Click a thumbnail to caption → embed—no vector database.
caption → embed
Click to caption
Vision model writes a caption, then the caption is embedded.
Pick an image to run caption then embed.