This is truly fantastic work Christian.
I really want to dig into this properly (haven’t had an opportunity this weekend). I will give it my attention over the next few days.
This is truly fantastic work Christian.
I really want to dig into this properly (haven’t had an opportunity this weekend). I will give it my attention over the next few days.
Excellent.
Current version is now in 25.5pr3 on the website:
For next versions the parameters for constructors will be optional and I add a simple Ask() function if you don’t like to do the loop yourself with decode.
Ask(Sampled as LlamaSamplerMBS, prompt as String, MaxTokens as Integer = 100) as String
Yes that would be much nicer to have that option.
I have this afternoon off work so going to dig into your examples.
I’m just getting started experimenting with this awesome library Christian.
Do I need to do anything special when my app quits to unload a model and free up RAM or does the plugin handle that for me?
sorry to heard that.. what happen to your right hand in may 2026??
Xojo uses reference counting.
So once the objects go out of scope, it should free itself.
So anyone made something cool using llama?
I’ve been on leave for about a week so haven’t had really any chance to dig into it yet.
It’s my next goto project though!
As the new beta is there, the focus has shifted to deliver macOS 26 design first… But it is next on the list.