Compass Overview

Federal Frontier Compass — the digital twin UI for infrastructure visibility, AI-powered queries, and graph visualization across the Federal Frontier Platform.

Compass is the Federal Frontier Platform’s digital twin user interface. It provides platform operators with a unified view of infrastructure state by querying the Federal Frontier Ontology (FFO) — a TypeDB knowledge graph that models clusters, nodes, deployments, services, findings, images, and their relationships.

URL: https://compass.vitro.lan

Current versions: API v5.13.79 Frontend v5.1.17

Capabilities

Compass provides five primary capabilities for platform operators:

Capability Description
AI Chat Natural language queries against infrastructure state, powered by Ollama and 150+ MCP tools
Instance Graph Interactive ReactFlow visualization of FFO entities and relationships
Storage Dashboard Ceph cluster health, pool usage, OSD status, and monitor visibility
Workitem Tracking View and manage workitems associated with deployments and findings
MCP Integration Access to 12 MCP servers exposing 150+ tools across the platform

Architecture

Compass is composed of two independently deployed components:

graph TD FE[Compass Frontend
Next.js + BPJS
Port 3000/nginx] --> API[Compass API
FastAPI
Port 8000] API --> DB[TypeDB
FFO] API --> MCP[12 MCP Servers
150+ tools] style FE fill:#2d3748,stroke:#4299e1,color:#e2e8f0 style API fill:#2d3748,stroke:#4299e1,color:#e2e8f0 style DB fill:#2c7a7b,stroke:#38b2ac,color:#e2e8f0 style MCP fill:#1a365d,stroke:#4299e1,color:#e2e8f0

Frontend

The frontend is a Next.js application using BlueprintJS as the component library. It renders the chat interface, instance graph, storage dashboard, and workitem views. The frontend communicates exclusively with the Compass API — it never queries TypeDB or MCP servers directly.

API

The Compass API is a Python FastAPI application that serves as the backend for all Compass functionality. It handles:

  • Chat processing — Pattern-matching queries for instant results, LLM-based queries for everything else
  • TypeDB queries — All FFO reads and writes go through the API’s TypeDB driver connection
  • MCP tool invocation — The API maintains connections to 12 MCP servers and orchestrates tool calls on behalf of the LLM
  • PostgreSQL lookups — MCP server registry and metadata stored in PostgreSQL

MCP Server Integration

Compass integrates with 12 MCP servers that collectively expose over 150 tools. The MCP server registry is stored in PostgreSQL (mcp_servers table) with a hardcoded fallback in mcp_tools.py. When the LLM determines it needs external data to answer a query, it selects appropriate tools from the available set and the API executes the calls.

The MCP servers span infrastructure domains including:

  • FFO (ontology queries)
  • Keycloak (identity and access)
  • OpenStack / Kolla (compute and networking)
  • Ceph (storage)
  • Kubernetes (workload orchestration)
  • ArgoCD (GitOps)
  • And others