Record schemas#

This page is generated from the code and must not be edited by hand.

The engine writes a small number of kinds of record. Each has a schema, a machine-readable description that two people could implement the record from independently. The schemas are generated from the engine’s own models and checked against them, so these pages cannot drift from what the engine writes.

Observations and interpretations never share a record: a fact the engine saw and a conclusion it drew are different records, so neither can be mistaken for the other.

Record types#

Record	What it is
`assertion`	A published edge.
`candidate`	A proposed edge, before the decoder has chosen a globally consistent subset.
`claim`	A statement the engine derived.
`component`	One independently comparable normalized value drawn from an entry.
`dictionary_binding`	What the Dictionary made of one component, held apart from the component.
`dictionary_entry`	What a string denotes when it appears in this data pool.
`entry`	One thing in the data pool, filesystem object or promoted derived stream.
`equivalence_level`	A named equivalence relation that rules and projections may evaluate at.
`finding`	What a flagging rule raised about one subject, for a person’s attention.
`journal_event`	One structured entry in the development build’s run journal.
`name_shape`	Where the parts of one entry’s name sat, held apart from what they were.
`observation`	Something read from storage.
`quality_figure`	One quality figure of one run: a count with its denominator, a duration, or a peak memory.
`record_event`	One human decision, appended and never rewritten.
`role`	A relationship an entity can play inside one interpretation.
`root`	One indexed root of a data pool.
`rule`	A named, versioned, toggleable, parameterised statement licensing one inference.
`rule_failure`	A rule unable to decide on one subject, recorded against the rule.
`run_record`	Everything needed to explain, compare and reproduce one run.
`scan_error`	An error encountered while reading, retained as data rather than raised.
`theme`	A materialised subject-matter group: everything sharing one value on one axis.
`theme_kind`	An axis along which the data pool can be explored.