feat(kb): implement shared ingestion executor and simplify delete operations

Backend changes:
- Add shared thread pool _ingest_executor to prevent global thread pool exhaustion
- Implement graceful shutdown with timeout mechanism (30s) for ingestion executor
- Fix rename_collection_api cross-user data pollution by adding user_id filter
- Improve rename_collection_api transaction safety with pre-validation
- Simplify delete_document_api: remove complex UploadedFile and physical file handling
- Physical file cleanup is now handled at collection level only

Frontend changes:
- Extend IngestionResult interface with embedding_count, vector_count, warnings, failed_step
- Add toast warning when embedding_count is 0 but chunks_count > 0
- Improve error handling for both document and web ingestion with partial status warnings

Author: sqhyz55

frontend/src/components/kb/knowledge-base-creation-dialog.tsx

@@ -33,9 +33,12 @@ import { CloudConnectDialog, CloudFile } from "./cloud-connect-dialog"
 interface IngestionResult {
   collection: string
   document_count: number
-  chunks_count: number
+  chunk_count: number
+  embedding_count?: number
+  vector_count?: number
   status: string
   message: string
+  warnings?: string[]
   failed_step?: string
 }
 
@@ -349,16 +352,29 @@ export function KnowledgeBaseCreationDialog({ open, onOpenChange, onSuccess }: K
           }))
         }
 
-        const result = isJsonRecord(parsed.data) ? parsed.data as unknown as IngestionResult : null
+        const result = isJsonRecord(parsed.data)
+          ? (parsed.data as unknown as IngestionResult)
+          : null
         if (!result) {
           throw new Error(t("kb.errors.uploadFailedFile", { name: file.name }))
         }
-        setIngestionResults(prev => [...prev, result])
 
-        if (result.status === "partial" && result.failed_step) {
-          throw new Error(result.message || t("kb.errors.failedAtStep", { step: result.failed_step }))
+        if (result.status === "error") {
+          throw new Error(result.message || t("kb.errors.uploadFailedFile", { name: file.name }))
+        }
+        if (result.status === "partial") {
+          toast.warning(
+            result.message + (result.warnings?.length ? ` Warnings: ${result.warnings.join(", ")}` : "")
+          )
+        }
+        if (result.status === "success" && (result.embedding_count ?? 0) === 0 && (result.chunk_count ?? 0) > 0) {
+          toast.error(
+            "文档上传成功，但 embedding 生成失败（" + result.chunk_count + " 个 chunks 未生成 embedding）。文档无法被搜索。请检查 embedding 模型配置和 API 状态。" +
+            (result.warnings?.length ? ` 警告: ${result.warnings.join(", ")}` : "")
+          )
         }
 
+        setIngestionResults(prev => [...prev, result])
         successfulCollections.push(collectionName)
         setCompletedUploadCount(i + 1)
         setUploadProgress(((i + 1) / selectedFiles.length) * 100)
@@ -449,6 +465,16 @@ export function KnowledgeBaseCreationDialog({ open, onOpenChange, onSuccess }: K
       if (!result) {
         throw new Error(t("kb.errors.webIngestFailed"))
       }
+
+      if (result.status === "error") {
+        throw new Error(result.message || t("kb.errors.webIngestFailed"))
+      }
+      if (result.status === "partial") {
+        toast.warning(
+          result.message + (result.warnings?.length ? ` Warnings: ${result.warnings.join(", ")}` : "")
+        )
+      }
+
       setWebIngestionResult(result)
       setWebIngestionProgress(100)
 
@@ -721,7 +747,7 @@ export function KnowledgeBaseCreationDialog({ open, onOpenChange, onSuccess }: K
                                       {result.document_count} {t("kb.dialog.fileUpload.processResult.createDocuments")}
                                     </Badge>
                                     <Badge variant="outline" className="text-xs">
-                                      {result.chunks_count} {t("kb.dialog.fileUpload.processResult.textChunks")}
+                                      {result.chunk_count} {t("kb.dialog.fileUpload.processResult.textChunks")}
                                     </Badge>
                                   </>
                                 )}

src/xagent/web/api/kb.py

@@ -518,6 +518,42 @@ async def wrapper(*args: Any, **kwargs: Any) -> Any:
 )
 
 
+def shutdown_ingest_executor() -> None:
+    """Shutdown the shared ingestion executor gracefully.
+
+    Called during application shutdown to ensure pending tasks complete
+    and resources are properly released. Uses a timeout to prevent blocking
+    application shutdown indefinitely.
+    """
+    logger.info("Shutting down ingestion executor...")
+    try:
+        import threading
+
+        shutdown_complete = threading.Event()
+
+        def wait_for_shutdown() -> None:
+            _ingest_executor.shutdown(wait=True)
+            shutdown_complete.set()
+
+        shutdown_thread = threading.Thread(target=wait_for_shutdown, daemon=True)
+        shutdown_thread.start()
+
+        if shutdown_complete.wait(timeout=30):
+            logger.info("Ingestion executor shutdown complete")
+        else:
+            logger.warning(
+                "Executor shutdown timed out after 30s; forcing shutdown. "
+                "Some ingestion tasks may be incomplete."
+            )
+            _ingest_executor.shutdown(wait=False)
+    except Exception as e:
+        logger.error("Error during executor shutdown: %s", e)
+        try:
+            _ingest_executor.shutdown(wait=False)
+        except Exception:
+            pass
+
+
 class CloudFile(BaseModel):
     provider: str
     fileId: str

src/xagent/web/app.py

@@ -562,6 +562,12 @@ async def shutdown_event() -> None:
     if sandbox_mgr:
         await sandbox_mgr.cleanup()
 
+    # Shutdown shared ingestion executor
+    from .api.kb import shutdown_ingest_executor
+
+    shutdown_ingest_executor()
+    logger.info("Ingestion executor shutdown completed")
+
 
 # Frontend is now served by Next.js at http://localhost:3000
 # This backend only provides API endpoints