Update app.py

app.py

@@ -617,6 +617,20 @@ AVATAR_HTML = """
 </div>
 """
 
+def _audio_html(path: str) -> str:
+    """Convert audio file path to autoplay HTML data URI."""
+    if not path:
+        return "<span></span>"
+    try:
+        import base64
+        data = Path(path).read_bytes()
+        b64 = base64.b64encode(data).decode()
+        Path(path).unlink(missing_ok=True)
+        return f'<audio autoplay style="display:none" src="data:audio/mp3;base64,{b64}"></audio>'
+    except Exception:
+        return "<span></span>"
+
+
 def _init_state(memory, lang, history):
     """Ensure state values are never None."""
     if memory is None:
@@ -697,9 +711,7 @@ def build_app():
                         btn_en = gr.Button("🇬🇧 English",  variant="primary")
                         btn_ar = gr.Button("🇸🇦 العربية", variant="primary")
                         btn_ur = gr.Button("🇵🇰 اردو",    variant="primary")
-                    lang_audio_out = gr.Audio(
-                        label="Agent", autoplay=True,
-                        visible=False, interactive=False)
+                    lang_audio_out = gr.HTML("<span></span>")  # audio via HTML data URI
 
                 # ── Main chat interface (hidden until lang chosen) ─────
                 with gr.Group(visible=False) as chat_group:
@@ -724,11 +736,7 @@ def build_app():
                     with gr.Row():
                         send_btn    = gr.Button("Send ➤", variant="primary")
                         change_btn  = gr.Button("🔄 Change language", variant="secondary")
-                    audio_out     = gr.Audio(
-                        label="Reem speaks",
-                        autoplay=True,
-                        interactive=False,
-                    )
+                    audio_out     = gr.HTML("<span></span>")  # audio via HTML data URI
 
                     # Welcome audio hidden autoplay on load
                     welcome_audio_html = gr.HTML(
@@ -786,24 +794,24 @@ def build_app():
                 gr.update(visible=False),
                 gr.update(visible=True),
                 badge,
-                audio_path,
+                _audio_html(audio_path),
                 memory,
             )
 
         def handle_lang_mic(audio, memory):
             memory, _, _ = _init_state(memory, None, None)
             if audio is None:
-                return ("", gr.update(), gr.update(), "", None, memory)
+                return ("", gr.update(), gr.update(), "", "<span></span>", memory)
             sr, data = audio
             import io, soundfile as sf
             buf = io.BytesIO()
             sf.write(buf, data, sr, format="WAV")
             transcript = transcribe_audio_bytes(buf.getvalue())
             if not transcript:
-                return ("", gr.update(), gr.update(), "", None, memory)
+                return ("", gr.update(), gr.update(), "", "<span></span>", memory)
             detected = detect_lang(transcript)
             if not detected:
-                return ("", gr.update(), gr.update(), "", None, memory)
+                return ("", gr.update(), gr.update(), "", "<span></span>", memory)
             return _set_language(detected, memory)
 
         lang_mic.stop_recording(
@@ -834,21 +842,21 @@ def build_app():
         def handle_voice(audio, lang_code, memory, history):
             memory, lang_code, history = _init_state(memory, lang_code, history)
             if audio is None:
-                return history, None, memory, ""
+                return history, "<span></span>", memory, ""
             sr, data = audio
             import io, soundfile as sf
             buf = io.BytesIO()
             sf.write(buf, data, sr, format="WAV")
             raw = buf.getvalue()
             if len(raw) < MIN_AUDIO_BYTES:
-                return history, None, memory, ""
+                return history, "<span></span>", memory, ""
             transcript = transcribe_audio_bytes(raw)
             if not transcript or len(transcript.strip()) < 3:
                 return history, None, memory, "⚠️ No clear speech detected."
             _, new_history, audio_path = process_query(
                 transcript, memory, lang_code or "en", history)
             mem_info_txt = _mem_info(memory)
-            return new_history, audio_path, memory, mem_info_txt
+            return new_history, _audio_html(audio_path), memory, mem_info_txt
 
         voice_input.stop_recording(
             handle_voice,
@@ -860,11 +868,11 @@ def build_app():
         def handle_text(text, lang_code, memory, history):
             memory, lang_code, history = _init_state(memory, lang_code, history)
             if not text or not text.strip():
-                return history, None, memory, "", ""
+                return history, "<span></span>", memory, "", ""
             _, new_history, audio_path = process_query(
                 text, memory, lang_code or "en", history)
             mem_info_txt = _mem_info(memory)
-            return new_history, audio_path, memory, mem_info_txt, ""
+            return new_history, _audio_html(audio_path), memory, mem_info_txt, ""
 
         def _mem_info(memory: ConversationMemory) -> str:
             ec = memory.entity_cache