Datacendia — Enterprise Edition

⚠️ Private Repository — This is the enterprise monorepo containing all commercial features. The open-source Community Edition is at datacendia-core (Apache 2.0).

The Defensible AI Platform — every decision, defensible.

Datacendia is the only AI platform where every decision is auditable, explainable, and forensic-grade, independently verifiable. Multiple AI agents deliberate — argue, dissent, and challenge each other — then every decision is recorded in a cryptographically signed, immutable evidence packet. Open-core. Sovereign-first. Self-hosted or cloud. Your data, your keys, your proof.

Repository Landscape

Repo	Purpose	License	Visibility
datacendia-components (this repo)	Enterprise monorepo — full platform (frontend + backend)	Commercial	Private
datacendia-core	Community Edition — open-source core engine	Apache 2.0	Public
datacendia-marketing	Marketing website (datacendia.com)	Proprietary	Private
decision-governance-infrastructure	DDGI framework specification	CC BY 4.0	Public

┌─────────────────────────────────────────────────────────────┐
│                    DATACENDIA PLATFORM                       │
│                                                              │
│   ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────┐   │
│   │ Strategy │  │  Risk    │  │Compliance│  │ Dissent  │   │
│   │  Agent   │──│  Agent   │──│  Agent   │──│  Agent   │   │
│   └────┬─────┘  └────┬─────┘  └────┬─────┘  └────┬─────┘   │
│        │             │             │             │           │
│        └─────────────┴──────┬──────┴─────────────┘           │
│                             │                                │
│                    ┌────────▼────────┐                       │
│                    │ Council Engine  │  ← Multi-agent        │
│                    │  (Deliberation) │    deliberation        │
│                    └────────┬────────┘                       │
│                             │                                │
│                    ┌────────▼────────┐                       │
│                    │ Decision Ledger │  ← Immutable,         │
│                    │  (Merkle-signed)│    cryptographic       │
│                    └─────────────────┘                       │
└─────────────────────────────────────────────────────────────┘

Why Datacendia?

• Deliberation over dictation — Multiple AI perspectives challenge each other before any recommendation. No single-model black box.
• Immutable audit trail — Every decision is Merkle-signed with full reasoning chains. Export forensic-grade, independently verifiable evidence packets.
• Sovereign by default — Runs entirely on your infrastructure. Air-gapped deployable. No data leaves your network.
• Compliance-native — Architecture aligned to SOC 2, HIPAA, GDPR, NIST 800-53, Basel III, EU AI Act. Controls implemented, formal certifications available on enterprise contract.
• 9 enterprise infrastructure integrations — Kafka, Temporal, OPA, OpenBao, NeMo Guardrails, RAPIDS, Flink CEP, Triton, Confidential Computing. All opt-in with embedded fallbacks.

What the Council Found

We ran the platform against eleven real decision types across eleven industries. In each case, the pivotal insight emerged from cross-examination between agents with different mandates — not from any single agent's analysis.

Walkthrough	Scenario	What the Council Found
Financial Services	$2.3B CRE acquisition, Basel III capital planning	Credit Analyst reframes "22% discount" as 2.2% discount to fair value; Compliance Officer catches SR 11-7 model risk blocking issue that would halt the deal
Healthcare	Sepsis prediction AI tool, FDA SaMD classification	Patient Safety Officer calculates alert fatigue reduces effective sensitivity to 54.8% — worse than current manual screening at 68%
Government	$47M IT modernization, FAR Part 15 source selection	Program Manager shows "cheaper" incumbent costs $22.1M more over 10 years due to COBOL lifecycle debt; Legal Counsel identifies 3 GAO protest vulnerabilities
Defense	HA/DR + freedom-of-navigation, 8-agent JOPP	OPSEC Officer reveals "lower risk" COA is actually worse for security — no exclusion zone lets adversary collection platforms operate at close range
Energy	1,400 MW grid deficit, 8 minutes to cascading blackout	Renewable Optimizer identifies that discharging full battery now creates a catastrophic 2,840 MW deficit at sunset 90 minutes later
Legal	$28M FLSA class action, 10-day settlement deadline	Ethics Counsel surfaces Rule 2.1 duty: lawyers must tell the board their $12M authority is unrealistic, even if candor costs the client relationship
Sports	€47M player transfer, UEFA FFP compliance	CMO's risk-adjusted pricing adds €9.2M to effective cost (ACL history); Governance agent turns potential FFP sanction into voluntary agreement
Manufacturing	Brake caliper defect, $42K/hour line-down penalty	Maintenance Engineer identifies multi-factor root cause where every individual parameter was "within limits" but the combination exceeded tolerance
Insurance	$1.4M workers' comp claim, fraud score 74/100	Chief Actuary decomposes reserve: $984K is provider-ring inflation on top of $416K legitimate claim — reframes from single-claim to $28M ring exposure
Real Estate	$94M office-to-residential conversion, 76% vacancy	Chief Appraiser cuts $30M from developer's valuation; Market Analyst shows conversion pipeline is 80% of annual absorption — market window is closing
Pharmaceutical	Phase III cardiovascular trial, interim safety signal	Biostatistician shows DILI confidence interval is 30× wide — stopping for efficacy is a regulatory trap because safety database is too small for NDA

How Pivotal Insights Emerge

Every finding above came from one of three agent interaction patterns:

1. A specialist reframes a number. The Credit Analyst turns 22% into 2.2%. The Patient Safety Officer turns 87% sensitivity into 54.8% effective sensitivity. The CMO turns €47M into €56.2M risk-adjusted. These reframings change the decision because they reveal the number the decision-maker should be looking at, not the one that was presented.

2. A compliance agent catches a blocking prerequisite. The SR 11-7 model risk finding. The 21 CFR Part 11 audit trail requirement. The host nation exclusion zone request. These are binary — they don't change the recommendation, they gate it. A single-model prompt rarely surfaces prerequisites because it's optimizing for the answer, not the preconditions.

3. Cross-examination reveals second-order effects. The OPSEC Officer inverts the risk assessment. The Renewable Optimizer models the sunset crisis. The Discovery Specialist shows the remaining 40% of production makes the case worse. These emerge specifically because one agent challenges another's framing and forces a deeper analysis that the original agent's mandate wouldn't have produced.

This structure — not the technology, but the pattern of how insights emerge from multi-agent deliberation — is why a council of specialized agents produces different results than a single model with a longer prompt. The single model has all the knowledge. It lacks the adversarial structure that forces knowledge to be applied from competing perspectives.

Pricing & Tiers

This repository (datacendia-components) contains all paid tiers. The free Community tier is datacendia-core (Apache 2.0).

Tier	Price	What You're Actually Buying
Community	Free	A working AI governance platform you can try. Self-hosted. No SLA. No support. Regex PII. Basic Council. Enough to demo to your CISO.
Pilot	$50K/yr	We run it for you. Managed deployment, 99.5% SLA, priority support, full deliberation engine, 90-day money-back guarantee.
Foundation	$150K–$500K/yr	Production compliance. Full compliance engines (Basel III, EU AI Act, cross-jurisdiction), ML-based PII (Presidio), Echo/Gnosis evidence, OmniTranslate (26 languages), expanded verticals.
Enterprise	$500K–$1.5M/yr	Advanced risk. COLLAPSE stress testing, 12 adversarial agents, Shadow Council, sovereign LLM providers, SSO/MFA, SIEM, ZK proofs, multi-tenant.
Strategic	$1.5M+/yr	Nation-scale. Air-gapped, data diode, TPM attestation, federated mesh, post-quantum crypto, defense-grade, portable instances.

Feature Comparison

Capability	Community	Pilot	Foundation	Enterprise	Strategic
Council Engine (multi-agent deliberation)	✅	✅	✅	✅	✅
Immutable Audit Ledger (Merkle-signed)	✅	✅	✅	✅	✅
CendiaGateway (AI governance proxy)	✅	✅	✅	✅	✅
CendiaReplay (decision playback)	✅	✅	✅	✅	✅
DCII Services (Truth, Notary, Witness)	✅	✅	✅	✅	✅
PII Detection	Regex	Regex	ML (Presidio)	ML + Custom	ML + Custom
Industry Verticals	18 basic	18 basic	30 expanded	30 full	30 full + custom
Managed Platform & SLA	—	✅ 99.5%	✅ 99.9%	✅ 99.95%	✅ Custom
Compliance Engines (Basel III, EU AI Act)	—	—	✅	✅	✅
Echo/Gnosis Evidence & Audit Replay	—	—	✅	✅	✅
OmniTranslate (26 languages)	—	—	✅	✅	✅
COLLAPSE Stress Testing	—	—	—	✅	✅
Shadow Council & Red Team	—	—	—	✅	✅
SSO/MFA (Keycloak)	—	—	—	✅	✅
SIEM Integration (Splunk, Sentinel)	—	—	—	✅	✅
Sovereign LLM Providers (offline)	—	—	—	✅	✅
Sovereign Online Toggle	—	—	—	✅	✅
Zero-Knowledge Proofs	—	—	—	✅	✅
Air-Gapped Deployment	—	—	—	—	✅
Data Diode & TPM Attestation	—	—	—	—	✅
Federated Mesh (multi-org)	—	—	—	—	✅
Post-Quantum Cryptography	—	—	—	—	✅

Sovereign Online Toggle (Enterprise+)

Enterprise and Strategic tiers support fully sovereign/air-gapped deployments. A single environment variable disables all external cloud services:

DATACENDIA_ONLINE_MODE=false     # Master toggle — disables all cloud AI + external services
DATACENDIA_CLOUD_AI_FALLBACK=error  # 'error' (auditor-safe default) or 'local' (silent Ollama fallback)

When offline: cloud AI → hard 503 or local fallback, external data → cached datasets, notifications → internal event bus only. System validates at startup that local LLM providers are configured — the passing validation is an audit artifact.

See GET /api/v1/health/sovereign for real-time sovereign mode status.

🚀 Quick Start

Try it now (zero config)

# Clone and launch — no .env, no setup, pre-seeded demo data
git clone https://github.com/datacendia/datacendia-components.git
cd datacendia-components
docker compose -f docker-compose.demo.yml up

Open http://localhost:5173 → log in as sarah.chen@acme.demo (dev auth bypass, no password).

You'll see a pre-seeded Acme Corporation with 5 users, 6 Council agents, 5 deliberations (completed + in-progress), 8 decisions, 12 months of metrics, and a full audit trail. ~6GB RAM required.

Development Setup

Prerequisites

• Node.js 20.x or later
• Docker & Docker Compose
• PostgreSQL 16+
• Redis 7+
• Neo4j 5+ (for knowledge graph)
• Ollama (for local AI/LLM — or Triton/NIM via INFERENCE_PROVIDER)
• NVIDIA GPU (optional — required for RAPIDS, Triton, Confidential Computing)

# Clone the repository
git clone https://github.com/datacendia/datacendia-components.git
cd datacendia-components

# Copy environment files
cp .env.example .env
cp backend/.env.example backend/.env

# Install all dependencies (Prisma client auto-generates via postinstall)
npm install
cd backend && npm install && cd ..

# Start infrastructure with unified compose (RECOMMENDED)
docker-compose -f docker-compose.unified.yml --profile core up -d

# Run database migrations
cd backend && npx prisma migrate deploy && cd ..

# Seed demo data (optional)
npm run db:seed

# Start frontend and backend locally
npm run dev              # Frontend - http://localhost:5173
cd backend && npm run dev # Backend - http://localhost:3001

Docker Compose Profiles

Profile	Services	RAM Required
core	PostgreSQL, Redis, Neo4j, Ollama	8GB
sovereign	+ Druid, ClickHouse, MinIO, Keycloak, etc.	32GB
observability	+ Prometheus, Grafana, Loki, Tempo	48GB
security	+ Wazuh, Infisical, Step-CA	64GB
nvidia	+ Triton, NeMo Guardrails, RAPIDS	32GB + GPU
events	+ Kafka, Temporal, Temporal UI	16GB
policy	+ OPA, OpenBao, Flink	8GB
full	Everything	96GB+

# Core only (minimal for development)
docker-compose -f docker-compose.unified.yml --profile core up -d

# Core + Sovereign services
docker-compose -f docker-compose.unified.yml --profile core --profile sovereign up -d

# Full stack
docker-compose -f docker-compose.unified.yml up -d

Service URLs

Service	URL	Credentials
Frontend	http://localhost:5173	See .env
Backend API	http://localhost:3001	-
Neo4j Browser	http://localhost:7474	See .env (NEO4J_USER / NEO4J_PASSWORD)
MinIO Console	http://localhost:9001	See .env (MINIO_ROOT_USER / MINIO_ROOT_PASSWORD)
Grafana	http://localhost:3002	See .env (GRAFANA_USER / GRAFANA_PASSWORD)
Keycloak	http://localhost:8180	See .env (KEYCLOAK_ADMIN / KEYCLOAK_ADMIN_PASSWORD)
Triton Inference	http://localhost:8000 (HTTP), :8001 (gRPC)	-
Kafka	localhost:9092	-
Temporal UI	http://localhost:8088	-
OpenBao/Vault	http://localhost:8200	OPENBAO_TOKEN
OPA	http://localhost:8181	-
Flink Dashboard	http://localhost:8081	-

Security: Copy .env.example to .env and set strong, unique passwords before running. Never commit .env to version control.

📁 Project Structure

datacendia-components/
├── src/                    # React frontend (Vite + TypeScript + Tailwind)
│   ├── components/         # Reusable UI components
│   ├── pages/              # Page components (routes)
│   ├── lib/                # Utilities, API clients, hooks
│   └── services/           # Frontend services
├── backend/                # Node.js backend (Express + Prisma)
│   ├── src/
│   │   ├── routes/         # API endpoints (160 route files)
│   │   ├── services/       # Business logic (456 service files)
│   │   │   ├── inference/  # InferenceProvider (Ollama, Triton, NIM)
│   │   │   ├── guardrails/ # NeMo Guardrails engine
│   │   │   ├── kafka/      # Kafka producer, consumer, event bridge
│   │   │   ├── temporal/   # Temporal.io workflow orchestration
│   │   │   ├── opa/        # Open Policy Agent engine
│   │   │   ├── vault/      # OpenBao/Vault secrets management
│   │   │   ├── gpu/        # RAPIDS analytics + Confidential Computing
│   │   │   ├── streaming/  # Flink CEP real-time processing
│   │   │   ├── verticals/  # 29 industry vertical implementations
│   │   │   └── sovereign/  # 11 sovereign architecture patterns
│   │   ├── middleware/     # Auth, logging, security
│   │   ├── security/       # PolicyEngine, KeycloakAuth, KMS, HSM
│   │   └── config/         # Database, Redis, Neo4j, inference config
│   └── prisma/             # Database schema & migrations (194 models)
├── tests/                  # Test suites (Vitest + Playwright)
├── infrastructure/         # PostgreSQL HA scripts
├── grafana/                # Dashboard & datasource provisioning
├── docs/                   # Technical documentation (70+ files)
└── docker/                 # Docker configurations

🧪 Testing

Current status: 262 test files, 205,754+ tests, 99.99% passing (2 pre-existing env-dependent failures)

# All tests (frontend + backend + integration)
npm test

# Frontend unit tests
npm run test

# Backend tests
npm run test:backend

# Tests with coverage
npm run test:coverage

# E2E tests (Playwright)
npm run test:e2e

# Type checking (all)
npm run typecheck:all

# Linting (all)
npm run lint:all

Test Categories

Category	Files	Description
Backend unit	165	Services, routes, middleware
Integration/E2E	33	Full platform, edge cases, Playwright
AI Validation	5	LLM quality, bias/ethics, load, air-gap
Enterprise	6	Schema, security, performance, i18n
Frontend	4	Auth, routing, components, i18n
Contract	1	Consumer pact tests

Tests gracefully skip when optional services (Ollama, backend, frontend) are offline.

🏗️ Building

# Build frontend
npm run build

# Build backend
cd backend && npm run build

# Build Docker images
docker build -t datacendia/frontend:latest .
docker build -t datacendia/backend:latest ./backend

🚀 Deployment

Cloud/Standard Deployment

# Production deployment with Docker Compose
docker compose -f docker-compose.production.yml up -d

See DEPLOYMENT.md for detailed instructions.

Air-Gapped / On-Premise Deployment

For sovereign deployments on isolated networks:

# Build deployment package (on machine with internet)
.\scripts\build-airgapped-package.ps1 -Version "1.0.0"

# With LLM models included (~15-25GB)
.\scripts\build-airgapped-package.ps1 -Version "1.0.0" -IncludeModels

This creates a self-contained package with:

• All Docker images as .tar.gz files
• Configuration templates
• Installation scripts for Linux/Windows
• LLM models (optional)

See Air-Gapped Deployment Guide for complete instructions.

📊 Platform Scale

Metric	Count
Backend service files	456 (implementation, excl. index/types)
Backend source files (total)	1,024
Backend route files	160
Backend test files	262
Backend middleware	17
Frontend pages	209
Frontend components	92
Frontend source files	504 (325 TSX + 179 TS)
Prisma models	194
Prisma enums	141
Prisma schema files	13
Passing tests	205,755
Industry verticals	30
Collapse agents	19 specialized
Docker Compose files	4 (dev, demo, production, nvidia)
Compliance frameworks	10
Supported jurisdictions	17
AI agent presets	50+
Documentation files	359
TypeScript errors	0

🏛️ Cendia™ Product Catalog

Core Suite (The "Brain")

Product	Description
CendiaChronos™	Enterprise Time Machine — replay past decisions, simulate future crisis scenarios
Ghost Board™	Rehearse high-stakes board meetings against AI avatars
CendiaPreMortem™	AI analyzes why your decision will fail before you execute it
CendiaPredict™	NEW — Forward-looking quantitative risk scoring: "73% chance of regulatory challenge in 9 months" with evidence
CendiaRewind™	NEW — Counterfactual decision replay: "If we'd chosen Option B, we'd be 15% better off"
CendiaRecall™	Decision Outcome Tracker — predicted vs actual, bias detection, lessons learned
Decision Debt™	Real-time dashboard of stuck decisions and financial cost of delay
CendiaLive™	Watch AI agents deliberate in real-time with animated avatars
CendiaReplay™	Watch past deliberations unfold like a movie
CendiaEcho™	Decision Outcome Engine — track what actually happened after each decision
CendiaLens™	AI Interpretability — token confidence, reasoning chains, bias detection
CendiaCollapse™	Adversarial Policy Stress-Testing — 19 specialized agents stress-test decisions
CendiaPulse™	Mission control — real-time agent activity, compliance, risk scoring
CendiaCrisis™	Incident Response Center — detection to resolution with complete audit trail
CendiaROI™	Prove the ROI of governance with real deliberation throughput and quality metrics
CendiaDCII™	Decision Crisis Immunization Infrastructure — IISS, 9 primitives, media auth, timestamps
CendiaGateway™	AI Governance Gateway — reverse proxy for all AI traffic with PII detection, policy enforcement, DCII signing
The Governance Receipt™	Printable, cryptographically verified artifact proving every AI interaction was governed — for regulators and auditors
CendiaOrchestrate™	Service Orchestration Workflow Builder — visual drag-and-drop of 60+ platform services into reusable workflows with persistent save/load and run simulation

Trust Layer (The "Shield")

Product	Description
CendiaOversight™	Real-time Regulatory Radar — FDA, GDPR, DORA frameworks with policy gates
CendiaNotary™	Cryptographic Signing Authority — customer-owned keys
CendiaVault™	Unified Evidence Storage — decision packets, audit ledger, evidence bundles
CendiaProvenance™	Full decision lineage & evidence export — forensic-grade, independently verifiable
CendiaCrucible™	Adversarial Stress Testing — attack decisions with simulated threats
CendiaRedTeam™	Every agent becomes a devil's advocate
SGAS™	Synthetic Governance Agent System — 5 agent classes at societal scale
CendiaCourt™	Formal AI dispute resolution with precedent tracking
CendiaSandbox™	Test against proposed regulations before they become law
CendiaZKP™	Prove compliance without revealing proprietary logic or data
CendiaInsure™	Direct liability coverage per AI decision with real-time risk scoring
CendiaQuantumKMS™	Quantum-resistant cryptographic signatures (Dilithium, SPHINCS+, Falcon)
CendiaCarbon™	Reduce AI carbon footprint with intelligent workload scheduling
CendiaJurisdiction™	17-jurisdiction compliance engine for cross-border data transfers
CendiaCompliance™	Real-time monitoring for 10 compliance frameworks

📚 Documentation

Deployment

Document	Description
Quick Reference	Copy-paste commands cheat sheet
Deployment Guide	Standard deployment instructions
Docker Guide	Complete Docker configuration reference
Air-Gapped Deployment	Offline/on-premise deployment
PostgreSQL HA Guide	Database high availability setup
Infrastructure Setup	Local infrastructure services

Use Cases & Research

See What the Council Found for the full pivotal insights table with all eight walkthroughs.

Document	Description
Benchmark: Council vs. Single-Model	Worked comparison — 3.2× more risk factors, 6× regulatory citations, dissent surfacing
Community vs Enterprise	Open-core boundary — what's free, what's commercial

Sales & Outreach Resources

Document	Description
Case Studies	Platform demonstrations across 9 industries
Financial Services Demo	$2.3B CRE acquisition — Basel III deliberation
Healthcare Demo	FDA SaMD — alert fatigue analysis
Outreach Templates	Cold email/DM templates by vertical
HN Launch Post	Draft Show HN post

Technical

Document	Description
Platform Audit (Mar 2)	Cross-repo audit — 33 findings, 14 remediated, CI/security/deps
Platform Audit (Feb 18)	Service verification audit
Enterprise Readiness	Production audit & compliance
API Documentation	Backend API reference
Architecture Diagrams	System architecture
Product Bible	Product vision and features
Real-World Value	Detailed use case examples
Complete Service Matrix	Full service catalog with pricing
Verticals	29 industry verticals deep dive

🔐 Environment Variables

See .env.example for required environment variables:

# Database
DATABASE_URL=postgresql://user:pass@localhost:5432/datacendia

# Redis
REDIS_URL=redis://localhost:6379

# Neo4j
NEO4J_URI=bolt://localhost:7687
NEO4J_USER=neo4j
NEO4J_PASSWORD=password

# AI/LLM Inference (choose one provider)
INFERENCE_PROVIDER=ollama              # ollama | triton | nim
OLLAMA_BASE_URL=http://localhost:11434
TRITON_URL=localhost:8001              # Triton gRPC
TRITON_HTTP_URL=http://localhost:8000  # Triton HTTP
NIM_URL=http://localhost:8000          # NVIDIA NIM

# Auth
JWT_SECRET=your-secret-key

# ── Optional Infrastructure (all disabled by default) ──

# NeMo Guardrails
NEMO_GUARDRAILS_ENABLED=false          # true to activate
NEMO_GUARDRAILS_MODE=hybrid            # server | embedded | hybrid
NEMO_GUARDRAILS_URL=http://localhost:8080

# Apache Kafka
KAFKA_ENABLED=false
KAFKA_BROKERS=localhost:9092

# Temporal.io
TEMPORAL_ENABLED=false
TEMPORAL_ADDRESS=localhost:7233
TEMPORAL_NAMESPACE=datacendia

# OpenBao/Vault
OPENBAO_ENABLED=false
OPENBAO_ADDR=http://127.0.0.1:8200
OPENBAO_TOKEN=

# Open Policy Agent
OPA_ENABLED=false
OPA_MODE=embedded                      # server | embedded
OPA_URL=http://localhost:8181

# NVIDIA RAPIDS
RAPIDS_ENABLED=false
RAPIDS_URL=http://localhost:8787

# Confidential Computing
CC_ENABLED=false
CC_ATTESTATION_URL=http://localhost:8443

# Apache Flink CEP
FLINK_ENABLED=false
FLINK_MODE=embedded                    # cluster | embedded
FLINK_URL=http://localhost:8081

🛡️ Security

• See SECURITY.md for security policy and vulnerability reporting
• All secrets must be stored in environment variables
• Never commit .env files or API keys

🤝 Contributing

See CONTRIBUTING.md for contribution guidelines.

📄 License

Copyright © 2024-2026 Datacendia, Inc. See LICENSE for details.

Community Edition components are available under open-source terms. Enterprise features require a commercial license — contact sales@datacendia.com.

---

📋 Changelog (click to expand)

March 2, 2026 — Platform Audit Remediation

• Cross-repo audit — 33 findings across all 4 repos, 14 remediated same-day
• datacendia-core CI overhaul — Rewrote ci.yml (concurrency, Prisma generate, community build, status gate), added security.yml (CodeQL, TruffleHog, dependency audit), added dependabot.yml
• Dependency vulns reduced 31→6 — fast-xml-parser override → 5.4.1, multer → 2.1.0, added @aws-sdk/xml-builder override
• Removed 2.5GB caselaw data from git tracking (91,725 files)
• Marketing HTTPS enabled — Uncommented Force HTTPS redirect in .htaccess
• CI hardened — Lint now blocks merges (removed continue-on-error), security audit blocks on critical vulns
• Housekeeping — Fixed SECURITY.md versions, moved @types/* to devDependencies, deleted empty fix-ds.ts

March 2, 2026 — The Governance Receipt™ & CendiaGateway Housekeeping

• The Governance Receipt™ — Named artifact: the printable, cryptographically verified document a CISO hands to a regulator. Replaces "AI Manifest" branding. Export as HTML (PDF), CSV, JSON via POST /governance-receipt/export
• Prisma generate automated — postinstall hook in backend/package.json runs prisma generate automatically after npm install. Also added to build script and both enterprise setup scripts (.ps1 / .sh)
• Browser extension icon tooling — generate-icons.html renders the gold shield SVG onto canvas at 16/48/128px and provides PNG download buttons. README updated with prominent warning that Chrome requires real PNGs
• GovernanceReceipt type alias — exported alongside legacy AIManifest for backwards compatibility
• CendiaGateway™ and The Governance Receipt™ added to product catalog

February 26, 2026 — Infrastructure Upgrade (9 Enterprise Components)

NVIDIA Inception Program Member 🟢

• InferenceProvider Abstraction — Unified IInferenceProvider interface with Ollama, Triton, and NVIDIA NIM backends
• NeMo Guardrails — 9 default rails (jailbreak, hallucination, bias, PII, topic enforcement)
• NVIDIA RAPIDS / cuGraph — GPU-accelerated bias analysis, graph analytics, anomaly detection (CPU fallback)
• Confidential Computing — GPU attestation, session management, CC evidence generation
• Apache Kafka — 7 topic categories, in-memory fallback, EventBridge integration
• Temporal.io — 6 built-in workflows, embedded execution fallback
• OpenBao/Vault — KV v2, transit encryption, PKI, dynamic DB credentials
• Open Policy Agent — 8 embedded policies, EU AI Act, HIPAA minimum necessary
• Apache Flink CEP — Sliding-window engine, 6 default rules

All 9 components are opt-in (disabled by default) with embedded fallbacks.

March 16, 2026 — Service Orchestration Workflow Builder

• CendiaOrchestrate™ — Visual drag-and-drop workflow builder with 60+ services across Foundation, Enterprise, and Strategic tiers
• ServiceRegistry — Full platform service catalog with inputs, outputs, and configuration schemas for every registered service
• WorkflowPersistenceService — localStorage CRUD for workflows (create, read, update, delete, duplicate, export/import, run logging)
• Shared workflow types — Workflow, WorkflowStep, WorkflowConnection, ServiceDefinition, ConfigField, PortDefinition
• Infrastructure audit — Full cross-repo sync verified: pages, routes, nav, components, contexts, APIs, schemas all aligned
• 4 shared components synced — RedTeamReportPanel, SimilarDecisionsPanel, CendiaStampSeal, EvidencePackageDownload
• Sidebar nav — Workflows entry added to enterprise navigation group

March 10, 2026 — Complete Vertical Deep Testing

• All 30 verticals deep-tested — 647 domain-specific tests across 9 test files
• Basel III Engine — CET1/AT1/Tier2 capital, credit/market/operational RWA, LCR, NSFR, large exposures, stress tests
• Defense vertical — 24 agents, 26 council modes, 5 compliance frameworks (FedRAMP/CMMC/ITAR/NIST/LOAC)
• Industrial Services — 10 expanded schemas with SUNAFIL/OSHA/ASME regulatory rules
• 205,754 tests passing across 262 test files

February 18, 2026 — Predictive Intelligence

• CendiaPredict™ — Forward-looking quantitative risk scoring
• CendiaRewind™ — Counterfactual decision replay
• 204,751 tests passing — 0 TypeScript errors

February 17, 2026 — Platform Integrity Audit

• CendiaRecall™ — Decision outcome tracking with bias detection
• 666 Math.random() calls eliminated — Replaced with deterministic, reproducible computations
• 311/311 backend services rated 10/10 — Zero placeholder code

February 12, 2026 — DCII Framework

• CendiaIISS™ — Institutional Immune System Score (0–1000 scale)
• CendiaMediaAuth™ — Synthetic media authentication (C2PA signing)
• CendiaTimestamp™ — RFC 3161 external timestamp authority

February 7, 2026 — Enterprise Platinum

• PostgreSQL HA — Primary/replica with PgBouncer, WAL archiving, auto-failover
• Grafana auto-provisioning — Dashboards and datasources on startup
• 202,500+ tests passing

January 28, 2026 — Platform v4.5

• Unified Docker Compose — Single file with profiles for all services
• Defense & National Security Vertical — 24 agents, 35 council modes
• CendiaQuantumKMS™ — Post-quantum cryptography (Dilithium, SPHINCS+, Falcon)
• Sports/Football Vertical — Transfer governance with UEFA FFP, FIFA Agent Regs

---

Built with ❤️ by the Datacendia team