GemmaEmbed: text, image and audio embeddings in Swift with Metal

I've open sourced GemmaEmbed, a Swift package that runs Google's EmbeddingGemma 2 locally on Apple Silicon. It produces text, image and audio embeddings in the same vector space, so you can compare a text query with photos or audio clips. You can also embed an ordered mix of text, images and audio as one document. The package implements its own Metal kernels and has no dependencies beyond the Apple SDK. Inference runs in your process using GGUF weights that you download separately. The current...