Gemini completion returning endless line breaks as others have reported

Gemini API
, ,
1

I am calling Gemini 2.0 Flash and Gemini 1.5 Pro, and in my calls to each, they’re reliably returning a HUGE number of line breaks.

Here is the full trace of an instance of a conversation with 2.0 Flash, which should make all the parameters I’m using visible: https://us.cloud.langfuse.com/project/cm618klud027agh9gdkwu30bo/traces/9a310499-f130-4359-a8b5-b9d4f954738e?timestamp=2025-03-02T17%3A35%3A14.073Z

Please note the 2nd and 3rd Assistant messages, where the line breaks are visible.

This seems to have also been reported but unresolved here:

3

Hi @adambossy,

Welcome to the forum!

I was trying to replicate the issue you described using the same prompts, and it seemed to be working on my end.

Before I escalate the issue further, I wanted to check in with you and see if you’re still experiencing it. Also, I’m curious if you’ve experimented with the temperature setting (other than the default of 1). Let me know if you think adjusting that might help.

How often are you encountering this issue, and does it seem to happen with any specific types of prompts or after a certain amount of usage?

Any additional details you can provide would be very helpful!

Cheers!!