Skip to content

DevS69 SDETKit

From noisy CI evidence to a reviewable decision, a diagnosis, and the exact next command.

DevS69 SDETKit is a release-confidence CLI for deterministic ship/no-ship decisions with machine-readable evidence. It runs your repository's own checks in its own environment, reduces a failure to the first one that matters, and tells you what to run next. It never applies, merges or dismisses anything for you.

Primary outcome: know if a change is ready to ship.

Canonical first path: python -m sdetkit gate fast -> python -m sdetkit gate release -> python -m sdetkit doctor.

Which version these docs describe

These docs follow main, the 2.0.0 release candidate. The latest published package is sdetkit==1.2.0. The verdicts, the fix loop (diagnose run --prove) and checks that run in the project's own environment are on main only until 2.0.0 is released; the current product delta lists every difference.

Get started

python -m venv .venv && source .venv/bin/activate
python -m pip install sdetkit==1.2.0

python -m venv .venv && source .venv/bin/activate
python -m pip install "git+https://github.com/sherif69-sa/DevS69-sdetkit"

Then, in your repository:

python -m sdetkit gate fast --format json --stable-json --out build/gate-fast.json
python -m sdetkit gate release --format json --out build/release-preflight.json
python -m sdetkit doctor --format json --out build/doctor.json

Each command writes one JSON artifact that people, bots and CI read the same way. Start in 5 minutes walks through the first run.

One verdict per run

Every run ends in one of four decisions, recorded in the artifact's verdict.decision:

Verdict Meaning What to do
SHIP Every check ran against this repository and passed. Continue to gate release, or ship.
REVIEW Nothing blocks, but a person should look first (findings that do not block, or a proven fix waiting for approval). Read the findings, then decide.
NOT_RUN The checks could not run against this repository, for example because the project is not installed in the environment that ran them. Nothing was proven either way. Run the setup command in next_action.next_command, then run the gate again.
NO_SHIP A check failed. Start from next_action: the first failure, its file, and the command to reproduce it.

A failing run names its first failure instead of the loudest wrapper error:

decision: NO-SHIP
classification: test
first_failure: FAILED tests/test_release_contract.py::test_wheel_smoke
affected_file: tests/test_release_contract.py
next_command: python -m pytest -q tests/test_release_contract.py -o addopts=
authority: review-first

From a failing gate to a reviewed fix

After a failing gate fast, one command diagnoses the failure, proves the narrowest mechanical fix in a throwaway copy of the repository, and writes a decision card. A person approves or rejects it; nothing is applied for you.

python -m sdetkit diagnose run --prove
python -m sdetkit diagnose decide --card build/sdetkit/remediation-loop/decision-card.json --approve --reviewer <you>

See the remediation loop and first-failure triage.

Proven on real projects

gate fast has been run on seven well-known open-source projects at pinned commits: pallets/itsdangerous, markupsafe and click, python-attrs/attrs, pytest-dev/pluggy, python-poetry/poetry-core and psf/requests. Every fresh clone reads NOT_RUN with the setup command it needs; after setup, five read SHIP and two read NO_SHIP for real failing tests. The external repository proof records every result and how to reproduce it.

Where to go next

This page is the product homepage/router: everything a first-time user needs is linked above. Historical and transition-era references remain intentionally secondary to first-time adoption; use the archive index only after the canonical first path is operating.

SDETKit needs Python 3.10 or newer. Use an isolated environment (a virtualenv or pipx).