Find visually similar images based on a reference image. Uses AI-powered visual similarity (CLIP embeddings) to find images with similar content, style, or composition. The Razuna asset grid displays the returned thumbnails. Use thumbnail_url whenever provided, including for documents and URL bookmarks; file type does not determine thumbnail availability. The grid includes expandable descriptions and keywords. In ChatGPT, respond in plain text only and refer to the grid. Never add image groups, carousels, web previews, or a second gallery, even when labeled illustrative. If the client cannot display the grid, use only the exact returned thumbnail URLs and Razuna links. Do not search the web or generate illustrative replacement images. A returned URL is not proof that the client loaded the image; report a preview failure only if observed.