S12 — the customer's release gate

One script that says red or green before a change reaches a live shop. This sample is for a customer's team (or a partner's) who ship changes to a shop — an app, a manifest, a price or tax rule — and want a pass or fail in their pipeline before every release.

When to use it: before every release, in your pipeline, against a staging shop.

release_gate.sh --baskets baskets.json [--app DIR] [--manifest FILE] runs, against a staging shop:

your recorded baskets (with the totals they expect),
the money kit (preview = booking),
the self-check on your app.

It exits 0 only if all are green, 1 if any is red, 2 if it could not run. Recipe: recipes/release_gate.md.

⚠ Staging only: it books sales and closes till days.

The files

File	What it is
`release_gate.sh`	the gate; the block marked `HERE` is the only part a customer changes
`baskets.norecord.json` → `baskets.json`	five baskets without, then with, their recorded answers
`record_expect.py`	fills the answers from a clean run's own record
`pipeline.example.yml`	the gate as a pipeline job (an example; not run by a CI service)
`tests/test_s12.py`	13 laptop tests with a stub shop

What you need

The money kit, tools/moneykit/, installed on the staging shop's server and pointed at your shop:

a shop file (cashier, enrolled device, POS profile, terminal);
if your console runs a script's text, the kit's folder.

Pass both through GATE_KIT_SETTINGS (kit_dir:…;shop:…;record:…). GATE_KIT_SCRIPT names the script as your console command does (default tools/moneykit/kit.py).

How to run it

Write your baskets without answers (baskets.norecord.json). Run the gate once with --record on a clean staging shop, then fill in the answers from that run's own record with record_expect.py. That gives baskets.json.
Change the block marked HERE in release_gate.sh to point at your staging shop.
Run the gate, or put it in your pipeline (pipeline.example.yml):

./release_gate.sh --baskets baskets.json --app path/to/your/app

What you see

First a green run, then a run with one recorded answer one unit off (the self-check's rule lines trimmed):

GATE   VERDICT PASS
GATE self-check GREEN
GATE   G001 OK  "two leather cases"  expected 2,598.00  booked 2,598.00  preview 2,598.00
GATE   G002 OK  "two leather cases, 10% markdown"  expected 2,338.20  booked 2,338.20  preview 2,338.20
GATE baskets GREEN: 5 of 5 matched
GATE open POS days: 0 GREEN
GATE RESULT GREEN — safe to release

GATE   G002 RED "two leather cases, 10% markdown"  expected 2,339.20  booked 2,338.20  preview 2,338.20  <- diverged: grand_total: recorded answer 2339.20, booked 2338.20
GATE baskets RED: 4 of 5 matched
GATE RESULT RED — do not release

A red basket names itself, with the figure you expected, the figure booked and the figure previewed.

Technical notes

What was run, and where

The output above is what the gate printed on our test bench (2026-10-07; the self-check's rule lines trimmed). The red run was without --app, so its self-check was skipped.

The money kit runs in the staging shop's bench.

What was tested

Where	What
`tests/test_s12.py`	13 laptop tests with a stub shop
our test bench's own run (not in the kit)	16 checks on our test bench

This page in the kit