Obviously the step 1 was the most difficult. It takes a lot of trial and error to learn how to prompt LLM correctly. What helped is to provide code snippets and let AI to build user interface and put it all together.
I have quite a few extensions in mind that I’m going to develop the similar way. For example there is no reason Lovable.dev or similar platform wouldn’t be able to produce simple point of sale interface built on top of Manager as an extension. In fact, quite a few topics in ideas category can be now solved using extensions.
We need an extension developer AI model same as Theme Enhancer. The issue currently we have is that Chatgpt and other AI’s doesn’t have information about Manager extensions and API2. So, we have to feed that information first. It becomes too much hassle, and it makes too many mistakes.
So if that model is already trained on Api2 and extensions then it will be much easy to develop extensions even with no programming skills.
The extension was OK, it was issue with Manager itself when invoice had amount is words feature enabled. That’s why it worked for me because I didn’t test it with that option enabled.
It can’t be done automatically. Perhaps there could an extension which could do it in batch but in the end, it seems like QR code generation will be always extra step.
You first create the invoice
Review it’s correct
Then make it an e-invoice (this is because at this point it’s reported to tax authority)
