Skip to content

gosset usage ​

How much of each dataset your account has seen, how many predictions it has run, and how close both are to your plan's limits.

bash
gosset usage

Real output, trimmed to a few rows:

Unique records seen in the last 365 days

  catalysts           [####################....]     833 /   1,000
  companies           [################........]   3,790 /   5,790
  deals               [#####################...]  17,653 /  19,710
  events              [##......................]   5,248 /  55,050
  news                [########................]  10,433 /  30,720
  …

Monthly allowance (last 30 days)

  ptrs       …
  timeline   …

Each row is one dataset: records seen, out of the yearly limit. assets is the drugs command and conference_results is abstracts. A row that reads seen (unlimited on your plan) has no cap. The prediction rows show calls used in the last 30 days, out of your allowance.

The last line sums it up: Within quota., or which dataset or allowance is used up.

bash
gosset usage --json     # the full report, machine-readable

What it counts ​

Distinct records, not requests. Asking for the same 200 drugs every morning gives a number that stops growing; walking through a dataset gives one that doesn't. Rate limits already bound how fast you can ask; this bounds how much of the database an account ends up holding.

A record counts once per window, however often it comes back. There are two windows:

WindowApplies to
rolling 365 daysrecords seen, per dataset (limit)
rolling 30 daysrecords seen, per dataset (limit_month), and prediction calls (ptrs, timeline)

Limits ​

Limits are set per dataset and per plan, as a share of each dataset. They are shares rather than fixed numbers because the datasets differ in size by orders of magnitude. Small datasets have a floor, so ordinary browsing never hits a limit.

In --json, each dataset has an entry like this (trimmed):

json
{
  "entity_type": "catalysts",
  "seen": 833,
  "limit": 1000,
  "seen_month": 502,
  "limit_month": 1000,
  "binding_window": null,
  "over_quota": false,
  "window_days": 365,
  "month_window_days": 30
}
FieldMeaning
seen, limitrecords seen in the last 365 days, and the cap
seen_month, limit_monththe same for the last 30 days
binding_windowwhich window is the one you have hit: month, year or null
over_quotatrue once a cap is reached. This is the field to watch
ungatedset when no cap applies on your plan

Prediction allowances are under features, one entry each for ptrs and timeline, with used, limit, remaining and over_quota.

If you're over ​

Requests for that dataset are refused until the window rolls forward or the plan changes. Nothing else is affected: the quota is per dataset, so exhausting trials leaves drugs working. A used-up prediction allowance returns HTTP 429 with an explanation; see Predictions.

Narrowing the query is usually the better fix anyway: --limit 500 pages through a dataset, while a filter answers the question.

bash
# walks through the data               # answers the question
gosset trials --limit 500              gosset trials --drug tulisokibart --phase 3

Gosset Documentation