prepare_data

Scope: builder · Signature: prepare_data(dataset_id: str, action: {columns|eda|leakage|triage|cancel}, params: dict[str, Any] | None = None, idempotency_key: str | None = None) -> Any

Use to make a dataset ready to train: column roles, EDA, leakage triage.

action="columns": params target_column, id_column (the row key), columns {name: {is_ignored, is_key, is_fairness, is_base_time, type, domain, format, delimiter, description} | null to reset}, merged into the current settings. domain: describe_platform("domains") lists each (what it makes, alone and with a companion); mt = tag list, wt = weighted tokens, with delimiter comma | whitespace | semicolon | pipe | tab. get_eda(section= "columns") has the value checks and suggestions. "eda": compute the profile (target, time_column, sample_size, refresh). "leakage": rescan (target, time_column). Both run on the server; one still running after ~20 s returns status queued/running and a wait_handle (wait on it, then read with get_eda); "cancel" stops one that runs far too long. "triage": finding_id, acknowledged, note (why it is not a leak) — a PROPOSAL a person confirms on the Leakage tab. Pass a returned idempotency_key back only to retry a call whose response was lost.