Consider it done. The open-source AI agent that works out of the box · 想到,就能做到。开源、开箱即用的 AI Agent。
TypeScript agentai-agentai-assistantandroidclaude-codecodexelectroniosllmmacosreact-nativetypescriptwindows
/**
* ThinkingText
* ---------------------------------------------------------------------------
* Lightweight inline presentation for model reasoning. Codex commonly wraps
* short reasoning updates in Markdown strong markers; rendering those markers
* literally is noisy, while the full chat Markdown pipeline is too heavy and
* allows block layouts that do not belong inside compact activity rows.
*
* Supported syntax is intentionally narrow and deterministic:
* - paired `**text**` → medium-weight text (500, per docs/design-rules/cindy-design-system.md);
* - paired backtick runs → inline monospace text;
* - malformed / unmatched delimiters remain literal text.
*
* Lists, headings, links, images, and HTML are never interpreted here.
*/
import { Fragment, memo, useMemo } from 'react';
import { tokenizeThinkingText } from '@cindy/maker-shared/thinking-text';
export { tokenizeThinkingText } from '@cindy/maker-shared/thinking-text';
export type { ThinkingTextToken } from '@cindy/maker-shared/thinking-text';
/** Shared compact renderer for live, expanded, and standalone thinking views. */
export const ThinkingText = memo(function ThinkingText({ content }: { content: string }) {
const tokens = useMemo(() => tokenizeThinkingText(content), [content]);
return (
<>
{tokens.map((token, index) => {
if (token.kind === 'strong') {
return (
<span key={`strong-${index}`} data-thinking-markup="strong" className="font-medium">
{token.value}
</span>
);
}
if (token.kind === 'code') {
return (
<code
key={`code-${index}`}
data-thinking-markup="code"
className="font-mono not-italic"
>
{token.value}
</code>
);
}
return <Fragment key={`text-${index}`}>{token.value}</Fragment>;
})}
</>
);
});