Roadmap

The runtime ships as a provider layer first; agents, tools, RAG, and memory come after that foundation is reliable.

MVP scope

Built first:

@stewmore/expo-ai-core
@stewmore/expo-ai-apple-foundation-models
@stewmore/expo-ai-android-aicore
@stewmore/expo-ai-cloud

MVP features:

• iOS system model provider
• Android system model provider
• Cloud fallback adapter
• Capability detection
• Normalized errors
• Privacy metadata
• Basic structured output
• Basic streaming where native support is available

Deferred: LiteRT-LM, model catalog, model downloads, local RAG, native tools, agent loop, long-term memory, and Genkit/LangChain bridges.

Phases

Phase	Delivers
1. System Provider MVP	Core API, iOS + Android adapters, capability detection, generate, sessions, error normalization, privacy metadata
2. Cloud Fallback + Structured Output	Cloud adapter, provider routing, generateObject, JSON validation + repair, provider metadata on every result
3. Streaming	Native event streaming, AsyncIterable wrapper, cancellation, timeout handling
4. Task APIs	summarize, rewrite, proofread, image input where available, provider-specific feature mapping
5. LiteRT-LM BYOM	LiteRT-LM provider, model file manager, remote download, checksum verification, compatibility metadata
6. Local Context / RAG	Local search abstraction, app-managed index, generateWithContext
7. Tools + Agent Loop	Tool registry, native iOS tools where available, JS-emulated Android tools, tool-call validation, loop limits, permissions
8. Backend Framework Bridges	Optional Genkit, LangChain, and Vercel AI SDK backend bridges

What v1 proves

• iOS can call Apple Foundation Models through Expo.
• Android can call Gemini Nano through AICore / ML Kit through Expo.
• The same TypeScript API works across both.
• Apps can detect capabilities before showing features.
• Results include provider and privacy metadata.
• Cloud fallback is explicit and controllable.
• Structured output works consistently enough for app features.