Build a voice assistant without selling your soul to Big Tech
Kalliope wires speech input to actions through YAML config, not cloud APIs.

What it does Kalliope is a Python framework for building personal voice assistants on Linux (including Raspberry Pi). You define “signals” — voice commands, scheduled events, MQTT messages, GPIO triggers — and map them to “neurons,” which are actions that execute in response. The whole flow is declared in YAML, no code required for basic setups.
The interesting bit The neuron marketplace lets you compose functionality like Lego blocks, but the real kicker is the escape hatch: when the pre-built modules don’t cut it, you write your own neuron in Python and plug it in. It’s open-source under GPLv3, meaning your assistant stays yours — no subscription, no surprise deprecation, no voice data shipped to a server farm in Virginia.
Key highlights
- Multi-language speech recognition and synthesis
- Runs on Debian-based Linux, including Raspberry Pi
- Signal sources: voice, cron-like scheduling, MQTT, GPIO events
- Neuron marketplace for community-contributed modules
- Companion Android app available
- “Easy-peasy” hello-world gets you talking in minutes
Caveats
- README claims “all Linux Debian based distribution” support but doesn’t specify exact dependencies or hardware requirements beyond “a microphone”
- Last copyright date in README is 2018; project vitality is unclear from this source
- Build badge still points to Travis CI, which shut down its free tier years ago
Verdict Tinkerers who want a Jarvis on a Pi without Alexa’s eavesdropping should look here. If you need turnkey reliability or Windows support, this isn’t your ride.