ArcaQ V1.1: From 15 to 17 Agents — OLM, OOC, Causal-HTG & One-Click Connectors

Why 17 agents matter

ArcaQ is not a chatbot. It is a sovereign decision-intelligence operating system: a constellation of specialized agents that collaborate under human governance to turn organizational knowledge into auditable, explainable decisions. With V1.1 we promote two capabilities that were previously implicit into full, first-class agents: OLM (Ontology Lifecycle Management) and OOC (Operational Ontology Contract).

This is not a marketing rebrand. Each agent is a separate deployable service with its own API, health checks, ReBAC policies, and event contract. Raising OLM and OOC to agent status means they now participate in the agent auction bus, publish audit events, and scale independently on Kubernetes.

OLM: Ontology-as-a-Product

OLM brings product-lifecycle discipline to enterprise ontologies. Every ontology is versioned, stateful, and traceable through a formal workflow:

1. DRAFT — proposals created by curators or autonomous agents.
2. IN_REVIEW — four-eyes validation by independent SMEs.
3. APPROVED — promoted to the production knowledge graph.
4. DEPRECATED — retained for historical queries, no new assertions.
5. ARCHIVED — immutable, auditable, and queryable via bi-temporal SPARQL.

The lifecycle is enforced by SHACL constraints and semantic contracts, so a deprecated class cannot silently power a live compliance report, and an unapproved property cannot leak into production inference.

OOC: Meaning as Code

Data contracts validate shape. Operational Ontology Contracts validate meaning. OOC lets domain experts write human-readable YAML that ArcaQ compiles into:

• OWL T-Box axioms — classes, property characteristics, logical restrictions.
• SHACL A-Box shapes — runtime validation rules for instance data.
• Multilingual definitions — native labels and definitions per locale.
• Lineage policies — retention, ownership, and deprecation rules.

The result is an enforceable contract that governs semantics across systems. A turbine sensor reading valid xsd:decimal is not enough; OOC defines what "operationalTemperature" means, how it relates to "thermalLimit", and what inferences are valid across plants, languages, and time.

Causal-HTG Orchestrator: From Correlation to Causation

The new Causal-HTG orchestrator sits between the Heterogeneous Graph Transformer (HGT) and the causal governance pipeline. It listens to platform events — KG triple changes, ontology updates, data-sufficiency thresholds — and orchestrates:

• Auto-HGT triggers: debounced retraining when the graph topology changes.
• Auto-causal triggers: PC/GES/FCI discovery when datasets cross statistical thresholds.
• P1/P2/P3 promotion funnel: candidate edges progress from machine suggestion to SME-certified fact.
• Causal quarantine: disputed edges are isolated, routed to reviewers, and never enter production until certified.

Every stage emits events such as hgt.candidate.detected, causal.discovery.completed, causal.edge.quarantined, causal.edge.certified, and sme.review.overdue, so the entire pipeline is observable, auditable, and interruptible by humans.

One-Click Connectors: Zero-Friction Integration

ArcaQ-Connect now ships with a catalog of pre-configured connectors for the tools enterprises already use: Microsoft Teams, Slack, Jira, GitHub, Salesforce, Confluence, PostgreSQL, MySQL, Redis, Kafka, S3, SAP, Prometheus, and more.

Each connector supports three activation modes:

• Direct API key for fast, self-contained access.
• OAuth for delegated, revocable enterprise authorization.
• MCP (Model Context Protocol) for agent-driven interaction, validated against an allowlist and falling back to HTTP/SSE automatically.

Sources are auto-discovered, auto-labeled, and tagged with connection state — so users see at a glance whether a source is discovered, testing, active, or degraded without touching a configuration file.

Session Reliability You Can Trust

Sovereign AI is useless if users lose their session in the middle of a critical review. V1.1 hardens the UI layer with:

• A reusable _authFetch wrapper that refreshes tokens on 401, retries once, and redirects to /login?error=session_expired if refresh fails.
• Global HTMX interceptors for 401/403 responses in base.html.
• A visible "Session expired" toast and safe fallback for the Cmd+K overlay when DOM elements are absent.
• Null guards on every getElementById and querySelector call in dashboard scripts.

AIEO/SEO Refresh: Made for AI Search Engines

As AI crawlers become gatekeepers of discoverability, we have refreshed every public page for AI Engine Optimization (AIEO) and classic SEO:

• Optimized <title> and meta descriptions under recommended character limits.
• Open Graph and Twitter Cards on every content page.
• JSON-LD Article, TechArticle, BreadcrumbList, and SoftwareApplication structured data.
• Semantic heading hierarchy with one h1 per page.
• Alt text on informative images and empty alt on decorative ones.
• Updated ai-ready.json, ai.txt, and llms.txt to reflect the 17-agent constellation and new capabilities.

Availability

ArcaQ Enterprise V1.1 is deployed in production and available for new on-premise installations. All 26 patent claims remain validated, all processing stays inside customer infrastructure, and air-gap deployments are supported without GPU requirements.

Read the deep-dives: OLM — Ontology Lifecycle Management, OOC — Operational Ontology Contracts, Causal-HTG Orchestrator, and Connectors for Autonomous Agents.