Usage
See every request, its cost, cache, speed and the spending forecast.
The Usage page shows your requests from the log. The content of prompts and answers is not there. See Data and privacy.
Range and filters
- Pick a range: today, yesterday, this week, this month, or your own dates.
- The range reaches at most 90 days back, the same as how long detailed requests are kept.
- Filter by API key and model.
Columns
| Column | Content |
|---|---|
| Time | When the request came in |
| Model | The model used |
| Key | Name of the API key |
| Client | The calling application, when detected |
| Input tokens | Input token count, and the share read from cache |
| Output tokens | Output token count |
| Cost | In rupiah |
| Paid | Balance, Plan, or Plan + Balance |
| First token | Time until the first token arrived, in milliseconds |
| Speed | Tokens per second |
Analysis tab
The Analysis tab explains why your spending changed compared with the previous period. The difference is split in order into three causes: number of requests (volume), tokens per request, then rate and model mix. The three add up exactly to the spending difference; rounding goes to the last cause.
Chart and forecast
- The chart shows input tokens per day, split between tokens from cache and new tokens.
- The forecast of spending to the end of the month and the days of balance left need at least 3 days of usage.
- Requests paid by a plan allowance still show, with the Paid column marked Plan.