Hi everybody,
To help provide more insights into what the team is working on, we are going to use this thread to post a brief summary of some of the notable changes / new features / fixes / etc. that we are releasing in Firebase Studio.
Jump to:
Cheers,
Kirupa
Feature UpdatesWhen you bring your own API key, we now provide “thinking” summaries when you are using our AI assistance in code mode:
This will provide better insights into what our AI agent is doing.
When your Flutter app preview is loaded, you can connect to and start debugging immediately. We automatically now add a VSCode launch configuration file to enable this.
We added Thumbs Up/Down to the chat in our prototyper:
This will help us better tune our AI assistant to provide better quality responses over time.
Bug Fixes
General AI Chat Interface Updates
Fixed several minor issues, including CSS tweaks for the model picker and disabling spellcheck for code input areas.
Improved Error Handling
Unhandled promise rejections are now caught gracefully to prevent unexpected crashes.
Smarter Tab Locking
Tab locking behavior is now more predictable—webview-only groups lock only when not in the first column.
Updated Firebase Dependency
The Firebase dependency in our Next.js template has been bumped to 11.9.1, bringing in various upstream fixes.
AI Assistant Enhancements
Several improvements were made to better handle edge cases, provide clearer error notifications, and boost reliability.
Quieter Telemetry
Telemetry logs no longer clutter your browser console.
Edit App Flow Resilience
Added a retry step in the Edit App agent flow if model responses are incomplete—more resilient and robust.
Better Image Handling
Image attachments are now stored on disk and referenced in chat payloads for more efficient processing.
Monospace & Icon Themes Return
Official Monospace and file icon themes are back, now built directly in the Nix environment.
Reduce Prototyping Out-of-Memory Errors
Increased the Node.js heap size for our prototyping tools—this will reduce the occurrence of out-of-memory errors.
Firebase Studio Integration
Added a new “Firebase Studio” category to VSCode commands for managing public ports more intuitively.
Fewer Toast Notifications
Reduced the frequency of Toast notifications unless they’re truly necessary.
GitHub Workspace Naming Fix
Importing GitHub workspaces now avoids suggesting overly long names that could cause confusing errors.
Fewer Unwanted VM Shutdowns
Introduced a heartbeat mechanism to prevent VM shutdowns while you’re actively using the system.
@kirupa Amazing! Thanks for the update, it’s great to see this post.
I immediately noticed the four new buttons on the home screen of the community forum here as well.
Thanks for sharing and keeping us up to date.
Can you elaborate on this?
Start Debugging Sooner
When your app preview is loaded, you can start debugging immediately. We automatically now add a VSCode launch configuration file to enable this.Blockquote
Glad you noticed and liked the updates. I revised the description to be more clear. This is for Flutter Android preview apps. When your preview loads, you can now connect to and debug immediately.
Thanks for sharing here.
Feature Updates
Prototyping Status and Address Bar UpdatesDuring prototyping, we add loading state and update the address bar value for greater clarity on what the app is doing
TTS SupportGenerated Apps can use the new Text-To-Speech (TTS) capabilities with the gemini-2.5-flash-preview-tts model.
Updated Image Generation ModelPrototyper will use the googleai/gemini-2.0-flash-preview-image-generation model for image generation
Bug Fixes Gemini API Key Prompting / Authentication Fix
Fixed an issue where the prototyper failed to prompt users for a Gemini API key during the “Edit App” flow when required.
Code Rollback Stability
Resolved a bug where the rollback logic in the prototyper didn’t correctly insert a message after performing a Git revert, which could result in stale code being reapplied.
Live Preview Refresh
The prototyper now automatically refreshes the preview iframe after applying code changes via auto-fix, ensuring the latest updates are immediately visible.
Token Limit Warnings
Added chat UI warnings in the prototyper when the prompt or total token count approaches or exceeds the model’s token limits.
File Cache Management
Introduced a size limit to the prototyper context engine file cache. The system now automatically evicts the largest files when the limit is exceeded to maintain performance and stability.
Annotation Export Optimization
Set a maxWidth for exported annotations in the prototyper to reduce file and thread size during export.
These two features alone are a huge improvement to the users experience!
File cache management on the backend is great as well to keep things chugging along.
Thanks for the updates!
Great updates. Thanks for putting in the work and for keeping the communication lines open.
I know it hasn’t been formally announced yet, but…
The new Diff check feature in the prototyper is incredible!
Just what we needed. Keep it up!
Feature UpdatesWe Improved how file changes are displayed by showing added/removed lines counts and providing a detailed diff view with before/after content
Added more file paths (airules.md, GEMINI.md, .gemini/styleguide.md, etc.) for our AI assistant to check when looking for AI rules configuration:
When prototyping, your apps can use our new Carousel and Collapsible UI components to the using Radix UI and Embla Carousel
Bug Fixes
Stabilized Preview Server Launch
Fixed an issue where the preview server start command could repeatedly restart (“thrash”) when interacting with the UI iframe.
Smarter File Management during Prototyping
Large context state and thread history files are now automatically rotated and trimmed to improve performance. Image attachments from trimmed history are safely saved to disk.
Faster Response to Cancel Actions
Generation processes now respond more quickly when users cancel operations (e.g., by clicking “Stop”).
Easier Navigation Near Token Limits
Added a “See Files” link to the token usage warning. When you’re close to token limits in Prototyping mode, this makes it easier to jump to the code view
bro as i have used next js and react for webapp but actually gemini prototyper is nto supported for react native then will it work in future and can i get some date so that i can postpond my project in that date
You can create a new React Native app from our templates and use the AI chat to build your app. It is a different starting point than the prototyper, but it should still give you great results. Have you tried that?
i tired it i just said to make a react native but actually it said i am unable to do it as i can do only next js project etc soo i just asked you
Just want to thank the Firebase Studio team again. You all work hard to get these releases out early and often.
I suggest not using the prototyper but instead creating the app from our templates: https://studio.firebase.google.com/new/react-native-expo
Amazing! Great updates.
Questions:
Can you elaborate on how these work?
Support for More AI Rules Configurations
Added more file paths (airules.md, GEMINI.md, .gemini/styleguide.md, etc.) for our AI assistant to check when looking for AI rules configuration:
okay… btw thanks for help
Keeps getting better and better. Thanks, Firebase Studio team!
Haha, I personally thank you for fixing rules. It works flawlessly. Thank you again!
Thank everyone, the problem has been resolved
