Docs

Usage, sources and limits

Where every number comes from, and what Overtem can’t know.

Every usage figure in Overtem carries its source and when it was last read.

Sources

• Provider reported: the provider told us directly, such as Codex’s rate-limit window percentage and reset time.
• Counted locally: tokens added up from session metadata on your Mac. Accurate for what’s on disk, but not the provider’s own accounting.
• Billing reported: cost reported by an API billing source.
• Not reported / unavailable: the selected source has no valid reading, needs authentication or does not support that product’s allowance. Overtem says so instead of showing zero or estimating a quota.

Reset countdowns and alerts

Countdowns only appear when the provider reports a reset time. Alerts you set (for example at 80% of a window) fire from those same readings.

Heatmaps and recaps

Heatmaps, trends and weekly recaps are built on your Mac from your history. They stay local unless you choose to sync totals to your profile.

Overtem’s own usage

Analysis uses a budget you set. Input estimates and provider-reported token counts are labelled separately; an estimate is not a billing total. Routine monitoring makes no model calls. API organization usage and cost require a supported billing connection with the provider’s admin key and stay separate from subscription quotas.

Machine readings

CPU is shown for all cores combined, and per core as a percentage of one core. Memory separates app, wired, compressed and cached files. Swap in use is separate from swap activity. A reading that isn’t valid is marked as such, never shown as zero.