I uploaded triton as part of the whisper requirements, but ran into problems with whisper licensing and behaviour (downloading models at run time without warning users), and have focused the effort on whipser.cpp the last few months to see if that is an easier way to get audio transcription to work out of the box in Debian. As the whisper.cpp package have not yet cleared NEW, it is not clear if this approach will work any better, so I kept triton in the archive for now in case I would have to go back to looking at whisper.
JFYI -- Happy hacking Petter Reinholdtsen
