About
A full-stack AI knowledge assistant demonstrating production RAG engineering over ancient wisdom texts.
DHARMA (Divine Healing And Reflective Mindfulness Assistant) is a production-quality RAG system that helps users explore philosophical questions using verses from the Bhagavad Gita and Patanjali Yoga Sutras — with intent-aware answers, inline citations, and multi-turn conversation memory.
This is a software engineering portfolio project — not religious authority, medical advice, or a replacement for scholarly study of primary texts.
The knowledge base includes 867 verses (700 Bhagavad Gita + 167 Yoga Sutras) with translation and explanation text embedded using BAAI/bge-small-en-v1.5 and stored in PostgreSQL with pgvector.
Your question is preprocessed and routed by intent. Metadata lookup resolves explicit chapter/verse references; otherwise hybrid retrieval (pgvector + BM25 + cross-encoder reranking) finds relevant passages. Groq LLM synthesizes a structured Markdown answer with inline citations, related follow-up questions, and scripture source cards.
Multi-turn sessions remember recent context so follow-up questions like “explain more” or “what did he teach?” build on prior answers without starting over.