OpsProductObservability

Private operations console

Nidopr Monitor

One pane of glass for the whole business, alerting to your phone, no vendor lock-in.

Nidopr Monitor (PWA)
Nidopr Monitor preview
Private operations console — screenshots; live walkthrough on request.
40automated production health checks
3 sinksDiscord, email, web-push to phone
~48×endpoint fix: 4.8s → 0.1s
3-wayvisitor classification — bot / human / internal

A custom observability console for the Nidopr platform: a single installable app that watches the whole business, classifies who’s visiting, and pushes alerts straight to a phone. Built and operated as part of running Nidopr in production solo — no third-party monitoring vendor, no per-host billing, no lock-in.

Problem

A production system with a scraping layer, an ML pricing model, a public app, and dozens of unattended nightly jobs has a lot of surfaces that can fail quietly. Off-the-shelf monitoring means stitching together vendors, paying per host or per metric, and still not covering the business-specific things that actually matter — like whether a search actually returned results, or whether last night’s data refresh completed. The goal was one place to see all of it, owned end to end.

Build

A self-hosted observability layer purpose-built for the platform:

  • 40 automated production health checks covering the live surfaces of the business, not just generic infra pings.
  • Installable PWA that runs on desktop and on a Galaxy Fold — the console travels in a pocket.
  • Three alert sinks: Discord, email, and web-push directly to the phone, so a failure reaches a person wherever they are.
  • Visitor analytics with bot / human / internal classification — separating real prospective users from crawlers and the operator’s own traffic, so the numbers mean something.
  • A search → view → contact behavior funnel to see where real users drop off.
  • A self-healing client that recovers on its own rather than going dark.

Numbers

  • 40 health checks running against production.
  • 3 independent alert channels (Discord, email, web-push to phone).
  • Visitor traffic split three ways: bot / human / internal.
  • A profiled endpoint taken from 4.8s to 0.1s — roughly a 48× speedup — found and fixed using the console’s own instrumentation.

That last one is the point of the whole tool: the observability layer didn’t just report a slow endpoint, it gave the visibility to locate and verify the fix.

Stack

Installable PWA (desktop + Galaxy Fold) · web-push notifications · Discord and email alert sinks · self-hosted visitor analytics with bot/human/internal classification · a search-to-contact behavior funnel · a self-healing client · endpoint performance profiling. Part of the Nidopr platform, operated in production.

Stack

PWAWeb PushDiscordEmailVisitor analyticsPerformance profiling

← All work

Have data? Let’s make it think.

Open to senior / lead data & AI roles, and to Vizlogic consulting engagements.