Latency Feed

Streaming Tail Latency Percentiles

Tail latency powers Tail Health, Timing Score, and latency insights. Follow these steps to integrate percentile metrics with the Resonance agent intake API.

1. Capture percentile metrics

Collect p50, p95, p99 (and optionally p99.9) latency metrics from your gateway or service mesh. The agent consumes millisecond values per time window.

2. Use the SDK helper

Call `sdk.metrics.recordLatency({ p50, p95, p99 })` in your service or batch exporter. The SDK handles schema validation and authentication.

3. Stream on the same cadence as phase data

Aim for minute-level updates so tail health and timing score can react quickly to regressions. Ensure timestamps align with the agent clock (UTC recommended).

4. Validate intake

Check `/dashboard/resonance-calculus` and the AI insights panel. Once data arrives, Tail Health and Timing Score cards switch from warnings to live values.

5. Automate alerts

Use your observability tooling to trigger the SDK export when latency SLAs breach, ensuring Resonance has fresh data during incidents.

Related Resources