🚀 OmniRAG: Multi-Source Conversational AI Assistant

A production-style Retrieval-Augmented Generation (RAG) chatbot that enables conversational AI over multiple knowledge sources including documents, websites, REST APIs, and YouTube transcripts.

📌 Overview

OmniRAG is a multi-source conversational RAG application built with LangChain LCEL, Groq Llama 3.3 70B, HuggingFace Embeddings, and FAISS. Unlike traditional RAG systems limited to PDFs, OmniRAG lets users build a unified knowledge base from multiple data sources and query it through a natural conversational interface.

✨ Features

• 📄 Chat with PDF documents
• 📝 Upload DOCX, TXT, and CSV files
• 🌐 Query website content
• 🔗 Retrieve information from REST APIs
• ▶️ Chat with YouTube transcripts
• 💬 Conversational memory
• 🔍 Semantic search with HuggingFace Embeddings
• ⚡ Fast inference using Groq Llama 3.3 70B
• 🎯 MMR-based retrieval
• 📚 Metadata-aware document processing
• 📂 Multiple document upload support

🏗️ Architecture

User
 │
 ▼
Streamlit UI
 │
 ▼
PDF / DOCX / TXT / CSV / Website / API / YouTube
 │
 ▼
LangChain Document Loaders
 │
 ▼
RecursiveCharacterTextSplitter
 │
 ▼
HuggingFace Embeddings
 │
 ▼
FAISS Vector Store
 │
 ▼
MMR Retriever
 │
 ▼
Groq Llama 3.3 70B
 │
 ▼
AI Response

🛠️ Tech Stack

• Python
• Streamlit
• LangChain (LCEL)
• Groq (Llama 3.3 70B)
• HuggingFace Embeddings
• FAISS
• RunnableWithMessageHistory
• RecursiveCharacterTextSplitter

📂 Supported Sources

• ✅ PDF
• ✅ DOCX
• ✅ TXT
• ✅ CSV
• ✅ Websites
• ✅ REST APIs
• ✅ YouTube Transcripts

⚙️ Installation

git clone https://github.com/Vinay-Rai/OmniRAG-Multi-Source-Conversational-AI-Assistant.git
cd OmniRAG-Multi-Source-Conversational-AI-Assistant

python -m venv venv

# Windows
venv\Scripts\activate

pip install -r requirements.txt

🔑 Environment Variables

Create a .env file:

GROQ_API_KEY=YOUR_GROQ_API_KEY
HF_TOKEN=YOUR_HUGGINGFACE_TOKEN

▶️ Run

streamlit run app.py

🚀 Future Improvements

• User authentication
• Persistent vector databases (Qdrant/Chroma)
• OCR support
• Image & audio RAG
• SQL database integration
• GitHub & Notion connectors
• LangGraph multi-agent workflows
• Streaming responses
• Source citations

🤝 Acknowledgements

This project was inspired by the educational content of Krish Naik on Generative AI, LangChain, and Retrieval-Augmented Generation (RAG). The implementation extends those concepts with support for multiple heterogeneous knowledge sources, conversational memory, and a unified ingestion pipeline.

Special thanks to the teams behind: - LangChain - Groq - HuggingFace - FAISS - Streamlit

🔗 Links

• Live Demo: https://omnirag-multi-source-conversational-ai-assistant.streamlit.app/
• GitHub: https://github.com/Vinay-Rai/OmniRAG-Multi-Source-Conversational-AI-Assistant

👤 Author

Vinay Rai

• GitHub: https://github.com/Vinay-Rai
• LinkedIn: https://www.linkedin.com/in/vinay-rai

---

⭐ If you found this project useful, consider giving it a star!