[Bug] Message Creator crashes on German Umlaute (Workaround provided)

,

Today I played arount with the new “Message Creator” node, introduced in KNIME AP 5.5.0. There is also an example in the hub which is “draft” as of today but I could get it to work (the data is not included so I had to use my own one).

What I’m expecting very often, is an error message like “ERROR LLM Prompter 6:337 Execute failed: ‘utf-8’ codec can’t decode byte 0xfc in position 211: invalid start byte”. Playing around a little bit, it looks like the prompt length in the message creator is causing this issue. With very short prompts, it seems to work. I already checked for “invisible” characters and skipped every special character in the prompt when I figured out, that there is one “ü” in the prompt. I went back to a longer prompt and used the string manipulation node with replaceUmlauts($prompt$,true) and that solved the issue.

Using the same LLM prompter node with the same config and without the message creater (so simple text wrangling) doesn’t have any issues with the German Umlaute or other special characters so the issues seems to be in this specific new node. Workaround clear but of course it would be good to fix this bug.

1 Like

moving to features & ideas and tagging as bug :slight_smile:

2 Likes

Update: Crash also happens on French accents like in “à”

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.