Elysia Tools
Navigation
Workflow Playbook
Derive a reviewable JSON Schema draft from representative JSON samples or XLSX rows, then inspect types, requiredness, nulls, and outliers before treating it as a contract.
Hubs
Schema inference is most useful when the original contract is incomplete but the data shape is visible. Gather several JSON records or a workbook sheet with representative rows, including missing fields, nulls, empty values, boundary cases, and known anomalies. Keep the source files and the exact sheet or sample set with the draft.
Use json-schema-generator for JSON examples. Use xlsx-sheet-to-json-schema when the evidence starts with worksheet headers and rows. When both sources describe the same object, produce comparable drafts instead of treating one format as automatically authoritative.
A repeated field is not automatically required, and a numeric-looking value may be an identifier rather than a number. Check each field's type, requiredness, nullability, enum candidates, date format, units, and unusual values. Blank spreadsheet cells and mixed JSON values deserve their own review notes.
The accepted output is a reviewable schema draft with source evidence and explicit assumptions. After approval, it can feed the json-validate or json-to-typed-code-generators workflow. Until then, it is evidence about observed data, not the final business contract.
Workflow playbook
Separate representative records from one-off anomalies, record the JSON files or workbook sheet and sample range, and preserve the original inputs before inference.
Run the JSON branch over more than one meaningful example and retain the generated draft together with the sample set, observed types, and any manual schema edits.
Use the header row and representative XLSX rows to derive property names, primitive types, nullable behavior, enum candidates, and basic constraints; note blank cells and formatting traps.
Compare the branch outputs field by field, rerun the chosen inference after deliberate edits, and record decisions for types, requiredness, nulls, and outliers before handing the draft to validation or model generation.
Use json-schema-generator for JSON examples and xlsx-sheet-to-json-schema for worksheet headers and rows. If both describe the same payload, preserve both drafts and reconcile differences explicitly rather than silently choosing one.
No. Sample frequency is evidence about the observed data, not a business rule. Mark requiredness only after checking optional records, missing columns, blank cells, and the producer's expectations.
Keep the generated output as a draft until types, nullability, enum candidates, constraints, and outliers have been reviewed against representative inputs and documented assumptions.