Jacob Lee
Jacob Lee
Seems something still might be up with the types @nickscamara?
Bumping in #6488
Do you need to use `.stream()` for this or could use `.batch()` instead? Catching streaming errors is much more nuanced (e.g. what to do with partial responses).
> @jacoblee93 I think I see what's going on: when using streaming, completionWithRetry is used in ChatOpenAI._streamResponseChunks, which awaits (and should retry) the creation of the stream iterable. However, when...
I think for token counting logic we should absolutely not error the entire call on parsing failures, so yes that's definitely a bug.
Should have some mechanism here - will think through it a bit more.
Closing since it looks incomplete and abandoned but please reopen if not
Thanks for flagging - will try to take a look soon.
Yeah only certain providers support tool calling. It's a model-specific feature so I'm not sure there's much we can do about it as I don't have access to Yandex models....