Table of Contents

AsiBackbone API Glossary

This page is the implementation-side glossary for the AsiBackbone package family. It maps the architecture vocabulary taught in ASI Backbone Learning to concrete AsiBackbone.* APIs and product behavior.

For canonical educational definitions, terminology lineage, and comparisons with established architecture concepts, use the ASI Backbone Learning Architecture Glossary and Terminology and Established Architecture Concepts.

Important

Learning owns the teaching definition of the architecture concepts. This repository owns exact type names, namespaces, API contracts, package-specific semantics, implementation invariants, and released runtime behavior. Learning does not override the product contract documented here or in the generated API reference.

Terms such as responsibility handshake or acknowledgment describe accountability workflows in the software model. They do not create legal protection, legal advice, regulatory compliance, or a substitute for organizational review.

Concept-to-API mapping

Learning concept Current AsiBackbone API / product mapping Product-specific semantics
Policy decision pipeline IGovernancePolicyEvaluator<TContext>, IGovernanceConstraint<TContext>, GovernanceDecision, audit and capability surfaces “Governance spine” remains architectural positioning; there is no single GovernanceSpine type.
Intent / request GovernanceEvaluationContext carries proposed operation data The product does not require one universal Intent base type. Hosts map application-specific proposal data into the evaluation context.
Actor context IGovernanceActorContext Core remains host-neutral. Authentication stays host-owned; integrations adapt authoritative host identity into the Core abstraction.
Policy context IGovernanceEvaluationContext, GovernanceEvaluationContext The context is the product evaluation input. Hosts remain responsible for supplying trustworthy actor, operation, resource, region, risk, and policy metadata.
Constraint IGovernanceConstraint<TContext>, ConstraintEvaluationResult Constraints contribute decision information; they do not own the governed side effect.
Policy evaluation IGovernancePolicyEvaluator<TContext> Evaluation composes constraint results into a product decision and does not execute the protected operation.
Decision policy IGovernanceDecisionPolicy<TContext> Optional host policy can reshape or raise the composed decision after constraint evaluation.
Decision outcome GovernanceDecision, GovernanceDecisionOutcome The released enum includes Allowed, Warning, Denied, Deferred, AcknowledgmentRequired, and EscalationRecommended. A decision is not execution authority.
Acknowledgment AcknowledgmentRequest, AcknowledgmentResponse; ASP.NET Core challenge support includes AcknowledgmentChallenge Acknowledgment records acceptance of a challenge or responsibility statement. It is not authentication, authorization, legal protection, or an automatic execution grant.
Decision receipt DecisionReceipt The type carries structured decision evidence. Durability, retention, signing, integrity protection, and backend storage remain separate concerns.
Audit ledger AuditLedgerRecord, IGovernanceAuditLedgerStore The product exposes storage-ready records and contracts; concrete persistence behavior depends on the selected provider and host configuration.
Decision receipt sink IDecisionReceiptSink A sink receives decision receipt. It does not by itself imply durable or tamper-evident storage.
Scoped capability / capability grant CapabilityGrant, CapabilityGrantValidator The product models bounded execution authority through grant data and validation. A token format alone does not create least privilege.
Host-owned execution Host-Owned Execution Enforcement and host/gateway code There is intentionally no universal executor owned by Core. The host retains control of the real side effect.
Operational gateway AI Agent Gateway Scenario, Robotics Operational Gateway A gateway is an integration pattern, not one mandatory base type. It validates current decision/authority state before external execution.
Policy version GovernanceDecision PolicyVersion is the product's readable policy-generation label.
Policy fingerprint GovernanceDecision PolicyHash is the current product field used for the effective-policy fingerprint.
Reason codes GovernanceDecision, ConstraintEvaluationResult Reason codes are machine-readable product explanations and should remain safe to persist or expose according to host policy.
Correlation ID GovernanceHttpRequestCorrelation plus correlation fields on governance records Correlation connects request, decision, audit, lifecycle, and telemetry records; it is not an authorization credential.
Operation result OperationResult Product operation success/failure is deliberately separate from governance outcome. A policy denial is not the same thing as infrastructure failure.

Implementation-specific additions

Some public product concepts are intentionally more detailed than the smallest Learning examples.

Warning outcome

The current GovernanceDecisionOutcome enum includes Warning. A warning permits continuation through the governance layer while retaining warning reasons. Learning may use a five-outcome teaching model when that keeps a tutorial focused; that simplification does not change the released enum.

Audit lifecycle, outbox, emission, and signing

The package family includes implementation surfaces for audit lifecycle records, durable outbox handling, governance emission, signing, and verification. These are product/runtime concerns documented in this repository. Learning may teach the architectural reason for those boundaries, but exact provider behavior, configuration, retries, storage semantics, and cryptographic posture remain product-owned.

Product invariants

The following statements are implementation boundaries, not replacement teaching definitions:

  • An Allowed decision means the governance layer permits continuation; the host still owns whether and how execution occurs.
  • Acknowledgment does not grant access-control permission and does not automatically become execution authority.
  • Decision receipt is structured evidence, not a claim of durability, immutability, or tamper evidence by itself.
  • Capability grants must be validated according to the host's configured execution-boundary rules.
  • Policy version and policy fingerprint are distinct product fields with different meanings.
  • Governance outcome and package operation result remain separate concepts.