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

# Monitoring with Sentinels

> Find faults, drill into requests, and track fault rate over time

Once sentinels are configured, the Portal gives you a few complementary ways to monitor how they’re behaving in real traffic.

## Track fault rate at the Intent Group level

Open an Intent Group and use the **Overview** tab to monitor the **Fault Rate** chart over time. This helps you spot regressions and measure improvements after changes.

## Review recent faults (Events)

The Portal includes an **Events** page (`/events`) with a **Faults** table showing:

* date
* application
* intent
* session + request identifiers
* the faulting sentinel name
* whether a correction was applied (`Corrected`)

Clicking a fault opens the underlying session details so you can see the full context around the failing request.

## Drill down by sentinel

Open `Observe > Sentinels` and click into a Sentinel to see its **Requests**: the traffic it has
flagged, so you can answer "what kinds of requests is this sentinel failing on?"

## Debug a single case deeply (Request page)

On any Request detail page, use the **Evaluation** section to:

* inspect the per-sentinel results for that request
* run a **manual one-off evaluation** with a different set of sentinels (see [Evaluating a Request](/observe/sentinels/request_evaluations))

## Enabling evaluations and corrections

Sentinel evaluation behavior is also influenced by configuration:

* **Evaluations** can be enabled/disabled in configuration.
* **Apply Corrections** is only available when evaluations are enabled, and indicates that Maitai should automatically apply corrections based on evaluation outcomes.

## Notifications (optional)

If your Portal has notifications enabled, you can configure application-level preferences under `Settings > Notifications` (for example, choosing which applications should emit real-time fault notifications).