Hooked

Hooked is an API-first lab for spinning up ultra-high performing hooks for TikTok, Reels, Shorts, and YouTube long form. The MVP ships a dark, neon-tinged dashboard plus REST endpoints Idolforge (or any other app) can call directly.

Stack

• Next.js 14 App Router + TypeScript
• Tailwind CSS for the cyber/creator UI
• Prisma + SQLite (easily portable to Postgres)
• NextAuth Credentials flow with bcrypt hashing
• OpenAI GPT-4o-mini integration (swap for any provider inside lib/services/aiClient.ts)
• SQLite doesn’t support native enums/JSON, so arrays/enums are stored as JSON strings in Prisma models and hydrated via helper utilities.

Key features

• Email/password auth (/signup, /login)
• Dashboard showing avatars + latest hooks
• Avatar builder with presets, tag inputs, and multi-platform focus selection
• Hook Lab that generates 10-hook stacks per platform, saves results, and exposes them via API
• Trend service abstraction with mocked data + LLM clustering (ready for real APIs)
• External POST /api/hooks/generate for Idolforge or other partners (API key based)

Getting started

1. Install deps (requires Node 18+):

   npm install

2. Copy env template:

   cp .env.example .env

3. Fill the env vars:
   • DATABASE_URL – e.g. file:./dev.db for local SQLite
   • NEXTAUTH_SECRET – any long random string
   • OPENAI_API_KEY – optional. Without it, mocked hooks are returned
   • HOOKED_API_KEY – secret token for external API access
4. Prep the database:

   npx prisma migrate dev --name init

   (run npx prisma generate whenever the schema changes)
5. Start the dev server:

   npm run dev

6. Visit http://localhost:3000/signup to create your first account.

API

POST /api/hooks/generate

• Internal (requires session cookie)

  {
    "avatarProfileId": "cuid",
    "platforms": ["TIKTOK", "YOUTUBE_SHORT"],
    "workingTitle": "",
    "description": "",
    "goal": "go viral",
    "useTrends": true
  }

  Responds with hooks persisted in Prisma. Each hook includes text, styleTags, predictedScore, trendAlignment, rationale, timestamps, etc.
• External (API key mode)

  {
    "apiKey": "HOOKED_API_KEY",
    "avatarProfile": {
      "name": "Neon Cherub",
      "type": "AI_INFLUENCER",
      "toneOfVoice": "dark, poetic, seductive",
      "niche": "futuristic afro-club music",
      "likes": ["latex fashion", "underground raves"],
      "personalityTraits": ["mysterious", "witty", "rebellious"],
      "primaryLanguage": "English",
      "secondaryLanguages": ["French"],
      "platformFocus": ["YOUTUBE_SHORT", "TIKTOK"]
    },
    "contentIdea": {
      "platform": "TIKTOK",
      "workingTitle": "How I built my cyberpunk afro-rave sound",
      "description": "Breaking down my sound design for a new afro-futurist club track with distorted drums and alien vocals.",
      "goal": "go viral"
    },
    "useTrends": true
  }

  Responds with hook JSON but does not persist to the DB.

Other useful routes

• POST /api/auth/signup – create user
• GET /api/avatars / POST /api/avatars
• GET /api/hooks – latest saved hooks for the authenticated user
• GET /api/content-ideas – last 20 ideas + avatars

Services & extensibility

• lib/services/trendService.ts – drop in real YouTube/TikTok APIs later; today it mocks trending topics and uses an LLM to map angles.
• lib/services/hookGenerator.ts – core orchestrator pulling avatar/content/trend context, building the system prompt, parsing JSON, and saving output.
• lib/services/aiClient.ts – single integration point for swapping out the AI vendor.

TODOs

• Integrate real trend APIs + georegional filters
• Rate limiting on /api/hooks/generate
• Advanced scoring models + historical performance feedback
• Multi-language hook output toggle
• Avatar collaboration & sharing between teams

Running tests / lint

This MVP doesn’t ship tests yet. Before production hardening, add e2e coverage (Playwright) plus integration tests for the hook generator mocks, along with ESLint/Prettier in CI.