AI-propose every mapping document a pipeline default needs from one pasted sample; with config_values, re-check those documents with the reader's pins instead (no model call).
Authorizations
Bearer authentication header of the form Bearer <token>, where <token> is your auth token.
Body
Mapping-default pipeline to propose for, e.g. jsonl_mapping_default or jsonl_mapping_business_default.
Client-dialect schema inferred from the pasted records.
One pasted record; evaluated on the server for the preview.
Send the sample's values to the model as well as its structure.
step -> alias -> target pointer -> the reader's own binding, which wins over the model.
step -> alias -> mapping/1 document from an earlier proposal; re-checked with pins, no model call.
Response
Successful Response
step -> alias -> mapping/1 document, ready for furnish.viaSample.
Required target fields nothing binds; the mapping cannot be saved until each is bound.
Fields the model noted it could not bind that the target does not require.
Every step produced a document free of problems besides missing fields.
Each mapping step's target, roots, pins and target models, so the proposal can be drawn on the graph.
What went wrong, when verified is false.