Meal Image Nutrition Analysis Webhook with OpenAI

How it works

1. 1.The meal_image_webhoook node accepts HTTP POST at path fb1795ae-7290-48be-8c52-27b50eb32691 with response deferred to the response node.
2. 2.The analyze_image node sends the binary property image (base64) to chatgpt-4o-latest with instructions to identify foods, estimate portions, compute nutrition, and return a single JSON object.
3. 3.The extract_results chain LLM node uses gpt-4o-mini to convert the analysis text into an object matching a manual JSON schema (mealName, calories, macros, sodium, confidenceScore, healthScore).
4. 4.An Auto-fixing Output Parser wraps the Structured Output Parser to retry corrections when output does not match the schema, reusing gpt-4o-mini.
5. 5.The respond_to_webhook node returns JSON.stringify($json.output) as the JSON response body.

Use cases

• Backend service for a calorie-tracking app (e.g., Lovable) that uploads meal photos and receives structured nutrition estimates.
• On-demand food-photo analysis endpoint for any frontend that POSTs a base64 image named image.

Requirements

• Valid OpenAI credentials configured for both the analyze_image node and the OpenAI Chat Model node.
• Incoming requests must include the image as binary data under the property name image.
• Manual activation of the workflow, since it is saved with active: false.

Setup and testing in n8n

1. Activate the workflow, POST a food photo to the webhook URL with an image field, and confirm a JSON nutrition payload is returned.
2. POST a non-food image and verify the response reports confidenceScore 0 and mealName Unidentifiable.
3. Check execution logs to confirm the structured parser and auto-fixing step succeed on irregular model output.

Service Value

Ideal as a smart automation service combining integrations and AI to produce ready-to-use results.