🕉️📿 Gemma-3-1B — Spiritual ISKCON Q&A Assistant
A lightweight, respectful, and spiritually-aligned chatbot trained to preserve and share Guru wisdom.
📖 Project Purpose
To enhance spiritual accessibility and preserve Guru teachings, this service-learning project aims to:
✨ Develop an intelligent chatbot trained on ISKCON Q&A audio transcripts using state-of-the-art NLP
✨ Transcribe, preprocess, and organize religious Q&A content from Chaitanya Guru into a RAG-optimized knowledge base
✨ Provide accurate, context-specific responses to questions about Hinduism, philosophy, and devotional life
✨ Support devotees and learners globally by disseminating authentic wisdom through a digital medium
This model is text-to-text fine-tuned on our curated and ethically sourced Q&A dataset.
✨ Features
✅ Responses inspired by Vaishnava philosophy & teachings
✅ Lightweight ✅ 1B parameters ✅ 4-bit quantized
✅ Runs locally on < 8GB RAM devices
✅ Supports RAG for improved factual accuracy
✅ Suitable for desktop, mobile, and edge deployment
⚠️ Guidance & Respect for Teachings
📌 This AI does not replace a spiritual master, temple authority, or scriptural study.
It aims to:
- Support curiosity & learning
- Maintain respect for the Guru-shishya tradition
- Avoid personal interpretation or misinformation
For deeper guidance, devotees are encouraged to consult qualified ISKCON teachers.
🛠️ Technical Details
- Base Model:
unsloth/gemma-3-1b-it - Fine-Tuning Method: LoRA (4-bit QLoRA)
- Format: GGUF for local inference
- Data: Scraped Q&A transcripts from ISKCON discourses
- RAG-Ready: QA pairs stored in vector DB