Now that Gemini Flash can output images in amongst text, need to figure out how to model that

See https://simonwillison.net/2025/May/7/gemini-images-preview/ - it's now possible to get a response back from gemini-2.0-flash-preview-image-generation which includes inline images like this:

{
  "candidates": [
    {
      "content": {
        "parts": [
          {
            "text": "## Vegetarian \"Butter Chicken\" with Chickpeas\n\nThis recipe delivers the rich, creamy flavors of classic butter chicken, substituting tender chickpeas for chicken. It's a satisfying and flavorful vegetarian main course perfect for any night of the week.\n\n**Yields:** 4-6 servings\n**Prep time:** 20 minutes\n**Cook time:** 35 minutes\n\n**Ingredients:**\n\n*   **For the Chickpeas:**\n  *   2 (15-ounce) cans chickpeas, drained and rinsed\n    "
          },
          {
            "inlineData": {
              "mimeType": "image/png",
              "data": "iVBORw0KGgoAAAANSUhEUgAAAtgAAAQAC..."
            }
          },
          {
            "text": "\n  *   1 tablespoon olive oil\n    "
          },
          {
            "inlineData": {
              "mimeType": "image/png",
              "data": "iVBORw0KGgoAAAANSUhE..."
            }
          },
          {
            "text": "\n  *   1 teaspoon garam masala\n    "
          },
          {
            "inlineData": {
              "mimeType": "image/png",
              "data": "iVBORw0KGgoAAAANS..."
            }
          }
        ]
      }
    }
  ]
}

Full example here: https://gist.github.com/simonw/55894032b2c60b35f320b6a166ded493

I think we need to store these in the DB, even though they can be really big (that single response was 41MB!)

No repository file or test is named; start with the linked Gemini response example and inspect the existing response-storage entry point. Decide how mixed text and inline images should be represented and persisted, including the 41MB response case; done means the model and storage behavior are defined and covered by tests.