Parse The Data
Local only

Field guide

Prepare JSONL before an AI pipeline finds the mistakes.

Data workspace

Ready in this tab

Source

0 B Processed locally

Paste data or choose a file to begin.

Analysis canvas

Ready when you are.

  1. Add a local file or paste data.
  2. Analyze format, quality, and likely sensitive fields.
  3. Review, clean, and export the in-tab working copy.
01 / Parse

Know the shape of the data.

Detect delimiters, validate structured text, and surface useful error locations.

02 / Review

Spot quality and privacy risks.

Measure completeness, duplicates, mixed values, and likely identifiers before sharing.

03 / Prepare

Clean a working copy locally.

Trim, deduplicate, mask, infer a schema, and export without changing the original file.

Focused utilities

Start with the job in front of you.

Every route opens the same private engine with guidance tailored to that format.

CSV cleanerProfile + clean tables JSON validatorValidate + inspect records JSONL converterPrepare streaming data Schema generatorInfer a data contract Delimiter detectorFind hidden separators PII scannerCatch obvious exposure

Practical guide

A reliable JSONL preparation loop

Each line should be a complete JSON value, normally an object. Keep records compact, use UTF-8, avoid a surrounding array, and validate the exact file you intend to upload rather than an earlier spreadsheet export.

  • 01Parse every non-empty line independently and stop on the first error.
  • 02Compare observed keys and value types across all records.
  • 03Remove exact duplicates and investigate near-duplicates separately.
  • 04Scan prompts, completions, metadata, and free text for sensitive data.
Does Parse The Data upload my files?

No. Parsing, profiling, cleanup, and export run inside your browser tab. The application does not send dataset contents to our server or any third party.

What file types are supported?

The first release supports CSV, TSV, JSON, and JSON Lines (JSONL or NDJSON), including pasted text and local files.

Can it detect every privacy risk?

No. The privacy scan uses deterministic patterns and column-name hints. It can find common emails, phone numbers, IP addresses, SSNs, and payment-card patterns, but it is not a substitute for a formal privacy review.

Designed for restraint

The server delivers the tool. It never receives the dataset.

Parsing runs in a separate browser thread. There are no accounts, upload endpoints, third-party scripts, advertising trackers, or cloud storage in the analysis path.

Use the PII scan as an early warning—not as legal advice or a compliance certification.