Update app.py

app.py

@@ -5,6 +5,7 @@ import pandas as pd
 import os, random, tempfile, json
 import re
 import threading
+import uuid
 
 # --- API設定 ---
 API_KEY = os.getenv("API_KEY")
@@ -24,6 +25,38 @@ levels = [1, 2, 3, 4, 5]
 # ✅ ログの同時書き込みガード（ログは全ユーザーで共有）
 _log_lock = threading.Lock()
 
+# --- セッション state 初期化 ---
+def _new_session_state():
+    return {
+        "session_id": str(uuid.uuid4()),
+        "used_passages": set(),
+        "question_count": 0,     # ✅ submit毎に増える（=採点済みの問題数）
+        "user_id": None,
+        "action_log": [],        # ✅ 1問の間の選択変更履歴（submit時に一緒に保存）
+    }
+
+# --- ログ追記（CSVのみ） ---
+def log_to_csv(entry: dict):
+    with _log_lock:
+        df = pd.DataFrame([entry])
+        if os.path.exists(LOG_FILE):
+            df.to_csv(LOG_FILE, mode="a", index=False, header=False)
+        else:
+            df.to_csv(LOG_FILE, index=False)
+
+# --- イベントログ（start / select / submit / submit_result） ---
+def log_event(event: str, state: dict, **kwargs):
+    now = (datetime.utcnow() + timedelta(hours=9)).strftime("%Y-%m-%d %H:%M:%S")
+    entry = {
+        "time": now,
+        "event": event,
+        "session_id": state.get("session_id"),
+        "user_id": state.get("user_id"),
+        "question_count_scored": state.get("question_count"),
+        **kwargs
+    }
+    log_to_csv(entry)
+
 # --- AIで問題生成 ---
 def generate_question(text):
     prompt = f"""
@@ -81,7 +114,7 @@ def adaptive_test(prev_level, prev_correct):
         return levels[idx - 1]
     return prev_level
 
-# --- passage取得（✅ used_passages は state から渡す） ---
+# --- passage取得（used_passages は state から渡す） ---
 def get_passage(level, used_passages):
     subset = passages_df[passages_df["level"] == level]
     available = [pid for pid in subset["passage_id"] if pid not in used_passages]
@@ -91,30 +124,13 @@ def get_passage(level, used_passages):
     text = subset[subset["passage_id"] == passage_id]["text"].iloc[0]
     return passage_id, text
 
-# --- ログ追記（✅ CSVのみ。Dataset保存はしない） ---
-def log_to_csv(entry):
-    with _log_lock:
-        df = pd.DataFrame([entry])
-        if os.path.exists(LOG_FILE):
-            df.to_csv(LOG_FILE, mode="a", index=False, header=False)
-        else:
-            df.to_csv(LOG_FILE, index=False)
-
-# --- セッション state 初期化 ---
-def _new_session_state():
-    return {
-        "used_passages": set(),
-        "question_count": 0,
-        "user_id": None,
-        "action_log": [],
-    }
-
-# --- 開始ボタン動作 ---
+# --- 開始ボタン動作（startログを残す） ---
 def start_test(student_id, state):
-    # 新規開始時は常に初期化
+    # 新規開始時は常に初期化（session_idも更新）
     state = _new_session_state()
 
     if not student_id or student_id.strip() == "":
+        # （必要なら invalid start も残せるが、今回は残さない）
         return (
             state,
             "", "", 0, "", None,
@@ -131,31 +147,64 @@ def start_test(student_id, state):
     question = generate_question(text)
     displayed_time = datetime.utcnow() + timedelta(hours=9)
 
+    # ✅ startイベントをCSVに記録
+    log_event(
+        "start",
+        state,
+        initial_level=level,
+        passage_id=passage_id,
+        displayed_time=displayed_time.isoformat()
+    )
+
     return (
         state,
         text, question, level, passage_id, None,
         "", True, displayed_time.isoformat(), 1, state["user_id"]
     )
 
-# --- 選択肢変更イベント（操作履歴を保存） ---
+# --- 選択肢変更イベント（変更も全部CSVに残す） ---
 def log_choice_change(choice, question_number, user_id, state):
     if not isinstance(state, dict):
         state = _new_session_state()
+
     if choice:
+        t_iso = (datetime.utcnow() + timedelta(hours=9)).isoformat()
+
+        # ✅ メモリ上の操作履歴（submit_resultにまとめて入れる）
         state["action_log"].append({
             "action": "choice",
             "choice": choice,
-            "time": (datetime.utcnow() + timedelta(hours=9)).isoformat()
+            "time": t_iso
         })
+
+        # ✅ selectイベントをCSVに都度記録（変更も全部残る）
+        log_event(
+            "select",
+            state,
+            question_number=question_number,
+            choice=choice
+        )
     return state
 
-# --- 回答送信 ---
+# --- 回答送信（submitログ + 結果ログ） ---
 def next_step(prev_level, user_answer, question_text, passage_text,
               displayed_time, question_number, user_id, passage_id, state):
 
     if not isinstance(state, dict):
         state = _new_session_state()
 
+    # ✅ submitイベントを「ボタン押下ごとに」必ずCSVに記録（未選択でも残す）
+    log_event(
+        "submit",
+        state,
+        question_number=question_number,
+        reading_level=prev_level,
+        passage_id=passage_id,
+        choice=user_answer,
+        displayed_time=displayed_time,
+        submit_status="ok" if user_answer else "no_answer"
+    )
+
     if not user_answer:
         return (
             state,
@@ -164,25 +213,26 @@ def next_step(prev_level, user_answer, question_text, passage_text,
         )
 
     submit_time = datetime.utcnow() + timedelta(hours=9)
+
+    # ✅ 採点済み数を増やす（=この問題は確定）
     state["question_count"] += 1
 
     correct = check_answer_with_ai(passage_text, question_text, user_answer)
     new_level = adaptive_test(prev_level, correct)
 
-    # ✅ ログ保存（actionsに選択履歴を入れる）
-    entry = {
-        "user_id": user_id,
-        "question_number": state["question_count"],
-        "reading_level": prev_level,
-        "passage_id": passage_id,
-        "question": question_text,
-        "user_answer": user_answer,
-        "correct": correct,
-        "displayed_time": displayed_time,
-        "submit_time": submit_time.strftime("%Y-%m-%d %H:%M:%S"),
-        "actions": json.dumps(state["action_log"], ensure_ascii=False)
-    }
-    log_to_csv(entry)
+    # ✅ submit_result（採点結果）をCSVに記録（actionsも含める）
+    log_event(
+        "submit_result",
+        state,
+        question_number=state["question_count"],
+        reading_level=prev_level,
+        passage_id=passage_id,
+        question=question_text,
+        user_answer=user_answer,
+        correct=correct,
+        submit_time=submit_time.strftime("%Y-%m-%d %H:%M:%S"),
+        actions=json.dumps(state["action_log"], ensure_ascii=False)
+    )
 
     # 最終問題なら結果だけ大きく表示
     if state["question_count"] >= 5:
@@ -201,7 +251,7 @@ def next_step(prev_level, user_answer, question_text, passage_text,
     next_question = generate_question(next_text)
     next_display_time = datetime.utcnow() + timedelta(hours=9)
 
-    # ✅ 新しい問題のために選択操作ログをリセット
+    # ✅ 次の問題のために選択操作ログをリセット
     state["action_log"] = []
 
     feedback = "✅ Correct!" if correct else "❌ Incorrect."
@@ -251,7 +301,7 @@ with gr.Blocks() as demo:
     feedback_display = gr.Markdown()
 
     hidden_level = gr.Number(visible=False)
-    hidden_passage = gr.Textbox(visible=False)
+    hidden_passage = gr.Textbox(visible=False)          # ※使っていないが互換で残す
     hidden_display_time = gr.Textbox(visible=False)
     hidden_question_number = gr.Number(visible=False)
     hidden_user_id = gr.Textbox(visible=False)
@@ -269,14 +319,14 @@ with gr.Blocks() as demo:
         ]
     )
 
-    # 選択肢変更ログ
+    # ✅ 選択肢変更ログ（変更も全部残す）
     user_answer.change(
         fn=log_choice_change,
         inputs=[user_answer, hidden_question_number, hidden_user_id, session_state],
         outputs=[session_state]
     )
 
-    # 回答送信
+    # ✅ 回答送信（submitログ + resultログ）
     submit_btn.click(
         fn=next_step,
         inputs=[hidden_level, user_answer, question_display, text_display,
@@ -289,7 +339,7 @@ with gr.Blocks() as demo:
         ]
     )
 
-    # 表示のON/OFF制御
+    # ✅ 表示のON/OFF制御
     def toggle_visibility(show):
         v = bool(show)
         return (