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

> ## Agent Instructions
> Unkey is two separate products. Compute builds, deploys, and runs apps behind a gateway. API Management issues API keys, enforces rate limits, manages identities and permissions, and reports usage. Say which product a page belongs to; a reader can use either without the other.
> Every Unkey API endpoint is an HTTP POST to https://api.unkey.com/v2/{service}.{procedure} with a root key in the Authorization: Bearer header. Root keys are workspace scoped.
> Error codes have the form err:{system}:{category}:{specific} and each has a page at /errors/{system}/{category}/{specific}.
> The word environment means production or preview in Compute. Rate limiting has four meanings on this site; the glossary lists them.

# Observability

> The signals Unkey records for a running app and where to find each one.

We record four kinds of data about your <Tooltip tip="A Compute app: a deployable service inside a project. Not 'your application' in general.">app</Tooltip>, with nothing to install in your code:

* Every request the gateway passes to your app.
* Everything your instances print to stdout and stderr.
* CPU, memory, disk, and network for each instance.
* Each build step and its output.

<Columns cols={2}>
  <Card title="Request logs" icon="arrow-right-arrow-left" href="/docs/compute/observe/requests">
    One row per proxied request: method, path, status, and the split between gateway and instance latency. Headers and bodies on request.
  </Card>

  <Card title="Runtime logs" icon="terminal" href="/docs/compute/observe/runtime-logs">
    stdout and stderr from every instance, parsed into severity, message, and searchable attributes.
  </Card>

  <Card title="Metrics" icon="chart-line" href="/docs/compute/observe/metrics">
    Requests per second, latency percentiles, CPU, memory, disk, network, instance count, and instance events.
  </Card>

  <Card title="Build logs" icon="hammer" href="/docs/compute/observe/build-logs">
    Each step of a build, whether it was cached, and its output.
  </Card>
</Columns>

## Where to look in the dashboard

* **Request and runtime logs:** open a project and click **Requests** or **Logs** in the sidebar. Filter by app, <Tooltip tip="A production or preview environment of a Compute app, not the dashboard label on a key.">environment</Tooltip>, <Tooltip tip="One built and running version of an app in one environment.">deployment</Tooltip>, or region.
* **Metrics and build steps:** open the app, click **Deployments**, and open a deployment. Its overview has the traffic and resource charts and the build steps. Its **Network** view shows regions and instances with live request rates.

## Rejected requests aren't in the request log

The request log only has requests that reached your app. Requests rejected by a policy, or that the gateway couldn't route (an unknown hostname, or no running instance), aren't there. So your request log can show less traffic than your callers sent. See [Request logs](/docs/compute/observe/requests).

## Retention

Each kind of data is kept for a fixed time, shown below. The dashboard can show all of it.

The [analytics API](/docs/compute/observe/analytics-api) has a second limit: your plan's log query range, 3 days on Starter, 7 on Pro, and 14 on Business. You can read back whichever is shorter. A query that reaches further fails with [`err:user:bad_request:query_range_exceeds_retention`](/docs/errors/user/bad_request/query_range_exceeds_retention) instead of returning partial results.

| Data | Stored for | Queryable |
| - | - | - |
| Request log rows | 7 days | Dashboard: 7 days. Analytics API: the shorter of 7 days and your plan's range |
| Per-minute request counts and latency percentiles | 14 days | Charts, no plan range |
| 5 and 15 minute request aggregates | 30 days | Charts, no plan range |
| Hourly request aggregates | 90 days | Charts, no plan range |
| Daily request aggregates | 365 days | Charts, no plan range |
| Runtime logs | 90 days from ingestion | Dashboard: 90 days. Analytics API: the shorter of 90 days and your plan's range |
| Build steps and build logs | 3 months | Dashboard only |
| 15 second resource metrics | 7 days | Charts, no plan range |
| Per-minute resource metrics | 30 days | Charts, no plan range |
| Hourly resource metrics | 90 days | Charts, no plan range |
| Daily resource metrics | 365 days | Charts, no plan range |
| Monthly resource metrics | 1825 days | Charts, no plan range |
| Raw resource samples | 95 days | Charts, no plan range |
| Instance events | 90 days | Charts, no plan range |

For example, on Starter the analytics API reaches back 3 days into runtime logs that are kept for 90. On Business it reaches 7 days of request logs, not 14, because that's how long they're kept. See [Compute limits](/docs/compute/configure/limits).

## Query the data yourself

You can query request and runtime logs with SQL through the analytics API, using the `gateway_requests_v1` and `runtime_logs_v1` tables, to build your own dashboards or alerts. See [Query gateway requests and runtime logs](/docs/compute/observe/analytics-api). Build steps and resource metrics are only in the dashboard.