model and ui update

medbot/config.py

@@ -1,3 +1,2 @@
-ME_LLAMA_MODEL = "clinicalnlplab/me-llama",
-
+ME_LLAMA_MODEL = "clinicalnlplab/me-llama"
 FALLBACK_MODEL = "meta-llama/Llama-2-7b-chat-hf" 

medbot/interface.py

@@ -1,10 +1,24 @@
 import gradio as gr
 from .handlers import respond, reset_chat
 
+def user_message_handler(user_message, chat_history):
+    # Immediately show the user's message
+    chat_history = chat_history + [(user_message, None)]
+    return "", chat_history
+
+def bot_response_handler(chat_history):
+    # Get the last user message
+    user_message = chat_history[-1][0]
+    # Get the bot's response
+    _, updated_history = respond(user_message, chat_history[:-1])
+    # Replace the last (user, None) with (user, bot_response)
+    chat_history[-1] = updated_history[-1]
+    return chat_history
+
 def build_interface():
     with gr.Blocks(theme="soft") as demo:
         gr.Markdown("# 🏥 Dr Bavana's Medical Assistant ")
-        gr.Markdown("By Blue Planet Infosolutions Pvt Ltd")
+        gr.Markdown("By Techindia Solutions AI")
         with gr.Row():
             with gr.Column(scale=4):
                 chatbot = gr.Chatbot(height=500)
@@ -24,7 +38,12 @@ def build_interface():
             ],
             inputs=msg
         )
-        msg.submit(respond, [msg, chatbot], [msg, chatbot])
-        send_btn.click(respond, [msg, chatbot], [msg, chatbot])
+        # Step 1: Show user message instantly
+        msg.submit(user_message_handler, [msg, chatbot], [msg, chatbot]).then(
+            bot_response_handler, chatbot, chatbot
+        )
+        send_btn.click(user_message_handler, [msg, chatbot], [msg, chatbot]).then(
+            bot_response_handler, chatbot, chatbot
+        )
         reset_btn.click(reset_chat, [], [chatbot, msg])
     return demo 

medbot/model.py

@@ -28,7 +28,7 @@ class ModelManager:
                 device_map="auto"
             )
 
-    def generate(self, prompt, max_new_tokens=400, temperature=0.5, top_p=0.9):
+    def generate(self, prompt, max_new_tokens=800, temperature=0.5, top_p=0.999):
         self.load()
         inputs = self.tokenizer(prompt, return_tensors="pt")
         if torch.cuda.is_available():