mixle.semantics module

Minimal domain-neutral semantics for uncertain values and inference artifacts.

The contracts in this module describe what a quantity and result mean. They deliberately exclude inquiry graphs, domain laws, solvers, jobs, and storage. Operational metadata is serialized for audit but excluded from semantic identity, so moving an unchanged problem between backends cannot change it.

class CalibrationArtifact(id: 'str', target_identity: 'str', method: 'str', metrics: 'Mapping[str, float]', slice: 'Mapping[str, str]'=<factory>, schema_version: 'str' = '1.0.0')

Bases: object

Parameters:

• id (str)

• target_identity (str)

• method (str)

• metrics (Mapping[str, float])

• slice (Mapping[str, str])

• schema_version (str)

class CapabilityExtension(id: 'str', owner_project: 'str', input_schema_uri: 'str', output_schema_uri: 'str', maturity: 'str', schema_version: 'str' = '1.0.0')

Bases: object

Parameters:

• id (str)

• owner_project (str)

• input_schema_uri (str)

• output_schema_uri (str)

• maturity (str)

• schema_version (str)

class ConstraintSpec(lower: 'float | None' = None, upper: 'float | None' = None, lower_inclusive: 'bool' = True, upper_inclusive: 'bool' = True, allowed_values: 'tuple[str | int | float | bool, ...]' = (), schema_version: 'str' = '1.0.0')

Bases: object

Parameters:

• lower (float | None)

• upper (float | None)

• lower_inclusive (bool)

• upper_inclusive (bool)

• allowed_values (tuple[str | int | float | bool, ...])

• schema_version (str)

accepts(value)

Whether value (a scalar, or a non-empty vector of scalars) satisfies this constraint.

Admissibility is decided by explicit checks, never by IEEE comparison behavior: NaN used to be accepted by a bounded constraint because BOTH ordered comparisons against it are false, and an empty list was accepted because the validation loop ran zero times – a vacuous container certified as satisfying every bound. A bounded constraint also has no ordering to apply to a Boolean or a non-numeric value (bool only compares because it subclasses int), so those are rejected rather than silently ranked.

Parameters:

value (Any)

Return type:

bool

class DecisionArtifact(id: 'str', alternatives: 'tuple[str, ...]', selected: 'str', utility: 'Mapping[str, float]', posterior_identity: 'str', risk_measure: 'str', assumptions: 'tuple[str, ...]' = (), schema_version: 'str' = '1.0.0')

Bases: object

Parameters:

• id (str)

• alternatives (tuple[str, ...])

• selected (str)

• utility (Mapping[str, float])

• posterior_identity (str)

• risk_measure (str)

• assumptions (tuple[str, ...])

• schema_version (str)

class LikelihoodSpec(id: 'str', family: 'str', observation_ids: 'tuple[str, ...]', parameters: 'Mapping[str, Any]'=<factory>, discrepancy_ref: 'str | None' = None, schema_version: 'str' = '1.0.0')

Bases: object

Parameters:

• id (str)

• family (str)

• observation_ids (tuple[str, ...])

• parameters (Mapping[str, Any])

• discrepancy_ref (str | None)

• schema_version (str)

class ObservationSpec(id: 'str', value_spec_id: 'str', content_digest: 'str', data_ref: 'str', likelihood: 'LikelihoodSpec', measurement_uncertainty: 'float | None' = None, unit: 'str | None' = None, schema_version: 'str' = '1.0.0')

Bases: object

Parameters:

• id (str)

• value_spec_id (str)

• content_digest (str)

• data_ref (str)

• likelihood (LikelihoodSpec)

• measurement_uncertainty (float | None)

• unit (str | None)

• schema_version (str)

class PosteriorArtifact(id: 'str', values: 'tuple[ValueSpec, ...]', observations: 'tuple[ObservationSpec, ...]', likelihood: 'LikelihoodSpec', method: 'str', random_seed: 'int', summary: 'Mapping[str, Any]', uncertainty: 'tuple[UncertaintyComponent, ...]', sample_digest: 'str | None' = None, diagnostics: 'Mapping[str, Any]'=<factory>, sample_ref: 'str | None' = None, backend_id: 'str | None' = None, job_id: 'str | None' = None, schema_version: 'str' = '1.0.0')

Bases: object

Parameters:

• id (str)

• values (tuple[ValueSpec, ...])

• observations (tuple[ObservationSpec, ...])

• likelihood (LikelihoodSpec)

• method (str)

• random_seed (int)

• summary (Mapping[str, Any])

• uncertainty (tuple[UncertaintyComponent, ...])

• sample_digest (str | None)

• diagnostics (Mapping[str, Any])

• sample_ref (str | None)

• backend_id (str | None)

• job_id (str | None)

• schema_version (str)

property identity: str

The content digest fixed at construction – a durable record cannot drift into a new identity.

class PredictiveArtifact(id: 'str', posterior_identity: 'str', target_value_ids: 'tuple[str, ...]', content_digest: 'str', uncertainty: 'tuple[UncertaintyComponent, ...]', method: 'str', backend_id: 'str | None' = None, schema_version: 'str' = '1.0.0')

Bases: object

Parameters:

• id (str)

• posterior_identity (str)

• target_value_ids (tuple[str, ...])

• content_digest (str)

• uncertainty (tuple[UncertaintyComponent, ...])

• method (str)

• backend_id (str | None)

• schema_version (str)

class PriorSpec(id: 'str', family: 'str', parameters: 'Mapping[str, Any]', unit: 'str' = '1', source_ref: 'str | None' = None, schema_version: 'str' = '1.0.0')

Bases: object

Parameters:

• id (str)

• family (str)

• parameters (Mapping[str, Any])

• unit (str)

• source_ref (str | None)

• schema_version (str)

property identity: str

The content digest fixed at construction – a durable record cannot drift into a new identity.

class TraceEvent(trace_id: 'str', sequence: 'int', event_type: 'str', semantic_identity: 'str', payload: 'Mapping[str, Any]', occurred_at: 'str', schema_version: 'str' = '1.0.0')

Bases: object

Parameters:

• trace_id (str)

• sequence (int)

• event_type (str)

• semantic_identity (str)

• payload (Mapping[str, Any])

• occurred_at (str)

• schema_version (str)

class TraceSink(*args, **kwargs)

Bases: Protocol

class TransformKind(*values)

Bases: StrEnum

class TransformSpec(kind: 'TransformKind' = <TransformKind.IDENTITY: 'identity'>, scale: 'float' = 1.0, offset: 'float' = 0.0, lower: 'float' = 0.0, upper: 'float' = 1.0, schema_version: 'str' = '1.0.0')

Bases: object

Parameters:

• kind (TransformKind)

• scale (float)

• offset (float)

• lower (float)

• upper (float)

• schema_version (str)

log_abs_det_jacobian(value)

Log absolute derivative of forward at a natural-space value.

Parameters:

value (float)

Return type:

float

class UncertaintyComponent(id: 'str', kind: 'UncertaintyKind', measure: 'str', value: 'float | None' = None, artifact_digest: 'str | None' = None, unit: 'str' = '1', schema_version: 'str' = '1.0.0')

Bases: object

Parameters:

• id (str)

• kind (UncertaintyKind)

• measure (str)

• value (float | None)

• artifact_digest (str | None)

• unit (str)

• schema_version (str)

class UncertaintyKind(*values)

Bases: StrEnum

class ValueRole(*values)

Bases: StrEnum

class ValueSpec(id: 'str', role: 'ValueRole', unit: 'str', dtype: 'str' = 'float64', shape: 'tuple[int, ...]'=(), transform: 'TransformSpec' = <factory>, constraint: 'ConstraintSpec | None' = None, prior: 'PriorSpec | None' = None, value: 'Any' = None, expression: 'str | None' = None, dependencies: 'tuple[str, ...]'=(), description: 'str' = '', schema_version: 'str' = '1.0.0')

Bases: object

Parameters:

• id (str)

• role (ValueRole)

• unit (str)

• dtype (str)

• shape (tuple[int, ...])

• transform (TransformSpec)

• constraint (ConstraintSpec | None)

• prior (PriorSpec | None)

• value (Any)

• expression (str | None)

• dependencies (tuple[str, ...])

• description (str)

• schema_version (str)

property identity: str

The content digest fixed at construction – a durable record cannot drift into a new identity.

load_reference_fixture()

Load the packaged cross-project value/prior/observation fixture.

Return type:

dict[str, Any]