Jacob Lee

Thanks! Looks great overall - I think it'd be a bit more user-friendly to just initialize the client within the message store class rather than having the user import directly...

Hey I don't quite understand this one - it just echoes back what the agent calls the tool with?

Seems like there's some other components in the Python version to make it functional: https://github.com/hwchase17/langchain/blob/master/langchain/tools/human/tool.py Could you add them and add an integration test?

Hey @hiroshinishio and @haouarihk, we just pushed a refactor of the recursive splitter - going to close for now but can you please reopen if you see this again?

Thank you!

Maybe `StructuredInputParameterParser` is a clearer name?

Updated! Found that `gpt-3.5-turbo` was a little bit worse at picking up the "description" with this new JSON Schema format - kept getting silly responses like: `Based on the input...

Bump as well!

Will check this out this weekend! Thank you for the PR :)

This looks really cool but yeah I think moving out of the worker/blocking on things like embeddings wouldn't be worth it yet - I'm looking into raw WebLLM support first...