检索结果过程添加

frontend-web/src/open-canvas/api/ai-writing.ts

@@ -29,7 +29,7 @@ export async function startWritingSession(req: AIWritingRequest): Promise<string
 }
 
 const SSE_EVENT_TYPES: ProgressEventType[] = [
-  'connected', 'step_start', 'step_done',
+  'connected', 'step_start', 'step_done', 'research_progress',
   'materials_ready', 'outline_chunk', 'outline_ready',
   'draft_chunk', 'draft_ready',
   'review_chunk', 'review_item_done', 'review_ready',

frontend-web/src/open-canvas/components/ai-writing/AIWritingPanel.tsx

@@ -327,7 +327,7 @@ export function AIWritingPanel() {
          !(state.status === 'writing_outline' && (state.outlineStreamingText || state.outlineThinkingText)) &&
          !(state.status === 'reviewing' && state.streamingReviewItems.length > 0) && (
           <div className="px-4 pb-4">
-            <WritingSkeletonCard status={state.status} />
+            <WritingSkeletonCard status={state.status} researchProgress={state.researchProgress} />
           </div>
         )}
 
@@ -376,7 +376,9 @@ export function AIWritingPanel() {
               {isRunning && (
                 <span className="inline-block size-1.5 rounded-full bg-blue-500 animate-pulse" />
               )}
-              {STATUS_LABEL[state.status] ?? state.status}
+              {state.status === 'researching' && state.researchProgress
+                ? state.researchProgress
+                : STATUS_LABEL[state.status] ?? state.status}
             </>
           )}
         </div>

frontend-web/src/open-canvas/components/ai-writing/WritingSkeletonCard.tsx

@@ -3,6 +3,8 @@ import type { WritingStatus } from '../../types/ai-writing';
 
 interface WritingSkeletonCardProps {
   status: WritingStatus;
+  /** 素材收集阶段的实时进度文案（仅 researching 阶段使用） */
+  researchProgress?: string;
 }
 
 // ── 基础骨架元素 ──────────────────────────────────────────────────────────────
@@ -29,13 +31,13 @@ function ThinkingDots({ color = 'bg-blue-400' }: { color?: string }) {
 
 // ── 素材收集骨架屏 ────────────────────────────────────────────────────────────
 
-function MaterialsSkeleton() {
+function MaterialsSkeleton({ researchProgress }: { researchProgress?: string }) {
   return (
     <div className="space-y-3 p-4">
-      {/* 标题行 */}
+      {/* 标题行：展示素材收集专家的实时进度 */}
       <div className="flex items-center gap-2 text-xs text-blue-500 font-medium">
         <ThinkingDots color="bg-blue-400" />
-        <span>素材收集专家搜索中…</span>
+        <span>{researchProgress || '素材收集专家搜索中…'}</span>
       </div>
 
       {/* 关键词标签占位 */}
@@ -198,21 +200,28 @@ function RevisingSkeleton() {
 
 // ── 状态映射 ──────────────────────────────────────────────────────────────────
 
+// researching / awaiting_material_confirm 由 WritingSkeletonCard 单独处理（需透传进度文案）
 const SKELETON_MAP: Partial<Record<WritingStatus, React.FC>> = {
   // 实际运行状态
-  researching:             MaterialsSkeleton,
   writing_outline:         OutlineSkeleton,
   writing_draft:           DraftSkeleton,
   reviewing:               ReviewSkeleton,
   revising:                RevisingSkeleton,
   // 过渡等待状态（用户提交后、后端事件到达前的空白期）
-  awaiting_material_confirm:  MaterialsSkeleton,   // 重新搜索 / 过渡到大纲
   awaiting_outline_confirm:   OutlineSkeleton,     // 过渡到写作
   awaiting_draft_confirm:     DraftSkeleton,       // 提交修改意见或编辑审核后
   awaiting_review_confirm:    RevisingSkeleton,    // 采纳编辑意见后
 };
 
-export function WritingSkeletonCard({ status }: WritingSkeletonCardProps) {
+export function WritingSkeletonCard({ status, researchProgress }: WritingSkeletonCardProps) {
+  // 素材收集阶段单独渲染，以便透传实时进度文案
+  if (status === 'researching' || status === 'awaiting_material_confirm') {
+    return (
+      <div className="rounded-xl border bg-card shadow-sm overflow-hidden">
+        <MaterialsSkeleton researchProgress={researchProgress} />
+      </div>
+    );
+  }
   const Skeleton = SKELETON_MAP[status];
   if (!Skeleton) return null;
   return (

frontend-web/src/open-canvas/hooks/useAIWriting.ts

@@ -34,6 +34,7 @@ const INITIAL_STATE: AIWritingState = {
   streamingDraftSections: [],
   outlineStreamingText: '',
   outlineThinkingText: '',
+  researchProgress: '',
   streamingReviewItems: [],
   revisionSummary: null,
   pendingRevisionContext: null,
@@ -64,14 +65,19 @@ export function useAIWriting() {
           case 'step_done':
             return base;
 
+          case 'research_progress':
+            // 素材收集专家实时进度：仅更新最新提示文案
+            return { ...base, researchProgress: event.message ?? '' };
+
           case 'resumed':
             return { ...base, currentPause: null };
 
           case 'materials_ready':
-            // 仅缓存素材数据，不弹干预卡（等 await_user 事件再弹）
+            // 仅缓存素材数据，不弹干预卡（等 await_user 事件再弹）；进度文案已完成可清空
             return {
               ...base,
               materials: (event.materials ?? []) as AIMaterial[],
+              researchProgress: '',
             };
 
           case 'outline_chunk': {
@@ -345,6 +351,7 @@ export function useAIWriting() {
           currentPause: null,
           status: nextStatus,
           // Clear stale streaming state so new-cycle detection in chunk handlers works correctly
+          ...(nextStatus === 'researching' ? { researchProgress: '' } : {}),
           ...(nextStatus === 'writing_outline' ? { outlineStreamingText: '', outlineThinkingText: '' } : {}),
           ...(nextStatus === 'writing_draft'   ? { streamingDraftSections: [], streamingDraftTitle: '' } : {}),
           ...(nextStatus === 'reviewing'        ? { streamingReviewItems: [] } : {}),

frontend-web/src/open-canvas/types/ai-writing.ts

@@ -67,6 +67,7 @@ export type ProgressEventType =
   | 'connected'
   | 'step_start'
   | 'step_done'
+  | 'research_progress'
   | 'materials_ready'
   | 'outline_chunk'
   | 'outline_ready'
@@ -201,6 +202,8 @@ export interface AIWritingState {
   // 流式大纲（正文 / 思考分开）
   outlineStreamingText: string;
   outlineThinkingText: string;
+  // 素材收集实时进度（最新一条提示文案）
+  researchProgress: string;
   // 流式审核（逐条到达）
   streamingReviewItems: StreamingReviewItem[];
   // 修改摘要

offline-backend-20260512/backend/packages/harness/deerflow/agents/ai_writing/nodes/researcher.py

@@ -30,7 +30,21 @@ async def researcher_node(state: AIWritingState, config: RunnableConfig) -> dict
     intervention = state.get("last_user_intervention") or {}
     extra_keywords: list[str] = intervention.get("_extra_keywords") or intervention.get("extra_keywords") or []
 
+    # SSE 实时进度推送（与 outline/draft 节点共用 sse_queue 机制）
+    cfg = config.get("configurable", {})
+    sse_queue = cfg.get("sse_queue")
+    sse_session_id: str = cfg.get("sse_session_id", "")
+
+    async def _emit(message: str) -> None:
+        if sse_queue is None:
+            return
+        await sse_queue.put(("research_progress", {
+            "session_id": sse_session_id,
+            "message": message,
+        }))
+
     # Step 1: 提取关键词
+    await _emit("正在生成检索词…")
     kw_result = await llm_json(
         EXTRACT_KEYWORDS_SYSTEM,
         EXTRACT_KEYWORDS_USER.format(
@@ -49,7 +63,25 @@ async def researcher_node(state: AIWritingState, config: RunnableConfig) -> dict
     ai_writing_config = config.get("configurable", {}).get("ai_writing_config", {})
     provider = get_search_provider(ai_writing_config)
 
-    search_tasks = [provider.search(kw, max_results=3) for kw in keywords[:6]]
+    search_keywords = keywords[:6]
+    total = len(search_keywords)
+    await _emit(f"检索词已生成，开始检索 {total} 个方向…")
+
+    # 并发搜索，但每完成一个就推送进度（计数在单线程 asyncio 下无需加锁）
+    completed = 0
+
+    async def _search_one(kw: str) -> list[SearchResult]:
+        nonlocal completed
+        try:
+            return await provider.search(kw, max_results=3)
+        finally:
+            completed += 1
+            if completed < total:
+                await _emit(f"已完成 {completed}/{total} 次检索，正在继续检索相关内容…")
+            else:
+                await _emit(f"已完成全部 {total} 次检索")
+
+    search_tasks = [_search_one(kw) for kw in search_keywords]
     all_results_nested: list[list[SearchResult]] = await asyncio.gather(*search_tasks, return_exceptions=True)
 
     # 合并去重
@@ -82,6 +114,7 @@ async def researcher_node(state: AIWritingState, config: RunnableConfig) -> dict
             ))
 
     # Step 3: 相关度排序（如素材较多，让 LLM 排序）
+    await _emit("检索完成，正在筛选排序素材…")
     if len(raw_materials) > _MAX_MATERIALS:
         materials_list = "\n".join(
             f"- ID: {m['id']}  标题: {m['title']}  摘要: {m['content'][:100]}"