fix: enhance session management and output handling in startup and response functions

app.py

@@ -25,13 +25,16 @@ if phone_waiting_sound.channels > 1:
     sound_samples = sound_samples.reshape((-1, phone_waiting_sound.channels)).mean(axis=1)
 sound_samples = sound_samples.astype(np.float32) / 32768.0  # Normalize to [-1, 
 def startup(*arg):
-    
+    session_identifier = arg[1] if len(arg) > 1 else None
 
     yield (phone_waiting_sound.frame_rate, sound_samples)
     STARTUP_MESSAGE = "สวัสดีค่ะ พลอย 1577Homeshopping ยินดีให้บริการค่ะ"
     yield from synthesize_text(STARTUP_MESSAGE)
     time.sleep(1)
-    yield AdditionalOutputs([{"role": "assistant", "content": STARTUP_MESSAGE}])
+    yield AdditionalOutputs(
+        [{"role": "assistant", "content": STARTUP_MESSAGE}],
+        session_identifier,
+    )
 
 custom_css = """
 /* Overall Gradio page styling: hot pink background */
@@ -91,10 +94,15 @@ def response(
     conversation_history = conversation_history or []
     start_time = time.time()
     session_identifier = session_id or ""
+    generated_session_id = False
     if not session_identifier:
         session_identifier = str(uuid.uuid4())
+        generated_session_id = True
         print(f"[WARN] Missing session_id; generated temporary session {session_identifier}")
 
+    if generated_session_id:
+        yield AdditionalOutputs(snapshot_history(conversation_history), session_identifier)
+
     if not audio or audio[1] is None or not np.any(audio[1]):
         print("No audio input detected; skipping response generation.")
         print(f"------------------------")
@@ -143,7 +151,7 @@ def response(
     print(f"User: {transcription}")
     if is_valid_turn(user_turn):
         conversation_history.append(user_turn)
-        yield AdditionalOutputs(snapshot_history(conversation_history))
+        yield AdditionalOutputs(snapshot_history(conversation_history), session_identifier)
 
     # print("Conversation history:", conversation_history)
 
@@ -223,7 +231,10 @@ def response(
                             first_chunk_sent = True
                             text_buffer = ""
                             delimiter_count = 0
-                            yield AdditionalOutputs(snapshot_history(conversation_history))
+                            yield AdditionalOutputs(
+                                snapshot_history(conversation_history),
+                                session_identifier,
+                            )
 
                     i += 1
         if text_buffer.strip():
@@ -247,7 +258,10 @@ def response(
                 text_buffer = ""
                 delimiter_count = 0
                 
-                yield AdditionalOutputs(snapshot_history(conversation_history))
+                yield AdditionalOutputs(
+                    snapshot_history(conversation_history),
+                    session_identifier,
+                )
 
     except Exception as e:
         print(f"An error occurred during response generation or synthesis: {e}")
@@ -257,7 +271,10 @@ def response(
         except Exception as synth_error:
             print(f"Could not synthesize error message: {synth_error}")
         assistant_turn["content"] = (assistant_turn.get("content", "") + f" [Error: {e}]").strip()
-        yield AdditionalOutputs(snapshot_history(conversation_history))
+        yield AdditionalOutputs(
+            snapshot_history(conversation_history),
+            session_identifier,
+        )
 
     total_latency = time.time() - start_time
     print(f"Total:      {total_latency:.4f}s")
@@ -276,6 +293,12 @@ def initialize_session_id():
 with gr.Blocks(css=custom_css, theme=gr.themes.Soft(primary_hue="orange", secondary_hue="orange")) as demo:
     gr.HTML("""<h1 style='text-align: center'>1577 Voicebot Demo</h1>""")
     session_state = gr.State(value=None)
+    session_display = gr.Textbox(
+        label="Session ID",
+        value="",
+        interactive=False,
+    )
+
     with gr.Row():
         with gr.Column(scale=1, elem_classes=["phone-column"]):
             audio = WebRTC(
@@ -340,9 +363,12 @@ with gr.Blocks(css=custom_css, theme=gr.themes.Soft(primary_hue="orange", second
         time_limit=8192
     )
 
+    def _sync_history_and_session(history, session_identifier):
+        return history, session_identifier, session_identifier
+
     audio.on_additional_outputs(
-        lambda history: history,
-        outputs=[conversation_history],
+        _sync_history_and_session,
+        outputs=[conversation_history, session_state, session_display],
         queue=True,
         show_progress="hidden"
     )

backend/systemprompt.py

@@ -196,8 +196,8 @@ You are "Ploy" (พลอย), a friendly and professional sales expert for 1577
 
 **Step C: Address Collection (Strict Sequential Flow)**
 Collect data **one field at a time**. Confirm the user's input before moving to the next.
-1. **ชื่อจริง:** (Verify spelling if unclear).|
-2. **นามสกุล:** (Verify spelling if unclear).|
+1. **ชื่อจริง[Assume Thai Name]:** (Ask user to speak clearly if unclear).|
+2. **นามสกุล[Assume Thai last name]:** (Ask user to speak clearly if unclear).|
 3. **Province (จังหวัด):**|
 4. **District (เขต/อำเภอ):**|
 5. **Sub-district (แขวง/ตำบล):**|