I ran out of shakespeare.diy credits in the middle of it working and tried to top it up before it noticed, now this is happening on clicking "play" (credits are positive again).
I'm hesitant to refresh if it'll lose its current chunk of work and context. what's the best way to handle this, nostr:npub18ams6ewn5aj2n3wt2qawzglx9mr4nzksxhvrdc4gzrecw7n5tvjqctp424 ?
ah okay, i opened in a new tab and was able to "resume"!
checking.
I spent $57 and then "this project has encountered an error, please restart". i restarted and chat history is wiped :)
the project is still there, right? just the context is missing?
we believe your previous bug is fixed. let me know if you see the issue again.
as for this one, checkign on it.
additionally, having chat history is sometimes actually a really good thing. when contexts get too large, things start getting expensive. starting from scratch is a good idea and allows things to work better. i do this often when using mkstack from the command line.
project still there, yep! and I started up the chat again and it seems to be still as competent as previously ...I think... me and it are trying to figure out some stuff with zaps and replaceable events....
how did you end up making out? did you deploy anything for me to play with? :)
I only had about an hour in the morning and didn't finish it. I think I need to open an Ngrok tunnel to my local catallax-enabled relay and get MKStack to point the client at it. a project for this weekend!
zaps still seemed rocky when I called it quits earlier but I hope to have it all sorted next week.
thanks for checking in!
i can confirm that zaps are a pain in the butt. i have an open issue to make them better and more consistent. you're probably having one of several issues that seem to always reoccur. soon we'll get a template setup for zaps so this dosn't happen.
here are some issues that i've personally seen.
AI struggles to figure out how to properly implement zaps, struggling with kind:9734 (sending those to relays instead of the LNURL provider), publishing the kind:9735 zap receipt properly, the Lightning address endpoint request formatting, Zap amounts in milisats, Zap buttons using and detecting WebLN on desktop (Alby) and still sending attempting to execute the Lightning invoice even though WebLN is used. NWC struggles as well, trying to maintain the connection and re-establish the connection.
yep, all these and a couple more; real buzzkill after getting so far 🫤
I'll probably wait for those zap templates before dumping another pile of money in the hole. thanks for letting me know! I look forward to it
