app-observability

The app-observability skill helps users stand up Grafana Cloud's application, frontend, and AI observability products on top of a shared OpenTelemetry (OTLP) plus Mimir/Loki/Tempo/Pyroscope pipeline. It solves the problem of wiring instrumentation and verifying that telemetry actually flows end-to-end: Application Observability provides RED metrics and service maps from OTel spanmetrics, Frontend Observability adds real-user monitoring via the Faro Web SDK, and AI Observability adds LLM/vector-DB monitoring through OpenLIT.

The skill walks through concrete workflows: configuring a Grafana Alloy OTLP receiver and forwarding to Grafana Cloud, then verifying receipt with health checks and metrics queries; instrumenting apps in Node.js, Python, Java, and Go with required resource attributes (service.name, service.namespace, deployment.environment); instrumenting a React frontend with Faro plus TracingInstrumentation for browser-to-backend trace correlation, Core Web Vitals, and session replay; and adding LLM cost/token/latency and hallucination/toxicity monitoring by initializing OpenLIT. It includes verification steps using TraceQL, PromQL, and LogQL, plus troubleshooting for missing service-map edges and collect-endpoint errors. Credentials such as the Grafana Cloud API key and instance ID are handled through environment variables and standard publisher scopes.

Target users are backend and frontend engineers, SREs, and platform teams setting up APM, debugging telemetry gaps, monitoring LLM cost drift, or correlating frontend errors to backend traces. The shell commands it runs are setup, health-check, and read-only verification operations against official Grafana tooling.