Tests: what QuorX checks before every change
The test suites and how they grew, the platform packs' self-tests, the UI test that drives the desktop app, and the last results kept on this computer.
Every change to QuorX goes through three kinds of tests. None of them needs an OpenRouter key or spends money: they drive QuorX's real code with no model, or with a scripted one.
| Command | What it tests | Takes |
|---|---|---|
npm test | The agent loop, the safety rules, Stop, the terminal, approvals, the desktop tool, News | About a minute |
npm run test:packs | Each platform pack's self-test that this computer can run: real toolchains, real builds, real debuggers | Minutes, more with more toolchains |
npm run test:ui | The QuorX desktop app itself, driven like a person would, with a screenshot of every step | A few minutes |
What runs
The biggest files are the agent (test-agent: the loop, tools, checks and reviewer) and safety (test-safety: the rules for commands, the folders that are off limits, Stop, background jobs, the debugger). test-posix tests the same safety layer the way it works on macOS and Linux; on Windows, the parts that need those systems skip.
Over time
Counted from the test files at each change merged into QuorX's main branch: the test(…) calls in the files npm test runs, and in the platform packs' self-tests.
Pack self-tests
Every platform pack has self-tests, and npm run test:packs runs them. A pack's logic (parsers, rules, notes) is tested on any computer; the tests that build and run real apps need that pack's toolchain. The Platforms report rates each pack by the best run on record.
| Pack | Level | Result | Self-tests passed |
|---|---|---|---|
| QuorX on Windows: Stop, rules, desktop keys | Passed | passed | 16 |
| Android | Passed | passed | 11 |
| Apple apps | Logic tested | logic tested | 8 |
| Every project | Logic tested | logic tested | 0 |
| C and C++ | Logic tested | logic tested | 0 |
| .NET (C#) | Logic tested | logic tested | 0 |
| Electron | Passed | passed | 8 |
| Flutter and Dart | Passed | passed | 5 |
| Go | Logic tested | logic tested | 0 |
| Godot | Passed | passed | 4 |
| Windows installers | Passed | passed | 3 |
| Java and Kotlin | Passed | passed | 5 |
| Node.js and TypeScript | Passed | passed | 1 |
| Python | Passed | passed | 6 |
| Rust | Passed | passed | 3 |
| Tauri | Passed | passed | 7 |
| Unity | Passed | passed | 4 |
No rows at this level.
UI test
The UI test (npm run test:ui) starts the QuorX desktop app with its own data folder and port, drives its window the way a person would, and saves a screenshot of every step. Its last published run passed 9 of 9 steps. Scenarios: settings, news, cli, running.
On this computer
The last npm test, npm run test:packs and npm run test:ui run in this computer's QuorX folder, if any. Each run keeps only its counts and the names of failed tests, in logs/test-results.json on this computer.
Your own numbers show here when you open this page in QuorX's News.
Help test
Tests pass on the machines they ran on. The more systems they run on, the more they're worth: a pack's self-test that skips here may be the one that fails on another. Running the kit and pasting its report into a "Pack test report" issue adds that system to the record.
> npm test > npm run test:packs Report: pack-test-report.md (paste it into a "Pack test report" issue) >