Llama-server client in 22kB

[clort81]

https://github.com/clort81/bloppadd/blob/main/README.md

Not great, still old completion api. Wont properly parse thinking tags for GLM and some other models.
But it's small and actually fast, and you can actually type and hilight long texts without browser bogging down to unusable.
I'd be curious to see what others can make of it.