Import a sample

Infer a GBDDL recipe from sample data, then review and edit the result. Provide a JSON sample through the import journey. Inference maps its structure and values to a GBDDL text recipe; review every inferred physical and smart type before using it. Inference is a starting point and does not store or charge for the sample. GBDDL validation and pricing are shared across Chef, the HTTP API and MCP. Consult the recipe language, output formats, and credit model for the supported syntax and behavior.

API contract

Call POST /api/v1/gbddl/infer?input_format=json&output_format=json with an object containing sample (the JSON value to inspect). Choose the desired supported data format through output_format. Optional types maps table.column names to smart types that override inference. API keys need recipes:write.

The response includes GBDDL recipe text, column reports, warnings and validation errors. Inference does not save a recipe or charge credits. Review the result, validate it and explicitly save or submit it afterward. Inference derives types from sample values; it does not learn statistical distributions from a production dataset.