Twelve engines around one question: what has to be true before intelligence becomes action?
SwiftOPSA separates source, evidence, semantic reasoning, authority, governance, assurance, human oversight, decisions, telemetry, replay, agent-control enforcement, and commercial entitlement into explicit operational components.
01 · Ingestion
Source registration, hashing, provenance, extraction, evidence, and semantic indexing.
02 · Governance
Authorized rules, deterministic evaluation, scope, conflicts, and ruleset fingerprints.
03 · Authority
Authority identity, source, precedence, scope, status, delegation, and resolution.
04 · Evidence
Integrity verification, provenance, chain-of-custody events, and evidence identity.
05 · Decision
Attributable decision records linked to evidence, authority, evaluations, AI interactions, and approvals.
06 · Replay
Hash-chain verification, matter replay, subject replay, and governed historical reconstruction.
07 · Telemetry
Runtime health for schemas, contracts, engines, storage, audit, and diagnostics.
08 · Runtime State
Manifest hashes, audit head, capabilities, configuration state, and runtime fingerprint.
09 · Semantic Reasoning
Proposition indexing, relevance, support, contradiction, insufficiency, and attributable reasoning traces.
10 · AI Assurance
System inventory, risk assessment, permit-review-block gates, approvals, incidents, and red-team evidence.
11 · Agent Control Gateway
Agent identity, delegated authority, executable policy, human review, signed action receipts, and fail-closed tool execution.
12 · Licensing & Entitlements
Signed customer licenses, feature entitlements, installation activation, usage quotas, validation, and commercial runtime enforcement.
The architecture is the chain, not the cards
The engines become commercially meaningful when they operate as one governed path: Source → Provenance → Evidence → Authority → Governance → Semantic Reasoning → AI Assurance → Delegated Agent Authority → PERMIT / REVIEW / BLOCK → Human Decision → Action Receipt → Audit → Replay.
Portable governance meaning
The durable architecture is not intended to be limited to PHP or IONOS. Canonical record contracts and engine responsibilities can survive migrations to other languages, PostgreSQL, private cloud, containers, Android, APIs, and enterprise infrastructure.
What remains private
The public pages explain the architecture. Operational data, customer documents, governance rules, evidence, AI assessments, credentials, licensing records, and runtime administration remain behind authenticated controls and are intentionally excluded from public indexing.
Independent Owner Access
Release 9.0.12 preserves a dedicated owner identity outside ordinary operator credentials. A protected one-time bootstrap establishes the owner as Eric, then forces creation of a stronger permanent password and independent MFA. Runtime credentials, governance data, automation agents, evidence, licensing, replay, and public surfaces remain separated by the existing security boundaries.
Governed Automation Layer
Release 9.0.9 adds nine bounded bots above the twelve-engine runtime: adversarial assurance, inline gateway interception, policy coverage, AI intake risk tiering, reasoning-trace review, cryptographic replay, execution anomaly guard, license/quota monitoring, and public-surface privacy checks. These bots do not create their own authority. High-risk access is routed to a human JIT decision, and critical runtime anomalies can quarantine side-effect agent calls at the gateway.
Do not make every new model start governance from zero.
SwiftOPSA is designed to make the evidence, authority, policy, approval, and replay layer persistent even as the intelligence underneath it evolves.