ai-core/chat-experience

Stream AI chat responses by passing the chat() function into toServerSentEventsResponse on a TanStack Start route, then connect the useChat client hook with fetchServerSentEvents to handle real-time UIMessage rendering.

TanStack AI handles multimodal content and thinking parts by rendering UIMessage components that explicitly support text, image, tool-call, and reasoning outputs, allowing you to surface model logic and multimedia directly inside the useChat interface.

Yes, TanStack AI chat hooks support building streaming chat experiences across React, Vue, Solid, Svelte, and Preact, allowing you to wire framework-specific fetchers for SSE or HTTP stream conversations.

The best way to configure model adapters is by applying adapter modelOptions within your chat() setup, which ensures the streaming client correctly maps generation parameters and message structures to the useChat hook.

Chat streams can break if middleware disrupts the streaming lifecycle, but TanStack AI provides middleware-driven lifecycle tracking to properly intercept and execute analytics without terminating the server-sent events connection.

You render tool-call and image parts by mapping the multimodal UIMessage parts array received from the fetchServerSentEvents stream, extracting the specific tool-call and image structures for UI display within your client component.