PG1204 opened a new issue, #7195: URL: https://github.com/apache/texera/issues/7195
### What happened? For `question-answering`, `table-question-answering`, `zero-shot-classification`, `sentence-similarity`, and `text-ranking`, the per-task codegen builds a structured payload (context passage, table dict, candidate labels, sentence list). But when `_post_with_fallback` routes to a third-party provider via `_call_provider`, the OpenAI-compatible and unknown-provider branches send only `prompt_value` as a chat message and the structured fields are silently dropped. The user gets a hallucinated chat answer instead of extractive QA / similarity scores, with no indication anything was lost. Only `hf-inference` (which accepts the raw pipeline dict) works correctly. File: `.../huggingFace/codegen/HuggingFaceCodegenBase.scala` (`_call_provider`). ### How to reproduce? Run `question-answering` with a model served by an OpenAI-compatible provider (not hf-inference). The context column is ignored and the result is a generic chat answer rather than an extractive one. ### Version/Branch 1.3.0-incubating-SNAPSHOT (main) ### Commit Hash (Optional) _No response_ ### What browsers are you seeing the problem on? _No response_ ### Relevant log output ```shell ``` -- 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]
