How to add AI food recognition to your app in 10 minutes

A while back I was building a personal calorie tracker and hit a wall: I needed an API that could look at a photo of food and return structured nutritional data. After months of experimenting with existing solutions — too expensive, too inaccurate, or barely documented — I ended up building my own.

This is that API: CaloAI. In this tutorial I'll show you how to integrate it in minutes.

👉 CaloAI on RapidAPI

What CaloAI does

Send a photo of any meal. Get back a clean JSON with:

Calories, proteins, fats, carbs
A health score (1–10)
(PRO) Detected dish components, allergen alerts, diet tags (keto, vegan, low-carb), AI nutritionist verdict, and a confidence score

Everything is powered by Google Gemini 3.5 Flash under the hood, with a built-in caloric consistency check — meaning macros always add up mathematically (Calories = Protein×4 + Carbs×4 + Fat×9), no hallucinated numbers.

Setup

Sign up on RapidAPI
Subscribe to CaloAI (free tier available)
Copy your X-RapidAPI-Key

That's it. One endpoint: POST https://caloai.p.rapidapi.com/v1

Request structure

Field	Type	Required	Description
`image_url`	string	✅	Publicly accessible URL of the food photo
`extra`	string	❌	Optional context — weight, meal type, hidden ingredients

The extra field is a secret weapon. If your user knows something the camera can't see ("exactly 300g, added olive oil"), pass it here and the model treats it as ground truth — dramatically improving accuracy.

Python example

import requests

url = "https://caloai.p.rapidapi.com/v1"

headers = {
    "x-rapidapi-key": "YOUR_API_KEY",
    "x-rapidapi-host": "caloai.p.rapidapi.com",
    "Content-Type": "application/json"
}

payload = {
    "image_url": "https://example.com/my-lunch.jpg",
    "extra": "Weight is around 300g, added olive oil"  # optional
}

response = requests.post(url, json=payload, headers=headers)
data = response.json()

r = data["response"]
print(f"Calories: {r['calories']} kcal")
print(f"Protein: {r['proteins']}g | Carbs: {r['carbs']}g | Fat: {r['fats']}g")
print(f"Health score: {r['healthy_points']}/10")

Response (Basic plan):

{"status":"success","response":{"proteins":38,"fats":52,"carbs":20,"calories":652,"healthy_points":3}}

JavaScript example

const response = await fetch("https://caloai.p.rapidapi.com/v1", {
  method: "POST",
  headers: {
    "x-rapidapi-key": "YOUR_API_KEY",
    "x-rapidapi-host": "caloai.p.rapidapi.com",
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    image_url: "https://example.com/breakfast.jpg"
  })
});

const { response: data } = await response.json();

console.log(`${data.calories} kcal — Health: ${data.healthy_points}/10`);

Unlocking PRO features

On PRO, ULTRA, or MEGA plans the response expands with deeper analytical fields — ready to plug straight into your UI without extra processing:

{"status":"success","response":{"proteins":38,"fats":52,"carbs":20,"calories":652,"healthy_points":3,"confidence_score":95,"detected_dishes":["fried eggs","crispy bacon","cheddar cheese"],"diet_tags":["keto-friendly","low-carb"],"potential_allergens":["eggs","dairy"],"ai_verdict":"A solid high-protein option, but quite high in saturated fats. Consider pairing with fresh vegetables next time."}}

Field	What it's useful for
`confidence_score`	Show a data quality indicator in your UI
`detected_dishes`	Break down mixed plates into components
`diet_tags`	Filter meals by dietary preference
`potential_allergens`	Safety warnings for your users
`ai_verdict`	Ready-to-display copy, no prompt engineering needed

Real example: daily meal logger

Here's a minimal script that analyzes multiple meals and sums up daily totals:

import requests

def analyze_meal(image_url, extra=None):
    headers = {
        "x-rapidapi-key": "YOUR_API_KEY",
        "x-rapidapi-host": "caloai.p.rapidapi.com",
        "Content-Type": "application/json"
    }
    payload = {"image_url": image_url}
    if extra:
        payload["extra"] = extra

    r = requests.post("https://caloai.p.rapidapi.com/v1", json=payload, headers=headers)
    return r.json()["response"]

meals = [
    ("https://example.com/breakfast.jpg", "oat porridge with banana"),
    ("https://example.com/lunch.jpg", None),
    ("https://example.com/dinner.jpg", "about 400g"),
]

totals = {"calories": 0, "proteins": 0, "carbs": 0, "fats": 0}

for image_url, context in meals:
    result = analyze_meal(image_url, context)
    print(f"✓ {result['calories']} kcal | Health: {result['healthy_points']}/10")
    for key in totals:
        totals[key] += result[key]

print(f"\n📊 Daily totals:")
print(f"  Calories: {totals['calories']} kcal")
print(f"  Protein: {totals['proteins']}g | Carbs: {totals['carbs']}g | Fat: {totals['fats']}g")

Error handling

The API returns clear error messages:

response = requests.post(url, json=payload, headers=headers)
data = response.json()

if data["status"] == "error":
    print(f"Error: {data['error']}")
else:
    result = data["response"]
    # process result...

Common errors:

403 — missing image_url field
403 — malformed or inaccessible image URL (make sure the URL is publicly accessible)

Tips

Better photos = better results. Single dish, good lighting, top-down angle works best.
Use extra whenever you can. Even rough context like "about 300g" or "added butter" improves accuracy.
healthy_points ranges from 1 (deep-fried, highly processed) to 10 (whole foods, lean protein, nutrient-dense). Great for gamification in fitness apps.
ai_verdict is UI-ready copy — no need to write prompts yourself, just render it directly.

Wrap up

That's one endpoint, a couple of headers, and you've got AI food recognition in your app.

If you're building something with CaloAI — a calorie tracker, fitness dashboard, smart food diary — I'd love to hear about it in the comments.

👉 Try CaloAI on RapidAPI →