Operations Guide

Repeatable health, warm-up, benchmark, and performance-triage workflows for operators and agents.

Highest-value workflows:

• add warm-up before first speech
• preserve clientId in SDK initialization
• benchmark warm performance with real audio
• read the local dashboard before speculating about latency

Recommended operator loop:

vox doctor
vox warmup start
vox transcribe bench /path/to/audio.wav 5
vox perf dashboard --client <integration>

1. vox doctor
2. vox warmup start
3. vox transcribe bench /path/to/audio.wav 5
4. vox perf dashboard --client <integration>

Recommended client naming

Use stable product-surface IDs instead of per-user or per-session IDs:

• vox-cli
• menu-bar
• browser-extension
• editor-plugin

This keeps dashboard slices meaningful over time.

Performance triage order

When a user reports that transcription feels slow:

1. confirm whether the report is about hot-path inference or cold-path readiness
2. inspect inferenceMs before speculating about model quality
3. inspect totalMs and modelLoadMs to separate warm-up cost from steady-state cost
4. compare samples by clientId and route

Contributor checklist

• keep clientId, route, and modelId intact in telemetry
• avoid hiding model lifecycle inside helpers that make latency opaque
• prefer repeatable file-based benchmarks before changing live-session behavior

See Observability for metric interpretation and Quickstart for installation recovery.