fix(core): override streaming callback if streaming attribute is set

[ccurme]

Currently the presence of a streaming callback handler enables streaming, even when a chat model has a streaming attribute that is explicitly set to False:

llm = MyChatModel(streaming=False)

await model.ainvoke(..., config={"callbacks": [_AstreamEventsCallbackHandler()]})  # streams

The streaming attribute is not on BaseChatModel but it is on most major chat models, including OpenAI, Anthropic, and Google GenAI. Its intended purpose is to enable streaming under the hood when calling .invoke.

OpenAI and Anthropic implement streaming as a boolean attribute with no sentinel value, so here we disable streaming if that attribute is explicitly set to False.

N.B. BaseChatModel does implement disable_streaming for the complementary case, and this currently works:

llm = MyChatModel(disable_streaming=True)

await model.ainvoke(..., config={"callbacks": [_AstreamEventsCallbackHandler()]})  # does not stream

but the behavior with streaming is unintuitive, so we update it here.

Ideally we would have a single bool | None = None attribute instead of these two booleans.

[codspeed-hq]

CodSpeed Performance Report

Merging #33351 will not alter performance

_{Comparing cc/streaming_callback (cdebe2d) with master (4c38157)}

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

Summary

✅ 13 untouched
⏩ 21 skipped¹

Footnotes

1. 21 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩