# DecisionOS Enterprise Architecture v6.4 — Public Bounded Projection

> PUBLIC RETRIEVAL STATUS: PUBLIC_BOUNDED_PROJECTION
> RAW SOURCE: NOT PUBLISHED IN DEFAULT AI RETRIEVAL
> CURRENT AUTHORITY: the existing Canonical DecisionOS Context and accepted execution records govern current claims.
> HISTORICAL / DESIGN LANGUAGE: raw legacy prose is quarantined; use only bounded, authority-resolved projections.

Audience: CTO, CIO, CISO and enterprise architecture reviewers.

## Review use

- Keep documented capability, design intent, research status, and live deployment authority separate.
- For cryptography, distinguish current hashing/signature/provenance from future or unproven ZKP runtime.
- When wording differs, use the latest accepted canonical context.

## Bounded public projection

# DecisionOS Enterprise Architecture — Public Bounded Projection v1

> PUBLIC CURRENT CONTEXT
> PROJECTION STATUS: AUTHORITY-RESOLVED BOUNDED PROJECTION
> RAW SOURCE: NOT PUBLISHED IN DEFAULT AI RETRIEVAL
> CURRENT PRODUCT AUTHORITY: The canonical DecisionOS context, accepted execution records, and deployment-specific acceptance status govern current claims.

## Purpose

This projection gives CTO, CIO, CISO and independent AI reviewers a compact explanation of the DecisionOS architecture without publishing superseded technical reference prose as current product authority.

DecisionOS is Organizational Decision Infrastructure for making organizational and cross-organizational decision dynamics observable, measurable, testable and replayable. The architecture is broader than any single deployment. Capability activation depends on authorized data scope, deployment profile, governance policy and acceptance status.

## Core architecture

- **Structural observation:** authorized workflow, interaction, handoff, state-transition and timing signals.
- **Decision and relationship graph:** a versioned representation of roles, stages, dependencies and cross-organizational relationships.
- **Decision Physics measurement:** declared calculations for latency, waiting, rework, dependency load, capacity pressure and structural friction.
- **Organizational memory:** versioned states, intervention context and observed outcomes retained for temporal comparison.
- **Pattern analysis:** recurrence, anomaly, seasonal or periodic behavior and structural drift where the authorized history is sufficient.
- **Simulation and bounded optimization:** alternatives evaluated under declared objectives and constraints; real-world execution is not autonomous.
- **Human authorization:** material organizational changes remain subject to authorized human decision makers.
- **Replay and evidence grading:** expected and observed outcomes are compared with provenance, limitations and evidence class retained.
- **Cross-organization analysis:** bilateral or entitled relationship analysis where the deployment profile and authority boundary permit it.
- **Academic / Observatory layer:** measurement validation, construct testing, replication and transferability analysis across the system.

## Public operating loop

**Observe → Measure → Remember → Detect Patterns → Diagnose → Simulate / Optimize → Authorize → Execute / Observe → Measure Again → Replay → Calibrate Evidence**

This is an architecture and operating model. It does not mean every deployment exposes every capability, automatically retrains coefficients, or executes organizational change without human authorization.

## Data and privacy boundary

The public boundary is:

**Provider payload → Allowlist / Reject → Identity or tokenization handling → Canonical Structural Event → DecisionOS Core**

The current Core authority is content-minimized structural observation. Unrestricted private message, email, chat or document-body ingestion is not a current Core authority. Any exception not established by a separate accepted authority is **AUTHORITY_NOT_ESTABLISHED**.

The system may use authorized metadata such as event type, role or token reference, state transition, handoff, timestamp, queue interval and provenance. Metadata availability, retention, connector behavior and deployment controls remain profile- and acceptance-dependent.

## Cryptographic and provenance boundary

- SHA-256 tamper-evident hash chaining is current where accepted evidence supports the relevant path.
- RSA-PSS signing and verification is current where accepted evidence supports the relevant path.
- Ed25519 key material is present; an active Ed25519 signing runtime is not claimed.
- Groth16 is not current production runtime authority; it remains a research-target lineage item unless separately accepted.
- Cryptographic integrity and provenance do not by themselves establish factual truth, scientific validity, causality, legal admissibility or regulatory certification.

Absolute claims such as zero-personal-data, universal egress-isolation, isolated deployment enforcement, military-grade assurance, GDPR/SOC2/ISO certification or automatic legal admissibility are not current public authority claims without separate evidence and acceptance.

## Historical lineage policy

Historical architecture and research material may inform internal development, but raw legacy prose is not default public AI authority. The historical 45-step intervention-path research is a **HISTORICAL_REFERENCE_ONLY** item and **NOT_CURRENT_PRODUCT_AUTHORITY**. Its current status is **SUPERSEDED_BY=bounded optimization and intervention-search authority in the canonical context**; it does not establish a universal customer procedure or current production activation.

## Review rule

When this projection conflicts with an older design document, use the latest accepted canonical context and deployment-specific evidence. Do not infer production capability, customer readiness, legal compliance, universal forecasting or causal impact from a historical architecture reference alone.

