Options to pass through frames (transcription, llm context, etc...)

[lucasrothman]

Could there be parameters added to some frame processors to pass through frames instead of capturing them? For things like TranscriptionFrames and OpenAILLMContext frames it would sometimes be useful to have them available to upstream processors. Is there a specific reason or advantage to having frames captured instead of just handled and passed through?

I know context aggregators would have to be adjusted, but ultimately my use case is to get the last user message from context/transcription but I have no good way to tell when that will be updated. The UserStoppedSpeakingFrame comes before the context is updated, and constantly checking the difference between context stored as a variable in my frame processor and the context aggregator seems like a hacky way of doing it.