improve display

utils.py

@@ -61,20 +61,42 @@ def load_leaderboard() -> pd.DataFrame:
         ds = load_dataset(SCORES_REPO, token=HF_TOKEN)
         df = ds["train"].to_pandas()
 
-        display_cols = [
-            "model_id",
-            "model_name",
-            "overall_score",
-            "meter_score",
-            "verse_score",
-            "focus_score",
-            "evaluated_at",
-        ]
-        display_cols = [c for c in display_cols if c in df.columns]
-        df = df[display_cols]
+        # Build column mapping
+        col_map = {
+            "model_id": "Model ID",
+            "model_name": "Model Name",
+            "overall_score": "Overall",
+            "meter_score": "Meter",
+            "verses_score": "Verse",
+            "focus_score": "Focus",
+            "evaluated_at": "Evaluated At",
+        }
+        display_cols = [c for c in col_map if c in df.columns]
+        df = df[display_cols].copy()
 
+        # Sort by overall score descending before formatting
         if "overall_score" in df.columns:
             df = df.sort_values("overall_score", ascending=False).reset_index(drop=True)
+            df.insert(0, "Rank", range(1, len(df) + 1))
+
+        # Format percentages
+        for col in ["overall_score", "meter_score", "verses_score", "focus_score"]:
+            if col in df.columns:
+                df[col] = df[col].apply(
+                    lambda x: f"{x * 100:.1f}%" if pd.notna(x) else "N/A"
+                )
+
+        # Format dates
+        if "evaluated_at" in df.columns:
+            df["evaluated_at"] = pd.to_datetime(df["evaluated_at"], errors="coerce")
+            df["evaluated_at"] = df["evaluated_at"].dt.strftime("%Y-%m-%d %H:%M")
+
+        # Rename to human-readable names
+        df.rename(columns=col_map, inplace=True)
+
+        # Ensure column order matches empty leaderboard
+        final_cols = [c for c in _empty_leaderboard().columns if c in df.columns]
+        df = df[final_cols]
 
         return df
     except Exception as e:
@@ -82,19 +104,104 @@ def load_leaderboard() -> pd.DataFrame:
         return _empty_leaderboard()
 
 
+
 def _empty_leaderboard() -> pd.DataFrame:
     return pd.DataFrame(
         columns=[
-            "model_id",
-            "model_name",
-            "overall_score",
-            "meter_score",
-            "verse_score",
-            "focus_score",
-            "evaluated_at",
+            "Rank",
+            "Model ID",
+            "Model Name",
+            "Overall",
+            "Meter",
+            "Verse",
+            "Focus",
+            "Evaluated At",
         ]
     )
 
+def load_leaderboard() -> pd.DataFrame:
+    """Load the latest scores from the scores dataset, with caching and formatting."""
+    global _leaderboard_cache
+
+    # Return cached result if still fresh
+    if _leaderboard_cache is not None:
+        df, cached_at = _leaderboard_cache
+        if time.time() - cached_at < CACHE_TTL_SECONDS:
+            return df
+
+    if not HF_TOKEN:
+        _leaderboard_cache = (_empty_leaderboard(), time.time())
+        return _leaderboard_cache[0]
+
+    try:
+        files = list_repo_files(SCORES_REPO, repo_type="dataset", token=HF_TOKEN)
+        score_files = [f for f in files if f.startswith("scores/") and f.endswith(".json")]
+
+        records = []
+        for f in score_files:
+            try:
+                path = hf_hub_download(
+                    repo_id=SCORES_REPO,
+                    filename=f,
+                    repo_type="dataset",
+                    token=HF_TOKEN,
+                )
+                with open(path, "r", encoding="utf-8") as file:
+                    record = json.load(file)
+                records.append(record)
+            except Exception as e:
+                print(f"Could not load score file {f}: {e}")
+                continue
+
+        if not records:
+            _leaderboard_cache = (_empty_leaderboard(), time.time())
+            return _leaderboard_cache[0]
+
+        df = pd.DataFrame(records)
+
+        # Build column mapping
+        col_map = {
+            "model_id": "Model ID",
+            "model_name": "Model Name",
+            "overall_score": "Overall Score",
+            "metric_a": "Metric A",
+            "metric_b": "Metric B",
+            "evaluated_at": "Evaluated At",
+        }
+        display_cols = [c for c in col_map if c in df.columns]
+        df = df[display_cols].copy()
+
+        # Sort by overall score descending before formatting
+        if "overall_score" in df.columns:
+            df = df.sort_values("overall_score", ascending=False).reset_index(drop=True)
+            df.insert(0, "Rank", range(1, len(df) + 1))
+
+        # Format percentages
+        for col in ["overall_score", "metric_a", "metric_b"]:
+            if col in df.columns:
+                df[col] = df[col].apply(
+                    lambda x: f"{x * 100:.1f}%" if pd.notna(x) else "N/A"
+                )
+
+        # Format dates
+        if "evaluated_at" in df.columns:
+            df["evaluated_at"] = pd.to_datetime(df["evaluated_at"], errors="coerce")
+            df["evaluated_at"] = df["evaluated_at"].dt.strftime("%Y-%m-%d %H:%M")
+
+        # Rename to human-readable names
+        df.rename(columns=col_map, inplace=True)
+
+        # Ensure column order matches empty leaderboard
+        final_cols = [c for c in _empty_leaderboard().columns if c in df.columns]
+        df = df[final_cols]
+
+        _leaderboard_cache = (df, time.time())
+        return df
+    except Exception as e:
+        print(f"Could not load scores: {e}")
+        _leaderboard_cache = (_empty_leaderboard(), time.time())
+        return _leaderboard_cache[0]
+
 
 def request_model(model_id: str, request: gr.Request) -> str:
     """