Spaces:
Running
Running
File size: 3,235 Bytes
d2ce5f7 954e0aa d2ce5f7 00c3bab d2ce5f7 00c3bab d2ce5f7 00c3bab d2ce5f7 00c3bab d2ce5f7 00c3bab f412b66 d2ce5f7 f412b66 d2ce5f7 00c3bab d2ce5f7 f412b66 00c3bab | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 | import re
from loguru import logger
from src.rag.rag import get_llm
from src.config import settings
# Load English NSFW word list from config settings (kept out of committed code)
LOCAL_NSFW_WORDS = set(settings.local_nsfw_words)
# Load Arabic NSFW word list from config settings (kept out of committed code)
LOCAL_ARABIC_NSFW_WORDS = set(settings.local_arabic_nsfw_words)
def is_local_nsfw(text: str) -> bool:
text_clean = text.strip().lower()
# Check Arabic NSFW words
words = text_clean.split()
for w in words:
if w in LOCAL_ARABIC_NSFW_WORDS:
return True
# Substring checks for high-signal Arabic NSFW roots
arabic_substrings = {"سكس", "بورن", "شرموط", "منيوك", "قحبة", "قحبه", "متناك"}
for root in arabic_substrings:
if root in text_clean:
return True
# Substring checks for high-signal English NSFW roots
high_signal_substrings = {"porn", "nude", "sex", "vagina", "penis", "clitoris"}
for root in high_signal_substrings:
if root in text_clean:
return True
# Check for direct matches or common word boundary matches for English
for word in LOCAL_NSFW_WORDS:
# If it was already checked as a high-signal substring, skip
if word in high_signal_substrings:
continue
pattern = rf"\b{re.escape(word)}\b"
if re.search(pattern, text_clean):
return True
# Also check obfuscated variations like b**tch, f**k
obfuscated = word[0] + r"\*+" + word[-1] if len(word) > 2 else ""
if obfuscated and re.search(rf"\b{obfuscated}\b", text_clean):
return True
# Check common obfuscated patterns (e.g., f*ck, b*tch, f**k)
# Match any word that contains asterisks inside it
if "*" in text_clean:
# Additional safety check for common swear structures
for word in ["fuck", "bitch", "shit", "cunt", "asshole", "bastard"]:
parts = list(word)
pattern_parts = [parts[0]]
for char in parts[1:-1]:
# Allow the character itself, or one or more asterisks
pattern_parts.append(rf"({re.escape(char)}|\*+)")
pattern_parts.append(parts[-1])
pattern = rf"\b{''.join(pattern_parts)}\b"
if re.search(pattern, text_clean):
return True
return False
def validate_topics_batch(texts: list[str]) -> list[str]:
"""
Validates a batch of topics.
For each topic text:
Checks against the local English and Arabic NSFW lists.
Returns:
A list of results: "ALLOWED" or "not safe for work words".
"""
logger.info(f"Validating batch of {len(texts)} topics...")
results = ["ALLOWED"] * len(texts)
for i, text in enumerate(texts):
if is_local_nsfw(text):
results[i] = "not safe for work words"
return results
def validate_topic_input(topic: str) -> str:
"""
Synchronous validation fallback (e.g. for non-async contexts).
"""
if is_local_nsfw(topic):
return "not safe for work words"
return "ALLOWED"
|