FriedrichAI: Offline AI · published · build 24846386

This update focuses on the underlying FriedrichAI backend, voice quality, and reliability.
The biggest visible change is a new local text-to-speech voice powered by Supertonic, with significantly more natural speech than the previous default. The voice model is bundled locally, works fully offline, and does not require an internet connection or external service. FriedrichAI will still fall back to the existing Piper voice where appropriate.
This update also includes a large batch of backend improvements: better response completion, improved handling of longer and numbered answers, expanded diagnostics and status reporting, cleaner error handling, and support for cancelling an active generation.
Local speech-to-text support has also been wired into the backend using a separate offline runtime. The user-facing microphone controls are not enabled yet; those will arrive with the upcoming UI work.
Other changes include improved medical-guidance handling for voice output, additional backend reliability work, and preparation for future UI and Friedrich Worlds development.
As always, FriedrichAI remains local-first: no account, no cloud dependency, and no telemetry required.