0.45.0: Embedding models

Llama now runs embedding models, starting with EmbeddingGemma 2. They turn text into vectors for search and retrieval, through the server's /v1/embeddings endpoint. They carry an Embedding chip, and their model page links to a guide on how to use them.

• Sign in with Hugging Face to download gated models
• Say when a gated model needs access instead of blaming the network
• Add a Load row to the model page to warm a model up before use
• Let decision models take inputs up to 4k tokens on every Mac
• Update the engine to b11671, adding support for Nemotron 3.5 Super VL
• Fit Qwen 3.8 Flash Next on 128 GB Macs
• Show the Liquid AI logo for LFM models
• Use standard-size controls and a one-line token field in Settings
• Show focus on the shortcut recorder and cancel it on a click outside