langchain Agents

[zlapp]

Thanks for the great repository.
Does langcorn support langchain agents?

[nb-programmer]

I tried to use an Agent. Initially I got an error due to input_key missing. But I then modified the api file with input_keys. It worked, but I don't know how memory is handled, because the API has a memory input as well

[msoedov]

Thx for the feedback @zlapp ! I just added and tested an example agent https://github.com/msoedov/langcorn/blob/main/examples/ex7_agent.py. It works

[knitish619]

I tried to use an Agent. Initially I got an error due to input_key missing. But I then modified the api file with input_keys. It worked, but I don't know how memory is handled, because the API has a memory input as well

what modifications have you made in api file?

[nb-programmer]

I tried to use an Agent. Initially I got an error due to input_key missing. But I then modified the api file with input_keys. It worked, but I don't know how memory is handled, because the API has a memory input as well

what modifications have you made in api file?

It was a while ago and just a quick workaround, but this is what I had changed (in api.py of the package).

[knitish619]

how did you handle memory ?

[msoedov]

You can store memory on the client side and pass it every time with the request. The updated memory state will be returned from the response.

[nb-programmer]

Just as an FYI, it is possible to add chat history memory in the chain that will store in a DB itself, without passing the burden to client
https://python.langchain.com/docs/modules/memory/agent_with_memory_in_db

You need to pass the session id to the xyzChatMessageHistory class instantiation.

message_history = RedisChatMessageHistory(
    url="redis://localhost:6379/0", ttl=600, session_id="my-session"
)

memory = ConversationBufferMemory(
    memory_key="chat_history", chat_memory=message_history
)