text-to-image-search

Natural language image search lets you find images by describing what you want in plain text. This skill converts your description into semantic vectors and matches them against indexed images using CLIP or VLM Captions, returning visually relevant results without manual tagging.

CLIP Direct provides fast semantic matching for straightforward visual descriptions. VLM Captions handles complex, domain-specific queries by generating detailed image captions first, then matching against your request for higher accuracy on nuanced searches.

Yes. This skill uses Milvus as a scalable vector database to index images and perform semantic search across large collections. Milvus handles the indexing and retrieval infrastructure, enabling efficient search even as your image library grows.

You need an image collection to index, access to OpenAI GPT-4o for caption generation, a Milvus instance for vector storage, and Python libraries for embeddings and semantic matching. The skill handles the indexing and search once these dependencies are configured.

Yes. The dual-mode design supports both fast general searches via CLIP and complex domain-specific queries via VLM Captions, which generates contextual descriptions tailored to specialized vocabularies like medical imaging or product photography.