Diagnose slow or stuck createStreamableUI updates

TL;DR: Find the createStreamableUI call. Symptom: a generative UI built with createStreamableUI updates sluggishly or the final state never arrives. The stream seems stuck.

Fix

1. Find the createStreamableUI call. Expected: you have the value in hand.

2. Trace every code path (including early returns and catch blocks) and check whether .done() is reached on all of them. Expected: this specific failure stops.

3. Exceptions thrown between update() and done() are the classic. Expected: this specific failure stops.

4. Fix: import { createStreamableUI } from '@ai-sdk/rsc'; const stream = createStreamableUI('initial'); try { stream.update('working...'); // ... do work ... stream.done('final'); } catch (e) { stream.done('failed'); } return stream.value; Rules: 1. Expected: this specific failure stops.

5. done() takes the final value. .update() and .append() are for intermediate states. Expected: this specific failure stops.

6. If updates are slow but do finish, check that you are not awaiting the full model response before the first update. Expected: this specific failure stops.

Details

Symptom: a generative UI built with createStreamableUI updates sluggishly or the final state never arrives. The stream seems stuck. Cause: the streamable was never closed. Confirmation: 1. 2. If any path returns stream.value without calling stream.done() first, that is the bug. Always close with .done(), even on error paths. try/finally is the safest shape. 3. This applies to the @ai-sdk/rsc streamable UI path. The newer tool-part generative UI pattern (tools returning data rendered by React components) does not have this failure mode; prefer it for new code. 4. Stream early, stream often. Verification: exercise the error path deliberately and confirm the UI reaches its final state instead of hanging.

When to use

You hit exactly this: Diagnose slow or stuck createStreamableUI updates in Diagnose.

When not to use

A different error, or the same symptom in a different tool. This page only covers the failure above.

Compatibility

Diagnose.

Root cause

The client keeps waiting for more parts because the stream is still open server-side.

From the AI SDK troubleshooting docs (Unclosed Streams entry), verified 2026-09-26: the docs show createStreamableUI from @ai-sdk/rsc and the fix as calling .done() to close the stream.