Update intent_utils.py

intent_utils.py

@@ -14,16 +14,16 @@ from transformers import (
     AutoConfig,
 )
 from log import log
-from core import INTENT_MODELS
+from core import llm_models
 
 async def detect_intent(text, project_name):
-    project_model = INTENT_MODELS.get(project_name)
-    if not project_model:
+    llm_model_instance = llm_models.get(project_name)
+    if not llm_model_instance or not llm_model_instance.intent_model:
         raise Exception(f"'{project_name}' için intent modeli yüklenmemiş.")
 
-    tokenizer = project_model["tokenizer"]
-    model = project_model["model"]
-    label2id = project_model["label2id"]
+    tokenizer = llm_model_instance.intent_tokenizer
+    model = llm_model_instance.intent_model
+    label2id = llm_model_instance.intent_label2id
 
     inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True)
     outputs = model(**inputs)
@@ -96,17 +96,13 @@ def background_training(project_name, intents, model_id, output_path, confidence
             if accuracy < confidence_threshold or total < 5:
                 log(f"⚠️ Yetersiz performanslı intent: '{intent_name}' — Doğruluk: {accuracy:.2f}, Örnek: {total}")
 
+        # Eğitim sonrası model ve tokenizer'ı diske kaydet
         model.save_pretrained(output_path)
         tokenizer.save_pretrained(output_path)
         with open(os.path.join(output_path, "label2id.json"), "w") as f:
             json.dump(label2id, f)
 
-        INTENT_MODELS[project_name] = {
-            "model": model,
-            "tokenizer": tokenizer,
-            "label2id": label2id,
-        }
-        log(f"✅ Intent eğitimi tamamlandı ve '{project_name}' modeli yüklendi.")
+        log(f"✅ Intent eğitimi tamamlandı ve '{project_name}' için model disk üzerinde hazır.")
 
     except Exception as e:
         log(f"❌ Intent eğitimi hatası: {e}")