project-f0c97df3

byAbhilash

I'm building CiteTrack, an AI-visibility tracker (AEO/GEO monitoring tool). Full specs are in my project docs — read these first before writing any code: - 01-PRD.md (problem, scope, MVP boundaries) - 08-Source-Code-Overview.md (stack + folder structure) - 09-Database-Schema.md (Convex schema) - 10-API-Documentation.md (internal + external API contracts) - 02-User-Stories.md (flows to support) Stack: Next.js (App Router) on Vercel, Convex for backend/database, Clerk for auth, Dodo Payments for billing, Resend for email. AI providers: OpenAI + Anthropic only for v1 (Gemini + Perplexity come in Phase 1 — don't build for them yet). Task: scaffold the project. 1. Initialize a Next.js app with the folder structure from 08-Source-Code-Overview.md 2. Set up Convex and implement schema.ts exactly as defined in 09-Database-Schema.md 3. Wire up Clerk auth (sign up / sign in, tied to the `users` table via clerkId) 4. Build the onboarding form: business name, description, up to 5 questions (validation: name required, min 1 / max 5 questions — per 11-Test-Plan.md P0 cases) 5. Implement `businesses.create` and `questions.addBatch` mutations per 10-API-Documentation.md 6. Stub out `scans.runScan` action — loop questions × [OpenAI, Anthropic], call each provider with a plain unframed question, save raw response to `scanResults` 7. Implement `scoring.detectMention` per the test cases in 13-Automated-Tests.md (case-insensitive, handles suffix variations like "Inc"/"LLC", no partial-word false positives) 8. Results table UI showing per-question, per-model mention status Don't build: Gemini/Perplexity integration, scheduled scans, email digest, or the paywall yet — that's Phase 1, listed explicitly out-of-scope in 01-PRD.md section 6. Ask me before making architecture decisions not already specified in the docs.