Convert JSON objects to TOML locally, using table sections or inline tables. Retain supported number tokens exactly and get clear errors for nulls, duplicate keys or integers outside TOML’s range.
Conversion happens locally in a disposable worker, with no uploads, accounts or saved input history. The shared JSON tokenizer retains source number tokens, while explicit compatibility checks prevent silent omission of unsupported data.
The root must be an object. Nested objects, arrays, strings, booleans and supported numbers are converted. TOML has no null type, so null is rejected at any depth with a JSON-style path. Root arrays/primitives, duplicate JSON keys and unpaired surrogate strings are also rejected rather than repaired or omitted.
Integer tokens are checked with BigInt and written without Number conversion, within TOML’s signed 64-bit integer range of -9223372036854775808 through 9223372036854775807. Larger integer tokens must be quoted in JSON to retain them as text. Fractional or exponent tokens are also written as authored, but TOML consumers generally interpret floats as binary64 and may round. Obvious float overflow or nonzero underflow to zero is rejected.
Table layout writes nested object sections with quoted paths, putting assignments before child sections to keep scope correct. Inline layout writes nested objects as inline tables. Arrays of objects use inline tables in both layouts, not double-bracket array-of-table sections. Empty objects remain tables and an empty root produces a comment-only TOML document. Output order may differ from input.
All keys are quoted, so dots, spaces, empty names and prototype-like names remain literal data. Strings are decoded from JSON and escaped as TOML basic strings; Unicode scalar values and control characters are handled explicitly. Date-like strings remain quoted strings and are never inferred as dates. Original escape spelling may change.
It converts supported data to a TOML 1.0-compatible subset, not the original JSON layout, spelling of string escapes or member order. Number tokens are retained within the documented numeric limits; a TOML reader may interpret their numeric type or precision differently. No JSON Schema, application configuration schema or runtime behavior is checked.
Input is limited to 200,000 UTF-16 units, 50,000 JSON tokens, 20,000 values and 100 nested containers. Output is limited to 1,000,000 units. A five-second deadline includes worker loading; cancel, edits and page exit terminate work and stale replies are ignored. Errors expose no partial output. JSON and TOML remain on your device, and files download as UTF-8 text.
