🕉️ Gita Assistant: AI Spiritual Guide based on Bhagavad Gita

"Your personal AI guide to the timeless wisdom of Bhagavad Gita"

“Yoga is the journey of the self, through the self, to the self.” — Bhagavad Gita

An AI-powered spiritual assistant inspired by the timeless wisdom of the Bhagavad Gita, designed to provide guidance, self-reflection, and mindfulness through teachings of Lord Krishna.

📖 Model Description

The Gita Assistant is an AI-powered spiritual guide that provides personalized advice and solutions based on the sacred teachings of the Bhagavad Gita. This is a fully independent model fine-tuned specifically for understanding and responding to spiritual queries using Gita wisdom.

Developed by: Shubhagaman
Model type: RAG (Retrieval Augmented Generation) System
Language(s): English, Hindi, Sanskrit (Multilingual support)
License: MIT
Finetuned from: sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2
Repository: Gita Assistant
Embeddings Model: Gita Embeddings

🎯 What This Model Does

The Gita Assistant helps users find spiritual guidance by:

🔍 Understanding spiritual questions in multiple languages
📚 Retrieving relevant passages from Bhagavad Gita
💭 Providing contextual advice based on ancient wisdom
🧘 Offering practical solutions for modern life problems

🏗️ Model Architecture

This project consists of two main components:

1. Gita Assistant (This Repository)

Purpose: Main application and user interface
Contains:
- RAG pipeline implementation
- FAISS search index (geeta_finetuned.index)
- Gita text chunks (geeta_chunks.pkl)
- Application code and examples

2. Gita Embeddings (Separate Repository)

Purpose: Specialized text embedding model
Features:
- Fine-tuned on Bhagavad Gita content
- Multilingual understanding (English, Hindi, Sanskrit)
- Domain-optimized for spiritual texts

🚀 Quick Start

Installation

pip install sentence-transformers faiss-cpu torch transformers numpy

📊 Training Data

Model Fine-tuned On:

Primary Source: Complete text of Bhagavad Gita
Languages: English and Hindi translations
Chapters: All 18 chapters with 700 verses

Training Approach:

Domain-specific fine-tuning
Multilingual sentence embeddings
Spiritual context optimization

🌍 Intended Use Cases

✅ Appropriate Uses

Personal spiritual guidance and self-reflection
Academic study of the Bhagavad Gita
Counseling and mental well-being support
Cultural and religious education
Meditation and mindfulness practices

❌ Limitations

Not a substitute for professional psychological help
Limited to Gita teachings (does not cover other scriptures)
Cultural context may vary across interpretations
AI-generated guidance should be considered supplementary

⚠️ Ethical Considerations

Spiritual Sensitivity: Treats religious content with utmost respect
Cultural Context: Acknowledges diverse interpretations
Privacy: No personal data collection or storage
Transparency: Clearly identifies as AI-generated guidance

🔧 Technical Details

🧠 Model Specifications

Base Architecture: Transformer-based embeddings
Embedding Dimensions: 384
Max Sequence Length: 512 tokens
Supported Languages: English, Hindi, Sanskrit

📁 File Structure

Gita-assistant/
├── app.py # Main application code
├── geeta_chunks.pkl # Gita text chunks
├── geeta_finetuned.index # FAISS search index
├── requirements.txt # Dependencies
├── README.md # This file
└── examples/ # Usage examples
├── basic_usage.py
└── advanced_usage.py

📦 Dependencies

python sentence-transformers>=2.2.2 faiss-cpu>=1.7.4 torch>=1.9.0 transformers>=4.21.0 numpy>=1.21.0

🌟 Example Interactions

🧭 Example 1: Life Guidance

User Question:

"I'm feeling lost in my career, what should I do?"

Gita Assistant Response:

Based on the wisdom of Bhagavad Gita:

"You have the right to perform your prescribed duties, but you are not entitled to the fruits of your actions." (Chapter 2, Verse 47)

"Perform your duty equipoised, O Arjuna, abandoning all attachment to success or failure." (Chapter 2, Verse 48)

"It is better to engage in one's own occupation, even though one may perform it imperfectly, than to accept another's occupation and perform it perfectly." (Chapter 18, Verse 47)

Remember: The Gita teaches us to focus on our duties without attachment to results.

🧘 Example 2: Emotional Support

User Question:

"How to handle anger and frustration?"

Gita Assistant Response:

Based on the wisdom of Bhagavad Gita:

"From anger comes delusion, from delusion confusion of memory, from confusion of memory loss of intellect, and from loss of intellect one perishes." (Chapter 2, Verse 63)

"A person who is not disturbed by the incessant flow of desires can achieve peace." (Chapter 2, Verse 70)

"When a person gives up all varieties of desire for sense gratification, which arise from mental concoction, and when his mind finds satisfaction in the self alone, then he is said to be in pure transcendental consciousness." (Chapter 2, Verse 55)

Remember: The Gita teaches us to focus on our duties without attachment to results.

🔄 Updating and Improving the Model

➕ Adding New Training Data

from sentence_transformers import SentenceTransformer, InputExample, losses
from torch.utils.data import DataLoader
import pickle

# Load your current model
model = SentenceTransformer('Shubhagaman/Gita-embeddings')

# Prepare new training examples
with open('geeta_chunks.pkl', 'rb') as f:
    chunks = pickle.load(f)

train_examples = []
for chunk in chunks[:500]:
    train_examples.append(InputExample(texts=[chunk, chunk]))

# Fine-tune further
train_dataloader = DataLoader(train_examples, shuffle=True, batch_size=8)
train_loss = losses.MultipleNegativesRankingLoss(model)

model.fit(
    train_objectives=[(train_dataloader, train_loss)],
    epochs=1,
    show_progress_bar=True
)

# Save updated model
model.save_pretrained('./Gita-embeddings-updated')

🧱 Recreating Search Index

import faiss
import numpy as np

# After updating embeddings, recreate FAISS index
def create_new_index(model, chunks, index_path='geeta_updated.index'):
    embeddings = model.encode(chunks)
    dimension = embeddings.shape[1]
    index = faiss.IndexFlatIP(dimension)
    
    faiss.normalize_L2(embeddings)
    index.add(embeddings.astype('float32'))
    
    faiss.write_index(index, index_path)
    print(f"✅ New index created with {index.ntotal} vectors")
    return index

📊 Training Data

Model Fine-tuned On:

Primary Source: Complete text of Bhagavad Gita
Languages: English and Hindi translations
Chapters: All 18 chapters with 700 verses

Training Approach:

Domain-specific fine-tuning
Multilingual sentence embeddings
Spiritual context optimization

🌍 Intended Use Cases

✅ Appropriate Uses

Personal spiritual guidance and self-reflection
Academic study of the Bhagavad Gita
Counseling and mental well-being support
Cultural and religious education
Meditation and mindfulness practices

❌ Limitations

Not a substitute for professional psychological help
Limited to Gita teachings (does not cover other scriptures)
Cultural context may vary across interpretations
AI-generated guidance should be considered supplementary

⚠️ Ethical Considerations

Spiritual Sensitivity: Treats religious content with utmost respect
Cultural Context: Acknowledges diverse interpretations
Privacy: No personal data collection or storage
Transparency: Clearly identifies as AI-generated guidance

🔧 Technical Details

🧠 Model Specifications

Base Architecture: Transformer-based embeddings
Embedding Dimensions: 384
Max Sequence Length: 512 tokens
Supported Languages: English, Hindi, Sanskrit

📁 File Structure

Gita-assistant/
├── app.py                          # Main application code
├── geeta_chunks.pkl                # Gita text chunks
├── geeta_finetuned.index           # FAISS search index
├── requirements.txt                # Dependencies
├── README.md                       # This file
└── examples/                       # Usage examples
    ├── basic_usage.py
    └── advanced_usage.py

📦 Dependencies

python
sentence-transformers>=2.2.2
faiss-cpu>=1.7.4
torch>=1.9.0
transformers>=4.21.0
numpy>=1.21.0

🌟 Example Interactions

🧭 Example 1: Life Guidance

User Question:

"I'm feeling lost in my career, what should I do?"

Gita Assistant Response:

Based on the wisdom of Bhagavad Gita:

"You have the right to perform your prescribed duties, but you are not entitled to the fruits of your actions." (Chapter 2, Verse 47)

"Perform your duty equipoised, O Arjuna, abandoning all attachment to success or failure." (Chapter 2, Verse 48)

"It is better to engage in one's own occupation, even though one may perform it imperfectly, than to accept another's occupation and perform it perfectly." (Chapter 18, Verse 47)

Remember: The Gita teaches us to focus on our duties without attachment to results.

🧘 Example 2: Emotional Support

User Question:

"How to handle anger and frustration?"

Gita Assistant Response:

Based on the wisdom of Bhagavad Gita:

"From anger comes delusion, from delusion confusion of memory, from confusion of memory loss of intellect, and from loss of intellect one perishes." (Chapter 2, Verse 63)

"A person who is not disturbed by the incessant flow of desires can achieve peace." (Chapter 2, Verse 70)

"When a person gives up all varieties of desire for sense gratification, which arise from mental concoction, and when his mind finds satisfaction in the self alone, then he is said to be in pure transcendental consciousness." (Chapter 2, Verse 55)

Remember: The Gita teaches us to focus on our duties without attachment to results.

🔄 Updating and Improving the Model

➕ Adding New Training Data

from sentence_transformers import SentenceTransformer, InputExample, losses
from torch.utils.data import DataLoader
import pickle

# Load your current model
model = SentenceTransformer('Shubhagaman/Gita-embeddings')

# Prepare new training examples
with open('geeta_chunks.pkl', 'rb') as f:
    chunks = pickle.load(f)

train_examples = []
for chunk in chunks[:500]:
    train_examples.append(InputExample(texts=[chunk, chunk]))

# Fine-tune further
train_dataloader = DataLoader(train_examples, shuffle=True, batch_size=8)
train_loss = losses.MultipleNegativesRankingLoss(model)

model.fit(
    train_objectives=[(train_dataloader, train_loss)],
    epochs=1,
    show_progress_bar=True
)

# Save updated model
model.save_pretrained('./Gita-embeddings-updated')

🧱 Recreating Search Index

import faiss
import numpy as np

# After updating embeddings, recreate FAISS index
def create_new_index(model, chunks, index_path='geeta_updated.index'):
    embeddings = model.encode(chunks)
    dimension = embeddings.shape[1]
    index = faiss.IndexFlatIP(dimension)
    
    faiss.normalize_L2(embeddings)
    index.add(embeddings.astype('float32'))
    
    faiss.write_index(index, index_path)
    print(f"✅ New index created with {index.ntotal} vectors")
    return index

🤝 Contributing

We welcome contributions from:

Spiritual scholars and practitioners
AI/ML developers
Multilingual translators
Open-source enthusiasts

Contribution Guidelines

Fork the repository

Create a feature branch:

git checkout -b feature/amazing-feature

Commit your changes:
```
git commit -m "Add amazing feature"
```

Push to the branch:

git push origin feature/amazing-feature

Open a Pull Request

📜 Citation

If you use this model in your research or application, please cite:

@software{gita_assistant_2025,
  title = {Gita Assistant: AI Spiritual Guide based on Bhagavad Gita},
  author = {Shubhagaman},
  year = {2024},
  url = {https://huggingface.co/Shubhagaman/Gita-assistant},
  note = {Multilingual RAG system for spiritual guidance}
}

📞 Support

Issues: GitHub Issues
Discussions: Hugging Face Discussions
Email: [[email protected]]

🙏 Acknowledgments

The eternal wisdom of Bhagavad Gita
Hugging Face for the open-source ecosystem
SentenceTransformers library
The spiritual community for guidance and feedback

📄 License

This project is licensed under the MIT License — see the LICENSE file for details.

✨ May this assistant help you on your spiritual journey. ✨ 🕉️ Lokah Samastah Sukhino Bhavantu — "May all beings be happy and free"

🔗 Related Repositories

Gita Embeddings Model — The specialized embedding model used by this assistant
Bhagavad Gita Dataset — The training dataset used for fine-tuning

🆕 Latest Updates

v1.0 (Current): Initial release with multilingual support
Planned: Enhanced answer generation, more languages, mobile app

Last updated: December 2024

Downloads last month: -; Downloads are not tracked for this model. How to track