fix(uploads): retry transient DirectUploadErrors at outer KB level

The KB outer retry only triggered on isNetworkError, missing
transient 5xx from S3/Azure (DirectUploadError code
DIRECT_UPLOAD_ERROR or MULTIPART_ERROR). Adds isTransientUploadError
and retries on it, restoring resilience for small-file presigned
PUTs against flaky cloud storage.

Author: waleedlatif1

apps/sim/app/workspace/[workspaceId]/knowledge/hooks/use-knowledge-upload.ts

@@ -4,6 +4,7 @@ import { sleep } from '@sim/utils/helpers'
 import { useQueryClient } from '@tanstack/react-query'
 import {
   DirectUploadError,
+  isTransientUploadError,
   LARGE_FILE_THRESHOLD,
   MULTIPART_MAX_RETRIES,
   MULTIPART_RETRY_BACKOFF,
@@ -266,7 +267,8 @@ export function useKnowledgeUpload(options: UseKnowledgeUploadOptions = {}) {
           return buildUploadedFile(file, toAbsoluteUrl(filePath))
         }
 
-        if (isAbortError(error) || !isNetworkError(error) || attempt >= MULTIPART_MAX_RETRIES) {
+        const retryable = isNetworkError(error) || isTransientUploadError(error)
+        if (isAbortError(error) || !retryable || attempt >= MULTIPART_MAX_RETRIES) {
           throw error
         }
 

apps/sim/lib/uploads/client/direct-upload.ts

@@ -63,6 +63,16 @@ export class DirectUploadError extends Error {
   }
 }
 
+/**
+ * Transport-level upload errors worth retrying at the outer level: timeouts,
+ * 5xx from the storage backend, and per-part failures whose inner retry was
+ * exhausted. Excludes deterministic failures (`PRESIGNED_URL_ERROR`,
+ * `FALLBACK_REQUIRED`) and aborts.
+ */
+export const isTransientUploadError = (error: unknown): boolean =>
+  error instanceof DirectUploadError &&
+  (error.code === 'DIRECT_UPLOAD_ERROR' || error.code === 'MULTIPART_ERROR')
+
 const calculateUploadTimeoutMs = (fileSize: number): number => {
   const sizeInMb = fileSize / (1024 * 1024)
   const dynamicBudget = BASE_TIMEOUT_MS + sizeInMb * TIMEOUT_PER_MB_MS