About
Last updated
What this site is
yizhiweb.top is an independent collection of 102 browser-based utilities for working with text, numbers, encodings, data formats, time, images and files. Every tool runs entirely in your browser. There is no account, no upload, and no server-side processing — the code that transforms your input is the code that runs on your own machine, and nothing you type or drop into a tool is transmitted anywhere.
The site covers the awkward, unglamorous part of everyday technical work: moving digits between 22 numeral systems, stripping or diagnosing diacritics, finding the invisible character that breaks a database lookup, encoding Base64 or URL values, hashing a file, converting timestamps, resizing an image, or turning a CSV into JSON. These tasks come up constantly, are usually served by single-purpose sites that do not explain themselves, and are frequently done wrong as a result.
Why it exists
A tax form uses Thai digits. A pasted spreadsheet column refuses to match because it carries a non-breaking space. A search fails because one string is precomposed and the other is decomposed. A JWT is “decoded” in a way that proves nothing about whether it is genuine. In each case the tool exists, but the rule behind it is hidden — so the user cannot check the answer, and cannot tell a correct result from a plausible one.
This site publishes the rule as well as the result. Each tool page states which standard or algorithm it applies, links to the relevant specification where one exists, and documents the cases it deliberately refuses rather than guessing. The methodology page lists the basis and tests for the tools whose rules are easiest to get wrong, and the guides carry the longer explanation of when a result is trustworthy and when it is not.
Who runs it
It is a small independent project, not a company and not a media brand. The site is built and maintained by a small editorial and engineering team; the authors page describes the split between writing and engineering and how accuracy is verified. There is no anonymous content farm behind it, and no outsourced content production — the prose on a tool page was written by the people who implemented the engine that runs beneath it.
We are deliberately specific about what we are not: we are not lawyers, doctors or financial advisers, and a calculator is not a substitute for professional advice. The terms page states the limits plainly.
How accuracy is checked
Every engine is tested at four levels, and the tests run before every deployment:
- Unit tests — known inputs with expected outputs, including the RFC test vectors for standards-based tools (TOTP against RFC 6238, punycode against a reference implementation, Base64 against RFC 4648).
- Property and round-trip tests — hundreds of thousands of assertions that a conversion and its inverse reproduce the input, and that invariants hold over random data.
- Oracle tests — independent implementations of the same calculation, written differently, cross-checked against the engine.
- Browser tests — every control on every tool page is driven in a real browser, so a change that breaks a button is caught before it ships.
When a tool cannot be made correct, it is not published. A converter that silently produces a plausible wrong answer is worse than one that refuses, which is why several tools report “out of range” or “not valid” rather than returning something approximately right.
What we cover
The catalogue is organised into thirteen families: scripts and numerals, diacritics, phonetics, text utilities, encoding, developer and data formats, converters and calculators, colour, CSS, date and time, images, documents, and QR and barcodes. A full list is on the tools page, and the families are grouped under seven top-level domains for browsing.
Coverage grows where a real need is visible and the result can be verified. It does not grow to fill a keyword list: a tool that would duplicate something already well served, or that would have to guess at an answer, is left out on purpose.
Privacy, concretely
“Nothing is uploaded” is not a slogan here — it is a property of the architecture. The site is static: there is no application server, no database, and no endpoint that receives tool input. The tools are JavaScript that runs in the page, and the file-based tools read the file through the browser's local File API and process it on a canvas or in a worker on your machine. You can verify this by opening the network panel and using any tool: no request carries your data.
The only third-party code is the advertising and analytics described in the privacy policy. That is why tools which handle secrets — the encryption tool, the password generator, the JWT decoder — are explicitly designed so that the sensitive value is never placed in a field that could be read by anything other than the local script.
How the site is funded
Hosting and development are paid for by advertising. The site carries display advertising through
Google AdSense, which is why there is an ads.txt file and a consent mechanism for
readers in regions that require one. There is no paywall, no account tier, and no charge for any
tool.
Advertising is kept out of the part of the page you are trying to use: no ad is placed inside a tool's input area, and the layout reserves space so that an ad loading late does not move the content you are reading. If an advertisement ever appears to interfere with a tool, that is a bug and we would like to hear about it.
Corrections
If a result is wrong, we want to know. A correction is more useful with the input that produced the problem and the answer you expected — with those two things the engine can usually reproduce the case and add it to the test suite. The contact page has the address. Corrections that change a tool's behaviour are treated as bugs, not suggestions; corrections to prose are made under the editorial policy.
Contact
Corrections, bug reports and questions are welcome at henrivnts@gmail.com. The contact page describes what to include so that a report can be acted on.
Questions about this page? Contact us.