"model_name" missing in "response_metadata" in JavaScript SDK but is present in Python SDK

[rors41]

Checked other resources

• I added a very descriptive title to this issue.
• I searched the LangChain.js documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.
• I am sure that this is a bug in LangChain.js rather than my code.
• The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).

Example Code

Python example code:

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model="gpt-4o")
print(llm.invoke("hello world").model_dump_json())

Javascript example code:

import { ChatOpenAI } from "@langchain/openai"

(async () => {
	const model = new ChatOpenAI({ model: "gpt-4o", apiKey: process.env.OPENAI_API_KEY });
	const response = await model.invoke("hello world");
	console.log(response)
})()

Error Message and Stack Trace (if applicable)

No response

Description

Hey, I've noticed that the response from simple LLM OpenAI call in Python Langchain is different from Javascript Langchain.
We are using LiteLLM proxy and model_name field is helpful because it is the real name of the model that was used.

Is there a reason why this field is not present in the JS response?

Python Langchain:

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model="azure.gpt-4o") # litellm proxy model name
print(llm.invoke("hello world").model_dump_json())

Response:

{
  "content": "Hello! How can I assist you today?",
  "additional_kwargs": {
    "refusal": null
  },
  "response_metadata": {
    "token_usage": {
      "completion_tokens": 9,
      "prompt_tokens": 9,
      "total_tokens": 18,
      "completion_tokens_details": null,
      "prompt_tokens_details": null
    },
    "model_name": "gpt-4o-2024-08-06", <- actual model name
    "system_fingerprint": "fp_",
    "finish_reason": "stop",
    "logprobs": null
  },
  "type": "ai",
  "name": null,
  "id": "run-",
  "example": false,
  "tool_calls": [],
  "invalid_tool_calls": [],
  "usage_metadata": {
    "input_tokens": 9,
    "output_tokens": 9,
    "total_tokens": 18,
    "input_token_details": {},
    "output_token_details": {}
  }
}

Javascript Langchain:

import { ChatOpenAI } from "@langchain/openai"

const model = new ChatOpenAI({ model: "azure.gpt-4o", apiKey: process.env.OPENAI_API_KEY, configuration: { baseURL: process.env.OPENAI_API_BASE }}); // litellm proxy model name
const response = await model.invoke("hello world");
console.log(response)

Response:

{
  "id": "chatcmpl-id",
  "content": "Hello! How can I assist you today?",
  "additional_kwargs": {
    "function_call": null,
    "tool_calls": null
  },
  "response_metadata": {
    "tokenUsage": {
      "promptTokens": 9,
      "completionTokens": 9,
      "totalTokens": 18
    },
    "finish_reason": "stop",
    "usage": {
      "completion_tokens": 9,
      "prompt_tokens": 9,
      "total_tokens": 18,
      "completion_tokens_details": null,
      "prompt_tokens_details": null
    },
    "system_fingerprint": "fp_"
  },
  "tool_calls": [],
  "invalid_tool_calls": [],
  "usage_metadata": {
    "output_tokens": 9,
    "input_tokens": 9,
    "total_tokens": 18,
    "input_token_details": {},
    "output_token_details": {}
  }
}

The same behaviour can be observed in the examples from the official docs:

• JS - no model name in the response - https://js.langchain.com/docs/integrations/chat/openai/#invocation
• Python - model name present in the response - https://python.langchain.com/docs/integrations/chat/openai/#invocation

System Info

Node version: v20.18.0
Platform: macOS Sonoma 14.7.1
langchain: 0.3.6
langchain/openai: 0.3.14

[dosubot]

The absence of the model_name field in the JavaScript SDK's response, compared to the Python SDK, might be due to differences in how the LangChain.js library processes API responses from OpenAI. The JavaScript implementation does not explicitly exclude model_name during response processing, but it might not include it by default in the response metadata.

The _identifyingParams function in LangChain.js includes model_name as part of the identifying parameters for the model, which suggests that it is recognized within the library's configuration ^[1]. However, this does not guarantee its presence in the response metadata.

To address this discrepancy, you might need to manually include the model_name in the response by modifying the response handling logic or by ensuring that the configuration explicitly requests this field. If the model_name is crucial for your application, consider checking if there are configuration options or updates in the library that might affect its inclusion in the response.

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}

[anadi45]

Hey @rors41 I have made the change to include model_name in response_metadata.

[rors41]

Thanks, works as expected