ModelsLab

The generation/transformation tools (text-to-image, image-to-video, etc.) are clearly distinct. However, the four retrieval tools overlap heavily: fetch-generation explicitly handles images, video, and music, directly duplicating fetch-image, fetch-video, and fetch-audio, forcing the agent to guess which poll endpoint to use.

A partial source-to-target pattern (text-to-image, image-to-video, video-to-video) and a fetch-* pattern exist, but they are mixed with bare noun phrases like chat-completion, dubbing, music-generation, and list-models, so no single convention dominates.

23 tools is on the heavy side but justified for a multimodal platform spanning image, video, audio, music, and LLM generation, with each generation tool plus its fetch counterpart earning a place.

The surface covers discovery (list-models, list-providers), generation across all major modalities, editing/inpainting, and async retrieval. Minor gaps remain around LLM beyond chat-completion and lifecycle management, but core workflows are covered.