Compass UI Guide

Visual walkthrough of the Federal Frontier Compass interface — navigation, storage dashboard, AI chat, ontology browser, and graph visualization.

Compass UI Guide

This guide walks through the Compass user interface as seen by a platform operator. Compass is accessed at https://compass.vitro.lan and provides a unified view of all Federal Frontier Platform infrastructure.

Top Navigation Bar

The top navigation bar provides access to all major sections:

Ontology Browser (Left Sidebar)

The left sidebar displays the Digital Twin Browser — a hierarchical tree view of all entity types in the FFO ontology, organized by domain:

Infrastructure Domain

• Cluster — Kubernetes clusters (e.g., vitroai-fmc)
• Node — Worker and control-plane nodes
• Network — Cluster networking configuration
• StorageClass — Kubernetes storage classes
• Registry — Container image registries (e.g., Harbor)
• Image — Container images tracked in the ontology

Security Domain

• Finding — Security findings from scanners (Trivy, compliance checks)
• Vulnerability — CVEs linked to container images
• Threat — Threat intelligence entries
• Mitigation — Applied mitigations for findings

Compliance Domain

• Control — NIST 800-53, CMMC, and other framework controls
• Framework — Compliance frameworks (NIST, CMMC, FedRAMP)
• Assessment — Compliance assessment results

Workload Domain

• Deployment — Kubernetes deployments
• Service — Kubernetes services
• Pipeline — CI/CD pipelines
• Database — Managed databases

Identity Domain

• Principal — IAM users (Keycloak principals)
• IAM Role — Role definitions
• Group — User groups
• Service Account — Kubernetes service accounts

Each entity type shows a count and can be expanded to list individual entities with their attributes.

Ceph Storage Dashboard

The Storage tab provides a comprehensive view of the Ceph distributed storage cluster:

Cluster Overview

• Cluster name and FSID — e.g., “VitroAI Quincy Ceph” with unique FSID
• Storage capacity bar — Visual gauge showing used vs. total (e.g., 254.27 GB / 1.46 TB, 17.0% used)
• Available space — Remaining capacity at a glance

Summary Cards

Three summary cards provide instant status:

• Pools — Total number of storage pools (e.g., 5)
• OSDs Up — Object Storage Daemons status (e.g., 3/3 up)
• Monitors — Ceph monitor count (e.g., 3)

Storage Pools Table

Example pools in a typical deployment:

Object Storage Daemons (OSDs)

Each OSD is displayed as a card showing:

• OSD ID (e.g., osd.0, osd.1, osd.3)
• Status — “up” with green indicator
• Device Class — Storage device type (hdd, ssd, nvme)
• Weight — CRUSH weight for data distribution (typically 1.0000)

Ceph Monitors

Monitor nodes are listed with their addresses, state (leader, peon), and health status.

Health Banner

A prominent health indicator appears at the top:

• HEALTH_OK (green) — All components healthy
• HEALTH_WARN (yellow) — Warnings present (e.g., undersized PGs, clock skew)
• HEALTH_ERR (red) — Critical issues requiring attention

AI Chat Interface

The Chat panel (right side of the screen) provides the Federal Frontier AI Assistant — a conversational interface powered by LLM-based tool calling.

How It Works

1. Type a natural language question in the input box
2. The system either matches a query template (instant response) or routes to the LLM with 150+ MCP tools
3. Results are formatted as markdown tables for readability
4. A copy button on each response lets you copy the content to clipboard

Example Queries

Infrastructure queries:

• “List the containers on hyperv-1” → Calls kolla_list_containers → Returns table with container name, image, status
• “How many clusters do we have?” → Matches template → Queries FFO TypeDB
• “Show me the nodes” → Matches template → Returns node names and types

Storage queries:

• “Ceph health” → Matches template → Returns cluster health, capacity, pool status
• “List OSDs” → Matches template → Returns OSD names, status, device class
• “How is the storage?” → Matches template → Returns capacity breakdown

Identity queries:

• “List users in FAS realm” → Matches template → Calls Keycloak MCP → Returns user table
• “Who are the users?” → Matches template → Returns principal list from FFO

Operations queries:

• “List MCP servers” → Matches template → Returns table of all 12 servers with tool counts
• “Kolla services” → Matches template → Calls Kolla MCP → Returns service-to-container mapping
• “Show ArgoCD apps” → Routes to LLM → Calls ArgoCD MCP tools

Response Formatting

All tabular results render as proper HTML tables with:

• Sortable column headers
• Row highlighting on hover
• Responsive horizontal scrolling for wide tables
• Dark theme styling consistent with the Compass UI

Chat Architecture

User query
    │
    ▼
Template matcher (keyword patterns)
    │
    ├── Match found → Execute directly (instant)
    │   ├── TypeQL query → FFO TypeDB
    │   ├── API call → /integrations/servers
    │   └── MCP tool → JSON-RPC to MCP server
    │
    └── No match → LLM path
        ├── Load 150+ tools from DB + MCP_SERVERS dict
        ├── Send to Ollama/vLLM with tools array
        ├── LLM returns tool_calls
        ├── Execute tool calls against MCP servers
        └── Format results as markdown tables

Instance Graph (Explorer Tab)

The Explorer tab renders an interactive graph visualization using ReactFlow:

• Nodes represent FFO entities (clusters, nodes, deployments, findings, etc.)
• Edges represent TypeDB relations between entities
• Color coding differentiates entity types at a glance
• Click any node to inspect its attributes in a detail panel
• Zoom/pan to navigate large graphs
• Limited to 100 nodes to maintain performance

Schema Browser

The Schema tab provides a read-only view of the FFO TypeDB schema:

• Entity types with their attributes and value types
• Relation types with their role players
• Inheritance hierarchies
• 7 domains, 28 object types as reported by the schema browser

Integrations Panel

The Integrations tab shows all registered MCP servers:

• Server name, description, URL, tool count, enabled/disabled status
• Currently 12 servers with 150+ tools total
• Servers can be enabled/disabled without redeployment
• Tool discovery happens at startup from the Postgres mcp_servers table