Add position announcements

Enable position announcements on an existing queue: callers hear their place in line and estimated wait time. Want to deflect long waits instead of announcing them? See Deflecting long waits below.

1. Turn on announcements on the queue and tune the timing.

   {
     "id": "q_01j160r23gw5zkpfjme3fwzs5k", "name": "Support",
     "announce_position": 1,
     "announce_first_sec": 15,
     "announce_interval_sec": 30
   }

   • announce_first_sec: wait this long before the first announcement.
   • announce_interval_sec: the repeat interval; set to 0 to announce once.

2. (Optional) set the announcement language per queue. Otherwise the account/system default is used.

   { "id": "q_01j160r23gw5zkpfjme3fwzs5k", "language": "es-ES" }

3. Test it. Call into a busy queue. After ~15s you’ll hear your position and estimated wait (“you are caller number 3, estimated wait 5 minutes”); you’ll only hear it again when your position actually changes, so the hold music isn’t interrupted by a number that hasn’t moved.

Tip

Announcements are push-based and change-aware: the brain only breaks the audio when your position genuinely changes, which keeps stream-mode hold music effectively gapless.

Why it sounds right

Numbers are spoken with the correct grammar for each language. This includes gender-correct number agreement across languages like Hebrew, Arabic, Spanish, French and Russian (in Hebrew, for instance, the position number agrees masculine with “number” while estimated-wait minutes agree feminine with “minutes”). Each unique line is synthesized once and cached.

Deflecting long waits instead of just announcing them

Telling a caller their estimated wait time (EWT) is one option; skipping the wait entirely is another. A queue can deflect a caller before they ever join, if their estimated wait would exceed a threshold:

{
  "id": "q_01j160r23gw5zkpfjme3fwzs5k",
  "ewt_threshold_sec": 300,
  "ewt_aht_sec": 180,
  "ewt_dest_kind": "voicemail",
  "ewt_dest_id": "vm_01jg8asp59ps9rtt7azhrhvcre"
}

If the same estimated-wait calculation used in the spoken announcement predicts a wait past ewt_threshold_sec (using ewt_aht_sec as the average handle-time estimate), the caller is routed straight to ewt_dest_* instead of holding at all. See Overflow & Exit Destinations for the full mechanics, including how it fails open on a transient error.