Backend Health in the Dashboard

03 May 2026 • 4 minute
Carmen Dou

Carmen Dou

Software Engineer

Backend Health view in the InsForge dashboard

InsForge now has a Backend Health view in the dashboard — Observability and Backend Advisor, live in every project.

Observability

Every InsForge project runs on a dedicated EC2 instance. The dashboard now exposes its real-time signals as first-class cards on the project page — no SSH, no agent install, no separate metrics stack.

  • Five metrics: CPU usage, memory usage, disk usage, network in, network out.
  • Four ranges: 1h, 6h, 24h, 3d, switchable in one click.
  • Threshold gradients: cards transition from neutral to red as CPU crosses 60%, memory crosses 85%, and disk crosses 90%.
Observability section

The gradient is the point. A glance at the dashboard tells you whether anything is hot, before you read a single axis label.

CPU at threshold

Backend Advisor

Backend Advisor scans every active project daily across security, performance, and health. We surface findings inline in the dashboard — no separate report, no email digest to chase down.

  • Severities: critical, warning, info — filterable from a dropdown.
  • Categories: tabbed by area, each with its own count.
  • Re-scan: trigger a fresh scan from the section header; results poll in within 30 seconds. Last scan time sits next to the title so you always know how stale the view is.
  • Copy All: pull every finding currently displayed onto your clipboard as one batched prompt.
  • No-scan empty state: a fresh project shows "No scan yet" with a direct prompt to run the first scan.
  • Retention: findings stay for 30 days; resolved issues drop off automatically on the next scan.
Backend Advisor list

The scan looks for concrete patterns: slow queries above 1s, connection pools past 80%, unused indexes older than a week, low cache hit ratios, dead tuple bloat, sequence exhaustion, autovacuum blockers, and security misconfigurations like missing RLS policies on public tables.

Click any row to expand the finding — the rule that fired, the affected resource, the description, and the exact remediation. Each expanded card surfaces a Copy Remediation button, and that is where the AI-agent piece lives.

Expanded advisor finding

Copy Remediation as a Prompt

The Copy Remediation button does not put a sentence on your clipboard — it puts a full agent prompt: severity and category in the header, then the rule ID, the affected resource, the human-readable description, and a ready-to-run remediation snippet.

Paste into Cursor, Claude Code, or any agent shell and the agent has everything it needs to apply the fix without re-investigating the issue. The header Copy All copies every visible finding as one batched prompt — useful right after a fresh scan when there are several criticals to clear at once.

CLI Still Works

We kept the CLI as the way to script diagnostics into CI or local agent loops:

bash
npx @insforge/cli diagnose advisor --severity critical
npx @insforge/cli diagnose metrics --range 1h

Read the CLI launch post for the full diagnose workflow and the companion insforge-debug agent skill.

Get Started