Skip to main content

Architecture Overview

Compute Intelligence Graph (CIG) is a self-hosted and managed infrastructure intelligence platform. It provides visibility into cloud environments through automated discovery, graph-based relationship modeling, live/demo graph exploration, and AI-assisted analysis.

Core Philosophical Pillars​

  1. Discovery-First: Infrastructure state comes from discovery and graph ingestion, not from UI assumptions.
  2. Graph-Native Analysis: Relationships are first-class citizens. CIG stores "Resource A depends on Resource B" as data, not as a sidebar label.
  3. Human-in-the-Loop Intelligence: Chat, templates, and agents use the graph plus semantic retrieval to answer questions with actual infra context.
  4. Live/Demo Source Control: The Dashboard can switch between the real graph and a seeded demo workspace without changing the UI flow.

System Topology​

The system is organized into four primary functional layers:

1. Ingestion & Discovery Layer​

  • Discovery Orchestrator (@cig/discovery): Manages scheduling and lifecycle of discovery jobs.
  • Cartography Service (services/cartography): A Python-based service that pulls cloud inventory into the graph.
  • Graph Engine (@cig/graph): A Neo4j-backed layer that defines the schema, scope filters, and traversal logic.
  • Demo Workspace Seeder: Seeds a shared demo graph and semantic namespace for demo mode and local development.

2. Domain & API Layer​

  • Canonical API (@cig/api): A Fastify-based server providing:
    • REST: Resources, graph snapshots, discovery status, costs, security, actions, chat, demo, and node-management endpoints.
    • Graph queries: Read-only graph query execution plus controlled graph refinement proposals.
    • WebSocket: Real-time discovery and node status streaming.
    • Metrics: Prometheus integration for system health.
  • Auth Service (@cig/auth): Handles session management, JWT verification, and integration with Authentik and Supabase.

3. Intelligence & UX Layer​

  • CIG Dashboard (apps/dashboard): A Next.js 14+ application using the App Router. It serves as the primary visualization and management surface.
  • AI Agents (@cig/agents & @cig/chatbot):
    • Provides a RAG-based chatbot for infrastructure querying.
    • Uses specialized agents to perform graph reasoning, semantic retrieval, and controlled refinement proposals.
  • CLI (@cig-technology/cli): The public operator tool for local setup, demo provisioning, environment syncing, and direct API interaction.

4. Infrastructure & Runtime Layer​

  • IaC (@cig/iac): Terraform modules for networking, Neo4j, and shared AWS data-plane configuration.
  • Infra (@cig/infra): SST-based delivery mechanism for ECS/Fargate containers.

High-Level Data Flow​

Detailed Sections​