QuizGen / src /streamlit_app.py
pluem9
fixminilmTFIDF
3d94eac
Raw History Blame Contribute Delete
16 kB
import os
import sys
import io
import csv
import json
import re
import random
from typing import List, Dict, Optional
import traceback
import streamlit as st
import pandas as pd
from dotenv import load_dotenv
load_dotenv()
st.write("DEBUG ANTHROPIC_MODEL:", os.getenv("ANTHROPIC_MODEL"))
st.write("DEBUG ANTHROPIC_API_KEY exists:", bool(os.getenv("ANTHROPIC_API_KEY")))
LLM_AVAILABLE = False
generate_quiz_llm = None
llm_error = None
try:
from utils.utils_quiz_llm import generate_quiz_llm # type: ignore
LLM_AVAILABLE = True
except Exception as e:
LLM_AVAILABLE = False
llm_error = str(e)
# ========= Paths / Imports =========
BASE_DIR = os.path.dirname(os.path.abspath(__file__)) # .../src
UTILS_DIR = os.path.join(BASE_DIR, "utils") # .../src/utils
if BASE_DIR not in sys.path:
sys.path.append(BASE_DIR)
# ---- Baseline (ของเดิม) ----
from utils.loader import load_to_text
from utils.qg_baseline import (
generate_quiz as generate_quiz_baseline,
to_csv_rows as to_csv_rows_baseline,
)
def mcq_to_baseline_rows(
questions,
q_types=None,
default_type="MCQ",
seed: Optional[int] = None,
):
"""
แปลง list ของคำถามให้กลายเป็นรูปแบบเดียวกับ baseline
(No, Type, Question, OptionA..D, Answer, Explanation)
"""
# 1) เคสที่ engine (เช่น LLM) สร้าง type/question/options/answer มาให้ครบแล้ว
# -> ไม่ต้องดัดรูปแบบคำถาม แค่จัด options ให้ครบ 4 ตัว แล้วส่งเข้า to_csv_rows_baseline
if questions and all(isinstance(q, dict) and "type" in q for q in questions):
quiz = []
for q in questions:
q_type = str(q.get("type", default_type)).upper()
stem = str(q.get("question", ""))
ans = str(q.get("answer", ""))
opts = [
str(o)
for o in q.get("options", [])
if isinstance(o, (str, int, float)) and str(o).strip()
]
# ให้มี answer อยู่ใน options ด้วย
if ans and all(ans.lower() != o.lower() for o in opts):
opts.insert(0, ans)
if not opts and ans:
opts = [ans]
# pad ตัวเลือกให้ครบ 4
while len(opts) < 4:
opts.append("—")
opts = opts[:4]
quiz.append({
"type": q_type,
"question": stem,
"options": opts,
"answer": ans,
"explanation": str(q.get("explanation", "") or ""),
})
return to_csv_rows_baseline(quiz)
# 2) เคสเก่า ๆ (เช่น TF-IDF / MiniLM) ที่ยังให้มาแค่ question/answer/options
# ตรงนี้จะดัดจาก MCQ ตาม q_types (ถ้ายังอยากใช้ logic เดิมก็ใส่ต่อจากตรงนี้ได้)
# ถ้าไม่อยากดัดอะไรเลย สามารถให้กลับไปเป็น MCQ ล้วน ๆ แบบนี้:
quiz = []
for q in questions:
stem = str(q.get("question", ""))
ans = str(q.get("answer", ""))
opts = [
str(o)
for o in q.get("options", [])
if isinstance(o, (str, int, float)) and str(o).strip()
]
if ans and all(ans.lower() != o.lower() for o in opts):
opts.insert(0, ans)
if not opts and ans:
opts = [ans]
while len(opts) < 4:
opts.append("—")
opts = opts[:4]
quiz.append({
"type": default_type,
"question": stem,
"options": opts,
"answer": ans,
"explanation": "",
})
return to_csv_rows_baseline(quiz)
# ---- Smart TF-IDF ----
SMART_AVAILABLE = False
generate_quiz_smart: Optional[callable] = None
smart_error: Optional[str] = None
try:
from utils.utils_quiz_quality import generate_quiz as generate_quiz_smart # type: ignore
SMART_AVAILABLE = True
except Exception as e:
SMART_AVAILABLE = False
smart_error = str(e)
# ---- Smart AI (MiniLM) ----
AI_AVAILABLE = False
generate_quiz_ai: Optional[callable] = None
ai_error: Optional[str] = None
try:
from utils.utils_quiz_embedding import generate_quiz as generate_quiz_ai # type: ignore
AI_AVAILABLE = True
except Exception as e:
AI_AVAILABLE = False
ai_error = f"{e}\n{traceback.format_exc()}"
# ---- LLM (Claude 3) ----
LLM_AVAILABLE = False
generate_quiz_llm = None # type: ignore
llm_error: Optional[str] = None
try:
from utils.utils_quiz_llm import generate_quiz_llm # type: ignore
LLM_AVAILABLE = True
except Exception as e:
LLM_AVAILABLE = False
llm_error = str(e)
# ========= Page Config / CSS =========
st.set_page_config(page_title="Quiz Generation", page_icon="🧠", layout="wide")
st.markdown("""
<style>
.block-container { max-width: 980px; margin: 2rem auto; }
.upload-title { display:flex; gap:12px; align-items:center; margin-bottom:8px; font-weight:700; font-size:22px; }
div[data-testid="stFileUploaderDropzone"]{ border:2px dashed rgba(0,0,0,.25); border-radius:14px; padding:26px;
transition:border-color .2s, background .2s; }
div[data-testid="stFileUploaderDropzone"]:hover{ border-color:#4f8bf9; background:rgba(79,139,249,0.06); }
.hint{ color:#8a8a8a; font-size:13px; margin-top:4px; }
.file-chip{ display:inline-flex; align-items:center; gap:8px; background:rgba(79,139,249,0.12); color:#1f4fbf;
border-radius:999px; padding:6px 12px; font-size:13px; }
.kpi { display:grid; grid-template-columns: repeat(3, minmax(0,1fr)); gap:12px; margin:8px 0 4px 0; }
.kpi .card{ border:1px solid rgba(0,0,0,.08); border-radius:12px; padding:14px 16px; background:var(--secondary-background-color); }
.kpi .title{ font-size:12px; color:#666; }
.kpi .value{ font-size:20px; font-weight:700; margin-top:2px; }
.badge{ display:inline-block; padding:3px 8px; border-radius:999px; font-size:12px; border:1px solid rgba(0,0,0,.15); }
</style>
""", unsafe_allow_html=True)
# ========= Helpers =========
def text_to_slides(txt: str) -> List[str]:
if not txt:
return []
parts = [p.strip() for p in txt.split("\n\n") if p.strip()]
if len(parts) <= 2:
import re
parts = re.split(r"(?:^|\n)(?:Slide|สไลด์)\s*\d*[::-]?", txt, flags=re.IGNORECASE)
parts = [p.strip() for p in parts if p and p.strip()]
return [p if len(p) <= 5000 else p[:5000] for p in parts]
def render_outputs(df: pd.DataFrame, rows: List[Dict], qc_info: Optional[Dict] = None) -> None:
counts = df["Type"].value_counts().to_dict() if not df.empty else {}
k_cloze = counts.get("CLOZE", 0)
k_tf = counts.get("T/F", 0)
k_mcq = counts.get("MCQ", 0)
st.markdown('<div class="kpi">', unsafe_allow_html=True)
st.markdown(f'<div class="card"><div class="title">รวมข้อ</div><div class="value">{len(df)}</div></div>', unsafe_allow_html=True)
st.markdown(f'<div class="card"><div class="title">CLOZE</div><div class="value">{k_cloze}</div></div>', unsafe_allow_html=True)
st.markdown(f'<div class="card"><div class="title">T/F • MCQ</div><div class="value">{k_tf} • {k_mcq}</div></div>', unsafe_allow_html=True)
st.markdown('</div>', unsafe_allow_html=True)
if qc_info is not None:
st.info(f"QC report: {qc_info}")
tab_table, tab_csv, tab_json = st.tabs(["📋 ตารางคำถาม", "🧾 พรีวิว CSV", "🧩 JSON"])
with tab_table:
st.dataframe(df, use_container_width=True)
with tab_csv:
csv_buf = io.StringIO()
df.to_csv(csv_buf, index=False, sep=";", quoting=csv.QUOTE_ALL)
csv_str = csv_buf.getvalue()
st.code(csv_str[:1000] + ("\n... (ตัดท่อนี้ในการแสดงตัวอย่าง)" if len(csv_str) > 1000 else ""), language="csv")
st.download_button("⬇️ ดาวน์โหลด CSV", csv_str.encode("utf-8-sig"),
file_name="quiz_generated.csv", mime="text/csv", use_container_width=True)
with tab_json:
json_str = json.dumps(rows, ensure_ascii=False, indent=2)
st.code(json_str[:2000] + ("\n... (ตัดท่อนี้ในการแสดงตัวอย่าง)" if len(json_str) > 2000 else ""), language="json")
st.download_button("⬇️ ดาวน์โหลด JSON", json_str.encode("utf-8"),
file_name="quiz_generated.json", mime="application/json", use_container_width=True)
# ========= UI =========
st.markdown("<h1 style='margin:0 0 10px 0;'>Quiz Generation</h1>", unsafe_allow_html=True)
st.caption("อัปโหลดสไลด์หรือไฟล์ข้อความ → ตั้งค่า → เลือก Engine → กดปุ่มเพื่อสร้าง Quiz อัตโนมัติ")
st.markdown('<div class="upload-title">📘 Upload File</div>', unsafe_allow_html=True)
uploaded = st.file_uploader("ลากไฟล์มาวาง หรือกดเพื่อเลือกไฟล์",
type=["txt", "pdf", "docx", "csv"],
label_visibility="collapsed", accept_multiple_files=False)
st.markdown('<div class="hint">รองรับ .txt .pdf .docx .csv • ขนาดไม่ควรเกิน ~50MB</div>', unsafe_allow_html=True)
text = ""
if uploaded:
st.markdown(f'<div class="file-chip">📄 {uploaded.name}</div>', unsafe_allow_html=True)
st.divider()
uploaded.seek(0)
text = load_to_text(uploaded)
st.subheader("📄 Preview")
st.text_area("เนื้อหาไฟล์ (ตัวอย่าง)", text[:5000] if text else "(ไฟล์ไม่มีข้อความอ่านได้)",
height=260, label_visibility="collapsed")
if not text:
st.info("ไม่พบข้อความที่อ่านได้จากไฟล์ที่อัปโหลด (อาจเป็นสแกนภาพหรือไฟล์ถูกล็อก).")
st.divider()
st.subheader("⚙️ Quiz Options")
col1, col2, col3 = st.columns([1, 1, 2])
# ---------- ENGINE SELECT ----------
engine_choices = ["Baseline"]
if SMART_AVAILABLE:
engine_choices += ["Smart (TF-IDF)"]
if AI_AVAILABLE:
engine_choices += ["Smart AI (MiniLM)"]
if LLM_AVAILABLE:
engine_choices += ["LLM (Claude 3)"]
engine = col1.selectbox("Engine", engine_choices, index=0, key="engine_select")
n_questions = col1.slider("จำนวนข้อ", 5, 50, 10)
q_types = col2.multiselect("ประเภทคำถาม", ["CLOZE", "T/F", "MCQ"], default=["CLOZE", "T/F", "MCQ"])
seed_value = col2.number_input("Seed (ผลลัพธ์คงที่)", min_value=0, value=42, step=1)
if engine.startswith("Smart"):
level = col3.selectbox("Difficulty", ["easy", "medium", "hard"], index=1)
if "AI" in engine:
st.markdown('<span class="badge">Smart AI (MiniLM) engine: sentence embedding</span>', unsafe_allow_html=True)
else:
st.markdown('<span class="badge">Smart TF-IDF engine: weighted keywords + QC</span>', unsafe_allow_html=True)
else:
level = None
st.markdown('<span class="badge">Baseline engine</span>', unsafe_allow_html=True)
# ---------- WARNINGS ----------
if not SMART_AVAILABLE and smart_error:
st.warning(f"⚠️ Smart (TF-IDF) engine โหลดไม่สำเร็จ: {smart_error}")
if not AI_AVAILABLE and ai_error:
st.warning(f"⚠️ Smart AI (MiniLM) engine โหลดไม่สำเร็จ: {ai_error}")
if not LLM_AVAILABLE and llm_error:
st.warning(f"⚠️ LLM (Claude 3) engine โหลดไม่สำเร็จ: {llm_error}")
generate_btn = st.button("🚀 สร้าง Quiz", use_container_width=True, disabled=not uploaded)
# ========= Generate =========
if generate_btn:
if not uploaded:
st.warning("กรุณาอัปโหลดไฟล์ก่อน")
elif not text:
st.error("ไม่พบข้อความสำหรับสร้างแบบทดสอบ")
else:
slides = text_to_slides(text)
if not slides:
st.error("ไม่สามารถแยกสไลด์จากไฟล์ได้")
else:
if engine == "Smart (TF-IDF)" and SMART_AVAILABLE and generate_quiz_smart:
with st.spinner("กำลังสร้างคำถามด้วย Smart TF-IDF engine…"):
out = generate_quiz_smart(
slides,
n_questions=n_questions,
level=level,
seed=seed_value,
q_types=q_types,
)
rows = mcq_to_baseline_rows(
out.get("questions", []),
q_types=q_types,
default_type="MCQ",
seed=seed_value,
)
df = pd.DataFrame(rows)
render_outputs(df, rows, qc_info=out.get("qc_report", {}))
elif engine == "Smart AI (MiniLM)" and AI_AVAILABLE and generate_quiz_ai:
with st.spinner("กำลังสร้างคำถามด้วย Smart AI (MiniLM) engine…"):
out = generate_quiz_ai(
slides,
n_questions=n_questions,
level=level,
seed=seed_value,
q_types=q_types,
)
rows = mcq_to_baseline_rows(
out.get("questions", []),
q_types=q_types,
default_type="MCQ",
seed=seed_value,
)
df = pd.DataFrame(rows)
render_outputs(df, rows, qc_info=out.get("qc_report", {}))
elif engine == "LLM (Claude 3)" and LLM_AVAILABLE and generate_quiz_llm:
try:
with st.spinner("กำลังสร้างคำถามด้วย LLM (Claude 3)…"):
out = generate_quiz_llm(
slides,
n_questions=n_questions,
level=level,
seed=seed_value,
q_types=q_types,
)
rows = mcq_to_baseline_rows(
out.get("questions", []),
q_types=q_types,
default_type="MCQ",
seed=seed_value,
)
df = pd.DataFrame(rows)
except Exception as e:
st.error(f"LLM (Claude 3) มีปัญหา: {e}")
rows = []
df = pd.DataFrame()
else:
render_outputs(df, rows, qc_info=out.get("qc_report", {}))
else:
with st.spinner("กำลังสร้างคำถามด้วย Baseline engine…"):
quiz = generate_quiz_baseline(
text,
n_questions,
mix=tuple(q_types),
seed=seed_value,
)
rows = to_csv_rows_baseline(quiz)
df = pd.DataFrame(rows)
render_outputs(df, rows)