Skip to content

Latest commit

 

History

History
199 lines (151 loc) · 10.3 KB

README.md

File metadata and controls

199 lines (151 loc) · 10.3 KB

appdevelopsk.github.io

Production deployment verification (no GitHub Actions)

The production domain is https://appdevelopsk.com, but this repository does not contain a deploy-provider configuration or CNAME. Do not assume merging main publishes this domain. Check the hosting provider's project settings for the actual source repository, production branch, deployed commit and production domain before making changes to DNS or hosting.

From a clean checkout of the intended production commit:

python3 -m unittest discover -s scripts -p 'test_*.py' -v
node --test scripts/test_catalog_search_events.cjs
node --test scripts/test_guide_store_events.cjs
python3 scripts/diagnose_growth_deploy.py --root . --output growth-deploy-diagnosis.json
python3 scripts/diagnose_growth_routes.py --output growth-route-diagnosis.json
python3 scripts/check_growth_live.py > growth-live-smoke.json

The route diagnosis additionally compares /apps and /apps.html, recording final URLs, public cache headers, canonical tags and hashes to uncover split cache/rewrite behavior. Inspect both results before editing redirects; the command makes no hosting or DNS changes.

The deploy diagnosis checks HTTP 200 and compares the raw HTML bytes of the homepage, app catalog, five priority app support pages, and five press locales to the checked-out commit. It records hashes, release markers and fetch errors, not page bodies or user data. A byte mismatch is not necessarily failed deployment if the host minifies/injects content; consult the semantic growth smoke and inspect the hosting deployment logs. A matching milestone meta tag does not establish the exact deployed commit. Do not equate site-to-Play clicks with installs.

After publication, manually verify catalog search, zero-result reset, app referrals and real GA4 DebugView events, then reconcile actual first-time installers in Google Play Console. Keep tracking in the release verification issue until those checks have evidence.

App support search snippet deployment gate (2026-11-13)

The Debate Me, BrainBurst, and TaleTime titles and descriptions were updated on main, but merging source is not proof that the public host received it. Run this read-only check after an authorized deployment and compare the local source against the public HTML (titles, descriptions, Open Graph descriptions, canonical URLs, and duplicate tags):

node --test scripts/test_support_serp_snippets_20261113.cjs
python3 -m unittest discover -s scripts -p 'test_support_serp_live.py' -v
python3 scripts/check_support_serp_live.py > support-serp-live.json

A nonzero exit means the live host is stale, unavailable, or inconsistent with source. Inspect support-serp-live.json and the hosting provider's deployed branch/SHA before changing DNS or announcing the release. The tool does not request indexing, change the site, or prove search clicks or app installs.

Global press/creator landing pages — 2026-11-13

Source contains five intended localized press routes: /press/ (English), /press/es/ (Spanish), /press/pt-br/ (Brazilian Portuguese), /press/ja/ (Japanese), and /press/id/ (Indonesian). This is source coverage, not proof of production delivery or indexing. No automated outreach or social publication follows from a merge.

Before promoting these URLs to partners:

python3 -m unittest scripts.test_press_kit -v
python3 -m unittest scripts.test_diagnose_growth_deploy -v
python3 scripts/diagnose_growth_deploy.py --root . --output growth-deploy-diagnosis.json

The deployment comparison includes all five press pages (12 pages in total). Confirm the hosting provider's actual production source/ref/commit, inspect public HTTP 200, redirects, canonical/hreflang and all five Play links, and verify press_app_click in GA4 DebugView. A source merge or static test is not a live deployment check. Record results in issue #51. Only promote the published URL after these checks pass; otherwise use verified live app support/store pages.

Five-language catalog discovery UI (2026-11-13)

The /apps page has browser-language-aware interface labels in English, Japanese, Spanish, Brazilian Portuguese and Indonesian, plus an explicit language selector. Category labels, five featured-app descriptions, finder status, empty state, and selected search synonyms are localized. App card names and descriptions remain English. No SEO hreflang pages are claimed and canonical stays /apps. The allowlisted ?lang=es etc. query is UI preference, not an indexed translation. The ?goal=... filter is preserved during manual language changes.

Search text is never sent to analytics; existing coarse search events remain. The only new event is catalog_language_selected, carrying the selected locale and placement without search text or sensitive identifiers.

Run checks: node --test scripts/test_catalog_search_events.cjs node --test scripts/test_catalog_locale.cjs python3 scripts/check_growth_live.py > growth-live-smoke.json

Before claiming a live release, verify the hosting provider deployed SHA and open /apps?goal=quick-games&lang=es and /apps?lang=ja in a real browser. Confirm 92 cards, filters, language switching, console errors, real store destinations, and exactly one catalog_language_selected per manual choice in GA4 DebugView. Record source-vs-live evidence in GitHub issue #51. A merge is not deployment.

VPS production publish gate for the five-language app catalog

Source-only PRs are not production deployments. The repository now contains scripts/publish_app_discovery.sh (dry-run by default) and scripts/check_app_discovery_live.sh (read-only production smoke). The catalog locale selector is part of apps.html and is in the publish file allowlist.

On the authorized deployment machine with the correct SSH alias, verified domain mapping and a clean checkout of the intended main commit:

bash scripts/publish_app_discovery.sh --dry-run

Review every destination/changed file and the current public host configuration. Only after approving that exact deploy, invoke:

bash scripts/publish_app_discovery.sh --apply bash scripts/check_app_discovery_live.sh

The live smoke now requires the five catalog languages and their browser-side event contract, plus previously released growth pages. Check mobile browser behavior, GA4 DebugView and Play acquisition data separately. This workflow cannot be run against the owner's VPS from a disconnected ChatGPT container; no source merge or local test proves that /apps is updated on the public host.

Press-route and catalog deployment diagnosis (2026-11-13)

The read-only scripts/diagnose_growth_routes.py checks the public origin's /apps vs /apps.html behavior and all five creator locale routes: English, Spanish, Brazilian Portuguese, Japanese and Indonesian.

For each press route, it records HTTP status, final URL, public cache headers and a SHA-256 of the body, plus boolean checks for the expected HTML language, canonical URL, all reciprocal hreflang entries (including x-default), five store CTAs, and the press click analytics contract. It also requires the public sitemap to list all five routes. The app catalog must include the five-language UI selector and catalog_language_selected event source. A HTTP 200 response with a stale or incorrect fallback page is not a passing release.

Run on the authorized deployment machine with network access:

python3 -m unittest discover -s scripts -p 'test_diagnose_growth_routes*.py' -v
python3 scripts/diagnose_growth_routes.py --output growth-route-diagnosis.json

An exit code of 1 indicates one or more missing/ambiguous public routes, a stale catalog, or an apps alias mismatch. Review the per-route checks in the JSON file and hosting/VPS release logs before changing redirects or DNS. An HTTP failure reported in this tool is an observation from the machine running it, not definitive proof the origin is offline to every user. This script makes no publishing, cache-purge, tracking or DNS changes.

Press-to-store GA4 event unification (2026-11-13)

All five press locales retain the original press_app_click event for both app detail links and Play outbound links. A real Play click now additionally fires the website-wide download_click event with allowlisted event data: store=google_play, location=press, placement=press_kit_card, source=press_kit, app_slug, app_package, locale. A detail-page click does not emit download_click. These represent outbound links not installs. No raw search terms, URLs, private user data or unique identifiers are sent.

Run source tests:

node --test scripts/test_press_store_events.cjs
python3 -m unittest scripts.test_press_kit -v

Following an authorized VPS publish, the existing live smoke now requires unified press click source on each of the five routes. Use GA4 DebugView to verify one press_app_click per real press link click and one download_click per Play click, without counting app detail clicks as installs. Store acquired users must be verified separately in Play Console.

Fail-closed offline preflight before VPS publishing (2026-11-13)

The app-discovery publish command now runs a local Python stdlib-only source integrity check before both --dry-run and --apply. It fails and sends no files if a broken Google Play package link, duplicated/missing catalog card, incorrect schema count, missing 5-language catalog interface, incorrect press canonical/hreflang/sitemap, missing support destination or incorrect press-to-store analytics identity is detected.

Run from the intended production checkout:

python3 -m unittest scripts.test_preflight_app_discovery scripts.test_publish_preflight_gate -v
python3 scripts/preflight_app_discovery.py
bash scripts/publish_app_discovery.sh --dry-run

The JSON report marks source_only: true. A PASS proves only that the checked-out source files agree; it does not prove that the domain is already updated, that the Play listing is available in every country, or that first-time app installs increased. After reviewing the dry-run destination and file list on the authorized host, use the existing --apply flag, then the live-smoke scripts. Preserve a rollback copy or backup before applying. Do not execute an unintended VPS transfer just because the source preflight passes.