Migration wing-command to Tinyfish SDK

wing-command/.env.example

@@ -1,17 +1,7 @@
-# ===========================================
-# Wing Command — Environment Variables
-# ===========================================
+# TinyFish Web Agent API key 
+# Get yours at: https://agent.tinyfish.ai/api-keys
+TINYFISH_API_KEY="your_api_key_here"
 
-# Supabase (required) — https://supabase.com/dashboard
-NEXT_PUBLIC_SUPABASE_URL=https://your-project.supabase.co
-NEXT_PUBLIC_SUPABASE_ANON_KEY=your-anon-key
-SUPABASE_SERVICE_ROLE_KEY=your-service-role-key
-
-# TinyFish Scraper (required) — https://docs.tinyfish.ai
-TINYFISH_API_KEY=your-tinyfish-api-key
-# TINYFISH_API_URL=https://agent.tinyfish.ai/v1/automation/run  # sync (default)
-# TINYFISH_API_URL=https://agent.tinyfish.ai/v1/automation/run-sse  # SSE streaming
-
-# Upstash Redis (optional, recommended) — https://upstash.com
-# UPSTASH_REDIS_REST_URL=https://your-redis.upstash.io
-# UPSTASH_REDIS_REST_TOKEN=your-redis-token
+# Groq API key 
+# Get yours at: https://console.groq.com
+GROQ_API_KEY="your_api_key_here"

wing-command/.gitignore

@@ -23,8 +23,9 @@ yarn-debug.log*
 yarn-error.log*
 
 # Local env files
-.env*.local
 .env
+.env.local
+.env.*.local
 
 # Vercel
 .vercel

wing-command/README.md

@@ -1,37 +1,62 @@
-# Wing Scout - Super Bowl LX War Room
+# Wing Command
 **Live Demo: https://wings-command.up.railway.app/**
 
-**Flavor-first, mesmerizing, hyper-local chicken wing tracker for Super Bowl LX (Feb 8, 2026).**
+**Flavor-first, hyper-local chicken wing tracker powered by TinyFish**
 
-A "War Room" interface that scouts the best chicken wings near you in real-time using AI-powered parallel scraping from DoorDash, Uber Eats, Grubhub, and Google.
+A "War Room" interface that scouts the best chicken wings near you in real-time. Pick a flavor persona, enter your zip, and parallel AI browser agents hit DoorDash, Uber Eats, Grubhub, and Google simultaneously — streaming results back as each source completes.
 
 ## Visual Identity
 
-- **Theme:** "Midnight Turf" - Dark #050505 background with subtle grass texture grid
-- **Accents:** Neon Green (#39FF14) glows, stadium lighting effects
-- **Typography:** Bebas Neue (scoreboard style) + Inter (body)
-- **Cards:** Glassmorphism "Scout Cards" with backdrop blur
-- **Animations:** Framer Motion parallax, floating particles, "tackle-in" card entrances
+- **Theme:** "Midnight Turf" — dark `#050505` background with subtle grass texture grid
+- **Accents:** Neon Green (`#39FF14`) glows, stadium lighting effects
+- **Typography:** Bebas Neue (scoreboard) + Inter (body)
+- **Cards:** Glassmorphism Scout Cards with backdrop blur
+- **Animations:** Framer Motion parallax, floating particles, tackle-in card entrances
 
 ## Architecture
 
 ```
-Frontend (Next.js 14 App Router)
-  |-- page.tsx           -> War Room hero + flavor selector + results grid
-  |-- HeroVisuals.tsx    -> Parallax field lines + floating wing/confetti particles
-  |-- FlavorSelector.tsx -> 3 flavor personas with pulsing selection
-  |-- ZipSearch.tsx      -> Stadium-light illuminated zip input
-  |-- WingGrid.tsx       -> Glassmorphism Scout Cards grid
-
-Backend (API Route)
-  |-- /api/scout         -> Main endpoint (zip + flavor)
-  |-- lib/tinyfish-scraper.ts     -> TinyFish parallel scraping engine
-  |-- lib/geocode.ts     -> Nominatim (OpenStreetMap) geocoding
-  |-- lib/cache.ts       -> Upstash Redis caching layer
-
-Data
-  |-- Supabase (PostgreSQL + PostGIS)
-  |-- Upstash Redis (15-min TTL cache)
+┌─────────────────────────────────────────────────────────┐
+│                     Browser (Client)                    │
+│                                                         │
+│  FlavorSelector → ZipSearch → WingGrid (streams in)     │
+└──────────────────────────┬──────────────────────────────┘
+                           │ GET /api/scout?zip=&flavor=
+                           │ (SSE — results stream as agents finish)
+┌──────────────────────────▼──────────────────────────────┐
+│                   Next.js App Router                    │
+│                                                         │
+│  /api/scout/route.ts                                    │
+│    │                                                    │
+│    ├─ 1. geocode.ts ──► Nominatim (free, no key)        │
+│    │                                                    │
+│    ├─ 2. Groq LLM ────► generates 5–7 source URLs       │
+│    │      llama-3.3-70b-versatile                       │
+│    │                                                    │
+│    └─ 3. TinyFish SDK ─► Promise.allSettled             │
+│           client.agent.stream({ url, goal })            │
+│           │                                             │
+│           ├── Agent → DoorDash city wings page          │
+│           ├── Agent → Uber Eats city wings page         │
+│           ├── Agent → Grubhub city wings page           │
+│           └── Agent → Google search results             │
+│                                                         │
+│           Each agent fires EventType.COMPLETE           │
+│           → spots sent to client immediately via SSE    │
+└─────────────────────────────────────────────────────────┘
+
+No database. No cache. Pure in-memory — results live only for the request.
+```
+
+### TinyFish SDK event flow
+
+```
+client.agent.stream({ url, goal })
+  │
+  ├── EventType.STARTED   → agent confirmed running
+  ├── EventType.PROGRESS  → live step updates (shown in UI)
+  └── EventType.COMPLETE
+        └── RunStatus.COMPLETED → event.result → parse restaurants → SSE → client
 ```
 
 ## Flavor Personas
@@ -44,68 +69,46 @@ Users pick a team/flavor before searching:
 | **The Classicist** | Buffalo, Hot, Mild, Traditional, Cayenne | 🦬 |
 | **The Sticky Finger** | Honey BBQ, Garlic Parm, Teriyaki, Korean | 🍯 |
 
-Spots are scored 0-100 against the selected persona.
+Spots are scored 0–100 against the selected persona.
 
 ## Scraping Flow
 
 1. User enters ZIP + selects Flavor Persona
 2. **Geocode** via Nominatim (free, no API key)
-3. **Parallel scrape** (`Promise.allSettled`):
-   - DoorDash search results
-   - Uber Eats search results
-   - Grubhub search results
-   - Google search (hidden gem detection)
-4. **Deduplicate** by normalized name + address
-5. **Score** against flavor persona keywords
-6. **Cache** in Redis (15-min TTL) + persist to Supabase
+3. **Groq** (`llama-3.3-70b-versatile`) generates 5–7 real search URLs for the location
+4. **Parallel scrape** (`Promise.allSettled`) — one TinyFish browser agent per URL:
+   - DoorDash city wings page
+   - Uber Eats city wings page
+   - Grubhub city wings page
+   - Google search results
+5. **Deduplicate** by normalized name + address
+6. **Score** against flavor persona keywords
+7. Results stream back to the client as each agent completes
 
 ## Setup
 
 ### Prerequisites
 
-- Node.js >= 18
-- Supabase project (free tier works)
+- Node.js 22.x
 - TinyFish API key
-- Upstash Redis (optional but recommended)
+- Groq API key
 
 ### Environment Variables
 
-Create `.env.local`:
-
-```env
-# Required - Server
-SUPABASE_SERVICE_ROLE_KEY=your_service_role_key
-TINYFISH_API_KEY=your-tinyfish-api-key
-
-# Required - Client
-NEXT_PUBLIC_SUPABASE_URL=https://your-project.supabase.co
-NEXT_PUBLIC_SUPABASE_ANON_KEY=your_anon_key
-
-# Optional - Caching (highly recommended)
-UPSTASH_REDIS_REST_URL=https://your-redis.upstash.io
-UPSTASH_REDIS_REST_TOKEN=your_redis_token
-
-# Optional - Custom TinyFish endpoint
-TINYFISH_API_URL=https://agent.tinyfish.ai/v1/automation/run
+```bash
+cp .env.example .env.local
 ```
 
-### Database Setup
+Then fill in:
 
-Run the schema in your Supabase SQL editor:
+```env
+# TinyFish (required) — https://agent.tinyfish.ai/api-keys
+TINYFISH_API_KEY=your-tinyfish-api-key
 
-```bash
-# The schema file is at:
-supabase/schema.sql
+# Groq (required) — https://console.groq.com
+GROQ_API_KEY=your-groq-api-key
 ```
 
-This creates:
-- `wing_spots` table with `flavor_tags` (TEXT[]), `menu_json` (JSONB), and PostGIS spatial indexing
-- `geocode_cache` table (permanent zip-to-lat/lng mapping)
-- `scrape_queue` table (for background cron jobs)
-- `menus` table (cached restaurant menus)
-- PostGIS trigger for auto-computing `location` from `lat`/`lng`
-- Row Level Security policies
-
 ### Install & Run
 
 ```bash
@@ -115,97 +118,64 @@ npm run dev
 
 Open http://localhost:3000
 
-## Render.com Deployment (Migration from Vercel)
+## Deployment
 
-### Why Render?
+The app is deployed on Railway. Any Node-compatible host works — no external services to provision.
 
-Vercel serverless functions have a **60-second timeout** (300s on Hobby with Fluid Compute). Parallel scraping across 4 platforms can exceed this. Render Web Services have **no timeout limit**.
+### Render.com (alternative)
 
-### Render Setup
+Vercel serverless functions have a 60-second timeout. Parallel scraping across 4 platforms can exceed this. Render Web Services have no timeout limit.
 
-1. **Create a Web Service** on Render
-   - Connect your GitHub repository
+1. Create a Web Service — connect your GitHub repo
    - **Build Command:** `npm install && npm run build`
    - **Start Command:** `npm start`
-   - **Environment:** Node
-   - **Plan:** Starter ($7/mo) or Free (with limitations)
-
-2. **Environment Variables**
-   - Add all env vars from the `.env.local` section above
-   - Set `NODE_ENV=production`
-
-3. **Render Config (`render.yaml`)**
+2. Add environment variables:
 
 ```yaml
 services:
   - type: web
-    name: wing-scout
+    name: wing-command
     runtime: node
     buildCommand: npm install && npm run build
     startCommand: npm start
     envVars:
       - key: NODE_ENV
         value: production
-      - key: NEXT_PUBLIC_SUPABASE_URL
-        sync: false
-      - key: NEXT_PUBLIC_SUPABASE_ANON_KEY
-        sync: false
-      - key: SUPABASE_SERVICE_ROLE_KEY
-        sync: false
       - key: TINYFISH_API_KEY
         sync: false
-      - key: UPSTASH_REDIS_REST_URL
-        sync: false
-      - key: UPSTASH_REDIS_REST_TOKEN
+      - key: GROQ_API_KEY
         sync: false
 ```
 
-4. **Next.js Standalone Output**
-   - `next.config.mjs` includes `output: 'standalone'` which is required for Render deployment
-
-### Cron Job (Optional)
-
-Set up a Render Cron Job to pre-populate the database:
-
-- **Schedule:** `0 */4 * * *` (every 4 hours)
-- **Command:** `python scraper/scrape_wings.py`
-- This pre-scrapes 150+ major US zip codes
+`next.config.mjs` already includes `output: 'standalone'` for Render compatibility.
 
 ## Project Structure
 
 ```
-wing-scout/
+wing-command/
 ├── app/
 │   ├── layout.tsx              # Root layout (Bebas Neue + Inter fonts)
 │   ├── page.tsx                # War Room main page
-│   ├── globals.css             # Midnight Turf theme styles
+│   ├── globals.css             # Midnight Turf theme
 │   ├── loading.tsx             # Loading state
 │   ├── error.tsx               # Error boundary
 │   └── api/
 │       ├── scout/route.ts      # GET /api/scout?zip=xxxxx&flavor=classicist
-│       └── menu/route.ts       # GET /api/menu?spot_id=xxxxx
+│       └── discover/route.ts   # URL discovery via Groq
 ├── components/
 │   ├── HeroVisuals.tsx         # Parallax + floating particles (Framer Motion)
 │   ├── FlavorSelector.tsx      # 3 flavor persona cards with pulse animation
 │   ├── ZipSearch.tsx           # Stadium-lit glowing zip input
 │   ├── WingGrid.tsx            # Scout Cards grid (glassmorphism)
 │   └── ui/                     # Reusable UI primitives
 ├── lib/
-│   ├── tinyfish-scraper.ts              # TinyFish scraper (parallel, flavor-aware)
-│   ├── types.ts                # TypeScript definitions
+│   ├── types.ts                # TypeScript definitions (pure in-memory, no DB)
 │   ├── utils.ts                # Flavor scoring, dedup, formatting
-│   ├── supabase.ts             # Database client
-│   ├── cache.ts                # Upstash Redis caching
-│   ├── geocode.ts              # Nominatim geocoding
-│   ├── env.ts                  # Environment validation
-│   └── menu.ts                 # Menu fetching
-├── supabase/
-│   └── schema.sql              # Full database schema
-├── scraper/
-│   └── scrape_wings.py         # Python cron pre-scraper
+│   ├── geocode.ts              # Nominatim geocoding (no API key needed)
+│   └── env.ts                  # Environment validation
 ├── tailwind.config.ts          # Midnight Turf theme
-├── next.config.mjs             # Standalone output for Render
-└── package.json                # framer-motion, lucide-react, etc.
+├── next.config.mjs             # Standalone output
+└── package.json
 ```
 
 ## Constraint Checklist
@@ -214,18 +184,17 @@ wing-scout/
 |-----------|--------|
 | Google Places API used? | NO (OSM/Nominatim) |
 | Yelp API used? | NO |
-| Vercel deployment? | NO (Render.com) |
+| External database or cache used? | NO (pure in-memory) |
 | UI "Mesmerizing"? | YES (Framer Motion, glassmorphism, neon glow) |
-| Menu scraping parallel? | YES (`Promise.allSettled` across 4 sources) |
-| Flavor persona filtering? | YES (keyword matching, 0-100 scoring) |
+| Scraping parallel? | YES (`Promise.allSettled` across 4–7 sources) |
+| Flavor persona filtering? | YES (keyword matching, 0–100 scoring) |
 
 ## Tech Stack
 
 - **Framework:** Next.js 14 (App Router), TypeScript, Tailwind CSS
 - **Animations:** Framer Motion
 - **Icons:** Lucide React
-- **Database:** Supabase (PostgreSQL + PostGIS)
-- **Cache:** Upstash Redis
-- **Scraper:** TinyFish
-- **Geocoding:** Nominatim (OpenStreetMap)
-- **Deployment:** Render.com (Web Service)
+- **AI / LLM:** Groq (`llama-3.3-70b-versatile`) for URL discovery
+- **Browser Agents:** TinyFish SDK (`client.agent.stream`)
+- **Geocoding:** Nominatim (OpenStreetMap) — no API key required
+- **Deployment:** Railway / Render.com