Neural & AI API Reference

Gemini in the cloud, Gemini Nano on-device, ML Kit vision and language, speech in and out

This document covers conversational reasoning, on-device generative tasks, speech recognition and synthesis, multimodal vision, and hardware-secured API key storage. Each entry documents its Inputs (arguments, with defaults), its Outputs (every returned field) and its Functions (what each callable takes and returns).

Reference

Hook	What it reads
useAppFunctions	Exposes on-device actions and hardware capabilities to system AI and Gemini Assistant.
useEmbeddings	On-device vector embeddings and semantic cosine similarity scoring via Tensor EdgeTPU.
useGemini	Cloud Gemini chat with real multi-turn history.
useGeminiNano	Gemini Nano running on the phone, with no network and no API key.
useGenAITasks	Four focused text tasks that run locally: summarise, proofread, rewrite, describe.
useNaturalLanguageAI	Translation, language detection, smart replies and entity extraction, all offline.
useSpeech	Speaking text aloud with the voices the phone has installed.
useSpeechAI	Turning speech into text, on the device or in the cloud.
useVisionAI	Nine on-device vision capabilities, plus cloud scene understanding.