Skip to content

datacendia/datacendia-core

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

354 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Datacendia Core

NVIDIA Inception Program Member

The Defensible AI Platform

Cryptographic proof and forensic trails for every AI decision.

When your AI makes a $500M acquisition recommendation, can you prove in court exactly what it considered, what it dissented on, and who signed off? Datacendia can.

CI Security License: Apache 2.0 TypeScript Node.js React PostgreSQL Redis Docker NVIDIA Inception

Try It Now → · Live Demo · Architecture · Verticals · Free Pilot · Security Audit · Standards · Contributing


Try It in 60 Seconds

Don't want to install anything? Try the live demo at app.datacendia.com — no signup required.

# 1. Install Ollama (https://ollama.com) and pull a model:
ollama pull llama3.2:3b          # Fast — 2GB, good for demos
# ollama pull qwen3:32b          # Production — 20GB, much better quality

# 2. Run the platform:
git clone https://github.com/datacendia/datacendia-core.git
cd datacendia-core
docker compose -f docker-compose.demo.yml up -d

# 3. Open http://localhost:5173
#    Demo login: sarah.chen@acme.demo (no password needed)

What you'll see: A pre-seeded Council dashboard with 5 real deliberations across 5 industries -- Energy grid emergency, Manufacturing safety defect, $1.7B CRE acquisition, Veterans IT modernization, and SaMD medical device deployment. Full agent transcripts, cross-examinations, and cryptographically signed decision packets.

⚡ Pre-built images available — first run should take under 2 minutes

Docker pulls pre-built images from GHCR (ghcr.io/datacendia/datacendia). If you haven't pulled an Ollama model, AI agents will produce no output — the demo data is still fully browseable.

# Pull the latest all-in-one image directly
docker pull ghcr.io/datacendia/datacendia:latest

API Docs: Interactive OpenAPI documentation is available at localhost:3001/api/docs once the backend is running. Raw spec at /api/docs.json.


What is Datacendia?

Datacendia is the only AI platform where every decision is auditable, explainable, and independently verifiable at forensic grade. Multiple AI agents with distinct perspectives deliberate on your behalf -- then every interaction is recorded in an immutable, cryptographically signed audit ledger.

This is not another chatbot. It's an operating system for enterprise decisions.

+-----------------------------------------------------------+
|                    YOUR DECISION                          |
|                                                           |
|    +----------+  +----------+  +----------+              |
|    | Financial|  | Legal    |  | Risk     |  ...agents   |
|    | Analyst  |  | Counsel  |  | Assessor |              |
|    +----+-----+  +----+-----+  +----+-----+              |
|         |             |             |                     |
|         +--------+----+-------------+                     |
|                  |                                        |
|         +--------v---------+                              |
|         |  THE COUNCIL     |  <- Multi-agent consensus    |
|         |  Deliberation    |                              |
|         +--------+---------+                              |
|                  |                                        |
|         +--------v---------+                              |
|         | DECISION PACKET  |  <- Signed, Merkle-rooted   |
|         | + Audit Trail    |     forensic-grade, independently verifiable        |
|         +------------------+                              |
+-----------------------------------------------------------+

See It in Action

Datacendia Council Deliberation -- five demo decisions loaded

The Council deliberation dashboard with five pre-seeded decisions -- docker compose -f docker-compose.demo.yml up -d

Key Capabilities

Currently Implemented (Can Demo Today)

  • CendiaGateway -- AI governance proxy with PII detection, policy enforcement, and SIEM integration
  • The Council -- Multi-agent deliberation with configurable agent panels
  • Immutable Audit Ledger -- Every decision cryptographically signed with Merkle tree integrity
  • Cryptographic Services -- 8 frontend UIs for forensic-grade verification:
    • CendiaVerify -- Public verification portal (/verify) for third-party receipt validation
    • CendiaEvidence -- Evidence package download (ZIP, standalone HTML verifier, JSON)
    • CendiaGapScan -- Compliance gap scanner across 8 regulatory frameworks
    • CendiaStamp -- Cryptographic seal SVG renderer with Ed25519 signatures
    • CendiaPrecedent -- TF-IDF similar decisions panel with precedent matching
    • CendiaRedTeam -- 6-vector adversarial analysis report with gate decision
    • CendiaEscrow -- Shamir Secret Sharing (3-of-5) with VDF time-locks
    • CendiaReplay -- Decision Replay Theater with full deliberation playback
  • Service Orchestration Workflow Builder -- Visual drag-and-drop workflow builder with 17 Foundation services
  • 30 Industry Vertical Frameworks -- Financial, Healthcare, Legal, Defense, Insurance, Energy, Government, Manufacturing, Pharmaceutical, Sports, and 19 more
  • Enterprise Security -- JWT authentication, role-based access control, encryption at rest/in transit
  • WCAG 2.1 AA Accessibility -- 57 passing tests for enterprise accessibility compliance
  • Multi-Tenant Architecture -- Database-level isolation with configurable tenant separation
  • Basic White-Labeling -- CSS-based theming with brand customization
  • SIEM Integration -- Splunk, Datadog, Sentinel, ELK for security event forwarding
  • Performance Monitoring -- Core Web Vitals tracking with real-time metrics
  • GDPR Compliance -- Data subject rights implementation with audit trails

Enterprise Infrastructure (Opt-In)

  • Apache Kafka -- Durable event streaming (KAFKA_ENABLED=true)
  • Temporal.io -- Durable workflow orchestration (TEMPORAL_ENABLED=true)
  • OPA -- Data-driven policy-as-code (OPA_ENABLED=true)
  • OpenBao/Vault -- Secrets management (OPENBAO_ENABLED=true)
  • NeMo Guardrails -- LLM input/output rail evaluation (NEMO_GUARDRAILS_ENABLED=true)
  • NVIDIA RAPIDS -- GPU-accelerated analytics (RAPIDS_ENABLED=true)
  • Apache Flink CEP -- Real-time complex event processing (FLINK_ENABLED=true)

Roadmap Features (In Development)

  • HSM Key Management -- Hardware Security Module integration (Q2 2026)
  • Enterprise Connectors -- SAP, Salesforce, Oracle integrations (Q3 2026)
  • 99.99% Uptime SLA -- Multi-region infrastructure deployment (Q2 2026)
  • Distributed Tracing -- Jaeger/Zipkin integration (Q2 2026)
  • 7-Year Audit Retention -- Long-term storage configuration (Q2 2026)
  • Advanced Analytics -- Predictive monitoring and business intelligence (Q4 2026)

See enterprise-capabilities-audit.md for complete capability matrix and enterprise-roadmap.md for implementation timeline.


CendiaGateway™ -- AI Governance Proxy

CendiaGateway governs every AI interaction in your organization -- API calls, browser-based AI usage, and internal systems. Three coverage layers provide 100% governance:

Layer 1: API Gateway (Reverse Proxy)

Sits between your systems and AI providers. Every request is PII-scanned, policy-enforced, and cryptographically signed.

Employee/System → CendiaGateway → AI Provider (OpenAI, Anthropic, Google, Ollama, etc.)
                       ↓
           PII Scan → Policy Engine → DCII Signing → Audit Ledger

Supported providers: OpenAI, Anthropic, Google AI, Azure OpenAI, Mistral, Cohere, Groq, Ollama

Endpoints:

  • POST /api/v1/gateway/v1/chat/completions -- OpenAI-compatible
  • POST /api/v1/gateway/v1/messages -- Anthropic-compatible
  • POST /api/v1/gateway/proxy/:provider/* -- Any provider

Layer 2: Browser Extensions

Content-script extensions that monitor 15+ AI websites (ChatGPT, Claude, Gemini, Copilot, Perplexity, DeepSeek, Poe, etc.). Intercepts prompts before submission, blocks critical PII instantly.

Browser Status
Chrome / Edge / Brave / Arc ✅ Manifest V3
Firefox ✅ Manifest V2
Safari ✅ Web Extension

Enterprise deployment via Group Policy, Intune, MDM, or Google Workspace Admin.

Layer 3: HTTP Forward Proxy (Network-Level)

Browser-agnostic, network-level AI traffic interception. Configured via PAC file or system proxy settings. Works with any browser, any app.

# Start the proxy
curl -X POST http://localhost:3001/api/v1/gateway/proxy/start

# Configure browsers via PAC file:
# http://gateway-host:3001/api/v1/gateway/proxy/pac

PII Detection

Detects and blocks 10 PII types before they reach AI providers:

Type Action
SSN, Credit Card, Medical Record, Bank Account, Passport Blocked
Email, Phone, IP Address, Date of Birth Redacted or warned

Recommended Deployment

  1. IT blocks direct AI websites at the firewall (5 minutes)
  2. Employees use the internal AI portal routed through CendiaGateway
  3. Browser extension on managed devices catches stragglers
  4. All interactions: PII scanned, policy enforced, cryptographically signed

See browser-extension/README.md for detailed deployment instructions.

Federation Infrastructure

"One agreement, all member organizations covered." Create a federation, add member orgs, push shared AI governance policies, and generate consolidated compliance reports — all with cryptographic signing.

# Create a federation
curl -X POST http://localhost:3001/api/v1/gateway/federation \
  -H "Content-Type: application/json" \
  -d '{"name": "FEPCMAC", "adminOrgId": "fepcmac-hq", "regulatoryFramework": "DS 115-2025-PCM"}'

# Add member organizations
curl -X POST http://localhost:3001/api/v1/gateway/federation/{id}/members \
  -d '{"organizationId": "cmac-aqp", "orgName": "Caja Municipal de Arequipa", "orgCode": "CMAC-AQP"}'

# Push shared policies (block SSN across all members)
curl -X POST http://localhost:3001/api/v1/gateway/federation/{id}/policies \
  -d '{"name": "Block SSN", "action": "block", "piiTypes": ["ssn"]}'

# Consolidated compliance dashboard
curl http://localhost:3001/api/v1/gateway/federation/{id}/dashboard

# Generate signed compliance report
curl -X POST http://localhost:3001/api/v1/gateway/federation/{id}/reports \
  -d '{"reportType": "compliance", "periodStart": "2026-01-01", "periodEnd": "2026-03-31"}'

Federation API: 12 endpoints for federation CRUD, member management, shared policies, per-member stats, risk scoring, and SHA-256 + HMAC signed compliance reports.


Getting Started

Option 1: Demo Mode (Recommended for first look)

See Try It in 60 Seconds above. Pre-seeded data, no configuration needed.

Option 2: Development Setup (~45 minutes first time)

Prerequisites: Node.js 20+, Docker, Ollama

# 1. Pull an Ollama model (do this first — it takes time)
ollama pull llama3.2:3b          # Fast demo model (2GB download)
# For production quality, use: ollama pull qwen3:32b (20GB download, 30-60 min)

# 2. Clone and configure
git clone https://github.com/datacendia/datacendia-core.git
cd datacendia-core
cp backend/.env.example backend/.env

# 3. Start infrastructure (Postgres, Redis, Neo4j)
docker compose -f docker-compose.dev.yml up -d

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

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

# 6. Start the platform (two terminals)
npm run dev              # Terminal 1 — Frontend: http://localhost:5173
cd backend && npm run dev # Terminal 2 — Backend:  http://localhost:3001
Ollama Model Guide
Model Size Quality Best For
llama3.2:3b 2 GB Good Quick demos, testing
llama3.2:8b 5 GB Better Development
qwen3:32b 20 GB Excellent Production (default in .env)
llama3.3:70b 43 GB Best Enterprise (needs 48GB+ RAM/VRAM)

The .env defaults to qwen3:32b. To use a smaller model, edit backend/.env:

OLLAMA_MODEL=llama3.2:3b
OLLAMA_MODEL_FAST=llama3.2:3b

Requirements

Requirement Version Required
Node.js 20.x or later Yes
Docker & Docker Compose Latest Yes
Ollama Latest Yes (or Triton/NIM)
PostgreSQL 16+ Via Docker
Redis 7+ Via Docker
Neo4j 5+ Optional (knowledge graph)
NVIDIA GPU CUDA 12+ Optional (RAPIDS, Triton, CC)

Architecture

datacendia-core/
|-- src/                          # React frontend (Vite + TypeScript + Tailwind)
|   |-- components/               # 92 reusable UI components
|   |   |-- crypto/               # CendiaEvidence, CendiaStamp
|   |   |-- council/              # CendiaPrecedent, CendiaRedTeam, SimilarDecisions
|   |   +-- ui/                   # ServiceInfoDropdown, shared UI
|   |-- pages/                    # 240 page components
|   |   |-- public/               # CendiaVerify (/verify)
|   |   |-- cortex/crypto/        # CendiaEscrow
|   |   |-- cortex/crown/         # Echo, Gnosis, RedTeam
|   |   |-- cortex/dcii/          # Memory, Notary, Truth, Witness, Similarity, Timestamp
|   |   |-- cortex/workflows/     # Service Orchestration Workflow Builder
|   |   +-- cortex/monitor/       # CendiaPulse live agent monitor
|   |-- config/                   # Service info definitions (22 services)
|   |-- types/                    # Shared TypeScript types (Workflow, ServiceDefinition)
|   +-- services/                 # ServiceRegistry (17), WorkflowPersistence, API clients
|-- backend/                      # Node.js backend (Express + Prisma)
|   |-- src/
|   |   |-- services/
|   |   |   |-- council/          # Council deliberation engine
|   |   |   |-- inference/        # LLM provider abstraction (Ollama/Triton/NIM)
|   |   |   |-- guardrails/       # NeMo Guardrails engine
|   |   |   |-- kafka/            # Apache Kafka event streaming
|   |   |   |-- temporal/         # Temporal.io workflow orchestration
|   |   |   |-- opa/              # Open Policy Agent
|   |   |   |-- vault/            # OpenBao/Vault secrets management
|   |   |   |-- gpu/              # RAPIDS analytics + Confidential Computing
|   |   |   |-- streaming/        # Flink CEP real-time processing
|   |   |   |-- gateway/          # CendiaGateway™ AI governance proxy
|   |   |   +-- verticals/        # Industry vertical modules
|   |   |-- routes/               # API route files (domain-grouped)
|   |   |-- security/             # Casbin RBAC, Keycloak SSO
|   |   +-- middleware/           # Auth, rate limiting, security
|   +-- prisma/                   # Database schema and models
|-- browser-extension/            # AI governance browser extensions
|   |-- chrome/                   # Chrome/Edge/Brave/Arc (Manifest V3)
|   |-- firefox/                  # Firefox (Manifest V2)
|   +-- safari/                   # Safari (Web Extension)
+-- docker-compose.dev.yml        # Development infrastructure

Tech Stack

Layer Technology
Frontend React 18, TypeScript, Vite, Tailwind CSS, shadcn/ui (240 pages, 92 components)
Backend Node.js, Express, TypeScript
Database PostgreSQL 16 + Prisma ORM
Cache Redis 7
Graph Neo4j 5
Vector Qdrant
LLM Ollama / NVIDIA Triton / NVIDIA NIM
Events Apache Kafka (opt-in)
Workflows Temporal.io (opt-in)
Policy OPA + Casbin
Secrets OpenBao/Vault (opt-in)
Guardrails NeMo Guardrails (opt-in)
Analytics NVIDIA RAPIDS / cuGraph (opt-in)

Industry Verticals

Datacendia ships with 30 industry vertical definitions. Each vertical provides domain-specific agents, compliance frameworks, decision schemas, and knowledge bases.

Vertical Compliance Frameworks Status
Financial Services Basel III, MiFID II, SOX, GLBA, DORA 100%
Healthcare HIPAA, FDA SaMD, HITRUST, JCAHO 100%
Legal ABA Ethics, privilege gates, citation enforcement 100%
Insurance NAIC, Solvency II, bias/fairness engine 100%
Government FAR, FISMA, GPRA, FedRAMP architecture 100%
Energy NERC CIP, IEC 62443, FERC 100%
Manufacturing ISO 9001, IATF 16949, OSHA 100%
Defense CMMC, NIST 800-171, ITAR architecture 100%
EU Banking Basel III CRR/CRD, EU AI Act 100%
Sports FIFA Agent Regs, UEFA FFP, Premier League PSR 100%
+ 19 more Education, Pharma, Retail, Real Estate, etc. Framework

Infrastructure Integrations

All infrastructure components are opt-in -- disabled by default, zero impact when off. Every component has an embedded fallback for air-gapped deployment.

Component Purpose Activation
Apache Kafka Durable event streaming KAFKA_ENABLED=true
Temporal.io Durable workflow orchestration TEMPORAL_ENABLED=true
OPA Data-driven policy-as-code OPA_ENABLED=true
OpenBao/Vault Secrets management, PKI, transit encryption OPENBAO_ENABLED=true
NeMo Guardrails LLM input/output rail evaluation NEMO_GUARDRAILS_ENABLED=true
NVIDIA RAPIDS GPU-accelerated bias analysis & graph analytics RAPIDS_ENABLED=true
Confidential Computing GPU attestation, data-in-use protection CC_ENABLED=true
Apache Flink CEP Real-time complex event processing FLINK_ENABLED=true

Security

📄 Q1 2026 Platform Audit Report -- Comprehensive security audit across all repositories. 2 Critical, 6 High findings identified and all resolved with commit-level traceability. View the audit →

  • Sovereign-first -- Runs fully on-premise, air-gapped capable
  • Casbin RBAC/ABAC -- Role and attribute-based access control
  • OPA Policies -- GDPR, HIPAA, SOX, EU AI Act enforcement
  • Immutable Audit Ledger -- Merkle tree integrity, cryptographic signatures
  • Post-Quantum KMS -- Dilithium, SPHINCS+ support
  • Confidential Computing -- NVIDIA H100/H200 GPU attestation

Testing

# Backend tests
cd backend && npm test

# Frontend tests
npm test

# Type checking
npx tsc --noEmit --skipLibCheck          # Frontend
cd backend && npx tsc --noEmit --skipLibCheck  # Backend

Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

Quick rules:

  • PRs to any code in this repo are welcome
  • Follow Conventional Commits
  • Add tests for new functionality
  • Run npx tsc --noEmit before submitting

See COMMUNITY.md for the full open-source boundary definition.


Free Pilot Program

Datacendia uses an open-core model. This repository is the Community Edition (Apache 2.0, free forever).

We're offering a free 90-day guided pilot to qualified organizations navigating AI governance challenges (EU AI Act, HIPAA, Basel III, DORA, etc.). We'll help you deploy, configure for your industry, and prove value -- no commitment, no credit card.

What You Get Community (This Repo) Guided Pilot (Free)
Council Engine (multi-agent deliberation)
CendiaGateway (AI governance proxy)
Immutable Audit Ledger (Merkle-signed)
DCII Services (Truth, Notary, Witness, Timestamp)
CendiaReplay (decision playback)
30 Industry Vertical Frameworks
White-glove deployment assistance --
Custom agent configuration for your industry --
Priority support (Slack/email) --
Compliance report for your regulatory framework --

Request a Free Pilot · enterprise@datacendia.com

Looking for Enterprise or Strategic tier features? See TIER-MAPPING.md for the complete service-by-service breakdown, or contact us for pricing.


Standards & Governance

Datacendia implements the Datacendia Decision Governance Infrastructure (DDGI) -- a vendor-neutral framework for treating institutional decisions as auditable lifecycle artifacts.

Resource Description
DDGI Framework Vendor-neutral governance specification (CC BY 4.0)
DCII White Paper Reference implementation specification
API Specification OpenAPI 3.0 (59 endpoints)
Compliance Mapping Regulation-to-primitive matrix

DDGI is being prepared for submission to ISO/IEC JTC 1/SC 42 (Artificial Intelligence) as a New Work Item Proposal. See standards body engagement for details.


Troubleshooting

AI agents produce no output / Council deliberation is empty

Ollama must be running on your host machine and have at least one model pulled.

# Check Ollama is running:
curl http://localhost:11434/api/tags

# Pull a model if none are listed:
ollama pull llama3.2:3b

Demo mode connects to http://host.docker.internal:11434 -- this requires Ollama running on the host, not in Docker.

Docker build fails or is very slow
  • Ensure Docker has at least 6 GB RAM allocated (Docker Desktop → Settings → Resources)
  • First build downloads all npm dependencies and compiles TypeScript -- this takes 5-15 min
  • Subsequent builds use cache and start in seconds
Prisma migration fails
# If migrate deploy fails, try db push (works for fresh databases):
cd backend && npx prisma db push

The Prisma schema uses split files (requires Prisma 5+). Run npx prisma --version to verify.

Port conflicts

The dev stack uses non-standard ports to avoid conflicts with local services:

Service Port
Frontend 5173
Backend API 3001
PostgreSQL 5433 (not 5432)
Redis 6380 (not 6379)
Neo4j 7474 / 7687

If ports are still in use, edit the port mappings in docker-compose.dev.yml.


License

Apache License 2.0 -- See LICENSE for details.

Copyright 2024-2026 Datacendia, LLC


NVIDIA Inception Program Member

Built by Datacendia · DDGI Framework · NVIDIA Inception Program Member

Try the Platform · Request a Free Pilot · enterprise@datacendia.com

Last updated: April 2026