Groq API generates a lot of extra output with built-in functions

[umasse]

Describe the bug
问题描述
I have configured the Groq API with the new Llama 3.1 70B model. Most of the time it works well, but sometimes I get too much output of nonsensical words. It seems to happen with other models as well. It just happened with the right-click menu "Explain".

To Reproduce
如何复现
Steps to reproduce the behavior:

1. Configure Groq API and set it as default module in ChatGPTBox options. See here: Unable to use Groq API as custom model #780
2. On any page, select a piece of text.
3. Right-click and choose "Explain".
4. See the results, which are nonsensical and too long.

Expected behavior
期望行为
Show a simple explanation of the selected text, in English.

Screenshots
截图说明
Example using the function "Explain" with Llama-3.1-70B-Versatile. I selected "klein-jail" from this page, just as a random test, and right-click menu function "Explain":
https://www.smbc-comics.com/comic/outside
The output is somewhat ok for the first 2 lines, then becomes non-sensical.

I would expect an output more like this:

Please complete the following information):
请补全以下内容

• OS: Linux
• Browser: Brave
• Extension Version: v2.5.8

[josStorer]

I believe this is not an issue with chatgptbox, you can reproduce it on the Groq official website.

[umasse]

Ah, that's interesting!
And in the playground I can see that it actually happens also randomly. Sometimes the answer is perfectly fine.
I have reported it to Groq. Thanks for testing.