Evaluation Engine

The evaluation engine runs client-side (in the SDK) after fetching the flag configuration from the API. There's no server round-trip per evaluation — the logic runs locally.

Evaluation flow

When you call getFlag(key, context), the engine follows this decision tree:

1. No environment config — if the flag has no configuration for this environment, return the default value with reason: 'not_found'
2. Disabled — if the flag is disabled in this environment, return false for boolean flags or the default value for other types, with reason: 'disabled'
3. Evaluate rules — iterate through targeting rules in order. If a rule matches, return its returnValue with reason: 'rule_match'
4. Default — if no rules matched and the flag is enabled:
   • Boolean flags return true (enabled = on)
   • Other types return the flag's defaultValue
   • Reason is 'default'

Example

Given a boolean flag new-checkout that is enabled with two rules:

Evaluating with different contexts:

Pages

• Operators — all 10 operators with examples
• Percentage Rollouts — consistent hashing algorithm
• Rule Ordering — first-match-wins semantics