> ## Documentation Index
> Fetch the complete documentation index at: https://docs.orq.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Monitors
> Build custom dashboards that chart cost, latency, errors, and evaluator results from Traces, Metrics, and Logs.
## **What Are Monitors**
A **Monitor** is a custom dashboard made of widgets. Each widget charts one query over **Traces**, **Metrics**, or **Logs**, so a team can keep the numbers it cares about on a single page: cost per model, p95 latency, evaluator pass rate, and more.
Every **Monitor** has a scope:
* **Workspace**: charts data across every **Project** in the workspace. Only workspace admins can create, edit, or delete workspace **Monitors**.
* **Project**: charts data from a single **Project**.
The scope is set at creation and cannot be changed afterwards.
## **Creating a Monitor**
Head to **Observability** in the **AI Studio**, select the **Monitors** tab, then choose **+ Monitor**.
Enter a **Name** and an optional **Description** of what the **Monitor** tracks.
Select **Workspace** or a **Project**. Workspace admins see **Workspace** as the default. Other members only see the **Projects** they have access to.
Under **Start from template**, keep **Blank monitor** or start from a ready-made dashboard. See [Templates](#templates).
**Default range** is the time window the dashboard opens with. **Automatic** uses the standard range. Ranges beyond the plan's data retention are not offered.
Choose **Create monitor**. The dashboard opens straight away.
### Templates
| Template | What it shows | Default range |
| --------------------- | ------------------------------------------------------------------------------------------------------------------------------- | ------------- |
| **GenAI overview** | Total requests, cost, tokens, and errors, requests over time by model, p95 latency, and cost and errors by model. | Last day |
| **Cost analysis** | Total cost, requests, and tokens, cost over time by model, and cost by model and by **Project**. | Last 7 days |
| **Latency** | p50, p95, and p99 latency over time, and the slowest models and providers at p95. | Last day |
| **Evaluator quality** | Evaluator runs, pass rate, guardrails triggered, pass rate and average score over time, and guardrail block rate per evaluator. | Last 7 days |
Every widget created from a template can be edited, moved, or deleted like any other.
## **Using the Dashboard**
The dashboard header holds the controls for the whole **Monitor**:
* **Time range**: pick a preset range or a custom range. The range applies to every widget and is kept in the page URL, so a link to the dashboard opens the same window. Ranges are capped at the plan's data retention.
* **Refresh**: reloads every widget.
* **+ Widget**: opens the widget editor. See [Adding a Widget](#adding-a-widget).
* **Monitor actions**: **Settings**, **Duplicate**, and **Delete**.
Drag a widget by its header to move it, or drag its edge to resize it. Once the layout changes, choose **Save layout** to keep it, or **Cancel** to discard it.
### Widget actions
Open the menu on a widget to:
* **Edit**: reopen the widget in the editor.
* **Duplicate**: add a copy to the same **Monitor**.
* **Create alert**: open the [**Alerts**](/ai-studio/observability/alerts) form with the widget's metric, filters, and **Project** already filled in. A widget's target line becomes the alert threshold. Available for GenAI metrics from **Traces**. Alerts have no split, so **Split results by** is not carried over, and a workspace **Monitor** needs a **Project** picked in the form.
* **Delete**: remove the widget.
Time series widgets drawn as lines or bars also show a toggle in the header to switch between a line chart and a bar chart.
## **Adding a Widget**
The widget editor has the configuration on top and a live **Preview** of the widget below, using the dashboard's current time range.
* **Dataset**: **Traces**, **Metrics**, or **Logs**. Only the datasets available to the workspace are listed.
* **Measure**: the metric to chart, for example `Requests`, `LLM cost`, `Latency p95`, or `Evaluator pass rate`. See [Measures](#measures).
* **Aggregate**: how values are combined, such as count, sum, average, or a percentile. The options depend on the measure.
* **Interval**: the bucket size for time series, **Auto**, **Minute**, **Hour**, or **Day**.
Under **Where**, choose **+ Filter** to restrict the query, for example to one model, provider, or **Deployment**. Up to 20 filters per widget.
Under **Split results by**, select one or more dimensions, such as **Model**, **Provider**, **Project**, **Agent**, or **Tool**, to get one series or row per value. Leave it empty to chart all results together.
Enter a **Title**, then pick a **Visualization**. Only visualizations that fit the query are listed:
| Visualization | When it is available |
| --------------------------------------------- | -------------------------------------------------------------- |
| **Line chart**, **Bar chart**, **Area chart** | Values over time, with or without a split. |
| **Top list**, **Table** | A single value per split value. Requires **Split results by**. |
| **Big number** | A single value for the whole range. No split. |
Expand **Display options** to fine tune the widget. Which options appear depends on the visualization:
* **Sort** and **Limit**: show the **Top** or **Bottom** values, and how many (top lists and tables).
* **Value precision**: decimal places, from 0 to 6. Leave blank for automatic.
* **Show overall**, **Show others**, **Show percentage**: add a total row, group remaining values into one row, or show each value's share (top lists and tables).
* **Target line**: a reference value drawn on the chart (time series).
Check the **Preview**, then choose **Save**. The widget is added to the **Monitor**.
### Measures
The **Traces** dataset provides the GenAI measures below.
| Area | Measures |
| -------------- | -------------------------------------------------------------------------------------------------------------------------------------- |
| **Usage** | `Requests`, `Tokens` |
| **Cost** | `LLM cost` |
| **Errors** | `Errors`, `Error rate` |
| **Latency** | `Latency (avg)`, `Latency p50`, `Latency p95`, `Latency p99` |
| **Streaming** | `Time to first token (avg)`, `Time to first token p50`, `Time to first token p95`, `Time to first token runs` |
| **Evaluators** | `Evaluator runs`, `Evaluator pass rate`, `Evaluator failure rate`, `Evaluator score (avg)`, `Evaluator errors`, `Evaluator error rate` |
| **Guardrails** | `Guardrail runs`, `Guardrails triggered`, `Guardrail block rate` |
## **Managing Monitors**
The **Monitors** list shows each **Monitor** with its scope, widget count, last update, and who updated it.
* **Search** by name or description, and **Filter** by **Scope**.
* Open the menu on a row to open, **Duplicate**, or **Delete** a **Monitor**. A duplicate copies every widget.
* To rename a **Monitor**, change its description, or change its default range, open it and choose **Settings** from the menu.
## **Permissions**
Viewing, creating, editing, and deleting **Monitors** each depend on the member's role. Members without edit access see dashboards read only: widgets cannot be moved, resized, added, or edited. Workspace **Monitors** additionally require a workspace admin to create, edit, or delete them. **Create alert** on a widget also requires permission to create **Alerts**.