Excited to finally tackle a refactor of my gmfarcaster transcript creation pipeline When I first built @warpee.eth I created a workflow that takes a YouTube URL as input, asks user (me) to add the names of the hosts/guests, the series name, and a unique episode identifier, and it then proceeds to: - generate the transcript - create embeddings for the transcript - updates the global directory of all episodes - push all the data to a place where it can be accessed by warpee and web apps The problems with this approach: - I have to run it locally and monitor it for errors - because of this I usually only do it once every 1-2 weeks, so Warpee doesn’t usually have the most recent transcripts - it’s also a batch operation for new episodes only so I can’t easily go back and update or fix transcripts for older episodes (ie for spelling errors or missing embeddings) I’ve been wanting to move this to a hosted server somewhere so I can run it from anywhere at anytime, but I’ve been procrastinating it bc it requires a refactor to be compatible with queuing. With Claude code all of a sudden this seems so much easier.