Feed Me Recipes

Create Anylist recipes from Instagram recipe posts.

How It Works

iOS Shortcut
     ↓
POST /ingest { "url": "https://instagram.com/p/..." }
     ↓
┌─────────────────────────────────────────┐
│  Recipe Ingest Service (Docker)         │
│                                         │
│  1. Validate bearer token               │
│  2. Fetch post via Apify                │
│  3. Parse recipe with Claude Haiku      │
│  4. Create recipe in AnyList            │
│  5. Send push notification via ntfy.sh  │
└─────────────────────────────────────────┘

Quick Start

# Install dependencies
pnpm install

# Copy and configure environment
cp .env.example .env
# Edit .env with your credentials

# Run locally
pnpm dev

# Test
curl -X POST http://localhost:3000/ingest \
  -H "Authorization: Bearer $API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://www.instagram.com/p/ABC123/"}'

API

POST /ingest

Accepts an Instagram URL and processes it into a recipe.

Request:

POST /ingest HTTP/1.1
Authorization: Bearer <API_TOKEN>
Content-Type: application/json

{
  "url": "https://www.instagram.com/p/ABC123/"
}

Response:

HTTP/1.1 202 Accepted

{
  "status": "processing",
  "message": "Recipe ingestion started"
}

GET /health

Health check (no auth required).

HTTP/1.1 200 OK
{"status": "ok"}

Environment Variables

Variable	Description
API_TOKEN	Bearer token for authenticating requests
APIFY_TOKEN	Apify API token for Instagram scraping
ANYLIST_EMAIL	AnyList account email
ANYLIST_PASSWORD	AnyList account password
ANTHROPIC_API_KEY	Claude API key for recipe parsing
NTFY_TOPIC	ntfy.sh topic for notifications
PORT	Server port (default: 3000)

Apify Token Permissions

Create a scoped token at Apify Console → Settings → Integrations:

• Actor: nH2AHrwxeTRJoN5hX (apify/instagram-post-scraper)
  • Run permission

No dataset permissions required (uses sync API).

Project Structure

├── src/
│   ├── index.ts      # Express server + routes
│   ├── apify.ts      # Instagram fetching (Apify sync API)
│   ├── parser.ts     # Claude Haiku recipe parsing
│   ├── anylist.ts    # AnyList recipe creation
│   └── notify.ts     # ntfy.sh notifications
├── scripts/
│   └── postinstall.js  # Patches anylist-napi ESM bug
├── Dockerfile
├── package.json
└── tsconfig.json

Deployment

Docker

docker build -t feed-me-recipe .

docker run -d \
  --name feed-me-recipe \
  --env-file .env \
  -p 3000:3000 \
  feed-me-recipe

Recommended: Cloudflare Tunnel

Expose the service securely without opening ports:

1. Create tunnel in Cloudflare Zero Trust dashboard
2. Add public hostname pointing to http://localhost:3000
3. Optionally configure Zero Trust access policies (or rely on bearer token auth)

Recipe Parsing

Claude Haiku extracts structured recipe data from Instagram captions:

{
  "is_recipe": true,
  "confidence": 0.95,
  "recipe": {
    "name": "Sheet Pan Beef Kefta Wraps",
    "servings": "4 servings",
    "prepTime": 20,
    "cookTime": 15,
    "ingredients": [
      {"name": "ground beef", "quantity": "1.5 lbs"},
      {"name": "garlic", "quantity": "8 cloves", "note": "minced"}
    ],
    "steps": [
      "Preheat oven to 400°F...",
      "Mix beef with spices..."
    ]
  }
}

Posts without recipes return {"is_recipe": false, "reason": "..."}.

Known Issues

• anylist-napi prepTime/cookTime values save as 0 (upstream bug)

Future Improvements

• Confidence threshold → flag low-confidence for manual review
• Job queue for reliability
• Recipe deduplication
• Web UI for review/edit
• Multi-user support