June 2020 Summaries

Firebase can extend TensorFlow Lite deployments on mobile and edge devices by providing over-the-air model delivery, production performance monitoring, and A/B testing without requiring developers to build these capabilities independently. Teams can upload and publish TensorFlow Lite models through the Firebase Console or Model Management API, allowing models to be updated independently of app releases, downloaded on demand to reduce app size, and integrated into mobile applications through Firebase’s Android and iOS libraries. Firebase Performance Monitoring can record inference, preprocessing, and post-processing times across real user devices, helping developers identify performance differences by device type or operating system. Firebase Remote Config and A/B Testing can assign different model versions to user segments, measure business outcomes such as conversions or retention, and support broader rollout of the stronger-performing model. Android and iOS codelabs demonstrate these workflows using a handwritten-digit recognition application.

Firebase has introduced substantial improvements to Cloud Firestore and Cloud Storage Security Rules to make policies easier to write, read, debug, and scale. New Set support enables concise comparisons of document fields through operations such as checking required keys, intersections, differences, and membership, while new Map methods including diff() and get() simplify detecting changed fields and safely retrieving values with defaults. Functions can now use up to 10 local variables declared with let, allowing repeated access-control conditions and document paths to be organized into clearer reusable logic. A ternary operator also adds conditional if/else-style expressions for more compact authorization rules. In addition, the maximum compiled Rules file size has increased from 64 KB to 256 KB after performance optimizations, expanding capacity while preserving the low-latency decisions required by Security Rules.

Firebase describes its efforts to support increasingly complex mobile game development through tools, SDKs, Unity workflows, tutorials, and developer-focused updates announced around the Google for Games Developer Summit. The company highlights Gameloft, a global studio with more than 190 games and 80 million monthly players, which adopted Firebase Crashlytics after struggling to obtain actionable crash data, reproduce failures, and resolve technical issues quickly. Crashlytics groups crashes by root cause, shows their player impact, and supplies contextual details that help teams prioritize and investigate problems. Gameloft reported that using the service reduced the number of users encountering crashes by 10% overnight in Overdrive City, while also contributing to higher Play Store ratings and a 16% increase in session duration.

Firebase announced the public release of its Crashlytics SDK following a beta period adopted by tens of thousands of apps and used by hundreds of millions of end users. The release provides a migration path away from Fabric dependencies, while the legacy Fabric SDK is being deprecated and will stop sending crash reports to Firebase dashboards on November 15, 2020, along with beta Firebase Crashlytics versions. Developers must upgrade to Firebase Crashlytics version 17.0.0 or later for Android, 4.0.0 or later for iOS, or 6.15.0 or later for Unity to continue receiving crash reports, while new users are directed to the getting-started documentation and support resources.