ML Kit’s Prompt API: Unlock Custom On-Device Gemini Nano Experiences

Android's on-device generative AI has reached a significant milestone with the Alpha release of the ML Kit GenAI Prompt API. This new API empowers developers to send natural language and multimodal requests to Gemini Nano, offering greater control and flexibility. Partners like Kakao are already leveraging this API to build innovative experiences. The Prompt API moves beyond pre-built functions, enabling custom, app-specific generative AI use cases. It utilizes Gemini Nano for local data processing, ensuring offline capabilities and enhanced user privacy. Key use cases include image understanding for classification, intelligent document scanning, and transforming data for user interfaces. Developers can also use it for content prompting, analysis of reviews, and information extraction from emails. Implementing the Prompt API requires minimal code, allowing developers to create custom prompts and set generation parameters. The API currently performs best on Pixel 10 devices running Gemini Nano nano-v3, which shares an architecture with Gemma 3n. This shared foundation facilitates easier prototyping for developers, who can also experiment with Gemma 3n locally or via Google AI Studio. Developers can find detailed implementation guidance and samples on GitHub.