Update app.py
Browse files
app.py
CHANGED
|
@@ -1,9 +1,9 @@
|
|
| 1 |
-
# app.py
|
| 2 |
import torch
|
| 3 |
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
| 4 |
import time
|
| 5 |
import json
|
| 6 |
import hashlib
|
|
|
|
| 7 |
from datetime import datetime, timedelta
|
| 8 |
import threading
|
| 9 |
from queue import Queue
|
|
@@ -32,6 +32,7 @@ class TranslationResponse(BaseModel):
|
|
| 32 |
processing_time: float
|
| 33 |
character_count: int
|
| 34 |
status: str
|
|
|
|
| 35 |
|
| 36 |
class TranslationCache:
|
| 37 |
def __init__(self, cache_duration_minutes: int = 60):
|
|
@@ -99,6 +100,127 @@ class TranslationQueue:
|
|
| 99 |
thread = threading.Thread(target=worker)
|
| 100 |
thread.start()
|
| 101 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 102 |
class MultilingualTranslator:
|
| 103 |
def __init__(self, cache_duration_minutes: int = 60):
|
| 104 |
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
|
@@ -120,46 +242,148 @@ class MultilingualTranslator:
|
|
| 120 |
except Exception as e:
|
| 121 |
logger.error(f"Error loading model: {e}")
|
| 122 |
raise
|
| 123 |
-
|
| 124 |
-
def translate_text(self, text: str, source_lang: str, target_lang: str) -> Tuple[str, float]:
|
| 125 |
-
"""Translate text from source to target language"""
|
| 126 |
-
start_time = time.time()
|
| 127 |
-
|
| 128 |
-
# Check cache first
|
| 129 |
-
cached_result = self.cache.get(text, source_lang, target_lang)
|
| 130 |
-
if cached_result:
|
| 131 |
-
return cached_result, time.time() - start_time
|
| 132 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 133 |
try:
|
| 134 |
# Set source language for tokenizer
|
| 135 |
self.tokenizer.src_lang = source_lang
|
| 136 |
|
| 137 |
# Encode input
|
| 138 |
-
encoded = self.tokenizer(text, return_tensors="pt").to(self.device)
|
| 139 |
|
| 140 |
-
# Generate translation
|
| 141 |
generated_tokens = self.model.generate(
|
| 142 |
**encoded,
|
| 143 |
forced_bos_token_id=self.tokenizer.get_lang_id(target_lang),
|
| 144 |
-
max_length=
|
| 145 |
-
|
| 146 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 147 |
)
|
| 148 |
|
| 149 |
# Decode result
|
| 150 |
translation = self.tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)[0]
|
| 151 |
|
| 152 |
-
#
|
| 153 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 154 |
|
| 155 |
processing_time = time.time() - start_time
|
| 156 |
-
logger.info(f"
|
| 157 |
|
| 158 |
-
return
|
| 159 |
|
| 160 |
except Exception as e:
|
| 161 |
logger.error(f"Translation error: {e}")
|
| 162 |
-
return f"Translation error: {str(e)}", time.time() - start_time
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 163 |
|
| 164 |
# Language mappings for M2M100 model
|
| 165 |
LANGUAGE_MAP = {
|
|
@@ -236,7 +460,7 @@ LANGUAGE_MAP = {
|
|
| 236 |
translator = MultilingualTranslator(60)
|
| 237 |
|
| 238 |
# Create FastAPI app
|
| 239 |
-
app = FastAPI(title="Multilingual Translation API", version="
|
| 240 |
|
| 241 |
# Add CORS middleware
|
| 242 |
app.add_middleware(
|
|
@@ -249,11 +473,11 @@ app.add_middleware(
|
|
| 249 |
|
| 250 |
@app.get("/")
|
| 251 |
async def root():
|
| 252 |
-
return {"message": "Multilingual Translation API", "status": "active"}
|
| 253 |
|
| 254 |
@app.post("/api/translate")
|
| 255 |
async def api_translate(request: TranslationRequest):
|
| 256 |
-
"""API endpoint for translation"""
|
| 257 |
if not request.text.strip():
|
| 258 |
raise HTTPException(status_code=400, detail="No text provided")
|
| 259 |
|
|
@@ -264,7 +488,7 @@ async def api_translate(request: TranslationRequest):
|
|
| 264 |
raise HTTPException(status_code=400, detail="Invalid language codes")
|
| 265 |
|
| 266 |
try:
|
| 267 |
-
translation, processing_time = translator.translate_text(request.text, source_code, target_code)
|
| 268 |
|
| 269 |
return TranslationResponse(
|
| 270 |
translation=translation,
|
|
@@ -272,7 +496,8 @@ async def api_translate(request: TranslationRequest):
|
|
| 272 |
target_language=request.target_lang,
|
| 273 |
processing_time=processing_time,
|
| 274 |
character_count=len(request.text),
|
| 275 |
-
status="success"
|
|
|
|
| 276 |
)
|
| 277 |
except Exception as e:
|
| 278 |
raise HTTPException(status_code=500, detail=f"Translation error: {str(e)}")
|
|
@@ -280,7 +505,7 @@ async def api_translate(request: TranslationRequest):
|
|
| 280 |
# Alternative endpoint for form data (compatibility with WordPress)
|
| 281 |
@app.post("/api/translate/form")
|
| 282 |
async def api_translate_form(request: Request):
|
| 283 |
-
"""Alternative endpoint that accepts form data"""
|
| 284 |
try:
|
| 285 |
form_data = await request.form()
|
| 286 |
text = form_data.get("text", "")
|
|
@@ -308,7 +533,7 @@ async def api_translate_form(request: Request):
|
|
| 308 |
raise HTTPException(status_code=400, detail="Invalid language codes")
|
| 309 |
|
| 310 |
try:
|
| 311 |
-
translation, processing_time = translator.translate_text(text, source_code, target_code)
|
| 312 |
|
| 313 |
return {
|
| 314 |
"translation": translation,
|
|
@@ -316,7 +541,8 @@ async def api_translate_form(request: Request):
|
|
| 316 |
"target_language": target_lang,
|
| 317 |
"processing_time": processing_time,
|
| 318 |
"character_count": len(text),
|
| 319 |
-
"status": "success"
|
|
|
|
| 320 |
}
|
| 321 |
except Exception as e:
|
| 322 |
raise HTTPException(status_code=500, detail=f"Translation error: {str(e)}")
|
|
@@ -337,7 +563,9 @@ async def health_check():
|
|
| 337 |
"status": "healthy",
|
| 338 |
"device": str(translator.device),
|
| 339 |
"model": translator.model_name,
|
| 340 |
-
"cache_size": len(translator.cache.cache)
|
|
|
|
|
|
|
| 341 |
}
|
| 342 |
|
| 343 |
if __name__ == "__main__":
|
|
|
|
|
|
|
| 1 |
import torch
|
| 2 |
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
| 3 |
import time
|
| 4 |
import json
|
| 5 |
import hashlib
|
| 6 |
+
import re
|
| 7 |
from datetime import datetime, timedelta
|
| 8 |
import threading
|
| 9 |
from queue import Queue
|
|
|
|
| 32 |
processing_time: float
|
| 33 |
character_count: int
|
| 34 |
status: str
|
| 35 |
+
chunks_processed: Optional[int] = None
|
| 36 |
|
| 37 |
class TranslationCache:
|
| 38 |
def __init__(self, cache_duration_minutes: int = 60):
|
|
|
|
| 100 |
thread = threading.Thread(target=worker)
|
| 101 |
thread.start()
|
| 102 |
|
| 103 |
+
class TextChunker:
|
| 104 |
+
"""کلاس برای تقسیم متن طولانی به بخشهای کوچکتر"""
|
| 105 |
+
|
| 106 |
+
@staticmethod
|
| 107 |
+
def split_text_smart(text: str, max_chunk_size: int = 400) -> List[str]:
|
| 108 |
+
"""تقسیم هوشمند متن بر اساس جملات و پاراگرافها"""
|
| 109 |
+
if len(text) <= max_chunk_size:
|
| 110 |
+
return [text]
|
| 111 |
+
|
| 112 |
+
chunks = []
|
| 113 |
+
|
| 114 |
+
# تقسیم بر اساس پاراگرافها
|
| 115 |
+
paragraphs = text.split('\n\n')
|
| 116 |
+
current_chunk = ""
|
| 117 |
+
|
| 118 |
+
for paragraph in paragraphs:
|
| 119 |
+
# اگر پاراگراف خودش بزرگ است، آن را تقسیم کن
|
| 120 |
+
if len(paragraph) > max_chunk_size:
|
| 121 |
+
# ذخیره قسمت فعلی اگر وجود دارد
|
| 122 |
+
if current_chunk.strip():
|
| 123 |
+
chunks.append(current_chunk.strip())
|
| 124 |
+
current_chunk = ""
|
| 125 |
+
|
| 126 |
+
# تقسیم پاراگراف بزرگ
|
| 127 |
+
sub_chunks = TextChunker._split_paragraph(paragraph, max_chunk_size)
|
| 128 |
+
chunks.extend(sub_chunks)
|
| 129 |
+
else:
|
| 130 |
+
# بررسی اینکه آیا اضافه کردن این پاراگراف از حد تجاوز میکند
|
| 131 |
+
if len(current_chunk) + len(paragraph) + 2 > max_chunk_size:
|
| 132 |
+
if current_chunk.strip():
|
| 133 |
+
chunks.append(current_chunk.strip())
|
| 134 |
+
current_chunk = paragraph
|
| 135 |
+
else:
|
| 136 |
+
if current_chunk:
|
| 137 |
+
current_chunk += "\n\n" + paragraph
|
| 138 |
+
else:
|
| 139 |
+
current_chunk = paragraph
|
| 140 |
+
|
| 141 |
+
# اضافه کردن آخرین قسمت
|
| 142 |
+
if current_chunk.strip():
|
| 143 |
+
chunks.append(current_chunk.strip())
|
| 144 |
+
|
| 145 |
+
return chunks
|
| 146 |
+
|
| 147 |
+
@staticmethod
|
| 148 |
+
def _split_paragraph(paragraph: str, max_chunk_size: int) -> List[str]:
|
| 149 |
+
"""تقسیم پاراگراف بزرگ به جملات"""
|
| 150 |
+
# تقسیم بر اساس جملات
|
| 151 |
+
sentences = re.split(r'[.!?]+\s+', paragraph)
|
| 152 |
+
chunks = []
|
| 153 |
+
current_chunk = ""
|
| 154 |
+
|
| 155 |
+
for sentence in sentences:
|
| 156 |
+
if not sentence.strip():
|
| 157 |
+
continue
|
| 158 |
+
|
| 159 |
+
# اضافه کردن علامت نقطه اگر حذف شده
|
| 160 |
+
if not sentence.endswith(('.', '!', '?')):
|
| 161 |
+
sentence += '.'
|
| 162 |
+
|
| 163 |
+
if len(sentence) > max_chunk_size:
|
| 164 |
+
# جمله خودش خیلی بلند است - تقسیم بر اساس کاما
|
| 165 |
+
if current_chunk.strip():
|
| 166 |
+
chunks.append(current_chunk.strip())
|
| 167 |
+
current_chunk = ""
|
| 168 |
+
|
| 169 |
+
sub_chunks = TextChunker._split_by_comma(sentence, max_chunk_size)
|
| 170 |
+
chunks.extend(sub_chunks)
|
| 171 |
+
else:
|
| 172 |
+
if len(current_chunk) + len(sentence) + 1 > max_chunk_size:
|
| 173 |
+
if current_chunk.strip():
|
| 174 |
+
chunks.append(current_chunk.strip())
|
| 175 |
+
current_chunk = sentence
|
| 176 |
+
else:
|
| 177 |
+
if current_chunk:
|
| 178 |
+
current_chunk += " " + sentence
|
| 179 |
+
else:
|
| 180 |
+
current_chunk = sentence
|
| 181 |
+
|
| 182 |
+
if current_chunk.strip():
|
| 183 |
+
chunks.append(current_chunk.strip())
|
| 184 |
+
|
| 185 |
+
return chunks
|
| 186 |
+
|
| 187 |
+
@staticmethod
|
| 188 |
+
def _split_by_comma(sentence: str, max_chunk_size: int) -> List[str]:
|
| 189 |
+
"""تقسیم جمله طولانی بر اساس کاما"""
|
| 190 |
+
parts = sentence.split(', ')
|
| 191 |
+
chunks = []
|
| 192 |
+
current_chunk = ""
|
| 193 |
+
|
| 194 |
+
for part in parts:
|
| 195 |
+
if len(part) > max_chunk_size:
|
| 196 |
+
# قسمت خودش خیلی بلند است - تقسیم اجباری
|
| 197 |
+
if current_chunk.strip():
|
| 198 |
+
chunks.append(current_chunk.strip())
|
| 199 |
+
current_chunk = ""
|
| 200 |
+
|
| 201 |
+
# تقسیم اجباری بر اساس طول
|
| 202 |
+
while len(part) > max_chunk_size:
|
| 203 |
+
chunks.append(part[:max_chunk_size].strip())
|
| 204 |
+
part = part[max_chunk_size:].strip()
|
| 205 |
+
|
| 206 |
+
if part:
|
| 207 |
+
current_chunk = part
|
| 208 |
+
else:
|
| 209 |
+
if len(current_chunk) + len(part) + 2 > max_chunk_size:
|
| 210 |
+
if current_chunk.strip():
|
| 211 |
+
chunks.append(current_chunk.strip())
|
| 212 |
+
current_chunk = part
|
| 213 |
+
else:
|
| 214 |
+
if current_chunk:
|
| 215 |
+
current_chunk += ", " + part
|
| 216 |
+
else:
|
| 217 |
+
current_chunk = part
|
| 218 |
+
|
| 219 |
+
if current_chunk.strip():
|
| 220 |
+
chunks.append(current_chunk.strip())
|
| 221 |
+
|
| 222 |
+
return chunks
|
| 223 |
+
|
| 224 |
class MultilingualTranslator:
|
| 225 |
def __init__(self, cache_duration_minutes: int = 60):
|
| 226 |
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
|
|
|
| 242 |
except Exception as e:
|
| 243 |
logger.error(f"Error loading model: {e}")
|
| 244 |
raise
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 245 |
|
| 246 |
+
# تنظیمات بهینه برای ترجمه متنهای بلند
|
| 247 |
+
self.max_chunk_size = 350 # حداکثر طول هر قسمت
|
| 248 |
+
self.min_chunk_overlap = 20 # همپوشانی بین قسمتها
|
| 249 |
+
|
| 250 |
+
def translate_chunk(self, text: str, source_lang: str, target_lang: str) -> str:
|
| 251 |
+
"""ترجمه یک قسمت کوچک از متن"""
|
| 252 |
try:
|
| 253 |
# Set source language for tokenizer
|
| 254 |
self.tokenizer.src_lang = source_lang
|
| 255 |
|
| 256 |
# Encode input
|
| 257 |
+
encoded = self.tokenizer(text, return_tensors="pt", truncation=True, max_length=512).to(self.device)
|
| 258 |
|
| 259 |
+
# Generate translation with optimized parameters
|
| 260 |
generated_tokens = self.model.generate(
|
| 261 |
**encoded,
|
| 262 |
forced_bos_token_id=self.tokenizer.get_lang_id(target_lang),
|
| 263 |
+
max_length=1024, # افزایش طول خروجی
|
| 264 |
+
min_length=10, # حداقل طول خروجی
|
| 265 |
+
num_beams=5, # افزایش تعداد beam ها برای کیفیت بهتر
|
| 266 |
+
early_stopping=True,
|
| 267 |
+
no_repeat_ngram_size=3, # جلوگیری از تکرار
|
| 268 |
+
length_penalty=1.0, # تنظیم جریمه طول
|
| 269 |
+
repetition_penalty=1.2, # جلوگیری از تکرار کلمات
|
| 270 |
+
do_sample=False, # استفاده از روش قطعی
|
| 271 |
+
temperature=0.7, # کنترل تنوع
|
| 272 |
+
pad_token_id=self.tokenizer.pad_token_id,
|
| 273 |
+
eos_token_id=self.tokenizer.eos_token_id
|
| 274 |
)
|
| 275 |
|
| 276 |
# Decode result
|
| 277 |
translation = self.tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)[0]
|
| 278 |
|
| 279 |
+
# پاکسازی ترجمه از کاراکترهای اضافی
|
| 280 |
+
translation = translation.strip()
|
| 281 |
+
|
| 282 |
+
return translation
|
| 283 |
+
|
| 284 |
+
except Exception as e:
|
| 285 |
+
logger.error(f"Chunk translation error: {e}")
|
| 286 |
+
return f"[Translation Error: {str(e)}]"
|
| 287 |
+
|
| 288 |
+
def translate_text(self, text: str, source_lang: str, target_lang: str) -> Tuple[str, float, int]:
|
| 289 |
+
"""ترجمه متن با پشتیبانی از متنهای طولانی"""
|
| 290 |
+
start_time = time.time()
|
| 291 |
+
|
| 292 |
+
# بررسی کش برای کل متن
|
| 293 |
+
cached_result = self.cache.get(text, source_lang, target_lang)
|
| 294 |
+
if cached_result:
|
| 295 |
+
return cached_result, time.time() - start_time, 1
|
| 296 |
+
|
| 297 |
+
try:
|
| 298 |
+
# اگر متن کوتاه است، مستقیماً ترجمه کن
|
| 299 |
+
if len(text) <= self.max_chunk_size:
|
| 300 |
+
translation = self.translate_chunk(text, source_lang, target_lang)
|
| 301 |
+
|
| 302 |
+
# ذخیره در کش
|
| 303 |
+
self.cache.set(text, source_lang, target_lang, translation)
|
| 304 |
+
processing_time = time.time() - start_time
|
| 305 |
+
logger.info(f"Short text translation completed in {processing_time:.2f} seconds")
|
| 306 |
+
|
| 307 |
+
return translation, processing_time, 1
|
| 308 |
+
|
| 309 |
+
# تقسیم متن طولانی به قسمتهای کوچکتر
|
| 310 |
+
chunks = TextChunker.split_text_smart(text, self.max_chunk_size)
|
| 311 |
+
logger.info(f"Split long text into {len(chunks)} chunks")
|
| 312 |
+
|
| 313 |
+
# ترجمه هر قسمت
|
| 314 |
+
translated_chunks = []
|
| 315 |
+
for i, chunk in enumerate(chunks):
|
| 316 |
+
logger.info(f"Translating chunk {i+1}/{len(chunks)} (length: {len(chunk)})")
|
| 317 |
+
|
| 318 |
+
# بررسی کش برای هر قسمت
|
| 319 |
+
chunk_translation = self.cache.get(chunk, source_lang, target_lang)
|
| 320 |
+
|
| 321 |
+
if not chunk_translation:
|
| 322 |
+
chunk_translation = self.translate_chunk(chunk, source_lang, target_lang)
|
| 323 |
+
# ذخیره قسمت در کش
|
| 324 |
+
self.cache.set(chunk, source_lang, target_lang, chunk_translation)
|
| 325 |
+
|
| 326 |
+
translated_chunks.append(chunk_translation)
|
| 327 |
+
|
| 328 |
+
# کمی استراحت بین ترجمهها برای جلوگیری از بارگذاری زیاد
|
| 329 |
+
if i < len(chunks) - 1:
|
| 330 |
+
time.sleep(0.1)
|
| 331 |
+
|
| 332 |
+
# ترکیب قسمتهای ترجمه شده
|
| 333 |
+
final_translation = self._combine_translations(translated_chunks, text)
|
| 334 |
+
|
| 335 |
+
# ذخیره نتیجه نهایی در کش
|
| 336 |
+
self.cache.set(text, source_lang, target_lang, final_translation)
|
| 337 |
|
| 338 |
processing_time = time.time() - start_time
|
| 339 |
+
logger.info(f"Long text translation completed in {processing_time:.2f} seconds ({len(chunks)} chunks)")
|
| 340 |
|
| 341 |
+
return final_translation, processing_time, len(chunks)
|
| 342 |
|
| 343 |
except Exception as e:
|
| 344 |
logger.error(f"Translation error: {e}")
|
| 345 |
+
return f"Translation error: {str(e)}", time.time() - start_time, 0
|
| 346 |
+
|
| 347 |
+
def _combine_translations(self, translated_chunks: List[str], original_text: str) -> str:
|
| 348 |
+
"""ترکیب قسمتهای ترجمه شده به یک متن یکپارچه"""
|
| 349 |
+
if not translated_chunks:
|
| 350 |
+
return ""
|
| 351 |
+
|
| 352 |
+
if len(translated_chunks) == 1:
|
| 353 |
+
return translated_chunks[0]
|
| 354 |
+
|
| 355 |
+
# ترکیب قسمتها با در نظر گیری ساختار اصلی متن
|
| 356 |
+
combined = []
|
| 357 |
+
|
| 358 |
+
for i, chunk in enumerate(translated_chunks):
|
| 359 |
+
# پاکسازی قسمت
|
| 360 |
+
chunk = chunk.strip()
|
| 361 |
+
|
| 362 |
+
if not chunk:
|
| 363 |
+
continue
|
| 364 |
+
|
| 365 |
+
# اضافه کردن فاصله مناسب بین قسمتها
|
| 366 |
+
if i > 0 and combined:
|
| 367 |
+
# اگر قسمت قبلی با نقطه تمام نمیشود، نقطه اضافه کن
|
| 368 |
+
if not combined[-1].rstrip().endswith(('.', '!', '?', ':', '؛', '.')):
|
| 369 |
+
combined[-1] += '.'
|
| 370 |
+
|
| 371 |
+
# بررسی اینکه آیا نیاز به پاراگراف جدید داریم
|
| 372 |
+
if '\n\n' in original_text:
|
| 373 |
+
combined.append('\n\n' + chunk)
|
| 374 |
+
else:
|
| 375 |
+
combined.append(' ' + chunk)
|
| 376 |
+
else:
|
| 377 |
+
combined.append(chunk)
|
| 378 |
+
|
| 379 |
+
result = ''.join(combined)
|
| 380 |
+
|
| 381 |
+
# پاکسازی نهایی
|
| 382 |
+
result = re.sub(r'\s+', ' ', result) # حذف فاصلههای اضافی
|
| 383 |
+
result = re.sub(r'\.+', '.', result) # حذف نقطههای تکراری
|
| 384 |
+
result = result.strip()
|
| 385 |
+
|
| 386 |
+
return result
|
| 387 |
|
| 388 |
# Language mappings for M2M100 model
|
| 389 |
LANGUAGE_MAP = {
|
|
|
|
| 460 |
translator = MultilingualTranslator(60)
|
| 461 |
|
| 462 |
# Create FastAPI app
|
| 463 |
+
app = FastAPI(title="Multilingual Translation API", version="2.0.0")
|
| 464 |
|
| 465 |
# Add CORS middleware
|
| 466 |
app.add_middleware(
|
|
|
|
| 473 |
|
| 474 |
@app.get("/")
|
| 475 |
async def root():
|
| 476 |
+
return {"message": "Multilingual Translation API v2.0", "status": "active", "features": ["long_text_support", "smart_chunking", "cache_optimization"]}
|
| 477 |
|
| 478 |
@app.post("/api/translate")
|
| 479 |
async def api_translate(request: TranslationRequest):
|
| 480 |
+
"""API endpoint for translation with long text support"""
|
| 481 |
if not request.text.strip():
|
| 482 |
raise HTTPException(status_code=400, detail="No text provided")
|
| 483 |
|
|
|
|
| 488 |
raise HTTPException(status_code=400, detail="Invalid language codes")
|
| 489 |
|
| 490 |
try:
|
| 491 |
+
translation, processing_time, chunks_count = translator.translate_text(request.text, source_code, target_code)
|
| 492 |
|
| 493 |
return TranslationResponse(
|
| 494 |
translation=translation,
|
|
|
|
| 496 |
target_language=request.target_lang,
|
| 497 |
processing_time=processing_time,
|
| 498 |
character_count=len(request.text),
|
| 499 |
+
status="success",
|
| 500 |
+
chunks_processed=chunks_count
|
| 501 |
)
|
| 502 |
except Exception as e:
|
| 503 |
raise HTTPException(status_code=500, detail=f"Translation error: {str(e)}")
|
|
|
|
| 505 |
# Alternative endpoint for form data (compatibility with WordPress)
|
| 506 |
@app.post("/api/translate/form")
|
| 507 |
async def api_translate_form(request: Request):
|
| 508 |
+
"""Alternative endpoint that accepts form data with long text support"""
|
| 509 |
try:
|
| 510 |
form_data = await request.form()
|
| 511 |
text = form_data.get("text", "")
|
|
|
|
| 533 |
raise HTTPException(status_code=400, detail="Invalid language codes")
|
| 534 |
|
| 535 |
try:
|
| 536 |
+
translation, processing_time, chunks_count = translator.translate_text(text, source_code, target_code)
|
| 537 |
|
| 538 |
return {
|
| 539 |
"translation": translation,
|
|
|
|
| 541 |
"target_language": target_lang,
|
| 542 |
"processing_time": processing_time,
|
| 543 |
"character_count": len(text),
|
| 544 |
+
"status": "success",
|
| 545 |
+
"chunks_processed": chunks_count
|
| 546 |
}
|
| 547 |
except Exception as e:
|
| 548 |
raise HTTPException(status_code=500, detail=f"Translation error: {str(e)}")
|
|
|
|
| 563 |
"status": "healthy",
|
| 564 |
"device": str(translator.device),
|
| 565 |
"model": translator.model_name,
|
| 566 |
+
"cache_size": len(translator.cache.cache),
|
| 567 |
+
"max_chunk_size": translator.max_chunk_size,
|
| 568 |
+
"version": "2.0.0"
|
| 569 |
}
|
| 570 |
|
| 571 |
if __name__ == "__main__":
|