powder-o/MCP-RAGbot

RAG Chatbot with MCP Tools

A Retrieval-Augmented Generation (RAG) chatbot that uses MCP (Model Context Protocol) tools to interact with a ChromaDB vector store and Groq's Deepseek model.

Features

• Vector Store: ChromaDB for document storage and similarity search
• LLM: Deepseek model via Groq API
• MCP Tools: FastMCP-based tools for document management and search
• CLI Interface: Simple command-line chat interface
• Document Management: Add documents, files, and search the knowledge base

Setup

1. Install Dependencies:

   pip install -r requirements.txt
   

2. Set up Environment:

   cp .env.example .env
   # Edit .env and add your Groq API key
   

3. Get Groq API Key:

   • Sign up at Groq
   • Create an API key
   • Add it to your .env file

Usage

Start the Chat Interface

python cli_chat.py

Available Commands

• /help - Show help message
• /add - Add a document to the knowledge base
• /addfile - Add a file to the knowledge base
• /search - Search the knowledge base
• /info - Show collection information
• /clear - Clear conversation history
• /quit or /exit - Exit the chat

Natural Chat

You can also just ask questions naturally. The AI will automatically:

• Search the knowledge base for relevant information
• Use that context to provide informed responses
• Invoke MCP tools as needed

MCP Tools

The system includes these MCP tools:

• search_documents: Search for relevant documents
• add_document: Add new document content
• add_file: Add a file's content to the vector store
• get_collection_info: Get collection statistics
• delete_document: Remove documents from the store

Testing

Run the test script to verify everything works:

python test_rag.py

Architecture

• vector_store.py: ChromaDB integration and document management
• mcp_server.py: MCP server with RAG tools
• groq_client.py: Groq API client for Deepseek model
• mcp_client.py: MCP client and agent logic
• cli_chat.py: Command-line chat interface
• test_rag.py: Basic functionality tests

graph TD
%% User Entry
      USER[User] --> CLI[cli_chat.py]

      %% Main Flow
      CLI --> CLIENT[mcp_client.py]
      CLIENT --> GROQ[groq_client.py]
      CLIENT --> SERVER[mcp_server.py]

      %% Core Components
      SERVER --> VECTOR[vector_store.py]
      GROQ --> API[Groq API]
      VECTOR --> CHROMA[(ChromaDB)]
      VECTOR --> EMBED[SentenceTransformer]

      %% Main Functions
      SERVER --> SEARCH[search_documents]
      SERVER --> ADD[add_document]
      SERVER --> ADDFILE[add_file]

      %% Vector Operations
      ADD --> CHUNK[Text Chunking]
      CHUNK --> STORE[Store Embeddings]
      SEARCH --> RETRIEVE[Retrieve Context]

      %% Data Flow
      STORE --> CHROMA
      RETRIEVE --> CHROMA
      RETRIEVE --> GROQ
      GROQ --> CLI

      %% Styling
      classDef user fill:#e1f5fe
      classDef app fill:#fff3e0
      classDef external fill:#e8f5e8
      classDef data fill:#f3e5f5

      class USER user
      class CLI,CLIENT,SERVER,VECTOR app
      class API,CHROMA,EMBED external
      class CHUNK,STORE,RETRIEVE data

The system automatically manages the RAG pipeline, searching for relevant context when needed and providing informed responses based on your knowledge base.