"Ask AI" button breaks UI when using streamed output in AI Explorers

I've been facing the same error,
When I enabled the OpenAI provider and checked in chat explorer with the "streamed" option
Following the screenshot below

Thank you!

Not sure, but this https://www.drupal.org/project/ai/issues/3519196 🐛 Handle errors for API explorers when a provider not configured Active might be related.

This issue specifically occurs when you do not assign the Stop Sequences at the right side. To ensure the UI doesn't break maybe a default value for a stop sequence should solve this.

In my case, I am using Gemini Provider.

I think the cause of this issue is that in case of streamed response the error messages are not handled properly. Maybe there is a format in which it expects the response to be and treats the error message the same way which causes this issue.

The issue was with stream.js file in the ai_api_explorer module, it was injecting a whole DOM content in the $message field in the ChatExplorer.php. I added validations to this part to ensure that the response that is sent back is not HTML as it indicates an error and have give a proper message to check the Stop Sequences because that the reason this UI breaking is happening.

I dont get the issue with the required fields filled and streamed checked. But when I don't fill the required field and streamed checked the issue is reproduced both in the issue branch and 1.2.x. please check the ss and update the steps to reproduce if I missed any step.

Thanks @techmantejas - found one issue in my code review, anyone can check and I'll merge it after.

"Invalid response received. Please check your configuration." Then I guess this much should be fine

https://git.drupalcode.org/issue/ai-3521390/-/commit/15fd6df6b60aca9bc5a...

Please check this

I could replicate the issue, and can confirm that checking out this MR also fixes the layout issue on my side.

However, the streaming functionality itself does not seem to be working for me. Unsure if this is something wrong with local configuration or an existing problem. Using gpt-4o model.

Getting merged now! Thanks all!

Also getting backported to 1.1.x

Thanks!

Remove outdated backport tag