Member Junction
    Preparing search index...

    Thresholds for clustering ad-hoc task graphs and identifying promotion candidates.

    interface PromotionDiscoveryOptions {
        clusterSimilarityThreshold?: number;
        evolutionMinSimilarity?: number;
        existingWorkflowMatchThreshold?: number;
        fingerprintOptions?: DAGFingerprintOptions;
        maxGraphsToCluster?: number;
        minClusterSize?: number;
        minNodesPerGraph?: number;
        minSuccessRate?: number;
        similarityWeights?: DAGSimilarityWeights;
    }
    Index

    Properties

    clusterSimilarityThreshold?: number

    Minimum pairwise similarity to group two executed graphs into the same cluster. Default 0.75.

    evolutionMinSimilarity?: number

    Lower similarity bound [evolutionMinSimilarity, existingWorkflowMatchThreshold) where a cluster is flagged as a candidate evolution/v2 of an existing saved Workflow. Default 0.40.

    existingWorkflowMatchThreshold?: number

    Similarity threshold above which a cluster is considered a duplicate of an already-saved Workflow (indicating a routing/discovery gap rather than a net-new workflow). Default 0.85.

    fingerprintOptions?: DAGFingerprintOptions
    maxGraphsToCluster?: number

    Maximum number of executed task graphs to cluster in a single invocation (most recent runs are prioritized when the input exceeds this cap). Default MAX_CLUSTERING_GRAPHS (500).

    minClusterSize?: number

    Minimum number of executions in a cluster to qualify for promotion. Default 3.

    minNodesPerGraph?: number

    Ignore trivial 1-node graphs when looking for multi-step workflow promotions. Default 2.

    minSuccessRate?: number

    Minimum success rate across cluster members to qualify for promotion. Default 0.80.

    similarityWeights?: DAGSimilarityWeights