I am coding a backend server in python using lancgain and I need to attach a callback to track each time the chain has been run using track_feedback(). I am using the “callbacks” argument on the with_config() in the self.runnable. But it is getting triggered many times per runnable execution. Could the children chains be somehow inheriting the call?

class BaseInputs(BaseModel):
    """Base inputs model."""
    app: str 
    run_id: str
    score: int
    tool: str
    asset: str

class GA4CallbackHandler(BaseCallbackHandler):
    def on_chain_start(self, serialized, inputs, **kwargs):
        if not self.has_run:
            print(track_feedback("123456.7654321", 'app', 'tool', 'asset', 'standard', "user", "0").status_code)
            print("feedback_sent")

class FeedbackChain(Chain):
    
    def __init__(self) -> None:                
        self.runnable = ( 
            {
                "app" : itemgetter("app"),
                "run_id" : itemgetter("run_id"),
                "score" : itemgetter("score"),
                "tool": itemgetter("tool"),
                "asset": itemgetter("asset")                
            } 

            | RunnableLambda(self.send_feedback)
        ).with_config(run_name="feedback",callbacks=[GA4CallbackHandler()]).with_types(input_type=BaseInputs)

The reults is many triggers of the status code and “feedback_sent” string on console for each feedback sent. I printed the “serialized” parameter and it seems to be executing for each input (so once for “app”, another for “run_id”…)
How can I make this execute only once at the start?

I tried using a flag on the handler class but then it only executes once and subsequent executions never run.
I also tried attaching the callback to the RunnableLambda inside the self.runnable chain, that worked but it looks dirty. I’d like it to be attached to the whole chain.