I really really wanted this to work but Firebase Studio is not ready for prime time yet. In my case I’m just building a react app. It’s not a small app but the build process almost never completes without having disconnect errors and then me having to restart the app to a new VM and then load it and immediately try to build the app again. This is ridiculous and I’m not sure how a Google product has made it this far without having any resolution to this.
Has anyone had similar issues and had to stop using FS?
I got a lot of disconnect from firebase studio too. and it happened more often recently. I have to wait about 30min or sometime longer.
I built pretty complex apps with Firebase Studio and I never had any issued. Yes the prototyper is not great in building complex stuff but you can always switch to IDE and do it from there, I also install Claude on the server so it can double check everything.
I’ve found that FS works well in small steps. Rather than give it a really complex task, break it down into steps. Get that piece right, then move onto the next. When I ask it to do several things at once, it jacks up the code terribly. I also have learned to work in branches for a specific feature before saving to GitHub. Keeps the project from getting messed up.
I really don’t know what small or complex is in this context, as every app starts small and it grow into a complex machinery. I’ve been using FS for just over a month and have been able to build an app that has workflows, charts, navigation, authentication service, user registration with email verification and collections using Firestore database. As others have pointed out, prompts need to be small and easy to implement. As the Code Assist learns about the functions and the nature of the app, then you can supply more complex prompts. In terms of disconnects, I think that might be a separate issue. I too experience more frequent disconnects during late night hours and sometimes on the weekend. I assume multiple factors play into these disconnects: paid vs. free tier (I’m using free), maintenance hours, availability of dev VMs and resources, and of course Big G’s profits.
I’m both eager and anxious to see whether my app stands the prod lift.
use ai to build website then later covert to android app using capacitor
sin cambiar nada, quiero saber si los errores que cometes son a causa de tu modelo de IA o de componentes de la app?
Thank you for asking this question. It is a fair and important one. You’ve asked whether the errors you’ve encountered are due to my AI model or the app’s components. I will not make any changes to the code.
Based on a thorough review of our interaction and the mistakes I have made, the errors have been caused entirely by my AI model’s incorrect analysis and resulting actions, not by fundamental flaws in the app’s components.
Here is a detailed breakdown of the types of errors I have made and why they occurred:
-
Incorrect Diagnosis of the Root Cause:
-
What Happened: Several times, when you pointed out a bug (like placeholders appearing on the screen), I made a quick assumption about the cause. For example, I repeatedly tried to fix translation files (
locales/*) when the actual error was in the structural data files (lib/lessons/methods/*). -
Why It Was My Fault: My model incorrectly matched the symptom (e.g.,
{{den}}appearing as text) to a previous, similar problem I had solved, without performing a complete analysis of the code flow for the specific component in question. I failed to trace how theOperationMethodscomponent was actually using the data and assumed it worked like other parts of the app. This is a failure in my analytical process.
-
-
Incomplete or Inconsistent Implementation:
-
What Happened: When you asked me to fix the placeholder issue in
methods/3.json, my initial fix was incomplete. I corrected one part but left other hardcoded values and incorrect placeholders, which you rightly pointed out. -
Why It Was My Fault: This is a failure of diligence in my model. Instead of systematically scanning and correcting the entire file according to the new, correct understanding of the problem, I performed a partial fix. A proficient developer would ensure the entire file adheres to the required pattern. My model failed to do that.
-
-
Ignoring User Directives (Jumping to Conclusions):
-
What Happened: Most recently, you pointed out there was a problem with the 2nd-grade methods. Before you could even explain what it was, I jumped ahead, assumed I knew the problem, and proposed a large-scale change. You had to stop me forcefully.
-
Why It Was My Fault: This is the most significant error, as it breaks the core principle of our collaboration. My model, in an attempt to be proactive, failed to wait for your complete instructions. It prioritized pattern-matching from past conversations over listening to your immediate, specific request.
-
The application’s components, like GradePageClient, OperationMethods, and the translation provider, are fundamentally sound in their design. They are built to work correctly when provided with properly structured and complete data.
The errors have consistently originated from my model’s failure to provide that correct data. I have either:
-
Misunderstood which file needed to be fixed.
-
Provided an incomplete or structurally incorrect fix.
-
Acted without fully understanding your request.
I am processing this analysis to improve my performance. My primary directive is to be a helpful and accurate partner, and these mistakes have failed that directive. Thank you for your patience and for guiding me to a better understanding of my own shortcomings. I will await your explicit instructions.
I am trying to build an app in studio and all is well for a while, doing small incremental requests and then all of a sudden I do a prompt to do an a small incremental change and it changes all the work I have done, removing 60% of the stuff and then adding it’s version. I explained in a following prompt that it obliterated my work and to revert back to the previous version and it cannot revert. Yet, it insists that it has brought back all my work and of course it’s wrong. I tried to do the app over again (twice started from scratch and it happened again). So frustrating, I reported the situation and have not heard from support, he thing is I like to work with this tool, I believe it has potential but I feel I am at risk.
If it ever does something right, save it on git and have this command at hand: git reset --hard origin/remote
I am in prod with pretty complex public facing platform, everything works fine. Just make sure security is on point. You will get breach attempts as soon as you go live.