Conversational RAG With PDF Uploads and Chat History

A Streamlit web application that lets you upload PDF documents and have a conversational Q&A session with their content. Built using Retrieval-Augmented Generation (RAG) with LangChain, Groq, and ChromaDB.

How It Works

1. PDF Upload — Upload one or more PDF files through the web interface.
2. Document Processing — PDFs are parsed, split into chunks (5000 chars with 500 overlap), and embedded into a ChromaDB vector store using HuggingFace's all-MiniLM-L6-v2 model.
3. Question Contextualization — Each new question is reformulated into a standalone query using the chat history, so follow-up questions like "tell me more about that" resolve correctly.
4. Answer Generation — Relevant document chunks are retrieved from the vector store and passed to the Groq LLM (llama-3.1-8b-instant) to generate a concise answer.
5. Chat History — Conversation history is maintained per session, enabling multi-turn dialogue about the uploaded documents.

Architecture

User Question
      |
      v
[ Contextualize Question ] -- chat history --> Standalone Question
      |
      v
[ ChromaDB Retriever ] -- vector search --> Relevant Document Chunks
      |
      v
[ Groq LLM (llama-3.1-8b-instant) ] -- context + question --> Answer
      |
      v
  Response displayed in Streamlit UI

Prerequisites

• Python 3.10+
• A Groq API key

Setup

1. Clone the repository

   git clone https://github.com/sothulthorn/RAG-Q-A-Conversation.git
   cd RAG-Q&A-Conversation

2. Create and activate a virtual environment

   python -m venv .venv
   
   # Windows
   .venv\Scripts\activate
   
   # macOS/Linux
   source .venv/bin/activate

3. Install dependencies

   pip install -r requirements.txt

4. Create a .env file (optional, for HuggingFace token)

   HF_TOKEN=your_huggingface_token_here
   

Usage

1. Start the application

   streamlit run app.py

2. Enter your Groq API key in the password field.

3. Upload one or more PDF files using the file uploader.

4. Ask questions about the uploaded documents in the text input field.

5. Use Session IDs to maintain separate conversation threads (default: default_session).

Project Structure

RAG-Q&A-Conversation/
├── app.py              # Main Streamlit application
├── requirements.txt    # Python dependencies
├── .env                # Environment variables (create manually)
└── README.md           # This file

Dependencies

Package	Purpose
streamlit	Web UI framework
langchain-classic	RAG chain orchestration
langchain-chroma	ChromaDB vector store integration
langchain-community	PDF loader, chat message history
langchain-core	Prompts, runnables, base classes
langchain-groq	Groq LLM integration
langchain-huggingface	HuggingFace embedding models
langchain-text-splitters	Document chunking
python-dotenv	Environment variable loading
pypdf	PDF parsing backend
chromadb	Vector database
sentence-transformers	Embedding model runtime

Configuration

Parameter	Value	Location
Embedding model	all-MiniLM-L6-v2	app.py:27
LLM model	llama-3.1-8b-instant	app.py:39
Chunk size	5000 characters	app.py:67
Chunk overlap	500 characters	app.py:67
Max answer length	3 sentences	System prompt

Troubleshooting

Groq API errors

Ensure your Groq API key is valid and has available quota at console.groq.com.