I think that AGPLv3 is a good license.
BTW, I have a suggest about a feature. Indeed I'm a bit afraid to store all my knowledge in a database. What about a possibly corruption of the sqlite file ?
So, it should be great to keep all markdown files and use sqlite to find an item. dnote could sync the sqlite file and markdown files from a directory ?
sqlite would not store the file, but just the name/shasum/whatever usefull from the markdown.
what do you think about ?

You mean storing data as files, but maintaining indices in SQLite for searching, etc? Sounds like we can get best of both worlds--files and SQLite. A trade-off is keeping them in sync would add complexity. Maybe we can have a tool that translates SQLite notes into files. What is your thought?

AGPL is definitely solid but I wonder if a more permissive license might help adoption and long-term success. Apache or MIT, probably.

A trade-off is keeping them in sync would add complexity. Maybe we can have a tool that translates SQLite notes into files. What is your thought?

Sure, but if the database file is corrupted I suppose the user would not be able to export his datas to files ?
I suggest you this process:
when the user write a markdown file with any editor, the tmp file is stored in sqlite and in a data directory. Same process when he update any note.
On dnote sync, the server update his database and export all notes in files.
So in case of a corrupted database, the server or the client can rebuilt it with an import from the data directory.

AGPL is definitely solid but I wonder if a more permissive license might help adoption and long-term success. Apache or MIT, probably.

Depend what you want. With MIT I can use your code in a propritary software.

Thanks for the interesting idea. Let's use it as data point for the future direction.