Commercial voice assistants have always sent audio to the cloud for processing — that was the known trade-off from day one. But for a while, some devices offered a middle ground: an opt-out setting that kept certain voice interactions local, never leaving the hardware. That option is gone now. In early 2025, it was quietly removed. The official reason was AI: newer generative features require more processing power than the device itself can provide, so everything goes to the cloud, without exception, and without a way to decline.
The convenience was never the problem. A voice assistant is genuinely useful for the small, repetitive things — timers, music, reminders. But those tasks don't require a server farm. They never did. The cloud dependency isn't a technical necessity for a kitchen timer. It's a data strategy. "Always-on" and "always uploading" are two different things, and they don't have to come as a package.
Sage is built around the tasks that make a voice assistant genuinely useful day-to-day — not an AI chatbot, not a smart home hub. Just the practical stuff, done reliably, without the surveillance footprint.
Sage runs entirely on a Raspberry Pi in Python. The goal was to build something that lives in the kitchen like a small appliance — always available, never demanding, and completely understandable by anyone who wants to look at the code.
Keeping it Python-only wasn't just a language preference — it was a constraint that keeps the project readable, accessible, and easy to fork. The entire stack, from wake word detection through command parsing and response, is written in a single language.
The whole point of Sage is that you can see exactly what it does and doesn't do with your voice. Closed-source privacy claims are just claims. Open source is a receipt.
The project will be released publicly on GitHub once it reaches a stable, usable state. The plan is to make it easy enough to run that anyone with a Raspberry Pi and a reason to distrust their smart speaker can set it up themselves.