CSV Doctor — Offline Encoding, Delimiter & Header Check

One CSV in, a plain report on its bytes and shape, nothing leaves this machine.

CSV Doctor is an offline CSV checker: open one local HTML file, choose a CSV, and read what is actually in it before another tool chokes on it. It reports byte order marks, UTF-8 validity with the offset of the first bad byte, line endings, the delimiter, header problems and ragged records. No account, no subscription, no telemetry, no upload.

WHO IT IS FOR: bookkeepers, shop operators and anyone whose CSV export "looks fine" but will not import: Stripe, Shopify, bank and warehouse exports, spreadsheet saves, files handed over by a vendor.

HOW IT WORKS: 1. Unzip and open CSVDoctor.html in a desktop browser. 2. Choose a CSV (up to 10 MB) or paste text; leave the delimiter on automatic or pick comma, semicolon, tab or pipe. 3. Press Check this CSV and read the findings table (OK / WARN / FAIL), the numbered header list and the field-count histogram with ragged record numbers. 4. Optionally prepare a .txt or .json report for download. Nothing is written until you choose to download.

IN-APP EXAMPLES (synthetic rows only, id=001 / sku=WIDGET): Clean comma CSV (all OK); UTF-8 BOM + semicolons + CRLF (BOM warning, semicolon detected); Ragged rows + duplicate header (blank header in column 5, "qty " vs "qty" reported, records 2 and 3 ragged); Invalid UTF-8 byte + tabs (bad byte at offset 31, tab detected); Unbalanced quote (never-closed quoted field flagged).

Also included: a shape check against the published input limits of Table Tidy and CSV Compare from this store (comma only, size, rows, columns, unique nonblank headers). It is a shape check, not a promise about those tools.

You will get a ZIP: CSVDoctor.html, README.md, PRIVACY.md, LICENSE.txt, VERSION.txt. Includes a single-user license for personal and commercial work.

Reports describe the bytes you supplied. CSV Doctor does not repair, reformat or validate the meaning of your data, and a clean report is not a guarantee that another program will accept the file. Review results before relying on them.

Limits: one file at a time up to 10 MB (larger files are refused, not sampled); delimiter detection reads the first 200 records, all other counts read the whole file up to 200,000 records and 200 columns; exact case-sensitive comparison, with a case/space-insensitive duplicate-header check reported but never applied; UTF-8 and UTF-16 (with BOM) only. Tested in Microsoft Edge with the five built-in examples (clean comma; BOM+semicolon; ragged headers; invalid UTF-8+tab; unclosed quote). Firefox not yet verified.

Version 0.1.0.