QtCS25 -Qt AI Inference API PoC

Session Summary

Draft discussion topics

Go through the arose questions
Demo the current "QuteAI" example app
Go through the example "QuteAI" application SW and the AI Inference API SW
Discuss how new Inference API could be used internally in Qt for implementing:
- Accessibility features (Screen Reader, Voice commands etc.)
- Automatic conversion of texts for Internalization
Discuss how the API could be improved
- adding of "enabled" to control when AI model can/cannot run to enable conditional execution
- adding of "DataFunnel" class to allow any kind of data sources and data processing elements
Discuss how to prototype a graphical editor for AI/data pipelines
- e.g. extend https://github.com/arjafari93/QBluePrint
Keep the "Pioneering Local AI on RISC-V: Building Industry-Ready AI PCs for the Next Computing Era" Lightning talk (by Yuning Liang & Danfeng Zhang)

Tommi Mänttäri

Tinja Paavoseppä

Questions got answered
QuteAI demo app was demonstrated
- How Ollama, Whisper, and diffusion model AI services with needed modes were started
- How the app is used for ASR, Text2Image, Image2Text, Text2Text and TTS
Did not have time to go the app and the API SW, nor any discussion about internal use of the new API, nor the coming changes to the API and the graphical editor prototyping
Kept the "Pioneering Local AI on RISC-V: Building Industry-Ready AI PCs for the Next Computing Era" Lightning talk (by Yuning Liang & Danfeng Zhang)

Links