How voldemort9999/subxvoice-app works
SUBXVoice is a cross-platform desktop dictation app: the user holds a hotkey, audio is captured and streamed for speech recognition, transcript text is cleaned or rewritten, then inserted at the cursor with the clipboard restored. The graph also preserves the dashboard, local history, account/backend, native platform support, and planned voice-agent capabilities. Relationships among unsampled modules are limited to README-documented behavior; optional or pending capabilities are dashed.
Main entry points: workflow.ts Dictation workflow · pipeline.ts Transcript pipeline · App.tsx Dashboard UI · loop.ts Voice command agent
Components 20
Desktop experience
- Desktop shell
apps/desktop/src/main/index.ts - Dashboard UI
apps/desktop/src/renderer/src/App.tsx - Product pages
apps/desktop/src/renderer/src/pages - Live dictation pill
apps/desktop/src/renderer/src/pill/Pill.tsx - Account and sign-in
apps/desktop/src/main/auth.ts - Local recordings and history
apps/desktop/src/main/recordings.ts
Dictation pipeline
- Dictation workflow
apps/desktop/src/main/workflow.ts - Transcript pipeline
packages/core/src/pipeline.ts - Speech recognition
packages/providers/src/stt/gnani.ts - Text rewrite
packages/providers/src/llm/openrouter.ts - Dictionary and rules
packages/core/src/rules.ts
Platform integration
- macOS platform
packages/platform-mac/src - Windows platform
packages/platform-win/src - macOS native helper
native/mac-helper/Sources/main.swift
Providers and backend
- Supabase functions
supabase/functions - Provider selection
packages/providers/src/index.ts
Voice agent
- Voice command agent
packages/core/src/agent/loop.ts - Desktop actions
packages/platform-win/src/commands.ts - MCP integration
packages/mcp/src/index.ts
Other components
- User
Connections 24
- User → Desktop shell: uses
- Desktop shell → Dashboard UI: opens
- Dashboard UI → Product pages: routes to
apps/desktop/src/renderer/src/App.tsx - User → Dictation workflow: holds hotkey
- Dictation workflow → Transcript pipeline: processes audio
- Transcript pipeline → Speech recognition: transcribes
packages/core/src/pipeline.ts - Transcript pipeline → Text rewrite: rewrites text
packages/core/src/pipeline.ts - Transcript pipeline → Dictionary and rules: applies rules
packages/core/src/pipeline.ts - Transcript pipeline → Local recordings and history: reads vocabulary
packages/core/src/pipeline.ts - Dictation workflow → Local recordings and history: saves recordings
- Dictation workflow → macOS platform: uses platform
- Dictation workflow → Windows platform: uses platform
- macOS platform → macOS native helper: delegates capture
- Dictation workflow → Live dictation pill: shows live status
- Speech recognition → Supabase functions: may authenticate via
- Text rewrite → Supabase functions: may authenticate via
- User → Account and sign-in: signs in
- Account and sign-in → Supabase functions: requests account services
- Dashboard UI → Local recordings and history: shows history
- Provider selection → Speech recognition: creates provider
packages/providers/src/index.ts - Provider selection → Text rewrite: creates provider
packages/providers/src/index.ts - User → Voice command agent: speaks commands
- Voice command agent → Desktop actions: runs actions
- Voice command agent → MCP integration: may use tools