⚡ Bolt: Replace blocking time.sleep with asyncio.sleep in get_frame_image

.jules/bolt.md

@@ -16,3 +16,7 @@
 ## 2024-06-28 - Optimize array lookups in high-frequency React event handlers
 **Learning:** Even if data is pre-processed/memoized (like cached `parsedTimestamps`), performing an O(N) linear search on that array inside a high-frequency event handler (like `<video onTimeUpdate>`, which fires multiple times per second) can cause UI stuttering for large arrays.
 **Action:** When searching sorted arrays (like ordered video timestamps) in high-frequency event loops, replace linear iteration loops with an O(log N) binary search to minimize main thread blocking and ensure smooth UI execution.
+
+## 2024-05-24 - Replace blocking time.sleep with asyncio.sleep in async FastAPI routes
+**Learning:** FastAPI async routes must not use blocking I/O calls like `time.sleep()`. This completely blocks the main event loop, causing the entire application to hang and fail to process other concurrent requests.
+**Action:** Always use `await asyncio.sleep()` in `async def` route handlers when waiting or polling, so the event loop can process other tasks.

backend/main.py

@@ -12,7 +12,6 @@
 from typing import List, Dict, Any
 from pydantic import BaseModel, Field
 from pathlib import Path
-import time
 
 from scanner import scan_directory_for_videos
 from extractor import async_extract_keyframes_parallel, async_get_video_duration
@@ -858,8 +857,10 @@ async def get_frame_image(directory_path: str, frame_index: int, project: Projec
 
     max_wait = 30
     waited = 0
+    # ⚡ Bolt: Use asyncio.sleep instead of time.sleep to avoid blocking the main event loop
+    # Impact: Allows server to handle concurrent requests while polling for frame creation
     while not file_path.exists() and waited < max_wait:
-        time.sleep(1)
+        await asyncio.sleep(1)
         waited += 1
 
     if file_path.exists():