- Subagents: when to split work and how to keep the parent lean
The subagent pattern in AI SDK 7: parent delegates via a tool, subagent runs with its own context, returns a summary via toModelOutput.
- RAG with streaming and cited sources in AI SDK 7
End-to-end RAG workflow: embed the query, retrieve chunks, streamText with context, and surface sources the UI can render.
- Multi-provider routing: customProvider, gateway, and fallbacks
Route across providers in AI SDK 7: the AI Gateway for one-key access, customProvider for named routing with fallbacks, and per-request selection.
- useChat client in App Router: DefaultChatTransport, sendMessage, manual input state
How the client side of useChat looks in AI SDK 7: transport object, sendMessage with a text message, and input state you manage yourself.
- Provider API keys: env wiring, gateway, and keeping keys off the client
Key hygiene for AI SDK 7 apps: providers read standard env vars, the gateway needs one key, custom providers take keys explicitly, and keys never ship to the browser.
- Next.js App Router chat route: streamText to UI message stream wiring
The exact server wiring for an App Router chat endpoint in AI SDK 7: streamText with convertToModelMessages, abortSignal, and createUIMessageStreamResponse.
- AI SDK on Expo and TanStack Start: quickstart-verified setup notes
AI SDK 7 has first-party quickstarts for Expo and TanStack Start: same ai core, framework-native server routes, standard env key config.