> ## Knowledge Base Index
> Fetch the complete knowledge base index at: https://doc.ecomx.co/sitemap.xml
> Use this file to discover available pages before exploring further.
> Pure-Markdown content can be obtained by appending a '.md' suffix to the content URLs listed in the sitemap (without the trailing slash).

# Incidents and Diagnostics

AgentWatch attaches evidence so you can understand what changed and distinguish automatic detection from manual actions.

## Incident types

* **Channel traffic drop:** AI-referred sessions for one Shopify-reported channel dropped significantly.
* **Global Catalog drop:** store-level Global Catalog search requests dropped significantly.
* **Health check failure:** the same health check failed repeatedly.

## Severity

* **Warning:** the signal crossed the warning threshold or a health check failed twice consecutively.
* **Critical:** the signal crossed the critical threshold or a health check failed three times consecutively.

Severity describes the observed signal. It does not prove a Shopify or external AI platform outage.

## Status

* **Open:** the incident is active and has not been acknowledged.
* **Acknowledged:** someone has reviewed the incident, but the underlying signal is still considered active.
* **Resolved:** the incident was resolved manually or met automatic recovery criteria.

Acknowledging an incident does not pause monitoring or suppress future evidence. Resolving manually closes the record but does not change products, traffic sources, or external services. Reopen a resolved incident if it still needs investigation.

## Traffic and catalog anomaly rules

AgentWatch requires eight consecutive completed days: seven days for a median baseline and one current day for evaluation.

For AI-referred channel sessions:

* The seven-day median must be at least 20 sessions.
* A drop of 30% or more opens a Warning.
* A drop of 60% or more opens or escalates to Critical.

For Global Catalog search requests:

* The seven-day median must be at least 100 requests.
* A drop of 30% or more opens a Warning.
* A drop of 60% or more opens or escalates to Critical.

If there is not enough consecutive history or the baseline is below the minimum volume, AgentWatch does not create a traffic or catalog incident.

Automatic recovery requires two consecutive completed days at or above 80% of the incident baseline.

## Filter incident history

Use Status, Severity, Type, and Channel filters to narrow the table. Up to 100 matching incidents are shown, newest first.

Select **Evaluate latest metrics** to run anomaly rules against the latest stored metrics. This does not fetch new analytics; run an analytics sync first if the latest sync timestamp is old. Manual evaluations have a one-minute cooldown.

## Incident details

Open an incident to review:

* Baseline, observed value, and percentage change.
* Affected channel, catalog signal, or health check.
* Opened, acknowledged, and resolved timestamps.
* Resolution reason.
* Evidence timeline.
* Alert-delivery history.
* Sanitized diagnostic packet.

## Share diagnostics with support

Select **Copy diagnostic packet** and include it in your support request. The packet is designed to avoid secrets and customer-level data. Never add Shopify access tokens, passwords, API keys, or customer information to the message.

Next: [Alerts and settings](https://help.ecomx.co/en/articles/17607668-alerts-and-settings)