Voice control

The whole point of Hey Caddy is to keep your hands on the club. The voice layer is two modes: hands-free (always-on listener with a wake word) and tap-to-talk (tap-and-speak). Both route through the same intent parser.

Two modes

Hands-free

Tap the floating gold mic to start. The listener runs continuously until you stop it (or end the round). On a wake-word match, it captures the next sentence, dispatches it, and speaks back a confirmation. Best for solo walks with AirPods in.

Wake-word variants we accept: Hey Caddy, Hey Kati, Hey Katie, Hey Kaddy. Apple’s on-device speech recognition isn’t perfect, especially in wind or in a cart — these variants catch most common mishearings.

Tap-to-talk

On Glance there’s a Tap to talkbutton in the bottom bar. On every other tab there’s a floating gold mic. Tap, speak one sentence, release. No wake word needed. More reliable in noisy environments.

Phrase library

Score a hole

• “Got a par on 7.”
• “Bogey on 12 with two putts.”
• “Birdied 14, one putt from 8 feet.”
• “Double bogey, three putt from 40.”
• “Hit driver 280 off the tee, had 130 in, hit pitching wedge to 15 feet.”

The parser extracts hole, score (named or numeric), putts, club, distance, fairway-hit, GIR. Any of those fields can be omitted — only score is required.

Score for someone else

• “Mike got a 5 on 12.”
• “Tom made par.”

Player names match against the active round’s roster. If the parser isn’t sure which player you meant, it asks once.

Caddy advice

• “Par 4, 153 over water, what should I hit?”
• “145 from the rough, what club?”
• “Wind’s in my face, 160 to pin.”

The caddy uses your hole, current GPS yardage, your club distances, the wind from the conditions card, and recent shots from this round. One club + one cue, under 35 words. No markdown, no emoji, plain text.

Presses

• “Team B is pressing the front.”
• “We’re pressing the overall.”
• “Team B presses the front and the overall.” (multi-segment)
• “Aloha press.” (shorthand: overall, hole 17 or 18, doubled stake)

Junk

• “Got a sandie.”
• “Greenie on 5.” (or just respond to the par-3 prompt)
• “Polie on 14.”
• “Mike had a snake.”

Wolf

• “I’m wolf, taking Mike.”
• “Lone wolf.” (go alone vs the field)
• “Blind wolf.” (declare lone before anyone tees off — bigger multiplier)

Round flow

• “Move to hole 8.” (jump current hole, useful when you skip or play out of order)
• “Recap my round.” (post-round summary)

Speech-to-text quirks

• The listener uses Apple’s on-device recognizer (no audio leaves the device for transcription).
• The audio session is set to .record only during listening. Combining .playAndRecord with the listener triggered Siri / Voice Control to hijack “Hey” commands — known and avoided.
• The TTS playback uses .playback / .spokenAudio so confirmations come through your AirPods cleanly even after the listener was active.

What if the parser gets it wrong?

Two recovery paths:

1. Open Scorecard → tap the wrong cell → edit the score directly.
2. If the wake-word listener is misfiring on background noise, tap the bar at the bottom to stop it and switch to tap-to-talk.