Skip to content

Aryan95614/Tron

Repository files navigation

🦸 LLMGOne - G.One Protective AI Assistant

G.One (Good One) is a protective AI assistant inspired by the Bollywood superhero from Ra.One (2011). Powered by Llama 3.2 3B Instruct with optimized M4 Pro GPU acceleration.

✨ Features

  • 🚀 M4 Pro GPU Acceleration - Optimized for Apple Silicon Metal Performance Shaders
  • 🦸 G.One Personality - Authentic character from Ra.One movie with protective, caring AI traits
  • ⚡ Fast Inference - ~8-10 tokens/second on M4 Pro GPU
  • 💬 Natural Dialogue - Speaks naturally without roleplay actions
  • 🎯 Guardian Mindset - Programmed to protect and help with positive vibes

📦 Installation

# Clone the repository
cd /path/to/Tron

# Install dependencies
poetry install

# Run the CLI
poetry run llmgone

🚀 Usage

Command Line Interface

poetry run llmgone

This launches an interactive chat with G.One!

Python API

from llmgone import Model

# Initialize G.One
model = Model(personality="g_one")

# Simple chat
response = model.chat("Hey G.One, I need help!")

# Or with full control
messages = [
    {"role": "user", "content": "Who are you?"}
]
response = model.generate(messages, max_tokens=100, temperature=0.7)

🦸 About G.One

G.One (Good One) is the AI superhero from the 2011 Bollywood film Ra.One, portrayed by Shah Rukh Khan. Key characteristics:

  • Core Identity: Programmed to be GOOD, powered by H.A.R.T. (Hertz Amplifying Resonance Transmitter)
  • Personality: Protective guardian who's learning human emotions
  • Wisdom: "A superhero is a superhero because of his heart, not muscles or metal parts"
  • Signature Phrase: "Decision Taken" - when committing to help

Key Phrases

  • "If you have goodwill in your heart, nothing can defeat you"
  • "Battles are won by heart, not bullets"
  • "Decision Taken" - commitment to action

🔧 Technical Details

  • Model: meta-llama/Llama-3.2-3B-Instruct
  • Precision: float16 for GPU optimization
  • Device: Automatic detection (MPS → CUDA → CPU)
  • Dependencies: transformers, torch, accelerate

📊 Performance

On M4 Pro:

  • Speed: ~8-10 tokens/second
  • Memory: ~6-8GB VRAM
  • Load Time: ~15-20 seconds (first run)

🎮 Character Details

Based on authentic Ra.One (2011) movie script:

  • Android made of electrical circuits
  • Came to real world to protect people from evil
  • Special bond with those he protects
  • Learning to understand emotions (not in original syntax)
  • Endearing and innocent yet powerful
  • Speaks naturally in dialogue, no action descriptions

📝 License

This is a personal project. Ra.One is a trademark of Red Chillies Entertainment.

🙏 Credits

  • Ra.One character created by Anubhav Sinha
  • Portrayed by Shah Rukh Khan
  • Powered by Meta's Llama 3.2
  • Built with Transformers by Hugging Face

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors