Loading
Every upstream Gemini attempt this deployment made, as recorded by the ledger. No prompt, no response and no key is stored, so none can be shown.
| Purpose | Calls | Failed | Tokens today | Avg |
|---|---|---|---|---|
| rank-matchesOne call ranks the whole exhibitor list for one buyer's needs. Cached per needs, digest version, model and prompt version, so a second view of the same deck costs nothing. | none today | — | — | — |
| deep-analysisOne dossier for one company, on the stronger model, from the attendee's “Deeper analysis” button. Own daily cap (GEMINI_DEEP_DAILY_CAP) and cached 30 days. | none today | — | — | — |
| needs-chatOne call per TYPED or SPOKEN turn in “Refine this with AI”. Taps, the wrap-up card and the opening cost nothing. Own daily cap (GEMINI_CHAT_DAILY_CAP; 0 ships typed turns dark). A refinement also spends a second rank-matches call. | none today | — | — | — |
rank-matches, deep-analysis and needs-chat are always listed, at zero if need be. Their call counts come from the server’s own per-purpose counters, which cover every event this deployment serves; tokens and latency are read from this event’s ledger, so they read “—” for a call recorded against another event.
| Time | Purpose | Model | Latency | Tokens | Result |
|---|---|---|---|---|---|
No calls recorded | |||||
Aggregated from the same ledger rows, grouped by the event each call was made for and by what it was for.
The ledger writes each call against the request’sevent, so these rows say which event spent the calls. There is no cost column: MatchIt has no billing model and no price list, and a ringgit figure here would be invented. What the platform actually spends is read from the provider’s own console against these call and token counts.