DanielLeens commented on issue #9998:
URL: https://github.com/apache/seatunnel/issues/9998#issuecomment-4849337915

   I traced the current custom LLM response path here, and this does look like 
a real issue rather than a config mistake.
   
   The problematic path is the custom provider normalization step: the value 
extracted from `custom_response_parse` is currently funneled through 
list-oriented handling before `LLMTransform` turns it into the final output 
field. That is why a response whose `choices[0].message.content` is logically a 
single string can still end up written as an array-looking string in the output 
column.
   
   There is active follow-up for this now in PR #11235.
   
   So this issue should stay open until that fix is merged into `dev`. If you 
need a temporary workaround before then, and your endpoint is 
OpenAI-compatible, using the built-in `OPENAI` provider path instead of the 
`CUSTOM` path may help avoid this specific normalization behavior.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to