Update app.py

app.py

@@ -91,8 +91,8 @@ primary_client = InferenceClient("meta-llama/Llama-3.3-70B-Instruct", token=HF_T
 print("Primary model (LLaMA-3.3-70B-Instruct) client initialized.")
 
 # Initialize InferenceClient for fallback model (Gemma-2-9b-it)
-fallback_client = InferenceClient("openai-community/gpt2", token=HF_TOKEN)
-print("Fallback model (gpt2) client initialized.")
+fallback_client = InferenceClient("meta-llama/Llama-4-Scout-17B-16E-Instruct", token=HF_TOKEN)
+print("Fallback model (Llama-4-Scout-17B-16E-Instruct) client initialized.")
 
 
 # Load spacy model for sentence splitting
@@ -1184,7 +1184,7 @@ if __name__ == "__main__":
 
     DESCRIPTION = """
     # LLM with Tools (DuckDuckGo Search, Date Calculation, Business Info RAG, Hugging Face Dataset Logging) and Two-Tier Model System
-    Ask me anything! I can perform web searches, calculate dates, retrieve business information using RAG, and conversation data will be logged to a Hugging Face Dataset. I use a primary LLaMA-70B model for complex queries and a fallback Gemma-2-9b-it model for simpler ones and RAG synthesis.
+    Ask me anything! I can perform web searches, calculate dates, retrieve business information using RAG, and conversation data will be logged to a Hugging Face Dataset. I use a primary LLaMA-70B model for complex queries and a fallback Llama-4-Scout-17B-16E-Instruct model for simpler ones and RAG synthesis.
     """
 
     demo = gr.ChatInterface(