Settings
The engine's constants, your class priors, and which store this deployment is really using.
Class priors
How much each risk class contributes to the score. Normalised before use, so only the relative sizes matter.
- 1.00
- 1.00
- 1.00
- 1.00
- 1.00
- 1.00
- 1.00
- 1.00
- 1.00
- 1.00
Engine constants
- version
- herbarium-grade/1.0.0
- saturation
- 1.2
- gap threshold
- 35%
- classes
- 10
Tier weights
- monograph
- 1
- standard
- 0.95
- survey
- 0.85
- empirical
- 0.8
- theoretical
- 0.65
- position
- 0.45
Status weights
- read
- 1
- reading
- 0.55
- queued
- 0.2
- parked
- 0.1
- rejected
- 0
A rejected sheet contributes nothing. A queued sheet contributes 0.2 of a read sheet's weight, so unread material opens a gap but cannot close one.
Store health
checking
This session
- ownership
- anonymous http-only cookie
- fingerprint
- a42f9dd1
The fingerprint is a one-way summary of your token, shown so you can recognise the session without the token ever being displayed. There are no accounts, so clearing cookies means losing access to these volumes — export first if they matter.
Abuse controls
- 40 writes per minute per session.
- 20 mounts per minute per session.
- 12 volume changes per five minutes per session.
Counters live in the database, not in process memory, so a cold start cannot reset them. This is best-effort on serverless: two concurrent requests for a brand-new bucket can both insert, so the count can briefly under-report. It reliably stops a single misbehaving client. A hard guarantee needs a hosted limiter in front of these routes.