Lab Assistant AI

A RAG-based AI chatbot for the CSE department's Advanced AI & Deep Learning Lab. Students can ask questions about lab timings, hardware, coursework, policies, and troubleshooting — and get accurate, context-grounded answers powered by the Grok API.

Tech Stack

• Backend: Python, FastAPI
• LLM: Grok API (xAI) — grok-3-mini
• RAG Pipeline: LangChain + ChromaDB + sentence-transformers
• Frontend: HTML / CSS / JavaScript (SSE streaming)

Project Structure

LabAssistantAi/
├── .env                       # Grok API key (gitignored)
├── config.py                  # Settings (API URL, model name, system prompt)
├── rag.py                     # RAG pipeline (embedding, vector store, chain)
├── main.py                    # FastAPI server with SSE streaming endpoint
├── requirements.txt           # Python dependencies
├── knowledge_base/
│   └── lab_manual.md          # Lab knowledge base document
└── static/
    ├── index.html             # Chat UI
    ├── style.css              # Dark theme styling
    └── script.js              # Chat logic with SSE streaming

Setup

Prerequisites

• Python 3.10+
• A Grok API key from console.x.ai

1. Clone and install dependencies

cd LabAssistantAi
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

2. Add your Grok API key

Edit the .env file:

XAI_API_KEY=xai-your-actual-key-here

3. Run the app

source venv/bin/activate
uvicorn main:app --reload --port 8000

4. Open in browser

Go to http://localhost:8000

The first run downloads the embedding model (~80MB) and indexes the knowledge base into ChromaDB. Subsequent starts load from disk instantly.

Configuration

Edit config.py to change:

• MODEL_NAME — switch between Grok models (e.g., grok-3-mini, grok-2)
• CHUNK_SIZE / CHUNK_OVERLAP — tune document chunking
• SYSTEM_PROMPT — customize the assistant's behavior

To update the knowledge base, edit knowledge_base/lab_manual.md and restart the server. The vector store rebuilds automatically when the document changes.