> ## Documentation Index
> Fetch the complete documentation index at: https://docs.hubtalk.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Alerts

> Messages about campaign and agent failures in Telegram, Slack or the cabinet: how to create an alert, connect a chat and set up noise control.

# Alerts

The **Admin → Alerts** section reports failures of your campaigns and agents: a campaign stopped, calls do not go through, the agent could not transfer a call, call results do not reach your system. A message comes to Telegram, to Slack or stays in the cabinet.

The platform checks campaigns and agents every minute, so an alert comes no later than a minute after the event.

The section has two tabs:

* **Alerts** — what to watch and where to send it. Each alert is an object (a campaign or an agent), a set of events and the chats to send to.
* **Delivery log** — what went out and when, what was held back and what failed.

![The Alerts section with no alerts](https://mintcdn.com/hubtalk/0CLLnqL-ushSlhv0/images/v4_alerts_page_en.png?fit=max\&auto=format\&n=0CLLnqL-ushSlhv0\&q=85\&s=ce74aeb1f5219177605312d6cdfa1182)

## New alert

Click **New alert**. The window has five blocks.

![The New alert window: what to watch and the campaign events](https://mintcdn.com/hubtalk/0CLLnqL-ushSlhv0/images/v4_alerts_new_en.png?fit=max\&auto=format\&n=0CLLnqL-ushSlhv0\&q=85\&s=b75f88dd1e55bfeb15f200fec93ba888)

1. **Name.** It fills in by itself, for example "Alerts for All campaigns"; you can change it.
2. **What to watch** — **Calling campaign** (its progress and the trunk it dials through) or **Agent** (its calls and what it depends on). Below, pick a campaign or an agent — the list has a search — or **All campaigns** / **All agents**. "All" also covers the ones created later.
3. **Alerts** — what to watch on the chosen object, see the list below. A new alert already has the critical events ticked; **Select all** ticks the rest. Some events have a threshold: its field appears under the event once it is ticked. On the right you see how many objects the event watches, for example "Trunks of your campaigns: 1" or "Webhooks of your agents: 2".
4. **Send to** — one or more destinations, see [below](#send-to). An alert without a destination is not saved.
5. **Noise control** — repeats and quiet hours, see [below](#noise-control).

**Create alert** saves it and turns it on. Messages are written in the language of your cabinet: the language is saved with the alert.

In the list each alert has an **Alert enabled** switch and a menu with **Edit** and **Delete**. A deleted alert stops at once; the delivery log keeps its records. If the campaign or agent of an alert was deleted, the row says "The campaign or agent was deleted — the alert watches nothing".

## What you can watch

Each event has a level, shown as a coloured dot: **Critical: calls stop**, **Warning** or **Info**. Thresholds can be changed: a percentage from 1 to 99, a window from 1 to 1440 minutes, a count from 1 to 100. You cannot add an event type of your own.

### Campaign

| Event | When it comes | Default threshold | Level |
| - | - | - | - |
| **Campaign stop** | The campaign is paused, finished or cancelled | — | Info |
| **No calling during working hours** | The campaign is running and has contacts ready to call, but no new calls start | No new calls for 20 min | Critical |
| **Many calls without a connection** | Many call attempts end without a conversation. Technical failures count: no answer, busy, voicemail and a declined call do not | More than 60% of dials in 30 min | Warning |
| **Calling is delayed: every line is busy** | Calls postponed because every line is busy grow two checks in a row | — | Warning |
| **Many numbers that do not exist** | The carrier often answers that the number does not exist | More than 10% of dials in an hour | Info |

### Agent

| Event | When it comes | Default threshold | Level |
| - | - | - | - |
| **Agent failure during a call** | The agent could not produce a reply or perform an action. One alert per call | — | Critical |
| **The call could not be transferred to a person** | Nobody answered at the transfer number or the transfer was rejected. One alert per call | — | Critical |
| **The agent’s calls end with an error** | Calls the platform marked as failed reached the threshold | 5 calls in 15 min | Warning |
| **Calls end because of silence** | After the call connects no speech is heard, and the calls end | More than 15% of answered calls in 60 min | Warning |
| **Possible dissatisfaction in conversations** | Call analysis finds dissatisfaction of the people called | More than 20% of analyzed calls in 60 min | Info |
| **An automatic check flagged a call** | A check of the agent opened an [incident](https://docs.hubtalk.ai/v4/platform/incidents.md) on the call with a notification. One alert per incident | — | Warning |
| **Call results cannot be delivered** | Deliveries of the agent's webhook fail in a row | 5 failed deliveries in a row | Critical |
| **A knowledge base file could not be processed** | A file or knowledge base of the agent went into an error. Comes once, at the moment of the error | — | Warning |

![Agent events with default thresholds](https://mintcdn.com/hubtalk/0CLLnqL-ushSlhv0/images/v4_alerts_agent_en.png?fit=max\&auto=format\&n=0CLLnqL-ushSlhv0\&q=85\&s=3f1432cdbec049dfcfafab1c7ab4b3c4)

### Both campaign and agent

| Event | When it comes | Default threshold | Level |
| - | - | - | - |
| **Calls cannot be placed** | The last calls through the phone connection of the campaign or agent were all refused by telephony | 5 failed calls in a row | Critical |
| **Telephony settings were not applied** | A change of the phone connection was saved but not applied. Comes at once, without waiting for a failed call | — | Critical |

**Shares need a sample large enough.** Events with a percentage wait for at least 10 calls (20 for Many numbers that do not exist), so a couple of failed calls raise no alarm. An event on a share or a count comes at most once per its window.

**The first run sends no history.** A new alert reports events that happen after it was created.

## Send to

Chats are connected right in the alert window, in the **Send to** block. You can add as many as you like, each with its own name. A chat is saved only if it accepted a test message.

![Send to and Noise control](https://mintcdn.com/hubtalk/0CLLnqL-ushSlhv0/images/v4_alerts_send_en.png?fit=max\&auto=format\&n=0CLLnqL-ushSlhv0\&q=85\&s=ffe3cf15ef4343650eff1642c2e49f78)

* **HubTalk AI** — a checkbox. Nothing leaves the platform: the alert lands in the delivery log, and the **Alerts** menu item shows a counter of new ones.
* **Telegram** — open the Telegram list → **Add Telegram**:

  1. Add the bot to the group for alerts. The bot's name is shown in the connection window; a click opens its page in Telegram. If you use a channel, make the bot an admin and allow it to post messages.
  2. Open the group or channel in Telegram Web. In the address bar copy the number after `#` together with the minus, for example `-1001234567890`.
  3. Paste the number into **Chat id** and click **Check and save**. The bot sends a test message.

  No bot or token of your own is needed.
* **Slack** — open the Slack list → **Add Slack** and paste an Incoming Webhook URL like `https://hooks.slack.com/services/…` (in Slack: the app → Incoming Webhooks → Add New Webhook to Workspace). One webhook — one Slack channel.

Check a connected chat with **Send test** — the result is in the delivery log. **Delete chat** removes it from every alert of the organization; the alerts themselves stay. The chat id and the webhook URL are stored encrypted and shown only in part.

**Telegram:**

* **The group became a supergroup** — its ID changed, connect the chat again.
* **The bot was removed from the chat or lost its rights** — the log shows "Telegram refused the message".
* **In a group with topics** messages go to the General topic; you cannot pick a topic.

## Noise control

* **Repeats** — the pause between identical events on one object: **Every time**, **At most once per 5 min**, **10 min** (the default), **30 min** or **At most once per hour**. It is counted per event and object; held-back alerts show in the delivery log.
* **Quiet hours** — 22:00 to 08:00 by default, in the organization time zone; the window may cross midnight. During it non-critical alerts are held and come as one summary when quiet hours end. The summary has up to 20 lines, the rest as "and N more".
* **Critical: send at night too** — on when quiet hours are on: events of the **Critical: calls stop** level go out at once.

Quiet hours do not apply to HubTalk AI — it wakes nobody. Repeats do apply.

One event — one alert, even if the check reads the same period again.

## The message

A message has a title, two or three lines of facts and a link to the cabinet. The caller's number never gets into it. The link leads to where the problem is seen:

| Event | What the link opens |
| - | - |
| Possible dissatisfaction in conversations | The agent's calls with a negative sentiment |
| Calls end because of silence | The agent's calls ended because of silence |
| The agent’s calls end with an error | The agent's calls marked as failed |
| Many calls without a connection | The campaign's calls without a connection |
| Many numbers that do not exist | The campaign's calls with an invalid number |
| Agent failure, failed transfer, a flagged call | The call itself |
| Campaign stop, no calling, calling delayed | The campaign card |
| Telephony | The **Telephony** section |
| Call results cannot be delivered | The agent editor |
| A knowledge base file could not be processed | The **Knowledge bases** section |

Call history filters by whole days, so the link opens such calls for the whole day.

## Delivery log

The **Delivery log** tab is a table with the **Time**, **Alert and event**, **Destination** and **Status** columns and a status filter. The log is kept for 30 days.

* **Delivered** — the message was accepted. A note "Included in the quiet hours summary" means it went out in the morning summary.
* **Held back** — not sent now: "The repeat alert was not sent: the set interval has not passed yet" or "Sending is postponed for quiet hours".
* **Failed** — with the reason: "The alert channel is not connected", "Telegram refused the message" (the bot is not in the chat, a wrong ID, no right to post), "Slack refused the message" (the webhook was revoked or the URL is wrong), "The message could not be sent because of a network error".

**Show message** in a row opens the message text. For one that was not sent — the text that would have gone out.

If an alert did not arrive, look at its row in the log first: the reason is there.

## Who can do what

* **An administrator and a developer** create, change and delete alerts, connect and delete chats, send tests.
* **A manager and a member** see the alerts and the log marked "View only. Alerts can be created, changed and deleted by a developer or an administrator."

Each organization sees only its own alerts, chats and log. More on roles — [Roles](https://docs.hubtalk.ai/v4/platform/roles.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.