Tools

Reading

Turning bytes into something with a shape, and saying so when it cannot.

In the app

Paste, drop or open a file and the Head answers with a verdict before anything else is drawn — “Valid JSON”, “Not valid JSON yet · N fixable”, “Looks like a log · N lines”.

A log or a CSV asks its reading question beside your text: the pattern or the separator as chips, the match count, a live preview. Read as records folds the text and opens the list; Keep as text is a real answer with a door back.

The reading question — The shape it matched, the count, the chips — then Read as records. · 7 s

Across the two shapes

A verb is implemented once per shape, not once per format. That is why a CSV gets the same tools as a webhook payload: both collapse to one of two shapes before any verb runs.

How the shapes differ for parse: adapter-provided.

document · JSON, YAML, XML
records · NDJSON, CSV, TSV, Markdown table, Log

A worked example

The site’s own samples, run through detection when this page was built: the format the registry chose and the confidence it gave. 0.95 is proof — a parser read the whole text; 0.60 is a shape recognised without a parse, which is what a broken JSON document scores and why it gets a verdict and a fix list rather than a tree.

clean-order-example.json · json · 0.95
checkout-api.log · log · 0.95
kibana-export.csv · csv · 0.90
broken-webhook-example.json · json · 0.60
checkout-api.ndjson · ndjson · 0.95
kibana-export.md · markdown · 0.95
orders-export.csv · csv · 0.67
stripe-event.pprint.txt · json · 0.60