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

# Dashboard Overview

> Understanding your AI usage at a glance

The dashboard gives you a real-time view of your AI usage across all your connected providers. Every request that flows through the Metrion proxy is captured and surfaced here — tokens consumed, costs incurred, latency measured, and errors caught.

## Tabs

The dashboard is organized into five tabs. Each tab loads its data independently, so switching between them is instant after the first load.

<Tabs>
  <Tab title="Overview">
    The Overview tab is your home base. It combines three rows of widgets into a single unified panel.

    **Row 1 — Request summary**

    Three cards sit side by side:

    * **Requests** — total request count for the selected period, with a dual sparkline (successful requests in green, errors in red) and a period comparison badge showing the percentage change versus the previous equivalent period.
    * **Requests by model** — a bar breakdown of the top models by request count for the period.
    * **Errors** — total error count, error rate, and a bar breakdown by HTTP status code.

    **Row 2 — Charts**

    Two stacked area charts display token usage over time and cost by model side by side. A **Top Providers** panel on the right shows cost share per provider with a percentage bar. For time ranges of 24 hours or less, charts use per-request data bucketed by hour. For longer ranges, charts use daily aggregates.

    **Row 3 — Sparklines**

    Three compact sparklines give you a quick trend view:

    * **Costs** — spending over time for the period, with a period comparison badge.
    * **Requesters** — unique `x-metrion-user` identifiers active over time.
    * **Average latency** — mean response time over time, with a period comparison badge.

    <Info>
      Period comparison badges (showing `+12%` or `−5%`) compare the selected period against the immediately preceding period of the same duration. For example, if you select the last 7 days, the badge compares against the 7 days before that. Badges are hidden if the previous period has no data.
    </Info>
  </Tab>

  <Tab title="Models">
    The Models tab breaks down your usage by individual model.

    A stacked area chart shows **cost by model** over time. Below it, a comparison table lists every model you used during the period with the following columns:

    | Column      | Description                            |
    | ----------- | -------------------------------------- |
    | Model       | Display name and model ID              |
    | Requests    | Total request count                    |
    | Tokens      | Total tokens with a relative bar       |
    | Cost        | Total cost in your selected currency   |
    | Cost / req  | Average cost per request               |
    | I/O ratio   | Ratio of input tokens to output tokens |
    | Avg latency | Mean response time in milliseconds     |

    The most cost-efficient model is highlighted with a badge to help you spot savings opportunities.
  </Tab>

  <Tab title="Latency">
    The Latency tab shows how fast your models respond.

    A percentile chart plots P50, P75, P95, and P99 latency over the selected period. Below it, a per-model table shows:

    | Column   | Description              |
    | -------- | ------------------------ |
    | Model    | Model name               |
    | Requests | Number of timed requests |
    | Avg      | Mean latency             |
    | Min      | Fastest request          |
    | Max      | Slowest request          |
    | P95      | 95th-percentile latency  |

    A second table lists the **20 slowest requests** from the period, with their model, latency, HTTP status, token count, cost, and timestamp.
  </Tab>

  <Tab title="Errors">
    The Errors tab shows failed requests — any response with an HTTP status code of 400 or above.

    Three metric cards summarize the period: total errors, error rate (as a percentage of all requests), and an overall health status (OK / Warning / Degraded).

    A **timeline chart** plots 4xx and 5xx errors side by side over the period so you can spot spikes. A **breakdown panel** beside it lists each HTTP status code observed, with a count, percentage, and color-coded bar.

    Below those panels, a table lists the **last 50 error messages** recorded, with columns for timestamp, model, HTTP status code, error message text, latency, and requester.
  </Tab>

  <Tab title="Sources">
    The Sources tab shows usage broken down by the `x-metrion-user` header value you pass in your requests.

    A stacked area chart plots cost per source over time (top 8 sources). A table below it shows each source with:

    | Column      | Description                                       |
    | ----------- | ------------------------------------------------- |
    | Source      | The `x-metrion-user` value, or "Direct" if absent |
    | Requests    | Total request count                               |
    | Tokens      | Total tokens consumed                             |
    | Cost        | Total cost                                        |
    | Cost / req  | Average cost per request                          |
    | Latency     | Average response time                             |
    | Errors      | Error rate (orange above 5%, red above 10%)       |
    | Last active | Timestamp of most recent request                  |
    | Share       | Percentage of total traffic                       |

    Sources with no activity in the last 7 days are marked as **Inactive**.
  </Tab>
</Tabs>

## Time range

Use the date picker in the header to control the period displayed across all tabs.

**Quick ranges:** `1h`, `4h`, `8h`, `12h`, `1d`, `7d`, `30d`, `90d`

**Custom range:** Open the calendar to pick any start and end date. You can set a specific hour for both the start and end time, giving you hour-level precision.

| Plan | Maximum history |
| ---- | --------------- |
| Free | 7 days          |
| Pro  | 90 days         |

## Filters

Click **Filters** to open the filter panel. You can combine multiple filters — they apply across all tabs and persist in the URL so you can share a filtered view.

| Filter      | Description                                                    |
| ----------- | -------------------------------------------------------------- |
| Model       | Select one or more specific models                             |
| Provider    | Filter by provider (Anthropic, OpenAI, Gemini, Mistral, Grok)  |
| HTTP status | Filter by specific response status codes observed in your data |
| Requester   | Filter by `x-metrion-user` value                               |
| Cost range  | Set a minimum and/or maximum cost per request                  |
| Token range | Set a minimum and/or maximum token count per request           |

A purple badge on the Filters button shows how many filters are currently active.

<Tip>
  Use the Model and Provider filters together to isolate a single model's performance, or use the HTTP status filter to focus the Errors tab on a specific failure type like `429 Too Many Requests`.
</Tip>

## Currency

The cost graphs and metrics support three currencies: **USD**, **EUR**, and **CHF**. Change your preferred currency in your account settings. The conversion applies only to cost displays — all underlying data is stored in USD.