google-ai-edge/gallery ? reverse-engineered prompt

Build me a mobile app that's a gallery for showcasing on-device AI. The main idea is that users can download and run powerful open-source language models like Gemma completely offline on their phones, ensuring their data stays private.

The app should have several modes. I want a standard AI chat, but with a special "thinking mode" so users can see the model's reasoning. It also needs to be multimodal so users can ask questions about images from their camera or gallery. Add a tool for transcribing and translating audio recordings in real-time. Finally, include a "prompt lab" for advanced users to experiment with different models and settings, and a way for them to manage and benchmark the models they've downloaded onto their device.

Please build this as a native Android app using Kotlin. Feel free to use Google AI Edge for the on-device ML parts and check the latest docs if you need to.