Skip to content

Latest commit

 

History

History
294 lines (234 loc) · 7.57 KB

AI_FEATURE.md

File metadata and controls

294 lines (234 loc) · 7.57 KB

AI Nutrition Summary Feature

Overview

This feature uses OpenAI's GPT-4.1-mini to generate natural language nutrition summaries from meal descriptions. Users can input text like "two scrambled eggs with toast and butter" and receive estimated calorie counts and macronutrient breakdowns.

Files Added

controllers/nutritionController.ts

  • Handles POST requests to generate AI nutrition summaries
  • Sends user's meal description to OpenAI API
  • Returns a concise paragraph (max 50 words) with estimated calories, carbs, protein, and fat
  • Includes reasonable assumptions when quantities are unclear

routes/nutritionRoutes.ts

  • Defines the route /api/ai/nutrition/summary
  • Requires authentication via JWT token
  • Maps POST requests to the nutrition controller

Implementation Steps Completed

1. Database Schema Update

Added aiSummary column to the FoodLog model in prisma/schema.prisma:

model FoodLog {
  // ... existing fields
  aiSummary   String?  @map("ai_summary")
  // ... rest of model
}

Apply schema changes to Supabase:

npx prisma db push
npx prisma generate

The db push command syncs your schema directly to the database without creating migration files. You should see:

🚀  Your database is now in sync with your Prisma schema.

2. Updated Interfaces

Modified CreateFoodLogInput in queries.ts:

interface CreateFoodLogInput {
  // ... existing fields
  aiSummary?: string  // Added
}

3. Installed Dependencies

npm install openai

4. Environment Variables

Added to .env:

OPENAI_API_KEY=sk-your-key-here

5. Mounted Route

Updated server.ts:

import nutritionRoutes from "./routes/nutritionRoutes.js";
app.use("/api/ai", nutritionRoutes);

6. Authentication

Modified nutritionRoutes.ts to require authentication:

router.use(authMiddleware);

7. Rebuild and Restart

npm run build
npm start

API Endpoint

Generate Nutrition Summary

Endpoint: POST /api/ai/nutrition/summary

Headers:

Authorization: Bearer <jwt_token>
Content-Type: application/json

Request Body:

{
  "text": "two scrambled eggs with toast and butter"
}

Response:

{
  "summary": "Your meal of two scrambled eggs with toast and butter is roughly 300-350 kcal, with about 20g protein, 20-25g carbs, and 15-20g fat. This estimate assumes one slice of toast with a pat of butter and standard-sized eggs. A balanced, satisfying start!"
}

Error Responses:

400 Bad Request:

{
  "error": "text is required"
}

502 Bad Gateway:

{
  "error": "no_content"
}

500 Server Error:

{
  "error": "server_error",
  "detail": "error message"
}

Testing

1. Start the Server

npm run build
npm start

2. Get Authentication Token

curl -X POST http://localhost:5001/auth/login \
  -H "Content-Type: application/json" \
  -d '{"email": "john@example.com", "password": "password123"}'

Save the token from the response.

3. Test AI Nutrition Summary

Using curl:

TOKEN="your_token_here"

curl -X POST http://localhost:5001/api/ai/nutrition/summary \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"text": "chicken breast with rice and broccoli"}'

Using Thunder Client (VS Code):

  1. Method: POST
  2. URL: http://localhost:5001/api/ai/nutrition/summary
  3. Headers:
    • Authorization: Bearer YOUR_TOKEN
    • Content-Type: application/json
  4. Body (JSON): 
    {
  "text": "grilled salmon with quinoa and asparagus"
}

4. Create Food Log with AI Summary

After getting the AI summary, include it when creating a food log:

Using curl:

curl -X POST http://localhost:5001/api/food-logs \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "loggedDate": "2025-10-02",
    "mealType": "lunch",
    "foodName": "Chicken with rice and broccoli",
    "calories": 450,
    "protein": 40,
    "carbs": 50,
    "fat": 10,
    "servingSize": 1,
    "servingUnit": "plate",
    "aiSummary": "Your meal of chicken breast with rice and broccoli is approximately 450 kcal with 40g protein, 50g carbs, and 10g fat. A well-balanced, nutritious meal!"
  }'

Using Thunder Client:

  1. Method: POST
  2. URL: http://localhost:5001/api/food-logs
  3. Headers:
    • Authorization: Bearer YOUR_TOKEN
    • Content-Type: application/json
  4. Body (JSON): 
    {
  "loggedDate": "2025-10-02",
  "mealType": "dinner",
  "foodName": "Grilled Chicken",
  "calories": 400,
  "protein": 50,
  "carbs": 10,
  "fat": 15,
  "servingSize": 1,
  "servingUnit": "plate",
  "aiSummary": "Your grilled chicken dinner has about 400 kcal with 50g protein, 10g carbs, and 15g fat. A lean, high-protein meal!"
}

5. Verify Storage

Check via API:

curl -H "Authorization: Bearer $TOKEN" \
  http://localhost:5001/api/food-logs/daily/2025-10-02

The response should include the aiSummary field.

Check via Tableviewer:

  1. Open http://localhost:5001 in browser
  2. Navigate to food_logs table
  3. Hard refresh the page (Ctrl+Shift+R)
  4. You should see the ai_summary column with your saved data

Usage Workflow

Recommended Implementation (Frontend)

  1. User enters meal description in a text field
  2. Frontend calls /api/ai/nutrition/summary to get AI estimation
  3. Frontend displays the summary to the user
  4. User can edit/confirm the values
  5. Frontend includes aiSummary in the POST request to /api/food-logs
  6. AI summary is saved with the food log entry
  7. When viewing past meals, the stored AI summary is displayed

Token Efficiency

To minimize OpenAI API costs:

  • Only generate summaries when explicitly requested by the user
  • Store the summary in the database on first generation
  • Return stored summaries on subsequent fetches (don't regenerate)
  • Consider adding a "Regenerate Summary" button for updates only

Important: The /api/ai/nutrition/summary endpoint only generates text - it does NOT save anything to the database. The frontend must explicitly include the aiSummary field when calling /api/food-logs to persist it.

Important Notes

  1. Authentication Required: All requests must include a valid JWT token
  2. OpenAI API Key: Ensure OPENAI_API_KEY is set in .env
  3. Model: Uses gpt-4.1-mini for cost efficiency
  4. Temperature: Set to 0.2 for consistent, less creative responses
  5. Accuracy: AI estimates are approximations and may not be precise
  6. Storage: The aiSummary field is optional in the database
  7. Two Separate Endpoints:
    • /api/ai/nutrition/summary - generates AI text (does NOT save)
    • /api/food-logs - saves food log (include aiSummary in body to persist)

Error Handling

The controller handles:

  • Missing or invalid text input (400)
  • OpenAI API failures (500)
  • Empty responses from OpenAI (502)

All errors are logged to the console for debugging.

Troubleshooting

Column not appearing in tableviewer

If the ai_summary column doesn't appear in the tableviewer after running npx prisma db push:

  1. Hard refresh the browser (Ctrl+Shift+R)
  2. Check Supabase dashboard → Table Editor → food_logs to verify column exists
  3. Restart the server after schema changes

Token expired errors

JWT tokens expire after 7 days, or may expire for other reasons. Get a new token by calling the login endpoint again.

OpenAI API errors

Check that OPENAI_API_KEY is correctly set in .env and that you have API credits in your OpenAI account.