TraceTriage

TraceTriage is a web app (with optional CLI agent) that continuously samples and ranks distributed traces to highlight the few that actually matter. Most observability stacks drown teams in high-cardinality traces, expensive ingestion bills, and low-signal alerts. This product sits on top of OpenTelemetry and your existing backend (Tempo/Jaeger/Datadog/New Relic) to identify “incident-shaped” trace patterns: sudden latency shifts, new error paths, dependency fan-out explosions, and rare-but-costly endpoints. It auto-builds a short incident brief: suspected service, top exemplars, recent deploy correlation, and the minimal set of traces to inspect. It’s a combination traditional + AI app: deterministic detection for reliability, plus LLM summarization to produce readable triage notes and suggested next queries. The goal is not another full observability platform—just a focused triage layer that reduces MTTR and trace spend.