Look into what the options are for configuring entrypoints
mwalbeck opened this issue · comments
Hi @Kixunil I had a closer look and I can include Talked as an executable quite easily in the pyproject file. Though that would be for running the development server, but I think it's good to include it as it can be useful for testing. The standard way to run Talked using uwsgi is to call uwsgi and telling it to run Talked, so more or less the way I currently do it in my example systemd unit file.
You can then either configure uwsgi using a config file or commandline arguments. Looking more into it, I'll also be removing uwsgi as a dependency in the pyproject file, as it doesn't make sense to have it there. uwsgi is more of an external and optional dependency that is recommended for production use.
The proper place to specify uwsgi as a dependency is in the debian package and then use uwsgi to run Talked whenever you start it using systemd.
I'll also just add a quick note on the hosting Talked on a subroot. It's not tested but should without a doubt work if you put Nginx or Apache in front of uwsgi, uwsgi is more of an application server that happens to also support HTTP, so configuring it in the webserver config would make more sense. Since Talked doesn't really care about the external endpoint provided by the webserver it should work.
Yeah that is reasonable. Maybe I will even mark uwsgi as recommended (not depends) and conditionally launch it or the binary directly. With API exposed internally only subroot should be non-issue anyway. I now think I may have said it confusingly during the all. It's only requirement for APIs that are exposed to outside via nginx.
Okay, fantastic!