Estimate which room Christophe is currently in using local Hue presence sensors plus optional D-Link/EZVIZ snapshots. Use when asked things like “which room am I in?”, “find me in the apartment”, “room presence now”, or recurring room-location checks.
Estimate room presence with a deterministic local workflow:
python3 /Users/cri/clawd/skills/apartment-room-locator/scripts/locate_room.py \
--capture-dlink \
--capture-ezviz \
--ezviz-motion-seconds 12 \
--ezviz-fps 1 \
--save /Users/cri/clawd/tmp/room_locate_latest.json
Notes:
--capture-ezviz now does sequence snapshots + motion detection.--capture-dlink now applies the D-Link vision-field map by default (if available at /Users/cri/clawd/data/dlink_vision_field_map.json) to skip known wall/ceiling blind zones before selecting the D-Link frame.--dlink-no-vision-map.--ezviz-code → env → local secret files.EZVIZ_CODE/Users/cri/clawd/data/ezviz-room-motion-history.jsonlUse these fields in your user-facing summary:
inference.roominference.confidenceinference.basisinference.reasons[]rooms_with_lights_on[]presence_signals[] (top recent rows)captures paths (if available)Always provide:
If confidence < 0.6, explicitly say it is tentative. If cameras show no person, say “not visually confirmed” and keep the estimate sensor-led.