Refactor analyze_variability_json to streamline top 10 results extraction by consolidating DataFrame operations. Update UI button click handling to ensure consistent output across UI and MCP functionalities.

app.py

@@ -793,17 +793,17 @@ def analyze_variability_json(
     )
 
     if result_df is not None and not result_df.empty:
-        # CV(%) が NaN の場合は除外して上位10件を抽出
-        top10 = result_df.dropna(subset=["CV(%)"]).sort_values(
-            by="CV(%)", ascending=False
-        ).head(10)
-
-        # 必要な列だけ残す
-        results = top10[["項目名", "平均", "CV(%)", "スパイク数", "不安定判定"]].to_dict(orient="records")
+        # 必要なカラムだけに絞り込んで上位10件
+        top10 = (
+            result_df.dropna(subset=["CV(%)"])
+            .sort_values(by="CV(%)", ascending=False)
+            .head(10)[["項目名", "平均", "CV(%)", "スパイク数", "不安定判定"]]
+        )
+        results = top10.to_dict(orient="records")
     else:
         results = []
 
-    return {
+    return result_df, summary, {
         "meta": {
             "datetime_str": datetime_str,
             "window_minutes": window_minutes,
@@ -914,18 +914,11 @@ with gr.Blocks(css=".gradio-container {overflow: auto !important;}") as demo:
             summary_output4 = gr.Textbox(label="サマリー")
             json_output4 = gr.Json(label="JSON結果")
 
-            # ✅ UI用
-            run_btn4.click(
-                analyze_variability,
-                inputs=[datetime_str_global, window_minutes_global, cv_threshold4, jump_threshold4, mad_sigma4],
-                outputs=[result_df4, summary_output4, json_output4]
-            )
-
-            # ✅ MCP用
+            # ✅ UI + MCP 共通
             run_btn4.click(
                 analyze_variability_json,
                 inputs=[datetime_str_global, window_minutes_global, cv_threshold4, jump_threshold4, mad_sigma4],
-                outputs=[json_output4]
+                outputs=[result_df4, summary_output4, json_output4]
             )
 
 if __name__ == "__main__":