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

# Alerts

> Be told when a policy blocks, Neon escalates, a provider fails, or traffic changes.

Be told when something needs a person, instead of watching Sessions. An alert rule fires on an event or a metric and notifies the channels you choose.

For example, a rule on **Policy escalation** for all agents, sent to Slack, posts a message each time a policy escalates a call for review. The alert links to the sessions it covers.

## The Alerts page

**Alerts** in the sidebar lists the alerts your rules fired, with their **Severity** (Info, Warning or Critical) and **State** (**Firing** or **Resolved**). Filter by state, severity or agent. Open an alert to see its condition, the observed value and threshold, its scope and when it fired, and jump to **View filtered sessions** or **Open session**.

Xenovia also raises one built-in alert that needs no rule: a critical, organization-wide alert when Neon's checks can't be reached.

![Alerts table with fired time, severity, state, alert, description, scope and count](https://mintcdn.com/xenovia/RO5UF1Tk8L9B7yPt/images/screens/alerts-page.png?fit=max\&auto=format\&n=RO5UF1Tk8L9B7yPt\&q=85\&s=a39f830dda7b4af22c855f311d1c4bb2)

The Alerts page

## Add a rule

Rules live in **Settings › Alerts**; **Configure alert rules** on the Alerts page takes you there. Admins manage rules and channels; members can read them.

**1. Trigger and scope**

Choose **Add rule**. Give it a **Name**, pick its **Alert condition** (an event or a metric, below), its **Scope** (**All agents** or one agent) and its **Severity** (**Warning** or **Critical**). Optional **Filters** narrow it to a sub-agent, provider, model, policy rule or guardrail.

**2. Notification**

Choose its **Notification channels** (**In-app**, **Email**, **Slack**) and a **Notification cooldown** (**5 minutes**, **15 minutes** or **1 hour**): the minimum time between repeat notifications for the same alert.

A rule's type can't be changed after creation. When Nova is on, **Create with Nova** drafts a rule for you to finish.

## Event rules

An event rule fires each time the event happens.

| Trigger | Fires when |
| - | - |
| **Policy block**, **Policy redaction**, **Policy escalation** | A policy blocks, redacts or escalates a call |
| **Guardrail block**, **Guardrail escalation** | A Neon guardrail blocks or escalates a call |
| **High latency** | A call takes longer than the threshold you set, in milliseconds |
| **High guardrail score** | A guardrail's score is above the threshold you set, from 0 to 1 |
| **Run error** | A run fails. It doesn't fire for failed attempts the runtime moved on from |
| **Request blocked**, **Request escalated**, **Request error** | A whole request ended blocked, escalated or in error |
| **Request served by a fallback** | A [fallback](https://docs.xenovia.io/providers/fallbacks.md) answered a request |
| **Every provider failed** | A request ended in `fallbacks_exhausted` |
| **Compliance check starts failing** | A [compliance](https://docs.xenovia.io/organization/compliance.md) check fails. Listed only when Compliance is on |

## Metric rules

A metric rule fires when a measure over an **Evaluation window** (5 minutes to 24 hours) crosses a threshold. Pick the **Aggregation** (such as Mean or p95), the **Comparison** and the **Threshold**. **Group by** fires one alert per agent, policy rule, guardrail or provider, and **Minimum sample size** keeps a rule quiet below that many requests.

| Category | Metrics |
| - | - |
| LLM calls | **LLM latency**, **Time to first token**, **Token usage**, **LLM cost** |
| Enforcement | **Policy block rate**, **Guardrail score**, **Guardrail blocks**, **Escalations** |
| Turns and sessions | **Request error rate**, **Turn blocked rate**, **Turn volume**, **No traffic** |
| System | **Fallback rate**, **Provider failure rate** |

## Channels

Set up channels under **Notification channels** in **Settings › Alerts**:

* **In-app**: the alert shows on the Alerts page.
* **Email**: choose **Organization users** and add **Additional emails**.
* **Slack**: paste an **Incoming webhook URL** for the channel.

**Send test** sends a test message to a configured email or Slack channel.

## What next?

> [card] [Sessions](https://docs.xenovia.io/observability/sessions.md)
>
> Investigate what an alert points to.

> [card] [Policies](https://docs.xenovia.io/control/policies.md)
>
> Decide what escalates.

> [card] [Fallbacks](https://docs.xenovia.io/providers/fallbacks.md)
>
> Alert on provider failover.


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