Member Junction
    Preparing search index...

    Interface ExperimentRunOptions

    Tunables for an orchestration run. All optional with sensible deterministic defaults; tests vary them to drive wave/prune/budget behavior precisely.

    interface ExperimentRunOptions {
        agentRunID?: string;
        budget?: Budget;
        concurrency?: number;
        experimentID?: string;
        keepTopK?: number;
        relativePruneThreshold?: number;
        sessionName?: string;
    }
    Index

    Properties

    agentRunID? budget? concurrency? experimentID? keepTopK? relativePruneThreshold? sessionName?

    Properties

    agentRunID?: string

    Optional MJ: AI Agent Runs id that owns/drives this session (plan §9.1).

    budget?: Budget

    Optional override of the session's budget (otherwise taken from ModelingPlanSpec.ProposedBudget). Useful for tests.

    concurrency?: number

    Max iterations executed concurrently within a wave (bounded concurrency, §8.3). Also bounds the default wave size. Defaults to DEFAULT_WAVE_CONCURRENCY.

    experimentID?: string

    Optional pre-resolved MJ: Experiments id to create the session under. When omitted, the orchestrator creates a new durable Experiment definition.

    keepTopK?: number

    Keep the top-K iterations on the leaderboard after each wave; iterations ranked below K are pruned (their Status set to Pruned). When omitted, no top-K cap is applied (only the relative-threshold prune, if set, runs).

    relativePruneThreshold?: number

    Prune any completed iteration whose Score is below bestScore * relativePruneThreshold (a fraction in [0,1]). E.g. 0.8 prunes anything scoring under 80% of the current best. When omitted, no relative prune is applied.

    sessionName?: string

    Optional human-readable session name (defaults to a name derived from the plan goal).

    Settings

    Member Visibility

    On This Page

    Properties