Cruise

Set a lead target and a monthly budget per city. Cruise holds them.

It clones a campaign that already works into a new city — resolving the geo key, the catalogue product set and the creative for that city rather than copying the old one — then watches every city against its target and its budget, and stops the ones that are done.

This is a full working build: real authentication, a real SQLite database, a real rules/scoring engine, and a mock (clearly isolated) Meta Marketing API layer so the product runs end-to-end without live ad spend or real API credentials.

Data provenance (updated 2026-09-01). The campaign, ad set, targeting, spend and result data in this build is pulled from the live Meta account "Used Cars Sales" (10153410655459601, CarDekho) over its trailing 30 days — see scripts/real-account-data.ts. An earlier version of this build shipped with an invented nine-city dataset; that has been removed. Catalogues and product sets are real too (scripts/real-catalogue-data.ts): the account runs one pan-India vehicles catalogue, Catalog_Vehicles-Pan-India-20-July, with a dynamic product set per city filtered on city — Ahmedabad 1,230 vehicles, Nashik 1,054, Chandigarh 460, Mohali 412. Those product counts are now the inventory figures for those cities, so the only city still carrying a generated inventory number is Delhi NCR, which has no product set because its campaign is a lead form rather than a catalogue ad.

---

1. Quickstart

Requirements: Node.js 18.18+ (20 LTS recommended), npm.

npm install
cp .env.example .env
npm run setup     # creates the SQLite schema and seeds 9 cities / ~8,800 vehicles / 22 campaigns / 30 days of performance + leads
npm run build
npm start          # production build, http://localhost:3000

or for local iteration:

npm install
cp .env.example .env
npm run setup
npm run dev         # http://localhost:3000

To wipe and regenerate demo data at any point: npm run db:reset.

On a restrictive/corporate network: better-sqlite3 ships prebuilt native binaries for all major platforms and does not need to compile anything — but npm's default install step still attempts a node-gyp rebuild, which tries to download Node.js headers from nodejs.org even though it ends up unused. If npm install fails with a header-download/403 error, run npm install --ignore-scripts instead — the prebuilt binary loads directly and everything above still works.

Demo credentials

npm run db:seed gives each seeded user a different random password and prints them once, at the end of the run. They are not stored anywhere else and cannot be looked up later — copy them from the terminal, or reset them from Settings → Users & Roles.

Before sharing a deployed URL: change those passwords, or delete the seeded accounts and create real ones. A published credential on an install that can spend money in a live ad account is the whole ballgame.

Role	Email	Can do
Admin	[email protected]	Everything: campaigns, budgets, integrations, users, settings
Marketing Manager	[email protected]	Create/edit campaigns, approve/reject AI recommendations, run inventory imports — no billing/user/integration admin
Marketing Manager	[email protected]	Same as above (second manager seeded for the Users table demo)
Viewer	[email protected]	Read-only: dashboards, reports, copilot

---

2. What's implemented

Scope for this build, per the agreed "Must-Have + AI Copilot + Vehicle-level scoring" depth:

여기까지가 README 앞부분입니다. 나머지는 원문에서 읽을 수 있습니다. README 전문 보기 ↗