Dataset utilities
Inspect JSONL before you train or import it.
Parse every non-empty line independently, detect duplicates and common conversation shapes, and optionally scan records for supported PII and credential patterns.
JSONL dataset
Paste or select a local fileOpen a .jsonl or .txt file
The browser reads the selected file directly; this site has no upload endpoint.
Dataset report
Client-side analysis{L}
Ready to validateEach non-empty line must be independent JSON. The tool does not send the dataset anywhere.
Syntax plus dataset-level signals
The validator reports invalid lines, duplicates, mixed shapes, empty message/prompt/completion content, very large records and optional supported PII or credential-like matches.
Shape detection is descriptive
The tool recognizes common messages, prompt/completion and input/output record shapes. It does not assume that one vendor schema is universally correct.