FullStackGPT - LangChain Web Applications

A collection of AI-powered web applications built with LangChain and Streamlit, demonstrating various use cases of Large Language Models (LLMs) for document analysis, quiz generation, and private AI assistants.

Features

1. DocumentGPT

• Upload PDF, TXT, or DOCX files
• AI-powered document Q&A using RAG (Retrieval-Augmented Generation)
• Vector embeddings with FAISS for semantic search
• Real-time streaming responses

2. QuizGPT

• Automatic quiz generation from uploaded documents
• Multiple choice questions based on document content
• Interactive quiz interface with scoring

3. PrivateGPT

• Local LLM integration for private document analysis
• No data sent to external servers
• Secure and confidential document processing

Tech Stack

• Framework: Streamlit
• LLM Integration: LangChain, OpenAI GPT
• Vector Database: FAISS
• Embeddings: OpenAI Embeddings with caching
• Document Processing: UnstructuredFileLoader
• Language: Python 3.11

Project Structure

fullstackGPT/
├── Home1.py                    # Main entry point
├── home.py                     # Homepage with demo
├── pages/
│   ├── 01_DocumentGPT.py      # Document Q&A application
│   ├── 02_QuizGPT.py          # Quiz generation application
│   └── 03_PrivateGPT.py       # Private LLM application
├── utils.py                    # Shared utility functions
├── notebook.ipynb              # Development notebook
├── requirements.txt            # Python dependencies
├── prompt.json                 # Prompt templates (JSON)
├── prompt.yaml                 # Prompt templates (YAML)
├── cache.db                    # Local cache database
└── .cache/                     # Embeddings cache directory

Installation

1. Clone the repository:

   git clone https://github.com/Shinhunjun/fullstackGPT.git
   cd fullstackGPT

2. Install dependencies:

   pip install -r requirements.txt

3. Set up OpenAI API key:

   export OPENAI_API_KEY="your-api-key-here"

4. Run the application:

   streamlit run Home1.py

Usage

1. DocumentGPT: Upload a document and ask questions about its content
2. QuizGPT: Upload a document to generate interactive quizzes
3. PrivateGPT: Use local LLM for confidential document analysis

Key Technologies

• LangChain: Framework for building LLM applications
• Streamlit: Interactive web application framework
• FAISS: Facebook AI Similarity Search for vector storage
• OpenAI: GPT models for natural language understanding
• RAG: Retrieval-Augmented Generation for accurate responses

Author

Hunjun Shin

License

This project is for educational purposes as part of Northeastern University coursework.