LatencySLA

LatencySLA is a real-time latency guardrail for teams running event-driven and microservice systems. It ingests traces/metrics, continuously computes p50/p95/p99 by route, tenant, and region, and triggers alerts only when user-impact is likely (e.g., sustained p99 breach plus error-rate shift). The app focuses on “latency budgets” per endpoint and release, showing which deploy, dependency, or queue caused the regression and how much of the budget is burned. It includes a lightweight canary checker that runs synthetic requests from multiple regions and correlates results with live telemetry to reduce false positives. The MVP is a web app with a small agent/collector and integrations for Slack/PagerDuty. This is a combination traditional + AI app: AI helps summarize incidents, propose likely culprits, and draft postmortems, but the core value is deterministic real-time monitoring and budgeting.