What caught my eye isn’t the birds, it’s the split. BirdNET-Go does the listening and classifying; Fugleramme just polls its API and redraws only when the birds change. Two local services, one boring interface, nothing leaving the network. That’s the shape I want for most of my self-hosted stuff, and it means you can point the frame at a detector you already run.
The other detail: the repo ships a CLAUDE.md, an AGENTS.md and a Claude skill called plate-review, yet none of the 800-plus cut-outs are AI-generated, only some retouched. That’s the division I’d copy: agents in the tooling, a human choosing what people actually look at. I’d start web-only on a spare screen before buying the panel.
The story — Fugleramme is an open-source Raspberry Pi bird frame by arnegiacomo. BirdNET-Go listens through a microphone and classifies species locally; Fugleramme matches each detection to one of over 800 hand-curated cut-outs from public-domain 1800s plates, sized by body mass, and renders them on an Inky Impression e-ink panel or any screen. Coverage favours the Nordics, British Isles and Germany. It installs via script or Docker, and the code is MIT-licensed. (Source)