PIXELOKAY / DATA & AI TOOLS

Data Schema Mapper

Compare two data structures, discover likely field matches, customize transformations, and deliver clean, validated output—without sending a record to a server.

✦ Smart Match + optional local AI◈ JSON / CSV / TSV / JSONL✓ Conversion diagnostics↓ Mapping & data export

YOUR DATA WORKSPACE

AI Data Schema Mapper

Import → Smart Match → Review → Validate → Export

Source records—Input dataset

Target fields—Destination schema

Mapped fields—Approved transformations

Preview errors—First preview rows

01 / SOURCE

Where data comes from

INPUT

Your records to transform. Nested JSON objects are supported; arrays are kept as values.

↑Drop your source file hereCSV, TSV, JSON, JSONL · 8 MB · 10,000 records

Or paste a dataset

No source loaded

Field quality summary appears after import.

02 / TARGET

Where data should go

OUTPUT

Import an example of the destination format. Field types and required status can be edited later.

↓Drop a target sample hereAt least one example record or column header

Or paste a target example

No target loaded

03 / FIELD MAPPING

Connect the right fields

Smart Match ranks field names, aliases and compatible types. Review every suggestion; nothing is silently approved.

AI suggestions are optional and require local model setup. Smart Match needs no download.

Target field · output type	Source field(s)	Transformation	Parameter / default	Review
Load source and target data, or try the example to start.

No fields

Rules: Copy, Trim, Upper/Lowercase, Number, Boolean, ISO date, Join, Split, Constant, Multiply, Replace, Prefix, Suffix, Round. Split parameter: delimiter|index. Join parameter: separator. Multiply: numeric factor. Replace: find|replacement. Round: decimal places 0–10.

04 / VALIDATION

Test your mapping

Preview up to 100 rows with clear conversion errors. Strict mode omits invalid rows from the exported dataset.

Strict: skip invalid rows

No preview yet.

Your transformed data preview will appear here.

0 records

05 / DELIVERY

Export and reuse

All downloads are generated inside your browser. No data is sent to PixelOkay.

Output format

Spreadsheet-safe CSV adds a leading apostrophe to text beginning with formula-like characters (such as + in phone numbers). Select JSON or JSONL if you need exact text-byte round trips.

Mapping files store field paths and rules, not source records. Reopen a mapping after loading matching source and target datasets.

How smart matching works

Smart Match uses deterministic field-name similarity and type compatibility without any model. Optional AI proposals run in your browser only after you explicitly load a local model. All AI suggestions require your approval and validation.

FOR DEVELOPERS, DATA TEAMS & CONTENT CREATORS

Turn mismatched columns into a repeatable workflow.

Compare schemas

See inferred fields and types from structured examples, including nested JSON properties.

Review suggestions

Build one-to-one field mappings, customize casts, and resolve missing or ambiguous matches.

Export with confidence

Inspect row-level conversion errors and save reusable rules for later datasets.

Frequently asked questions

Is this generative AI?

Smart Match uses deterministic local rules. Optional AI Suggest uses field paths, types and required flags, not record values, to propose unapproved mappings with a model that runs in your browser. WebLLM runtime assets load from a pinned CDN and model files download after you select a model; browser-built-in AI is experimental. Review each suggestion.

Does it support Excel spreadsheets?

Export your spreadsheet as CSV or TSV first. Native XLSX and relational joins are not currently supported.

Are data files uploaded?

Source records stay in this browser tab, and the optional AI prompt contains only field paths, types and required flags, not row values. The model runs in the browser; choosing WebLLM can load runtime files from a CDN and download model weights. No source row values are sent to an AI model.

What happens if data cannot be converted?

Conversion problems are reported by row and target field. You can choose lenient output with null values or strict output that excludes invalid rows.

Map CSV and JSON records to a target schema

Match source fields to a target structure, review transformations, and validate rows before exporting. Deterministic Smart Match is available without a model; optional AI suggestions use schema field metadata and remain unapproved until you review them.

Import a source CSV, TSV, JSON or JSONL dataset, then add a target sample or header that shows the required fields.

Run Smart Match or choose optional AI Suggest. Review every source-to-target path and select only supported transformations you need.

Preview the first 100 rows, run full validation when ready, then export JSON, JSONL or CSV plus any issue report or reusable mapping settings.

Example: Map an order file’s customer identifier, email address and creation date to the matching target paths. Confirm the date conversion, preview sample rows, validate the complete file and download JSONL.

Limits and privacy: Imports are limited to 8 MiB, 10,000 records and 150 fields; preview shows at most 100 records. Optional AI uses field paths, types and required flags, not record values. WebLLM runtime files load from a pinned CDN and model weights download only after you select a model; browser-built-in AI is experimental. Review all suggestions and exports.