Member Junction
    Preparing search index...

    Function DecisionConfigurationProblems

    • What is wrong with a Decision step's settings, as phrases naming each problem. Empty when it can run.

      The checks a task-graph Decision node gets at submit, exported so a Flow agent's Decision step is held to exactly the same ones rather than to a copy that drifts. Absent questions is not reported here — a task graph reports it as missing configuration — so a caller that requires them says so.

      Parameters

      • config: {
            promptName?: string;
            questions: Record<string, TaskGraphDecisionQuestion>;
            state?: string;
        }
        • OptionalpromptName?: string

          The MJ: AI Prompts row whose model bindings run the decision. Default Default Decision, the prompt agent decisions use.

        • questions: Record<string, TaskGraphDecisionQuestion>

          The questions, all answered in one call. The key is how conditions name the answer (decisions.<tempId>.<key>); the model never sees it, so put everything it needs in instructions and the option descriptions.

        • Optionalstate?: string

          What the questions are about. payload is the node's whole input: its own inputPayload merged with everything upstream produced. payload.<path> is one named value from it, typically an upstream step's output (payload.ticket). Default payload.

      Returns string[]