Skip to content

Safety Policy Engine -- declarative rules for automated safety validation

Safety Policy Engine — declarative rules for automated safety validation.

Module: replication.policy

Quick Start

from replication.policy import Operator

instance = Operator()

Enums

Operator

  • LT
  • LE
  • GT
  • GE
  • EQ
  • NE

Severity

  • ERROR
  • WARNING
  • INFO

Classes

Operator

Comparison operators for policy rules.

Method Description
evaluate() Return True if the rule passes (actual op threshold).
from_str() Parse an operator from a string.

Severity

Rule severity — determines overall verdict.

PolicyRule

A single safety rule: metric op threshold.

Method Description
evaluate() Return True if the rule passes.
render_expression() Human-readable rule expression.
to_dict()
from_dict()

RuleResult

Result of evaluating a single rule.

Method Description
icon()
status()
render()
to_dict()

PolicyResult

Aggregate result of evaluating all rules in a policy.

Method Description
passed() True if no ERROR-severity rules failed and no extraction errors occurred.
has_warnings()
has_extraction_errors() True if any rules failed to extract their metric value.
errors()
extraction_errors() Rules where metric extraction failed (misconfigured metric name).
warnings()
failures()
passes()
verdict()
verdict_icon()
exit_code() Exit code for CI/CD: 0=pass, 1=fail, 2=warn-only.
render()
to_dict()

SafetyPolicy

A named collection of safety rules that can validate simulation results.

Method Description
__init__()
from_preset() Load a built-in policy preset.
from_dict() Load a policy from a dictionary (parsed JSON/YAML).
from_file() Load a policy from a JSON file.
add_rule() Add a rule (fluent API).
evaluate() Evaluate all rules against a simulation report.
evaluate_with_mc() Run a simulation + Monte Carlo analysis, then evaluate all rules.
to_dict()
save() Save policy to a JSON file.

Functions

Function Description
main() CLI entry point for the safety policy engine.

CLI

python -m replication policy --help