AI Insights continuously monitors your clusters and surfaces issues with root cause analysis, remediation steps, and platform-aware actions — before they become incidents.
What are AI Insights?
AI Insights are proactively generated health findings for your Kubernetes clusters. Ankra’s AI scans clusters on a schedule, detects anomalies, and produces structured insights with:
- Root cause analysis — what went wrong and why
- Remediation commands — copy-pasteable kubectl commands
- Platform actions — fixes through the Ankra platform (stack changes, add-on configuration, resource navigation)
- Conversation starters — pre-built prompts to continue investigating with the AI Assistant
AI Insights differ from AI Incidents in that they are generated proactively by scheduled scans rather than triggered by alert rules.
Enabling AI Insights
AI Insights are controlled at the organisation level.
Navigate to Organisation Settings
Click your profile icon → Organisation → Settings → General.
Enable Proactive AI Insights
Toggle Proactive AI Insights on in the AI Features section.
When disabled, AI analysis still works on-demand via the AI Assistant (⌘+J), but automatic proactive insights are not generated.
Viewing AI Insights
Navigate to any cluster and click AI Insights in the sidebar. The page has three tabs:
Insights Tab
The main list of all insights for the cluster with filtering, search, and multiple view modes.
View Modes:
| Mode | Description |
|---|
| Cards | Insight cards with severity, title, root cause summary, and affected resources |
| Compact | Dense list view for scanning many insights quickly |
| Grouped | Insights grouped by category or namespace |
| Filter | Description |
|---|
| Status | Open, acknowledged, resolved, dismissed |
| Severity | Critical, warning, info |
| Namespace | Filter by Kubernetes namespace |
| Category | Filter by issue category |
| Date Range | Created after / before a specific date |
| Search | Full-text search across title and root cause |
Dashboard Tab
At-a-glance summary of your cluster’s insight health:
- Severity donut chart — distribution of open insights by severity
- Trend chart — daily new vs resolved insights over time
- Category breakdown — clickable chart that filters the insights list
- Status counts — open, acknowledged, and resolved totals
Analytics Tab
Deeper analytics with configurable time ranges (7, 30, or 90 days):
| Metric | Description |
|---|
| MTTR | Mean time to resolution across all resolved insights in the period |
| Severity Trend | Daily breakdown of new insights by severity level |
| Namespace Breakdown | Issue distribution across Kubernetes namespaces |
| Category Breakdown | Issue distribution by category |
| Trend Direction | Whether cluster health is improving, degrading, or stable |
Insight Detail Page
Click any insight to open its detail page. The detail page has four tabs:
Overview
- Severity and confidence score — how severe the issue is and how confident the AI is in its diagnosis
- Root cause analysis — detailed explanation of the underlying issue
- Root cause summary — short summary shown inline on cards
- Affected resources — list of impacted Kubernetes resources with direct links to their resource pages
- Conversation starters — click any starter to open the AI Assistant with the insight context loaded
- Remediation commands — copy-pasteable kubectl commands to fix the issue
- Platform actions — fixes you can apply through the Ankra platform:
| Action Type | Description |
|---|
| Stack change | Adjust Helm values or stack configuration |
| Add-on configuration | Modify add-on settings |
| Resource navigation | Jump to the affected resource in the dashboard |
| AI chat | Pre-built prompt to investigate further with the AI Assistant |
Health
- Health snapshot — cluster health score, pod stats, and node stats captured at the time the insight was detected
- Anomaly snapshot — the specific anomalies (up to 15) that triggered the insight
History
- Previous resolutions — if this issue has recurred, shows each past resolution with:
- When it was resolved
- Resolution type and notes
- Effectiveness (whether the fix held or the issue came back)
- Resolution context — additional metadata about how the issue was fixed
- Health at resolution — the cluster health state when the issue was last resolved
On the detail page, a Related Insights section shows other open insights that share the same namespace or category. This helps you spot patterns — for example, multiple issues in the same namespace may point to a common root cause.
| Match Reason | Description |
|---|
| Same namespace | Both insights affect resources in the same Kubernetes namespace |
| Same category | Both insights fall under the same issue category |
Managing Insights
Status Lifecycle
| Status | Description |
|---|
| Open | Newly detected, requires attention |
| Acknowledged | Seen by a team member, being investigated |
| Resolved | Issue has been fixed |
| Dismissed | Not actionable or a false positive |
Resolving an Insight
When resolving an insight, you can provide:
- Resolution notes — free-text description of what you did
- Resolution type — how the issue was resolved (manual fix, configuration change, etc.)
- Resolution context — additional metadata captured automatically
Ankra also captures a health snapshot at resolution time so you can compare the cluster state before and after the fix.
Bulk Actions
Select multiple insights and apply bulk status updates (acknowledge, resolve, dismiss) in one action.
Feedback
After reviewing an insight, provide feedback:
- Helpful — the analysis was accurate and useful. This indexes the insight’s root cause analysis and recommendations into the RAG knowledge base, improving future AI analysis quality.
- Not helpful — the analysis was inaccurate or not useful. This helps the AI learn and improve.
Adaptive Scanning
The maintenance scheduler adapts scan frequency based on cluster health:
| Cluster State | Scan Interval |
|---|
| Critical issues active | Every 60 seconds |
| Active non-critical issues | Every 3 minutes |
| Healthy (no active issues) | Every 10 minutes |
Keyboard Shortcuts
Navigate and manage insights entirely from the keyboard:
| Shortcut | Action |
|---|
j / k | Navigate between insights |
Enter | Open detail page |
p | Quick peek (drawer) |
a | Acknowledge |
r | Resolve |
d | Dismiss |
c | Ask AI |
/ | Focus search |
1–4 | Switch tabs (Insights, Dashboard, Analytics) |
v | Cycle view mode (Cards → Compact → Grouped) |
? | Show keyboard shortcuts help |
Keyboard shortcuts are disabled when the command palette, a dialog, or an input field is focused.
API
AI Insights are available via the REST API:
| Endpoint | Method | Description |
|---|
/api/v1/clusters/{id}/ai-insights | GET | List insights with filtering and pagination |
/api/v1/clusters/{id}/ai-insights/count | GET | Get insight counts by status |
/api/v1/clusters/{id}/ai-insights/digest | GET | Analytics digest (MTTR, trends, breakdowns) |
/api/v1/clusters/{id}/ai-insights/{insight_id} | GET | Get insight detail |
/api/v1/clusters/{id}/ai-insights/{insight_id} | PUT | Update insight status |
/api/v1/clusters/{id}/ai-insights/{insight_id}/related | GET | List related insights |
/api/v1/clusters/{id}/ai-insights/{insight_id}/feedback | POST | Submit feedback |
/api/v1/clusters/{id}/ai-insights/bulk-update | POST | Bulk update insight statuses |
| Parameter | Description |
|---|
status | Filter by status (comma-separated) |
severity | Filter by severity |
namespace | Filter by Kubernetes namespace |
category | Filter by issue category |
date_from | Created after (ISO 8601) |
date_to | Created before (ISO 8601) |
search | Full-text search in title and root cause |
sort | recent, severity, oldest, recurrence |
page | Page number (default 1) |
page_size | Results per page (1–100, default 20) |
| Parameter | Description |
|---|
days | Lookback period in days (1–90, default 7) |
