Archie - B.Arch RAG Chatbot

Intelligent study assistant for 5th-semester B.Arch students — powered by RAG, LLMs, and semantic search.

---

Project Overview

Archie is a specialized Retrieval-Augmented Generation (RAG) chatbot designed to help Bachelor of Architecture students efficiently access their 5th-semester course notes and materials. Instead of manually searching through extensive PDFs, students can now ask specific questions and receive contextual, academically-focused answers.

The system processes architectural course notes, creates semantic embeddings for intelligent search, and uses advanced language models to provide accurate responses.

---

What I Built

Core Architecture

• Modular RAG Pipeline: Complete document processing system from PDF ingestion to intelligent querying.
• Vector Database: Persistent storage using Chroma DB for efficient similarity search.
• LLM Integration: Connected Groq's Llama 3 70B model for high-quality response generation.
• Web Interface: Clean, responsive chat interface for easy student interaction.
• Monitoring System: Integrated LangSmith tracing for conversation tracking and system monitoring.

Technical Stack Used

• Document Processing: PyMuPDF for PDF text extraction, custom cleaning and chunking.
• Embeddings: HuggingFace's MiniLM for CPU-based semantic embeddings.
• Vector Storage: Chroma DB with persistence for fast retrieval.
• Language Model: Groq API with Llama 3 70B (llama3-70b-8192) for response generation.
• Backend: FastAPI for REST API and static file serving.
• Frontend: Vanilla HTML/CSS/JavaScript for chat interface.
• Monitoring: LangSmith integration for conversation and chain tracing.

---

System Architecture

b-arch-chatbot/
├── src/                   # Core processing modules
│   ├── loader.py          # PDF document loading
│   ├── cleaner.py         # Text preprocessing and cleaning
│   ├── splitter.py        # Document chunking with overlap
│   ├── embeddings.py      # Semantic embedding generation
│   ├── vectorstore.py     # Vector database operations
│   └── qa.py              # QA chain with prompt engineering
├── app/                   # Web application
│   ├── template/
│   │   └── index.html     # Student chat interface
│   ├── support.py         # Application helper functions
│   └── app.py             # FastAPI server with endpoints
├── scripts/
│   └── train.py           # Training pipeline orchestration
├── data/                  # Course PDF materials
├── chroma_db/             # Persistent vector database
└── requirements.txt       # Project dependencies

Running the System

Build / Train the Model

python -m scripts.train

Start the Chatbot

uvicorn app.app:app --reload

Future Enhancements

• Multi-semester Support: Cover all B.Arch semesters.
• Visual Content: Add support for architectural diagrams and drawings.
• Collaboration Features: Enable student discussion and shared notes.

---

Learn More

I’ve detailed the full story of building Archie – The B.Arch RAG Chatbot in this article:
Building Archie: A RAG-Based Chatbot for Architectural Education