Architecture - orcadevstack/orcaopta GitHub Wiki

Orcaopta Architecture

Orcaopta is structured as a modular, multi‑cloud reliability platform. Each subsystem contributes signals, audits, or actions to the unified cloud graph and healing controller.

High-Level Components

1. FastAPI Backend

Provides:

  • ML API endpoints
  • AI reasoning endpoints
  • Dashboard data endpoints
  • MCP server integration

2. AI Reasoning Engine

Located in src/orcaopta/ai/agent.py

  • Generates healing plans using LLM reasoning
  • Consumes unified cloud graph
  • Maps issues to remediation actions

3. Unified Cloud Graph Engine

Located in src/orcaopta/cloud/graph.py

  • Aggregates OpenStack, OVN, Ceph, Kubernetes, Terraform, ML, and RL signals
  • Produces a global topology representation

4. Self-Healing Controller

Located in src/orcaopta/controller/self_heal.py

  • Executes continuous healing cycles
  • Applies remediation actions
  • Logs healing events

5. ML & RL Engines

Located in src/ml/ and src/rl/

  • Anomaly detection
  • Forecasting
  • Autoscaling decisions
  • PPO RL agents

6. Dashboard

Located in src/dashboard/app.py

  • Cloud graph visualization
  • Healing timeline
  • AI plan viewer

7. Terraform, OpenStack, Kubernetes, OVN Audits

Located in:

  • src/orcaopta/cloud/openstack/
  • src/orcaopta/cloud/kubernetes/
  • src/orcaopta/cloud/terraform/
  • src/orcaopta/cloud/ovn/

Each module detects issues and provides healing actions.

Data Flow

  1. Collect signals
  2. Build cloud graph
  3. AI generates healing plan
  4. Controller executes actions
  5. Events logged
  6. Dashboard displays results