GraphiteGraphite by Augsidius
Using the API

Rate limits & usage

Rate limits

Limits apply per key and per organisation, in requests per minute (RPM):

Plan RPM (organisation)
Sandbox 20
Starter 120
Growth 600
Enterprise By contract

You can set a lower per-key limit in Graphite. Every response carries x-ratelimit-* headers; on 429 rate_limited, wait until x-ratelimit-reset-requests before retrying.

Input limits

Each call accepts a bounded amount of text. Longer inputs return 400 context_length_exceeded; split them or send a summary.

Model Max input per call Output per call
aug-evidence-qa 2,000 tokens (question + earlier turns) One answer, up to ~250 words, with citations
aug-extract 4,000 tokens Up to 150 entities
aug-care-model 3,000 tokens Up to 10 conditions, 12 missing observations, 10 actions
aug-cds 3,000 tokens Up to 8 cards

Tokens are counted as characters ÷ 4.

Daily clinical-item limit

Each key may receive a limited number of distinct clinical items (concepts and evidence statements) per day. The limit sits far above clinical use — a busy clinic uses a small fraction of it — and exists to protect the evidence base from bulk copying. When it is reached, calls return 429 exposure_limit until 00:00 UTC.

Plan Distinct items per key per day
Sandbox 3,000
Starter 25,000
Growth 150,000
Enterprise By contract

Contact support@augsidius.com if a legitimate workload needs more.

Units

Calls are priced per call in units — see Per-call pricing. The units for each call are in usage.augsidius_units. Sandbox calls are not billed. Failed calls and refusals are not billed.

Monthly allowance

Each plan includes a monthly allowance of units for production keys. Depending on the plan, usage beyond it is either billed as overage per 1,000 units or blocked with 429 quota_exceeded. Allowances reset on the 1st of each month (UTC).

Watching usage

  • Graphite → Usage: units and requests per day by model and environment, error rates, CSV export.
  • Graphite → Overview: this month's units against your allowance.
  • API: GET /v1/usage?from=2026-11-01&to=2026-11-30&group_by=model with a key that has usage:read.

Spend controls

In Graphite → Billing → Spend controls set a monthly cap and choose:

  • Alert only — owners and billing members are emailed at 50%, 80% and 100%;
  • Hard stop — production calls return 429 quota_exceeded once the cap is reached.