Observability FinOps AI
LLM Token & Cost Dashboard
Volver a todas las demos
A live dashboard that turns raw token usage into cost insight — per feature, per customer, per model — with trends and alerts.
Tokens are money, and they’re easy to lose track of. This demo shows a token and cost dashboard that makes AI spend visible and attributable, so a prompt tweak or a retry loop never becomes a surprise invoice.
What it demonstrates
- Token metrics — input and output tokens per request, rolled up by feature, customer, and model.
- Cost attribution — tokens converted to cost using model pricing, trended over time.
- Latency percentiles — p50/p95/p99 alongside cost, because speed and spend trade off.
- Alerts — thresholds that fire when cost or token usage spikes after a change.
How it works
- Each model call emits token counts as OpenTelemetry metrics with GenAI attributes.
- Metrics flow through the Collector into a time-series backend.
- Dashboards compute cost from tokens and pricing, broken down by dimension.
- Alert rules watch for anomalies — a doubling of average tokens, a cost spike, a latency regression.
Why it matters
“Costs went up 30%” becomes “this feature’s average tokens doubled after Tuesday’s prompt change.” You move from end-of-month surprises to proactive cost control — FinOps for AI.
This dashboard ships as part of our AI Observability package.