TypeError when using add_routes() with a RunnableWithMessageHistory wrapper

[rfoucart]

Hi all,

I don't know if I should post here or in the langchain-core issues (and I might do both).

I've got this error since I updated all the langchain-related packages.
It seems that the add_routes() method no longer works with a RunnableWithMessageHistory.
Here is the sample code :

import os

from fastapi import APIRouter
from langchain_anthropic.chat_models import ChatAnthropic
from langchain_core.chat_history import BaseChatMessageHistory, InMemoryChatMessageHistory
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain_core.runnables import RunnableWithMessageHistory
from langserve import add_routes

router = APIRouter(prefix="/llm", tags=["llm"])

store = {}


def get_session_history(session_id: str) -> BaseChatMessageHistory:
    if session_id not in store:
        store[session_id] = InMemoryChatMessageHistory()
    return store[session_id]


qa_system_prompt = """You are an assistant for question-answering on various topics.
If you don't know the answer, just say that you don't know and ask for more informations.
Answer in french.
"""
qa_prompt = ChatPromptTemplate.from_messages(
    [
        ("system", qa_system_prompt),
        MessagesPlaceholder("chat_history"),
        ("human", "{input}"),
    ]
)
llm = ChatAnthropic(model="claude-3-5-sonnet-20240620", temperature=0, api_key=os.getenv("ANTHROPIC_API_KEY"))
chain = qa_prompt | llm
conversational_chain = RunnableWithMessageHistory(
    chain,
    get_session_history,
    input_messages_key="input",
    history_messages_key="chat_history",
    output_messages_key="output",
)

if __name__ == "__main__":
    question = "What is the size of the Atlantic Ocean ?"
    response = conversational_chain.invoke(
        {"input": question},
        config={
            "configurable": {"session_id": "14"}
        },
    )
    print(f"Question: {question}\nAnswer: {response.content}")
    follow_up_question = "What is its deepest point ?"
    response = conversational_chain.invoke(
        {"input": follow_up_question},
        config={
            "configurable": {"session_id": "14"}
        },
    )
    print(f"Question: {follow_up_question}\nAnswer: {response.content}")
else:
    add_routes(app=router, runnable=conversational_chain, path="/chat")

If I run this script as main, it works perfectly as intented. If not, it throws this error :

Traceback (most recent call last):
  File "~/backend/app/llm_router.py", line 42, in <module>
    add_routes(app=router, runnable=conversational_chain, path="/chat")
  File "~/backend/.venv/lib/python3.12/site-packages/langserve/server.py", line 443, in add_routes
    api_handler = APIHandler(
                  ^^^^^^^^^^^
  File "~/backend/.venv/lib/python3.12/site-packages/langserve/api_handler.py", line 671, in __init__
    output_type_ = _resolve_model(
                   ^^^^^^^^^^^^^^^
  File "~/backend/.venv/lib/python3.12/site-packages/langserve/api_handler.py", line 338, in _resolve_model
    hash_ = model.schema_json()
            ^^^^^^^^^^^^^^^^^^^
  File "~/backend/.venv/lib/python3.12/site-packages/pydantic/v1/main.py", line 675, in schema_json
    cls.schema(by_alias=by_alias, ref_template=ref_template), default=pydantic_encoder, **dumps_kwargs
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "~/backend/.venv/lib/python3.12/site-packages/pydantic/v1/main.py", line 664, in schema
    s = model_schema(cls, by_alias=by_alias, ref_template=ref_template)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "~/backend/.venv/lib/python3.12/site-packages/pydantic/v1/schema.py", line 188, in model_schema
    m_schema, m_definitions, nested_models = model_process_schema(
                                             ^^^^^^^^^^^^^^^^^^^^^
  File "~/backend/.venv/lib/python3.12/site-packages/pydantic/v1/schema.py", line 581, in model_process_schema
    m_schema, m_definitions, nested_models = model_type_schema(
                                             ^^^^^^^^^^^^^^^^^^
  File "~/backend/.venv/lib/python3.12/site-packages/pydantic/v1/schema.py", line 622, in model_type_schema
    f_schema, f_definitions, f_nested_models = field_schema(
                                               ^^^^^^^^^^^^^
  File "~/backend/.venv/lib/python3.12/site-packages/pydantic/v1/schema.py", line 255, in field_schema
    f_schema, f_definitions, f_nested_models = field_type_schema(
                                               ^^^^^^^^^^^^^^^^^^
  File "~/backend/.venv/lib/python3.12/site-packages/pydantic/v1/schema.py", line 527, in field_type_schema
    f_schema, f_definitions, f_nested_models = field_singleton_schema(
                                               ^^^^^^^^^^^^^^^^^^^^^^^
  File "~/backend/.venv/lib/python3.12/site-packages/pydantic/v1/schema.py", line 926, in field_singleton_schema
    if issubclass(field_type, BaseModel):
       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "<frozen abc>", line 123, in __subclasscheck__
TypeError: issubclass() arg 1 must be a class

add_routes() fails with the following packages versions :

fastapi==0.112.3
langchain==0.2.16
langchain-anthropic==0.1.23
langchain-community==0.2.16
langchain-core==0.2.38
langserve==0.2.3
langsmith==0.1.114
pydantic==2.8.2
pydantic-core==2.20.1

but works fine with these versions :

fastapi==0.112.1
langchain==0.2.8
langchain-anthropic==0.1.20
langchain-community==0.2.7
langchain-core==0.2.19
langserve==0.2.2
langsmith==0.1.99
pydantic==2.8.2
pydantic_core==2.20.1

I put a breakpoint to see what field_type contains when the code breaks.
Here is the list of the successive field_type values I saw before the error :

• <class 'langchain_core.messages.base.BaseMessage'>
• <class 'langchain_core.messages.base.BaseMessage'>
• ForwardRef('Runnable')

I don't get where this bug comes from, any help would be appreciated.
Thanks !

[eyurtsev]

Should be fixed in the upcoming 0.3 release