Implementation:Langchain ai Langchain BaseChatOpenAI Create Chat Result

Knowledge Sources	LangChain
Domains	NLP, Data_Transformation
Last Updated	2026-02-11 00:00 GMT

Overview

Concrete tool for parsing OpenAI API responses into LangChain ChatResult objects provided by the LangChain OpenAI integration.

Description

The BaseChatOpenAI._create_chat_result() method takes a raw OpenAI API response (dict or openai.BaseModel) and constructs a ChatResult containing ChatGeneration objects. Each generation wraps an AIMessage with parsed content, tool calls, usage metadata (UsageMetadata with InputTokenDetails and OutputTokenDetails), and response metadata.

Usage

This is an internal method called by _generate() after receiving the raw API response. It handles both the Chat Completions API and the newer Responses API format.

Code Reference

Source Location

• Repository: langchain
• File: libs/partners/openai/langchain_openai/chat_models/base.py
• Lines: L1480-1596

Signature

def _create_chat_result(
    self,
    response: dict | openai.BaseModel,
    generation_info: dict | None = None,
) -> ChatResult:

Import

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

I/O Contract

Inputs

Name	Type	Required	Description
response	dict or openai.BaseModel	Yes	Raw API response from OpenAI
generation_info	dict or None	No	Additional generation metadata

Outputs

Name	Type	Description
return	ChatResult	Parsed result with ChatGeneration objects containing AIMessage, usage_metadata, and response_metadata

Usage Examples

Accessing Parsed Response Data

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model="gpt-4o-mini")
result = llm.invoke("Hello!")

# Content
print(result.content)  # "Hello! How can I help you?"

# Usage metadata (parsed by _create_chat_result)
print(result.usage_metadata)
# {'input_tokens': 8, 'output_tokens': 12, 'total_tokens': 20,
#  'input_token_details': {'cached': 0},
#  'output_token_details': {'reasoning': 0}}

# Response metadata
print(result.response_metadata)
# {'model_name': 'gpt-4o-mini', 'system_fingerprint': 'fp_...', 'finish_reason': 'stop'}

Related Pages

Implements Principle

• Principle:Langchain_ai_Langchain_Response_Parsing
• Heuristic:Langchain_ai_Langchain_Error_Context_Preservation