This release is a pre-release and may not be stable for production use.
The Python test runner built for precision.
▸ Zero plugins. Full arsenal.
Test Junkie 
Test Junkie is a Python test runner and test framework with built-in parallel test execution, exception-aware
retries, multi-layer parametrization, event listeners and HTML/XML/JSON reports — everything in one package, no plugins.
Use it from Python with Runner or from the terminal with the tj command.
Key Features
Not a wrapper around
unittest. Every capability below ships with the library — no plugins to install, no config files to write, no external orchestrators.
- Decorator-based:
@Suiteand@teston plain Python classes — no base class to extend, no config file to maintain - Lifecycle hooks:
@beforeClass,@afterClass,@beforeTest,@afterTeston the class, plus@beforeGroup/@afterGroup(via@GroupRules) that run once around a named set of suites - Built-in parallel execution: separate thread limits for suites (
-S/suite_multithreading_limit) and tests (-T/test_multithreading_limit);parallelized=Falsekeeps a suite or test out of the parallel pool, andpr=[...]stops specific suites or tests from ever overlapping - Retries:
retry=Non a suite or a test — only the failing parameter variants re-run;retry_on=[ConnectionError]retries just infrastructure noise (subclasses included),no_retry_on=[AssertionError]makes real bugs surface immediately;tj run --retry N/--no-retryoverride retries for one run - Multi-layer parametrization:
@Suite(parameters=[...])×@test(parameters=[...])= every combination, each variant tracked and retried on its own; parameters can be a function that builds the list at run time Rulesfor shared setup/teardown — definebefore_class/before_test/ … once, attach with@Suite(rules=...)to as many suites as you like- First-class
owner,feature,component,tagsandpriority— filter withtj run --owners,--features,--components,--tags-any/--tags-all(tags), orowners=/features=/components=/tag_config=inRunner.run() - Execution order per suite:
@Suite(order=TestOrder.RANDOM)(orALPHABETICAL,PRIORITY_ASC,PRIORITY_DESC) — e.g. shuffle one suite to surface order dependencies; the run prints the seed, andtj run --seed Nrepeats that exact order - Skipping:
skip=Trueor a function deciding at run time on any suite or test, andshortcuts.skip("reason")from inside a running test - Live event listeners: subclass
Listener, override any of its 21 events (on_failure,on_success,on_class_skip, …) and attach it with@Suite(listener=...) - Results as objects:
runner.get_executed_suites()returns typed suite/test objects with status, timing, retries, exceptions and metadata;Meta.update(...)adds to a test's metadata while it runs - Reports: HTML (with CPU and memory graphs when
monitor_resources=True/tj run -m), JUnit XML and JSON, written to the files or folders you pass ashtml_report=/xml_report=/json_report=;runner.run()also returns the results (.get_basic_report()has the counts per status and suite) - CLI:
tj run,tj audit,tj config,tj version— run and inspect suites without writing a runner script; settings can live in a projecttj.cfgorpyproject.toml;tj runexits non-zero on failures, errors or when nothing ran and marks failures on GitHub pull requests, ready for CI
And more →
Installation
pip install test-junkie
Quick example
from test_junkie.decorators import Suite, test, beforeClass
from test_junkie.runner import Runner
@Suite(feature="Login", owner="qa-team")
class LoginSuite:
@beforeClass()
def open_session(self):
self.session = {"user": "demo"}
@test(component="auth", tags=["smoke"], retry=2)
def valid_login(self):
assert self.session["user"] == "demo"
@test(component="auth", parameters=["", "wrong-password"])
def invalid_login(self, parameter):
assert parameter != "correct-password"
if __name__ == "__main__":
Runner([LoginSuite], html_report="report.html").run(test_multithreading_limit=4)
Or from the terminal: tj run -s path/to/tests -T 4 --html-report report.html
Python 2.7: the last release with Python 2.7 support is
0.8a.8. All versions from0.9a0onwards require Python 3.
Performance
Key takeaways (median of 30 runs; every run verified to have passed all of its tests):
- 1.4× faster than pytest, sequentially (1.97 s vs 2.74 s, 1,000 tests). Fastest of the four runners: 1.7× faster than Robot Framework (3.39 s) and just ahead of plain unittest (2.03 s).
- 6.7× faster than pytest-xdist in parallel (0.31 s vs 2.06 s, 1,000 tests, 10 workers). Tests run on threads inside one process, so there's no worker start-up cost, and no plugin to install.
- Parallelism that scales: going from 1 to 10 threads makes a 1,000-test run 6.2× faster. pytest-xdist peaks at ×1.7; pabot is slower than sequential Robot Framework at every setting.
- 6.2× smaller than pytest with the plugins for the same features (3,345 KB vs 20,680 KB installed, every dependency included). Parallel execution, parametrization, retries, listeners, ordering and HTML/XML/JSON reports ship in one package; pytest needs 9 plugins for them, which pull in 18 more packages. Robot Framework with plugins is 8,005 KB (2.4×) and covers 13 of the 19 features fully. Details: install size.
- ~1 ms framework overhead per test at 1,000 tests (+972 ms total), vs +1,028 ms for unittest, +1,741 ms for pytest and +2,395 ms for Robot Framework.
- 95 ms start-up for a 1-test run, on par with unittest (92 ms) and about a third of pytest (296 ms) or Robot Framework (308 ms).
Timings measured with test_junkie 0.9a5, pytest 9.1.1, Robot Framework 7.5 on Python 3.12.5 (Windows 11, Intel Core i9-10980HK); install sizes with test_junkie 0.9a6 and the latest plugins on 2026-10-07. Full results tables, parametrized and parallel runs, per-phase timings and the exact commands are in the full benchmark →
Getting Started
1 maintainer
Metadata
1 maintainer
Release files for test-junkie 0.9a6
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | |
|---|---|---|---|---|
| test_junkie-0.9a6.tar.gz | 100.9 kB | With provenance | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| test_junkie-0.9a6-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 208.7 kB
Release files / test_junkie-0.9a6.tar.gz
| Download URL | test_junkie-0.9a6.tar.gz |
|---|---|
| Size | 100.9 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
815840bacaa4efd24e381872e7bc56e4c171a38b3afc5ff1f4e1390bdc7c339e
|
|
BLAKE2b-256 checksum How to use checksums |
ff41569d43647cb34ff3aae7041817bb5106425ce483e14637d97dee86f2b03d
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 7, 2026.
Transparency logRelease files / test_junkie-0.9a6-py3-none-any.whl
| Download URL | test_junkie-0.9a6-py3-none-any.whl |
|---|---|
| Size | 107.8 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
4f4c4016ddbb219d10737e82eab78685cc3056e33964473dec44d03f61e6443b
|
|
BLAKE2b-256 checksum How to use checksums |
07365064eb41d62014f0dcf8171038bc1a22e8bf4503dffa70c4ca5b1b60e7cb
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 7, 2026.
Transparency logSecurity for test-junkie 0.9a6
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
Full provenance
All 2 files in this release have been uploaded with provenance metadata.
Attestation for test_junkie-0.9a6.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 7, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.
