register_model

Scope: builder · Signature: register_model(action: {register|record_release|import_predictions}, experiment_id: str | None = None, workspace_id: str | None = None, params: dict[str, Any] | None = None, idempotency_key: str | None = None, model_id: str | None = None) -> Any

Use to register a model, record what a person approved for a release, or import predictions.

action="register": experiment_id, params name, notes (a short description), risk_level (a person's answer only), release_record. action="record_release": params release_record on model_id (a registered model) or experiment_id (a run never registered: association, recommendation, learning-to-rank, transductive clustering). release_record: {question_id (the release gate's id from ask_gate), operating_point {summary, details, apply}, challenge {verdict, conditions [{text, answer, fallback}], document_id}, limitations [..]}; apply = {kind: association_shortlist, rule_ids} | {kind: cluster_labels, labels {id: name}} | {kind: decision_layer | anomaly_threshold | ranking_operating_point}; omit it when no surface carries the point (documented only). action="import_predictions" (last resort — get_playbook("byo")): workspace_id, params name, csv_text or file_path, reference_dataset_id, target, prediction_column, task_type, label_column, positive_class, sensitive_features. Promotion is a person's.