April 2025 Summaries

Firebase Data Connect, now renamed SQL Connect, has introduced preview Angular and React bindings that integrate generated GraphQL-based data operations with TanStack Query v5. Developers can enable framework-specific SDK generation in connector configuration to produce type-safe React hooks or Angular injectors from Data Connect queries and mutations, reducing the need for custom services or manual data-fetching logic. The bindings provide standard TanStack Query capabilities such as loading and error states, caching, request deduplication, mutations, and background updates, while preserving end-to-end typing from schema definitions to application components. Setup requires installing the relevant TanStack Query and Firebase binding packages, configuring a shared QueryClient, and optionally using the Firebase emulator, after which React components can call generated hooks and Angular components can consume generated query-result signals.

Firebase Data Connect, now renamed SQL Connect, introduces generated aggregate fields that let developers perform common database calculations directly in queries instead of defining SQL-backed `@view` types. Supported operations include row and non-null field counts, minimums, maximums, sums, and averages for numeric values, date and timestamp ranges, distinct-value retrieval and counting, grouped aggregates, and row- and group-level filtering through `where` and `having`. Using a movie-review example, the announcement shows how these fields can simplify tasks such as counting movies or ratings, finding release-year and review-date ranges, listing genres, calculating average ratings by genre, and aggregating related records across tables. More complex or nested calculations, such as averaging review counts per movie within each genre, still require `@view` and custom SQL. Because aggregates run on the server, they can improve application performance, reduce data-transfer costs, and limit client access to raw data.

Firebase Data Connect, now renamed SQL Connect, is generally available with pricing based primarily on service usage and a PostgreSQL database hosted on Cloud SQL, while vector similarity search can also incur embedding-generation charges through Vertex AI, now called Agent Platform. SQL Connect includes free monthly allowances of 10 GiB for network egress and 250,000 queries or mutations, after which egress follows Google Cloud Premium Tier transfer rates and operations cost $4 per million. A single SQL Connect operation can retrieve or modify many records, differentiating its pricing model from document-based Firestore reads and writes. Cloud SQL provides eligible default PostgreSQL instances with a three-month no-cost trial, limited to five trials per billing account and one per Firebase project, after which baseline costs begin around $9.37 per month depending on region and configuration. Direct Cloud SQL access, cross-region replicas, upgraded trial resources, private IPs, and read replicas may create additional Cloud SQL charges, while normal SQL Connect requests do not duplicate egress charges between the two services.

Android battery optimizations such as Doze mode can delay Firebase Cloud Messaging notifications, especially normal-priority messages, which are delivered opportunistically while a device screen is off. High-priority FCM messages can wake sleeping devices and are recommended for timely, user-visible alerts such as fraud warnings, chat messages, and delivery updates, while normal priority suits nonurgent notifications like bank statements. Developers can configure priority through the Firebase Admin SDK or FCM REST API, although Android-specific fields are unavailable in the Firebase console, and should avoid using high priority for non-visible updates because excessive misuse can cause throttling or deprioritization. Message handling in `onMessageReceived` should immediately render notifications, while longer work such as fetching images or remote content should use WorkManager, with expedited jobs for high-priority messages and regular work requests for normal-priority messages. Testing high-priority delivery in Doze mode using a device token and appropriately configured FCM payloads helps verify that notifications remain timely and reliable while respecting battery life.

Firebase App Hosting has reached general availability as a fully supported serverless platform for deploying production Angular and Next.js applications from GitHub repositories, automatically managing CI/CD, static generation, Cloud Run rendering, CDN caching, networking, scaling, and security. New production features include zero-downtime domain migration, Cloud Secret Manager integration, local emulation with configurable overrides and dynamic secret loading, improved build diagnostics, dashboard monitoring, and rollback support without rebuilding. The service also adds customization for Cloud Run resources, build and start commands, environment-variable scope, wildcard domains, VPC connectivity, and deployment regions in the United States, Taiwan, and the Netherlands. App Hosting now introduces a community adapter model with preview support for Nitro-based frameworks such as Nuxt and TanStack Start, plus Astro, alongside starter applications and improvements to Firebase’s JavaScript SDK for server-rendered apps, including FirebaseServerApp, Next.js cookie synchronization preview, and automatic SDK initialization across environments.

Genkit is an open-source framework for building AI-powered applications with unified access to models from Google, OpenAI, Anthropic, Ollama-supported open models, and other providers, alongside integrated development tooling and production observability for Google Cloud and Firebase. Following its stable Node.js release, Genkit has introduced an alpha Python SDK and beta Go SDK, offering consistent concepts for creating AI workflows, structured outputs, agentic tool use, plugins, testing, debugging, and evaluation. Python emphasizes concise asynchronous syntax and Pydantic-based typing for experimentation, while Go provides explicit, type-safe APIs aimed at scalable, increasingly production-ready applications. Available plugins support services including Google GenAI, Vertex AI—now called Agent Platform—Ollama, Firestore, Flask, Pinecone, pgvector, and telemetry tools, while the CLI and browser-based Developer UI enable developers to run flows, inspect execution traces, test datasets, and evaluate quality metrics. Both new SDKs remain below version 1.0 and may introduce breaking changes as Google works toward feature parity with Node.js and incorporates community feedback.

Google’s preview App Testing agent is a Gemini-powered feature in App Distribution designed to automate functional and UI testing for Android apps, addressing the cost and maintenance challenges of manual and traditional automated tests. Teams can define test goals, navigation hints, and optional success criteria in natural language, allowing the codeless agent to navigate apps, simulate user actions, and run tests concurrently across physical or virtual devices with different configurations. In an example travel app workflow, the agent searches for a dream trip and opens a result, while screenshots, action-by-action explanations, and pass-or-fail outcomes help developers investigate results. Although the agent can adapt to application changes and independently handle many flows, developers may need to refine tests with additional hints or smaller steps when navigation is inconsistent. Google is inviting users to try the feature in App Distribution and provide feedback to improve its reliability.

Firebase AI Logic, formerly Vertex AI in Firebase, expands Firebase’s generative AI capabilities with Gemini model access and tools designed for mobile and web applications, while Vertex AI has been renamed Agent Platform as of August 2026. New Gemini Live API support enables real-time, bidirectional audio and text conversations, including interruptions and streaming responses, with public-preview SDKs available for Android and Flutter. The platform now also supports React Native, allowing developers to integrate Gemini APIs into cross-platform applications, and Vertex AI Studio can generate Firebase SDK setup and initialization snippets for Android, Swift, web, and Flutter. These updates aim to simplify secure client-side AI integration and help developers create richer conversational and generative features using Firebase-managed infrastructure.

Firebase Studio is a preview cloud-based, agentic development environment from Google that combines Project IDX, Genkit, and Gemini in Firebase to support the full lifecycle of AI application development, from prototyping through deployment and collaboration. It enables users to generate functional Next.js AI web-app prototypes from natural-language prompts, images, or drawings, automatically integrating Genkit and a Gemini API key, while Gemini chat can help modify features, layouts, authentication, and AI flows. Developers can also work directly in a CodeOSS-based cloud IDE with Gemini-assisted coding, debugging, terminal access, and Firebase integrations, then preview applications through public URLs or QR codes and deploy them using Firebase App Hosting. The platform retains Project IDX capabilities, including customizable cloud VMs, emulators, templates, repository imports, and team templates, while adding agentic prototyping and deeper Firebase integration. Firebase Studio supports real-time workspace sharing and collaboration, plans early access to Gemini Code Assist agents for migration, model testing, and documentation, and offers a limited number of free workspaces depending on Google Developer Program membership, though some services may require billing.

Firebase Data Connect, now renamed SQL Connect, has reached general availability as a backend-as-a-service built on Cloud SQL for PostgreSQL, allowing developers to define GraphQL-based data models and operations while automatically generating secure endpoints and type-safe SDKs without custom resolvers. New capabilities include Gemini in Firebase support for generating schemas, aggregate queries, distinct and grouped results, atomic operations for counters and lists, and connections to existing Cloud SQL databases through managed migrations or brownfield deployment. SQL Connect also expands client support with generated SDKs for React and Angular, alongside new Android and iOS codelabs, and provides vector search plus integration with Agent Platform, formerly Vertex AI, for retrieval-augmented generation workflows. Its GA pricing includes free monthly allowances for network egress and up to 250,000 operations, followed by usage charges, while the underlying Cloud SQL PostgreSQL instance offers a three-month free trial before ongoing costs; vector-search applications may also incur embedding-generation fees.

At Cloud Next 2025, Firebase announced updates aimed at supporting the full lifecycle of AI application development, including the preview of Firebase Studio, a cloud-based agentic environment that combines Project IDX, Genkit, and Gemini tools to prototype, build, publish, and manage full-stack apps. A preview App Testing agent within App Distribution uses Gemini to generate, manage, and run natural-language-driven functional and UI tests across devices. Genkit expanded beyond its generally available Node.js offering with Python in Alpha and Go in Beta, while Vertex AI in Firebase added real-time Gemini Live API support, React Native compatibility, and Firebase SDK code generation through Vertex AI Studio; later updates renamed this offering Firebase AI Logic and renamed Vertex AI as Agent Platform. Firebase also made Data Connect generally available, providing a Cloud SQL Postgres-backed, GraphQL-based backend service with schema generation, expanded data operations, existing database connectivity, and Angular and React support. App Hosting reached general availability as a serverless hosting service for full-stack web applications, adding observability, rollback support, zero-downtime domain migration, a local emulator, infrastructure customization, broader regional deployment, and preview support for Nuxt and Astro.