Analysis, posture & rules

Statistics, estate posture and business-rule reconciliation.

AnalysisService runs statistical methods on data you send: forecasting, lag features, clustering, causal intervention and descriptive statistics.
PostureService serves the data-maturity picture that a fetch produced, maps the connected estate into a domain graph, and runs causal interventions over estate columns.
RulesService serves the business rules the reconciliation agent derived and stores the operator’s enforcement policy for each.

Summary

Service	RPC	Kind	Purpose
AnalysisService	AnalyzeTimeSeries	Unary	Backtest-selected forecast, stationarity and anomalies
AnalysisService	AnalyzeFeature	Unary	Significant autocorrelation lags
AnalysisService	ClusterRefinement	Unary	Density clustering of embedding vectors
AnalysisService	AnalyzeCausal	Unary	Do-intervention over a graph you state
AnalysisService	InferBayesian	Unary	Bayesian posterior (no network defined)
AnalysisService	AnalyzeDescriptiveStats	Unary	Mean, median, variance, skewness, kurtosis
AnalysisService	ClassifyESGRelevance	Unary	ESG relevance score for text
PostureService	GetHealthScore	Unary	Domain health, assessments, measures and contexts
PostureService	BuildDomainGraph	Unary	Map the connected estate into domains
PostureService	AnalyzeEstateCausal	Unary	Do-intervention over estate columns
PostureService	ListIncidents	Unary	Data-maturity assessments
PostureService	GetContext	Unary	Semantic context clusters
RulesService	ListRules	Unary	Derived business rules, with policies
RulesService	AnalyzeCode	Unary	Rules from the latest snapshot or from source code
RulesService	CheckCode	Unary	Model review of a code snippet
RulesService	UpdateRulePolicy	Unary	Set a rule’s enforcement policy

AnalysisService

Full name semantics.v1.AnalysisService.

These RPCs compute on the values in the request. They read no data source. Some errors are returned without a specific gRPC code, and clients see UNKNOWN for those.

AnalyzeTimeSeries

rpc AnalyzeTimeSeries(AnalyzeTimeSeriesRequest) returns (AnalyzeTimeSeriesResponse);

Kind: Unary.
Auth: Bearer session.
Errors: None. Too little data is reported in the response.

Runs an exploratory time-series analysis over data. The server chooses a Prophet decomposition or an ARIMA model by backtest, applies an ADF stationarity test (differencing a non-stationary series before modelling), and forecasts forecast_horizon steps. If no model fits, it falls back to linear regression and says so in recommended_model and anomalies. With 5 or fewer points, the response has only recommended_model: "Insufficient Data (Need > 5 points)".

Request: AnalyzeTimeSeriesRequest

Field	Type	Description
`data`	`repeated double`	The series, in time order. More than 5 points are needed.
`forecast_horizon`	`int32`	Steps to forecast. 0 or negative means 3.
`frequency`	`string`	Series frequency label. Not used by the server.
`domain`	`string`	Domain label. Not used by the server.
`telemetry_source`	`string`	Source label. Not used by the server.

Response: AnalyzeTimeSeriesResponse

Field	Type	Description
`forecast`	`repeated double`	Forecast values for the next `forecast_horizon` steps.
`anomalies`	`repeated string`	Findings and anomalies, as sentences.
`stationarity_test`	`string`	ADF result, for example `ADF: stationary (stat=… < crit5=…)`, or `ADF: inconclusive`. Empty on the fallback.
`recommended_model`	`string`	The chosen model, for example `ARIMA(1,1,0) [backtest-selected]`, or `Linear Regression (Fallback)`.

AnalyzeFeature

rpc AnalyzeFeature(AnalyzeFeatureRequest) returns (AnalyzeFeatureResponse);

Kind: Unary.
Auth: Bearer session.
Errors: None.

Computes the autocorrelation of target at lags 1 to 5 and reports each lag with |ACF| > 0.5 as significant, with a target_lag_<k> feature for it. Fewer than 10 points returns generated_features: ["Insufficient data"], and a constant series returns ["Constant target"].

Request: AnalyzeFeatureRequest

Field	Type	Description
`target`	`repeated double`	The target series. At least 10 points.
`external_vars`	`repeated double`	Not used by the server.

Response: AnalyzeFeatureResponse

Field	Type	Description
`significant_lags`	`repeated string`	For example `Lag-2 (ACF: 0.63)`.
`generated_features`	`repeated string`	Feature names such as `target_lag_2`, or a sentence when there are none.

ClusterRefinement

rpc ClusterRefinement(ClusterRefinementRequest) returns (ClusterRefinementResponse);

Kind: Unary.
Auth: Bearer session.
Errors: Returned without a specific code (clients see UNKNOWN) when vectors is empty, no vector has any dimensions, or clustering fails.

Clusters vectors with HDBSCAN (minimum cluster size 3, Euclidean distance), choosing parameters by silhouette. When the deployment has a re-ranking model, the clusters are refined with it, using BM25 scores of items against data_metadata (or telemetry_source) when items lines up with vectors. The dimension is taken from the first non-empty vector. Vectors of any other dimension are labelled noise (-1).

Request: ClusterRefinementRequest

Field	Type	Description
`items`	`repeated string`	Text items, one per vector. Used for BM25 scoring when the counts match.
`granularity`	`string`	`PARAGRAPH` or `COLUMN`. Not used by the server.
`vectors`	`repeated` `Embedding`	Required. The vectors to cluster.
`domain`	`string`	Context for re-ranking.
`access_level`	`string`	Context for re-ranking.
`telemetry_source`	`string`	Context for re-ranking, and the BM25 query when `data_metadata` is empty.
`data_metadata`	`string`	Context for re-ranking, and the BM25 query.
`business_unit`	`string`	Not used by the server.

Response: ClusterRefinementResponse

Field	Type	Description
`clusters`	`repeated int32`	A cluster label per input vector, in input order. `-1` is noise.
`noise`	`repeated int32`	Indices of the vectors labelled noise.
`health_score`	`double`	Cluster quality: the silhouette score, or the re-ranker’s cluster health when refinement ran.

AnalyzeCausal

rpc AnalyzeCausal(AnalyzeCausalRequest) returns (AnalyzeCausalResponse);

Kind: Unary.
Auth: Bearer session.
Errors: INVALID_ARGUMENT if an edge is refused: an endpoint missing from nodes, or an edge that would create a cycle. An intervention failure is returned without a specific code.

Builds the graph exactly as you state it and sets intervention_node to intervention_value (the do-operator). The change propagates along edges by multiplying by each effect_size. With no edges, nothing propagates and the effect is 0. That means “no structure was given”, not “no effect was measured”. For a graph built from the connected estate, use AnalyzeEstateCausal.

Request: AnalyzeCausalRequest

Field	Type	Description
`nodes`	`map<string, double>`	Node name to observed value.
`intervention_node`	`string`	The node to set.
`intervention_value`	`double`	The value to set it to.
`target_node`	`string`	The node whose change to report.
`edges`	`repeated` `CausalEdge`	The graph’s edges. Both endpoints must be in `nodes`, and the graph must stay acyclic.

Response: AnalyzeCausalResponse

Field	Type	Description
`effect`	`double`	Predicted change in `target_node`. 0 when the target is not reached or not named.
`new_target_value`	`double`	The target’s value after the intervention.

InferBayesian

rpc InferBayesian(InferBayesianRequest) returns (InferBayesianResponse);

Kind: Unary.
Auth: Bearer session.
Errors: FAILED_PRECONDITION on every call in the current release. No Bayesian network is defined on the instance, and the request cannot carry one.

Request: InferBayesianRequest

Field	Type	Description
`evidence`	`map<string, string>`	Variable to observed state.
`target_node`	`string`	The variable whose posterior to return.

Response: InferBayesianResponse

Field	Type	Description
`probabilities`	`map<string, double>`	State to posterior probability. Not returned in the current release.

AnalyzeDescriptiveStats

rpc AnalyzeDescriptiveStats(AnalyzeDescriptiveStatsRequest) returns (AnalyzeDescriptiveStatsResponse);

Kind: Unary.
Auth: Bearer session.
Errors: None. Empty data returns an empty response, with every field 0.

Request: AnalyzeDescriptiveStatsRequest

Field	Type	Description
`data`	`repeated double`	The values.

Response: AnalyzeDescriptiveStatsResponse

Field	Type	Description
`mean`	`double`	Mean.
`median`	`double`	Median.
`variance`	`double`	Variance.
`std_dev`	`double`	Standard deviation.
`skewness`	`double`	Skewness.
`kurtosis`	`double`	Kurtosis.

ClassifyESGRelevance

rpc ClassifyESGRelevance(ClassifyESGRelevanceRequest) returns (ClassifyESGRelevanceResponse);

Kind: Unary.
Auth: Bearer session.
Errors: Returned without a specific code if classification fails.

Scores the texts together for ESG relevance with a keyword classifier. No model is involved. The texts are also passed to the deployment’s knowledge-ingestion pipeline when one is configured.

Request: ClassifyESGRelevanceRequest

Field	Type	Description
`texts`	`repeated string`	The texts to score, as one group.

Response: ClassifyESGRelevanceResponse

Field	Type	Description
`relevance_score`	`double`	ESG relevance of the group.

PostureService

Full name semantics.v1.PostureService.

Posture comes from a fetch. After a fetch, GetHealthScore serves that run’s snapshot. Without a snapshot, the posture agent runs on the model plane. For how a fetch is started, see Fetch & traces. Long posture runs are visible live through ActivityService.

GetHealthScore

rpc GetHealthScore(GetHealthScoreRequest) returns (GetHealthScoreResponse);

Kind: Unary. It can take minutes when no snapshot exists and the posture agent has to run.
Auth: Bearer session.
Errors: FAILED_PRECONDITION if the instance has no data sources. UNAVAILABLE if the posture analysis cannot run. The server reports these rather than returning a grade it did not compute.

Returns the latest fetch’s posture snapshot, or runs the posture agent when there is none. Before it is returned, readiness figures are normalised, edges are cleaned, each assessment is assigned a domain, and each semantic context is grounded in the tenant’s knowledge corpus where one exists. graph_nodes and graph_edges are empty on this RPC.

Request: GetHealthScoreRequest

No fields.

Response: GetHealthScoreResponse

Also returned by BuildDomainGraph.

Field	Type	Description
`domains`	`map<string,` `DomainHealth>`	Health per business domain.
`assessments`	`repeated` `DataMaturityAssessment`	Data-maturity findings.
`sops`	`repeated` `OptimizationSOP`	Recommended measures.
`contexts`	`repeated` `SemanticContextItem`	Semantic context clusters.
`graph_nodes`	`repeated` `KnowledgeGraphNode`	The estate graph’s nodes. Set only by `BuildDomainGraph` with `include_connected_sources`.
`graph_edges`	`repeated` `KnowledgeGraphEdge`	The estate graph’s edges. Each edge’s evidence is in `relation`.

BuildDomainGraph

rpc BuildDomainGraph(BuildDomainGraphRequest) returns (GetHealthScoreResponse);

Kind: Unary.
Auth: Bearer session.
Errors: FAILED_PRECONDITION when include_connected_sources is false (there are no instance data files to map), when no data sources are configured, or when the server has no metadata store. INTERNAL if the configured connections cannot be listed.

With include_connected_sources: true, the server asks every configured connection for its catalogue (bounded at 30 seconds per source, concurrently), profiles columns, samples values except where excluded, and groups the discovered datasets into domains. A connection that refused, has no connector, or holds nothing still appears as a source node with its reason.

Sampled values are used only to infer structure. A sampled value never reaches the graph, a log line, an error, a lineage record or a model prompt: the graph carries counts and shapes, never a cell.

Request: BuildDomainGraphRequest

Field	Type	Description
`include_connected_sources`	`bool`	Deprecated. Ask the connected sources. The cockpit’s estate map now uses the platform’s shared estate service. This field is still served, and without it the RPC refuses.
`metadata_only_connection_ids`	`repeated string`	Connections to read as catalogue only, with no rows sampled. Sampling is on by default for every other connection.

Response

GetHealthScoreResponse with graph_nodes and graph_edges set.

AnalyzeEstateCausal

rpc AnalyzeEstateCausal(AnalyzeEstateCausalRequest) returns (AnalyzeEstateCausalResponse);

Kind: Unary. It builds the estate graph first, so it has the same cost as BuildDomainGraph.
Auth: Bearer session.
Errors: The same preconditions as BuildDomainGraph. INVALID_ARGUMENT if the coefficients cannot be wired, for example because an effect_size is not finite.

Runs a do-intervention over estate causal nodes: columns of connected datasets that came back with a measured value. The structure comes from the coefficients you assert. The estate proposes candidate pairs but never supplies a coefficient. Without at least one admitted edge, the RPC refuses (answered: false) rather than returning an effect of 0. Read answered first, and show basis_statement beside any number.

Request: AnalyzeEstateCausalRequest

Field	Type	Description
`metadata_only_connection_ids`	`repeated string`	As in `BuildDomainGraphRequest`. An excluded connection contributes no causal node.
`intervention_node`	`string`	Estate causal node to set: `<qualified dataset>.<column>`.
`intervention_value`	`double`	Value to set it to.
`target_node`	`string`	Estate causal node whose change to report.
`coefficients`	`repeated` `EstateCausalCoefficient`	Your structural coefficients. At least one must be admitted.

Response: AnalyzeEstateCausalResponse

Field	Type	Description
`answered`	`bool`	`false` when nothing was propagated. Read this first.
`refusal`	`string`	Why it declined, in full. `effect` and `new_target_value` are 0 and meaningless then.
`effect`	`double`	Predicted change in the target.
`new_target_value`	`double`	Target value after the intervention.
`weakest_basis`	`string`	The weakest basis among the edges used: `OPERATOR_ASSERTED` or `DECLARED_KEY_REPURPOSED`.
`basis_statement`	`string`	What that basis means. Show it verbatim beside the number.
`edges`	`repeated string`	Every edge that entered the graph, basis first.
`candidate_pairs`	`repeated string`	Every pair the estate proposed, priced or not.
`node_refusals`	`repeated string`	Columns that were not made nodes, with reasons. Structure only, no figures.
`edge_refusals`	`repeated string`	Coefficients that were not admitted, with reasons.
`nodes_added`	`int32`	Causal nodes built.
`columns_refused`	`int32`	Columns refused a node. Read `nodes_added` against it.

ListIncidents

rpc ListIncidents(ListIncidentsRequest) returns (ListIncidentsResponse);

Kind: Unary. It runs the posture agent.
Auth: Bearer session.
Errors: None. If the posture analysis cannot run, the list is empty. An empty list therefore does not mean “no incidents”.

Returns the posture agent’s data-maturity assessments. Entries that repeat the agent template’s example are dropped. domain is not set on this RPC.

Request: ListIncidentsRequest

No fields.

Response: ListIncidentsResponse

Field	Type	Description
`assessments`	`repeated` `DataMaturityAssessment`	The assessments.

GetContext

rpc GetContext(GetContextRequest) returns (GetContextResponse);

Kind: Unary. It runs the posture agent.
Auth: Bearer session.
Errors: None. If the posture analysis cannot run, it returns context_json: "{}" and no items.

Request: GetContextRequest

Field	Type	Description
`agent_type`	`string`	Not used by the server.

Response: GetContextResponse

Field	Type	Description
`context_json`	`string`	The semantic contexts as JSON.
`items`	`repeated` `SemanticContextItem`	The same contexts, typed.

RulesService

Full name semantics.v1.RulesService.

A rule has two halves with different authors. Fields such as name, logic_plain_english, proposed_code and reconciliation_status are findings of the reconciliation agent, and nothing on this API can edit them. policy is the operator’s decision about the rule. It is written through UpdateRulePolicy and survives the next fetch.

ListRules

rpc ListRules(ListRulesRequest) returns (ListRulesResponse);

Kind: Unary.
Auth: Bearer session.
Errors: Returned without a specific code if the rules agent cannot run (for example, no model plane).

Derives rules from the latest fetch session’s analysis, using a cached result when one exists, then merges in each rule’s stored policy. With no fetch session yet, it returns an empty list.

Request: ListRulesRequest

Field	Type	Description
`domain_filter`	`string`	Keep only rules whose `category` equals this, case-insensitively. Empty returns all.

Response: ListRulesResponse

Field	Type	Description
`rules`	`repeated` `Rule`	The rules.

AnalyzeCode

rpc AnalyzeCode(AnalyzeCodeRequest) returns (AnalyzeCodeResponse);

Kind: Unary.
Auth: Bearer session.
Errors: FAILED_PRECONDITION when no rules snapshot exists and source_code is empty. Model, template and parse failures are returned without a specific code.

If a fetch has written a rules snapshot, returns those rules and ignores the request. Otherwise the reconciliation agent derives rules from source_code.

Request: AnalyzeCodeRequest

Field	Type	Description
`source_path`	`string`	Not used by the server.
`source_code`	`string`	Code to derive rules from, used only when no snapshot exists.

Response: AnalyzeCodeResponse

Field	Type	Description
`found_rules`	`repeated` `Rule`	The rules.

CheckCode

rpc CheckCode(CheckCodeRequest) returns (CheckCodeResponse);

Kind: Unary.
Auth: Bearer session.
Errors: PERMISSION_DENIED if the snippet trips the rules agent’s guardrail.

Asks the model plane to review the snippet. valid reports whether the model answered, not whether the code is valid. Read report.

Request: CheckCodeRequest

Field	Type	Description
`source_code`	`string`	The code.
`language`	`string`	Its language.

Response: CheckCodeResponse

Field	Type	Description
`valid`	`bool`	`true` when the model answered. Not a verdict on the code.
`issues`	`repeated string`	Always empty.
`report`	`string`	The model’s review, or the error when the call failed.

UpdateRulePolicy

rpc UpdateRulePolicy(UpdateRulePolicyRequest) returns (UpdateRulePolicyResponse);

Kind: Unary.
Auth: Bearer session. The author is the verified session subject.
Errors: INVALID_ARGUMENT if rule_id or policy is missing. UNAUTHENTICATED if the session has no subject. FAILED_PRECONDITION if no rules have been derived yet. NOT_FOUND if rule_id is not in the current rules. INTERNAL if the policy cannot be stored.

Stores the operator’s enforcement policy for an existing rule. It does not create rules. updated_by and updated_at_unix are set by the server.

Request: UpdateRulePolicyRequest

Field	Type	Description
`rule_id`	`string`	Required. An existing rule’s `id`.
`policy`	`RulePolicy`	Required. Author fields are ignored.

Response: UpdateRulePolicyResponse

Field	Type	Description
`policy`	`RulePolicy`	The stored policy, with the server-set author and time.

Messages

Embedding

Field	Type	Description
`vector`	`repeated float`	One embedding vector.

CausalEdge

Field	Type	Description
`from`	`string`	Cause node.
`to`	`string`	Effect node.
`effect_size`	`double`	Structural coefficient: a unit change in `from` moves `to` by this much. An edge at 0 is a declared non-relationship; a missing edge is an unknown one.

EstateCausalCoefficient

Field	Type	Description
`from`	`string`	Estate causal node, `<qualified dataset>.<column>`, in the source’s own spelling. It must have a measured value.
`to`	`string`	Estate causal node.
`effect_size`	`double`	Your assertion of the coefficient. Must be finite.

DomainHealth

Field	Type	Description
`status`	`string`	`Stable`, `At Risk` or `Critical`.
`grade`	`string`	Letter grade, for example `A`, `B` or `C`.
`quality_stability_pct`	`double`	Quality stability percentage.
`risk_pii_pct`	`double`	PII risk percentage.
`lineage_pct`	`double`	Lineage coverage percentage.
`freshness_pct`	`double`	Freshness percentage.
`policy_pct`	`double`	Policy coverage percentage.
`cost_efficiency_pct`	`double`	Cost efficiency percentage.
`issues`	`string`	Issues, as text.
`uses_and_alignment`	`repeated` `MaturityIndicator`	Maturity indicators.
`data_and_architecture`	`repeated` `MaturityIndicator`	Maturity indicators.
`analysis_and_predictive`	`repeated` `MaturityIndicator`	Maturity indicators.
`tools_and_automation`	`repeated` `MaturityIndicator`	Maturity indicators.
`skills_and_fluency`	`repeated` `MaturityIndicator`	Maturity indicators.
`spatial_logistics`	`repeated` `MaturityIndicator`	Maturity indicators.
`empty_trip`	`repeated` `MaturityIndicator`	Always empty, on purpose. Nothing in the platform measures empty-trip running, so the server clears this field on every response. Display it as not measured.
`shadow_search_ingestion`	`repeated` `MaturityIndicator`	Maturity indicators.

MaturityIndicator

Field	Type	Description
`indicator_name`	`string`	Indicator name.
`rating`	`int32`	1 to 5.
`metric`	`string`	The metric.
`description`	`string`	Description.

DataMaturityAssessment

Field	Type	Description
`incident`	`string`	The finding.
`severity`	`string`	Severity.
`impact`	`string`	Impact.
`root_cause`	`string`	Root cause.
`blast_radius`	`string`	Blast radius.
`fix`	`string`	Suggested fix.
`domain`	`string`	The business domain it affects, matched against the fetch’s derived domains.

OptimizationSOP

Field	Type	Description
`type`	`string`	Measure type.
`title`	`string`	Title.
`action`	`string`	The action.
`reason_impact`	`string`	Why, and the expected impact.
`button_label`	`string`	Suggested action label.
`provenance`	`MeasureProvenance`	Where the wording came from. Only `MEASURE_PROVENANCE_MODEL_GENERATED` may be labelled AI-generated.

SemanticContextItem

Field	Type	Description
`cluster_name`	`string`	Context cluster name.
`attributes`	`string`	Attributes.
`relationships`	`string`	Relationships.
`database_objects`	`repeated string`	Database objects in the cluster.
`relevant_documents`	`repeated string`	Relevant documents.
`quality_pillars`	`repeated` `DataQualityPillar`	Data-quality checks.
`risk_checks`	`repeated` `RiskComplianceCheck`	Risk and compliance checks.

DataQualityPillar

Field	Type	Description
`name`	`string`	Pillar name.
`status`	`string`	`Pass`, `Warning` or `Fail`.
`description`	`string`	Description.

RiskComplianceCheck

Field	Type	Description
`check_name`	`string`	Check name.
`status`	`string`	`Pass`, `Warning` or `Fail`.
`details`	`string`	Details.

KnowledgeGraphNode

Field	Type	Description
`id`	`string`	Node id.
`label`	`string`	Display label.
`properties`	`map<string, string>`	Node properties.
`domain`	`string`	Business domain.
`access_level`	`string`	Access level.
`telemetry_source`	`string`	Telemetry source.
`data_metadata`	`string`	Metadata.
`business_unit`	`string`	Business unit.

KnowledgeGraphEdge

Field	Type	Description
`source`	`string`	Source node id.
`target`	`string`	Target node id.
`relation`	`string`	The edge’s evidence, as `KIND` or `KIND: detail`. `DECLARED_FOREIGN_KEY` is the source’s own constraint. `INFERRED_` kinds are FACE’s measurements. `GUESSED_` kinds are guesses and say so. Display the three differently.

Rule

Field	Type	Description
`id`	`string`	Rule id.
`name`	`string`	Rule name.
`logic_plain_english`	`string`	The rule in plain English.
`source_document`	`string`	Document the rule was found in.
`category`	`string`	For example `Logistics` or `Finance`.
`severity`	`string`	`Critical`, `High`, `Medium` or `Low`.
`maturity`	`string`	`High`, `Medium` or `Low`.
`inferred_objective`	`string`	The rule’s inferred objective.
`related_documents`	`repeated string`	Related documents.
`proposed_code`	`string`	Proposed implementation.
`rule_type`	`string`	Rule type.
`reconciliation_status`	`string`	Code-versus-policy verdict: `Aligned`, `Drift`, `Shadow` or `Missing`. Empty means nobody has judged it, which is not `Aligned`.
`implementation_file`	`string`	Where the rule is implemented in code.
`policy_document`	`string`	The governing policy document the agent grounded the rule in.
`reconciliation_confidence`	`optional double`	Unset on every response. Nothing measures a confidence for the verdict. Read `reconciliation_status`, and display an absent value as “not measured”.
`policy`	`RulePolicy`	The operator’s enforcement policy.

RulePolicy

Field	Type	Description
`enforcement_action`	`EnforcementAction`	What happens on a breach.
`tolerance`	`string`	Tolerance in the rule’s own units, as typed.
`policy_reference`	`string`	The clause the operator cites, for example `ISO-42001 §9.4`.
`updated_by`	`string`	Server-set. The verified subject that last wrote it.
`updated_at_unix`	`int64`	Server-set. 0 means never written.
`rationale`	`string`	Free-text justification.

Enums

EnforcementAction

Value	Meaning
`ENFORCEMENT_ACTION_UNSPECIFIED`	Not set.
`ENFORCEMENT_ACTION_AUDIT_ONLY`	Observe and record. Nothing is blocked.
`ENFORCEMENT_ACTION_REVIEW_REQUIRED`	Flag the breach for a human before the action proceeds.
`ENFORCEMENT_ACTION_BLOCK`	Refuse the action outright.

MeasureProvenance

Value	Meaning
`MEASURE_PROVENANCE_UNSPECIFIED`	Written before this field existed. Says nothing.
`MEASURE_PROVENANCE_MEASURED`	Read verbatim from a source the fetch pulled.
`MEASURE_PROVENANCE_DERIVED`	Computed deterministically from measured data.
`MEASURE_PROVENANCE_MODEL_GENERATED`	Written by a model.

Twins, simulation & claims Compliance, finance & operations