app.py

import os
import gradio as gr
from transformers import pipeline
import pdfplumber
import docx2txt
from docx import Document
from fpdf import FPDF
from langdetect import detect
import urllib.request

# === Настройки ===
os.makedirs("/app/models", exist_ok=True)
FONT_PATH = "DejaVuSans.ttf"

if not os.path.exists(FONT_PATH):
    urllib.request.urlretrieve(
        "https://github.com/dejavu-fonts/dejavu-fonts/raw/master/ttf/DejaVuSans.ttf",
        FONT_PATH
    )

# === Модели ===
def load_model(task, name):
    print(f"🔹 Загружается модель: {name}")
    return pipeline(task, model=name, cache_dir="/app/models")

models = {
    "en": load_model("summarization", "facebook/bart-large-cnn"),
    "ru": load_model("summarization", "IlyaGusev/mbart_ru_sum_gazeta"),
    "kz": load_model("summarization", "csebuetnlp/mT5_multilingual_XLSum"),
}

# === Чтение файлов ===
def read_file(file):
    if not file:
        return ""
    path = file.name.lower()
    text = ""
    try:
        if path.endswith(".pdf"):
            with pdfplumber.open(file.name) as pdf:
                for page in pdf.pages:
                    t = page.extract_text()
                    if t:
                        text += t + "\n"
        elif path.endswith(".docx"):
            text = docx2txt.process(file.name)
        else:
            text = file.read().decode("utf-8", errors="ignore")
    except Exception as e:
        return f"⚠️ Ошибка при чтении: {e}"
    return text.strip()

# === Определение языка ===
def detect_language(text):
    try:
        lang = detect(text)
        if lang.startswith("ru"):
            return "ru"
        elif lang.startswith("kk") or any(c in text for c in "әіңғүұқөһ"):
            return "kz"
        else:
            return "en"
    except:
        return "en"

# === Суммаризация ===
def summarize_text(text):
    if not text or len(text) < 50:
        return "⚠️ Недостаточно текста", "❌", "❌", 0, 0, "❌"

    lang = detect_language(text)
    model = models.get(lang, models["en"])
    flags = {"ru": "🇷🇺 Русский", "kz": "🇰🇿 Қазақ тілі", "en": "🇬🇧 English"}
    lang_label = flags.get(lang, "🌍 Unknown")
    model_label = getattr(model.model, "name_or_path", "Custom")

    chunk_size = 2500
    overlap = 200
    summaries = []

    for i in range(0, len(text), chunk_size - overlap):
        chunk = text[i:i+chunk_size]
        try:
            res = model(chunk, max_length=180, min_length=40, do_sample=False)
            summaries.append(res[0]['summary_text'])
        except Exception as e:
            summaries.append(f"[Ошибка в части {len(summaries)+1}: {e}]")

    summary = "\n\n".join(summaries).strip()
    src_len = len(text)
    sum_len = len(summary)
    comp = f"{round(100*(1 - sum_len/src_len),1)}%" if src_len else "0%"
    return summary, lang_label, model_label, src_len, sum_len, comp

# === Сохранение ===
def save_summary(summary):
    txt, docx, pdf = "summary.txt", "summary.docx", "summary.pdf"
    with open(txt, "w", encoding="utf-8") as f:
        f.write(summary)
    d = Document()
    d.add_heading("Резюме", 0)
    d.add_paragraph(summary)
    d.save(docx)
    pdf_doc = FPDF()
    pdf_doc.add_page()
    pdf_doc.add_font('DejaVu', '', FONT_PATH, uni=True)
    pdf_doc.set_font('DejaVu', '', 12)
    pdf_doc.multi_cell(0, 10, summary)
    pdf_doc.output(pdf)
    return txt, docx, pdf

# === Основная функция ===
def summarize_file(file, progress=gr.Progress()):
    text = read_file(file)
    if text.startswith("⚠️"):
        return text, "❌", "❌", 0, 0, "❌", None, None, None

    progress(0.2, "🧠 Анализ текста...")
    summary, lang, model, src, summ, comp = summarize_text(text)
    progress(0.8, "📄 Сохранение файлов...")
    txt, docx, pdf = save_summary(summary)
    progress(1, "✅ Готово!")
    return summary, lang, model, src, summ, comp, txt, docx, pdf

# === Интерфейс ===
with gr.Blocks() as demo:
    gr.Markdown("<h2 style='text-align:center'>🧠 Eroha Summarizer PRO</h2>")
    file_input = gr.File(label="📂 Загрузите документ (.pdf, .docx, .txt)")
    run_btn = gr.Button("🔍 Сгенерировать резюме", variant="primary")
    summary_out = gr.Textbox(label="🧾 Краткое резюме", lines=10)
    lang_out = gr.Textbox(label="🌍 Определённый язык")
    model_out = gr.Textbox(label="🧠 Используемая модель")
    src_len = gr.Number(label="📄 Длина исходного текста")
    sum_len = gr.Number(label="📝 Длина резюме")
    comp = gr.Textbox(label="📉 Степень сокращения")
    txt = gr.File(label="📄 TXT")
    docx = gr.File(label="📘 DOCX")
    pdf = gr.File(label="📕 PDF")

    run_btn.click(
        summarize_file,
        inputs=file_input,
        outputs=[summary_out, lang_out, model_out, src_len, sum_len, comp, txt, docx, pdf],
    )

demo.launch(server_name="0.0.0.0", server_port=7860)