feat: add navigable JSON summaries for tool output

Tool JSON output now starts with a compact navigable summary view.
Expandable tool output includes quick open-file and diff actions for changed files.
Reasoning headers strip stray HTML comments, and navigation tools stay compact.

packages/ui/src/components/chat/message/parts/ReasoningPart.tsx

@@ -40,6 +40,8 @@ const EXPANDED_CONTENT_TRANSITION = { duration: 0.2, ease: 'easeOut' as const };
 /** Strip common markdown syntax so the header preview reads as plain text. */
 const stripMarkdown = (text: string): string =>
     text
+        // Empty HTML comments are frequently appended by model tool wrappers.
+        .replace(/<!--\s*-->/g, '')
         // Fenced code blocks → keep inner text on one line
         .replace(/```[\w]*\n?([\s\S]*?)```/g, (_, inner: string) => inner.trim())
         // Inline code
@@ -476,84 +478,4 @@ const ReasoningPart = React.memo(({
     );
 });
 
-type MergedReasoningPartProps = {
-    parts: Part[];
-    onContentChange?: (reason?: ContentChangeReason) => void;
-    messageId: string;
-    streamPhase?: StreamPhase;
-};
-
-/**
- * Renders ALL reasoning parts for a message as a single collapsible block,
- * merging their text and spanning their combined time range.
- * This matches the VSCode Copilot pattern of showing one "Thought" block per turn.
- */
-export const MergedReasoningPart = React.memo(({
-    parts,
-    onContentChange,
-    messageId,
-    streamPhase,
-}: MergedReasoningPartProps) => {
-    const chatRenderMode = useUIStore((state) => state.chatRenderMode);
-
-    const mergedText = React.useMemo(() => {
-        return parts
-            .map((part) => {
-                const p = part as PartWithText;
-                return cleanReasoningText(p.text || p.content || '');
-            })
-            .filter((t) => t.length > 0)
-            .join('\n\n');
-    }, [parts]);
-
-    const mergedTime = React.useMemo(() => {
-        let earliestStart: number | undefined;
-        let latestEnd: number | undefined;
-
-        for (const part of parts) {
-            const time = (part as PartWithText).time;
-            if (typeof time?.start === 'number' && Number.isFinite(time.start)) {
-                if (earliestStart === undefined || time.start < earliestStart) {
-                    earliestStart = time.start;
-                }
-            }
-            if (typeof time?.end === 'number' && Number.isFinite(time.end)) {
-                if (latestEnd === undefined || time.end > latestEnd) {
-                    latestEnd = time.end;
-                }
-            }
-        }
-
-        return earliestStart !== undefined ? { start: earliestStart, end: latestEnd } : undefined;
-    }, [parts]);
-
-    const canBeStreaming = streamPhase === undefined || streamPhase !== 'completed';
-    const isStreaming = chatRenderMode === 'live' && canBeStreaming && parts.some(
-        (part) => typeof (part as PartWithText).time?.end !== 'number',
-    );
-
-    const throttledMergedText = useStreamingTextThrottle({
-        text: mergedText,
-        isStreaming,
-        identityKey: `${messageId}:reasoning-merged`,
-    });
-
-    const blockId = parts[0]?.id ?? `${messageId}-reasoning-merged`;
-
-    if (!throttledMergedText.trim()) {
-        return null;
-    }
-
-    return (
-        <ReasoningTimelineBlock
-            text={throttledMergedText}
-            variant="thinking"
-            onContentChange={onContentChange}
-            blockId={blockId}
-            time={mergedTime}
-            isStreaming={isStreaming}
-        />
-    );
-});
-
 export default ReasoningPart;