Skip to main content
SpeechKit is kombify’s local-first voice framework and runtime. One public project provides the Go voice kernel, a self-hosted server, agent tools, Android modules, and desktop clients for Windows and macOS. The current public release is v0.75.7, published 2026-10-05. SpeechKit is beta software: public APIs, configuration keys, and defaults can change between minor releases, so pin the version you tested.

​
Three modes and one recording service

Hands-Free is an activation and output layer over the three modes, not a fourth mode. Meeting is a separate recording service that shares only the speech-to-text transcriber with Dictation.

​
Where SpeechKit runs

​
Privacy boundaries you set

Two independent settings decide what SpeechKit may do:
  • Network Scope answers where the process may reach: all networks, local network only, or this device only. The backend enforces the boundary for speech providers, model downloads, updates, telemetry, server connections, and integrations. A stricter scope suspends incompatible settings instead of deleting them.
  • Retention Scope answers what outlives the work that produced it: retain, or zero retention. Zero retention sweeps the raw material when a session ends and keeps only the generated result.
The two are orthogonal. A device-only install still writes transcripts to disk unless you also choose zero retention.

​
Choose a target

Windows client

Install the Windows 10/11 x64 beta from the official release.

macOS client

Install the Apple Silicon Dictation beta and read its known limits.

Go framework

Embed the voice kernel, or build the server, Android, and browser targets.

Modes and data paths

Compare Dictation, Assist, Voice Agent, and provider boundaries.

Meeting

Record a call on two channels and produce a cited write-up.

Public repository

Browse source, examples, API contracts, and release notes.
The product site at speechkit.cc covers the same release from a product angle, including runnable starter prompts.