Update app.py

app.py

@@ -8,6 +8,9 @@ import re
 import threading
 import csv  # ★追加（csv 1行追記用）
 
+# ★追加：Flesch計測
+import textstat
+
 # --- API / HF 設定 ---
 API_KEY = os.getenv("API_KEY")
 BASE_URL = "https://openrouter.ai/api/v1"
@@ -41,6 +44,7 @@ LOG_COLUMNS = [
     "assigned_level",
     "passage_id",
     "original_level",
+    "flesch_score",     # ★追加：Group1=orig_lev, Group2=rewritten fre
     "action_time",
     "action_type",
     "page_text",
@@ -90,7 +94,8 @@ def download_log_csv(password: str) -> str:
 # ======================================================
 #  新しい教材管理：passages フォルダからランダム選択
 #  ※ used_passages は session_state に保持（グローバル禁止）
-#  ★変更：target level よりスコアが低い教材から選ぶ（excelのflesch_score）
+#  ★Group2：target level よりスコアが低い教材から選ぶ（excelのflesch_score）
+#  ★Group1：全教材からランダム選択
 # ======================================================
 
 def load_passage_file(text_id):
@@ -117,12 +122,10 @@ def get_title_from_excel(text_id):
 
 def get_new_passage_random(used_passages_set, target_level):
     """
+    ★Group2用：
     passages フォルダからランダムに教材を選び（pg◯.txt）、
     passage_information.xlsx の Text# の flesch_score を original_level として返す。
-
-    ★変更点：
     - ユーザーの target_level に対応する目標FREよりも低い（=難しい）教材のみから選ぶ
-      ※ flesch_score は passage_information.xlsx から取得
     """
     level_to_flesch = {1: 90, 2: 80, 3: 70, 4: 60, 5: 50}
     target_flesch = float(level_to_flesch[int(target_level)])
@@ -171,7 +174,54 @@ def get_new_passage_random(used_passages_set, target_level):
         orig_level = None
         title = None
     else:
-        orig_level = row.iloc[0]["flesch_score"]
+        orig_level = row.iloc[0].get("flesch_score", None)
+        title = row.iloc[0].get("Title", None)
+        if pd.isna(title):
+            title = None
+        else:
+            title = str(title)
+
+    return text_id, text, orig_level, title, used_passages_set
+
+
+def get_new_passage_random_any(used_passages_set):
+    """
+    ★Group1用：target_level による難易度フィルタなし
+    passages フォルダ内の全教材からランダムに選ぶ。
+    original_level (=flesch_score) は passage_information.xlsx から取得して返す。
+    """
+    files = glob.glob("passages/pg*.txt")
+    if not files:
+        return None, None, None, None, used_passages_set
+
+    all_ids = []
+    for f in files:
+        name = os.path.basename(f)
+        num = name.replace("pg", "").replace(".txt", "")
+        if num.isdigit():
+            all_ids.append(int(num))
+
+    if not all_ids:
+        return None, None, None, None, used_passages_set
+
+    available = [pid for pid in all_ids if pid not in used_passages_set]
+    if not available:
+        used_passages_set = set()
+        available = list(all_ids)
+
+    text_id = random.choice(available)
+    used_passages_set.add(text_id)
+
+    text = load_passage_file(text_id)
+    if text is None:
+        return None, None, None, None, used_passages_set
+
+    row = passage_info_df[passage_info_df["Text#"] == text_id]
+    if len(row) == 0:
+        orig_level = None
+        title = None
+    else:
+        orig_level = row.iloc[0].get("flesch_score", None)
         title = row.iloc[0].get("Title", None)
         if pd.isna(title):
             title = None
@@ -180,6 +230,7 @@ def get_new_passage_random(used_passages_set, target_level):
 
     return text_id, text, orig_level, title, used_passages_set
 
+
 # ======================================================
 #  Group1: 本文のみ抽出（書き換えなし）
 # ======================================================
@@ -250,22 +301,68 @@ def extract_main_body(text: str) -> str:
 
 # ======================================================
 #  Rewrite（同時実行制限付き） Group2で使用
+#  ★プロンプトを「改善後プロンプト」に置換
 # ======================================================
 
-def rewrite_level(text, target_level):
+def rewrite_level(text, target_level, original_fre):
     level_to_flesch = {1: 90, 2: 80, 3: 70, 4: 60, 5: 50}
     target_flesch = level_to_flesch[int(target_level)]
 
+    try:
+        original_fre_val = float(original_fre)
+    except Exception:
+        original_fre_val = float("nan")
+
     prompt = f"""
-Rewrite the following passage so it fits about {target_flesch} Flesch Reading Ease Score
-- Extract only the portions of the text that should be read as the main body,
-excluding the title, author name, source information, chapter number, annotations, and footers.
-- When outputting, make sure sections divided by chapters, etc., are clearly distinguishable by leaving a blank line between them.
-- Preserve the original meaning faithfully.
-- Do not add new information or remove essential information.
-- Output only the rewritten passage. Do not include explanations.
+The Flesch Reading Ease score is a numeric measure of text readability,
+where higher scores indicate easier readability and lower scores indicate more difficult text.
+
+In this task, we are trying to rewrite a given text into the target Flesch Reading Ease score
+and preserving the original meaning and information.
+
+Given the original draft (Flesch Reading Ease = {original_fre_val}):
+
+[TEXT START]
 {text}
-"""
+[TEXT END]
+
+Rewrite the above text to the difficulty level of:
+Flesch Reading Ease = {target_flesch}
+
+Follow the instructions below carefully.
+
+Content preservation:
+- Maintain the original meaning faithfully.
+- Do not add new information.
+- Do not remove important information.
+- Do not introduce interpretations or opinions that are not present in the original text.
+
+Scope of rewriting:
+- Rewrite ONLY the main body text.
+- Completely EXCLUDE titles, headings, chapter labels, author names, source information, footnotes, annotations, and introductions.
+- Do NOT include any text other than the rewritten main body under any circumstances.
+
+Readability control guidelines:
+- Make sentences shorter.
+- Prefer familiar, high-frequency vocabulary.
+- Use simple and direct sentence structures.
+- Avoid jargon; if technical terms are necessary, explain them clearly in simple language.
+- Minimize figurative language, idioms, and expressions whose meanings are not directly inferable.
+
+Language modernization:
+- Rewrite the text in clear, modern English.
+- Remove archaic expressions and unnatural or outdated syntax typical of older texts at all levels.
+
+Structure and formatting:
+- Preserve the original paragraph structure of the main text.
+- Insert exactly ONE blank line between paragraphs.
+- Do NOT create new section breaks, chapter divisions, or headings.
+
+Output constraints:
+- Output only the rewritten text.
+- Do not include explanations, comments, or metadata.
+- Do not include [TEXT START] and [TEXT END] in the output.
+""".strip()
 
     with _rewrite_sem:
         resp = client.chat.completions.create(
@@ -302,12 +399,19 @@ def split_pages(text, max_words=300):
 
 # ======================================================
 #  Start（session_stateでユーザー状態管理）
+#  ★Start後：入力はリロードまで固定（1回だけ）
 # ======================================================
 
 def start_test(student_id, level_input, group_input, session_state):
     action = "start_pushed"
     now = (datetime.utcnow() + timedelta(hours=9)).isoformat()
 
+    # ★Start押下時点で入力を固定（成功/失敗問わず、リロードまで不可）
+    lock_student = gr.update(interactive=False)
+    lock_group = gr.update(interactive=False)
+    lock_level = gr.update(interactive=False)
+    lock_start = gr.update(interactive=False)
+
     if not student_id or str(student_id).strip() == "":
         entry = {
             "user_id": None,
@@ -315,6 +419,7 @@ def start_test(student_id, level_input, group_input, session_state):
             "assigned_level": None,
             "passage_id": None,
             "original_level": None,
+            "flesch_score": None,
             "action_time": now,
             "action_type": action,
             "page_text": None
@@ -334,7 +439,8 @@ def start_test(student_id, level_input, group_input, session_state):
             gr.update(interactive=False, visible=False),
             gr.update(interactive=False, visible=True),
             gr.update(interactive=False, visible=False),
-            session_state
+            session_state,
+            lock_student, lock_group, lock_level, lock_start
         )
 
     user_id = str(student_id).strip()
@@ -349,13 +455,19 @@ def start_test(student_id, level_input, group_input, session_state):
         "assigned_level": level,
         "passage_id": None,
         "original_level": None,
+        "flesch_score": None,
         "action_time": now,
         "action_type": action,
         "page_text": None
     }
     save_log(entry)
 
-    pid, text, orig_lev, title, used_passages_set = get_new_passage_random(used_passages_set, level)
+    # ★変更：Group1は全教材、Group2はレベル制限あり
+    if group == 1:
+        pid, text, orig_lev, title, used_passages_set = get_new_passage_random_any(used_passages_set)
+    else:
+        pid, text, orig_lev, title, used_passages_set = get_new_passage_random(used_passages_set, level)
+
     if text is None:
         return (
             "",
@@ -370,13 +482,20 @@ def start_test(student_id, level_input, group_input, session_state):
             gr.update(interactive=False, visible=False),
             gr.update(interactive=False, visible=False),
             gr.update(interactive=False, visible=False),
-            session_state
+            session_state,
+            lock_student, lock_group, lock_level, lock_start
         )
 
     if group == 1:
         processed = extract_main_body(text)
+        measured_fre = orig_lev  # ★要件：Group1はpassage_informationのflesch_scoreを記録
     else:
-        processed = rewrite_level(text, level)
+        processed = rewrite_level(text, level, orig_lev)
+        # ★要件：Group2は書き換え後をtextstatで計測して記録
+        try:
+            measured_fre = float(textstat.flesch_reading_ease(processed))
+        except Exception:
+            measured_fre = None
 
     pages = split_pages(processed)
     total = len(pages)
@@ -397,6 +516,7 @@ def start_test(student_id, level_input, group_input, session_state):
         "assigned_level": level,
         "passage_id": pid,
         "original_level": orig_lev,
+        "flesch_score": measured_fre,   # ★追加
         "action_time": now2,
         "action_type": "page_displayed_1",
         "page_text": pages[0]
@@ -422,7 +542,8 @@ def start_test(student_id, level_input, group_input, session_state):
         prev_upd,
         next_upd,
         finish_upd,
-        session_state
+        session_state,
+        lock_student, lock_group, lock_level, lock_start
     )
 
 # ======================================================
@@ -441,6 +562,7 @@ def next_page(pages_json, current_page, total_pages, pid, orig_lev, session_stat
         "assigned_level": level,
         "passage_id": pid,
         "original_level": orig_lev,
+        "flesch_score": "",  # ★列維持
         "action_time": now,
         "action_type": "next_pushed",
         "page_text": None
@@ -463,6 +585,7 @@ def next_page(pages_json, current_page, total_pages, pid, orig_lev, session_stat
         "assigned_level": level,
         "passage_id": pid,
         "original_level": orig_lev,
+        "flesch_score": "",  # ★列維持
         "action_time": now2,
         "action_type": f"page_displayed_{new_page+1}",
         "page_text": pages[new_page]
@@ -503,6 +626,7 @@ def prev_page(pages_json, current_page, total_pages, pid, orig_lev, session_stat
         "assigned_level": level,
         "passage_id": pid,
         "original_level": orig_lev,
+        "flesch_score": "",  # ★列維持
         "action_time": now,
         "action_type": "prev_pushed",
         "page_text": None
@@ -530,6 +654,7 @@ def prev_page(pages_json, current_page, total_pages, pid, orig_lev, session_stat
         "assigned_level": level,
         "passage_id": pid,
         "original_level": orig_lev,
+        "flesch_score": "",  # ★列維持
         "action_time": now2,
         "action_type": f"page_displayed_{new_page+1}",
         "page_text": pages[new_page]
@@ -561,12 +686,18 @@ def finish_or_retire(pages_json, current_page, pid, orig_lev, action, session_st
         "assigned_level": level,
         "passage_id": pid,
         "original_level": orig_lev,
+        "flesch_score": "",  # ★列維持
         "action_time": now,
         "action_type": action,
         "page_text": None
     })
 
-    new_pid, new_text, new_orig_lev, title, used_passages_set = get_new_passage_random(used_passages_set, level)
+    # ★変更：Group1は全教材、Group2はレベル制限あり
+    if group == 1:
+        new_pid, new_text, new_orig_lev, title, used_passages_set = get_new_passage_random_any(used_passages_set)
+    else:
+        new_pid, new_text, new_orig_lev, title, used_passages_set = get_new_passage_random(used_passages_set, level)
+
     if new_text is None:
         return (
             "", "教材がありません", "", json.dumps([]), 0, "",
@@ -579,8 +710,13 @@ def finish_or_retire(pages_json, current_page, pid, orig_lev, action, session_st
 
     if group == 1:
         processed = extract_main_body(new_text)
+        measured_fre = new_orig_lev
     else:
-        processed = rewrite_level(new_text, level)
+        processed = rewrite_level(new_text, level, new_orig_lev)
+        try:
+            measured_fre = float(textstat.flesch_reading_ease(processed))
+        except Exception:
+            measured_fre = None
 
     new_pages = split_pages(processed)
     total = len(new_pages)
@@ -601,6 +737,7 @@ def finish_or_retire(pages_json, current_page, pid, orig_lev, action, session_st
         "assigned_level": level,
         "passage_id": new_pid,
         "original_level": new_orig_lev,
+        "flesch_score": measured_fre,  # ★追加
         "action_time": now2,
         "action_type": "page_displayed_1",
         "page_text": new_pages[0]
@@ -659,7 +796,6 @@ custom_css = """
 .gradio-container select {
     color: inherit !important;
 }
-
 /* ===============================
    ライトモード
    =============================== */
@@ -699,7 +835,6 @@ custom_css = """
         background-color: #eaeaea !important;
     }
 }
-
 /* ===============================
    ダークモード
    =============================== */
@@ -739,7 +874,6 @@ custom_css = """
         color: #e6e6e6 !important;
     }
 }
-
 /* ===============================
    ★Group選択：CSSのみで見やすく（EdgeでもOK）
    =============================== */
@@ -760,7 +894,6 @@ custom_css = """
   align-items: center !important;
   gap: 10px !important;
 }
-
 /* :has が効く環境は行全体ハイライト */
 @media (prefers-color-scheme: light){
   #group_radio label:has(input[type="radio"]:checked){
@@ -822,6 +955,7 @@ with gr.Blocks(css=custom_css) as demo:
 
     retire_btn = gr.Button("リタイア")
 
+    # ★変更：Start後に入力をロックするため、入力コンポーネントもoutputsに追加
     start_btn.click(
         fn=start_test,
         inputs=[student_id_input, level_input, group_input, session_state],
@@ -832,7 +966,8 @@ with gr.Blocks(css=custom_css) as demo:
             hidden_total_pages, hidden_passage_id,
             hidden_orig_lev, hidden_assigned_lev,
             prev_btn, next_btn, finish_btn,
-            session_state
+            session_state,
+            student_id_input, group_input, level_input, start_btn
         ]
     )