Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -7,17 +7,21 @@ from docx import Document
|
|
| 7 |
from fpdf import FPDF
|
| 8 |
from langdetect import detect
|
| 9 |
import urllib.request
|
|
|
|
| 10 |
|
| 11 |
-
# ===
|
| 12 |
os.makedirs("/app/models", exist_ok=True)
|
| 13 |
FONT_PATH = "DejaVuSans.ttf"
|
|
|
|
|
|
|
| 14 |
if not os.path.exists(FONT_PATH):
|
|
|
|
| 15 |
urllib.request.urlretrieve(
|
| 16 |
"https://github.com/dejavu-fonts/dejavu-fonts/raw/master/ttf/DejaVuSans.ttf",
|
| 17 |
FONT_PATH
|
| 18 |
)
|
| 19 |
|
| 20 |
-
# === Загрузка моделей ===
|
| 21 |
def load_model(task, model_name):
|
| 22 |
print(f"🔹 Загружается модель: {model_name}")
|
| 23 |
return pipeline(task, model=model_name, cache_dir="/app/models")
|
|
@@ -28,12 +32,13 @@ summarizers = {
|
|
| 28 |
"kz": load_model("summarization", "csebuetnlp/mT5_multilingual_XLSum")
|
| 29 |
}
|
| 30 |
|
| 31 |
-
# ===
|
| 32 |
def read_file(file):
|
| 33 |
if not file:
|
| 34 |
return ""
|
| 35 |
filename = file.name.lower()
|
| 36 |
text = ""
|
|
|
|
| 37 |
try:
|
| 38 |
if filename.endswith(".pdf"):
|
| 39 |
with pdfplumber.open(file.name) as pdf:
|
|
@@ -47,9 +52,10 @@ def read_file(file):
|
|
| 47 |
text = file.read().decode("utf-8", errors="ignore")
|
| 48 |
except Exception as e:
|
| 49 |
return f"⚠️ Ошибка при чтении файла: {e}"
|
|
|
|
| 50 |
return text.strip()
|
| 51 |
|
| 52 |
-
# === Определение языка ===
|
| 53 |
def detect_language(text):
|
| 54 |
try:
|
| 55 |
lang = detect(text)
|
|
@@ -62,17 +68,18 @@ def detect_language(text):
|
|
| 62 |
except:
|
| 63 |
return "en"
|
| 64 |
|
| 65 |
-
# === Суммаризация ===
|
| 66 |
def summarize_text(text):
|
| 67 |
if not text or len(text) < 50:
|
| 68 |
return "⚠️ Недостаточно текста для анализа.", "❌", "❌", 0, 0, "❌"
|
| 69 |
|
| 70 |
lang = detect_language(text)
|
| 71 |
model = summarizers.get(lang, summarizers["en"])
|
|
|
|
| 72 |
flags = {"ru": "🇷🇺 Русский", "kz": "🇰🇿 Қазақ тілі", "en": "🇬🇧 English"}
|
| 73 |
lang_label = flags.get(lang, "🌍 Unknown")
|
| 74 |
-
model_label = model.model.name_or_path if hasattr(model.model, "name_or_path") else "Custom"
|
| 75 |
|
|
|
|
| 76 |
chunk_size = 2500
|
| 77 |
overlap = 200
|
| 78 |
summaries = []
|
|
@@ -92,7 +99,7 @@ def summarize_text(text):
|
|
| 92 |
|
| 93 |
return summary, lang_label, model_label, src_len, sum_len, f"{compression}%"
|
| 94 |
|
| 95 |
-
# === Сохранение
|
| 96 |
def save_summary_as_txt(summary_text):
|
| 97 |
path = "summary.txt"
|
| 98 |
with open(path, "w", encoding="utf-8") as f:
|
|
@@ -117,7 +124,7 @@ def save_summary_as_pdf(summary_text):
|
|
| 117 |
pdf.output(path)
|
| 118 |
return path
|
| 119 |
|
| 120 |
-
# === Главная функция ===
|
| 121 |
def summarize_file(file):
|
| 122 |
text = read_file(file)
|
| 123 |
if text.startswith("⚠️"):
|
|
@@ -134,7 +141,16 @@ def summarize_file(file):
|
|
| 134 |
|
| 135 |
return summary, lang_label, model_label, src_len, sum_len, compression, txt_path, docx_path, pdf_path
|
| 136 |
|
| 137 |
-
# === Интерфейс ===
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 138 |
demo = gr.Interface(
|
| 139 |
fn=summarize_file,
|
| 140 |
inputs=gr.File(label="📂 Загрузите документ (.pdf, .docx, .txt)"),
|
|
@@ -144,14 +160,14 @@ demo = gr.Interface(
|
|
| 144 |
gr.Textbox(label="🧠 Используемая модель"),
|
| 145 |
gr.Number(label="📄 Длина исходного текста"),
|
| 146 |
gr.Number(label="📝 Длина резюме"),
|
| 147 |
-
gr.Textbox(label="📉
|
| 148 |
-
gr.File(label="📄 TXT"),
|
| 149 |
-
gr.File(label="📘 DOCX"),
|
| 150 |
-
gr.File(label="📕 PDF"),
|
| 151 |
],
|
| 152 |
-
title="🧠 Eroha Summarizer PRO (
|
| 153 |
-
description="🚀 Определяет язык (🇷🇺 / 🇰🇿 / 🇬🇧), создаёт краткое резюме и сохраняет в TXT, DOCX, PDF
|
| 154 |
-
theme=
|
| 155 |
allow_flagging="never"
|
| 156 |
)
|
| 157 |
|
|
|
|
| 7 |
from fpdf import FPDF
|
| 8 |
from langdetect import detect
|
| 9 |
import urllib.request
|
| 10 |
+
import gradio.themes as gt
|
| 11 |
|
| 12 |
+
# === 🗂️ Создание папок и установка шрифта ===
|
| 13 |
os.makedirs("/app/models", exist_ok=True)
|
| 14 |
FONT_PATH = "DejaVuSans.ttf"
|
| 15 |
+
|
| 16 |
+
# Если шрифт отсутствует — скачиваем
|
| 17 |
if not os.path.exists(FONT_PATH):
|
| 18 |
+
print("⬇️ Загружаю шрифт DejaVuSans.ttf ...")
|
| 19 |
urllib.request.urlretrieve(
|
| 20 |
"https://github.com/dejavu-fonts/dejavu-fonts/raw/master/ttf/DejaVuSans.ttf",
|
| 21 |
FONT_PATH
|
| 22 |
)
|
| 23 |
|
| 24 |
+
# === ⚙️ Загрузка моделей один раз при запуске ===
|
| 25 |
def load_model(task, model_name):
|
| 26 |
print(f"🔹 Загружается модель: {model_name}")
|
| 27 |
return pipeline(task, model=model_name, cache_dir="/app/models")
|
|
|
|
| 32 |
"kz": load_model("summarization", "csebuetnlp/mT5_multilingual_XLSum")
|
| 33 |
}
|
| 34 |
|
| 35 |
+
# === 📄 Функция чтения текстов из файлов ===
|
| 36 |
def read_file(file):
|
| 37 |
if not file:
|
| 38 |
return ""
|
| 39 |
filename = file.name.lower()
|
| 40 |
text = ""
|
| 41 |
+
|
| 42 |
try:
|
| 43 |
if filename.endswith(".pdf"):
|
| 44 |
with pdfplumber.open(file.name) as pdf:
|
|
|
|
| 52 |
text = file.read().decode("utf-8", errors="ignore")
|
| 53 |
except Exception as e:
|
| 54 |
return f"⚠️ Ошибка при чтении файла: {e}"
|
| 55 |
+
|
| 56 |
return text.strip()
|
| 57 |
|
| 58 |
+
# === 🌐 Определение языка ===
|
| 59 |
def detect_language(text):
|
| 60 |
try:
|
| 61 |
lang = detect(text)
|
|
|
|
| 68 |
except:
|
| 69 |
return "en"
|
| 70 |
|
| 71 |
+
# === 🧠 Суммаризация текста ===
|
| 72 |
def summarize_text(text):
|
| 73 |
if not text or len(text) < 50:
|
| 74 |
return "⚠️ Недостаточно текста для анализа.", "❌", "❌", 0, 0, "❌"
|
| 75 |
|
| 76 |
lang = detect_language(text)
|
| 77 |
model = summarizers.get(lang, summarizers["en"])
|
| 78 |
+
|
| 79 |
flags = {"ru": "🇷🇺 Русский", "kz": "🇰🇿 Қазақ тілі", "en": "🇬🇧 English"}
|
| 80 |
lang_label = flags.get(lang, "🌍 Unknown")
|
|
|
|
| 81 |
|
| 82 |
+
model_label = model.model.name_or_path if hasattr(model.model, "name_or_path") else "Custom"
|
| 83 |
chunk_size = 2500
|
| 84 |
overlap = 200
|
| 85 |
summaries = []
|
|
|
|
| 99 |
|
| 100 |
return summary, lang_label, model_label, src_len, sum_len, f"{compression}%"
|
| 101 |
|
| 102 |
+
# === 💾 Сохранение файлов ===
|
| 103 |
def save_summary_as_txt(summary_text):
|
| 104 |
path = "summary.txt"
|
| 105 |
with open(path, "w", encoding="utf-8") as f:
|
|
|
|
| 124 |
pdf.output(path)
|
| 125 |
return path
|
| 126 |
|
| 127 |
+
# === 🚀 Главная функция обработки ===
|
| 128 |
def summarize_file(file):
|
| 129 |
text = read_file(file)
|
| 130 |
if text.startswith("⚠️"):
|
|
|
|
| 141 |
|
| 142 |
return summary, lang_label, model_label, src_len, sum_len, compression, txt_path, docx_path, pdf_path
|
| 143 |
|
| 144 |
+
# === 🎨 Интерфейс Gradio ===
|
| 145 |
+
custom_theme = gt.Default(
|
| 146 |
+
primary_hue="purple",
|
| 147 |
+
secondary_hue="violet",
|
| 148 |
+
).set(
|
| 149 |
+
body_background_fill="#f8f6ff",
|
| 150 |
+
button_primary_background_fill="linear-gradient(90deg, #7e3ff2, #c084fc)",
|
| 151 |
+
button_primary_background_fill_hover="linear-gradient(90deg, #6b21a8, #9333ea)",
|
| 152 |
+
)
|
| 153 |
+
|
| 154 |
demo = gr.Interface(
|
| 155 |
fn=summarize_file,
|
| 156 |
inputs=gr.File(label="📂 Загрузите документ (.pdf, .docx, .txt)"),
|
|
|
|
| 160 |
gr.Textbox(label="🧠 Используемая модель"),
|
| 161 |
gr.Number(label="📄 Длина исходного текста"),
|
| 162 |
gr.Number(label="📝 Длина резюме"),
|
| 163 |
+
gr.Textbox(label="📉 Степень сокращения"),
|
| 164 |
+
gr.File(label="📄 Скачать TXT"),
|
| 165 |
+
gr.File(label="📘 Скачать DOCX"),
|
| 166 |
+
gr.File(label="📕 Скачать PDF"),
|
| 167 |
],
|
| 168 |
+
title="🧠 Eroha Summarizer PRO (Автономная версия)",
|
| 169 |
+
description="🚀 Определяет язык (🇷🇺 / 🇰🇿 / 🇬🇧), создаёт краткое резюме и сохраняет в TXT, DOCX, PDF с поддержкой кириллицы.",
|
| 170 |
+
theme=custom_theme,
|
| 171 |
allow_flagging="never"
|
| 172 |
)
|
| 173 |
|