ReviewIntel

Professional AI review intelligence SaaS for buyers, online sellers, and owner/admin operations.

ReviewIntel is platform-neutral. It works with manually pasted review text, TXT review batches, and quick screenshot uploads from Amazon, Walmart, Etsy, eBay, Shopify stores, AliExpress, Best Buy, TikTok Shop, Google Reviews, and future app-review sources. It does not scrape websites.

What is built

• Premium AI-themed landing page focused on the buyer moment of doubt: “Are these reviews real, and should I buy this?”
• Catchy first-screen hero with AI scan visuals, fake-review risk, buyer verdict, trust checks, review snippets, dark/light mode, and responsive layout
• Review analyzer with platform selector, buyer/seller mode, primary Paste Reviews workflow, multi-section bulk paste, TXT/CSV upload, Quick Scan Beta screenshots, preview, browser-side image compression, and optional screenshot stitching
• OpenAI Responses API route with structured JSON output and screenshot vision support when OPENAI_API_KEY is configured
• Real AI error handling: if OpenAI is configured but quota/billing fails, the API returns the OpenAI error instead of showing a mock result
• Results dashboard with product score, buyer recommendation, fake-review warnings, seller insights, keyword frequency, and issue categories
• Product comparison page for Buyer mode and Seller mode with 2-5 products, user priorities, winner badges, and report export
• Sponsored resources section for future affiliate tools, partner placements, and ecommerce service recommendations
• Login, sign up, Google login endpoint, password reset, email verification page, and local demo session fallback
• Separate Buyer, Seller, and private Admin/Developer experiences with route guards
• Stripe Checkout, Billing Portal, and signed webhook routes
• Supabase/PostgreSQL schema for users, profiles, subscriptions, uploaded images, review analyses, saved products, comparisons, usage tracking, app settings, billing history, sponsors, logs, and abuse reports
• Guest quota: 3 free analyses total before signup
• Free Buyer quota: 3 product analyses per UTC day, persisted in Supabase when production keys are configured
• Private Admin/Developer access bypasses all usage limits and can simulate Buyer Pro, Seller Starter, and Seller Pro for testing
• Confidence scoring by review volume: Low under 10 reviews, Medium from 10-50 reviews, High at 50+ reviews
• Rate limiting, max request/text/file protections, large-batch model chunking, upload validation, security headers, robots, sitemap, OpenGraph, and JSON-LD

How users scan a product

Open:

http://localhost:3000/analyze

Use:

• Quick Scan Beta: upload JPG/PNG screenshots for a fast mobile-style check.
• Deep Analysis: paste review batches or upload TXT/CSV for stronger confidence.
• Compare: open /compare to compare 2-5 products.

Product URLs are accepted as reference metadata only. ReviewIntel does not fetch or scrape the URL in this version.

Code Map

Main pages:

• app/page.tsx: homepage and first impression hero
• app/analyze/page.tsx: analyzer page shell
• app/results/page.tsx: latest result page shell
• app/compare/page.tsx: product comparison page
• app/pricing/page.tsx: pricing page
• app/admin-access/page.tsx: hidden admin unlock route
• app/admin/page.tsx: admin/developer control center

Main components:

• components/Header.tsx: top navigation
• components/AnalyzerForm.tsx: paste reviews, TXT/CSV upload, screenshot upload, scan animation, quota UI
• components/ResultsDashboard.tsx: separated Buyer and Seller result dashboards
• components/ResultsClient.tsx: reads latest session result and demo result loaders
• components/SellerIntelligenceTabs.tsx: interactive Seller Pro signal tabs for clusters, keywords, and risks
• components/SellerReportActions.tsx: functional Seller report export and local save actions
• components/SellerProCommandPanel.tsx: interactive Seller dashboard command board
• components/CompareForm.tsx: Buyer/Seller comparison workflow
• components/DeveloperControlCenter.tsx: admin system controls and plan simulation
• components/DeveloperQACenter.tsx: QA checklist and diagnostics
• components/SponsoredResources.tsx: future sponsored/affiliate resource cards

Core logic:

• app/api/analyze/route.ts: protected analysis API, OpenAI call, quota/rate limits, local fallback
• lib/prompts.ts: AI prompt architecture
• lib/reviewSchema.ts: structured JSON schema expected from AI
• lib/reviewIngestion.ts: paste/TXT/CSV normalization and valid review counting
• lib/localAnalyzer.ts: local analysis fallback when OpenAI is not configured
• lib/supabaseServer.ts: Supabase persistence, quota, profiles, analysis history
• lib/stripe.ts: Stripe checkout and billing portal helpers
• lib/account.ts and lib/clientAccount.ts: roles, plans, local account state, quota labels
• db/schema.sql: Supabase/PostgreSQL schema

Setup

npm install
cp .env.example .env.local
npm run dev

Open http://localhost:3000.

Environment

OPENAI_API_KEY=
OPENAI_MODEL=gpt-5.2
STRICT_OPENAI_ERRORS=true
DATABASE_URL=
NEXT_PUBLIC_APP_URL=http://localhost:3000

NEXT_PUBLIC_SUPABASE_URL=
NEXT_PUBLIC_SUPABASE_ANON_KEY=
SUPABASE_SERVICE_ROLE_KEY=

STRIPE_SECRET_KEY=
STRIPE_WEBHOOK_SECRET=
STRIPE_BUYER_PRO_PRICE_ID=
STRIPE_SELLER_STARTER_PRICE_ID=
STRIPE_SELLER_PRO_PRICE_ID=

NEXT_PUBLIC_SPONSORS_ENABLED=true

REVIEWINTEL_ADMIN_CODE=
REVIEWINTEL_ADMIN_CODE_HASH=
REVIEWINTEL_ADMIN_SESSION_SECRET=
[email protected]

The app can render without Supabase or Stripe in local demo mode, but public paid launch needs Supabase and Stripe configured. If OPENAI_API_KEY is set and OpenAI returns a quota, billing, model, or network error, ReviewIntel returns that error instead of a placeholder result. In local development only, /admin-access accepts reviewintel-dev-admin when no admin code environment variable is set.

Key Routes

• / landing page
• /analyze deep paste/TXT analyzer plus quick screenshot analyzer
• /results latest analysis dashboard
• /compare Buyer/Seller product comparison for 2-5 products
• /login, /signup, /auth/reset-password, /auth/verify-email
• /dashboard, /dashboard/customer (Buyer dashboard), /dashboard/seller
• /account, /pricing
• /admin-access private admin unlock route
• /admin System Control Center and Developer Mode

API

POST /api/analyze

{
  "productName": "Optional product or business name",
  "productUrl": "Optional reference URL. The API does not fetch it yet.",
  "platform": "shopify",
  "audience": "both",
  "reviewSections": [
    {
      "id": "batch-1",
      "title": "Recent negative reviews",
      "text": "Pasted review text...",
      "source": "paste"
    }
  ],
  "ingestionMode": "deep_paste",
  "imageAggregation": "individual",
  "images": [
    {
      "name": "reviews.png",
      "type": "image/png",
      "dataUrl": "data:image/png;base64,...",
      "size": 120000
    }
  ]
}

Returns structured JSON including:

• overall_summary
• positive_points
• negative_points
• common_complaints
• praised_features
• sentiment_score
• confidence_score
• buyer_recommendation
• seller_insights
• improvement_suggestions
• feature_requests
• quality_concerns
• packaging_issues
• durability_issues
• support_issues
• keyword_analysis

Supabase

1. Create a Supabase project.
2. Run db/schema.sql in the SQL editor.
3. Create a private storage bucket named review-screenshots.
4. Add Supabase environment variables to Vercel.
5. Wire auth callback URLs to /dashboard and /auth/verify-email.
6. Confirm SUPABASE_SERVICE_ROLE_KEY is present only in server-side environments. It powers persistent quota, profile creation, saved analysis history, and Stripe webhook updates.

Review Ingestion Roadmap

• Deep Analysis = Paste Reviews: primary path for long review batches, TXT uploads, multiple pasted sections, buyer decisions, and seller analytics.
• Quick Scan Beta = Screenshot Upload: lightweight path for mobile users and fast checks from a few visible reviews.
• Mixed Evidence: combines pasted review batches with quick screenshots.
• Product URL Import: reserved architecture only; no fetching or scraping is implemented yet.
• Ecommerce Connectors: future approved APIs, seller-owned store integrations, CSV feeds, and partner connectors.

Large pasted batches are accepted up to the configured server cap, then chunked before model analysis for stability. Admin accounts bypass quota/rate limits for testing, while public users remain protected by request size, file size, and daily usage gates.

Stripe

1. Create monthly prices for Buyer Pro ($9.99), Seller Starter ($29), and Seller Pro ($79).
2. Add STRIPE_BUYER_PRO_PRICE_ID, STRIPE_SELLER_STARTER_PRICE_ID, STRIPE_SELLER_PRO_PRICE_ID, and STRIPE_SECRET_KEY.
3. Set webhook URL to /api/stripe/webhook.
4. Add STRIPE_WEBHOOK_SECRET.
5. Webhook events upsert subscriptions and billing_history.

Admin / Developer Mode

Admin accounts do not use public signup/login. Open /admin-access, enter the private admin code, then use /admin for the System Control Center. Admin sessions are server-validated, hidden from search indexing, rate-limited on access attempts, and auto-expire.

• Toggle AI analysis, payments, sponsored resources, announcements, and maintenance-mode messaging
• Simulate Buyer Pro, Seller Starter, or Seller Pro without paying
• Reset quota instantly for local testing
• View API usage, failed webhook, screenshot, sponsor, and abuse-report placeholders

Launch Checks

npm run lint
npm run typecheck
npm run build

Before selling access, run one real /analyze request with a funded OpenAI project. A 429 insufficient_quota response means the app is correctly calling OpenAI, but the OpenAI billing/quota setup is not ready yet.

Sponsored Resources

Sponsored placements are intentionally designed as a small resource library, not banner ads. The current build uses lib/sponsors.ts placeholders and can be disabled with:

NEXT_PUBLIC_SPONSORS_ENABLED=false

The Supabase schema includes sponsors and sponsor_events tables for future admin add/edit/remove controls, affiliate URLs, active windows, placements, impressions, and clicks.

Vercel Deployment

1. Push the repo to GitHub.
2. Import the project in Vercel.
3. Add all production environment variables.
4. Deploy.
5. Confirm /robots.txt, /sitemap.xml, /api/analyze, and Stripe webhook routes.

Security Notes

• No direct scraping is implemented.
• API requests are rate-limited in memory for local/Vercel runtime protection.
• Uploads are limited to JPG/PNG data URLs and size-checked before analysis.
• Protected dashboards set Cache-Control: no-store.
• Production usage limits should move from memory to usage_tracking.