Title: Google AI Studio Build Agent returns “Invalid argument” even for simple prompts while Preview still works
Hello Google AI Studio team,
I am experiencing a persistent issue with the Build Agent in an existing, substantial Google AI Studio application project.
Environment
-
Model: Gemini 3.8 Flash
-
Browser: Firefox on Windows
-
Project: Existing application developed in Google AI Studio
-
Preview: Continues to work normally
Observed behavior
The Build Agent immediately returns:
Invalid argument
The interface displays:
Gemini 3.8 Flash — Ran for 0s
Even a simple prompt requesting only OK fails with the same error.
Read-only terminal diagnostics also fail with:
Invalid argument
The interface additionally reports:
1 error running the code
Previous interruption
An earlier operation reported:
-
Edited 1 file:
tmp/find_prompt.py -
Quota limits exceeded
I have not yet verified whether this temporary-file change affected the project or whether the latest Git checkpoint was created.
Why this matters
The application Preview still works, but the Build Agent cannot execute even simple prompts or read-only diagnostics.
I want to preserve the current project source files and any uncheckpointed changes. I have therefore avoided restoring an older version, deleting files, or attempting automated repairs.
Assistance requested
Could the team please investigate whether this behavior might be related to Build Agent execution, project snapshots, terminal execution, or the workspace session?
Please advise on the safest way to restore Build Agent functionality without losing existing project files or uncheckpointed changes.
I can provide screenshots and additional non-sensitive diagnostic details if required.
Thank you.