Input mapping load

Loads tables from Storage into a Workspace using a keboola/input-mapping Configuration\Table payload (snake_case) accepted unchanged. This mirrors what component code already produces and removes the need to translate to the SAPI camelCase /load shape.

The endpoint resolves the load type per input item via the canonical LoadTypeDecider with priority CLONE > VIEW > COPY. CLONE is gated by the bq-enable-clone feature on BigQuery; VIEW on Snowflake is gated by the input-mapping-read-only-storage feature.

Each input item must reference its source via source — the full source table id (e.g. in.c-bucket.mytable). where_column and where_values must be provided together (setting one without the other returns HTTP 400).

Dry-run (dryRun=true): preflight variant that resolves the load type per input item without creating a job. The dry-run runs the same validations the real load runs — a successful 200 dry-run guarantees the same payload submitted with dryRun=false will not be rejected by validation (modulo source state changing between calls). It returns HTTP 200 (not 202) with loadType (the type LoadTypeDecider would pick) and possibleLoadTypes (all types acceptable for the item; useful for driving a UI dropdown) per input item.

This operation is also available scoped to a Development Branch, at /branch/{branchId}/workspaces/{workspaceId}/input-mapping-load. The dryRun query parameter and both response shapes behave identically there.