session state

app.py

@@ -2,10 +2,8 @@ import spaces
 import gradio as gr
 import torch
 import subprocess
-import numpy as np
 import requests
-
-
+from gradio import State
 
 # Function to start the ochat server
 @spaces.GPU
@@ -45,28 +43,24 @@ def chat_with_ochat(message):
     except requests.RequestException as e:
         return f"Error: {e}"
 
-
-chat_history = []
-
-# Create a Gradio Blocks interface
+# Create a Gradio Blocks interface with session state
 with gr.Blocks(theme=gr.themes.Soft()) as app:
     gr.Markdown("## vLLM OpenChat-3.5 Interface")
     gr.Markdown("Run on your own machine using this command: ```docker run -it -p 7860:7860 --platform=linux/amd64 --gpus all \
-	registry.hf.space/macadeliccc-openchat-3-5-chatbot:latest python app.py```")
+    registry.hf.space/macadeliccc-openchat-3-5-chatbot:latest python app.py```")
 
     with gr.Row():
         input_text = gr.Textbox(label="Your Message", placeholder="Type your message here")
         submit_button = gr.Button("Send")
     output_chat = gr.Chatbot()
 
-    chat_history = []
+    chat_history = State([])  # Session state for chat history
 
-    def update_output(input_message):
-        global chat_history
+    def update_output(input_message, chat_history):
         server_response = chat_with_ochat(input_message)  # Server's response
         chat_history.append((input_message, server_response))
         return chat_history
 
-    submit_button.click(fn=update_output, inputs=[input_text], outputs=[output_chat])
+    submit_button.click(fn=update_output, inputs=[input_text, chat_history], outputs=[output_chat])
 
-app.launch()
+app.launch()