Analyse existing data
Turn samples and source schemas into reviewed reusable templates.
DataMaker can help propose templates from CSV or JSON samples, an OpenAPI document, a database schema or supported business-system metadata. The result is a starting point to review, not a guarantee that all formats, distributions and business rules were inferred.
From a file
Attach a sample in chat and state what you want to preserve:
Create a customer template from this CSV. Preserve the column names, use synthetic names and emails, and show ten generated rows for review. Do not save any real customer values as a reusable value list.
Inspect inferred types, null handling, nested structure and field relationships. CSV files are tables; they do not have spreadsheet sheets. For another format, ask the agent to explain any conversion it performs.
A PDF or an image can be attached too, and they are handled differently from data files. A PDF is read for the business fields it carries - dates, titles, reference numbers - so an invoice or a specification becomes a starting set of field names and example values. An image is described rather than parsed. Both produce a reading of a document, not a schema: check every extracted field against the real interface before building a template on it.
From a database
Select a configured source and identify the table. Review the analysis mode: direct-SQL templates read real source values, while relational synthetic templates propose generators from the structure. Inspect selected dependencies and foreign keys before accepting the proposal.
From OpenAPI, SAP or Oracle Fusion
Name the schema or entity you need and review required fields, keys and value restrictions. SAP uses OData metadata; Oracle Fusion uses describe resources. Not every semantic business rule is present in those documents.
Validate before reuse
Generate a small sample and check it against the receiving system or your test assertions. Correct categorical values, dates and relationships explicitly. Do not assume the analyser fitted a Gaussian or Pareto distribution or that automatic sensitive classification removed PII.
Save the reviewed template and retain the original specification when you need repeatability. See Schema discovery, Templates and Masking.