Update app.py

app.py

@@ -1,269 +1,178 @@
 import gradio as gr
 from openai import OpenAI
-from datasets import Dataset
-from datetime import datetime, timedelta
+from datasets import load_dataset, Dataset
+from datetime import datetime
 import pandas as pd
-import time, os, random, uuid, tempfile, json
+import time, os, uuid, tempfile
 
-# --- 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/lexile_test_logging"
-LOG_FILE = "logs.csv"
 
-client = OpenAI(base_url=BASE_URL, api_key=API_KEY)
+DATASET_REPO = "Toya0421/reading_exercise_logging"
 
-# --- 外部CSVとして管理する passsage_id / passage データ ---
-# （例）columns: passage_id, lexile_level, text
-passages_df = pd.read_csv("passage.csv")
+client = OpenAI(
+    base_url=BASE_URL,
+    api_key=API_KEY
+)
 
-levels = [300, 600, 850, 1050, 1250]
+# CSV読み込み（lexile_level → level）
+def load_passages():
+    df = pd.read_csv("passage.csv")
+    df["id"] = df["id"].astype(str)
+    df["level"] = df["level"].astype(int)   # ★修正
+    df["passage"] = df["passage"].astype(str)
+    return df
 
-# --- 状態 ---
-used_passages = set()
-question_count = 0
-current_user_id = None
-action_log = []  # ✅ 選択肢変更ログ
+passages_df = load_passages()
 
-# --- AIで問題生成 ---
-def generate_question(text):
-    prompt = f"""
-    Read the following passage and create ONE multiple-choice question with 4 options (A–D).
-    Only output the question and options.
-    Format:
-    Q: <question text>
-    A. <option>
-    B. <option>
-    C. <option>
-    D. <option>
-
-    Passage:
-    {text}
+
+# ---------------------------------
+# ★ levelを決定するロジック（学生番号から自動で決定）
+# ---------------------------------
+def decide_level(student_id: str):
     """
+    学生番号の末尾を使って 1〜5 に自動マッピング
+    """
+    try:
+        last_digit = int(student_id[-1])
+    except:
+        return 1  # 不正な値のときはレベル1
+
+    return last_digit // 2 + 1  # 0-1→1, 2-3→2, ..., 8-9→5
+
+
+# Passage 抽出
+def get_random_passage(level):
+    target_df = passages_df[passages_df["level"] == level]
+    if len(target_df) == 0:
+        return None
+    row = target_df.sample(1).iloc[0]
+    return row["id"], row["passage"]
+
+
+def generate_questions(passage_text):
+    prompt = f"""
+You are an English reading comprehension question generator.
+Create exactly 3 questions for the following passage.
+
+Return JSON in this format:
+{{
+  "q1": "...",
+  "q2": "...",
+  "q3": "..."
+}}
+
+Passage:
+{passage_text}
+"""
     response = client.chat.completions.create(
-        model="google/gemma-3-27b-it:free",
-        messages=[{"role": "user", "content": prompt}],
-        max_tokens=250,
-        temperature=0.7,
+        model="gpt-4o-mini",
+        messages=[{"role": "user", "content": prompt}]
     )
-    return response.choices[0].message.content.strip()
+    return response.choices[0].message["content"]
 
-# --- 正誤判定 ---
-def check_answer_with_ai(text, question, user_answer):
+
+def generate_answers(passage_text, questions_json):
     prompt = f"""
-    Read the passage and question below. Decide if the user's answer is correct.
-    Passage:
-    {text}
-    Question:
-    {question}
-    User Answer: {user_answer}
-    Respond with one word: "Correct" or "Incorrect".
-    """
+You are an AI that answers questions based on the passage.
+
+Return JSON in this format:
+{{
+  "a1": "...",
+  "a2": "...",
+  "a3": "..."
+}}
+
+Passage:
+{passage_text}
+
+Questions:
+{questions_json}
+"""
     response = client.chat.completions.create(
-        model="google/gemma-3-27b-it:free",
-        messages=[{"role": "user", "content": prompt}],
-        max_tokens=10,
-        temperature=0,
+        model="gpt-4o-mini",
+        messages=[{"role": "user", "content": prompt}]
     )
-    return response.choices[0].message.content.strip().lower() == "correct"
-
-# --- 自動難易度調整 ---
-def adaptive_test(prev_level, prev_correct):
-    idx = levels.index(prev_level)
-    if prev_correct and idx < len(levels) - 1:
-        return levels[idx + 1]
-    elif not prev_correct and idx > 0:
-        return levels[idx - 1]
-    return prev_level
-
-# --- passage取得 ---
-def get_passage(level):
-    subset = passages_df[passages_df["lexile_level"] == level]
-    available = [pid for pid in subset["passage_id"] if pid not in used_passages]
-    if not available:
-        available = list(subset["passage_id"])
-    passage_id = random.choice(available)
-    text = subset[subset["passage_id"] == passage_id]["text"].iloc[0]
-    return passage_id, text
-
-# --- ログ追記 & Push ---
-def log_to_csv_and_push(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)
-
-# --- 開始ボタン動作 ---
-def start_test(student_id):
-    global used_passages, question_count, current_user_id, action_log
-    used_passages = set()
-    question_count = 0
-    action_log = []
-
-    if not student_id or student_id.strip() == "":
-        return (
-            "", "", 0, "", "",
-            "⚠️ 学生番号を入力してからテストを開始してください",
-            False, "", "", ""
-        )
+    return response.choices[0].message["content"]
 
-    current_user_id = student_id.strip()
 
-    level = 850
-    passage_id, text = get_passage(level)
-    used_passages.add(passage_id)
+def log_to_huggingface(data):
+    tmp_path = f"/tmp/{uuid.uuid4()}.json"
+    ds = Dataset.from_pandas(pd.DataFrame([data]))
+    ds.to_json(tmp_path)
 
-    question = generate_question(text)
-    displayed_time = datetime.utcnow() + timedelta(hours=9)
+    commit_message = f"Add log {datetime.utcnow().isoformat()}"
 
-    return (
-        text, question, level, passage_id, "",
-        "", True, displayed_time.isoformat(), 1, current_user_id
-    )
+    try:
+        from huggingface_hub import upload_file
 
-# --- 選択肢変更イベント（ここで操作履歴を保存） ---
-def log_choice_change(choice, question_number, user_id):
-    global action_log
-    if choice:
-        action_log.append({
-            "action": "choice",
-            "choice": choice,
-            "time": (datetime.utcnow() + timedelta(hours=9)).isoformat()
-        })
-    return
-
-# --- 回答送信 ---
-def next_step(prev_level, user_answer, question_text, passage_text,
-              displayed_time, question_number, user_id, passage_id):
-
-    global question_count, used_passages, action_log
-
-    if not user_answer:
-        return (
-            "⚠️ Please select an answer!", passage_text, question_text, prev_level,
-            None, "", True, displayed_time, question_number, user_id, passage_id
+        upload_file(
+            path_or_fileobj=tmp_path,
+            path_in_repo=f"logs/{data['log_id']}.json",
+            repo_id=DATASET_REPO,
+            repo_type="dataset",
+            token=HF_TOKEN,
+            commit_message=commit_message
         )
+    except Exception as e:
+        return f"Error uploading: {str(e)}"
 
-    submit_time = datetime.utcnow() + timedelta(hours=9)
-    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に選択履歴がすべて入る)
-    entry = {
-        "user_id": user_id,
-        "question_number": question_count,
-        "lexile_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(action_log, ensure_ascii=False)
-    }
-    log_to_csv_and_push(entry)
-
-    # ✅ 最終問題なら結果だけ大きく表示
-    if question_count >= 5:
-        return (
-            f"<h1>🎯 Your Reading level: <strong>{new_level}L</strong></h1>",
-            "", "", new_level,
-            None, "", False, "", "", user_id, passage_id
-        )
+    return "Log saved."
 
-    # --- 次の問題へ ---
-    next_passage_id, next_text = get_passage(new_level)
-    used_passages.add(next_passage_id)
 
-    next_question = generate_question(next_text)
-    next_display_time = datetime.utcnow() + timedelta(hours=9)
+# ---------------------------------
+# ★ student_id だけ受け取り、内部で level を決定
+# ---------------------------------
+def run_reading_quiz(student_id):
 
-    # ✅ 新しい問題のために選択操作ログをリセット
-    action_log = []
+    # ★ここが最大の変更点
+    level = decide_level(student_id)
 
-    feedback = "✅ Correct!" if correct else "❌ Incorrect."
+    result = get_random_passage(level)
+    if result is None:
+        return "指定されたレベルの文章がありません", "", ""
 
-    return (
-        feedback + "\n➡️ Loading next question…",
-        next_text, next_question, new_level,
-        None, "", True, next_display_time.isoformat(), question_count + 1, user_id, next_passage_id
-    )
+    pid, passage = result
 
-# --- Gradio UI ---
-with gr.Blocks() as demo:
-    gr.Markdown("# 📘 Reading Level Test")
+    questions = generate_questions(passage)
 
-    student_id_input = gr.Textbox(label="Student ID", placeholder="例: B123456")
-    start_btn = gr.Button("▶️ Start Test")
+    answers = generate_answers(passage, questions)
 
-    text_display = gr.Textbox(label="Reading Passage", lines=8, interactive=False)
-    question_display = gr.Textbox(label="Question", lines=6, interactive=False)
-    user_answer = gr.Radio(["A", "B", "C", "D"], label="Your Answer")
-    submit_btn = gr.Button("Submit Answer")
+    log_data = {
+        "log_id": str(uuid.uuid4()),
+        "timestamp": datetime.utcnow().isoformat(),
+        "student_id": student_id,
+        "level": level,      # ★levelをログに保存
+        "passage_id": pid,
+        "passage": passage,
+        "questions": questions,
+        "answers": answers,
+    }
 
-    feedback_display = gr.Markdown()
+    status = log_to_huggingface(log_data)
 
-    hidden_level = gr.Number(visible=False)
-    hidden_passage = gr.Textbox(visible=False)
-    hidden_display_time = gr.Textbox(visible=False)
-    hidden_question_number = gr.Number(visible=False)
-    hidden_user_id = gr.Textbox(visible=False)
-    hidden_passage_id = gr.Textbox(visible=False)
-    test_visible = gr.State(False)
+    return passage, questions, answers
 
-    start_btn.click(
-        fn=start_test,
-        inputs=[student_id_input],
-        outputs=[
-            text_display, question_display, hidden_level, hidden_passage_id, user_answer,
-            feedback_display, test_visible, hidden_display_time,
-            hidden_question_number, hidden_user_id
-        ]
-    )
 
-    # ✅ 選択肢変更ログ
-    user_answer.change(
-        fn=log_choice_change,
-        inputs=[user_answer, hidden_question_number, hidden_user_id],
-        outputs=[]
-    )
+# ---------------------------------
+# ★ UI（level は削除して student_id のみ）
+# ---------------------------------
+with gr.Blocks() as app:
+    gr.Markdown("## 英語リーディング練習")
 
-    # ✅ 回答送信
-    submit_btn.click(
-        fn=next_step,
-        inputs=[hidden_level, user_answer, question_display, text_display,
-                hidden_display_time, hidden_question_number, hidden_user_id, hidden_passage_id],
-        outputs=[
-            feedback_display, text_display, question_display, hidden_level,
-            user_answer, hidden_passage, test_visible, hidden_display_time,
-            hidden_question_number, hidden_user_id, hidden_passage_id
-        ]
-    )
+    student_id_input = gr.Textbox(label="学生番号を入力")
 
-    # ✅ 表示のON/OFF制御
-    def toggle_visibility(show):
-        v = bool(show)
-        return (
-            gr.update(visible=v), gr.update(visible=v),
-            gr.update(visible=v), gr.update(visible=v)
-        )
+    btn = gr.Button("問題を生成")
+
+    passage_out = gr.Textbox(label="Passage", lines=6)
+    question_out = gr.Textbox(label="Questions (JSON)")
+    answer_out = gr.Textbox(label="Answers (JSON)")
 
-    test_visible.change(
-        fn=toggle_visibility,
-        inputs=test_visible,
-        outputs=[text_display, question_display, user_answer, submit_btn]
+    btn.click(
+        run_reading_quiz,
+        inputs=[student_id_input],
+        outputs=[passage_out, question_out, answer_out]
     )
 
-demo.launch()
+app.launch()