Tome

The power of AI at your fingertips, entirely hosted on your phone. We at Runic Solutions have integrated an LLM engine into an android phone, allowing for private and accessible intelligence.

Comment

Runic Tome!

Inspiration

We want AI that is private, sustainable, and accessible — not controlled by corporations or cloud servers. Our goal is to empower users with personal intelligence running directly on their devices. In short, we have harnessed the warhorse that is Google's Gemma 3, on Android, entirely offline.

What it does

Tome is an offline Small Language Model chat app.
It lets users:

  • Chat with an AI locally
  • Store and clear conversations
  • Keep all data on-device — no internet required post-model download

How we built it

  • Integrated Gemma 3 1B IT using MediaPipe GenAI
  • Built a mobile-friendly chat interface
  • Implemented fully local conversation storage
  • Ensured legal compliance with open-source and Gemma terms

Challenges we ran into

  • Memory + performance optimization for mobile
  • Fast model loading and response time
  • Handling chat history locally
  • Working within model licensing restrictions

Accomplishments

  • Fully offline AI chat on a phone
  • Clean and simple LLM-style UI
  • Strong privacy-first design
  • Ability to import custom models

What we learned

  • On-device AI demands smart optimization
  • Privacy can be a core feature
  • Licensing and responsible AI matter
  • Mobile hardware is capable of much more

Built With

Share this project:

Updates