Stack Overview

Next.js 15 · Vercel Edge · Upstash KV · Vercel Blob · Cloudinary · Discord

  Client browser
       │
       ▼
  antcpu.com/manda/          ← Static HTML · hosted on antcpu.com
  antcpu.com/manda/agent/    ← Static HTML · booking agent UI
       │
       │  fetch() calls
       ▼
  amandaland.vercel.app      ← Next.js 15 App Router on Vercel Edge
       │
       ├── /api/assets        ← Upstash KV read
       ├── /api/search        ← Upstash KV scan
       ├── /api/stats         ← Upstash KV read + Discord ping
       ├── /api/chat          ← Scripted engine · no external AI
       ├── /api/upload        ──┬── Vercel Blob write
       │                        ├── Cloudinary upload
       │                        ├── Upstash KV write
       │                        └── Discord webhook notify
       └── /api/auth          ← Cookie set/verify/delete

  Upload flow:
  ─────────────────────────────────────────────────────
  1. Client POSTs image to /api/upload
  2. File written to Vercel Blob → returns blobUrl
  3. File uploaded to Cloudinary → returns cloudinaryId
  4. Metadata written to Upstash KV:
       key: asset:{uuid}
       value: { id, filename, blobUrl, cloudinaryId,
                category, visibility, meta, uploadedAt }
  5. Event written to KV event log
  6. Discord webhook fires with filename + category
  7. Response returns full asset object

  Search flow:
  ─────────────────────────────────────────────────────
  1. GET /api/search?q=portrait
  2. KV scan returns all asset keys
  3. Filter by filename, title, category, EXIF match
  4. Returns { count, results[], defaults{} }

  Booking agent flow:
  ─────────────────────────────────────────────────────
  1. Client POSTs message to /api/chat
  2. Intent matched against keyword banks
  3. Multi-step booking flow tracked per session
  4. Reply returned in < 400ms
  5. On booking completion → /api/notify fires Discord