fix: count annotator history from both HF snapshot + local file to survive restarts

app.py

@@ -603,11 +603,30 @@ def mbench_a_start(annotator: str, state: dict):
                 gr.update(visible=False), gr.update(visible=False),
                 gr.update(visible=False),
                 "", "")
-    # Count how many tasks this annotator has already completed historically
+    # Count how many tasks this annotator has already completed.
+    # Check both:
+    # 1. MBENCH_A_COMPLETED (loaded from HF at startup + updated in-memory during this session)
+    # 2. The local annotation file (captures annotations made this session before any push)
     historical_count = sum(
         1 for anns in MBENCH_A_COMPLETED.values()
         if annotator in anns
     )
+    # Also scan the local file in case this session's annotations haven't been pushed yet
+    if ANN_FILE_MBENCH_A.exists():
+        with ANN_FILE_MBENCH_A.open() as f:
+            for line in f:
+                line = line.strip()
+                if not line:
+                    continue
+                try:
+                    r = json.loads(line)
+                    if r.get("annotator") == annotator and r.get("type") == "pairwise_mbench_a":
+                        tid = r.get("task_id", "")
+                        # Only count if not already counted in MBENCH_A_COMPLETED
+                        if tid in MBENCH_A_TASK_BY_ID and annotator not in MBENCH_A_COMPLETED.get(tid, []):
+                            historical_count += 1
+                except Exception:
+                    pass
 
     # Shuffle task order for this annotator
     order = list(range(len(MBENCH_A_TASKS)))