Which API and model each feature of the site uses, how often it was called, and what that cost (estimated from list prices — the invoice is in the OpenAI / Anthropic dashboards). Admins only.
Spend · estimated, all providers
What uses AI · one line per feature — click a row for models, route, cap and what it does
| Feature | API | Calls 7d | $ 7d | Calls 30d | $ 30d | Last used |
|---|
In a row's details, models in green were seen in the period; grey ones are wired but idle. A dashed outline means the model has no entry in the price sheet, so its cost shows as $0 — add it to ai_meter.PRICES.
By model · last 30 days
| API | Model | Calls | Errors | Tokens in | Tokens out | Audio min | Chars | Searches | Est. $ |
|---|
Per day · last 30 days, UTC
| Day | Calls | OpenAI $ | Anthropic $ | Total $ | |
|---|
Recent failures · newest first
No failed AI calls recorded. 🎉
| When | API | Model | Feature | Host | Error |
|---|
Price sheet used for the estimates
| Model | $ / 1M in | $ / 1M out | $ / audio min | $ / 1M chars | $ / search |
|---|
Edit app/ai_meter.py → PRICES when a price changes; rows are priced when they are written, so old rows keep the old price.