May 2024 Summaries

Firebase App Hosting is a new serverless hosting service for full-stack, server-rendered web applications built with frameworks such as Angular and Next.js, while the original Firebase Hosting remains the recommended option for static sites. App Hosting builds static assets with Cloud Build, deploys dynamic server-side content through Cloud Run, and distributes cached content via Cloud CDN, with automatic scaling and cost-oriented defaults. It uses framework-aware build adapters based on Cloud Native Buildpacks and supports GitHub-centered deployments that connect releases to specific commits and build applications in reproducible cloud environments. Its regional architecture is intended to improve fault isolation compared with the original Hosting service’s older global structure. Because App Hosting is still in preview and currently supports fewer frameworks, lacks preview channels, requires the Blaze pricing plan, and is less suited to purely static sites, Firebase recommends the original Hosting product for static applications, temporary pull-request previews, Spark-plan users, and projects using frameworks such as Flutter, Svelte, or Astro.

Firebase Data Connect, now renamed SQL Connect, introduces Query-Defined Infrastructure, a development approach that combines PostgreSQL, declarative data modeling, GraphQL queries, generated APIs, and type-safe client SDKs within the Firebase ecosystem. Unlike Firebase Realtime Database and Firestore, which allow clients to access databases subject to configurable security rules, SQL Connect generates secure server endpoints from a defined schema and a fixed set of application queries. The product aims to reduce repetitive work in traditional three-tier applications, where developers separately implement database schemas, server logic, APIs, authorization, and client-side types, while avoiding tradeoffs associated with backend-as-a-service systems and server-side web metaframeworks. Developers define relational models and authorized GraphQL operations, and SQL Connect creates PostgreSQL schemas and migrations, SQL queries, managed endpoints, authentication policies, and SDKs for web, iOS, and Android. By allowing untrusted clients to execute only pre-registered operations, the platform seeks to offer the rapid iteration of a unified backend service with the control, security, and maintainability of a dedicated application tier.

Gemini in Firebase, now available through the Firebase console and built on Gemini for Google Cloud, is an AI assistant designed to help developers answer product questions, generate and explain code, create tests, troubleshoot issues, and improve application quality while supporting enterprise security, privacy, and compliance needs. Its Crashlytics integration provides AI-generated crash summaries, suggested fixes, sample code, and best-practice guidance to accelerate debugging and triage. Firebase developed the service by fine-tuning Gemini with LoRA using Firebase documentation, internal support playbooks, and crowdsourced prompts, while using retrieval-augmented generation with a vector database to ground responses in current Firebase knowledge and Responsible AI services for content safety. The team evaluates quality through a curated set of “golden prompts,” custom metric dashboards, support-team testing, and measures such as relevance and factual consistency, alongside specialized prompting and Android-related retrieval data for Crashlytics insights. The post notes that, as of August 2026, Vertex AI has been renamed Agent Platform, and invites developers to use Gemini in Firebase and submit feedback.

FirebaseServerApp, available in Firebase JavaScript SDK version 10.10.0 and later, is designed to bridge Firebase Auth sessions between client-side and server-side rendering applications, reducing duplicated authentication logic, race conditions, and unnecessary asynchronous work. Initialized similarly to a standard FirebaseApp but supplied with a client Auth ID token, it allows server-side Firebase Auth to sign in the existing user automatically and supports server-side operations such as database queries during rendering. The recommended approach is to pass tokens from the browser to SSR requests, potentially through a service worker that intercepts requests, attaches authorization headers, and refreshes tokens proactively. Because initialization occurs asynchronously in the background, applications should wait for authStateReady() or monitor auth-state changes before accessing authenticated Firebase services, then redirect users when no valid current user exists. FirebaseServerApp also uses reference counting to reuse equivalent instances efficiently, with deletion or release-on-garbage-collection options intended to help manage cleanup in complex SSR frameworks such as Next.js.

Firebase highlights several updates for Flutter developers, including the evolution of Vertex AI for Firebase into Firebase AI Logic, which provides client SDKs for integrating Gemini-based generative AI into Flutter apps with native Dart code and without necessarily building a custom backend. Firebase’s Flutter SDKs now support Flutter web applications compiled to WebAssembly through improved js_interop integration, aiming to improve browser performance. Firestore for Flutter has added average and sum aggregation queries, which can reduce client-side processing and costs, along with support for range and inequality filters across multiple fields in a single query. The FlutterFire CLI also introduced an install command that uses published bills of materials to select compatible Firebase plugin versions, supports upgrades and branch-specific versions, and simplifies Firebase setup and maintenance. Firebase encourages developers to consult release notes and documentation and to contribute feedback, feature requests, and code through FlutterFire and UserVoice.

Firebase Data Connect, now renamed SQL Connect, is a preview backend-as-a-service that combines Firebase development workflows with a scalable, secure Cloud SQL for PostgreSQL database for applications requiring relational data models. Developers define data models and relationships through GraphQL schemas, while the service generates PostgreSQL database definitions and migrations, secure API endpoints, predefined queries and mutations, and strongly typed client SDKs for web and mobile apps. It supports relational joins, filtering, many-to-many relationships, and controlled access because clients can execute only server-defined operations. SQL Connect also supports semantic search through vector embeddings generated via Vertex AI, now called Agent Platform, enabling similarity queries alongside conventional database filters. Although it is not yet production-ready, Firebase invited developers to join a gated preview and explore its documentation and demonstrations.

At Google I/O 2024, Firebase announced an expansion toward AI-powered application development, alongside later naming updates that renamed Vertex AI for Firebase to Firebase AI Logic and Vertex AI to Agent Platform. Key launches included Firebase Data Connect, a private-preview service that connects apps to PostgreSQL on Cloud SQL and generates database infrastructure, APIs, typed SDKs, schema migrations, and optional vector-embedding and nearest-neighbor search capabilities from GraphQL-based definitions. Firebase Genkit entered beta as a server-side framework for composing, testing, evaluating, debugging, and deploying generative AI workflows, while Vertex AI for Firebase SDKs in public preview enabled Kotlin, Swift, Dart, and JavaScript clients to access Gemini models directly with App Check protections. Firebase also introduced preview support for App Hosting, a framework-aware serverless hosting service for Angular and Next.js applications, and added vector storage and KNN queries to Firestore. Operational updates included Remote Config feature rollouts with real-time monitoring against holdout groups, a Crashlytics-powered Release Monitoring dashboard, and general availability of Gemini in Firebase, including AI-generated Crashlytics insights to help developers diagnose and resolve app errors.

Firebase Genkit, now in public beta and later associated with Google’s renamed Agent Platform, is an open-source framework for JavaScript and TypeScript developers building AI-powered Node.js application backends, with Go support planned. It provides a code-centric approach for integrating language models, vector databases, embedding services, agents, evaluators, and cloud services, including Google Gemini, Ollama-hosted open models, Chroma, Pinecone, Firestore, PostgreSQL with pgvector, and Vertex AI evaluation tools. Genkit includes a browser-based developer UI and CLI for initializing projects, experimenting with prompts and AI components, tracing interactions, and debugging workflows locally, including support for accelerated inference with NVIDIA RTX GPUs and Gemma through Ollama. For production, it supports deployment to Google Cloud Run, Express servers, and Firebase Cloud Functions, with built-in logging, metrics, tracing, authentication, App Check, and OpenTelemetry-based observability integrations. Released under the Apache 2.0 license, its plugin architecture is intended to support additional models and services, while potential use cases include intelligent agents, customer support, content creation, coding assistants, translation, and AI evaluation; the beta may introduce breaking API changes before a planned 1.0 release.

Google announced Vertex AI for Firebase, later renamed Vertex AI in Firebase and subsequently Firebase AI Logic, to let mobile and web developers access Gemini models directly through Swift, Kotlin, Java, Dart, and JavaScript client SDKs without building a separate backend service. The offering simplifies onboarding through Firebase tooling and supports text generation, multimodal prompts containing images, PDFs, video, and audio, real-time streaming, structured JSON output, system instructions, and function calling. Firebase App Check helps restrict Gemini API access to legitimate apps and devices, while Cloud Storage for Firebase supports secure file uploads for multimodal requests and Firebase Security Rules control access to stored content. Firebase Remote Config enables developers to change prompts and model versions remotely, target user groups, and conduct A/B tests without requiring app updates. The announcement originally described public preview availability and planned general availability, while later updates note that Vertex AI was renamed Agent Platform and the Vertex AI Gemini API became the Agent Platform Gemini API in August 2026.

Firebase App Hosting is a public-preview serverless hosting service for server-rendered Angular and Next.js applications that automates builds, backend provisioning, CDN delivery, and scaling on Google Cloud. It connects to GitHub for continuous deployment, using Cloud Build to create static assets, Cloud Run to execute dynamic server-side content, and Cloud CDN to distribute cached content globally, while tracking deployed versions to specific commits. The platform includes configuration-free framework adapters, support for features such as Next.js streaming and Angular deferred loading, and optional controls over runtime settings, service accounts, secrets, and access permissions. Built on Google Cloud, it provides automatic scaling, high availability, Secret Manager integration, and compatibility with Firebase tools including the updated JavaScript SDK, AngularFire, Data Connect, and Genkit. Firebase positions App Hosting as a specialized option for modern full-stack server-rendered apps rather than a replacement for the original Firebase Hosting, which remains recommended for static web applications.

Firebase Crashlytics has introduced AI Assistance, a Gemini-powered feature designed to help developers diagnose and resolve Apple and Android app crashes more quickly through the Crashlytics dashboard. Available in the issue details page’s Insights section, the tool analyzes stack traces, issue variants, data trends, and custom keys to provide concise explanations of likely causes, suggestions for reproducing or investigating errors, potential resolution paths, and practices intended to prevent similar issues. Access requires an owner, editor, or Cloud AI Companion role in the Firebase project and acceptance of the Gemini terms of service. Firebase positions the feature as part of its broader Gemini in Firebase offering and encourages developer feedback to improve its use in crash-triage workflows.