Reviewed 2026-08-28

Build a Useful JSON Schema from Real Samples

Infer a draft, then replace accidental observations with deliberate rules.

{"id":42,"tags":["admin"]}

{"type":"object","properties":{"id":{"type":"integer"}}}

Reproducible method

1. Preserve the original and create the smallest representative sample.
2. Run a strict parse or validation before transformation.
3. Record options, compare counts and structure, then test in the receiving system.

Common error

One sample cannot reveal optional, nullable, or invalid cases.

Reduce a failure while retaining the problematic structure; this separates malformed input from unsupported behavior.

Technical limitation

Inference cannot discover undocumented business constraints.

A successful preview does not remove format ambiguity or downstream requirements.

Security and privacy

Use synthetic samples and require human schema review.

Local processing reduces transfer risk but cannot protect a compromised browser, unsafe extensions, clipboard history, or later misuse.

Verification checklist

• Check field, record, page, or byte counts.
• Review edge cases and error output.
• Retain the original until acceptance.
• Document assumptions for automation.

Open the related tool