How to import structured recipe data via new REST API without parser?

[moejo-jo]

Platform

Web

Describe the bug

When passing a fully structured recipe via /api/v1/recipes/import/paste (either as raw JSON-LD or as HTML with embedded <script type="application/ld+json">), the recipe-scrapers parser ignores or overwrites several fields: image and url are stored as null, tags are auto-generated instead of using the provided values, metric units are converted to English (gram, tablespoon), and in some cases the recipe content is rewritten/translated via an AI fallback.

To reproduce

1. Call POST /api/v1/recipes/import/paste with a JSON-LD recipe object containing image, url, recipeCategory and metric units
2. Retrieve the created recipe via GET /api/v1/recipes/{id}
3. Observe that image is null, url is null, tags are auto-generated, and units are in English

Expected behavior

Fields provided in the structured input should be stored as-is, without parser transformation. Previously this was possible via POST /api/trpc/recipes.create which accepted a fully structured JSON object and stored it directly. I would like to achieve the same result via the new /api/v1 REST API.
Is there a supported way to pass structured recipe data to the new API without going through the parser? If not, would a direct create endpoint (e.g. POST /api/v1/recipes) that bypasses the parser be considered?

Screenshots

No response

Environment

Norish version: v0.18.0-beta
Self-hosted via Docker

Additional context

I'm running an automated pipeline that imports recipes from Instagram Reels. An LLM parses the recipe from the video caption and produces structured data (name, description, image URL, video URL, ingredients with metric units, steps in German, custom tags). The data is correct before sending – the issue is that the parser overwrites or ignores it on import.

[kaneo-norish-org]

NOR-98

[mikevanes]

could you give me example data you use? I will expose a dedicated create endpoint, with the 18.1 release.
@moejo-jo

[moejo-jo]

@mikevanes
Here's an example of the structured data I generate per recipe. This comes from an LLM that parses Instagram Reel captions. I'd love a dedicated endpoint that accepts this structure directly and stores it as-is, without going through the parser.

A few notes on my current data format:

• fat, carbs and protein are passed as strings (this was required by the tRPC endpoint). I'm happy to switch to numbers if the new endpoint prefers that.
• ingredientId is always passed as an empty string "" since I don't have pre-existing ingredient IDs. It would be great if the new endpoint could make this field optional or simply ignore it.
• Even though I pass systemUsed: "metric" at both the recipe and ingredient level, Norish occasionally converts the recipe to "us" on its own. Should I include this as a separate bug report, or can I add it here?

Could you also share the full request body schema for the new endpoint? I'd love to know if there are any additional fields or properties I'm not currently using (e.g. categories, notes, or anything else the endpoint will support).

{
  "name": "Creamy Pumpkin Pasta with Vegan Mince",
  "description": "A quick and creamy pumpkin pasta with silken tofu, nutritional yeast and vegan mince. Perfect for meal prep and packed with protein.\n\n| **[View Original Recipe](https://www.instagram.com/reel/EXAMPLE/)** | [instagram:EXAMPLE_ID]",
  "url": "https://example.com/media/2026-04-16_Creamy-Pumpkin-Pasta/recipe.mp4",
  "image": "https://example.com/media/2026-04-16_Creamy-Pumpkin-Pasta/recipe.jpg",
  "servings": 1,
  "prepMinutes": 10,
  "cookMinutes": 20,
  "totalMinutes": 30,
  "notes": null,
  "systemUsed": "metric",
  "calories": 2125,
  "fat": "39",
  "carbs": "228",
  "protein": "190",
  "recipeIngredients": [
    {
      "amount": 0.5,
      "unit": "pc",
      "order": 0,
      "ingredientName": "Hokkaido Pumpkin",
      "ingredientId": "",
      "systemUsed": "metric"
    },
    {
      "amount": 200,
      "unit": "g",
      "order": 1,
      "ingredientName": "Silken Tofu",
      "ingredientId": "",
      "systemUsed": "metric"
    },
    {
      "amount": 40,
      "unit": "g",
      "order": 2,
      "ingredientName": "Nutritional Yeast",
      "ingredientId": "",
      "systemUsed": "metric"
    },
    {
      "amount": null,
      "unit": "a little",
      "order": 3,
      "ingredientName": "Oil",
      "ingredientId": "",
      "systemUsed": "metric"
    },
    {
      "amount": 300,
      "unit": "g",
      "order": 4,
      "ingredientName": "Protein Spaghetti",
      "ingredientId": "",
      "systemUsed": "metric"
    },
    {
      "amount": null,
      "unit": "to taste",
      "order": 5,
      "ingredientName": "Salt",
      "ingredientId": "",
      "systemUsed": "metric"
    }
  ],
  "steps": [
    {
      "step": "Dice the Hokkaido pumpkin and air fry at 180°C.",
      "order": 0,
      "systemUsed": "metric"
    },
    {
      "step": "Cook the pasta according to package instructions.",
      "order": 1,
      "systemUsed": "metric"
    },
    {
      "step": "Blend all remaining ingredients, add vegan mince and pasta water, mix in the pasta and enjoy.",
      "order": 2,
      "systemUsed": "metric"
    }
  ],
  "tags": [
    { "name": "new" },
    { "name": "vegan" },
    { "name": "insta" }
  ]
}

[mikevanes]

We have a dedicated create endpoint in the latest version @moejo-jo

[moejo-jo]

I had time to update to v0.18.1 and used the new API Point for importing via JSON format. So far it looks very good! Thank you for adding the endpoint @mikevanes.