https://retellai.com/?utm_source=discord&utm_medium=seo logo
Powered by
i have the follwing setup for an agent:Reminder me...
# ❓・ai-evy-support
k
@AI Evy i have the follwing setup for an agent: Reminder message 1 time every 120s end of call silence 2.2 m My desired outcome is that the call hangs upt if no input from the client is detected at all at the 2 min 12s mark. but it keeps for longer than that
a
Thanks for tagging me @K Rudy. I'll look through relevant sources and get back to you shortly.
### Answer The behavior you're experiencing is expected based on how the 
end_call_after_silence_ms
and reminder settings interact [(1) ]() . The 
end_call_after_silence_ms
setting triggers after agent speech when the user stays silent for the specified period . Your 2.2 minute (132,000 ms) setting means the call ends 132 seconds after the agent's last speech, not from the beginning of silence . Here's what's happening with your setup: 1. User is silent for 120 seconds (2 minutes) 2. Reminder triggers (since 
reminder_trigger_ms
is set to 120,000 ms) 3. Agent speaks the reminder 4. Now the 
end_call_after_silence_ms
timer starts 5. Call ends 132 seconds after the reminder if user remains silent This means your total silence time is actually *252 seconds (4 minutes 12 seconds) * - the 120s before the reminder plus the 132s after. To achieve your desired 2 minute 12 second (132 second) total silence timeout, you have two options: Option 1: Disable reminders and set 
end_call_after_silence_ms
to 132,000 ms (132 seconds) - Set 
reminder_max_count
to 0 - Set 
end_call_after_silence_ms
to 132000 Option 2: Keep the reminder but reduce the post-reminder silence timeout - Keep 
reminder_trigger_ms
at 120000 (120 seconds) - Set 
reminder_max_count
to 1 - Set 
end_call_after_silence_ms
to 12000 (12 seconds) The minimum value allowed for 
end_call_after_silence_ms
is 10,000 ms (10 seconds) , with a default of 600,000 ms (10 minutes) .
​ To learn more, see the sources I considered: [[1] Create Voice Agent](https://docs.retellai.com/api-references/create-agent) ​
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.
k
hmmm i tried option 2, but the end fo call sielnce seems to overwrite the reminder instrcutions, so the call actually ends at the end of call silence time
@ai evy hmmm i tried option 2, but the end fo call sielnce seems to overwrite the reminder instrcutions, so the call actually ends at the end of call silence time
@AI Evy hmmm i tried option 2, but the end fo call sielnce seems to overwrite the reminder instrcutions, so the call actually ends at the end of call silence time
a
Thanks for the follow-up, @K Rudy. I'll get back to you soon.
### Answer Based on the behavior you're describing, it appears that 
end_call_after_silence_ms
is triggering before the reminder has a chance to activate [(1) ]() . The 
reminder_trigger_ms
parameter triggers a reminder to the agent to speak if the user has been silent for the specified duration after some agent speech . The 
end_call_after_silence_ms
parameter ends the call if users stay silent for a period after agent speech . When you set 
end_call_after_silence_ms
to 12,000 ms (12 seconds) and 
reminder_trigger_ms
to 120,000 ms (120 seconds) , the shorter timeout (12 seconds) will execute first, ending the call before the 120-second reminder can trigger. To achieve your 2 minute 12 second goal, you'll need to use Option 1: - Set 
reminder_max_count
to 
0
(disables reminders) - Set 
end_call_after_silence_ms
to 
132000
(2 minutes 12 seconds) This ensures the call ends exactly 132 seconds after the agent's last speech when the user remains silent. The minimum value allowed for 
end_call_after_silence_ms
is 10,000 ms (10 seconds) , and the default is 600,000 ms (10 minutes) . The default value for 
reminder_max_count
is 1 (remind once) , and setting it to 0 disables the agent from reminding .
​ To learn more, see the sources I considered: [[1] Create Voice Agent](https://docs.retellai.com/api-references/create-agent) ​
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.
k
@AI Evy but why didnt it ovewrite it int he frist place when i had the reminder 1 120s + end of call silence 2.2 setup
a
Thanks for the follow-up, @K Rudy. I'll get back to you soon.
### Answer --- When you had 
reminder_trigger_ms
set to 120,000 ms (120 seconds) and 
end_call_after_silence_ms
set to 132,000 ms (2.2 minutes) , the reminder didn't get overwritten because 132 seconds is longer than 120 seconds [(1) ]() . Here's the timeline that occurred: 1. User is silent after agent speech 2. At 120 seconds of silence → reminder triggers (because 
reminder_trigger_ms: 10000
was reached) 3. Agent speaks the reminder 4. Timer for 
end_call_after_silence_ms
resets because the agent just spoke 5. User remains silent for another 132 seconds after the reminder 6. At 132 seconds post-reminder → call ends The key difference is that 
end_call_after_silence_ms
measures silence after agent speech . When the reminder fires at 120 seconds, the agent speaks, which resets the 
end_call_after_silence_ms
timer. So you got the full 132 seconds of additional silence time after the reminder. In your Option 2 attempt (reminder at 120s + end call silence at 12s) , the 12-second timeout was shorter than the 120-second reminder trigger, so the call ended at 12 seconds of silence before the reminder could ever activate . The reminder only works when 
end_call_after_silence_ms
is set to a value greater than 
reminder_trigger_ms
.
​ To learn more, see the sources I considered: [[1] Create Voice Agent](https://docs.retellai.com/api-references/create-agent) ​
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.
k
@AI Evy whe i update a knowledgebase, do i need to publish another version of the agents for the chnages to go live?
a
Thanks for the follow-up, @K Rudy. I'll get back to you soon.
### Answer Based on the available sources, when you update a knowledge base with auto-sync enabled for URL sources, the system automatically re-fetches all URLs every 24 hours to ensure the latest content is reflected [(1) ]() . For URL paths with auto-crawling enabled, the system automatically crawls all pages under each path every 24 hours . However, the sources don't explicitly state whether you need to publish a new agent version for knowledge base changes to take effect in production. The versioning documentation explains that published versions of agents lock in configuration and cannot be changed [(2) ]() , but it doesn't specifically address whether knowledge base updates require republishing. For the most accurate answer about whether knowledge base updates require publishing a new agent version, I'd recommend: Check out these potentially useful sources: - [Knowledge Base]() - [Setup versioning for agents]() Would you like me to provide a general answer based on my knowledge instead?
​ To learn more, see the sources I considered: [[1] Knowledge Base](https://docs.retellai.com/build/knowledge-base) [[2] Setup versioning for agents](https://docs.retellai.com/agent/version) ​
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.
k
@AI Evy yes, whats your answer
a
Thanks for the follow-up, @K Rudy. I'll get back to you soon.
### Answer Based on my understanding of how knowledge bases work with agents, when you update a knowledge base (by adding/removing URLs, documents, or text) , those changes should be reflected automatically for agents using that knowledge base without requiring you to publish a new version [(1) ]() [(2) ]() . The knowledge base operates as a separate resource that agents reference. When you link a knowledge base to an agent, the agent queries it during conversations to retrieve relevant information [(1) ]() . Updates to the knowledge base content—especially with auto-refreshing enabled for URLs (which re-fetches content every 24 hours) —should be available to the agent without republishing . However, if you're changing which knowledge bases are linked to an agent or modifying knowledge base configuration settings within the agent editor itself, you would likely need to publish a new version for those structural changes to take effect in production. In summary: - Content updates within an existing knowledge base (adding/editing URLs, documents, text) : No republishing needed - Linking/unlinking knowledge bases or changing KB settings on the agent: Republishing required
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.
​ To learn more, see the sources I considered: [[1] Knowledge Base](https://docs.retellai.com/build/knowledge-base) [[2] Empower AI with Dynamic Knowledge Bases | Retell AI](https://www.retellai.com/features/knowledge-base) ​
Thanks for the follow-up, @K Rudy. I'll get back to you soon.
k
@AI Evy could you exmplain trigger_timestamp from the batch call api
a
### Answer The 
trigger_timestamp
parameter in the batch call API is the scheduled time for sending the entire batch call, represented as a Unix timestamp in milliseconds [(1) ]() . If you omit 
trigger_timestamp
, the batch call will be sent immediately . If you include it, all calls in that batch will be queued and dispatched at the specified time [(1) ]() [(2) ]() . Important distinction: 
trigger_timestamp
controls when the entire batch starts executing, not when individual calls within the batch are made [(2) ]() . Currently, you cannot schedule individual calls within a batch to different times using the CSV upload . Each call in the batch will be initiated once the batch is triggered. The Batch Calling feature offers two modes [(3) ]() : - **Send Now**: Dispatch all calls immediately - **Scheduled Time**: Send out calls at a scheduled time (using 
trigger_timestamp
) The pricing is $0.005 per dial [(4) ]() [(3) ]() .
​ To learn more, see the sources I considered: [[1] Create Batch Call](https://docs.retellai.com/api-references/create-batch-call) [[2] If I upload Batch call from dashboard, does it know to call at a required time?](https://discord.com/channels/1186424376300679209/1338997190579847228) [[3] Batch Calling, New Docs, and More Improvements!](https://www.retellai.com/changelog/batch-calling-new-docs-and-more-improvements) [[4] Streamline Calls with Retell AI’s Batch Dialer | Retell AI](https://www.retellai.com/blog/boost-operation-efficiency-with-retell-ais-batch-calling) ​
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.
k
@AI Evy so, what happend if a set a batch call via api with trigger timestamp 1777272009511 But day is defined as "day": [ "Wednesday" ] Also, is that format correct for day, can i have an bject with only one day?
a
Thanks for the follow-up, @K Rudy. I'll get back to you soon.
### Answer --- The 
trigger_timestamp
(1777272009511 in your example) determines when the batch call begins executing [(1) ]() . However, the 
call_time_window
with 
day: ["Wednesday"]
restricts when individual calls within that batch can actually be placed . Here's what happens with your configuration: 1. The batch is triggered at timestamp 1777272009511 (which converts to approximately May 24, 2026) 2. Calls will only be placed on Wednesdays within the time windows you've defined 3. If the trigger timestamp falls on a non-Wednesday, calls will wait until the next Wednesday that matches your time window criteria Regarding the 
day
format: Yes, having an array with only one day is correct . The 
day
parameter accepts an array of weekday strings, so you can specify one or multiple days:
Copy code
json
"day": [
  "Wednesday"
]
Or multiple days:
Copy code
json
"day": [
  "Monday",
  "Wednesday",
  "Friday"
]
The 
call_time_window
object also requires 
windows
(defining start and end times in minutes since midnight) and 
timezone
. The windows use a half-open interval format 
[startMin, endMin)
where cross-midnight windows are not allowed . Example:
Copy code
json
"call_time_window": {
  "windows": [
    {
      "start": 540,
      "end": 1020
    }
  ],
  "timezone": "America/New_York",
  "day": [
    "Wednesday"
  ]
}
This would restrict calls to Wednesdays between 9:00 AM and 5:00 PM Eastern Time .
​ To learn more, see the sources I considered: [[1] Create Batch Call](https://docs.retellai.com/api-references/create-batch-call) ​
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.
k
@AI Evy so, in theory if i setup a batch call that has a unix time stamp equivalent to April 27, 2026 at 6:40:09.511 AM But day is only an object with Sunday, The batch will send no calls until Sunday May 3rd 2026?
a
Thanks for the follow-up, @K Rudy. I'll get back to you soon.
### Answer From April 27, 2026 (which is a Sunday) , I calculate that your scenario doesn't require waiting until May 3rd, 2026. April 27, 2026 is already a Sunday [(1) ]() . If your 
trigger_timestamp
is set to April 27, 2026 at 6:40:09.511 AM and your 
call_time_window
has 
day: ["Sunday"]
, the batch calls would be placed on *that same day (April 27, 2026) * as long as: 1. The time 6:40:09 AM falls within your defined 
windows
time range [(1) ]() [(2) ]() 2. The 
timezone
in your 
call_time_window
matches your intended timezone [(1) ]() The 
call_time_window
with 
day
restricts when individual calls can be placed [(2) ]() . Since April 27, 2026 is already a Sunday and matches your day restriction, calls would proceed immediately (assuming the time falls within your configured windows) [(1) ]() . If the trigger time (6:40 AM) falls outside your configured time windows, then calls would wait until the next valid Sunday time window, which could be later that same day or the following Sunday (May 4, 2026) [(1) ]() [(2) ]() .
​ To learn more, see the sources I considered: [[1] Create Batch Call](https://docs.retellai.com/api-references/create-batch-call) [[2] Create Batch calls](https://docs.retellai.com/deploy/make-batch-call) ​
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.
k
@AI Evy ok thanks for that but April 27, 2026 its actually a Monday
a
Thanks for the follow-up, @K Rudy. I'll get back to you soon.
### Answer You're absolutely right—thank you for the correction. April 27, 2026 is a Monday [(1) ]() [(2) ]() . Given your scenario where: - 
trigger_timestamp
= April 27, 2026 at 6:40:09.511 AM (a Monday) - 
day: ["Sunday"]
The batch will not send calls on April 27, 2026 because that day doesn't match your 
call_time_window
restriction . The calls will wait until the next Sunday that falls within your configured time windows . From April 27, 2026, I calculate the next Sunday is May 3, 2026. So yes, your batch calls would begin on Sunday, May 3, 2026, assuming the time windows you've configured include valid calling hours on that day . The 
call_time_window
with the 
day
parameter restricts when individual calls within the batch can actually be placed [(2) ]() , so even though the batch is triggered on April 27, the calls will queue until a day that matches your 
day
restriction arrives [(1) ]() [(2) ]() .
​ To learn more, see the sources I considered: [[1] Create Batch Call](https://docs.retellai.com/api-references/create-batch-call) [[2] Create Batch calls](https://docs.retellai.com/deploy/make-batch-call) ​
Feel free to tag @AI Evy with additional questions. I also appreciate feedback, it helps me improve.
PreviousNext
Powered by