Update app.py

app.py

@@ -6,9 +6,9 @@ from functools import lru_cache
 def load_hf_model(model_name):
     # Use the Gradio-built huggingface loader instead of transformers_gradio
     return gr.load(
-        name=f"deepseek-ai/{model_name}",
+        name=f"huggingface/deepseek-ai/{model_name}",
         src="huggingface",  # Changed from transformers_gradio.registry
-        api_name="/chat"
+        api_name="chat",
     )
 
 # Load all models at startup
@@ -21,32 +21,31 @@ MODELS = {
 # --- Chatbot function ---
 def chatbot(input_text, history, model_choice, system_message, max_new_tokens, temperature, top_p):
     history = history or []
-    
+
     # Get the selected model component
     model_component = MODELS[model_choice]
-    
+
     # Create payload for the model
-    payload = {
-        "messages": [{"role": "user", "content": input_text}],
-        "system": system_message,
-        "max_tokens": max_new_tokens,
-        "temperature": temperature,
-        "top_p": top_p
-    }
-    
+    payload = [
+        history,  # Pass the entire history 
+        input_text,
+        system_message,
+        max_new_tokens,
+        temperature,
+        top_p
+    ]
+
     # Run inference using the selected model
     try:
-        response = model_component(payload)  # The response is likely a dictionary
-        if isinstance(response, dict) and "choices" in response:
-            assistant_response = response["choices"][0]["message"]["content"]
-        else:
-            assistant_response = "Unexpected model response format."
+        response = model_component(payload) # the response now it is a tuple containing the updated history as the first element and the generated text as the second
+        updated_history, assistant_response = response[0], response[1]
+
+        history = updated_history
+
     except Exception as e:
         assistant_response = f"Error: {str(e)}"
-    
-    # Append user and assistant messages to history
-    history.append((input_text, assistant_response))
-    
+        history.append((input_text, assistant_response))
+
     return history, history, ""
 
 # --- Gradio Interface ---