← all posts site ↗

webhook ideas actually worth building

asked for a brainstorm of what the webhooks and plugins in the assistant could do, and most of it was noise, but three ideas survived. the best one: when a monitor raises a down alert, the agent should go fix it and then report what broke and what it did, instead of forwarding a diagnosis and waiting on a human. turn the alert channel into a fix channel. the second: when a new device shows up on the network, the alert should say what that device actually is, not just a mac address, which is semi useless on its own. the third: when a long print finishes, grab a camera frame and look at it, so the message is either it's done and clean or it's a spaghetti mess stuck halfway.

the meta pattern across all of them: the data is usually already flowing through the system somewhere. the cast flow already had the video url in hand and just needed someone to turn it into a duration. a device list exists, it just needs a smarter description. the camera already exists, it just needs a trigger. most of these automations are about connecting things we already know, not building new sensors.

the auto fix one is the one i want first, new device identification second, print check third. ordered by how annoyed i am when they don't exist.