Skip to main content
Pastures AI is the persistent chat drawer accessible via the sparkle icon (✦) in Rancher’s header bar. It provides AI-powered diagnostics grounded in real Kubernetes issue resolutions rather than generic LLM output.

Opening the drawer

Click the sparkle icon in Rancher’s top navigation bar from any page. The drawer slides in from the right and persists across page navigation.

Key capabilities

Context awareness

Suggestions adapt based on the current Rancher page and selected cluster.

Source citations

Every fix links back to the GitHub issue or KB article it was derived from.

Confidence scoring

Responses include a confidence percentage based on corpus match quality.

Fleet correlation

Banners indicate when the same issue appears across multiple clusters.

Fix verification

Post-fix monitoring confirms the issue is resolved.

Cluster memory

Indicator shows when the AI retains context from prior interactions with this cluster.

Drawer features

  • Expandable fix commands — code blocks with copy and run buttons
  • Action confirmation — destructive operations display a risk level before execution
  • Cluster memory indicator — shows when AI has prior context for the selected cluster
  • Ask AI buttons — appear inline on other pages to jump into the drawer with context

Engine API

POST /api/diagnose
endpoint
Sends a diagnostic query with cluster context and returns a grounded AI response.

Demo mode

Without an Engine connection, the AI drawer uses keyword matching to return canned responses for common scenarios:
KeywordDemo response topic
etcdetcd cluster health and quorum recovery
CrashLoopBackOffContainer restart loop diagnosis
upgradeVersion upgrade risk assessment
LonghornStorage replica and volume health
OOMOut-of-memory kill analysis
See the dedicated Pastures AI section for detailed documentation on each sub-feature.