A Python VA that runs locally and actually explains its brain

What it does

Stephanie listens for voice commands, converts speech to text, and tries to figure out what you want done. It’s built to automate daily tasks — calendar checks, weather, whatever modules you bolt on — through a local Python stack. You install it, add API keys to a config file, and run Index.py.

The interesting bit

Most assistants treat intent matching as proprietary magic. Stephanie ships its “brain” as a separate open-source project called Sounder, complete with a paper explaining the algorithm. That’s unusual transparency for a 797-star hobby project.

Key highlights

• Pure Python 3, Linux-first (needs portaudio19-dev)
• Modular architecture with a 3rd-party module ecosystem
• Intent prediction via the documented Sounder algorithm
• MIT licensed, though the author disclaims liability for community modules
• Active circa 2017 with Slack, Reddit, and Quora support channels

Caveats

• README shows its age: Python 3 and pip install instructions without version pinning, social links as primary support infrastructure
• No candidate images provided, so no screenshots or architecture diagrams visible in the repo itself
• Windows/macOS setup is absent from the README; Linux-only guidance

Verdict

Tinkerers who want a hackable voice assistant and care about understanding how it parses commands will find Stephanie worth a look. If you need polished cross-platform packaging or modern dependency management, this is not your project.