Usage

Usage tracks requests that pass through the selected Msty Nexus Runtime. Use it to understand what Nexus routed and how local or online models performed.

What Usage records

Nexus records one event for each gateway generation request that reaches a provider. Safe fields include:

• Provider and model IDs
• Client token ID
• Success or failure
• Provider-reported input, output, and cache token counters
• Tool names and counts
• Smart Route decisions and fallback reasons
• Acceleration and generation-speed counters
• Estimated cost where model pricing is known

Usage does not store prompt text, response text, reasoning text, tool arguments, credentials, headers, or raw provider bodies.

Views and filters

Usage can show:

• Summary totals and daily activity
• Token and tool-call heatmaps
• Provider, model, and client-token breakdowns
• Smart Route paths and fallbacks
• Prompt-cache reads, hit rate, savings, and reused tokens
• Local generation speed and acceleration history
• Reported or estimated cost

Filter by time range, provider, model, or breakdown dimension. The page follows the machine selected in the workspace switcher.

Local inference performance

For supported local runtimes, Nexus reports generation throughput and enough runtime context to distinguish model speed from queueing or setup time.

Treat short windows carefully. A small sample can be useful for debugging but is not a stable performance baseline.

Smart Route value

Smart Route insights use Usage evidence to compare routed traffic with the configured fallback. They may include request share, latency, quality evidence, online cost, and estimated savings when enough comparable traffic exists.

See Smart Routes for decision history and route-specific insights.

Retention

Usage retention is controlled by usage.retentionDays. Shorten it before a rollout when local policy requires a smaller history window.

Public Usage endpoints return aggregates and bounded safe records rather than raw request content.