PG1204 opened a new issue, #8486:
URL: https://github.com/apache/texera/issues/8486
### What happened?
`TextGenCodegen.parsePython` extracts the chat response with no guard at all:
return body["choices"][0]["message"]["content"]
`QaRankingCodegen.parsePython` guards only on key presence, at three sites:
if "choices" in body:
return body["choices"][0]["message"]["content"]
Both raise on a malformed provider response rather than degrading:
- `{"choices": []}` → `IndexError` (the `in body` check passes, the list is
empty)
- `{"choices": [{}]}` → `KeyError: 'message'`
- `{"choices": [{"message": {}}]}` → `KeyError: 'content'`
`parsePython` runs once per row, so a single malformed response from one
provider aborts the entire run. The native `hf-inference` paths in the same
functions already degrade politely via `body.get("answer", json.dumps(body))`,
so the two paths are inconsistent.
This path is reachable in practice: of the top 100 models by downloads, 45
for `text-generation`, 37 for `question-answering` and 72 for
`sentence-similarity` have a live third-party provider, so the chat branch does
execute for these tasks.
This is the same defect #7920 fixed in `ImageTaskCodegen`, raised in review
by @Ma77Ball. The remaining sites are in code merged earlier (#7798) and were
left out of that PR to keep the diff focused.
### How to reproduce?
Run the operator with `text-generation` (or any of the QA/ranking tasks)
against a model whose cheapest provider is a third-party chat provider, and
have that provider return a 200 with an empty `choices` list or a choice
missing `message`/`content`. The operator raises and the run aborts, instead of
writing the raw body into the result column.
### Version/Branch
1.4.0-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]
