Track product prices across the web. Get notified when a price drops more than your threshold.
Try it at dipcatch.eu.
- Public homepage at
/pitches the product and collects wait-list emails (table:waitlist_signups, IP-rate-limited 5/hr). - User creates a product and attaches one or more shop URLs to it. The probe step auto-extracts price/title/image from the page (host-specific adapter → JSON-LD → microdata → OpenGraph → generic), or accepts user CSS selectors when extraction can't find a price.
- App re-fetches each active shop on a schedule; the cheapest live shop sets the product's headline price.
- When the cheapest price drops more than
X(absolute or %) below a reference price, user gets a notification. - Dashboard lists every tracked product with image, name, cheapest shop, cheapest price, and per-product drilldown (price history chart + per-shop status).
- "Add shop" form: URL → live probe → preview/variant chooser → confirm. URLs are normalized + deduped per product.
- Optional per-product public share link (unguessable 32-char slug at
/p/{slug}, owner-rotatable,noindex,nofollow): exposes only product summary, eligible shops, and price-history chart — no notes, thresholds, or private fields.
- Product — the abstract item: title, image, currency, drop thresholds, denormalized
cheapest_shop_id+cheapest_price. - Shop — a URL at a specific webshop attached to a product; carries health (
ok/failing/dead), per-host failure counters, optional user-supplied CSS selectors, and the persistedadapter_keyhint. - PriceCheck — timestamped per-shop fetch result (price, currency, status, error if any).
- ProductCheapestHistory — segmented timeline of which shop was cheapest when, used for the price-history chart and the drop-reference window.
- PriceDropEvent — denormalized notification row carrying
price_check_id+triggered_by_shop_id; powers chart markers and the savings widget without JSON probing. - Notification — emitted when the cheapest shop's newest
PriceCheckprice drops by more than threshold vs. a 30-day median reference (falls back to initial price when <7 samples).
- PHP 8.5 + Laravel 13
- Filament v5 (admin panel + user-facing app panel) + Fortify auth (open registration, invitations optional)
- Livewire 4 + Flux 2
- Postgres on Laravel Cloud
- HTTP-only fetching via
ShopFetcher(robots.txt + SSRF guard + WAF detection + per-host rate limit + body cap), with a chain-of-responsibilityAdapterResolverover price extractors — no headless browser in v1 - Laravel Octane (FrankenPHP) app runtime
- Queue worker + scheduler on Laravel Cloud
- Mail + Filament in-app bell + web push (
laravel-notification-channels/webpush) - Pest 4, PHPStan (Larastan), Pint, Rector — full quality stack
Implementation-ready specs live in specs/. All v1 launch specs and post-launch refactors are shipped; their files are removed once fully implemented (see the strikethrough entries in the spec index for what shipped). notifications.md still has a deferred Phase 5 admin queue widget.
composer setup # install deps, copy env, key:generate, migrate, build assets
composer dev # queue + scheduler + pail + vite (concurrent); Herd serves the siteThe suite runs on Postgres only — the engine production runs — and in parallel:
createdb dipcatch_test # once; psql -c 'CREATE DATABASE dipcatch_test'
composer test # pest --parallel
composer test-serial # same suite, one process, for debuggingphpunit.xml.dist points at 127.0.0.1:5432 as postgres/postgres; override with the usual
DB_* variables in the environment when your local server differs. Each parallel worker gets its
own database and its own Redis prefix. Do not start a serial run the moment a parallel one ends —
the worker databases are still being torn down.
Herd serves the site at dipcatch.test, so composer dev starts no web server of its own: it runs
the queue listener, the scheduler, the log tailer and Vite. The scheduler matters locally — the price
rechecks, the daily digest, the adapter canary and the health heartbeat all reach you only through it.
To exercise production's runtime instead, run php artisan octane:start --watch yourself. It is still
a dependency and still what Laravel Cloud runs; the FrankenPHP binary downloads on first start and is
gitignored.
DipCatch targets Laravel Cloud — fully managed, no SSH/Docker.
- App runtime: Laravel Octane (FrankenPHP) — select the Octane runtime in the Laravel Cloud dashboard;
laravel/octane+config/octane.phpship in the repo. Deploys reload workers automatically. Remember: code changes only apply after a worker reload. - Postgres add-on attached →
DATABASE_URLinjected. - Queue worker (1 small instance),
redisdriver. Runsphp artisan queue:work --tries=1— everything (scrape checks, daily digests, notifications) runs on thedefaultqueue; split into dedicated queues later if volume demands it. - Scheduler enabled — runs
php artisan schedule:runminutely.
| Key | Purpose |
|---|---|
APP_URL |
Public base URL (used in invitation + push payload links). Production: https://dipcatch.eu. |
APP_KEY |
Standard Laravel key. |
DATABASE_URL |
Auto-injected by Laravel Cloud. |
MAIL_* |
Mail provider for invites + drop notifications + alerts. |
MAIL_FROM_ADDRESS / _NAME |
From-address shown to recipients. |
ADMIN_NAME / ADMIN_EMAIL / ADMIN_PASSWORD |
Seeded by AdminUserSeeder on db:seed. |
DIPCATCH_FETCHER_USER_AGENT |
Custom UA for ShopFetcher (default: a real-browser UA, since Cloudflare/Akamai block any UA that admits to being a bot). Override only when a shop's robots.txt demands a bot UA. |
DIPCATCH_FETCHER_RATE_LIMIT_PER_MINUTE |
Per-host fetch budget (default 30). Shared by probe + queued recheck paths. |
DIPCATCH_RECHECK_INTERVAL_HOURS / _JITTER_MINUTES |
Recheck cadence + jitter (defaults 6 / 30). |
DIPCATCH_SHOP_FAILING_AFTER / _DEAD_AFTER |
Health thresholds on the main failure counter (defaults 3 / 10). |
DIPCATCH_SHOP_FAILING_5XX_AFTER / _DEAD_5XX_AFTER |
Separate higher thresholds for transient upstream 5xx (defaults 10 / 30). |
VAPID_SUBJECT / VAPID_PUBLIC_KEY / VAPID_PRIVATE_KEY |
Web push (generate via php artisan webpush:vapid). |
FAILED_JOB_MONITOR_NOTIFIABLE |
Comma-separated admin emails for failed-job alerts. |
Health checks are exposed inside the admin panel at /admin/health-check-results. Active checks: environment, debug-mode, cache, database, schedule, used-disk-space, CPU load, security advisories, plus the custom LastSuccessfulScrapeCheck (warns if any active shop on an active product hasn't seen a successful fetch in 48h, fails after 96h) and CheckjebonFreshnessCheck (warns when the daily checkjebon price dataset is older than 48h, fails after 96h, hard-fails when empty while checkjebon shops exist).
Cache + queue + unique-job locks all run on Redis (CACHE_STORE=redis, QUEUE_CONNECTION=redis). The Redis connection backs withoutOverlapping() on the scheduler, the per-host ShopFetcher rate-limit bucket, and ShouldBeUnique locks on CheckShopPrice.
- Provision Postgres add-on.
- Set the env vars above.
- Push to
main→ Laravel Cloud builds + runs migrations. - Run
php artisan db:seed --class=AdminUserSeeder --forceonce to create the initial admin user (--forcepasses the production guard; the class scope skips local demo seeders). - Sign in at
/login, create invitations from/admin/invitations.