mreichhoff

for this one: > not all requests seem to go through the python server's handler I think there are a few things to look at: * the frontend is explicit...

> So it seems there is a bit of work needed on the core code to fully support base_path am I right? yes, that's correct, though I have this possibly...

nah, you're right: [service workers are https-only](https://developer.mozilla.org/en-US/docs/Web/API/Service_Worker_API/Using_Service_Workers#setting_up_to_play_with_service_workers), and I lean on those for offline support. Did you want to try writing in some `sed` in the dockerfile in this PR,...

thanks for the question! For signed in users with the AI permission (an allowlist model for now), when the user types in some Chinese or English text, or uploads a...

yes, both those make a lot of sense, as would making clearer when the user has made a flashcard from the sentence (currently it just shows up when clicking the...

thanks for the suggestion! It makes sense. Some questions: * would you imagine uploading a file? Or perhaps importing from Anki? Or I could set it up to use the...

seems like it should be a simple enough conversion, yeah...not sure when I'll get to it though, as I kinda like the numbers 😄

yep, agreed. The AI sentence generation does this, though the UI should be clearer, and I suspect AI would do well determining which definition of a word is present in...

ideally they'd exist in that file, yes, or be in the partitioned definitions data (I should probably do away with the partitioning for definitions and just load the whole dictionary...

got it, I suspect the issue is a mismatch of the characters between CEDICT and the components dataset I used: https://github.com/cjkvi/cjkvi-ids I will see about correcting this