-
Notifications
You must be signed in to change notification settings - Fork 10k
Description
What happened?
I asked gemini to perform the following tasks:
"go over all the 18 applications one by one, focus on academia only, which means looking at transcript, school grade, extracurrium activities and awards on │
│ STEM (e.g., coding, latin, math etc), and research achievement (e.g. admission and publications in highly recognizable research institute, top level │
│ competition awards, summarize and document the outstanding achievements of the applicant if there is any. Finally, produce a rank of all the applications in │
│ academia achievement."
Each application folder has a few pdf files like resume and etc. I have logged to my Google account, which has access to the Gemini 2.5 pro.
Gemini was able to process a few candidates without any issue, but then constantly return this error:
"API Error: request to https://cloudcode-pa.googleapis.com/v1internal:countTokens failed, reason: write ERANGE"
I asked it to "proceed". Sometimes it was able to continue. Other times, it just returned the same error again.
The error message is hard to understand. I am guessing this is some form of rate limiting?
What did you expect to happen?
Clearer error messages that are more actionable:
- switch the model
- reduce the task size
- raise the limit
?
Client information
- CLI Version: 0.3.3
- Git Commit: ddd4659
- Session ID: 61a4d62c-48e0-4bb6-b9cb-7dc0e479d0ae
- Operating System: darwin v22.17.1
- Sandbox Environment: no sandbox
- Model Version: gemini-2.5-pro
- Memory Usage: 516.0 MB
Login information
No response
Anything else we need to know?
No response