Update chat_demo.py

chat_demo.py

@@ -33,7 +33,7 @@ def start_server(command):
 
     return process
 
-server_process = start_server(["./llama.cpp/build/bin/llama-server", "-m" ,"./llama.cpp/build/ERNIE-4.5-0.3B-PT-UD-Q8_K_XL.gguf", "-c", "32000", "--jinja", "--no-mmap", "--port", "5100"])
+server_process = start_server(["./llama.cpp/build/bin/llama-server", "-m" ,"./llama.cpp/build/ERNIE-4.5-0.3B-PT-UD-Q8_K_XL.gguf", "-c", "32000", "--jinja", "--no-mmap", "--port", "5100", "--threads", "2"])
 
 
 cli = OpenAI(api_key="sk-nokey", base_url=BASE_URL)
@@ -93,9 +93,11 @@ with gr.Blocks() as demo:
     #download=gr.DownloadButton(label="Download Conversation", value=None)
     conv_state = gr.State()
     orig_path = gr.State()
+    chatbot = gr.Chatbot(placeholder="Have fun with the AI!")
     chat = gr.ChatInterface(
         openai_call,
         type="messages",
+        chatbot=chatbot,
         additional_inputs=[
             gr.Textbox("You are a helpful AI assistant.", label="System Prompt"),
             gr.Slider(30, 8192, value=2048, label="Max new tokens"),
@@ -104,11 +106,12 @@ with gr.Blocks() as demo:
         title="Edge level LLM Chat demo",
         description="In this demo, you can chat with sub-1B param range LLM - they are small enough to run with reasonable speed on most end user device. **Warning:** Do not input sensitive info - assume everything is public!"
     )
-    download_file = gr.File()
-    download_btn = gr.Button("Export Conversation for Download") \
-        .click(fn=gen_file, inputs=[conv_state], outputs=[download_file, orig_path]) \
-        .success(fn=clean_file, inputs=[orig_path])
-    download_file.download(on_download, None, None)
+    with gr.Accordion("Export Conversations"):
+        download_file = gr.File()
+        download_btn = gr.Button("Export Conversation for Download") \
+            .click(fn=gen_file, inputs=[conv_state], outputs=[download_file, orig_path]) \
+            .success(fn=clean_file, inputs=[orig_path])
+        download_file.download(on_download, None, None)
 
 try:
     demo.queue(max_size=10, api_open=True).launch(server_name='0.0.0.0')