🧠 nutrition-ai (Rust)

Analyze food images using Google Gemini AI with full nutritional breakdown – written in pure Rust.

🔄 Node.js developer? Use the official nutrition-ai npm package.

🍽️ Overview

This crate replicates the functionality of nutrition-ai in Rust.

• 📸 Upload food images as base64 (image/png or image/jpeg)
• 🤖 Google Gemini AI generates:
  • Calorie + macro breakdown per food item
  • Total calorie estimation
  • Nutritional balance and dietary tips
• 🛡️ Input validation and error handling
• 🔁 Auto-retries on 503 model overloads

🚀 Quick Example

use base64::engine::general_purpose::STANDARD;
use base64::Engine;
use nutrition_ai::{GeminiRequest, MimeType, generate_answer};

#[tokio::main]
async fn main() -> anyhow::Result<()> {
    let image_bytes = std::fs::read("img.png")?;
    let file_base64 = STANDARD.encode(&image_bytes);

    let req = GeminiRequest {
        file_base64,
        file_mime_type: MimeType::ImagePng,
        google_key: "<YOUR_GOOGLE_KEY>".to_string(),
        model: Some("Gemini2_0Flash".to_string()),
    };

    let result = generate_answer(req).await?;
    println!("{}", result);
    Ok(())
}

📥 API Parameters

🛠️ Supported Models

• Gemini1_0Pro
• Gemini1_5Pro
• Gemini1_5Flash (default)
• Gemini1_5Flash8B
• Gemini2_0Flash

Default fallback model if none is provided: Gemini1_5Flash

🧪 Error Handling

• ❌ Empty or invalid base64 input
• ❌ Unsupported MIME type
• ❌ Missing Google API key
• 🔁 Retries up to 3 times on Gemini API 503 (model overload)
• ❌ Missing or empty Gemini response content

🧑‍💻 Author

@virugamacoder

📄 License

MIT © 2025