My First RAG Assistant - Foundations of Agentic AI

GitHub Repository: https://github.com/sarthak-cs/RAG_chatbot

---

Project Summary

This project is a Retrieval-Augmented Generation (RAG) chatbot built using Python, LangChain, ChromaDB, HuggingFace embeddings, and Google Gemini 2.5 Pro.
The assistant can answer queries intelligently by retrieving relevant information from a custom dataset and generating context-aware responses.

Objective: To demonstrate the fundamentals of agentic AI by integrating retrieval and generative AI pipelines.

---

Key Features

• Document ingestion from .txt files
• Recursive text chunking for large datasets
• Embedding using HuggingFace MiniLM
• Storage and retrieval via Chroma vector database
• Smart, human-like responses with Gemini 2.5 Pro
• Console-based interactive chatbot interface

---

How It Works

1. Load a custom dataset
2. Split documents into manageable chunks
3. Generate embeddings and store in ChromaDB
4. For each user query:
   • Retrieve relevant chunks from the database
   • Pass context + query to Gemini 2.5 Pro
   • Return AI-generated answers

---

Demo / Run Instructions

pip install -r requirements.txt
python src/rag_chatbot.py

---

Environment Setup

Create a .env file with your Gemini API key:

GEMINI_API_KEY=your_api_key_here

---

Potential Enhancements

• Web-based GUI for easier user interaction
• Support for additional file types (PDF, CSV, DOCX)
• Logging and analytics of user queries
• Multi-source knowledge integration

Learning Outcomes

• Hands-on experience with RAG pipelines
• Practical use of LangChain and vector stores
• Integration with generative AI APIs
• Structuring and publishing AI projects professionally