Curious how it does on multi-page scanned PDFs vs. single screenshots? The ORT vision/decoder split is the part that usually makes or breaks CPU VLM OCR...
Seams we had a similar urge recently. I like the simplicity of your approach. Will post a show a little later about my approach to notetaking and LLM integration.
Similar for me, just no perplexity and I try to utilize more of Qwen for my coding, instead of cursor/claude, but i see my lazyness. This brings the advantage of having to go back to the basics aswell e.g. Search and Replace instead of just a brainless prompt...