Update README.md: Correct title and emoji, adjust color scheme

.gitignore

@@ -0,0 +1,3 @@
+.venv/
+*.un~
+.env

README.md

@@ -1,8 +1,8 @@
 ---
-title: Operation Data Analysis2
-emoji: 🐨
+title: Operation Data Analysis
+emoji: 🦀
 colorFrom: gray
-colorTo: red
+colorTo: gray
 sdk: gradio
 sdk_version: 5.44.1
 app_file: app.py

app.py

@@ -0,0 +1,173 @@
+from dotenv import load_dotenv
+load_dotenv()
+
+import os
+# 環境変数から取得
+SUPABASE_URL = os.environ.get('SUPABASE_URL')
+
+SUPABASE_KEY = os.environ.get('SUPABASE_KEY')
+
+import supabase
+table_threshold = "Threshold_data"
+table_sensor = "Sensor_data"
+table_troubleshooting = "Troubleshooting_collection"
+
+# クライアントの初期化
+supabase_client = supabase.create_client(SUPABASE_URL, SUPABASE_KEY)
+
+# データ取得 (初回のみ実行)
+threshold_data = supabase_client.table(table_threshold).select("*").execute()
+sensor_data = supabase_client.table(table_sensor).select("*").execute()
+troubleshooting_data = supabase_client.table(table_troubleshooting).select("*").execute()
+
+import pandas as pd
+# データフレームへの変換 (初回のみ実行)
+threshold_df = pd.DataFrame(threshold_data.data)
+sensor_df = pd.DataFrame(sensor_data.data)
+troubleshooting_df = pd.DataFrame(troubleshooting_data.data)
+
+# Convert 'datetime' column to datetime objects (初回のみ実行)
+sensor_df['datetime'] = pd.to_datetime(sensor_df['datetime'])
+
+
+# 閾値チェック関数の定義
+def check_thresholds(sensor_df_filtered, threshold_df): # Renamed parameter to clarify it's the filtered data
+    alerts = []
+
+    # '下限'と'上限'カラムを数値型に変換。変換できない値はNaNとする。
+    threshold_df['下限'] = pd.to_numeric(threshold_df['下限'], errors='coerce')
+    threshold_df['上限'] = pd.to_numeric(threshold_df['上限'], errors='coerce')
+
+    for _, row in threshold_df.iterrows():
+        metric = row["指標名"]
+        min_val = row["下限"]
+        max_val = row["上限"]
+        data_no = row["No."] # Get the 'No.' from threshold_df
+
+        # センサーデータに指標が存在しない場合はスキップ
+        if metric not in sensor_df_filtered.columns: # Use filtered data
+            continue
+
+        # センサーデータの該当カラムを数値型に変換。変換できない値はNaNとする。
+        sensor_metric_data = pd.to_numeric(sensor_df_filtered[metric], errors='coerce') # Use filtered data
+
+
+        for index, value in sensor_metric_data.items():
+            # Use the index from sensor_metric_data to get the timestamp from the filtered sensor_df passed to the function
+            # Ensure the index exists in the filtered sensor_df
+            if index in sensor_df_filtered.index:
+                timestamp = sensor_df_filtered.loc[index, "datetime"] if "datetime" in sensor_df_filtered.columns else index
+            else:
+                # Handle cases where the index might not be in the filtered dataframe (shouldn't happen with .copy() and .items())
+                continue
+
+
+            # 下限チェック
+            if pd.notna(min_val) and pd.notna(value) and value < min_val:
+                alerts.append({
+                    "timestamp": timestamp,
+                    "metric": metric,
+                    "value": value,
+                    "status": f"下限値 {min_val} 未満",
+                    "data no.": data_no # Add the 'data no.'
+                })
+
+            # 上限チェック
+            if pd.notna(max_val) and pd.notna(value) and value > max_val:
+                alerts.append({
+                    "timestamp": timestamp,
+                    "metric": metric,
+                    "value": value,
+                    "status": f"上限値 {max_val} 超過",
+                    "data no.": data_no # Add the 'data no.'
+                })
+
+    return pd.DataFrame(alerts)
+
+# Gradioインターフェースの構築
+import gradio as gr
+import pandas as pd
+import supabase
+import datetime # Import datetime here as it's used in run_troubleshooting
+
+# Assuming the data loading and check_thresholds function from the previous cell are available
+
+# トラブルシューティング実行関数の定義
+def run_troubleshooting():
+    try:
+        # Get current time and calculate the time 24 hours ago
+        current_time = datetime.datetime.now(datetime.timezone.utc)
+        time_24_hours_ago = current_time - datetime.timedelta(hours=24)
+
+        # Filter sensor data for the last 24 hours
+        # Use the globally available sensor_df and filter it each time the function is called
+        global sensor_df
+        recent_sensor_df = sensor_df[sensor_df['datetime'] >= time_24_hours_ago].copy()
+
+        # Ensure other dataframes are accessible (they are loaded globally once)
+        global threshold_df
+        global troubleshooting_df
+
+
+        # しきい値チェックの実行
+        alerts_df = check_thresholds(recent_sensor_df, threshold_df) # Pass the filtered data
+
+        # タイムスタンプごとのユニークなデータ番号の数をカウント
+        grouped_alerts = alerts_df.groupby('timestamp')['data no.'].nunique()
+        # 複数のデータ番号を持つタイムスタンプを抽出
+        multiple_data_nos_timestamps = grouped_alerts[grouped_alerts > 1].index.tolist()
+
+        # 複数の��ータ番号を持つタイムスタンプに該当するアラートをフィルタリング
+        filtered_alerts_df = alerts_df[alerts_df['timestamp'].isin(multiple_data_nos_timestamps)]
+
+        # タイムスタンプごとにデータ番号をリスト化
+        data_nos_by_timestamp = filtered_alerts_df.groupby('timestamp')['data no.'].unique().apply(list)
+
+        # 結果リストの作成
+        result_list = []
+        for timestamp, data_nos in data_nos_by_timestamp.items():
+            data_nos_str = ', '.join(map(str, data_nos))
+            result_list.append({"timestamp": timestamp, "data_nos": data_nos_str})
+
+        # 結果データフレームの作成
+        result_df = pd.DataFrame(result_list)
+
+        # If no alerts, return "異常ありません"
+        if result_df.empty:
+            return "過去24時間 異常ありません"
+
+        # トラブルシューティングデータフレームの指標番号リストを整数リストに変換
+        troubleshooting_indicator_lists = troubleshooting_df['指標No.'].str.split(',').apply(lambda x: [int(i) for i in x])
+        # 結果データフレームのデータ番号リストを整数リストに変換
+        result_data_nos_lists = result_df['data_nos'].str.split(', ').apply(lambda x: [int(i) for i in x])
+
+        # 出力テキストの生成
+        output_text = ""
+        for i, result_nos in enumerate(result_data_nos_lists):
+            result_timestamp = result_df.loc[i, 'timestamp']
+            for j, troubleshooting_nos in enumerate(troubleshooting_indicator_lists):
+                # 結果のデータ番号がトラブルシューティングの指標番号のスーパーセットであるか確認
+                if set(troubleshooting_nos).issubset(set(result_nos)):
+                    troubleshooting_situation = troubleshooting_df.loc[j, 'シチュエーション\n（対応が必要な状況）']
+                    troubleshooting_action = troubleshooting_df.loc[j, 'sub goal到達のために必要な行動\n（解決策）']
+
+                    output_text += f"Timestamp: {result_timestamp}\n"
+                    output_text += f"Trouble: {troubleshooting_situation}\n"
+                    output_text += f"Troubleshooting: {troubleshooting_action}\n"
+                    output_text += "-" * 20 + "\n" # 区切り線
+
+        return output_text
+    except Exception as e:
+        return f"エラーが発生しました: {type(e).__name__} - {e}"
+
+# Gradioインターフェースの設定
+iface = gr.Interface(
+    fn=run_troubleshooting,
+    inputs=None,  # No direct input needed as it uses existing dataframes
+    outputs="text",
+    title="Troubleshooting Information",
+    description="Displays troubleshooting information based on sensor and threshold data."
+)
+
+# Gradioインターフェースの起動
+iface.launch(mcp_server=True)

requirements.txt

@@ -0,0 +1,4 @@
+gradio[mcp]
+supabase
+numpy
+python-dotenv