LossLens: Automated Profit Troubleshooter for E-commerce & Shopify Stores

Live: https://ecom-ai-troubleshooter.vercel.app · Built by S Surya Narayan

When an online store has a losing week, LossLens finds out why. Answer four quick questions, then paste the store's weekly PnL and any supporting data you have (ad campaigns, gender, age, platform, state, product, fulfilment, catalogue changelog). LossLens runs a strict 8-step troubleshooting method and returns:

• A dashboard: the four problem metrics (RTO, Spend/GMV, COGS, Cancel) with their ₹ impact, state RTO shifts, campaign spend shifts, and a verdict for every layer (✅ / ⚠️ / 🔴)
• A cross-cut root cause: campaign × one other factor (gender, product, geo, catalogue or carrier)
• A prioritised action plan with a recovery target
• The full case study as markdown (copy or download)

The troubleshooting framework is in prompts/l3-diagnostic.md. I built it from 12+ live seller case studies.

How it works

Browser (Next.js page)  ──POST /api/analyze──▶  Serverless function (Vercel)
   paste data / sample                             ├─ system prompt = prompts/*.md
   charts (Recharts)  ◀── JSON ─────────────────── ├─ Gemini (free tier, JSON validated by zod)
   markdown report                                 └─ rate limit + input caps

• One model call returns both the narrative report and typed dashboard JSON, validated against lib/schema.ts.
• Missing data is never invented. Empty breakdowns show as "No data" and are listed under data gaps.
• The API key lives only on the server (Vercel environment variable).

Stack: Next.js 15 · TypeScript · Tailwind CSS v4 · Recharts · Google Gemini (free tier) · Zod

Deploy (no local install needed)

1. Get a free Gemini API key at https://aistudio.google.com/apikey. No billing needed; the site only uses free-tier models.
2. Put the code on GitHub: create a new repository → Add file → Upload files → drag in everything in this folder → Commit.
3. Deploy on Vercel: sign in at https://vercel.com with GitHub → Add New → Project → import the repository.
4. Before you click Deploy, open Environment Variables and add:
   • GEMINI_API_KEY = your free key
5. Click Deploy. You get a public https://<name>.vercel.app link.

Every later commit to GitHub, including edits made in GitHub's web editor, redeploys automatically.

Feature switches

Two add-on features can be turned off or back on without changing any code. Set the variable in Vercel → Settings → Environment Variables, then Redeploy.

Variable	off means	Default
NEXT_PUBLIC_FEATURE_STUDIO	Section Studio page, API, nav link and home promo are removed (/studio returns 404)	on
NEXT_PUBLIC_FEATURE_TOKENS	Set to on to bring back the demo token wallet, pricing page and per-use costs	off

To turn a feature back on, delete the variable (or set it to on) and redeploy. Balances and saved sections live in visitors' browsers, so they come back too.

Studio code is self-contained, so you can also remove it from the repo entirely if you want: app/studio/, app/api/studio/, components/studio/, lib/theme.ts.

Environment variables

Variable	Used by
GEMINI_API_KEY	Everything: troubleshooter, Section Studio, Ask Lens (free tier, aistudio.google.com/apikey)
WALLET_SECRET	Optional: a long random string used to sign token balances
ANALYZE_MODEL / THEME_FREE_MODEL / THEME_PRO_MODEL / CHAT_MODEL	Optional model overrides (default: free Gemini models)

Customise

What	Where
Your name and links in the header	lib/site.ts
The diagnostic framework	prompts/l3-diagnostic.md
Web-app output rules	prompts/web-app-addendum.md
Demo dataset (fictional)	lib/sampleData.ts
Dashboard fields	lib/schema.ts + components/Dashboard.tsx

Guardrails

• About 5 requests per IP every 10 minutes (best effort, per server instance)
• 60,000-character input cap and an output token cap
• Refusals, truncation and API errors return readable messages

Run locally (optional)

npm install
cp .env.example .env.local   # add your key
npm run dev