Sample App – On-Device AI Assistant for Android (github.com)

An innovative open-source Android app has been launched, demonstrating on-device AI capabilities using Gemma 4 and LiteRT-LM. This app allows users to chat, analyze images, and control their phone, all without internet access after the initial model download, enhancing privacy and portability. It showcases features like natural language conversations with real-time token streaming, multimodal image understanding (including object detection and OCR), and versatile phone automation tools, from sending SMS to adjusting device settings. Significantly, this app exemplifies the growing trend towards on-device AI, which reduces reliance on cloud services and boosts user privacy by ensuring that no data leaves the device. Key technical aspects include the use of Google's Gemma 4 model for speech recognition and image processing, along with a native ToolSet API for function calling, thereby streamlining user interactions. The application is built with Kotlin and incorporates a multi-module architecture for better organization and scalability, emphasizing contributions from the community to further enhance its capabilities. With a focus on efficiency and user autonomy, this app sets a new standard for mobile AI assistants.