Managing LangChainInstrumentor for Tracing in Phoenix Applications
Hi, I am really happy with the new feature to trace and download spans from a remotely running phoenix instance using px.Client. Once the tracing is initialised it keeps tracing all steps in the chains by using LangChainInstrumentor().instrument() - thanks again for the great work! If I am now done with a single run through my entire application, with every span traced as desired: Is there a way to stop the LangChainInstrumentor at the end of the script or to force a restart without shutting down the phoenix instance in the background? Scenarios:
Imagine you are running the application several times, and you are not so much interested in tracing the same things again, then it would be quite nice to start with an empty span_df again, no?
Likewise, if there are calls that should be excluded from the tracing, it would be quite nice to interrupt the tracing and start it at a later stage in the script again.
Is there already anything in place that I missed while reading the docs? Thank you! 🙏
