Latimal - Food AI for Delivery Platforms

The /report endpoint analyzes an entire restaurant menu and returns a comprehensive quality assessment. Use it to catch data issues before they reach customers.

What the report covers

Noise detection - promo text, prices, serving sizes, formatting junk embedded in item names
Duplicates - items that are the same dish under different spellings or transliterations
Non-food items - pharmacy, grocery, or merchandise items that shouldn't be in a food menu
Cuisine classification - breakdown of cuisine types across the menu
Quality score - single 0-100 number summarizing overall menu health

Basic usage

import requests

API_KEY = "YOUR_KEY"
BASE = "https://embed.statode.com"
headers = {"X-API-Key": API_KEY, "Content-Type": "application/json"}

resp = requests.post(f"{BASE}/report", headers=headers,
    json={"items": full_menu, "restaurant_name": "My Restaurant"})
report = resp.json()

print(f"Quality score: {report['quality_score']}/100")
print(f"Noisy items: {report['noise_analysis']['items_with_noise']}")
print(f"Duplicates: {report['dedup_summary']['total_duplicate_items']}")
print(f"Non-food items: {len(report['non_food_items'])}")

Setting up scheduled monitoring

Run the report daily or weekly to catch issues from menu imports:

import json
from datetime import datetime

def run_health_check(menu_items, restaurant_name):
    resp = requests.post(f"{BASE}/report", headers=headers,
        json={"items": menu_items, "restaurant_name": restaurant_name})
    report = resp.json()

    # Alert if quality drops
    if report["quality_score"] < 70:
        send_alert(f"{restaurant_name} quality score dropped to {report['quality_score']}")

    # Log for trend tracking
    log_entry = {
        "timestamp": datetime.now().isoformat(),
        "restaurant": restaurant_name,
        "score": report["quality_score"],
        "duplicates": report["dedup_summary"]["total_duplicate_items"],
        "noisy": report["noise_analysis"]["items_with_noise"],
        "non_food": len(report["non_food_items"])
    }
    with open("menu_health_log.jsonl", "a") as f:
        f.write(json.dumps(log_entry) + "\n")

    return report

Quality score interpretation

Score	Rating	Action
90-100	Excellent	No action needed
70-89	Good	Review flagged items when convenient
50-69	Needs attention	Clean up duplicates and noise this week
Below 50	Poor	Immediate cleanup required

Common issues

Promo text noise - Items like "NEW 50% OFF Chicken Biryani (Serves 2)" should just be "Chicken Biryani". The report flags these so you can clean your source data.

Duplicates from multi-source import - When the same restaurant feeds data from Zomato, Swiggy, and their POS, each source uses slightly different names. The report catches these overlaps.

Non-food items from POS categories - POS systems often include "Tissue Paper", "Packaging Charge", or "Delivery Fee" as menu items. The report identifies these for removal.

Missing cuisine signal - If the cuisine breakdown shows "unknown" for many items, the item names may be too noisy or abbreviated for classification.

Limits

Up to 2,000 items per report request
For larger menus, split by restaurant or outlet and run separate reports

Menu Health Monitoring