Implementation:Langchain ai Langchain BaseChatOpenAI Stream Implementation

Knowledge Sources	LangChain
Domains	Streaming, API_Integration
Last Updated	2026-02-11 00:00 GMT

Overview

Concrete tool for processing OpenAI streaming responses into LangChain chunks provided by the LangChain OpenAI integration.

Description

The BaseChatOpenAI._stream() method calls the OpenAI SDK with stream=True, iterates over the streaming response, and converts each OpenAI chunk into a ChatGenerationChunk containing an AIMessageChunk. It handles content deltas, tool call chunks, and optionally includes usage metadata (when stream_usage=True).

Usage

This is an internal method called by stream() and astream(). Users do not call it directly.

Code Reference

Source Location

Repository: langchain
File: libs/partners/openai/langchain_openai/chat_models/base.py
Lines: L1304-1377

Signature

def _stream(
    self,
    messages: list[BaseMessage],
    stop: list[str] | None = None,
    run_manager: CallbackManagerForLLMRun | None = None,
    *,
    stream_usage: bool | None = None,
    **kwargs: Any,
) -> Iterator[ChatGenerationChunk]:

Import

# Internal method — accessed via ChatOpenAI instance
from langchain_openai import ChatOpenAI

I/O Contract

Inputs

Name	Type	Required	Description
messages	list[BaseMessage]	Yes	Messages to send
stop	list[str] or None	No	Stop sequences
stream_usage	bool or None	No	Include token usage in chunks

Outputs

Name	Type	Description
return	Iterator[ChatGenerationChunk]	Chunks with AIMessageChunk containing incremental content and tool_call_chunks

Usage Examples

Observing Stream Chunks

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model="gpt-4o-mini", stream_usage=True)

for chunk in llm.stream("Hello!"):
    # Each chunk has incremental content
    if chunk.content:
        print(chunk.content, end="")
    # Last chunk may have usage_metadata
    if chunk.usage_metadata:
        print(f"\nTokens: {chunk.usage_metadata}")

Related Pages

Implements Principle

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment