Evaluate Condition • AI Agent SDK

evaluate_condition: Evaluate a condition expression with variable substitution.

evaluate_condition

Function

This function is defined in the evaluator module.

Evaluate a condition expression with variable substitution.

This is the shared condition evaluation function used by both AgentFlow and AgentTeam. It supports various condition formats.

Supported formats:

• Numeric comparison: “{{var}} > 80”, “{{var}} >= 50”, “{{var}} < 10”
• String equality: “{{var}} == approved”, “{{var}} != rejected”
• Contains check: “error in {{message}}”, “{{status}} contains success”
• Boolean: “{{flag}}” (truthy check)
• Nested property: “{{item.score}} >= 60”

Signature

def evaluate_condition(condition: str, variables: Dict[str, Any], previous_output: Optional[str]) -> bool

Parameters

condition · str · required

Condition expression with {{var}} placeholders.

variables · Dict · required

Dictionary containing variables for substitution.

previous_output · Optional

Optional output from previous step (substitutes {{previous_output}}).

Returns

Returns · bool

Boolean result of condition evaluation. Returns False on evaluation errors (fail-safe).

Usage

result = evaluate_condition("{{score}} > 80", {"score": 90})  # True
    result = evaluate_condition("{{status}} == approved", {"status": "approved"})  # True
    result = evaluate_condition("error in {{message}}", {"message": "An error occurred"})  # True

Uses

• re.findall
• get_nested_value
• re.match
• numeric_match.group
• string_match.group
• logger.warning

Used By

• ExpressionCondition.evaluate
• Task.evaluate_when

Source

View on GitHub

praisonaiagents/conditions/evaluator.py at line 131

---

Related Documentation

Evaluation Concept

LLM as Judge

Evaluation Loop