Exception: run not found when using Langfuse tracing on Langserve chains, even though the tracing works

[kubni]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

I have a python langchain langserve app.
Here is how i register a chain as an endpoint handler in server.py, with Langfuse's CallbackHandler being set as a callback.

add_routes(
    app,
    my_chain().with_config(RunnableConfig(callbacks=[CallbackHandler()])),
    path="/myPath",
    input_type=MyChainRequestSchema,
)

Here is the my_chain.py:
```python
def my_chain():
    prompt = PromptTemplate.from_template(
        """Some prompt"""
    )

    chain = {
        "prompt": prompt,
        "model": itemgetter("model"),
        "max_tokens": itemgetter("max_tokens"),
    } | RunnableLambda(llm_model)

    return chain

Here is the llm_model and get_llm in llms.py:

def make_llm_structured(
    llm: ChatOpenAI, response_schema: Type[BaseModel], include_raw: bool
):
    return llm.with_structured_output(response_schema, include_raw=include_raw)


def llm_model(
    input: dict[str, Any],
    config: dict[str, Any],
    response_schema: Optional[Type[BaseModel]] = None,
    include_raw: bool = False,
):
    filtered_input = {
        k: v for k, v in input.items() if k in VALID_ARGUMENTS_FOR_GET_LLM
    }
    print(f"\nFiltered input: {filtered_input}\n")
    base_llm = get_llm(**filtered_input, endpoint=endpoint)

    if response_schema:
        chain = RunnableLambda(itemgetter("prompt")) | make_llm_structured(
            llm=base_llm, response_schema=response_schema, include_raw=include_raw
        )
        return chain
    else:
        chain = RunnableLambda(itemgetter("prompt")) | base_llm

        return chain


def get_llm(*, model, endpoint: str, temperature=0.0, top_p=1.0, max_tokens=4096):
    print(f"\n[get_llm] Endpoint: {endpoint}...\n")
    if model in openai_models:
        config = openai_models[model]

        if model in ["gpt-o1", "gpt-o1-mini"]:
            llm = ChatOpenAI(
                model=config["model_name"],
                temperature=1,
            )

        else:
            llm = ChatOpenAI(
                model=config["model_name"],
                temperature=temperature,
                top_p=top_p,
            )
       return llm

This gives me proper langfuse tracing in the langfuse dashboard, but what also happens is that I get A LOT of errors originating from langchain.py, specifically its on_chain_start and on_chain_end callbacks. Like this:

Traceback (most recent call last):
  File "....../.venv/lib/python3.12/site-packages/langfuse/callback/langchain.py", line 273, in on_chain_start
    self.runs[run_id] = self.runs[parent_run_id].span(**content)
                        ~~~~~~~~~^^^^^^^^^^^^^^^
KeyError: UUID('402ae73d-c6ab-4a0d-a664-cb41a40712a5')

Traceback (most recent call last):
  File "....../.venv/lib/python3.12/site-packages/langfuse/callback/langchain.py", line 273, in on_chain_start
    self.runs[run_id] = self.runs[parent_run_id].span(**content)
                        ~~~~~~~~~^^^^^^^^^^^^^^^
KeyError: UUID('4c849e11-045c-4a9b-9c98-76299d9feace')

Traceback (most recent call last):
  File "....../.venv/lib/python3.12/site-packages/langfuse/callback/langchain.py", line 459, in on_chain_end
    raise Exception("run not found")
Exception: run not found

Traceback (most recent call last):
  File "....../.venv/lib/python3.12/site-packages/langfuse/callback/langchain.py", line 459, in on_chain_end
    raise Exception("run not found")
Exception: run not found

Traceback (most recent call last):
  File "....../.venv/lib/python3.12/site-packages/langfuse/callback/langchain.py", line 459, in on_chain_end
    raise Exception("run not found")
Exception: run not found

I thought that this happens because of an error with callback propagation, but I thought with_config is there specifically so I don't have to do it manually.

### Description

I am trying to implement Langfuse tracing for my Langcserve http handler chains.
Additional info is above the code blocks that its related to, above in the Example Code section.

### System Info

langchain==0.3.14
langchain-cli==0.0.31
langchain-community==0.3.14
langchain-core==0.3.29
langchain-google-genai==2.0.7
langchain-google-vertexai==2.0.9
langchain-openai==0.2.0
langchain-text-splitters==0.3.5

langfuse==2.57.12