AI Studio Applet repeatedly loses node_modules after container restart — Preview unusable

Google AI Studio
,
1

I am currently completely blocked from further development of an established application in Google AI Studio Build mode.

Application: WeddingComet
Applet ID: fea9ec5a-f0e2-47fc-adf0-0065fe234c..
Current application version: v2.0.50

The AI Studio Preview/dev environment repeatedly becomes unavailable, often after periods of inactivity, and cannot reliably recover.

This has now happened repeatedly, not as a one-time build error.

Symptoms and diagnostics:

  • Preview/dev server repeatedly disappears or fails to restart.
  • Recovery attempts have run for more than 1,300 seconds.
  • Refreshing AI Studio, restoring to a previous checkpoint, and even opening a new chat for the same Applet did not restore a stable environment.
  • The project source code and package.json are intact.
  • The AI Studio development container currently comes up with node_modules completely missing.
  • This causes:
    sh: 1: vite: not found
    Exit code 127
  • Previous automatic recovery attempts also encountered dependency-resolution failures and began modifying package.json in an attempt to repair the environment, forcing me to cancel and restore the project checkpoint.

I explicitly asked the AI Studio agent to diagnose the environment without modifying the application.

Its final assessment was:

  • WeddingComet source code: INTACT
  • package.json: INTACT
  • node_modules: MISSING
  • Root cause: CONTAINER / ENVIRONMENT
  • The recurring behavior is NOT normal or expected
  • Application-level fixes cannot permanently resolve the container lifecycle issue
  • “Escalation IS Required”

The agent stated that the managed AI Studio environment should automatically hydrate its dependencies when the container starts, but this is repeatedly failing for this Applet.

I have already submitted this through AI Studio’s bug/feedback mechanism, but I am posting here because the issue is completely blocking further development and there is no visible support ticket or status.

Could someone from the Google AI Studio engineering/platform team please inspect or reinitialize the managed development environment for this Applet?

I do not want the AI agent to keep modifying application code or dependency versions in order to compensate for an infrastructure/container problem.

I can provide the full Action History, screenshots, development App URL, and additional diagnostic details privately if needed.

Thank you.

2

Hello @Shlomo_Dvir ,

Thanks for reaching out. There is no mechanism at our end to “reinitialize the managed development environment for an Applet”. Nor do we have visibility on the series of steps which led the applet into its current ‘seemingly stuck’ state. Best way out for all such cases would be to remix it and work on the new ‘remixed’ applet.

3

Hi Sai,

Thank you for your response and for clarifying that there is currently no way to reinitialize the managed development environment of an existing Applet.

Since I posted the report, the Applet has started working again and I was eventually able to complete my changes and publish successfully. This suggests that, at least in my case, the problem was intermittent rather than a permanently corrupted Applet.

My main concern is with Remix as the recommended recovery path.

For a development or experimental project, creating a remixed Applet may be perfectly reasonable. In my case, however, this is an established application already running in production.

My concern is not only whether the source code is copied correctly. A production application also depends on its surrounding environment and integrations — database connections, secrets/environment variables, authentication configuration, GitHub integration, deployment configuration, Cloud Run association, custom domains, and potentially other Applet-specific settings.

If Remix does not preserve all of these, then recovering from a problem in the managed development environment would require me to manually reconstruct and revalidate the production environment. In that situation, Remix is effectively a migration to a new development environment rather than a recovery procedure.

Before using Remix as a fallback for a production application, I therefore need to understand exactly what is preserved.

Could you please clarify:

  • Does Remix preserve the current/latest source state of the Applet?

  • Are database connections and environment variables/secrets preserved?

  • Are authentication-related settings preserved?

  • Is the GitHub connection preserved?

  • Is the existing deployment / Cloud Run configuration preserved?

  • Are custom-domain and other production-related settings preserved?

  • If some of these are not preserved, is there another supported way to create a fresh managed development environment while keeping the existing production configuration intact?

At the moment I will continue using the existing Applet, since it has recovered and is working again.

However, for a production application, I would be reluctant to rely on Remix as the recovery mechanism unless I know in advance exactly what will need to be rebuilt or reconfigured afterward.

Thank you again for your help and clarification.
Shlomo.

4

The studio has been undergoing upgrade and should be back in service once the engineering team concluded their system update

5

Hi Ziam,

Thank you for the update.

Could you please let me know specifically which channels we should
monitor for confirmation that the AI Studio system upgrade has been
completed?

For example, will the update be posted in the Google AI Developers
Forum, an official status page, release notes, or another channel?

I’d like to know where to check so I can safely resume development
work once the upgrade is finished.

Thanks,
Shlomo