CEDX Test · Developer & low-code
Test runs the estate's test operations: 360 cases in 64 suites across four layers, every flake scored and quarantined or healed, and coverage gaps tied to the incidents they predict.
The top flake on the board fails 38% of its 12 attempts, holds inventory during a checkout race, and is not quarantined. The tool's first scored action is to quarantine it. That is what a test console is for.
Runs on demo data — Northline Test Ops is the demo workspace, not a customer.
What it is
64 suites split 26 unit, 16 integration, 15 E2E and 7 contract — each with cases, pass %, p95, flakes and an owner. pricing/edge runs 44 cases at 100% with a 4.1-second p95; billing/invoice-calc carries 54 at 98% with one flake. The layer chips cut the table to the stack you own.
360 cases with status, duration, retries and a flake score. POS offline queue flushes after reconnect tops the board at 94 — quarantined, 1m 28s, 3 retries. rolls back if POS queue wears a healed chip. 156 cases are assistant-authored, marked AI on the row.
320 runs with branch, status, pass %, failures and duration. run_4821 on main failed 12 cases at 91% pass; run_4818 on feat/checkout-hold passed 97 with zero failures in 15m 20s. The cancelled run shows 0 and 0 — recorded, not erased.
Product tour
Not a mockup and not a concept deck. This is what opens at /app/test.
01 — Overview
Pass rate 92%, 50 open flakes, 21 modules at high gap risk, 66% coverage — and alerts that connect the dots: payments-core sits at gap risk 88 with 4 recent incidents, and the main branch run failed 12 cases, 8 of them flaky. The do-this-next queue's first action is to quarantine the checkout race.
02 — Suites
1–25 of 64 suites with the mean pass at 92%. Contract suites punch above their size — api/routes-contract at 97% on 48 cases — while the flaky-case chip isolates the 46 that keep engineers up.
03 — Cases
Cases carry their suite, status, duration, retry count and flake score — the quarantined chip is a state, not a deletion. 3DS challenge resumes payment shows Failed at 71; eight cases wear the healed mark from the last 30 days.
04 — Runs
Runs list by branch with pass %, fail count and duration — release/2026.8 failed 22 cases at 88%, and the row says so. Median duration across the full pipeline is 11 minutes 32 seconds, printed on the card.
Who runs it
Roles, not references. We have no named customers yet, so nobody in these photographs is quoted, credited or claimed as one.
Owns the flake board: 50 open flakes, one quarantine decision at a time, and a pass rate that means something because the flakes are out of it.
92% pass · 50 open flakes
Write and review the cases — including the 156 assistant-authored ones, which arrive marked AI and get read like any other contribution.
360 cases · 156 AI-authored
Watch runs by branch before a release cut: release/2026.8 at 88% with 22 failures is a conversation, not a merge.
320 runs · median 11m 32sThe shape of it
Every figure below is legible in the captures above. Nothing here is a projection of your estate — it is the state of the demo data.
How it runs
Cases arrive from people and from the assistant — 156 of 360 carry the AI chip — into suites across unit, integration, E2E and contract layers.
320 runs in the window, median 11m 32s for the full pipeline, each tied to its branch: main, develop, release, feat or fix.
Failures and flakes split apart: 48 failed runs need triage, and every flake carries a score — the checkout race reads 91 at a 38% failure rate.
Selectors self-heal where they can — 8 in 30 days — and quarantine keeps the rest from blocking the pipeline while the fix lands.
One record
Test is not a separate CI silo. Its suites are the required checks the deploy gate consumes, running against the environments the estate clones.
Deploy's required checks are these suites — 13 deploys currently sit blocked behind failing ones, and both products say so.
CEDX SandboxE2E suites run against cloned environments — a failed seed apply explains a red suite before the code does.
CEDX FunctionsContract suites pin the routes and functions the runtime serves — api/routes-contract runs 48 cases at 97%.
CEDX BuilderApps assembled in Builder inherit the same gates: publish risk and test health are two views of one question.
Limits
Finding this out on the third call is worse for you than reading it here, and worse for us.
Start
Open it right now
Your first suites
Test with the rest of it
Questions
Yes. Every screenshot is a capture of the running build and you can open the same build at /app/test. It runs on demo data from the Northline Test Ops demo workspace.
A 0-to-99 score per case built from its attempt history. The checkout-race case scores 91 after failing 38% of 12 attempts; anything ≥55 joins the 50 open flakes on the overview card.
156 of the 360 cases were drafted by the assistant and carry an AI chip on the row. They are reviewed and owned like any other case — the chip is attribution, not an exemption.
When a UI selector drifts, the case repairs its own selector instead of failing — 8 times in the last 30 days, each marked healed on the row so a person can review the repair.
No certification has been issued. What we can evidence about hosting, encryption, tenant isolation and production access is written up on the security page.
Live build, demo data, no card. Then look up the flake score of the test your team re-runs until it goes green.