Add persistent base-model filter to discovery categories

catalog.py

@@ -51,6 +51,13 @@ def file_role(filename):
         role,label='configuration','Konfiguration'
     return dict(role=role,role_label=label,profile_eligible=role=='model')
 
+def is_derived_model(model):
+    """Use publisher metadata, never infer lineage from a repository name."""
+    card=model.get('cardData') or {}
+    if isinstance(card,dict) and (card.get('base_model') or card.get('base_model_relation')):return True
+    tags=model.get('tags') or []
+    return any(isinstance(tag,str) and (tag.startswith('base_model:') or tag.lower() in {'lora','peft','adapter','merge','mergekit','finetune','fine-tuned','quantized','gguf','gptq','awq'}) for tag in tags)
+
 class Catalog:
     def __init__(self,root):
         self.root=Path(root);self.lock=threading.RLock();self.job=None;self.cancel=threading.Event()
@@ -86,11 +93,16 @@ class Catalog:
         path=self.root/ident/('model'+PurePosixPath(entry['file']).suffix)
         if not path.is_file() or path.stat().st_size!=entry['size']:raise ValueError('Modelldatei fehlt oder ist unvollständig.')
         return entry
-    def search(self,q,kind,sort="downloads"):
+    def search(self,q,kind,sort="downloads",base_only=False):
         if kind not in KINDS or not isinstance(q,str) or len(q)>120:raise ValueError('Ungültige Suche.')
         if sort not in ('downloads','name','newest'):raise ValueError('Ungültige Sortierung.')
-        rows=metadata('/api/models?'+urllib.parse.urlencode(dict(search=q,filter=KINDS[kind],limit=20,sort='createdAt' if sort=='newest' else 'downloads',direction=-1)))
-        return {'models':[dict(repo=x['id'],downloads=x.get('downloads'),gated=x.get('gated',False),created_at=x.get('createdAt')) for x in rows]}
+        if not isinstance(base_only,bool):raise ValueError('Ungültiger Basismodellfilter.')
+        query=dict(search=q,filter=KINDS[kind],limit=100 if base_only else 20,sort='createdAt' if sort=='newest' else 'downloads',direction=-1)
+        if base_only:query['expand']=['downloads','createdAt','gated','cardData','tags']
+        rows=metadata('/api/models?'+urllib.parse.urlencode(query,doseq=True))
+        scanned=len(rows)
+        if base_only:rows=[x for x in rows if not is_derived_model(x)]
+        return {'models':[dict(repo=x['id'],downloads=x.get('downloads'),gated=x.get('gated',False),created_at=x.get('createdAt')) for x in rows[:20]],'base_only':base_only,'scanned':scanned}
     def files(self,repo):
         repo=repo_id(repo)
         with self.cache_lock: