SLI Configuration

A Service Level Indicator (SLI) is an automated measurement that runs on a schedule. Each SLI executes a script (via a CodeBundle) that returns a numeric metric, which is stored and evaluated for alerting.

Managing SLIs in the UI

SLIs are configured as part of an SLX in Workspace Studio > Tasks. Open the SLX Preview (eye icon) and select the Health tab to see live SLI data.

SLX Preview — Health tab showing SLI metric values over time

The Health tab displays:

SLI Values — a time-series chart of the metric, with a configurable time range (1 hour, 6 hours, 24 hours, etc.)
Debug Log — link to the raw SLI execution log for troubleshooting

The Metadata tab shows the SLX name, owners, resource group, and all tags (platform, cluster, namespace, resource type, access level) that drive alert routing and task discovery.

SLX Preview — Metadata tab showing SLX name, owners, group, and tags

When creating or editing an SLI through the UI, the platform presents the key fields described in the spec reference below — interval, CodeBundle, alert thresholds, and task triggers — without requiring you to write YAML directly.

Spec Reference

The sections below document the full SLI Custom Resource spec for users who manage SLX configuration through Git or need to understand the underlying data model.

SLI Spec Overview

apiVersion: runwhen.com/v1
kind: ServiceLevelIndicator
metadata:
  name: my-workspace--my-slx-sli
  labels:
    workspace: my-workspace
    slx: my-workspace--my-slx
spec:
  codeBundle:
    repoUrl: https://github.com/runwhen-contrib/rw-cli-codecollection
    pathToRobot: codebundles/k8s-namespace-healthcheck/sli.robot
    ref: main
  intervalSeconds: 60
  intervalStrategy: intermezzo
  locations:
    - northamerica-northeast2-01
  configProvided:
    - name: NAMESPACE
      value: production
  secretsProvided:
    - name: kubeconfig
      workspaceKey: kubeconfig
  displayUnitsLong: "percent available"
  displayUnitsShort: "%"
  alerts:
    warning:
      operator: "lt"
      threshold: "0.99"
      for: "5m"
    ticket:
      operator: "lt"
      threshold: "0.95"
      for: "10m"
    page:
      operator: "lt"
      threshold: "0.9"
  alertConfig:
    tasks:
      persona: eager-edgar
      sessionTTL: 10m

Spec Fields

Field	Type	Default	Description
`codeBundle`	object	required	Git reference to the SLI script
`intervalSeconds`	integer	`60`	How often the SLI runs (in seconds)
`intervalStrategy`	string	`intermezzo`	Scheduling strategy for the runner
`locations`	string[]	—	Runner location(s) where the SLI executes
`configProvided`	object[]	—	Environment variables passed to the script
`secretsProvided`	object[]	—	Workspace secrets mapped into the script
`servicesProvided`	object[]	—	Location service bindings
`displayUnitsLong`	string	—	Human-readable unit label (e.g. “percent available”)
`displayUnitsShort`	string	—	Short unit label, max 3 characters (e.g. ”%”, “ms”)
`alerts`	object	—	Threshold alert configuration (see below)
`alertConfig`	object	—	Automatic task-trigger configuration (see below)

CodeBundle

The codeBundle field points to the script that the SLI runs:

Field	Required	Description
`repoUrl`	Yes	Git repository URL
`pathToRobot`	Yes	Path to the `.robot` file within the repository
`ref`	Yes	Git ref — branch, tag, or commit (default: `main`)

Config and Secrets

configProvided — static environment variables:

configProvided:
  - name: NAMESPACE
    value: production
  - name: CONTEXT
    valueFrom:
      workspace: CONTEXT

Each entry supports either a literal value or a valueFrom reference that resolves from the workspace or SLX configuration.

secretsProvided — workspace secret mappings:

secretsProvided:
  - name: kubeconfig
    workspaceKey: kubeconfig

The workspaceKey references a secret stored in the workspace’s secret store.

Threshold Alerts (`spec.alerts`)

Threshold alerts fire when the SLI metric crosses a user-defined value. Your SLI script can return any numeric value — not just 0 or 1 — giving you full control over how the metric is evaluated.

Structure

Threshold alerts are defined per severity level:

alerts:
  warning:
    operator: "lt"
    threshold: "0.95"
    for: "5m"
  ticket:
    operator: "lt"
    threshold: "0.9"
    for: "10m"
  page:
    operator: "lt"
    threshold: "0.8"

Fields

Field	Required	Type	Description
`operator`	Yes	string	Comparison operator for the metric value
`threshold`	Yes	string	Numeric value to compare against (parsed as float64)
`for`	No	string	Duration the condition must hold before firing (e.g. `5m`, `1h`)

Operators

Word	Symbol	Meaning
`lt`	`<`	Less than
`le`	`<=`	Less than or equal to
`eq`	`==`	Equal to
`ge`	`>=`	Greater than or equal to
`gt`	`>`	Greater than
`ne`	`!=`	Not equal to

Severity Levels

Severity	Typical Use
`warning`	Informational — may not require immediate action
`ticket`	Should be investigated
`page`	Requires immediate response

You can define one, two, or all three. Each operates independently with its own operator, threshold, and duration.

The `for` Duration

When for is set, the condition must be continuously true for the specified duration before the alert fires. This prevents transient fluctuations from triggering false alerts.

Without for — fires immediately:

metric_name < 0.95

With for: "5m" — fires after 5 minutes:

last_over_time(metric_name[5m]) < 0.95

Alert Modes

Each threshold alert has a mode (managed in CRD status, not spec):

Mode	Behavior
`active`	Alert fires when threshold is breached (default)
`silenced`	Alert does not fire but is still visible in the UI
`disabled`	Alert is completely disabled and hidden

Task-Trigger Alerts (`spec.alertConfig`)

Task-trigger alerts are a separate, implicit mechanism. When configured, the platform automatically runs the SLX’s tasks whenever any sub-metric drops below 1.0. The operator and threshold are hardcoded — only the task execution behavior is configurable.

alertConfig:
  tasks:
    persona: eager-edgar
    sessionTTL: 10m

Field	Default	Description
`tasks.persona`	`eager-edgar`	Which AI Assistant persona runs the triggered task
`tasks.sessionTTL`	`10m`	Cooldown — no new task run until this time has elapsed since the last run session

Threshold Alert Examples

Availability (0.0 – 1.0)

alerts:
  warning:
    operator: "lt"
    threshold: "0.99"
    for: "5m"
  ticket:
    operator: "lt"
    threshold: "0.95"
    for: "10m"
  page:
    operator: "lt"
    threshold: "0.9"

Response Latency (milliseconds)

alerts:
  warning:
    operator: "gt"
    threshold: "200"
    for: "5m"
  page:
    operator: "gt"
    threshold: "500"

Unhealthy Pod Count

alerts:
  ticket:
    operator: "ge"
    threshold: "1"
  page:
    operator: "ge"
    threshold: "3"

Error Rate (percentage)

alerts:
  warning:
    operator: "gt"
    threshold: "1"
    for: "10m"
  ticket:
    operator: "gt"
    threshold: "5"
    for: "5m"
  page:
    operator: "gt"
    threshold: "10"

How SLI Alerts Are Evaluated

SLI script runs on the configured intervalSeconds schedule at the specified runner location
Metric is stored under the SLX’s metric name
Alert rules are generated from your spec.alerts thresholds and evaluated continuously
When a threshold is breached — immediately, or after the for duration — the alert fires
An issue is created and linked to the SLX
Task triggers execute automatically if alertConfig is configured

SLXs (Learn) — conceptual overview of SLXs
SLO Configuration — error budget alerting against SLI metrics
Task & Runbook Configuration — task execution configuration