InjectedState to tools expect the optional attributes of the state to be invoked.

[DevinduSamarasinghe]

Hi I am currently facing the above mentioned issue where I have a tool that, if no intervention is needed, can be invoked.
And my interruptor is as follows,

def rag_interruptor(state: RagState, config: RunnableConfig) -> Command[Literal["tools", "rag_agent"]]:
    """
    A node that checks if the tool should be interrupted based on the user's feedbback. 

    Args:
        state (RagState): The current state of the graph containing the user's feedback.
        config (RunnableConfig): Configuration for the runnable.

    Returns:
        RagState: The updated state of the graph with the tool interruption flag set based on the user's feedback.
    """
    last_message = state["messages"][-1]
    human_messages = [msg for msg in state["messages"] if hasattr(msg, 'type') and msg.type == 'human']
    last_human_message = human_messages[-1]
    last_tool_call = last_message.tool_calls[-1]
    
    print("ENTIRE STATE:", state)
    human_review = interrupt(
        {
            "question": "Are the details correct?",
            "Request": last_human_message.content,
        })

    action = human_review.get("action")
    feedback = human_review.get("feedback")
    
    print("human review:", human_review)
    print("action:", action)
    #conditions to check if the user wants to append, replace, keep or ignore the tool call entirely. 


    if action =="append":
        
        update = {
            "messages": {
                "role": "human",
                "content": last_human_message.content + "\n\n" + feedback,
                "id": last_human_message.id,
                "tool_calls": [
                    {
                        "id": last_tool_call["id"],
                        "name": last_tool_call["name"],
                        "args": {}
                    }
                ]
            },
            "interrupt_method": action,
            "human_feedback": {
                "query": feedback,
                "message_id": last_human_message.id
            }
        }
        
        return Command(
            goto="tools",
            update=update
        )
        
    elif action == "replace": 
        update = {
            "messages": [
                {
                    "role": "human",
                    "content": feedback,
                    "tool_calls": [
                        {
                            "id": last_tool_call["id"],
                            "name": last_tool_call["name"],
                            "args": {},
                        }
                    ],
                    "id": last_human_message.id,
                }
            ],
            "interrupt_method": action,
            "human_feedback": None
        }

        return Command(
            goto="tools",
            update=update
        )
        
    elif action == "keep":
        return Command(
            goto="tools"
        )

    elif action == "ignore":
        return Command(
            goto="rag_agent" 
        )
    
    else: 
        raise ValueError("Invalid action specified in human review.")

Now the problem is that I am using a tool with injectedState instead of arguments because it takes the entirety of the context.

@tool(description="Search the vector store for relevant documents. You may use the entirety of the query provided by the user. ")
def retrieve(state: Annotated[RagState,InjectedState], config: RunnableConfig) -> List[str]: 
    """
    Search the vector store for relevant documents based on the query. 
    
    Args:
        state (InjectedState): The current state of the graph.
        config (RunnableConfig): Configuration for the runnable.
    Returns:
        List[str]: A list of documents that match the query.    
    """
    human_messages = [msg for msg in state["messages"] if hasattr(msg, 'type') and msg.type == 'human']
    human_feedback = state.get("human_feedback", None)
    
    if not human_messages:
        return "No user query found."
    
    message = human_messages[-1].content
    
    if human_feedback:
        query = human_feedback.get("query", None)
        prompt = (
            f"{message}"
            f"in addition, {query}"
        )
    else: 
        prompt = message
        
    retrieved_docs = rag_store.similarity_search(prompt, k=2)

    #serialize all the documents into a string format
    serialized = "\n\n".join(
        (f"Source: {doc.metadata}\n" f"Content: {doc.page_content}") for doc in retrieved_docs
    )

    return serialized 

Now the issue is that, in both the options of replace and append, it worked perfectly as intended. But in the "keep" option, validation errors are coming from the tool, saying two attributes are missing. But those attributes are already Optional.

class RagState(MessagesState):
    tool_interruption: Optional[bool] = Field(
        default=True,
        description="Flag to indicate if the tool should be interrupted."
    )
    interrupt_method: Optional[Literal["replace", "append", "keep", "ignore"]] = Field(
        default=None,
        description="The additional prompt to see if the interrupt should replace, append or keep the current message."
    )

    human_feedback: Optional[dict[str, str]] = Field(
        default=None,
        description="Feedback from the user after tool execution and also it holds the feedback for the corresponding message."
    )

I don't want to delve into another update to provide updates as such. and the tool doesn't necessarily need those attributes as well if there is no update to be made via an interrupt. Any solutions to this?