Skip to content

Overview — Posture Overview

The Posture Overview (/dashboard) is the central command center of Theseus Guard.
It synthesizes the Governance Health Index (GHI), active cryptographic provenance verification, Zero-Data AI Data Plane posture, and real-time operations across the four DREP lanes (Detect, Route, Enforce, Prove).

1. Governance Health Index (GHI) Hero Section

At the top of the Posture Overview, the GHI Hero Section delivers an immediate, executive-level health assessment of all guarded AI interactions across the tenant:

1.1 GHI Score & Posture Gauge

  • GHI Score (0–100): A unified compliance metric calculated from measured operational samples. A status of Available displays the calibrated score, while empty observation windows report Insufficient Data (no fabricated baseline scores).
  • Posture Assessment: Visual classification indicating organizational AI safety status (e.g., Good, Strong, Needs Review).
  • Sample Count & Trend: Displays the total volume of evaluated events in the active window alongside a trend indicator.

1.2 GHI Component Subscores

  • Drift Stability: Measures data and policy divergence relative to the active constitutional baseline.
  • Sentinel Models Consensus: Multi-model agreement rate among active L2/L3 evaluator models during guard adjudication.
  • Review SLA Rate: Percentage of critical and high-severity policy escalations reviewed by human operators within target SLA windows (with average response latency tracking, e.g., 12m avg).

1.3 Immutable Ledger Verification Chain (Last 15 Blocks)

Directly integrated into the GHI Hero banner is a live cryptographic chain visualizer:

  • Block Nodes: Renders the last 15 sequential ledger entries with SHA-256 hash linking.
  • Verdict Color Coding: Nodes reflect final enforcement outcomes (green for compliant/allowed, amber for reviews/watchpoints, red for blocks/violations).
  • Interactive Drill-Down: Clicking any block navigates directly to its full cryptographic proof in Prove → Immutable Ledger.

2. Zero-Data AI Data Plane (ZT-ADP) Posture Card

Positioned immediately below the hero section, the ZT-ADP Posture Card displays tenant-wide sensitive data protection status:

  • Enforcement Mode: Displays active ZT-ADP policy state (Disabled, Observe, or Enforce).
  • Protected Scopes: Real-time protection indicators across Prompt, Tool, MCP, and Skill interaction surfaces.
  • Format-Preserving Blinding: Status of automated redaction and synthetic surrogate generation across 100+ sensitive entity classes (PII, PCI, HIPAA, credentials).

3. Top-Level Operational KPIs

A responsive KPI grid highlights operational velocity and immediate administrative requirements:

KPI Card Description Primary Action / Meaning
Active Guards Total number of compiled Guard kernels actively enforcing runtime policies. Confirms policy deployment coverage across agents.
Pending Actions Queued human reviews and unresolved policy conflicts awaiting operator sign-off. Directly impacts operational SLA and queue latency.
Critical Alerts Unresolved high-priority system alerts (e.g., probe degradation, compliance breaches). Requires triage in Prove → Alert Center.
Guard Last Sync Timestamp of the most recent policy sync across active tenant agents. Ensures agents are enforcing current policy snapshots.

4. DREP Command & Control Grid

The lower half of the dashboard is organized into four purpose-built operational panels corresponding to the platform's core governance lanes:

4.1 DETECT Lane — Real-Time Telemetry

  • Governed Request Volume: An interactive time-series line chart tracking governed agent requests per minute across recent operational windows.
  • Live Anomaly Detection: Surfaces volume spikes and behavioral anomalies as they occur, providing a direct gateway into Detect → Live Event Stream.

4.2 ROUTE Lane — Critical Human Reviews

  • Human-in-the-Loop Triage: Displays real-time pending critical reviews with severity badges (critical, high, medium), timestamps, target guard domains, and model identifiers.
  • Direct Queue Access: Clicking any alert item opens the corresponding review side-sheet in Route → Runtime Reviews to approve, reject, or safe-degrade the interaction.

4.3 ENFORCE Lane — Runtime Enforcement Distribution

  • Enforcement Verdicts: A comparative bar chart breaking down governed traffic by final L2/L3 decision outcomes (Compliant vs. Non-Compliant).
  • Fast-Path & Interceptions: Measures runtime enforcement actions applied at inference time (blocks, rewrites, and safe-degraded fallbacks).

4.4 DATA PRIVACY Lane — PII Detection & Masking (OWASP LLM02)

  • Hourly Detection & Masking Chart: Dual-line time-series visualizing total sensitive entities detected versus successfully masked over time.
  • Privacy Posture: Confirms zero data leakage into upstream frontier LLMs, with time-window filtering (24H, 7D, 30D).

5. Controls & Live Refresh

  • Timeframe Selector: Toggle analytics across 24H, 7D, or 30D historical windows.
  • Configurable Auto-Refresh: Select continuous polling intervals (Off, 5s, 15s, 30s, 1m, 5m) for live command center displays.
  • Manual Sync: Trigger an immediate on-demand refresh of all dashboard queries and ledger chains.
  • Onboarding Banner: Dynamically surfaces configuration checklists (Policy Studio authoring, Guard compilation, Agent DNA registration, and API key provisioning) until tenant onboarding is complete.

6. Role-Based Dashboard Access

Role Key Focus on Posture Overview
Admin Tenant configuration health, ZT-ADP mode, sync timestamps, and onboarding progress.
Approver Pending review counts, SLA countdowns, and critical alert triage.
Compliance Officer GHI score, ledger verification chain, and GDPR/PII masking coverage.
AI / ML Engineer Governed request throughput, model consensus, and L2/L3 verdict distributions.
Auditor Content-free ledger continuity, immutable block records, and compliance metrics.