Great work. I hope we will see a lot of preservation work done now that AI can help with that. I did something very similar recently: https://github.com/VorticonCmdr/pushover
HTML-in-canvas is a new original-trial feature of Chrome I wanted to experiment with. You don't need Chrome Canary. You can enable the feature in "normal" Chrome (148 to 154 see here https://developer.chrome.com/origintrials/#/view_trial/34784...) as well just go to chrome://flags/#canvas-draw-element and change it to "enabled"
Great work. Just Yesterday I thought about LLMzip and asked myself if this is something which could vastly improve HTML compression when done at Google scale and shipped with browsers. I haven't done any research though.
I mean neural compressors provide great compression, BUTT the issue is they are really slow like in my project it takes around 45 minutes for de compression of 100 mb so I doubt if it would be useful, also using a transformer in user's browser sounds like a heavy task.
I don't use a extra full-text search for my browser history. Why? Because usually searching the headlines and/or urls is enough.
In order to fulltext-search all the webpages one has to crawl and index all the pages first. While crawling might be easy (since you visit the pages anyways), indexing and then searching is a bit more tricky.
One idea for a useful homebrew plugin might be to make use of Google Custom Search. Since Google is quite good at crawling, indexing and searching this might be something to look into.
I am interested if you come up with something useable.
