Mube's Chatbot - FastAPI + Groq + RAG

A full-featured AI chatbot built with FastAPI, WebSocket streaming, SQLite persistence, and document-based RAG (Retrieval-Augmented Generation).
It supports normal conversation, multi-chat history, and Multi document Q&A over PDF/DOCX/TXT files.

✨ Features

• Real-time streaming chat responses
• Persistent chat history with multi-chat sidebar
• Upload and query PDF, DOCX, and TXT files
• Multi-document merge into one searchable knowledge base
• FAISS-based vector retrieval with context-aware prompting
• Responsive frontend with code block rendering

🧠 How It Works

1) Standard Chat Flow

• Browser gets a persistent user_id in local storage
• Frontend sends user message to backend through /ws (WebSocket)
• Backend creates chat on first message, stores messages in SQLite
• Prompt is built using recent history and sent to Groq model
• Assistant response streams back to UI in real time

2) Document (RAG) Flow

• User uploads document to /load_document/
• Backend parses file, chunks content, creates embeddings using Hugging Face Inference API model sentence-transformers/all-MiniLM-L6-v2
• Embeddings are saved in FAISS under vectors/<session_id>
• User asks questions via /query_document/
• Retriever pulls relevant chunks and model answers with context
• Q&A is saved to chat history when user_id and chat_id are provided

📁 Project Structure

chatbot-fastapi/
├── app.py
├── database.py
├── requirements.txt
├── chat.db
├── templates/
│   └── index.html
├── static/
│   ├── script.js
│   ├── style.css
│   └── mube-chat-bot-logo.jpeg
├── vectors/
├── temp_uploads/
├── Mube-chatbot(workflow).jpeg
└── README.md

⚙️ Installation

git clone <repo-url>
cd chatbot-fastapi
python3 -m venv fastapivenv
source fastapivenv/bin/activate
pip install -r requirements.txt

Create .env in project root:

GROQ_API_KEY=your_groq_api_key_here
HUGGINGFACEHUB_API_TOKEN=your_huggingface_token_here

Run the app:

python app.py

Open: http://localhost:8000

🛠 Tech Stack

• FastAPI
• Groq (llama-3.1-8b-instant)
• LangChain
• Embeddings: Hugging Face Inference API (sentence-transformers/all-MiniLM-L6-v2)
• FAISS
• SQLite
• Vanilla JavaScript + HTML + CSS

🚀 Workflow Diagram

▶️ YouTube Link

• Demo: https://youtu.be/FQVB1vDsxVg