Skip to content

ai_model

Data and lifecycle of the mail AI verdict. The ai_inference container does the classification; the ai_model connector (off by default) calls it; this app stores each case's AiAssessment, the label ledger the model is trained on, the registry of model versions, and the retrain, shadow, promotion and rollback workflow.

  • Models: AiAssessment, LabelRecord, LabelSyncState, ModelVersion, ModelEvent, RetrainJob, ShadowResult, ShadowCampaignEffect, ShadowReview.
  • REST: /api/ai-model/* (Admin and CISO read, Admin write).
  • Commands: ai_model_retrain, ai_model_backfill_labels, ai_model_reembed_ledger, ai_model_campaign_replay, ai_model_engine_replay, ai_model_text_gap (see the management commands reference).
  • Beat: ai-model-retrain-check, ai-model-materialise-stats, ai-model-server-sync and the connector's label-sync.

Full description of models, modules, endpoints and commands: Suspicious/Suspicious/ai_model/README.md. Connector configuration: connectors. Setup and troubleshooting: AI / ML Analyzers. Operator view: Profile & Settings.