Update app.py

app.py

@@ -1,4 +1,3 @@
-# top of the file
 import gradio as gr
 import os, time, re, json, base64, asyncio, threading, uuid, io
 import numpy as np
@@ -36,7 +35,7 @@ class WebSocketClient:
                 await self.websocket.send(f.read())
             await asyncio.gather(self.receive_messages(), self.send_audio_chunks())
         except Exception as e:
-            print(f"\U0001F534 WebSocket Connection Failed: {e}")
+            print(f"🔴 WebSocket Connection Failed: {e}")
 
     def run(self):
         asyncio.set_event_loop(self.loop)
@@ -91,7 +90,7 @@ def clear_transcript(cid):
         connections[cid].transcript = ""
     return ""
 
-# ============ Chat Assistant ============
+# Chat assistant logic
 def handle_chat(user_input, history, thread_id, image_url):
     if not OPENAI_API_KEY or not ASSISTANT_ID:
         return "❌ Missing secrets!", history, thread_id, image_url
@@ -116,7 +115,7 @@ def handle_chat(user_input, history, thread_id, image_url):
                 history.append({"role": "user", "content": user_input})
                 history.append({"role": "assistant", "content": content})
                 match = re.search(
-                    r'https://raw\\.githubusercontent\\.com/AndrewLORTech/surgical-pathology-manual/main/[\\w\\-/]*\\.png',
+                    r'https://raw\.githubusercontent\.com/AndrewLORTech/surgical-pathology-manual/main/[\w\-/]*\.png',
                     content
                 )
                 if match: image_url = match.group(0)
@@ -127,7 +126,17 @@ def handle_chat(user_input, history, thread_id, image_url):
     except Exception as e:
         return f"❌ {e}", history, thread_id, image_url
 
-# ============ Gradio UI ============
+def send_transcript_to_assistant(transcript, history, thread_id, image_url):
+    if not transcript.strip():
+        return gr.update(), history, thread_id, image_url
+    return handle_chat(transcript, history, thread_id, image_url)
+
+def clear_chat_and_transcript(client_id):
+    if client_id in connections:
+        connections[client_id].transcript = ""
+    return [], "", None, None
+
+# UI
 with gr.Blocks(theme=gr.themes.Soft()) as app:
     gr.Markdown("# 📄 Document AI Assistant")
 
@@ -135,10 +144,11 @@ with gr.Blocks(theme=gr.themes.Soft()) as app:
     <style>
     #ask-btn, #clear-chat-btn, #record-btn {
         font-size: 16px !important;
-        padding: 10px 20px !important;
+        padding: 10px 24px !important;
+        margin-top: 6px;
     }
-    #voice-transcript-box textarea {
-        width: 100% !important;
+    .record-button button {
+        display: none !important;
     }
     </style>
     """)
@@ -155,31 +165,20 @@ with gr.Blocks(theme=gr.themes.Soft()) as app:
 
         with gr.Column(scale=2):
             chat = gr.Chatbot(label="💬 Chat", height=460, type="messages")
-
             with gr.Row():
                 user_prompt = gr.Textbox(placeholder="Ask your question...", show_label=False, scale=6)
                 send_btn = gr.Button("Send", variant="primary", scale=2)
 
-            with gr.Accordion("🎤 Voice Transcription", open=True) as voice_section:
-                with gr.Row():
-                    voice_input = gr.Audio(label="", streaming=True, elem_id="record-btn")
-                    voice_transcript = gr.Textbox(label="Transcript", lines=2, interactive=False, elem_id="voice-transcript-box")
+            with gr.Accordion("🎤 Voice Transcription", open=False) as voice_section:
+                voice_input = gr.Audio(label="🎙️ Record", streaming=True, elem_classes="record-button")
+                voice_transcript = gr.Textbox(label="Transcript", lines=2, interactive=False)
+
                 with gr.Row():
-                    record_label = gr.Label("Record", visible=False)
                     ask_btn = gr.Button("🟢 Ask", elem_id="ask-btn")
                     clear_chat_btn = gr.Button("🧹 Clear Chat", elem_id="clear-chat-btn")
+                    record_btn = gr.Button("🎙️ Record", elem_id="record-btn")
 
     # Functional bindings
-    def send_transcript_to_assistant(transcript, history, thread_id, image_url):
-        if not transcript.strip():
-            return gr.update(), history, thread_id, image_url
-        return handle_chat(transcript, history, thread_id, image_url)
-
-    def clear_chat_and_transcript(client_id):
-        if client_id in connections:
-            connections[client_id].transcript = ""
-        return [], "", None, None
-
     send_btn.click(fn=handle_chat,
                    inputs=[user_prompt, chat_state, thread_state, image_state],
                    outputs=[user_prompt, chat, thread_state, image_state])
@@ -193,4 +192,4 @@ with gr.Blocks(theme=gr.themes.Soft()) as app:
                          outputs=[chat, voice_transcript, thread_state, image_state])
     app.load(fn=create_ws, outputs=[client_id])
 
-app.launch()
+app.launch()