> ## Documentation Index
> Fetch the complete documentation index at: https://docs.openchatwidget.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Reasoning

> How reasoning summaries appear in Open Chat Widget and how to turn them off

Reasoning is the widget's UI for AI SDK `reasoning` parts. When your backend streams reasoning summaries alongside the assistant response, Open Chat Widget shows a compact `Thinking...` row above the final answer. While the model is still working, that row animates subtly. After the stream finishes, it switches to a completed state like `Thought for 2s...`.

Clicking the reasoning row expands the streamed reasoning text. The expanded content is intentionally quiet and secondary so it does not overpower the main assistant response.

## What It Looks Like

* A small `Thinking...` disclosure appears above the assistant message while reasoning is streaming.
* A subtle caret shows whether the reasoning block is open or closed.
* When reasoning is done, the label changes to `Thought for Xs...`.
* The reasoning text stays collapsible and separate from the main response body.

## Backend Requirements

Your backend must stream AI SDK `reasoning` parts for the widget to show reasoning.

```ts theme={null}
const result = streamText({
  model: openai("gpt-5-mini"),
  messages: await convertToModelMessages(messages),
  providerOptions: {
    openai: {
      reasoningEffort: "medium",
      reasoningSummary: "detailed",
    },
  },
});

result.pipeUIMessageStreamToResponse(response, {
  sendReasoning: true,
});
```

If your backend only streams text, the widget renders a normal chat conversation without a reasoning block.

## Disable Reasoning

If you do not want reasoning to appear in the widget, pass `disableReasoning` on the root component:

```tsx theme={null}
<OpenChatWidget
  url="http://localhost:8787/api/chat"
  disableReasoning
/>
```

This only disables reasoning in the UI. Your backend contract does not need to change.