mcp_chatbot

The MCPChatbot example demonstrates how to integrate an MCP server pipeline with a customizable LLM (for example, Qwen) to build a chatbot that can call external tools through MCP servers. The project includes a built-in MCP server for Markdown processing and provides multiple interfaces: a simple CLI chatbot, two interactive terminal chat modes (regular and streaming), and a Streamlit web chatbot that visualizes MCP tool workflows. You can run single-prompt scenarios or engage in multi-turn conversations where the LLM optimizes calls to tools like the Markdown processor, all coordinated via MCP messages and a structured workflow trace.

Prerequisites:

• Python 3.10+
• Git
• (Optional) uv (Python wrapper for fast environment management)

Installation steps:

1. Clone the repository and navigate into it:

   git clone git@github.com:keli-wen/mcp_chatbot.git
   cd mcp_chatbot
   

2. Set up a virtual environment and install dependencies:

   # Create and activate a venv (example for macOS/Linux)
   uv venv .venv --python=3.10
   source .venv/bin/activate
   # Install dependencies
   pip install -r requirements.txt
   # If you prefer uv for faster installs
   uv pip install -r requirements.txt
   

3. Configure environment variables (example):

   cp .env.example .env
   # Edit .env to set your LLM API keys and paths
   

4. Ensure MCP server configuration is set to use the local uv path and the correct project paths in:
   • mcp_servers/servers_config.json Example entry:

   {
     "mcpServers": {
       "markdown_processor": {
         "command": "/path/to/your/uv",
         "args": [
           "--directory",
           "/path/to/your/project/mcp_servers",
           "run",
           "markdown_processor.py"
         ]
       }
     }
   }
   

5. (Optional) Run quick checks:

   bash scripts/check.sh
   

6. Run unit tests (optional):

   bash scripts/unittest.sh
   

Tips and common issues:

• Ensure paths in servers_config.json are absolute and correct for your environment.
• For Windows, refer to the Troubleshooting example in the README and adjust path formatting accordingly.
• Set the environment variables in .env to provide LLM API keys, model names, and folder paths used by the Markdown processor and result storage.
• The MCP pipeline supports multiple MCP tool calls per prompt and multi-turn chats; tune the LLM and tool configurations to match your use case.
• If you encounter API key errors, verify that LLM_BASE_URL and LLM_API_KEY (and related OLLAMA variables if using Ollama) are correctly set in .env.
• Use the provided examples (single_prompt, chatbot_terminal, chatbot_streamlit) to validate integration before extending with your own tools.