speechkit-voice

Overview ¶

Package main is the kombify SpeechKit Voice Server — a focused Linux container that exposes ONLY the Voice Agent mode (real-time audio-to-audio dialogue over WebSocket). Same Framework kernel and same code path as `speechkit-server`; the only difference is the built-in default mode set, which excludes Dictation and Assist HTTP routes so the binary stays small and the surface stays focused.

When to use which:

speechkit-server  → all three modes; one container handles
                    Dictation REST, Assist REST, Voice Agent WS.
                    Choose for single-pod deployments.
speechkit-voice   → only Voice Agent WS. Choose when scaling
                    voice independently from REST traffic, or
                    when running a stateful voice tier on
                    beefier nodes (GPU, more memory) than the
                    stateless REST tier needs.

The two are interchangeable from the device-target's `[server_connection]` perspective: ModeSource per mode lets the desktop app point Dictation/Assist at one URL and Voice Agent at another. See docs/server/README.md.