Crawler series

curl --request GET \
  --url 'https://truestat.io/api/v1/sites/{siteId}/ai/series' \
  --header 'Authorization: Bearer <token>'

{
  "data": [
    {
      "total": 338,
      "bucket": "2026-09-09 00:00:00",
      "indexing": 50,
      "training": 209,
      "answering": 2,
      "ai_referrals": 4
    }
  ],
  "meta": {
    "range": {
      "to": "2026-09-16T13:40:00.000Z",
      "tz": "UTC",
      "from": "2026-09-09T13:40:00.000Z",
      "clamped": false
    },
    "site_id": "6fed8980-a119-4a9c-a4d4-53e448ef9f9e"
  }
}

Crawler activity over time, with each bucket split by purpose — what the AI chart is drawn from.

Why the split lives here and not in /timeseries

/timeseries carries crawler_views, a single total per bucket. That is the line the visitor chart draws.

The split comes from its own query, and it is here because the distinction is the whole reason the AI section exists: a fetch that answers somebody's question right now and a fetch that harvests pages for a training run arrive the same way and mean opposite things. One is a reader you never see; the other is nobody at all. Collapsed into one line, that difference disappears.

So: one number for the chart, four for the argument.

total is more than the three parts

answering + indexing + training will usually come to less than total, and nothing is missing. Crawls whose purpose we do not recognise are counted in total and have no column of their own — a recognised crawler doing something unclassified, rather than an error.

Measured on a real site in one hour: 9 answering, 26 training, total 37.

Bucket size

unit takes minute, hour, day, week or month, and defaults to a size that suits the window — hours for a day, days for a month, weeks for a year. A year in minutes is half a million buckets, which is not a chart.

Buckets are cut in UTC

tz resolves the window here as it does everywhere: period=today&tz=Europe/Istanbul means your today, not the server's.

What it does not do is move the boundaries inside that window. A unit=day bucket starts at UTC midnight whatever tz you pass.

For this measurement it matters little — a crawler keeps no office hours, so an edge a few hours out changes the shape of the curve rather than what it says. It is written down because /timeseries does cut its buckets in tz, so the two charts disagree at the edges, and a caller comparing them deserves the reason rather than a mystery.

Authorizations

Path parameters

Query parameters

Response