RAG App with NVIDIA Nemotron!

A Retrieval-Augmented Generation (RAG) system built with Ollama and ChromaDB, featuring a Streamlit web interface for document upload and testing.

Features

• Document Ingestion: Upload and process PDF documents
• Vector Storage: Persistent document embeddings using ChromaDB
• RAG Query System: Ask questions about uploaded documents
• Streamlit UI: User-friendly web interface for testing
• Error Handling: Comprehensive exception handling throughout

Project Structure

├── config.yaml          # Configuration settings
├── common_utils.py       # Utility functions
├── ingestion.py         # DocumentIngestion class
├── LLM.py              # RAGLLMSystem class
├── streamlit_app.py    # Streamlit web interface
├── requirements.txt    # Python dependencies
└── README.md          # This file

Classes Overview

DocumentIngestion Class

• load_documents(doc_paths): Load PDF documents
• chunk_documents(docs): Split documents into chunks
• init_vector_store(): Initialize ChromaDB vector store
• add_documents_to_store(chunks): Add chunks to vector store
• process_documents(doc_paths): Complete pipeline from paths to vector store

RAGLLMSystem Class

• query(question): Query the RAG system
• add_documents(doc_paths): Add new documents to knowledge base
• get_vector_store_info(): Get vector store status information

Installation

1. Install dependencies:

pip install -r requirements.txt

2. Make sure Ollama is running with the required models:

# Install Ollama if not already installed
# Then pull the required models
ollama pull nemotron-mini
ollama pull nomic-embed-text

Usage

Option 1: Streamlit Web Interface (Recommended)

1. Run the Streamlit app:

streamlit run streamlit_app.py

2. Open your browser to the provided URL (usually http://localhost:8501)

3. Use the interface to:

   • Initialize the RAG system
   • Upload PDF documents
   • Ask questions about your documents

Option 2: Python Script

from LLM import RAGLLMSystem

# Initialize the system
rag_system = RAGLLMSystem()

# Add documents
doc_paths = ["path/to/your/document.pdf"]
rag_system.add_documents(doc_paths)

# Query the system
response = rag_system.query("What is this document about?")
print(response)

Configuration

Edit config.yaml to customize:

• Ollama Models: Change LLM and embedding models
• Chunk Settings: Adjust chunk size and overlap
• Vector Store: Modify collection name and directory
• Retrieval: Set number of documents to retrieve

Streamlit UI Features

Sidebar

• System Status: Shows if RAG system is active
• Document Upload: Upload multiple PDF files
• Vector Store Info: Display current configuration
• Clear Chat: Reset conversation history

Main Interface

• Chat Interface: Interactive Q&A with your documents
• SideBar: To Upload Files

Next Steps

• Add support for more document types (DOCX, TXT, etc.)
• Implement parsing of images and tables from documents
• Add chat history persistence
• Include tools like Firecrawl, MCP Server
• Add APIs using FastAPI and Agent-ify it using LangGraph