Published August 3, 2026

Jetpacker: Google's Intelligent Android App Showcase Built with Antigravity

Building GenAI features in an Android app usually means navigating through various models, APIs, and architecture choices. Google's Android team published a new technical series designed to walk developers through exactly those decisions.

At the center of the series is Jetpacker, a technical showcase app the team built from the ground up for this year's Google I/O using Antigravity. At its core, Jetpacker helps users plan, explore, and enjoy their next big adventure.

The app shows an overview of your trips, the itinerary of each trip, and details of each event on that trip, all following Android development best practices with a beautifully expressive Material UI design.

The series dives deep into each of the app's intelligent features, providing detailed implementation steps, code snippets, and architectural insights to help developers build their own intelligent Android applications.

On-Device, Cloud, and Hybrid Inference

Using an on-device model comes with no additional cloud inference costs, means you don't have to worry about internet connectivity, and lets users be confident that private information is processed locally, on the device, without any of their data being sent to the cloud.

In Jetpacker, the team chose on-device inference for three features:

  • The trip overview feature transforms a messy, multi-day itinerary into a concise, actionable summary using Gemini Nano through the ML Kit GenAI APIs, processing data locally on the device
  • The expense tracker automatically extracts structured data from receipt images to help users track travel spending, using the multimodal capabilities of Gemini Nano 4
  • The audio diary records, transcribes, and categorizes voice notes into relevant trip activities, powered by ML Kit Speech Recognition and GenAI Prompt APIs

Sometimes a use case requires AI models with greater world knowledge, a much larger context window, or greater ability in handling complex tasks. In that case, developers can switch from an on-device model to a cloud model instead.

Or, to get the best of both worlds, hybrid inference dynamically chooses either a cloud or on-device model at runtime. This lowers costs by moving inference to the device when it is available while still supporting all Android devices running the app.

In Jetpacker, several features use cloud or hybrid inference:

  • The place Q&A feature answers questions about specific locations by grounding responses in real-world data, using Firebase AI Logic integrated with Google Maps and web context
  • The review drafting feature composes detailed reviews for places the user has visited, leveraging both on-device and cloud models through Firebase AI Logic's new Hybrid inference API
  • Automatic chat translation dynamically translates chat messages in real time, demonstrating custom hybrid inference logic with specific considerations for when to choose each model

While not a feature you see in the app itself, the Android system integration opens up the app's core capabilities directly to the Android operating system. It uses the AppFunctions API to integrate with system-level intelligence.

Agenticness introduces a higher level of autonomy, enabling models to act as agents. Instead of a single inference call, an agent works towards a specific goal via an orchestration loop that allows it to reason, use tools, and adapt its path.

Depending on your requirements, these intelligent agents can run either in the cloud, directly on-device, or in a hybrid setup. For Jetpacker, the team added a booking assistant that automates end-to-end booking workflows directly within the application to streamline reservations.

It is built using A2UI and ADK running in the cloud, with the Android app functioning as a front-end to the multi-agentic system running in the cloud. In-app agentic workflows are listed as coming soon in the series.

The series also walks through the key architectural questions developers should ask when adding intelligence to an app:

  • Complexity: How complex is your setup? Are you doing a single inference call or do you need a more agentic flow?
  • In-app or Android System: Should your feature be built into your Android app or does it fit better as an Android system integration?
  • Cost and privacy: Can inference run on-device to save cost and protect user data, or does the feature need cloud-scale world knowledge?

For Flutter developers, the series is worth watching for a few reasons. It demonstrates how Google teams combine Antigravity, Firebase AI Logic, ML Kit, and ADK into a single production-style app, which mirrors the stack many Flutter teams are adopting for AI features.

The hybrid inference API in Firebase AI Logic is particularly relevant, since it offers a way to balance cost, privacy, and capability that applies equally well to Flutter mobile apps as to native Android ones.

The A2UI and ADK patterns used for the booking assistant are the same building blocks the Flutter ecosystem has been exploring for generative UI, making Jetpacker a useful reference for teams building agentic experiences in Flutter.

The Jetpacker series is available on the Android Developers Blog, with each post covering one feature area in depth, from on-device intelligence and cloud inference to system integration and agentic workflows.