Update app.py

app.py

@@ -26,4 +26,43 @@ def generate(prompt, history=[], temperature=0.2, max_new_tokens=256, top_p=0.95
         temperature = 1e-2
     top_p = float(top_p)
 
-    generate_kwargs
+    generate_kwargs = dict(
+        temperature=temperature,
+        max_new_tokens=max_new_tokens,
+        top_p=top_p,
+        repetition_penalty=repetition_penalty,
+        do_sample=True,
+        seed=42,
+    )
+
+    formatted_prompt = format_prompt(history)
+    formatted_prompt += f"<s>[INST] {prompt} [/INST]"  # Add the current prompt
+
+    stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
+    
+    output = ""
+    for response in stream:
+        output += response.token.text
+
+    # Adding the latest prompts and responses to history
+    history.append((prompt, output))
+
+    return output, history  # Pass the updated history back to the interface
+
+# Setting up the Chat Interface
+mychatbot = gr.Chatbot(
+    avatar_images=["./user.png", "./botm.png"], bubble_full_width=False, show_label=False, show_copy_button=True, likeable=True,
+)
+
+demo = gr.Interface(
+    fn=generate, 
+    inputs=[gr.Textbox(label="User Input", lines=2, placeholder="Enter your message here..."), gr.State()],
+    outputs=[gr.Textbox(label="AI Response", lines=2, placeholder="AI will respond here..."), gr.State()],
+    title="Mixtral 8x7b Chat",
+    theme="default",
+    initial_state=[(first_user_message, "")],  # Initialize the history with the first message
+    allow_screenshot=False,
+    allow_flagging="never"
+)
+
+demo.launch()