These settings let you fine‑tune how your agent behaves in edge cases like voicemail, user silence, and maximum call length. These safeguards help keep calls concise and professional.

Voicemail

Automatically terminate calls that reach voicemail to save time and cost.

Field	Description
Voicemail detection	Toggle enabled/disabled. When enabled, an ML model listens at the start of the call.
Wait duration	Milliseconds to wait from call start before making the voicemail decision. (e.g., 8000 = 8 s)
End message (optional)	What the agent says before hanging up if voicemail is detected (e.g., “Looks like I reached voicemail. Goodbye.”).

Idle user

Define how the agent handles long silences from the user.

Field	Description
Silence duration	Milliseconds the agent waits after its last message with no user response.
Number of tries	How many times to play an idle prompt before ending the call.
Idle messages	Enter multiple prompts; one is chosen at random each time (e.g., “Are we still connected?”, “Can you hear me?”).
End message	Final message before hanging up due to user inactivity.

Example:

1. After 8000 ms of no input, agent plays a random idle message.

2. If still no response, repeats up to 3 times (number of tries).

3. After final attempt, plays end message and ends the call.

Maximum call duration

Set an upper limit on total call length to control costs and user fatigue.

Field	Description
Maximum duration	Total call time allowed in milliseconds (e.g., 480000 = 8 min).
End message	What the agent says before ending a call that hits the max duration (e.g., “We’re approaching the end of our call time. Let’s wrap up.”).

Post call data

For calls that end via these safeguards, you can identify the reason in the post‑call object:

Field	Value
${callInfo.endReason}	voicemail, idle_end, or duration_end

Use this key to track how often each safeguard triggers and tune thresholds accordingly.