codeonbush:fix/openai-reasoning-content-streaming - Branch - langchain-ai/langchain

fix(openai): support reasoning_content in streaming delta chunks

#34152

Comparing

codeonbush:fix/openai-reasoning-content-streaming

(

82e6e5c

) with

master

(

12df938

)

Untouched: 6

Skipped: 28

Benchmarks

Skipped (28)

Passed

test_init_time

libs/partners/openai/tests/unit_tests/chat_models/test_responses_standard.py::TestOpenAIResponses

12.3 ms*12.3 ms

test_stream_time

libs/partners/openai/tests/integration_tests/chat_models/test_base_standard.py::TestOpenAIStandard

1.2 s*1.2 s

test_init_time

libs/partners/openai/tests/unit_tests/chat_models/test_base_standard.py::TestOpenAIStandard

12.3 ms*12.2 ms

test_stream_time

libs/partners/openai/tests/integration_tests/chat_models/test_responses_standard.py::TestOpenAIStandard

1.2 s*1.2 s

test_stream_time

libs/partners/openai/tests/integration_tests/chat_models/test_responses_standard.py::TestOpenAIResponses

857.8 ms*857.3 ms

test_init_time

libs/partners/openai/tests/unit_tests/chat_models/test_azure_standard.py::TestOpenAIStandard

1.7 s*1.7 s

Commits

Click on a commit to change the comparison range

Base

master

12df938

+0.14%

openai: support reasoning_content in streaming delta chunks This change adds support for the reasoning_content field in streaming delta chunks from the Chat Completions API. This field is used by: - LiteLLM when proxying Gemini models with reasoning_effort enabled - DeepSeek reasoning models (deepseek-reasoner) - vLLM with reasoning parser enabled - Other OpenAI-compatible providers that support reasoning/thinking output The reasoning_content is stored in additional_kwargs, consistent with how the Responses API handles reasoning content. Fixes #29513, #31326, #32981, #30580

82e6e5c

4 days ago

Home Terms Privacy Docs