Is there a way to get same message format in python and typescript?

OSS Product Help LangChain
1

As you can see in the code below, from the same piece of code in python and typescript the results are different, I need that both responses look the same.
I know I could convert manually the messages, but I wonder if there is already something in the langchain ecosystem that does this.



import {initChatModel} from "langchain/chat_models/universal";
const model = await initChatModel("ollama:granite4:micro", {
  temperature: 0.25
});

const response = await model.invoke("What is the capital of France?");
console.log(response.toJSON());

{
  lc: 1,
  type: 'constructor',
  id: [ 'langchain_core', 'messages', 'AIMessage' ],
  kwargs: {
    id: undefined,
    content: "The capital of France is Paris. It's known for its historical landmarks such as the Eiffel Tower, Louvre Museum, and Notre-Dame Cathedral.",
    tool_calls: [],
    response_metadata: {
      model: 'granite4:micro',
      created_at: '2025-11-10T10:35:50.78034Z',
      done: true,
      done_reason: 'stop',
      total_duration: 666412750,
      load_duration: 61871500,
      prompt_eval_count: 37,
      prompt_eval_duration: 62975917,
      eval_count: 33,
      eval_duration: 408072999
    },
    usage_metadata: {
      input_tokens: 37,
      output_tokens: 33,
      total_tokens: 70,
      input_token_details: {},
      output_token_details: {}
    },
    invalid_tool_calls: [],
    additional_kwargs: {}
  }
}

Python:

from langchain.chat_models import init_chat_model
model = init_chat_model("ollama:granite4:micro", temperature=0.25)

response = model.invoke("what is the capital of France?")

print("Response:", response.model_dump())

{
    "content": "The capital of France is Paris. It's located in the northern part of the country and serves as its political, cultural, and economic center.",
    "additional_kwargs": {},
    "response_metadata": {
        "model": "granite4:micro",
        "created_at": "2025-11-10T10:34:26.124632Z",
        "done": True,
        "done_reason": "stop",
        "total_duration": 608179417,
        "load_duration": 58786959,
        "prompt_eval_count": 37,
        "prompt_eval_duration": 58739791,
        "eval_count": 30,
        "eval_duration": 368132503,
        "model_name": "granite4:micro",
        "model_provider": "ollama",
    },
    "type": "ai",
    "name": None,
    "id": "lc_run--3e00c734-d3d2-4bde-b0dd-9d493c2fe50c-0",
    "tool_calls": [],
    "invalid_tool_calls": [],
    "usage_metadata": {"input_tokens": 37, "output_tokens": 30, "total_tokens": 67},
}
2

Hey! We don’t have an exported utility like .model_dump() in the JS version today, but we do use something like this to get a raw data structure for a message:

function modelDump(message: BaseMessage) {
  const { type, data } = message.toDict();
  return { ...data, type };
}

console.log(modelDump(message));
/**
 * {
 *.  type: 'human'
 *   content: 'Hello, how are you?',
 *   response_metadata: {},
 * }
 */