🔒On-Device InferenceRuns entirely on Apple Silicon. Your data never leaves the machine — no network requests, no API keys.
⚡Streaming GenerationAsync iterator interface for token-by-token streaming. Process responses as they're generated.
🧩Structured OutputTyped schemas with generation guides constrain output to exactly the shape you need.
🛠️Tool CallingGive the model tools to call during generation. Define schemas, implement handlers, get structured results.