Update app.py

app.py

@@ -1,308 +1,187 @@
 import gradio as gr
-from openai import OpenAI
-from datasets import Dataset
-from datetime import datetime, timedelta
 import pandas as pd
-import time, os, random, uuid, tempfile, json
-
-# --- API設定 ---
-API_KEY = os.getenv("API_KEY")
-BASE_URL = "https://openrouter.ai/api/v1"
-HF_TOKEN = os.getenv("HF_TOKEN")
-DATASET_REPO = "Toya0421/reading_exercise_logging"
-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")
-
-genres = [
-    "Literature","Science＆Technology","History","Social Science＆Society",
-    "Arts＆Culture","Religion＆Philosophy","Lifestyle＆Hobbies",
-    "Health＆Medicine","Education＆Reference"
-]
-
-# --- 状態 ---
-used_passages = set()
-current_user_id = None
-current_genre = None
-current_lexile = None
-action_log = []
-
-def rewrite_to_lexile(text, target_lexile):
-    prompt = f"""
-Rewrite the following passage so it fits about {target_lexile} Lexile.
-- Keep original meaning and length
-- Avoid figurative language
-- Use simple syntax
-- Output only the rewritten passage
-{text}
-"""
-    resp = client.chat.completions.create(
-        model="google/gemma-3-27b-it:free",
-        messages=[{"role": "user", "content": prompt}],
-        temperature=0.4,
-        max_tokens=1000
-    )
-    return resp.choices[0].message.content.strip()
-
-def split_pages(text, words=120):
-    w = text.split()
-    return [" ".join(w[i:i+words]) for i in range(0, len(w), words)]
-
-def get_new_passage(same_genre=True):
-    global used_passages, current_genre
-    df = passages_df[passages_df["genre"] == current_genre] if same_genre else passages_df[passages_df["genre"] != current_genre]
-
-    if len(df) == 0:
-        return None, None, None
-
-    available = [pid for pid in df["passage_id"] if pid not in used_passages]
-
-    if not available:
-        used_passages.clear()
-        available = list(df["passage_id"])
-
-    pid = random.choice(available)
-    row = df[df["passage_id"] == pid].iloc[0]
-    used_passages.add(pid)
-    return pid, row["text"], row["original_lexile_score"]
-
-def save_log(entry):
-    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)
-
-    all_logs = pd.read_csv(LOG_FILE)
-    tmp_dir = tempfile.mkdtemp()
-    tmp_path = os.path.join(tmp_dir, "data.parquet")
-    all_logs.to_parquet(tmp_path)
-    dataset = Dataset.from_parquet(tmp_path)
-    dataset.push_to_hub(DATASET_REPO, token=HF_TOKEN)
-
-# ============================
-#     Gradio 動作関数
-# ============================
-
-def start_test(student_id, genre, lexile):
-    global current_user_id, current_genre, current_lexile, used_passages, action_log
-    used_passages = set()
-    action_log = []
-
-    if not student_id or student_id.strip() == "":
-        return ("", "", "", 0, False, False, "", 0, "", "", 0)
-
-    current_user_id = student_id.strip()
-    current_genre = genre
-    current_lexile = int(lexile)
-
-    pid, text, orig_lex = get_new_passage(same_genre=True)
-    if text is None:
-        return ("教材がありません", "", "", 0, False, False, "", 0, "", "", 0)
-
-    rewritten = rewrite_to_lexile(text, current_lexile)
-    pages = split_pages(rewritten)
-    start_time = (datetime.utcnow() + timedelta(hours=9)).isoformat()
-
-    finish_enabled = (1 == len(pages))
-    prev_enabled = False
-    next_enabled = (len(pages) > 1)
-
-    return (
-        pages[0],
-        f"1 / {len(pages)}",
-        json.dumps(pages, ensure_ascii=False),
-        0,
-        finish_enabled,
-        prev_enabled,
-        next_enabled,
-        start_time,
-        len(pages),
-        pid,
-        orig_lex,
-        current_lexile
-    )
+import time
+import random
+from datasets import load_dataset, Dataset
+
+# --- Load passages CSV ---
+try:
+    passages_df = pd.read_csv("passage.csv")
+except:
+    passages_df = pd.DataFrame(columns=["genre", "passage_id", "page", "text"])
+
+# --- Logging dataset ---
+log_repo = "Toya0421/reading_exercise_logging"
+
+
+def start_test(student_id, selected_genre):
+    if not student_id:
+        return gr.update(value="⚠️ 学籍番号を入力してください", visible=True), gr.update(visible=False)
+
+    # そのジャンルから教材リストを抽出
+    genre_passages = passages_df[passages_df["genre"] == selected_genre]
+
+    if genre_passages.empty:
+        return "⚠️ 選択ジャンルに教材がありません", gr.update(visible=False)
+
+    # 最初の教材を選ぶ
+    passage_id = random.choice(genre_passages["passage_id"].unique())
+    pages = genre_passages[genre_passages["passage_id"] == passage_id].sort_values("page")
+
+    # 初期状態
+    state = {
+        "student_id": student_id,
+        "genre": selected_genre,
+        "passage_id": passage_id,
+        "pages": pages,
+        "page_index": 0,
+        "logs": []
+    }
 
-def next_page(pages_json, current_page, total_pages):
-    pages = json.loads(pages_json)
-
-    if current_page < len(pages) - 1:
-        current_page += 1
-        action_log.append({"action": "next", "time": (datetime.utcnow()+timedelta(hours=9)).isoformat()})
-
-    finish_enabled = (current_page == total_pages - 1)
-    prev_enabled = (current_page > 0)
-    next_enabled = (current_page < total_pages - 1)
-
-    return (
-        pages[current_page],
-        f"{current_page+1} / {total_pages}",
-        json.dumps(pages, ensure_ascii=False),
-        current_page,
-        finish_enabled,
-        prev_enabled,
-        next_enabled
-    )
+    return "", state, pages.iloc[0]["text"], enable_buttons(state)
 
-def prev_page(pages_json, current_page, total_pages):
-    pages = json.loads(pages_json)
-
-    if current_page > 0:
-        current_page -= 1
-        action_log.append({"action": "prev", "time": (datetime.utcnow()+timedelta(hours=9)).isoformat()})
-
-    finish_enabled = (current_page == total_pages - 1)
-    prev_enabled = (current_page > 0)
-    next_enabled = (current_page < total_pages - 1)
-
-    return (
-        pages[current_page],
-        f"{current_page+1} / {total_pages}",
-        json.dumps(pages, ensure_ascii=False),
-        current_page,
-        finish_enabled,
-        prev_enabled,
-        next_enabled
-    )
 
-def finish_or_retire(pages_json, current_page, pid, orig_lex, start_time, action):
-    global action_log
-    pages = json.loads(pages_json)
-    now = (datetime.utcnow()+timedelta(hours=9)).isoformat()
-
-    entry = {
-        "user_id": current_user_id,
-        "genre": current_genre,
-        "lexile_assigned": current_lexile,
-        "passage_id": pid,
-        "original_lexile": orig_lex,
-        "start_time": start_time,
-        "finished_time": now,
-        "actions": json.dumps(action_log, ensure_ascii=False),
-        "result": action
+def enable_buttons(state):
+    last_page = len(state["pages"]) - 1
+    pi = state["page_index"]
+
+    return {
+        "prev": gr.update(interactive=(pi > 0)),
+        "next": gr.update(interactive=(pi < last_page)),
+        "finish": gr.update(interactive=(pi == last_page)),  # ✅ 最終ページのみ押せる
+        "retire": gr.update(interactive=True)
     }
-    save_log(entry)
-
-    if action == "finished":
-        new_pid, new_text, new_orig_lex = get_new_passage(same_genre=True)
-    else:
-        new_pid, new_text, new_orig_lex = get_new_passage(same_genre=False)
-
-    if new_text is None:
-        return ("教材がありません", "", "", 0, False, False, False, "", 0, "", "", current_lexile)
-
-    rewritten = rewrite_to_lexile(new_text, current_lexile)
-    new_pages = split_pages(rewritten)
-    new_start = (datetime.utcnow()+timedelta(hours=9)).isoformat()
-    action_log = []
-
-    finish_enabled = (1 == len(new_pages))
-    prev_enabled = False
-    next_enabled = (len(new_pages) > 1)
-
-    return (
-        new_pages[0],
-        f"1 / {len(new_pages)}",
-        json.dumps(new_pages, ensure_ascii=False),
-        0,
-        finish_enabled,
-        prev_enabled,
-        next_enabled,
-        new_start,
-        len(new_pages),
-        new_pid,
-        new_orig_lex,
-        current_lexile
-    )
 
-# ============================
-#           UI
-# ============================
 
-with gr.Blocks() as demo:
-    gr.Markdown("## 📚 Reading Exercise")
+def next_page(state):
+    if state is None:
+        return None, "", enable_buttons(state)
 
-    student_id_input = gr.Textbox(label="学生番号")
-    lexile_input = gr.Number(label="Lexile（例: 900）")
-    genre_input = gr.Dropdown(choices=genres, label="ジャンル")
+    if state["page_index"] < len(state["pages"]) - 1:
+        state["page_index"] += 1
 
-    start_btn = gr.Button("スタート")
+    text = state["pages"].iloc[state["page_index"]]["text"]
+    return state, text, enable_buttons(state)
+
+
+def prev_page(state):
+    if state is None:
+        return None, "", enable_buttons(state)
+
+    if state["page_index"] > 0:
+        state["page_index"] -= 1
+
+    text = state["pages"].iloc[state["page_index"]]["text"]
+    return state, text, enable_buttons(state)
+
+
+def choose_next_passage(state, finished):
+    """finished = True → 読み終えた、False → リタイア"""
+
+    # ログ追加
+    state["logs"].append({
+        "student_id": state["student_id"],
+        "genre": state["genre"],
+        "passage_id": state["passage_id"],
+        "completed": finished,
+        "timestamp": time.time()
+    })
 
-    text_display = gr.Textbox(label="教材", lines=15, interactive=False)
-    page_display = gr.Textbox(label="進行状況", lines=1, interactive=False)
+    # ログ保存
+    save_logs(state["logs"])
 
-    hidden_pages = gr.Textbox(visible=False)
-    hidden_page_index = gr.Number(visible=False)
-    hidden_start_time = gr.Textbox(visible=False)
-    hidden_total_pages = gr.Number(visible=False)
-    hidden_passage_id = gr.Textbox(visible=False)
-    hidden_orig_lex = gr.Textbox(visible=False)
-    hidden_assigned_lex = gr.Textbox(visible=False)
+    # 次の教材選択
+    genre_passages = passages_df[passages_df["genre"] == state["genre"]]
+    remaining = genre_passages[genre_passages["passage_id"] != state["passage_id"]]
+
+    if remaining.empty:
+        return state, "✅ 同ジャンルの教材は終了しました", disable_all()
+
+    # 次の教材
+    new_passage_id = random.choice(remaining["passage_id"].unique())
+    pages = genre_passages[genre_passages["passage_id"] == new_passage_id].sort_values("page")
+
+    state["passage_id"] = new_passage_id
+    state["pages"] = pages
+    state["page_index"] = 0
+
+    return state, pages.iloc[0]["text"], enable_buttons(state)
+
+
+def disable_all():
+    return {
+        "prev": gr.update(interactive=False),
+        "next": gr.update(interactive=False),
+        "finish": gr.update(interactive=False),
+        "retire": gr.update(interactive=False)
+    }
+
+
+def save_logs(logs):
+    if not logs:
+        return
+    df = pd.DataFrame(logs)
+    dataset = Dataset.from_pandas(df)
+    dataset.push_to_hub(log_repo, private=False)
+
+
+# ==== UI ====
+
+with gr.Blocks() as demo:
+    gr.Markdown("### 📘 リーディングアプリ")
+
+    student_id = gr.Textbox(label="学籍番号を入力", placeholder="例: A12345")
+
+    genre_select = gr.Dropdown(
+        choices=[
+            "Literature", "Science＆Technology", "History",
+            "Social Science＆Society", "Arts＆Culture",
+            "Religion＆Philosophy", "Lifestyle＆Hobbies",
+            "Health＆Medicine", "Education＆Reference"
+        ],
+        label="ジャンルを選択"
+    )
+
+    start_btn = gr.Button("スタート")
+    warn = gr.Textbox(label="警告", visible=False)
 
-    with gr.Row():
-        prev_btn = gr.Button("前へ", interactive=False)
-        next_btn = gr.Button("次へ", interactive=False)
+    passage_state = gr.State()
+    passage_text = gr.Textbox(label="教材", lines=15, interactive=False)
 
-    finish_btn = gr.Button("読み終えた", interactive=False)
+    prev_btn = gr.Button("◀ 前へ")
+    next_btn = gr.Button("次へ ▶")
+    finish_btn = gr.Button("読み終えた")
     retire_btn = gr.Button("リタイア")
 
+    # --- Event binding ---
     start_btn.click(
-        fn=start_test,
-        inputs=[student_id_input, genre_input, lexile_input],
-        outputs=[
-            text_display, page_display,
-            hidden_pages, hidden_page_index,
-            finish_btn, prev_btn, next_btn,
-            hidden_start_time, hidden_total_pages,
-            hidden_passage_id, hidden_orig_lex, hidden_assigned_lex
-        ]
+        start_test,
+        [student_id, genre_select],
+        [warn, passage_state, passage_text, prev_btn, next_btn, finish_btn, retire_btn]
     )
 
     next_btn.click(
-        fn=next_page,
-        inputs=[hidden_pages, hidden_page_index, hidden_total_pages],
-        outputs=[
-            text_display, page_display,
-            hidden_pages, hidden_page_index,
-            finish_btn, prev_btn, next_btn
-        ]
+        next_page,
+        passage_state,
+        [passage_state, passage_text, prev_btn, next_btn, finish_btn, retire_btn]
     )
 
     prev_btn.click(
-        fn=prev_page,
-        inputs=[hidden_pages, hidden_page_index, hidden_total_pages],
-        outputs=[
-            text_display, page_display,
-            hidden_pages, hidden_page_index,
-            finish_btn, prev_btn, next_btn
-        ]
+        prev_page,
+        passage_state,
+        [passage_state, passage_text, prev_btn, next_btn, finish_btn, retire_btn]
     )
 
     finish_btn.click(
-        fn=lambda p, i, pid, o, st: finish_or_retire(p, i, pid, o, st, "finished"),
-        inputs=[hidden_pages, hidden_page_index, hidden_passage_id, hidden_orig_lex, hidden_start_time],
-        outputs=[
-            text_display, page_display,
-            hidden_pages, hidden_page_index,
-            finish_btn, prev_btn, next_btn,
-            hidden_start_time, hidden_total_pages,
-            hidden_passage_id, hidden_orig_lex, hidden_assigned_lex
-        ]
+        lambda s: choose_next_passage(s, True),
+        passage_state,
+        [passage_state, passage_text, prev_btn, next_btn, finish_btn, retire_btn]
     )
 
     retire_btn.click(
-        fn=lambda p, i, pid, o, st: finish_or_retire(p, i, pid, o, st, "retire"),
-        inputs=[hidden_pages, hidden_page_index, hidden_passage_id, hidden_orig_lex, hidden_start_time],
-        outputs=[
-            text_display, page_display,
-            hidden_pages, hidden_page_index,
-            finish_btn, prev_btn, next_btn,
-            hidden_start_time, hidden_total_pages,
-            hidden_passage_id, hidden_orig_lex, hidden_assigned_lex
-        ]
+        lambda s: choose_next_passage(s, False),
+        passage_state,
+        [passage_state, passage_text, prev_btn, next_btn, finish_btn, retire_btn]
     )
 
 demo.launch()