[Python] How to implement and execute human-in-the-loop with LangGraph agent for document...

I am building an agent using Langgraph, and I have a node that fetches documents from a vector store. I want to add a human-in-the-loop functionality. Specifically, if no relevant documents are found in the vector store, I want the agent to ask for approval to use another node that leverages a Large Language Model (LLM) with general knowledge to provide a response.

Can I implement this behavior using interrupts in Langgraph, or is there another recommended way to achieve this kind of flow with Langgraph's capabilities? Also, I would like to know how to properly invoke this graph with the human-in-the-loop logic in place.

Continue reading...