Module model 

The typed records of the economics subsystem: the declarations it reads (methodology, suites, tasks), the raw facts it records (runs), and the answer it derives (summary).

Raw facts and derived answers are different types on purpose. A run holds what the provider reported and what the success gates found, and nothing computed from them; a pair, a metric and a summary are recomputed from runs on every read, so a change to how a number is derived never needs the runs rewritten.

Every public type is prefixed Economics: the OpenAPI document keeps one flat namespace of schema names, and Token* and Evidence* already belong to other subsystems.

use majordomus_cli::economics::model::EconomicsClass;
assert!(EconomicsClass::Observed.is_measurement());
assert!(!EconomicsClass::Counterfactual.is_measurement());
assert_eq!(EconomicsClass::Derived.weakest(EconomicsClass::Estimated), EconomicsClass::Estimated);

Structs

EconomicsCheck

One success gate’s verdict on a run.

EconomicsClaimBinding

A published claim bound to the metric that must support it.

EconomicsClassDefinition

One measurement class as the methodology defines it.

EconomicsContextRun

One recorded measurement of the context suite: every seed the compiler compiled, counted, and every seed it refused, named with its reason rather than dropped.

EconomicsContextSeed

Token counts of one context-compilation seed.

EconomicsContextView

The context suite’s latest measurement, in brief.

EconomicsDistribution

The shape of a set of per-pair values.

EconomicsExclusion

One run excluded under the outlier rule.

EconomicsExplainInput

The input of economics.explain: the id of the one metric to explain. Nothing else is accepted, so a misspelt field is refused rather than silently ignored.

EconomicsExplanation

The answer of economics.explain: one metric, and everything it rests on.

EconomicsGateDefinition

One gate a run must pass to count as a success.

EconomicsHarness

The harness that ran a session, as it identified itself.

EconomicsHistoryPoint

A recorded measurement over time: one entry per recorded context run.

EconomicsHypothesis

A statement the benchmark exists to test. Never a result.

EconomicsInterval

A bootstrap interval, with everything needed to reproduce it.

EconomicsMethodology

.ai/repo/benchmarks/economics/methodology.yaml: what is measured, how, and when a result may be published.

EconomicsMethodologyChange

One entry of the methodology’s change log.

EconomicsMetric

One metric: its value, where it comes from, and how much it can bear.

EconomicsModelTotal

A model’s usage over a whole session, as the harness totalled it (modelUsage), including models the harness called on its own account.

EconomicsOrientation

What a session did before its first edit: the cost of getting oriented.

EconomicsOutcome

Whether the run did the task.

EconomicsOutlierPolicy

The rule for excluding runs, and the runs it excluded.

EconomicsPair

A control run and its treatment run, and what follows from them.

EconomicsPairCounts

Pair counts of a suite, so that nothing is dropped out of sight.

EconomicsPairing

How runs are paired and when a pair is comparable and valid.

EconomicsPublicationPolicy

When a quantitative total-token claim may be published. Every threshold must hold.

EconomicsQuery

The input of economics.summary: optional narrowing of the pairs and metrics. The verdict is always about all the evidence, whatever is narrowed, and a narrowed metric is never verified: the publication rule was evaluated over all of it, not over a slice.

EconomicsRequest

The token usage of one provider request, provider-native field names kept.

EconomicsRevision

The repository revision a run was recorded at.

EconomicsRun

One recorded run: a task, one variant, one repetition, every session of it. Raw facts only; nothing here is derived from another field.

EconomicsRunList

The answer of economics.runs: the raw facts every metric is computed from.

EconomicsRunView

One run in brief, with its derived totals.

EconomicsRunsQuery

The input of economics.runs: optional filters on suite, task and variant, all of which must match. An empty query lists every recorded run, and an unknown field is refused.

EconomicsSegment

A metric’s value within one segment of the pairs.

EconomicsSession

One session of a run: one harness invocation on one prompt, with every provider request it made, the harness’s own totals and the tools it called. A count the harness did not report is absent rather than zero.

EconomicsStatisticsPolicy

How per-pair results are aggregated into a headline value and its interval. Every parameter of the bootstrap is declared here, its generator seed included, so the same runs always give the same interval.

EconomicsSuite

A suite, as .ai/repo/benchmarks/economics/suites/<id>.yaml declares it: what is run, with which model and arms, and which tracked paths decide whether its evidence is still current. A live suite fills the session fields and a context suite the seed and tokenizer fields; the other kind’s fields stay absent and are not written back.

EconomicsSuiteView

A suite and the state of its evidence.

EconomicsSummary

The answer of economics.summary: every metric, where each comes from, and the one statement the evidence allows.

EconomicsTask

A task, as .ai/repo/benchmarks/economics/tasks/<id>/task.yaml declares it: the fixture a run starts from, the prompt of each session in order, and the hidden acceptance tests and verify command that decide whether the run succeeded.

EconomicsTaskSession

One session of a task: the prompt file it starts from.

EconomicsTokenizer

The identity of a tokenizer, recorded with every count it produced.

EconomicsUsage

A run’s totals, derived from its raw usage.

EconomicsVariant

One arm of the experiment: what the session runs with, and what it runs without.

EconomicsVerdict

The one sentence the evidence allows, and why it allows no stronger one.

Enums

EconomicsClass

How a number came to be. The classes are ordered from strongest to weakest, and a value computed from several inputs is never stronger than the weakest of them.

EconomicsFreshness

How recorded evidence stands against the repository as it is now.

EconomicsMetricStatus

Where a metric stands against the publication rule: whether its value may back a claim, is a result too thin to claim anything, or was never measured at all.

EconomicsPairStatus

Why a pair does or does not count.