What is a Supervisor?

A supervisor monitors child agents and restarts them when they fail. Inspired by Erlang/OTP supervision trees.

Why Supervisors?

In production, agents crash. Network timeouts, API errors, out-of-memory — failures are inevitable. Supervisors ensure your system recovers automatically.

Without Supervisors

Agent crashes → Nobody notices → System broken

With Supervisors

Agent crashes → Supervisor detects → Agent restarted → System healed

How It Works

1. You create a supervisor with a restart strategy
2. You add child agents to the supervisor
3. The supervisor monitors all children
4. When a child fails, the supervisor restarts it according to the strategy

# Create supervisor
naos supervisor create main --strategy one-for-one

# Add children
naos supervisor add-child main researcher
naos supervisor add-child main analyzer

# Check status
naos supervisor status main

Supervisor Properties