Digging into Wisprflow

Infra & archi:

Text inputs, electron app, baseten, openpipe, 2 models supabase auth

what does it store locally vs push home?

how does it get context?

how is it so fast with 2 api calls?

binary transport mechanism

other beta/trial features like:

Command mode/Wispr Lens/Flow Lens ask_llm, edit_text, draft_text, build_app (replit!)

Scaffolded but Empty (FlowLensHistory table) unmistakably a multi-turn AI chat conversation log

Interpretation: Wispr planned FlowLens as a richer multi-turn chat/agent interface (with screenshots, tool use, conversation history, server sync), separate from the simpler single-shot Command Mode. The database was created, the onboarding was designed, but the UI and model code were either never shipped or were pulled back. The current code collapses everything into the simpler “Command Mode” pattern — stateless, single-shot commands via /llm/command_mode_route.

if this were made into an android app, what’s the equivalent

like how would we get context etc etc

SugarDaddyCTA FallenChampion