{
  "title": "Find your images",
  "video_id": "QpNde4Jx6GQ",
  "published": "2026-06-09T16:02:21+0000",
  "complexity": "Easy",
  "old_url": "https://workflows-of-ai.com/workflow/find-your-images",
  "description": "Two Drupal recipes add an image search agent: AI describes and tags your media, then a database or vector search finds images by what they show.",
  "subtitle": "Install a recipe and let your Drupal CMS assistant find media by what is in the picture, not by file name or alt text.",
  "summary": "Two recipes add an image search agent to Drupal, one on database search and one on vector search, both built on an image classification recipe that describes and tags each media item. The database version only needs a default chat with tools model; the vector version also needs an embeddings provider and a vector database. The agent is tested in the Tool Explorer and then added to the Drupal CMS assistant as a sub-agent.",
  "categories": [
    "Drupal AI",
    "Agents"
  ],
  "intro_html": "<p>Learn how to build an AI-powered image search agent in Drupal that can find images based on what is actually shown in them.</p><p>In this video, I demonstrate two recipes:</p><p>Database Search - Uses AI-generated image descriptions and metadata stored in a traditional database for fast, simple searching.<br>Vector Search - Uses embeddings and semantic search to find images based on meaning and context, not just keywords.</p><p>Perfect for building agents that can answer questions like \"Find images with people on stage\" or \"Show photos of bicycles near water.\"</p><p>Recipes:<br>AI Agent Image Search (Database): <a href=\"https://www.drupal.org/project/ai_recipe_agent_image_search_db\">https://www.drupal.org/project/ai_recipe_agent_image_search_db</a><br>AI Agent Image Search (Vector): <a href=\"https://www.drupal.org/project/ai_recipe_agent_image_search_vector\">https://www.drupal.org/project/ai_recipe_agent_image_search_vector</a></p>",
  "intro_added": false,
  "chapters": [
    "0:00 Finding images by what they show",
    "1:18 Three levels of recipes",
    "3:00 Database versus vector search",
    "3:34 What each recipe needs",
    "5:29 Default chat with tools model",
    "6:55 Installing the recipe",
    "7:42 Uploading and classifying an image",
    "10:14 Trying the search tool",
    "11:39 Prompting the image search agent",
    "12:43 Connecting the Drupal CMS assistant"
  ],
  "corrections": [
    [
      "freely give",
      "FreelyGive"
    ],
    [
      "a llama",
      "Ollama"
    ],
    [
      "open AI",
      "OpenAI"
    ],
    [
      "and tropic",
      "and Anthropic"
    ],
    [
      "ABS or Azure",
      "AWS or Azure"
    ],
    [
      "PG vector",
      "pgvector"
    ],
    [
      "brush recipes",
      "Drush recipes"
    ]
  ],
  "ai_note": "The video and the introduction are Marcus's own, first published on workflows-of-ai.com. AI wrote the subtitle, the summary and the video chapters from the transcript, and corrected misheard names in YouTube's automatic captions."
}