Ryvl — seller market intelligence
Competitive pricing, competitor tracking and demand signals for Pakistani
e-commerce sellers. A seller connects their catalog and sees where they sit
against the market: who undercuts them, on what, and what to do about it.
This runs. It is a deployed Next.js 15 app plus a scraper that runs on a
cron. It is not an archive or a prototype.
Previously this file described the repo as a non-running archive extracted
out of a job board ("JobLo"). That was true in July 2026 and has been wrong
for months. If you are looking for that history it is in
docs/history/.
Repo layout
Two packages are live. Everything else is reference material.
| Path | What it is |
|---|
app/ | The app. Next.js 15 App Router + Supabase. Own package.json and lockfile. Most work happens here. |
scraper/ | Separate npm package. Scrapes marketplaces on a cron, writes with a service-role key. |
scraper/migrations/ | All SQL migrations live here — including the app's own tables, not just the scraper's. |
.github/workflows/ | CI and all cron triggers. At the repo root, not inside the app package. |
docs/ | Everything that isn't code. See docs/README.md. |
assets/landing-page/, assets/brand/ | Brand and landing-page source assets. See docs/ASSETS.md. |
vendor/agency-landing-template/, vendor/tailadmin-dashboard/ | Vendor UI templates kept as visual reference. Not built, not deployed, not imported. |
The two vendor template directories are the most common source of confusion
here: they are third-party downloads the live UI was modelled on. Nothing in
app/ imports from them.
Getting started
cd app
npm install
npm run dev
You need a .env.local with Supabase credentials. next build needs the two
public vars present but they do not have to be real:
NEXT_PUBLIC_SUPABASE_URL=https://placeholder.supabase.co \
NEXT_PUBLIC_SUPABASE_ANON_KEY=placeholder npm run build
If the build ever starts needing real credentials, that is a bug.
The verification bar
Before calling anything done — all four, from
app/:
npx tsc --noEmit # typecheck
npm run lint # one known <img> warning is expected
npm run test # vitest
npm run build # see env vars above
CI runs exactly these. The build is not optional: it is the only step
that catches Server/Client component boundary errors, and it has caught real
ones here that tsc, lint and the full test suite all passed.
How the app is put together
- Pages are Server Components. The established shape across
/dashboard/* and /apps/* is page.tsx (async, resolves the seller and
fetches) → SomethingView.tsx ('use client', pure rendering, takes data
as props) → loading.tsx (route-level skeleton). No page fetches its own
data from the client; after a mutation, call router.refresh().
- Data access lives in
src/lib/market-intel/, split into core/,
market/, seller/ and jobs/. These functions throw on error; route
handlers are thin wrappers that catch and shape the response.
- Migrations are applied by hand via the Supabase SQL Editor. There is no
migration tracking table — run
scraper/migrations/_check_applied.sql
(read-only) to see what is actually live.
CLAUDE.md has the fuller conventions and the known gotchas.
Where to look next
Known state, honestly
- Data coverage is no longer the binding constraint. As of 2026-09-15,
all 12 real seller categories have scraped rows (978 to 25,499 products
each, all refreshed within a day) — Daraz plus two batches of
Shopify/WooCommerce sources closed the gap OLX used to be the only cover
for. OLX itself is still disabled. See
ROADMAP.md Phase D's 2026-09-15
update for the full breakdown, and re-run
scraper/migrations/_check_coverage.sql
before relying on this if it's been a while — coverage moves with every
scraper run.
- There is no billing. Plan tiers exist in
entitlements.ts but demo mode
unlocks everything and no checkout exists.
- The account naming
market_analyst / market_accounts predates the
seller-focused positioning and is due a rename.
