fix: robust 429 retry (10 attempts, 60s max wait, clear error messages)

app.py

@@ -629,28 +629,45 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
     space_id = os.getenv("SPACE_ID", "ngbaoan/Final_Assignment_AI_agents_course")
     agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
 
-    # 1 — Fetch questions (retry up to 5× with backoff for 429)
-    yield "📡 Fetching GAIA questions…", None
+    # 1 — Fetch questions (retry up to 10× with backoff for 429)
+    import time
+    yield "📡 Fetching GAIA questions… (server may be busy, will retry automatically)", None
     questions_data = None
-    for attempt in range(1, 6):
+    last_error = None
+    for attempt in range(1, 11):
         try:
             resp = requests.get(QUESTIONS_URL, timeout=30)
             if resp.status_code == 429:
-                wait_sec = 10 * attempt          # 10s, 20s, 30s, 40s, 50s
-                yield f"⏳ Rate-limited (429). Waiting {wait_sec}s before retry {attempt}/5…", None
-                import time; time.sleep(wait_sec)
+                wait_sec = min(15 * attempt, 60)   # 15s, 30s, 45s, 60s max
+                if attempt == 10:
+                    last_error = "Server still rate-limiting after 10 attempts (429)."
+                    break
+                yield (
+                    f"⏳ Server busy (429). Waiting {wait_sec}s… "
+                    f"(attempt {attempt}/10 — this is normal, please wait)",
+                    None,
+                )
+                time.sleep(wait_sec)
                 continue
             resp.raise_for_status()
             questions_data = resp.json()
             break
         except Exception as exc:
-            if attempt == 5:
-                yield f"❌ Failed to fetch questions after 5 attempts: {exc}", None
-                return
-            import time; time.sleep(10 * attempt)
+            last_error = str(exc)
+            if attempt == 10:
+                break
+            wait_sec = min(15 * attempt, 60)
+            yield f"⚠️ Attempt {attempt}/10 failed: {exc}. Retrying in {wait_sec}s…", None
+            time.sleep(wait_sec)
 
     if not questions_data:
-        yield "❌ No questions received.", None
+        yield (
+            f"❌ Could not fetch questions from the scoring server.\n"
+            f"Reason: {last_error}\n\n"
+            f"💡 This is a server-side rate limit — NOT a bug in your code.\n"
+            f"Please wait a few minutes and click Run again.",
+            None,
+        )
         return
 
     total = len(questions_data)