Orthographic microservice telemetry and semantic causal graphs for high-scale distributed systems.
When an outage storm hits a fintech with 400 microservices, the incident channel explodes. 50 engineers open 12 tabs each: Datadog for APM charts, Jaeger for flame graphs, Grafana for latency curves, and Kubernetes pod logs. The biggest bottleneck during an outage isn't database IO—it's human visual triage under intense cortisol load.
“I don't need another flame graph with 50,000 rows. When checkout fails, I need to know which upstream service broke the circuit in under 30 seconds.”
Traditional observability dashboards present data in disjointed silos. Engineers spend 8 to 14 minutes just cross-referencing trace timestamps. Meanwhile, standard DOM-based SVG graphs freeze the browser when rendering >2,000 trace spans, inducing lag precisely when engineers are under extreme pressure.
Relay Mesh unifies topological dependency graphs and span latency waterfalls onto a single 60fps WebGL canvas. Using custom R-tree spatial indexing, we perform raycast hover detection on 25,000 nodes in sub-1ms. When an alert fires, a causal cone algorithm automatically highlights the root anomaly and dims nominal downstream noise.
Decompresses binary OpenTelemetry spans in a Web Worker, batching updates every 16ms.
Building great software is mostly about deciding what *not* to build. Here is what we originally thought would work, why it collapsed in real usability testing, and what we replaced it with.
Our initial exploration used dynamic D3 force-directed graph physics. When nodes spun up, the whole graph wiggled and repositioned itself dynamically.
We locked service ingress to rigid columns and rendered dependencies strictly as a left-to-right causal cone. No wandering nodes, no spring wobble. Absolute spatial permanence.
Try the real component states below. Everything runs client-side in React with real state machines.