty (no check, but prone to errors) --- ## Utilisation - **Use libraries (like litellm, ollama, smolagents) or tools (like Msty) to use PGPT with the OpenAI API interface:** Some example usages for external tools: - Usage in Msty: - To use the API in MSTY, add a remote server and add "<http://ip:port>" and your API key - Usage in Continue: - To use the API in Continue (Visual Studio/Pycharm Plugin) add this to your config: ```json { "model": "pgpt", "title": "PGPT", "systemMessage": "You are an expert software developer. You give helpful and concise responses.", "apiKey": "<your-api-key>", "apiBase": "http://<ip:port>", "provider": "openai" } ``` - Usage with OpenAI Library: - See openai_test_client.py for an example. --- ## License This project is licensed under the MIT License - see the LICENSE file for details.