Class RunClusterAnalysisAction

Action: "Run Cluster Analysis"

Thin wrapper over ClusteringEngine.RunPipeline. Fetches embedding vectors for an entity / entity document, clusters them (KMeans or DBSCAN), projects them to 2D/3D, optionally names clusters with an LLM, and — when a PersistAs name is supplied — persists the result to MJ: Cluster Analysis.

All business logic lives in the engine; this action only maps params and returns a summary.

Parameters (Input):

• EntityID: ID of the entity to cluster (alternative to EntityDocumentID).
• EntityDocumentID: specific Entity Document supplying vectors.
• Algorithm: 'kmeans' (default) or 'dbscan'.
• K: number of clusters for KMeans (default 4).
• Epsilon: DBSCAN neighbourhood radius (default 0.3).
• MinPoints: DBSCAN minimum points (default 3).
• DistanceMetric: 'cosine' (default) | 'euclidean' | 'dotproduct'.
• MaxRecords: max records to fetch (default 500).
• Filter: optional SQL filter applied to the source records.
• NameClusters: 'true' to run LLM cluster naming (default false).
• PersistAs: when set, the analysis is saved with this name.

Output params: AnalysisID (when persisted), ClusterCount, RecordCount, SilhouetteScore.