interrupt in loop

[vigneshmj1997]

Checked other resources

• This is a bug, not a usage question. For questions, please use GitHub Discussions.
• I added a clear and detailed title that summarizes the issue.
• I read what a minimal reproducible example is (https://stackoverflow.com/help/minimal-reproducible-example).
• I included a self-contained, minimal example that demonstrates the issue INCLUDING all the relevant imports. The code run AS IS to reproduce the issue.

Example Code

def ask_user(self, state):
        """Ask user question and get response from user"""
        temp = self._check_user_message(state) # Check for all Quality of LLM
        question = temp.get("message_to_user",None)
        if not question:
            raise Exception("question is None so code cannot proceed")
        response_from_user = interrupt(question)
        logger.info(f"Response from user {response_from_user}")
        return {
            "messages":[AIMessage(question),HumanMessage(response_from_user)]
        }

Error Message and Stack Trace (if applicable)

No response

Description

• This node goes in loop so the interrupt is triggered only first time and from second time it takes the values which has been already provided to it

System Info

langgraph = "^0.2.60"
langchain-community = "^0.3.13"
langchain = "^0.3.13"
langchain-core = "^0.3.28"
langchain-openai = "^0.2.14"
langgraph-checkpoint-postgres = "^2.0.9"
langgraph-checkpoint = "^2.0.9"
langchain-chroma = "^0.1.4"

[shengbo-ma]

Does it happen in a subgraph like #2870?

[Saisiva123]

This is the exact question that I've asked few days back:
#2927

In the documentation its mentioned that its the default behavior but there should be something that actually should interrupt the graph when invoked for the second time.

[vigneshmj1997]

@Saisiva123 Could you point me to the location in the documentation where it mentions that this is the default behavior?

I’ve identified an issue in the code at this location:
The problem is that the writes variable doesn’t retain the complete value of the entire subgraph when it’s called a second time.

I suspect that updating the value of __pregel_writes in RunnableConfig might solve the issue. Do you have any ideas on how to do this?

Additionally, I’m looking for a workaround to address this problem temporarily. If you know of any, please let me know

[vigneshmj1997]

Solved it have a look at this PR

its a single line change
Solution is instead of importing from the types.py
write your own function and use it

[vbarda]

Can someone provide a minimal reproducible example for the issue? https://stackoverflow.com/help/minimal-reproducible-example

We don't recommend using patches from the unofficial branches -- if there is an actual underlying issue we would want to solve it properly in the library