GitaGuru AI
RAG-Powered Bhagavad Gita Chatbot
An intelligent chatbot that delivers hallucination-free, cited answers from all 700 verses of the Bhagavad Gita. Built with a modern RAG architecture using FastAPI, OpenAI embeddings, ChromaDB, and deployed on AWS EC2 for production-grade availability.
The Problem
Why GitaGuru?
The Bhagavad Gita contains 700 verses across 18 chapters, covering profound wisdom on duty, dharma, devotion, and self-realization. Finding specific teachings or understanding their context is challenging with traditional search or reading methods.
Existing AI chatbots often hallucinate — generating plausible-sounding but fabricated answers. For sacred texts, accuracy and faithful citation are non-negotiable.
The Solution
RAG Architecture
GitaGuru uses Retrieval-Augmented Generation (RAG) to ground every answer in actual verse content. Your question is embedded, matched against pre-indexed verses in ChromaDB, and the most relevant verses are fed as context to the language model.
The result: every response includes exact chapter and verse citations, ensuring transparency and verifiability. No hallucinations — only the Gita's own words.
Capabilities
Key Features
Verse-Level Retrieval
Retrieves precise verses from all 18 chapters and 700 shlokas of the Bhagavad Gita, returning the exact verse along with its context.
Hallucination-Free Answers
Every response is grounded in actual verse content using RAG architecture — no fabricated or speculative answers. Each response cites the chapter and verse.
Semantic Understanding
OpenAI embeddings capture deep semantic meaning, enabling natural language queries like 'What does Krishna say about duty?' to surface the right verses.
Fast & Scalable
Containerized with Docker and deployed on AWS EC2 for reliable, low-latency responses. ChromaDB vector store enables millisecond-level retrieval.
Modern Chat Interface
A clean, intuitive chatbot UI built for conversational exploration of the Gita's teachings — ask follow-ups and dive deeper into any topic.
Cited & Transparent
Every answer includes chapter and verse citations, so you can verify the source and explore the original text independently.
System Design
Architecture
Query Flow: User Question → Embedding → ChromaDB Similarity Search → Top-K Verse Retrieval → Context Assembly → GPT-4o-mini Generation → Cited Response
Built With
Tech Stack
Backend
FastAPI
High-performance Python API framework powering the backend
AI/ML
OpenAI Embeddings
Text-embedding-3-small for semantic similarity search
Database
ChromaDB
Vector database for storing and querying Gita verse embeddings
AI/ML
GPT-4o-mini
Language model for generating grounded, cited responses
DevOps
Docker
Containerized deployment for consistent environments
Infrastructure
AWS EC2
Cloud compute for production hosting and availability
Language
Python
Core language for backend, data processing, and ML pipelines
Frontend
React
Modern component-based UI for the chatbot interface
Try GitaGuru AI
Ask any question about the Bhagavad Gita and receive cited, verse-level answers grounded in the original text.