Update app.py

app.py

@@ -108,12 +108,15 @@ def load_models():
         logger.error(f"❌ ASR model error: {e}")
         raise
     
-    # LLM Client
-    logger.info("🤖 Initializing Llama 3.1 client...")
+    # LLM Client - Using Novita AI provider for free Llama 3.1 access
+    logger.info("🤖 Initializing Llama 3.1 client via Novita AI...")
     try:
         if HUGGINGFACE_TOKEN:
-            LLM_CLIENT = InferenceClient(token=HUGGINGFACE_TOKEN)
-            logger.info("✅ LLM client initialized")
+            LLM_CLIENT = InferenceClient(
+                provider="novita",
+                api_key=HUGGINGFACE_TOKEN
+            )
+            logger.info("✅ LLM client initialized with Novita AI provider")
         else:
             logger.warning("⚠️ HF_TOKEN not set - recommendations will use fallback")
     except Exception as e:
@@ -606,25 +609,28 @@ def get_llama_recommendation(emotion_result: dict, retry_count: int = 0) -> str:
     prompt = compose_prompt(emotion_result)
     
     try:
-        logger.info(f"Calling Llama 3.1 (attempt {retry_count + 1})")
+        logger.info(f"Calling Llama 3.1 via Novita AI (attempt {retry_count + 1})")
         
-        # Use text_generation method (more compatible across versions)
-        response = LLM_CLIENT.text_generation(
-            prompt=prompt,
+        # Use chat.completions.create with Novita AI provider
+        completion = LLM_CLIENT.chat.completions.create(
             model=MODEL_NAME,
-            max_new_tokens=300,
+            messages=[
+                {
+                    "role": "user",
+                    "content": prompt
+                }
+            ],
+            max_tokens=300,
             temperature=0.7,
-            top_p=0.9,
-            return_full_text=False,
-            stop_sequences=["\n\n\n"]
+            top_p=0.9
         )
         
-        recommendation = response.strip()
+        recommendation = completion.choices[0].message.content.strip()
         
         if not recommendation:
             raise ValueError("Empty recommendation")
         
-        logger.info("✅ LLM recommendation generated")
+        logger.info("✅ LLM recommendation generated via Novita AI")
         return recommendation
         
     except Exception as e: