Lemmy Webhook - add webhook support to your Lemmy instance (github.com)

submitted 10 months ago by rikudou@lemmings.world to c/fediverse@lemmy.world

21 comments fedilink hide all child comments

Note that unless you're a Lemmy instance admin, this doesn't have much use to you.

Until this package came along, if you wanted a bot that responds to events, you had to manually traverse all comments/posts/whatever at a fixed interval. With this package you can actually react to events directly from the database. It's implemented in a very efficient way by connecting the package directly to the Lemmy database and using native Postgres features to get the events (LISTEN/NOTIFY if you want to get technical).

The webhooks themselves are inserted into a separate SQLite database (API is coming) and allow for both simple and complex filtering of the incoming data. The system is already in use by two of my bots, @ChatGPT@lemmings.world and @DallE@lemmings.world who now both receive the information about being tagged in a comment in seconds (the actual reply takes a little longer, but that's because of the nature of the bot).

Currently you can be notified about a post or a comment, other types are trivial to include as well.

Let me know what you think!

you are viewing a single comment's thread
view the rest of the comments

[-] rikudou@lemmings.world 3 points 10 months ago

Yep, it uses pushing from the postgres to the webhook processor instead of polling for data periodically by an app. So after every insert, an event is pushed using the native postgres listen/notify mechanism and then the webhook processor doesn't interact with the database at all.

[-] originalucifer@moist.catsweat.com 2 points 10 months ago

yeah, but im seeing a reference to the doubling of my standing-processing as now i have an insert-after-event that didnt exist before... is that right?

i mean i get that youre pushing the processing to a different, functional mechanism, but its still additive processing on the server that needs accounting... and seems expanded.

[-] rikudou@lemmings.world 1 points 10 months ago

Yeah, everything you do takes processing power. This is done in a way that minimises the impact. There's no insert-after-event that I'm aware of. Also I'm not sure what you mean by expanded.

[-] originalucifer@moist.catsweat.com 1 points 10 months ago

Yeah, everything you do takes processing power. So after every insert, an event is pushed using the native postgres listen/notify mechanism

right, i was just curious how much processing this is. it gets expensive quick on a large instance, efficiency matters. might be negligible, but i watch my services like a hawk.