app.py

import gradio as gr
import io, json, csv, uuid, random, re
import regex as re2, yake

random.seed(42)

# ====== المعالجة الأساسية ======
AR_STOP = set("في على من إلى عن مع لدى ذلك هذه هذا الذين التي الذي اللواتي اللواتيا أو أم إن أن كان تكون كانوا كانت كنت كنا كانا كانتِ ثم قد لقد ربما بل لكن لكنَّ إلا سوى حتى حيث كما لما لماّ لماَّ لماً ما ماذا لماذا متى أين كيف أي أيّ أيُّ هناك هنا هناكَ تلك ذلكم ذلكن أولئك هؤلاء هما هن هم أنتِ أنتَ أنتما أنتن أنتم أنا نحن هي هو هنَّ همَّ".split())
SENT_SPLIT = re2.compile(r"(?<=[\.!؟\?])\s+")

def clean_text_basic(txt:str)->str:
    txt = txt.replace('\r',' ').replace('\t',' ')
    txt = re.sub(r"\u200f|\u200e|\ufeff"," ",txt)
    txt = re.sub(r"\s+"," ",txt)
    txt = re.sub(r"\s*([\.\!\?؟،,:;؛])\s*", r"\1 ", txt)
    return txt.strip()

def extract_text_pdfminer(data: bytes) -> str:
    try:
        import pdfminer.high_level
        return pdfminer.high_level.extract_text(io.BytesIO(data)) or ""
    except Exception:
        return ""

def extract_text_pypdf(data: bytes) -> str:
    try:
        from pypdf import PdfReader
        out = []
        for p in PdfReader(io.BytesIO(data)).pages:
            out.append(p.extract_text() or "")
        return "\n".join(out)
    except Exception:
        return ""

def extract_text_from_pdf(data: bytes) -> str:
    txt = extract_text_pdfminer(data)
    if not txt or len(txt.strip())<10:
        txt = extract_text_pypdf(data)
    return clean_text_basic(txt)

def split_sentences(text:str):
    return [s for s in [s.strip() for s in SENT_SPLIT.split(text) if s.strip()] if len(s)>=25]

def top_keywords_yake(text:str, max_k=120, lan='ar'):
    kws=[kw for kw,_ in yake.KeywordExtractor(lan=lan, n=1, top=max_k).extract_keywords(text)]
    out,seen=[],set()
    for k in kws:
        k=k.strip()
        if not k or k in seen: continue
        if lan=="ar" and k in AR_STOP: continue
        if len(k)<2: continue
        seen.add(k); out.append(k)
    return out

def build_distractors(correct, pool, k=3):
    cand=[w for w in pool if w!=correct and len(w)>1]
    random.shuffle(cand)
    out=[]
    for w in cand:
        if len(out)==k: break
        w2=w.strip()
        if w2 and w2!=correct.strip(): out.append(w2)
    fillers=["—","-","—-"]
    while len(out)<k: out.append(random.choice(fillers))
    return out

def make_mcqs_from_text(text: str, n=8, lang='ar'):
    text = clean_text_basic(text)
    sents = split_sentences(text)
    if not sents: raise ValueError("النص قصير جدًا.")
    keywords = top_keywords_yake(text, 120, lang)
    if not keywords:
        toks = re.findall(r"[\p{L}\p{N}_]+", text)
        toks = [t for t in toks if not (lang=="ar" and t in AR_STOP)]
        from collections import Counter
        keywords=[w for w,_ in Counter(toks).most_common(80)]
    sent_for_kw={}
    for s in sents:
        for kw in keywords:
            if kw in s and kw not in sent_for_kw:
                sent_for_kw[kw]=s
    items=[]; used=set()
    pool=[kw for kw in keywords if kw in sent_for_kw]
    for kw in pool:
        if len(items)>=n: break
        s=sent_for_kw[kw]
        if s in used: continue
        blanked=s.replace(kw,"_____",1)
        choices=build_distractors(kw,[x for x in keywords if x!=kw],3)+[kw]
        random.shuffle(choices)
        ans=choices.index(kw)
        exp=f"مقتبس من الجملة: {s[:220]}" + ("..." if len(s)>220 else "")
        items.append({
            "id": str(uuid.uuid4())[:8],
            "question": blanked,
            "choices": choices,
            "answer_index": ans,
            "explanation": exp
        })
        used.add(s)
    if not items: raise RuntimeError("تعذر التوليد.")
    return items

def render_cards(items):
    html=[]
    for i,it in enumerate(items,1):
        li="".join(f"<li>{c}</li>" for c in it["choices"])
        ans=["A","B","C","D"][it["answer_index"]]
        html.append(f"""
        <article class="card">
          <header><span class="badge">س {i}</span><h3>{it['question']}</h3></header>
          <ol type="A" class="choices">{li}</ol>
          <details><summary>الإجابة</summary>
          <div class="answer"><b>الإجابة:</b> {ans}</div></details>
        </article>""")
    return "\n".join(html)

def to_files(items):
    json_bytes = io.BytesIO(json.dumps(items, ensure_ascii=False, indent=2).encode("utf-8")); json_bytes.name="mcqs.json"
    s=io.StringIO(); w=csv.writer(s)
    w.writerow(["id","question","A","B","C","D","answer_index","explanation"])
    for it in items:
        ch=it["choices"]
        w.writerow([it["id"], it["question"], *(ch+['']*(4-len(ch))), it["answer_index"], it["explanation"]])
    csv_bytes=io.BytesIO(s.getvalue().encode("utf-8")); csv_bytes.name="mcqs.csv"
    return json_bytes, csv_bytes

def pipeline(text, file, n, lang):
    src = (text or "").strip()
    if file is not None:
        b = file.read()
        name = file.name.lower()
        if name.endswith(".pdf"):
            src = extract_text_from_pdf(b)
        elif name.endswith(".txt"):
            src = clean_text_basic(b.decode("utf-8","ignore"))
        else:
            return "⚠️ ارفعي PDF أو TXT فقط.", None, None
    if not src: return "⚠️ أدخلي نصًا أو ملفًا.", None, None
    items = make_mcqs_from_text(src, n, lang)
    html = render_cards(items)
    j,c = to_files(items)
    return html,j,c

# ====== واجهة Gradio ======
theme = gr.themes.Soft()

try:
    with open("styles.css","r",encoding="utf-8") as f:
        css = f.read()
except FileNotFoundError:
    css = ""

with gr.Blocks(theme=theme, css=css) as demo:
    gr.HTML("<style>body{direction:rtl}</style>")
    gr.Markdown("## 🧠 مولّد أسئلة اختيار من متعدد (PDF / TXT / نص)")
    with gr.Row():
        with gr.Column(scale=1):
            t=gr.Textbox(label="النص",lines=8,placeholder="ألصقي النص هنا أو ارفعي ملف")
            f=gr.File(label="ملف PDF أو TXT",file_types=[".pdf",".txt"])
            n=gr.Slider(1,50,value=10,step=1,label="عدد الأسئلة")
            lang=gr.Dropdown(["ar","en"],value="ar",label="اللغة")
            b=gr.Button("توليد")
            j=gr.File(label="تحميل JSON")
            c=gr.File(label="تحميل CSV")
        with gr.Column(scale=2):
            out=gr.HTML(label="النتائج")
    b.click(pipeline,[t,f,n,lang],[out,j,c])

if __name__=="__main__":
    demo.launch()