My first idea was that it could be an local application, as in win32/linux/macos binary with Qt. But then all those hoops to bypass limitations makes no sense - you make a config file and you are done, or you can discover other applications on same PC...
The other idea could be a an Android or iPhone app.. but then author started to talk about tailscale and I got confused - surely they are not running tailscale active _all the time_ on their phone?
So now I am maybe thinking author means a "web app", as in a page full of javascript. This explains a need for extreme hoop jumping - the browser sandbox is very strict. But then I can't understand why not do the discovery on backend, which has no restrictions?
this post has _a lot_ of background missing.
I’d say I have the opposite of a narrow view of software one can write, having written everything from typical web/mobile apps to DBs, network protocols and VMs :)
My initial explorations were focused on the web, but the network-level scheme I landed on does indeed do discovery on the backend.
Appreciate this feedback. I’ve been in the weeds of this stuff for some time, so it helps to see the rubber meet the road so I know what I missed.
Do you need document storage? Hardcode `sftp://192.168.1.18/data` (or `sftp://data-server.local/data` if you want to be fancy), it's just for you. Do you need to send a notification? Pretty sure you know what your favorite method it, call it directly. Maybe add a local proxy that adds your token to avoid copying token around. Need ollama access? Don't mess with discovery, hardcode again...
So what does discovery give you?
