Coordinates a full session—screen, audio, transcription, desktop intent—and persists artifacts.

Orchestrator

Runs the session pipeline and stores results locally.

  • Path: workers/orchestrator/
  • API (port 8096): /session/start, /sessions, /sessions/{id}, /sessions/{id}/audio, /sessions/{id}/frames

Start Session

curl -sS -X POST http://localhost:8096/session/start \
  -H 'content-type: application/json' \
  -d '{"fps":1,"overlay":true,"auto_stop_seconds":60}' | jq

Environment

  • SCREEN_MONITOR_URL, AUDIO_RECORDER_URL, TRANSCRIBER_URL, DESKTOP_AUTOMATION_URL
  • OUTPUT_DIR, OVERLAY, FPS, AUTO_STOP_SECONDS, CALLBACK_URL

Output

  • session_{YYYYmmdd_HHMMSS}/ with audio_*.wav, frames_*.zip, transcript.txt, session.json
  • index.jsonl for quick listing