Sketch adds On‑Device LLM assistant for faster symbol generation and naming

Design · 3 min read

The built-in assistant runs locally on Apple silicon machines, offering suggestions for symbol names, variant naming conventions and responsive variant generation that respects an 8pt grid system. Because it runs on-device, teams worried about IP leakage can use the assistant in offline and air-gapped environments.

Sketch also added a 'batch explain' feature: run over a file to generate component docs and suggested token mappings, which speeds handoffs to engineering or design systems teams. The assistant integrates with Sketch Libraries and can propose migration paths when token scales or naming schemes change.

Power users noted the assistant handles common patterns well but still needs hand-curation for complex systems and cross-platform naming conventions. Sketch said it's monitoring usage to expand supported languages and platform conventions.