From 155e187400e0f7282d2dea5a2191405438011c84 Mon Sep 17 00:00:00 2001
From: david <david@truenas.local>
Date: Mon, 17 Aug 2026 11:12:44 -0400
Subject: [PATCH] fix: Require full name phrase match or verified adult
 database domain to eliminate non-adult product and object photos

---
 server.py | 37 +++++++++++++++++++++++++++----------
 1 file changed, 27 insertions(+), 10 deletions(-)

diff --git a/server.py b/server.py
index 5e41bf2..017f4dc 100644
--- a/server.py
+++ b/server.py
@@ -439,30 +439,47 @@ def is_valid_adult_candidate(img_url, title, desc, performer_name):
     url_lower = (img_url or '').lower()
     text_lower = f"{img_url} {title} {desc}".lower()
 
-    # 1. At least one significant token of the performer's name MUST appear in the URL, title or description
-    if not any(tok in text_lower for tok in name_tokens):
-        return False
-
-    # 2. Blacklist non-person, nature, stock photo, and unrelated domains
+    # 1. Blacklist non-person, ecommerce, stock photo, and unrelated domains
     blacklist_domains = [
         'pxhere.com', 'pixabay.com', 'freepik.com', 'alamy.com', 'wikipedia.org',
         'wikimedia.org', 'infoescola.com', 'ufrgs.br', 'todojujuy.com', 'gettyimages.com',
-        'shutterstock.com', 'tripadvisor.com', 'etsy.com', 'indiamike.com', 'toolstop.co.uk',
-        'homedepot.com', 'youtube.com', 'disney', 'marvel', 'cartoon', 'anime', 'uhrcenter.de'
+        'shutterstock.com', 'tripadvisor.com', 'etsy.com', 'etsystatic.com', 'pinterest.com',
+        'pinimg.com', 'ebay.com', 'amazon.com', 'indiamike.com', 'toolstop.co.uk',
+        'homedepot.com', 'youtube.com', 'disney', 'marvel', 'cartoon', 'anime', 'uhrcenter.de',
+        'swimxwin.com', 'cdiscount.com', 'dailymail.co.uk', 'dreamstime.com'
     ]
     if any(b in url_lower for b in blacklist_domains):
         return False
 
-    # 3. Blacklist non-human junk keywords (frogs, tools, shrines, boats, etc.)
+    # 2. Blacklist non-human junk keywords
     junk_keywords = [
         'sapo', 'frog', 'toad', 'shrine', 'temple', 'power-tool', 'drill', 'tractor',
         'boat', 'yacht', 'car-parts', 'engine', 'railroad', 'railway', 'hardware',
-        'wildlife', 'amphibian', 'reptile', 'jewelry', 'schmuck', 'panties', 'costume'
+        'wildlife', 'amphibian', 'reptile', 'jewelry', 'schmuck', 'panties', 'costume',
+        'tonsils', 'throat', 'golf-course', 'clubhouse', 'drone'
     ]
     if any(j in text_lower for j in junk_keywords):
         return False
 
-    return True
+    # 3. Verified adult biography / media domains
+    adult_domains = [
+        'babepedia.com', 'freeones.com', 'iafd.com', 'boobpedia.com',
+        'thenewsgod.com', 'famousbio.wiki', 'globalzonetoday.com', 'adultdvdempire.com',
+        'stashdb.org', 'theporndb.net', 'wikistarbio.com', 'babecelebs.com',
+        'pornstarbio.com', 'babe.today', 'indexxx.com', 'scrolller.com'
+    ]
+    is_adult_domain = any(ad in url_lower for ad in adult_domains)
+
+    # 4. Strict name token matching:
+    # If from an adult domain, at least 1 significant token must match.
+    # If from a general web domain, ALL tokens (or the full name phrase) must match.
+    if is_adult_domain:
+        return any(tok in text_lower for tok in name_tokens)
+    else:
+        # Full name phrase or all tokens must match
+        if clean_name in text_lower:
+            return True
+        return all(tok in text_lower for tok in name_tokens)
 
 def search_performer_candidate_images(name, aliases=None, limit=6):
     clean_name = clean_performer_search_name(name)