docs/how_to/qa_chat_history_how_to/

[giscus[bot]]

docs/how_to/qa_chat_history_how_to/

This guide previously used the RunnableWithMessageHistory abstraction. You can access this version of the documentation in the v0.2 docs.

https://python.langchain.com/docs/how_to/qa_chat_history_how_to/

[kimbstocker]

Hello, is this guide outdated? I cant seem to get it to work, noted the following import should be below

from langchain.chains.history_aware_retriever import create_history_aware_retriever
from langchain.chains.retrieval import create_retrieval_chain

[AndresO7]

Hello, how are you? I don’t know if you can help me, since I have already completed the entire RAG and I also manage memory with LangGraph. The problem is that I don't know how to stream the response, as that is not covered in this guide.

[dawnsen1150]

check out this page https://python.langchain.com/docs/tutorials/rag/

it shows how to stream the answer from the RAG chain

from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough

def format_docs(docs):
return "\n\n".join(doc.page_content for doc in docs)

rag_chain = (
{"context": retriever | format_docs, "question": RunnablePassthrough()}
| prompt
| llm
| StrOutputParser()
)

for chunk in rag_chain.stream("What is Task Decomposition?"):
print(chunk, end="", flush=True)

[AndresO7]

Thanks for the help, but if I understood that, the problem is that when trying to stream the data it is different when using langgraph, as it is in this documentation and I did it this way:

for event in app.stream(
{"input": "hello how are you?"},
config=config15,
stream_mode="messages",
):
if isinstance(event[0], AIMessageChunk):
print(event[0].content,end='',flush=True)
but the problem is that I get two responses, one from this part:

Contextualize question

contextualize_q_system_prompt = (
"Given a chat history and the latest user question "
"which might reference context in the chat history, "
"formulate a standalone question which can be understood "
"without the chat history. Do NOT answer the question, "
"just reformulate it if needed and otherwise return it as is."
)
contextualize_q_prompt = ChatPromptTemplate.from_messages(
[
("system", contextualize_q_system_prompt),
MessagesPlaceholder("chat_history"),
("human", "{input}"),
]
)
but I only want to get the final answer, the real answer

[epv44]

How does the create_history_aware_retriever handle long conversation history? Is there some sort of trimming that takes place by default? Is it possible to implement custom trimming of the message history?

[henrypmartin]

Hi All,

I am getting below error when I call app.invoke(). The error is thrown from
response = rag_chain.invoke(state) from call_model

pydantic.errors.PydanticSchemaGenerationError: Unable to generate pydantic-core schema for typing.Union[str, list[str]]. Set arbitrary_types_allowed=True in the model_config to ignore this error or implement __get_pydantic_core_schema__ on your type to fully support it.

[ryanma9629]

for message in reversed(state["messages"]): if message.type == "tool": recent_tool_messages.append(message) else: break
I think this should be 'continue' instead of 'break', although it makes no difference in this demo.