Update app.py

app.py

@@ -3,7 +3,7 @@ from openai import OpenAI
 from datasets import Dataset
 from datetime import datetime, timedelta
 import pandas as pd
-import time, os, random, tempfile, json
+import time, os, random, tempfile, json, glob
 
 # --- API / HF 設定 ---
 API_KEY = os.getenv("API_KEY")
@@ -14,8 +14,8 @@ LOG_FILE = "reading_logs.csv"
 
 client = OpenAI(base_url=BASE_URL, api_key=API_KEY)
 
-# --- passage.csv 読み込み ---
-passages_df = pd.read_csv("passage.csv")
+# --- passage_information.xlsx 読み込み (Text# と flesch_score 使用) ---
+passage_info_df = pd.read_excel("passage_information.xlsx")
 
 # --- 状態変数 ---
 used_passages = set()
@@ -23,19 +23,78 @@ current_user_id = None
 current_level = None
 
 
-def rewrite_level(text, target_level): # 変更点
+# ======================================================
+#  新しい教材管理：passages フォルダからランダム選択
+# ======================================================
 
+def load_passage_file(text_id):
+    """
+    passages/pg{text_id}.txt を読み込み、内容を返す。
+    """
+    path = f"passages/pg{text_id}.txt"
+    if not os.path.exists(path):
+        return None
+    with open(path, "r", encoding="utf-8") as f:
+        return f.read()
 
-    # --- Level → Flesch Reading Ease Score の変換表 ---
+def get_new_passage_random():
+    """
+    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。
+    """
+    global used_passages
+
+    # --- pg*.txt を取得 ---
+    files = glob.glob("passages/pg*.txt")
+    if not files:
+        return None, None, None
+
+    # --- ファイル名から Text# (整数) を抽出 ---
+    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))
+
+    # --- 未使用の ID を優先 ---
+    available = [pid for pid in all_ids if pid not in used_passages]
+    if not available:
+        used_passages.clear()
+        available = list(all_ids)
+
+    # --- ランダムに選択 ---
+    text_id = random.choice(available)
+    used_passages.add(text_id)
+
+    # --- テキスト読み込み ---
+    text = load_passage_file(text_id)
+    if text is None:
+        return None, None, None
+
+    # --- Excel から original_level (flesch_score) を取得 ---
+    row = passage_info_df[passage_info_df["Text#"] == text_id]
+    if len(row) == 0:
+        orig_level = None
+    else:
+        orig_level = row.iloc[0]["flesch_score"]
+
+    return text_id, text, orig_level
+
+
+# ======================================================
+#  Rewrite
+# ======================================================
+
+def rewrite_level(text, target_level):
     level_to_flesch = {
-        1: 90,   # 例: とても易しい
-        2: 70,   # 例: 易しい
-        3: 55,   # 例: 標準
-        4: 40,   # 例: やや難しい
-        5: 25    # 例: 難しい
+        1: 90,
+        2: 70,
+        3: 55,
+        4: 40,
+        5: 25
     }
 
-    # target_level を Flesch Score に変換（デフォルト 55）
     target_flesch = level_to_flesch[int(target_level)]
 
     prompt = f"""
@@ -43,10 +102,9 @@ Rewrite the following passage so it fits about {target_flesch} Flesch Reading Ea
 - Preserve the original meaning faithfully. 
 - Do not add new information or remove essential information. 
 - Output only the rewritten passage. Do not include explanations. 
-
-
 {text}
 """
+
     resp = client.chat.completions.create(
         model="google/gemma-3-27b-it:free",
         messages=[{"role": "user", "content": prompt}],
@@ -56,29 +114,14 @@ Rewrite the following passage so it fits about {target_flesch} Flesch Reading Ea
     return resp.choices[0].message.content.strip()
 
 
-def split_pages(text, words=120): #ページ分割
+def split_pages(text, words=120):
     w = text.split()
-    pages = [" ".join(w[i:i+words]) for i in range(0, len(w), words)]
-    return pages if pages else [text]
+    return [" ".join(w[i:i+words]) for i in range(0, len(w), words)] or [text]
 
 
-def get_new_passage_random():
-    """ジャンル関係なくランダムに取得"""
-    global used_passages
-    df = passages_df
-
-    available = [pid for pid in df["passage_id"].unique() if pid not in used_passages]
-    if not available:
-        used_passages.clear()
-        available = list(df["passage_id"].unique())
-
-    pid = random.choice(available)
-    row = df[df["passage_id"] == pid].iloc[0]
-
-    used_passages.add(pid)
-
-    return pid, row["text"], row.get("original_level", None)
-
+# ======================================================
+#  Save Log
+# ======================================================
 
 def save_log(entry):
     df = pd.DataFrame([entry])
@@ -86,7 +129,7 @@ def save_log(entry):
         df.to_csv(LOG_FILE, mode="a", index=False, header=False)
     else:
         df.to_csv(LOG_FILE, index=False)
-    # push to HF
+
     all_logs = pd.read_csv(LOG_FILE)
     tmp_dir = tempfile.mkdtemp()
     tmp_path = os.path.join(tmp_dir, "data.parquet")
@@ -95,13 +138,12 @@ def save_log(entry):
     dataset.push_to_hub(DATASET_REPO, token=HF_TOKEN)
 
 
-# ============================
-# コールバック
-# ============================
+# ======================================================
+#  Start
+# ======================================================
 
 def start_test(student_id, level_input):
-    """開始：ジャンル選択なし版"""
-    global current_user_id, current_level, used_passages # 変更点
+    global current_user_id, current_level, used_passages
     used_passages = set()
 
     action = "start_pushed"
@@ -109,7 +151,7 @@ def start_test(student_id, level_input):
 
     entry = {
         "user_id": student_id,
-        "assigned_level": current_level, # 変更点
+        "assigned_level": current_level,
         "passage_id": None,
         "original_level": None,
         "action_time": now,
@@ -119,17 +161,16 @@ def start_test(student_id, level_input):
     save_log(entry)
 
     if not student_id or str(student_id).strip() == "":
-        # 初期無効状態
         return (
             "", "", json.dumps([]), 0, "",
             0, "", None, None,
-            gr.update(interactive=False, visible=False),  # prev_btn
-            gr.update(interactive=False, visible=True),   # next_btn (visible but disabled)
-            gr.update(interactive=False, visible=False)   # finish_btn (hidden)
+            gr.update(interactive=False, visible=False),
+            gr.update(interactive=False, visible=True),
+            gr.update(interactive=False, visible=False)
         )
 
     current_user_id = str(student_id).strip()
-    current_level = int(level_input) # 変更点
+    current_level = int(level_input)
 
     pid, text, orig_lev = get_new_passage_random()
     if text is None:
@@ -141,22 +182,22 @@ def start_test(student_id, level_input):
             gr.update(interactive=False, visible=False)
         )
 
-    rewritten = rewrite_level(text, current_level) # 変更点
+    rewritten = rewrite_level(text, current_level)
     pages = split_pages(rewritten)
     total = len(pages)
 
-    # 最初のページ設定
-    prev_upd = gr.update(interactive=False, visible=False)
-    # next_btn は常に visible、ただしページ数1なら非表示して finish を表示する
     if total == 1:
+        prev_upd = gr.update(interactive=False, visible=False)
         next_upd = gr.update(interactive=False, visible=False)
         finish_upd = gr.update(interactive=True, visible=True)
     else:
+        prev_upd = gr.update(interactive=False, visible=False)
         next_upd = gr.update(interactive=True, visible=True)
         finish_upd = gr.update(interactive=False, visible=False)
 
     page_num = 1
     now = (datetime.utcnow() + timedelta(hours=9)).isoformat()
+
     entry = {
         "user_id": current_user_id,
         "assigned_level": current_level,
@@ -167,6 +208,7 @@ def start_test(student_id, level_input):
         "page_text": pages[0]
     }
     save_log(entry)
+
     return (
         pages[0],
         f"1 / {total}",
@@ -182,31 +224,30 @@ def start_test(student_id, level_input):
     )
 
 
+# ======================================================
+#  Next / Prev / Finish（以下は元コードのまま）
+# ======================================================
+
 def next_page(pages_json, current_page, total_pages, pid, orig_lev):
     now = (datetime.utcnow() + timedelta(hours=9)).isoformat()
 
-    action = "next_pushed"
-    
     entry = {
         "user_id": current_user_id,
-        "assigned_level": current_level, # 変更点
+        "assigned_level": current_level,
         "passage_id": pid,
         "original_level": orig_lev,
         "action_time": now,
-        "action_type": action,
+        "action_type": "next_pushed",
         "page_text": None
     }
     save_log(entry)
-    
+
     pages = json.loads(pages_json)
     if not pages:
-        # 安全措置
-        return (
-            "", "", json.dumps([]), 0, 
-            gr.update(interactive=False, visible=False),
-            gr.update(interactive=False, visible=False),
-            gr.update(interactive=False, visible=False)
-        )
+        return ("", "", json.dumps([]), 0,
+                gr.update(interactive=False, visible=False),
+                gr.update(interactive=False, visible=False),
+                gr.update(interactive=False, visible=False))
 
     new_page = min(current_page + 1, total_pages - 1)
 
@@ -222,72 +263,56 @@ def next_page(pages_json, current_page, total_pages, pid, orig_lev):
     }
     save_log(entry2)
 
-    # 最終ページに到達した場合は next ボタンではページ遷移して finish ボタンを表示する仕様にする
-    # ここでは new_page が最後のインデックスであれば prev を有効、next を隠し finish を表示
     if new_page == total_pages - 1:
-        # 記録
-        prev_upd = gr.update(interactive=True, visible=True)
-        next_upd = gr.update(interactive=False, visible=False)
-        finish_upd = gr.update(interactive=True, visible=True)
         return (
             pages[new_page],
             f"{new_page+1} / {total_pages}",
             json.dumps(pages),
             new_page,
-            prev_upd,
-            next_upd,
-            finish_upd
+            gr.update(interactive=True, visible=True),
+            gr.update(interactive=False, visible=False),
+            gr.update(interactive=True, visible=True)
         )
 
-    prev_enabled = (new_page > 0)
-    prev_upd = gr.update(interactive=prev_enabled, visible=prev_enabled)
-    next_upd = gr.update(interactive=True, visible=True)
-    finish_upd = gr.update(interactive=False, visible=False)
     return (
         pages[new_page],
         f"{new_page+1} / {total_pages}",
         json.dumps(pages),
         new_page,
-        prev_upd,
-        next_upd,
-        finish_upd
+        gr.update(interactive=(new_page > 0), visible=(new_page > 0)),
+        gr.update(interactive=True, visible=True),
+        gr.update(interactive=False, visible=False)
     )
 
 
 def prev_page(pages_json, current_page, total_pages, pid, orig_lev):
     now = (datetime.utcnow() + timedelta(hours=9)).isoformat()
-    
-    action = "prev_pushed"
-    
+
     entry = {
         "user_id": current_user_id,
-        "assigned_level": current_level, # 変更点
+        "assigned_level": current_level,
         "passage_id": pid,
         "original_level": orig_lev,
         "action_time": now,
-        "action_type": action,
+        "action_type": "prev_pushed",
         "page_text": None
     }
     save_log(entry)
-    
+
     pages = json.loads(pages_json)
     if not pages:
-        return (
-            "", "", json.dumps([]), 0,
-            gr.update(interactive=False, visible=False),
-            gr.update(interactive=False, visible=False),
-            gr.update(interactive=False, visible=False)
-        )
+        return ("", "", json.dumps([]), 0,
+                gr.update(interactive=False, visible=False),
+                gr.update(interactive=False, visible=False),
+                gr.update(interactive=False, visible=False))
 
     new_page = max(current_page - 1, 0)
 
-
-    # 最終ページから戻った場合は next を再表示して finish を隠す
     prev_upd = gr.update(interactive=(new_page > 0), visible=(new_page > 0))
-    # if we're now on last page (only possible if total_pages==1), show finish; otherwise show next
     next_visible = (new_page < total_pages - 1)
     next_upd = gr.update(interactive=next_visible, visible=next_visible)
     finish_upd = gr.update(interactive=(not next_visible), visible=(not next_visible))
+
     now2 = (datetime.utcnow() + timedelta(hours=9)).isoformat()
     entry2 = {
         "user_id": current_user_id,
@@ -317,7 +342,7 @@ def finish_or_retire(pages_json, current_page, pid, orig_lev, action):
 
     entry = {
         "user_id": current_user_id,
-        "assigned_level": current_level, # 変更点
+        "assigned_level": current_level,
         "passage_id": pid,
         "original_level": orig_lev,
         "action_time": now,
@@ -326,7 +351,6 @@ def finish_or_retire(pages_json, current_page, pid, orig_lev, action):
     }
     save_log(entry)
 
-    # 新しい教材取得
     new_pid, new_text, new_orig_lev = get_new_passage_random()
     if new_text is None:
         return (
@@ -337,20 +361,19 @@ def finish_or_retire(pages_json, current_page, pid, orig_lev, action):
             gr.update(interactive=False, visible=False)
         )
 
-    rewritten = rewrite_level(new_text, current_level) # 変更点
+    rewritten = rewrite_level(new_text, current_level)
     new_pages = split_pages(rewritten)
-    new_start = (datetime.utcnow() + timedelta(hours=9)).isoformat()
     total = len(new_pages)
 
-    # 新教材の最初のページ表示設定
-    prev_upd = gr.update(interactive=False, visible=False)
     if total == 1:
+        prev_upd = gr.update(interactive=False, visible=False)
         next_upd = gr.update(interactive=False, visible=False)
         finish_upd = gr.update(interactive=True, visible=True)
     else:
+        prev_upd = gr.update(interactive=False, visible=False)
         next_upd = gr.update(interactive=True, visible=True)
         finish_upd = gr.update(interactive=False, visible=False)
-    page_num = 1
+
     now2 = (datetime.utcnow() + timedelta(hours=9)).isoformat()
     entry2 = {
         "user_id": current_user_id,
@@ -358,7 +381,7 @@ def finish_or_retire(pages_json, current_page, pid, orig_lev, action):
         "passage_id": new_pid,
         "original_level": new_orig_lev,
         "action_time": now2,
-        "action_type": f"page_displayed_{page_num}",
+        "action_type": "page_displayed_1",
         "page_text": new_pages[0]
     }
     save_log(entry2)
@@ -371,29 +394,32 @@ def finish_or_retire(pages_json, current_page, pid, orig_lev, action):
         total,
         new_pid,
         new_orig_lev,
-        current_level, # 変更点
+        current_level,
         prev_upd,
         next_upd,
         finish_upd
     )
 
 
-# ============================
-# UI
-# ============================
+# ======================================================
+#  UI
+# ======================================================
 
 with gr.Blocks() as demo:
     gr.Markdown("# 📚 Reading Exercise")
 
     student_id_input = gr.Textbox(label="学生番号（必須）")
-    level_input = gr.Dropdown(choices=[1,2,3,4,5], label="あなたの Reading Level（Level Testの結果を選択）", value=3) # 変更点
+    level_input = gr.Dropdown(
+        choices=[1,2,3,4,5],
+        label="あなたの Reading Level（Level Testの結果を選択）",
+        value=3
+    )
 
     start_btn = gr.Button("スタート")
 
     text_display = gr.Textbox(label="教材", lines=18, interactive=False)
     page_display = gr.Textbox(label="進行状況", lines=1, interactive=False)
 
-    # hidden state
     hidden_pages = gr.Textbox(visible=False)
     hidden_page_index = gr.Number(visible=False)
     hidden_total_pages = gr.Number(visible=False)
@@ -404,25 +430,22 @@ with gr.Blocks() as demo:
     with gr.Row():
         prev_btn = gr.Button("◀ 前へ", interactive=False, visible=False)
         next_btn = gr.Button("次へ ▶", interactive=False, visible=False)
-        # finish_btn は UI 上では通常表示しない（最終ページで表示する）
         finish_btn = gr.Button("読み終えた", interactive=False, visible=False)
 
     retire_btn = gr.Button("リタイア")
 
-    # Start
     start_btn.click(
         fn=start_test,
         inputs=[student_id_input, level_input],
         outputs=[
             text_display, page_display,
             hidden_pages, hidden_page_index,
-            hidden_total_pages, hidden_passage_id, 
+            hidden_total_pages, hidden_passage_id,
             hidden_orig_lev, hidden_assigned_lev,
             prev_btn, next_btn, finish_btn
         ]
     )
 
-    # Next
     next_btn.click(
         fn=next_page,
         inputs=[
@@ -437,7 +460,6 @@ with gr.Blocks() as demo:
         ]
     )
 
-    # Prev
     prev_btn.click(
         fn=prev_page,
         inputs=[
@@ -452,7 +474,6 @@ with gr.Blocks() as demo:
         ]
     )
 
-    # Finish (最終ページで表示されるボタン)
     finish_btn.click(
         fn=lambda p, i, pid, o: finish_or_retire(p, i, pid, o, "finished"),
         inputs=[hidden_pages, hidden_page_index, hidden_passage_id, hidden_orig_lev],
@@ -465,11 +486,8 @@ with gr.Blocks() as demo:
         ]
     )
 
-    # Retire
     retire_btn.click(
-        fn=lambda p, i, pid, o: finish_or_retire(
-            p, i, pid, o, "retire"
-        ),
+        fn=lambda p, i, pid, o: finish_or_retire(p, i, pid, o, "retire"),
         inputs=[
             hidden_pages, hidden_page_index,
             hidden_passage_id, hidden_orig_lev
@@ -477,9 +495,9 @@ with gr.Blocks() as demo:
         outputs=[
             text_display, page_display,
             hidden_pages, hidden_page_index,
-            hidden_total_pages,
-            hidden_passage_id, hidden_orig_lev,
-            hidden_assigned_lev, prev_btn, next_btn, finish_btn
+            hidden_total_pages, hidden_passage_id,
+            hidden_orig_lev, hidden_assigned_lev,
+            prev_btn, next_btn, finish_btn
         ]
     )