henryklunaris/hey-jev

How henryklunaris/hey-jev works

Hey Jev is a macOS voice assistant: the user speaks in push-to-talk or wake-word mode, speech is transcribed locally, Jev classifies the request, and the assistant performs a Mac action or answers a question, then speaks a reply. The README also documents timers/reminders and a native status window with live countdowns and key settings. The excerpts show selected implementation details; timer and action orchestration links not directly visible are kept conservative.

Main entry points: siri.py Turn handling · assistant_ui.py Status window · secrets_store.py Key storage

Components 17

App experience

Status window assistant_ui.py
Key storage secrets_store.py

Other components

User
macOS
TypeSafe Jev
OpenRouter
Fish Audio

Connections 20

User → Audio capture: speaks to
Status window → Turn handling: starts runtime
Status window → Turn handling: sends controls
Audio capture → Wake detection: queues phrases
Wake detection → Local transcription: transcribes locally
Local transcription → Turn handling: provides text
User → Turn handling: submits text
Turn handling → Intent decisions: classifies request
Intent decisions → TypeSafe Jev: requests decisions
Turn handling → Mac commands: dispatches action
Mac commands → macOS: runs commands
Turn handling → Question answers: asks questions
Question answers → OpenRouter: requests answer
Turn handling → Timers and reminders: handles timer requests
Timers and reminders → Status window: provides countdowns
Turn handling → Reply selection: selects response
Reply selection → Speech synthesis: renders speech
Speech synthesis → Fish Audio: requests audio
Speech synthesis → macOS: plays audio
Status window → Key storage: reads and saves keys

More Python diagrams: growthepie/gtp-backend · gurknathe/pathfinding-algorithms · hal-ucsb/xarp · holobiomicslab/mimosa-ai · indexstorm/git-rec-back

Drawn by GitDiagram on September 26, 2026 from the repository's file tree, README and source files. Read as Markdown ·