tag 支持拖拽

src/components/WorkflowCanvas.tsx

@@ -99,10 +99,6 @@ import { canConnectByNodeConnectionSpec } from "@/utils/nodeConnectionSpec";
 
 const BOX_SELECTION_EXPAND_MIN_DISTANCE = 4;
 const MULTI_SELECTION_BOUNDS_PADDING = 40;
-const GROUP_NODE_Z_INDEX = 0;
-const DEFAULT_NODE_Z_INDEX = 1;
-const INITIAL_TOP_STACK_Z_INDEX = 100;
-const INITIAL_TOP_MEMBER_Z_INDEX = 1;
 
 type ScreenPoint = {
   x: number;
@@ -227,17 +223,15 @@ function getWorkflowNodeDimensions(node: Node): { width: number; height: number
 // wrapper is left untouched during a drag.
 const reactFlowNodeDimensionsCache = new WeakMap<Node, Node>();
 
-// Cache the visual-state-decorated (className + zIndex) node keyed on the RAW
-// store node. `applyNodeChanges` keeps the same reference for nodes that didn't
-// move this frame, so keying on the raw node lets us return the exact same
-// decorated object across frames — as long as the computed className/zIndex
-// signature is unchanged. Store nodes never carry className/zIndex themselves
-// (they're undefined), so without this cache the `{...node}` branch below minted
-// a new object for EVERY node on EVERY frame, breaking React Flow's per-node
-// `adoptUserNodes` reference check and re-rendering every wrapper during a drag.
+// Cache the className-decorated node keyed on the RAW store node. `applyNodeChanges`
+// keeps the same reference for nodes that didn't change this frame, so keying on the
+// raw node returns the exact same decorated object across frames as long as the
+// dim/skip className is unchanged — keeping React Flow's per-node wrapper untouched.
+// (z-index is no longer derived here; React Flow owns it via elevateNodesOnSelect +
+// parentId, so nodes carry no zIndex field and this overlay only touches className.)
 const nodeVisualStateCache = new WeakMap<
   WorkflowNode,
-  { decorated: WorkflowNode; className: string; zIndex: number }
+  { decorated: WorkflowNode; className: string }
 >();
 
 function withReactFlowNodeDimensions<T extends Node>(node: T): T {
@@ -624,11 +618,6 @@ export function WorkflowCanvas() {
   const [isBuildingWorkflow, setIsBuildingWorkflow] = useState(false);
   const [isMiniMapOpen, setIsMiniMapOpen] = useState(false);
   const [zoomPercent, setZoomPercent] = useState(100);
-  const [draggingGroupId, setDraggingGroupId] = useState<string | null>(null);
-  const [stackLayerMap, setStackLayerMap] = useState<Record<string, number>>({});
-  const [memberLayerMap, setMemberLayerMap] = useState<Record<string, number>>({});
-  const [topStackLayer, setTopStackLayer] = useState(INITIAL_TOP_STACK_Z_INDEX);
-  const topMemberLayerRef = useRef(INITIAL_TOP_MEMBER_Z_INDEX);
   const canvasGridColor = "var(--canvas-grid)";
   const miniMapMaskColor = "var(--surface-overlay)";
   const miniMapNodeColor = "var(--text-muted)";
@@ -643,41 +632,6 @@ export function WorkflowCanvas() {
   const lastCanvasPointerFlowPositionRef = useRef<ScreenPoint | null>(null);
   const [boxSelectionRect, setBoxSelectionRect] = useState<FlowRect | null>(null);
 
-  const bringNodeToFront = useCallback((node: Node) => {
-    const nodeGroupId = node.type === "group"
-      ? (node.data as { groupId?: string }).groupId
-      : (node as Node & { groupId?: string }).groupId;
-    const groupNodeId = nodeGroupId ? `group-node-${nodeGroupId}` : null;
-
-    if (node.type !== "group" && nodeGroupId && groupNodeId) {
-      topMemberLayerRef.current += 1;
-      const nextMemberLayer = topMemberLayerRef.current;
-      setMemberLayerMap((layers) => ({
-        ...layers,
-        [node.id]: nextMemberLayer,
-      }));
-
-      setTopStackLayer((current) => {
-        const nextStackLayer = current + 1;
-        setStackLayerMap((layers) => ({
-          ...layers,
-          [groupNodeId]: nextStackLayer,
-        }));
-        return nextStackLayer;
-      });
-      return;
-    }
-
-    setTopStackLayer((current) => {
-      const nextStackLayer = current + 1;
-      setStackLayerMap((layers) => ({
-        ...layers,
-        [node.id]: nextStackLayer,
-      }));
-      return nextStackLayer;
-    });
-  }, []);
-
   const onNodesChange = useCallback((changes: NodeChange<WorkflowNode>[]) => {
     if (!boxSelectionActiveRef.current) {
       storeOnNodesChange(changes);
@@ -967,27 +921,14 @@ export function WorkflowCanvas() {
     }
   }, [tutorialActive, nodes, setCenter]);
 
-  // Apply visual state that React Flow can turn into DOM classes and z-index.
+  // Overlay dim/skip CSS classes that React Flow turns into DOM classes.
+  // z-index is owned by React Flow itself (elevateNodesOnSelect for the
+  // selected/dragged node + parentId auto-elevation for group members), so it is
+  // no longer derived here — this overlay only touches className.
   const allNodes = useMemo(() => {
-    const draggingGroupNodeId = draggingGroupId ? `group-node-${draggingGroupId}` : null;
-    const groupStackLayerById = new Map<string, number>();
-    for (const node of nodes) {
-      if (node.type !== "group") continue;
-      const groupId = (node.data as { groupId?: string }).groupId;
-      const stackLayer = stackLayerMap[node.id];
-      if (groupId && stackLayer !== undefined) {
-        groupStackLayerById.set(groupId, stackLayer);
-      }
-    }
-
     return nodes.map((node) => {
       // Never dim Switch or ConditionalSwitch nodes themselves
       const canDimNode = node.type !== "switch" && node.type !== "conditionalSwitch";
-      const isDraggingGroupShell = Boolean(draggingGroupNodeId && node.id === draggingGroupNodeId);
-      const isDraggingGroupMember = Boolean(draggingGroupId && node.type !== "group" && (
-        node.groupId === draggingGroupId || node.parentId === draggingGroupNodeId
-      ));
-
       const isDimmed = canDimNode && dimmedNodeIds.has(node.id);
       const isSkipped = canDimNode && skippedNodeIds.has(node.id);
       const extraClasses = [
@@ -998,38 +939,18 @@ export function WorkflowCanvas() {
       // Preserve existing className if any, add/remove dimmed/skipped classes
       const baseClass = (node.className || "").replace(/\bswitch-dimmed\b/g, "").replace(/\bnode-skipped\b/g, "").trim();
       const newClass = extraClasses ? `${baseClass} ${extraClasses}`.trim() : baseClass;
-      const nodeGroupId = (node as Node & { groupId?: string }).groupId;
-      const groupStackLayer = nodeGroupId ? groupStackLayerById.get(nodeGroupId) : undefined;
-      let zIndex = node.type === "group"
-        ? stackLayerMap[node.id] ?? GROUP_NODE_Z_INDEX
-        : nodeGroupId && groupStackLayer !== undefined
-        ? groupStackLayer + (memberLayerMap[node.id] ?? DEFAULT_NODE_Z_INDEX)
-        : stackLayerMap[node.id] ?? DEFAULT_NODE_Z_INDEX;
-
-      if (isDraggingGroupShell) {
-        zIndex = Math.max(zIndex, topStackLayer);
-      } else if (isDraggingGroupMember) {
-        zIndex = Math.max(zIndex, topStackLayer + (memberLayerMap[node.id] ?? DEFAULT_NODE_Z_INDEX));
-      } else if (node.dragging) {
-        zIndex = Math.max(zIndex, topStackLayer + 1);
-      }
 
-      // Return a reference-stable decorated node. Keyed on the raw store node,
-      // so an unchanged node yields the identical object across frames and React
-      // Flow leaves its wrapper untouched; only the node whose className/zIndex
-      // actually changed (e.g. the one being dragged) gets a new reference.
+      // Reference-stable: keyed on the raw store node, so an unchanged node yields
+      // the identical object across frames and React Flow leaves its wrapper alone.
       const cached = nodeVisualStateCache.get(node);
-      if (cached && cached.className === newClass && cached.zIndex === zIndex) {
+      if (cached && cached.className === newClass) {
         return cached.decorated;
       }
-      const decorated =
-        node.className === newClass && node.zIndex === zIndex
-          ? node
-          : { ...node, className: newClass, zIndex };
-      nodeVisualStateCache.set(node, { decorated, className: newClass, zIndex });
+      const decorated = node.className === newClass ? node : { ...node, className: newClass };
+      nodeVisualStateCache.set(node, { decorated, className: newClass });
       return decorated;
     });
-  }, [nodes, dimmedNodeIds, draggingGroupId, memberLayerMap, skippedNodeIds, stackLayerMap, topStackLayer]);
+  }, [nodes, dimmedNodeIds, skippedNodeIds]);
 
   const absoluteNodePositions = useMemo(() => {
     const byId = new Map(allNodes.map((node) => [node.id, node]));
@@ -1269,7 +1190,6 @@ export function WorkflowCanvas() {
     };
 
     window.getSelection()?.removeAllRanges();
-    bringNodeToFront(node);
     selectSingleNode(node.id);
     setZoomPercent(Math.round(nextZoom * 100));
     setCenter(center.x, center.y, {
@@ -1277,7 +1197,7 @@ export function WorkflowCanvas() {
       zoom: nextZoom,
     });
 
-  }, [absoluteNodePositions, bringNodeToFront, nodes, selectSingleNode, setCenter]);
+  }, [absoluteNodePositions, nodes, selectSingleNode, setCenter]);
 
   const zoomNodeById = useCallback((nodeId: string) => {
     zoomNodeIntoFocusById(nodeId);
@@ -1414,6 +1334,15 @@ export function WorkflowCanvas() {
     [allNodes, batchNodes]
   );
 
+  // Pin every edge to zIndex 0 so it always renders below nodes. React Flow only
+  // auto-elevates an edge's zIndex when it's left undefined (edges touching grouped
+  // nodes get lifted to the node's band and end up above other nodes); setting an
+  // explicit value bypasses that. The ternary keeps unchanged edge references stable.
+  const reactFlowEdges = useMemo(
+    () => edges.map((edge) => (edge.zIndex === 0 ? edge : { ...edge, zIndex: 0 })),
+    [edges]
+  );
+
 
 
   const handleConnect = useCallback(
@@ -3040,28 +2969,27 @@ export function WorkflowCanvas() {
           <NodeDoubleClickZoomProvider value={{ zoomNodeById }}>
             <ReactFlow
               nodes={reactFlowNodes}
-              edges={edges}
+              edges={reactFlowEdges}
               onNodesChange={handleReactFlowNodesChange}
               onEdgesChange={onEdgesChange}
               onConnect={handleConnect}
               onConnectEnd={handleConnectEnd}
               onMoveStart={() => { isPanningRef.current = true; setHoveredNodeId(null); document.documentElement.classList.add("canvas-interacting"); }}
               onMoveEnd={() => { isPanningRef.current = false; document.documentElement.classList.remove("canvas-interacting"); syncZoomPercent(); }}
-              onNodeClick={(_event, node) => {
-                bringNodeToFront(node);
-              }}
               onNodeDragStart={(_event, node) => {
                 isDraggingNodeRef.current = true;
                 document.documentElement.classList.add("canvas-interacting");
                 document.documentElement.classList.add("canvas-node-drag-active");
-                bringNodeToFront(node);
-                setDraggingGroupId(node.type === "group" ? (node.data as { groupId?: string }).groupId ?? null : null);
+                // Selecting the dragged node lets React Flow's elevateNodesOnSelect
+                // raise it to the top. selectNodesOnDrag is false, so React Flow
+                // won't select it for us; only select when it isn't already, to
+                // preserve an existing multi-selection drag.
+                if (!node.selected) selectSingleNode(node.id);
               }}
               onNodeDragStop={(event, node) => {
                 isDraggingNodeRef.current = false;
                 document.documentElement.classList.remove("canvas-interacting");
                 document.documentElement.classList.remove("canvas-node-drag-active");
-                setDraggingGroupId(null);
                 handleNodeDragStop(event, node);
               }}
               onPaneClick={handlePaneClick}
@@ -3082,7 +3010,7 @@ export function WorkflowCanvas() {
               selectionKeyCode={["Space", "Control", "Meta"]}
               panOnDrag={canvasPanOnDrag}
               selectNodesOnDrag={false}
-              elevateNodesOnSelect={false}
+              elevateNodesOnSelect={true}
               nodesFocusable={false}
               nodeDragThreshold={5}
               nodeClickDistance={5}

src/components/composer/ComposerMentionInput.tsx

@@ -11,6 +11,7 @@ import {
   type CompositionEvent,
   type KeyboardEvent,
   type MouseEvent as ReactMouseEvent,
+  type PointerEvent as ReactPointerEvent,
   type ReactNode,
   type SyntheticEvent,
 } from "react";
@@ -172,6 +173,30 @@ function parseTokensFromDom(element: HTMLDivElement): ComposerPromptToken[] {
   return tokens;
 }
 
+function createReferenceTagLeadingVisual(
+  token: Extract<ComposerPromptToken, { type: "reference" }>
+): HTMLElement | null {
+  if (token.thumbnailUrl) {
+    const thumbnail = document.createElement("img");
+    thumbnail.src = token.thumbnailUrl;
+    thumbnail.alt = "";
+    thumbnail.draggable = false;
+    thumbnail.className = "h-4 w-4 shrink-0 rounded-[3px] object-cover";
+    return thumbnail;
+  }
+  if (token.mediaType === "audio" || token.mediaType === "video") {
+    const icon = document.createElement("span");
+    icon.contentEditable = "false";
+    icon.setAttribute("contenteditable", "false");
+    icon.className = "flex h-4 w-4 shrink-0 items-center justify-center rounded-[3px] bg-[var(--surface-active)] text-[var(--text-muted)]";
+    icon.innerHTML = token.mediaType === "audio"
+      ? '<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" class="h-3 w-3"><path d="M9 18V5l12-2v13"/><circle cx="6" cy="18" r="3"/><circle cx="18" cy="16" r="3"/></svg>'
+      : '<svg viewBox="0 0 24 24" fill="currentColor" class="h-3 w-3"><path d="M8 5v14l11-7z"/></svg>';
+    return icon;
+  }
+  return null;
+}
+
 function createReferenceTagElement(item: ComposerMentionReferenceItem): HTMLElement | null {
   const token = createReferenceToken(item);
   if (!token || token.type !== "reference") return null;
@@ -187,22 +212,28 @@ function createReferenceTagElement(item: ComposerMentionReferenceItem): HTMLElem
   tag.dataset.referenceUrl = token.url;
   if (token.thumbnailUrl) tag.dataset.referenceThumbnailUrl = token.thumbnailUrl;
   if (token.duration) tag.dataset.referenceDuration = String(token.duration);
-  tag.className = "mx-0.5 inline-flex h-5 max-w-[12rem] select-none items-center gap-1 rounded-md border border-[var(--border-subtle)] bg-[var(--surface-1)] px-1.5 align-middle text-xs leading-none text-[var(--text-primary)]";
+  tag.className = "mx-0.5 inline-flex h-5 max-w-[12rem] cursor-grab select-none items-center gap-1 rounded-md border border-[var(--border-subtle)] bg-[var(--surface-1)] px-1.5 align-middle text-xs leading-none text-[var(--text-primary)] active:cursor-grabbing";
 
-  if (token.thumbnailUrl) {
-    const thumbnail = document.createElement("img");
-    thumbnail.src = token.thumbnailUrl;
-    thumbnail.alt = "";
-    thumbnail.draggable = false;
-    thumbnail.className = "h-4 w-4 shrink-0 rounded-[3px] object-cover";
-    tag.appendChild(thumbnail);
-  }
+  const leadingVisual = createReferenceTagLeadingVisual(token);
+  if (leadingVisual) tag.appendChild(leadingVisual);
 
   const label = document.createElement("span");
   label.contentEditable = "false";
   label.setAttribute("contenteditable", "false");
   label.className = "min-w-0 truncate";
-  label.textContent = token.label;
+  // Hide the leading "@" visually while keeping it in the DOM text, so caret/offset math
+  // (which counts rendered text via Range.toString) stays aligned with the prompt string
+  // that includes "@". Stripping it from textContent shifts every downstream offset.
+  const atMatch = token.label.match(/^(@+)([\s\S]*)$/);
+  if (atMatch) {
+    const hiddenAt = document.createElement("span");
+    hiddenAt.textContent = atMatch[1];
+    hiddenAt.style.display = "none";
+    label.appendChild(hiddenAt);
+    label.appendChild(document.createTextNode(atMatch[2]));
+  } else {
+    label.textContent = token.label;
+  }
   tag.appendChild(label);
   return tag;
 }
@@ -261,6 +292,93 @@ function isReferenceTagNode(node: Node | null): node is HTMLElement {
   return node instanceof HTMLElement && node.dataset.composerReferenceTag === "true";
 }
 
+const REFERENCE_TAG_DRAG_THRESHOLD_PX = 4;
+
+function getReferenceTagAncestor(node: Node | null, root: HTMLElement): HTMLElement | null {
+  let current: Node | null = node;
+  while (current && current !== root) {
+    if (isReferenceTagNode(current)) return current;
+    current = current.parentNode;
+  }
+  return null;
+}
+
+function caretRangeFromPoint(x: number, y: number): Range | null {
+  const doc = document as Document & {
+    caretRangeFromPoint?: (x: number, y: number) => Range | null;
+    caretPositionFromPoint?: (x: number, y: number) => { offsetNode: Node; offset: number } | null;
+  };
+  if (typeof doc.caretRangeFromPoint === "function") {
+    return doc.caretRangeFromPoint(x, y);
+  }
+  if (typeof doc.caretPositionFromPoint === "function") {
+    const position = doc.caretPositionFromPoint(x, y);
+    if (!position) return null;
+    const range = document.createRange();
+    range.setStart(position.offsetNode, position.offset);
+    range.collapse(true);
+    return range;
+  }
+  return null;
+}
+
+/**
+ * Resolve the collapsed caret range for a drop point, snapping outside any reference tag
+ * so a dragged tag is never inserted inside another tag.
+ */
+function resolveReferenceDrop(
+  element: HTMLDivElement,
+  draggedTag: HTMLElement,
+  clientX: number,
+  clientY: number
+): Range | null {
+  const range = caretRangeFromPoint(clientX, clientY);
+  if (!range || !element.contains(range.startContainer)) return null;
+
+  const tagAncestor = getReferenceTagAncestor(range.startContainer, element);
+  if (tagAncestor === draggedTag) return null;
+  if (tagAncestor) {
+    const tagRect = tagAncestor.getBoundingClientRect();
+    if (clientX > tagRect.left + tagRect.width / 2) {
+      range.setStartAfter(tagAncestor);
+    } else {
+      range.setStartBefore(tagAncestor);
+    }
+    range.collapse(true);
+  }
+  return range;
+}
+
+/**
+ * Floating drag preview that follows the pointer — a clone of the whole tag
+ * (cover image + label), so the user drags a visual copy of the chip itself.
+ */
+function createReferenceDragPreview(tag: HTMLElement): HTMLElement {
+  const preview = document.createElement("div");
+  preview.style.position = "fixed";
+  preview.style.left = "0";
+  preview.style.top = "0";
+  preview.style.zIndex = "10090";
+  preview.style.pointerEvents = "none";
+  preview.style.opacity = "0.9";
+  preview.style.willChange = "transform";
+
+  const clone = tag.cloneNode(true) as HTMLElement;
+  clone.style.opacity = "1";
+  clone.style.margin = "0";
+  clone.style.boxShadow = "0 8px 24px rgba(0, 0, 0, 0.4)";
+  preview.appendChild(clone);
+
+  document.body.appendChild(preview);
+  return preview;
+}
+
+function positionReferenceDragPreview(preview: HTMLElement, clientX: number, clientY: number): void {
+  const width = preview.offsetWidth;
+  const height = preview.offsetHeight;
+  preview.style.transform = `translate(${clientX - width / 2}px, ${clientY - height - 2}px)`;
+}
+
 function getChildNodeIndex(element: HTMLDivElement, node: Node): number {
   return Array.prototype.indexOf.call(element.childNodes, node);
 }
@@ -423,6 +541,7 @@ export const ComposerMentionInput = forwardRef<HTMLDivElement, ComposerMentionIn
     const inputRef = useRef<HTMLDivElement>(null);
     const isComposingRef = useRef(false);
     const mentionTriggerIndexRef = useRef<number | null>(null);
+    const pointerDragCleanupRef = useRef<(() => void) | null>(null);
     const [menuPosition, setMenuPosition] = useState<{ left: number; top: number } | null>(null);
     const [textContextMenu, setTextContextMenu] = useState<TextContextMenuState | null>(null);
     const [activeReferenceIndex, setActiveReferenceIndex] = useState(0);
@@ -457,6 +576,99 @@ export const ComposerMentionInput = forwardRef<HTMLDivElement, ComposerMentionIn
       onChange(nextValue, textToTokens(nextValue));
     }, [onChange]);
 
+    const handleReferenceTagPointerDown = useCallback((event: ReactPointerEvent<HTMLDivElement>) => {
+      // Preserve the canvas-interaction guard for every pointer down on the editor.
+      event.stopPropagation();
+      if (readOnly || event.button !== 0 || !event.isPrimary) return;
+      const editor = inputRef.current;
+      if (!editor) return;
+      const tag = getReferenceTagAncestor(event.target as Node, editor);
+      if (!tag) return;
+
+      // Take over from contentEditable so the chip drags instead of placing a caret.
+      event.preventDefault();
+      closeReferenceMenu();
+      const pointerId = event.pointerId;
+      const startX = event.clientX;
+      const startY = event.clientY;
+      let dragging = false;
+      let preview: HTMLElement | null = null;
+      let handleMove: (moveEvent: PointerEvent) => void;
+      let handleUp: (upEvent: PointerEvent) => void;
+      let handleCancel: (cancelEvent: PointerEvent) => void;
+
+      const cleanup = () => {
+        window.removeEventListener("pointermove", handleMove);
+        window.removeEventListener("pointerup", handleUp);
+        window.removeEventListener("pointercancel", handleCancel);
+        tag.style.opacity = "";
+        if (preview) {
+          preview.remove();
+          preview = null;
+        }
+        pointerDragCleanupRef.current = null;
+      };
+
+      handleMove = (moveEvent) => {
+        if (moveEvent.pointerId !== pointerId) return;
+        if (!dragging) {
+          if (Math.hypot(moveEvent.clientX - startX, moveEvent.clientY - startY) < REFERENCE_TAG_DRAG_THRESHOLD_PX) return;
+          dragging = true;
+          editor.focus({ preventScroll: true });
+          tag.style.opacity = "0.35";
+          preview = createReferenceDragPreview(tag);
+        }
+        moveEvent.preventDefault();
+        if (preview) positionReferenceDragPreview(preview, moveEvent.clientX, moveEvent.clientY);
+        const dropRange = resolveReferenceDrop(editor, tag, moveEvent.clientX, moveEvent.clientY);
+        if (dropRange) {
+          // Move the real input caret to the drop point so it visibly follows the pointer.
+          const selection = window.getSelection();
+          if (selection) {
+            selection.removeAllRanges();
+            selection.addRange(dropRange.cloneRange());
+          }
+        }
+      };
+
+      handleUp = (upEvent) => {
+        if (upEvent.pointerId !== pointerId) return;
+        const wasDragging = dragging;
+        const dropX = upEvent.clientX;
+        const dropY = upEvent.clientY;
+        cleanup();
+        if (!wasDragging || !editor.contains(tag)) return;
+        const dropRange = resolveReferenceDrop(editor, tag, dropX, dropY);
+        if (!dropRange) return;
+        dropRange.insertNode(tag);
+        editor.normalize();
+        emitDomValue(editor);
+        window.requestAnimationFrame(() => {
+          if (!tag.isConnected) return;
+          const caretRange = document.createRange();
+          caretRange.setStartAfter(tag);
+          caretRange.collapse(true);
+          const selection = window.getSelection();
+          selection?.removeAllRanges();
+          selection?.addRange(caretRange);
+        });
+      };
+
+      handleCancel = (cancelEvent) => {
+        if (cancelEvent.pointerId !== pointerId) return;
+        cleanup();
+      };
+
+      pointerDragCleanupRef.current = cleanup;
+      window.addEventListener("pointermove", handleMove);
+      window.addEventListener("pointerup", handleUp);
+      window.addEventListener("pointercancel", handleCancel);
+    }, [closeReferenceMenu, emitDomValue, readOnly]);
+
+    useEffect(() => () => {
+      pointerDragCleanupRef.current?.();
+    }, []);
+
     const handleTextContextMenu = useCallback((event: ReactMouseEvent<HTMLDivElement>) => {
       event.preventDefault();
       event.stopPropagation();
@@ -814,7 +1026,7 @@ export const ComposerMentionInput = forwardRef<HTMLDivElement, ComposerMentionIn
             emitDomValue(event.currentTarget);
             onCompositionEnd?.(event);
           }}
-          onPointerDownCapture={stopCanvasInteraction}
+          onPointerDownCapture={handleReferenceTagPointerDown}
           onMouseDownCapture={stopCanvasInteraction}
           onWheelCapture={stopCanvasInteraction}
           style={{

src/components/composer/GenerationComposer.tsx

@@ -48,6 +48,7 @@ import {
 } from "@/utils/composerPricingContext";
 import { useAppConfigStore } from "@/store/appConfigStore";
 import { buildPreviewImageUrl } from "@/utils/previewImageUrl";
+import { buildVideoPosterUrl } from "@/utils/videoPosterUrl";
 import {
   getConcreteGenerationNodeType,
   inferCapabilityFromModel,
@@ -412,6 +413,7 @@ function ComposerEditor({
   void embedded;
   const { t } = useI18n();
   const previewImageSuffix = useAppConfigStore((state) => state.previewImageSuffix);
+  const videoPosterQuality = useAppConfigStore((state) => state.videoPosterQuality);
   const promptInputRef = useRef<HTMLDivElement>(null);
   const referenceAudioPreviewRef = useRef<HTMLAudioElement | null>(null);
   const referenceMaterialsRef = useRef<ComposerReferenceMaterial[]>([]);
@@ -962,18 +964,25 @@ function ComposerEditor({
       }];
     }
 
-    return referenceMaterials.map((material) => ({
-      key: `${material.type}-${material.id}-${material.url.slice(0, 24)}`,
-      label: material.name.replace(/^@/, ""),
-      value: `${material.name} `,
-      image: material.type === "image" ? buildPreviewImageUrl(material.url, previewImageSuffix) ?? material.url : undefined,
-      sourceNodeId: material.sourceNodeId,
-      referenceType: material.type,
-      url: material.url,
-      thumbnailUrl: material.thumbnailUrl,
-      duration: material.duration,
-    }));
-  }, [previewImageSuffix, referenceMaterials, t]);
+    return referenceMaterials.map((material) => {
+      const cover = material.type === "image"
+        ? buildPreviewImageUrl(material.url, previewImageSuffix) ?? material.url
+        : material.type === "video"
+          ? buildVideoPosterUrl(material.url, videoPosterQuality) ?? undefined
+          : undefined;
+      return {
+        key: `${material.type}-${material.id}-${material.url.slice(0, 24)}`,
+        label: material.name.replace(/^@/, ""),
+        value: `${material.name} `,
+        image: cover,
+        sourceNodeId: material.sourceNodeId,
+        referenceType: material.type,
+        url: material.url,
+        thumbnailUrl: cover,
+        duration: material.duration,
+      };
+    });
+  }, [previewImageSuffix, referenceMaterials, t, videoPosterQuality]);
 
   const runAction = useCallback(() => {
     if (isCurrentNodeRunning) {