Closed justinlevi closed 1 month ago
I would really love to see a chatbot example where you stream to a frontend implementation that accepts user input for an interrupt. This piece is really throwing me off.
I think this thread is related https://github.com/langchain-ai/langgraph/discussions/544
User Story: As a developer building a simple chatbot, I may want to interrupt a graph to get user approval before executing a specific tool. (ex. changing airline booking) I need an AI message sent to the user with the details of the proposed change and need to record this plus the users response in as a checkpoint before continuing on with the tool execution or not. I would expect the messages state to be managed by langgraph in this situation.
@justinlevi this should be fixed if you update to the latest version of langchain-anthropic
. Please update and let me know if you're still running into issues!
Yeah, this wasn’t related to anthropic as I was using OpenAI
On Tue, Oct 15, 2024 at 5:58 PM Vadym Barda @.***> wrote:
@justinlevi https://github.com/justinlevi this should be fixed if you update to the latest version of langchain-anthropic. Please update and let me know if you're still running into issues!
— Reply to this email directly, view it on GitHub https://github.com/langchain-ai/langgraph/issues/2075#issuecomment-2415237276, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAIQ6JJBYJGTXIR77UBBGM3Z3WFX3AVCNFSM6AAAAABPXEQZEWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDIMJVGIZTOMRXGY . You are receiving this because you were mentioned.Message ID: @.***>
Issue with current documentation:
https://langchain-ai.github.io/langgraph/tutorials/customer-support/customer-support/#utility
Running the notebook throws the following error:
Output before error
My python env:
Idea or request for content:
No response