For AI agents: the complete documentation index is available at https://docs.flezise.com/llms.txt, the full documentation bundle is available at https://docs.flezise.com/llms-full.txt, and this page is available as Markdown at https://docs.flezise.com/analytics/usage.md.
Docs

Usage

The Usage tab shows token consumption and platform activity for the selected period. At platform scope it aggregates every project; at project scope it covers the selected project only.

Open Settings → Usage (platform scope, #org-usage) or a project's Settings → Usage (project scope, #usage).

The Usage analytics tab showing metric cards for Spend, Fresh tokens, Cached read, Avg. fresh per request, and Cache hit rate, with a Token mix over time chart below

Metric cards

Five metric cards appear at the top of the tab.

CardWhat it shows
SpendTotal gateway cost in the selected period
Fresh tokensUncached input plus output tokens sent to and from providers
Cached readPrompt tokens served from the provider cache
Avg. fresh / requestMean input and output tokens per request, excluding cache reads
Cache hit rateCached tokens as a share of all prompt tokens

Fresh tokens and Cached read together make up the total prompt token count; Cache hit rate shows how much of that prompt traffic was served from cache rather than re-sent to the model.

Token consumption

The Token consumption section has a Token mix over time chart labeled "Uncached input, cache reads, and model output." Three series are plotted so cache-heavy periods are visible alongside fresh traffic.

Below the chart, the Tokens per model table lists models with columns Model, Requests, Fresh input, Cached read, Output, Cost, and Provider total. Fresh input, cache reads, and output are kept separate so cache-heavy runs do not inflate the fresh prompt count.

Platform activity

The Platform activity section shows four metric cards.

CardWhat it counts
Artifacts createdDocuments created in the selected period
API eventsRecorded platform API requests
Knowledge queriesRAG and search lookups against the knowledge base
AI generation runsDispatched AI generation jobs

Three bar charts below the cards plot artifacts, API calls, and knowledge queries over the selected period.

When something goes wrong

If the AI gateway is unreachable, an amber banner appears: "AI gateway is unreachable. Token totals are temporarily unavailable." The gateway-derived metric cards (Spend, Fresh tokens, Cached read, Avg. fresh / request, Cache hit rate) show dashes. Platform activity counts come from the application database and remain visible.

FAQ

What is the difference between Fresh tokens and Cached read?

Fresh tokens counts input and output tokens that the provider processed without a cache hit. Cached read counts input tokens that the provider served from its prompt cache at a reduced rate. The distinction matters for cost analysis: cached reads are billed differently from fresh input by most providers.

Where does the cost figure on Usage come from?

The Spend card on Usage uses the same gateway-priced cost as the Project spend card on Cost. It is a summary figure for context; the Cost tab has the full feature and model breakdown.

Next steps