ZWS: a URL shortener whose test script runs a linter, a formatter and a build
Project brief: Shorten URLs using invisible spaces. Individuals Organizations [Support this project with your organization][open-collective].
At a glance
- What is it?
- A short readme for a short project: URLs carried in invisible characters, two statistics badges, an OpenAPI schema served from a second hostname, a funding page with ten unfilled organisation slots, and a package manifest whose test command contains no tests.
- Who is it for?
- ZWS is a small, deliberately undocumented service, and the documentation is honest about its size rather than pretending otherwise. What it gives you is a public instance to try, an OpenAPI schema for building a client, a separate command line repository, and two badge endpoints you can embed anywhere.
- Can I use it commercially?
- Yes. Apache-2.0 is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 2 days ago.
- What is it written in?
- Mainly TypeScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 20, 2026, and from our analysis. They are not legal advice.
Star history
- zws-im/zws1,843 ★ · +2
| Repository | 2026-08-04 | 2026-11-20 |
|---|---|---|
| zws-im/zws | 1.8k | 1.8k |
2 Stars in the last 108 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
Two hostnames, and a badge that fetches the API for you
The public instance is at one address and the application programming interface sits at another, and the readme uses both without explaining the split. The documentation link is on the site host, while the badge images embedded at the top of the file point at an API host, and the badge service in turn is asked to fetch those statistics routes.
So the project relies on a third-party badge service to make live requests into its own API on behalf of whoever renders the readme. That is a small dependency with a specific consequence: the two counters at the top of the file are only as fresh as the badge service's cache, and the badge image URLs are hardcoded in the readme rather than generated.
The statistics themselves are two routes, one for the number of shortened URLs and one for the number of visited URLs, and they are described as implementing a published badge endpoint schema. That is the only part of the interface described concretely in the readme, which suggests the counters were designed to be embedded rather than queried.
The command line client is not in this repository. It has its own project with its own readme, and the only link to it is in the sentence inviting you to try the service online.
The readme states the premise and leaves the mechanism out
The entire technical claim fits in one line: URLs are shortened with invisible characters. The heading calls it a zero width shortener. There is a demo link, a link to the API documentation, and two statistic badges.
What is not in the file is the encoding. There is no description of which characters are used, how a URL is mapped onto them, how the decoder recovers the original, or what happens when a target system strips zero-width characters from text. For a project whose entire novelty is that the shortened form is not a short identifier at all, those are the questions a reader arrives with, and none of them are answered here.
There is an OpenAPI schema served from the documentation path, which is where the interface itself is described. That is a deliberate division: the mechanics stay undocumented, the request and response shapes do not. It also means a client can be written against the schema without knowing how the payload is carried, which is a reasonable boundary if the encoding is meant to stay an implementation detail.
One integration is credited by name: a screenshot and file upload tool added support for the service, which is the only third-party adoption the project mentions.
The funding page has ten organisation slots and no organisations in them
The financial contributors section is an open collective page with two subsections, individuals and organisations. The organisations part contains ten entries, and every entry is a bare link with an empty slot beside it.
That is what an uncustomised template looks like after commit. Each line is a link to an organisation profile's website field, and none of them has a logo or a name, so the readme currently advertises ten sponsors who are not there. It is a small blemish, and it is the kind that survives because nobody opens the funding section when they are reading the technical content.
The individuals section is a single link, also without a rendered image in the text.
The rest of the contributor section is short and specific. The project is maintained by one person, and it says contributions to the codebase are welcome. One integration is thanked by name. That is the whole account of who works on it, and for a service with no version history to speak of, that is worth knowing before you report a problem.
The only releases belong to a schemas package published in 2021
Three releases exist, and all three are for a separate schemas package rather than for the application. They were published on the same day in November 2021, within about twenty minutes of each other, at three consecutive patch versions.
That was nearly five years before the last commit, which is dated today. So the release history is not a history of this service at all: it is a snapshot of a schema package that was iterated on quickly one afternoon and then left alone, while the application itself continued to be developed without tagged releases.
Two practical consequences. If you are choosing a version to deploy, there is nothing to choose, because the interesting code has no tags. And if you are judging how actively the project is worked on, the release page will tell you the opposite of the truth, while the commit history tells you the truth: the last push is dated the current day.
The project is not archived, and its default branch is the conventional one, which is consistent with a small service that is still being touched.
The test script runs a linter, a format check and a build
The root manifest is a private workspace root named after the project, with the applications in a glob. Its script table has one entry that matters more than the others, and it is the test command: a chain of three checks that runs the linter, then a formatting check, then a build across the workspace. There is no test runner anywhere in it, and no test script in the manifest, and no test file in the tree.
So the honest statement is that this repository has no automated tests, and that the command named test means something else. That is not unusual in a project of this size, and having a single command that must pass before a commit is better than nothing. But anyone contributing needs to know that a green test run says the code lints, is formatted and compiles, and nothing more.
Two smaller details from the same table. Migrations have their own three commands, generate, check and apply, which is a schema-driven database setup rather than hand-written SQL. And the required runtime version is declared in a way that only warns on mismatch, so an unsupported Node version will not stop an install.
Two formatters, a migration tool in the wrong dependency list, and a hosting config
The tooling stack is generous for a project this small, and it overlaps.
There are two style tools. One handles linting with an unsafe-fix flag available for automatic corrections; the other handles formatting, and it is configured with a plugin that sorts fields in the package manifest and another that orders class names. Running both means formatting and linting have separate opinions about the same files.
Every development dependency is pinned to an exact version, which is the right call for a reproducible build and a sign the author expects contributors to install exactly what was tested.
The dependency list is where the picture gets odd. The migration tool, which is only needed while developing, sits in the runtime dependencies rather than the development ones, alongside a small environment validation library. A production install of the workspace therefore pulls a database toolkit it never runs.
The deployment side is conventional and inspectable: a process file for a platform that reads it, a hosting configuration naming the platform, a task runner configuration, and a dependency update bot configuration. A runtime version file pins the toolchain for contributors.
Editorial conclusion
ZWS is a small, deliberately undocumented service, and the documentation is honest about its size rather than pretending otherwise. What it gives you is a public instance to try, an OpenAPI schema for building a client, a separate command line repository, and two badge endpoints you can embed anywhere. What it does not give you is any account of how the invisible-character encoding works, which is the one thing a reader would come for.
So judge it on what you can check. The deployment surface is conventional and inspectable: a process file, a hosting configuration, a database with migrations, and two hostnames. The tooling is where the project's habits show, and one of them is worth knowing before you contribute: the test command runs the linter, the format check and a build, so there are no automated tests to satisfy. The runtime version requirement is advisory rather than enforced. And the only published releases belong to a schemas package from 2021, so the release history tells you nothing about the current code.
If you intend to run your own instance, the OpenAPI document and the migration tooling are the two places to start. If you intend to understand the encoding, expect to read the source.
Frequently asked questions
How does ZWS shorten a URL?
By carrying it in invisible characters rather than in a short identifier. The readme states the premise and links to an online demo and an OpenAPI document, but does not describe which characters are used or how a decoder recovers the original URL.
Does ZWS have automated tests?
No. The manifest has no test runner and the test script chains the linter, a formatting check and a workspace build. A passing run means the code lints, is formatted and compiles.
Where are the ZWS API docs and statistics?
The OpenAPI document is served from the documentation path on the main host, while the two statistics routes that feed badge images are served from an API host, with the badge service fetching them on request.
What does the ZWS release history show?
Three releases, all for a separate schemas package, all published on the same day in November 2021 at consecutive patch versions. The application itself has no tagged releases, and its last commit is much more recent.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/zws-im-zws)Star history chart for your README
[](https://hysenlabs.com/projects/zws-im-zws)Was this useful?