Default boundary
The README states voice, audio, and transcribed text stay on the machine unless a cloud AI provider is explicitly enabled.
local first dictation app
Local-first only works when every opt-in is visible before launch.
Use the planner to translate FluidVoice source facts into a concrete setup, privacy, model, and workflow decision.
Unofficial boundary
GPL-3.0 source
v1.6.1 release noted
Workflow
What this page helps decide.
Optional boundaries
AI enhancement, Fluid Intelligence, audio history, anonymous analytics, and beta builds are choices that should be documented.
Team checklist
Decide who can enable providers, whether history is saved, how audio budgets work, and when exports are allowed.
Decision matrix
What the planner checks before daily use.
| Area | Question | Planner output |
|---|---|---|
| Voice engine | Does the work need zero-download setup, low latency, multilingual support, Intel compatibility, or high accuracy? | Model recommendation with setup notes and disk expectations. |
| Permissions | Can users explain microphone, accessibility, Apple Events, and speech recognition access before approving prompts? | Onboarding checklist and risk labels. |
| Mode policy | Where should Command Mode, Write Mode, per-app prompts, and direct insertion be allowed? | Mode map with safe, review, and blocked zones. |
| Privacy | Which opt-ins are allowed: cloud enhancement, Fluid Intelligence, analytics, beta builds, local history, and exports? | Privacy boundary and data-retention notes. |