import torch
import gradio as gr
import json
import time
import random
from transformers import pipeline
import pycountry
from datetime import datetime
from pydantic import BaseModel, PydanticUserError, ConfigDict
from pydantic import BaseModel, ConfigDict
import json
class MyModel(BaseModel):
request: 'starlette.requests.Request'
model_config = ConfigDict(arbitrary_types_allowed=True)
from pydantic_core import core_schema
from starlette.requests import Request
def get_pydantic_core_schema(request_type, handler):
return core_schema.any_schema()
Request.__get_pydantic_core_schema__ = get_pydantic_core_schema
# Initialize pipelines with error handling
try:
lang_detector = pipeline("text-classification", model="papluca/xlm-roberta-base-language-detection")
#text_translator = pipeline("translation", model="facebook/nllb-200-distilled-600M", torch_dtype=torch.bfloat16)
text_translator = pipeline("translation", model="facebook/nllb-200-distilled-600M")
#text_translator = pipeline("translation", model="facebook/nllb-200-418M")
print("🚀 AI Translation Hub initialized successfully!")
except Exception as e:
print(f"⚠️ Error initializing models: {e}")
# Extended language support with emojis
LANGUAGES = {
'English': '🇺🇸',
'Spanish': '🇪🇸',
'French': '🇫🇷',
'German': '🇩🇪',
'Italian': '🇮🇹',
'Portuguese': '🇵🇹',
'Russian': '🇷🇺',
'Chinese (Simplified)': '🇨🇳',
'Japanese': '🇯🇵',
'Korean': '🇰🇷',
'Arabic': '🇸🇦',
'Hindi': '🇮🇳',
'Dutch': '🇳🇱',
'Swedish': '🇸🇪',
'Norwegian': '🇳🇴'
}
# Load language data with fallback
try:
with open('language.json', 'r') as file:
language_data = json.load(file)
except FileNotFoundError:
print("⚠️ Language data file not found. Using basic mapping.")
language_data = {'languages': []}
# Translation statistics
translation_stats = {
'total_translations': 0,
'languages_detected': set(),
'session_start': datetime.now()
}
def get_FLORES_code_from_language(language):
"""Enhanced FLORES code lookup with fallback mapping"""
# Remove emoji and extract clean language name
import re
# Remove emoji flags and extra spaces
clean_language = re.sub(r'[🇦-🇿]{2}\s*', '', language).strip()
for entry in language_data.get('languages', []):
if entry['Language'].lower() == clean_language.lower():
return entry['FLORES-200 code']
# Fallback mapping for common languages
fallback_mapping = {
'english': 'eng_Latn',
'spanish': 'spa_Latn',
'french': 'fra_Latn',
'german': 'deu_Latn',
'chinese (simplified)': 'zho_Hans',
'italian': 'ita_Latn',
'portuguese': 'por_Latn',
'russian': 'rus_Cyrl',
'japanese': 'jpn_Jpan',
'korean': 'kor_Hang',
'arabic': 'arb_Arab',
'hindi': 'hin_Deva',
'dutch': 'nld_Latn',
'swedish': 'swe_Latn',
'norwegian': 'nor_Latn'
}
return fallback_mapping.get(clean_language.lower())
def detect_language_confidence(text):
"""Get language detection with confidence score"""
if not text.strip():
return "Unknown", 0.0
try:
result = lang_detector(text)[0]
return result['label'], result['score']
except:
return "Unknown", 0.0
def translate_with_analytics(text, destination_language, show_confidence=True):
"""Enhanced translation with analytics and progress tracking"""
if not text.strip():
return "⚠️ Please enter some text to translate", "", ""
# Update statistics
translation_stats['total_translations'] += 1
# Simulate processing for dramatic effect
yield "🔍 Analyzing text...", "", ""
time.sleep(0.5)
# Detect source language with confidence
detected_lang, confidence = detect_language_confidence(text)
translation_stats['languages_detected'].add(detected_lang)
yield f"🧠 Detected language: {detected_lang.upper()} ({confidence:.1%} confidence)", "", ""
time.sleep(0.3)
# Get language codes
try:
lang = pycountry.languages.get(alpha_2=detected_lang)
src_code = f"{lang.alpha_3}_Latn" if lang else "eng_Latn"
except:
src_code = "eng_Latn"
dest_code = get_FLORES_code_from_language(destination_language)
if not dest_code:
yield f"❌ Unsupported target language: {destination_language}", "", ""
return
yield f"⚡ Translating to {destination_language.split(' ', 1)[-1] if ' ' in destination_language else destination_language}...", "", ""
time.sleep(0.5)
# Handle same language
if src_code == dest_code:
analytics = f"""
📊 **Translation Analytics**
- Source: {detected_lang.upper()} ({confidence:.1%} confidence)
- Target: Same language detected
- Action: No translation needed
- Processing time: <1s
"""
yield "✅ Translation complete!", text, analytics.strip()
return
# Perform translation
try:
start_time = time.time()
# Calculate appropriate max_length based on input length
input_length = len(text)
# Set max_length to be 1.5x input length with a minimum of 512 and maximum of 2048
max_length = max(512, min(2048, int(input_length * 1.5)))
translation = text_translator(
text,
src_lang=src_code,
tgt_lang=dest_code,
max_length=max_length,
do_sample=False, # For more consistent results
num_beams=4 # Better quality translation
)
processing_time = time.time() - start_time
result = translation[0]['translation_text']
# Generate analytics
import re
clean_dest_lang = re.sub(r'[🇦-🇿]{2}\s*', '', destination_language).strip()
analytics = f"""
📊 **Translation Analytics**
- **Source Language**: {detected_lang.upper()} ({confidence:.1%} confidence)
- **Target Language**: {clean_dest_lang}
- **Characters Processed**: {len(text):,}
- **Max Length Used**: {max_length}
- **Processing Time**: {processing_time:.2f}s
- **Session Translations**: {translation_stats['total_translations']}
- **Languages Detected**: {len(translation_stats['languages_detected'])}
"""
yield "✅ Translation complete!", result, analytics.strip()
except Exception as e:
yield f"❌ Translation failed: {str(e)}", "", ""
def clear_all():
"""Reset all fields"""
return "", "", "", ""
# Custom CSS for a modern, sleek interface
custom_css = """
.gradio-container {
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
}
.gr-button {
background: linear-gradient(45deg, #FF6B6B, #4ECDC4);
border: none;
border-radius: 25px;
color: white;
font-weight: bold;
transition: all 0.3s ease;
box-shadow: 0 4px 15px rgba(0,0,0,0.2);
}
.gr-button:hover {
transform: translateY(-2px);
box-shadow: 0 6px 20px rgba(0,0,0,0.3);
}
.gr-textbox {
border-radius: 15px;
border: 2px solid #e0e0e0;
transition: all 0.3s ease;
}
.gr-textbox:focus {
border-color: #667eea;
box-shadow: 0 0 15px rgba(102, 126, 234, 0.3);
}
.gr-dropdown {
border-radius: 15px;
border: 2px solid #e0e0e0;
}
.gr-panel {
background: rgba(255,255,255,0.95);
border-radius: 20px;
backdrop-filter: blur(10px);
box-shadow: 0 8px 32px rgba(0,0,0,0.1);
}
.gr-form {
background: transparent;
}
.gr-box {
border-radius: 15px;
background: rgba(255,255,255,0.9);
}
"""
# Create the interface
with gr.Blocks(css=custom_css, title="🌍 KS Translation Hub") as demo:
gr.HTML("""
🌍 KS Translation Hub
Powered by Advanced Neural Networks • Real-time Language Detection • 15+ Languages
""")
with gr.Row():
with gr.Column(scale=1):
gr.HTML("
📝 Input
")
input_text = gr.Textbox(
label="Enter text to translate",
placeholder="Type or paste your text here... 🖊️",
lines=8,
show_label=False
)
with gr.Row():
target_lang = gr.Dropdown(
choices=[f"{flag} {lang}" for lang, flag in LANGUAGES.items()],
label="🎯 Target Language",
value="🇪🇸 Spanish",
show_label=True
)
with gr.Column(scale=1):
gr.HTML("