Quick answer code snippets display broken (sometimes)

lelovsky

The quick answer code blocks seem to sometimes be broken now. Asking for code with the "?" results in an inconsistent syntax highlighting, which makes the reading experience slightly worse. Can't tell how to replicate the bug. Sometimes it works, sometimes not. Just try to make some searches with Quick Answer which should result in a code snippet.

Examples:

It's a minor bug, but it makes me sad :< Please, fix.

I expect the Quick Answer to reliably output pretty code blocks with proper highlighting. It used to work perfectly some time ago. I think it started to show broken results since around one to two weeks ago.

ed

There might be a bug in displaying formatted data. When I Do the search 'kibana operator set resources?' The example yaml file is not formatted, but while the answer was being generated I thought it was

I have tried in several browsers (Firefox, Safari and Brave) and the the same result on multiple queries

When it succeeds, it succeeds in all browsers. When if fails, if fails in all browsers. Attached are a few print screens demonstrating the issue which has already been confirmed by a Kagi staff.

Let me know if I can be of further help.
Kind regards,
Ed

I expected the answer to be formatted on all queries.

ed

I seem to be a duplicate of
https://kagifeedback.org/d/6224-quick-answer-code-snippets-display-broken-sometimes

marszalek-cc

Hi,
I see that more often lately - I do search with question mark -> I've got AI summary at top. And for very short amount of time all code is highlighted, and then it flashes and is not highlighted, like on attached screen. Code is also not highlighted when it takes longer amount of time to AI to finish summarizing. See attached image.

Bye

I expect code in summary properly formatted & highlighted.

marszalek-cc

I have this effect on newest firefox and chomium browsers.

ed

Same as me as described on the post bellow yours 😉
https://kagifeedback.org/d/6238-quick-answer-not-formatting-output

a0js

Something recently changed for AI Assistant and Quick Answers that causes the fancy rendering to start while the AI is typing. Previously, the output was rendered fancily after the AI was done. As a result when a triple backtick is encountered in what should still be a code block the rendering breaks.

This prompt in Quick Answers and The Assistant illustrates the problem:

Write an example rustdoc for a simple fibonacci function that gives the nth fibonacci number?

This is the result:

The code block should not terminate early if a triple backtick is encountered within a code block.

brunomueller93

Hey, there!

Since recently quick answer fails to display code correctly.

Tested on MacOS in Safari.

To replicate:

Search for: how to update transitive dependencies in maven?
See that the answer is generated correctly.
See that as generation continues, correctly generated code boxes are broken again.

Expected output:
Colourful code boxes that contain xml code and format it accordingly.

Actual output:
Code boxes are only shortly rendered correctly and then broken afterwards as the answer generation continues. Instead of displaying code boxes, it will display code as text like this:

xml <dependency> [...] </dependency>