PocketAI
Taking the vendor hardware out of a vendor demo
A hold-to-talk AI assistant on a Raspberry Pi 5 with a 3.5-inch touchscreen, a USB microphone and a USB sound card - built by removing the dependency on one specific audio board.
01Why the port matters more than the demo
The upstream project assumes a particular audio HAT with a physical button on it. That works beautifully for anybody who owns that exact board and not at all for anybody who does not, which is most people.
This version replaces the button with touch-and-hold on the screen itself, takes audio through whichever ALSA card is present rather than one hard-coded chip, and brings Chromium up in kiosk mode without a keyring prompt in the way. Portrait orientation is configurable for the display and the touch layer together, since getting one right and the other wrong is the usual outcome.
It is a small piece of work and it decides whether the hardware ends up in a drawer: the difference between a demo that needs one exact part and a device somebody actually carries.
Questions about this project?
I am happy to walk through the architecture, the parts that did not work, or the code itself.