Policy DSL (Planned v0.2)

YAML-based rule engine for declarative supervisor behaviour.

Status

Not yet implemented. The current supervisor uses hardcoded SupervisorPolicy. This spec describes the planned replacement.

Syntax

rules:
  - name: suppress_retry_storm
    regime: [NOMINAL, DEGRADED]
    condition:
      metric: R_bad
      layer: 1
      op: ">"
      threshold: 0.7
    action:
      knob: alpha
      scope: layer_1
      value: 0.3
      ttl_s: 5.0

  - name: restore_coordination
    regime: [DEGRADED, RECOVERY]
    condition:
      metric: R_good
      layer: 3
      op: "<"
      threshold: 0.3
    action:
      knob: K
      scope: global
      value: 0.1
      ttl_s: 10.0

Fields

Field	Type	Description
name	str	Rule identifier
regime	list[str]	Active only in listed regimes
condition.metric	str	R_good, R_bad, or any boundary variable
condition.layer	int	Layer index for R metrics
condition.op	str	">", "<", ">=", "<=", "=="
condition.threshold	float	Comparison value
action	object	ControlAction fields

Evaluation

Rules are evaluated in order. First matching rule fires. No chaining within a single step -- at most one rule fires per step per knob.

Regime Scoping

A rule only fires if the current regime is in its regime list. This prevents conflicting actions across regimes.

Integration

The binding spec references a policy file:

policy: policy.yaml

Resolved relative to the binding spec directory.