PG1204 commented on PR #7920: URL: https://github.com/apache/texera/pull/7920#issuecomment-5612427949
> @PG1204 Do you think a screenshot can be useful? > > @Ma77Ball Do you want to review it? @chenlica I looked into it, and I don't think a screenshot is possible for this one. The fix only takes effect when one of these three tasks routes to a third-party chat provider, and right now no model for visual-question-answering, document-question-answering, or zero-shot-image-classification has a live entry in HF's inferenceProviderMapping (I checked approx. 800, 250, and 1280 models respectively for each of the tasks). Without that, _resolve_providers always falls back to hf-inference and the chat branch is never entered, so there's no UI state that differs before and after. This is really a consistency fix: image-to-text and image-text-to-text already read choices[0].message.content in this same file, and #7798 did the same for the text tasks, now these three were the remaining gap, where a chat reply would land in the result column as a raw JSON envelope. I verified it by exercising the emitted Python directly: the three tasks return the chat content, native hf-inference responses parse exactly as before, and the untouched branches are unchanged. -- 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]
