Update app.py

app.py

@@ -57,10 +57,8 @@ Your role:
 
 Always respond in the same language as the user's question, except when translation is requested.
 When providing Mandarin text, include pinyin pronunciation in parentheses if helpful.
-Be encouraging and patient – the user is learning.
-"""
+Be encouraging and patient – the user is learning."""
 
-# Voice configuration options (mapped from React components)
 ACCENT_EMOJI = {
     "american": "🇺🇸", "british": "🇬🇧", "australian": "🇦🇺", "canadian": "🇨🇦",
     "irish": "🇮🇪", "scottish": "🏴󠁧󠁢󠁳󠁣󠁴󠁿", "indian": "🇮🇳", "south-african": "🇿🇦",
@@ -97,17 +95,11 @@ def transcribe_audio(audio_path, token=None):
         return f"[STT Error: {str(e)}]"
 
 def synthesize_speech(text, token=None, accent="chinese", gender="female", age=30):
-    """Convert text to speech with voice characteristics."""
     if not text:
         return None
     client = get_client(token)
-    # Construct a speaker prompt that Bark understands.
-    # Bark can be guided with a description like "A young female Chinese speaker".
-    # We'll combine the options into a natural description.
     age_desc = f"{age} year old" if age else ""
     voice_desc = f"{age_desc} {gender} {accent} speaker".strip()
-    # Prepend the description to the text, separated by a special token or just as context.
-    # Bark often uses a prompt like: "[Voice description] Text to speak."
     prompt = f"[{voice_desc}] {text}"
     try:
         audio_bytes = client.text_to_speech(prompt, model="suno/bark")
@@ -118,21 +110,28 @@ def synthesize_speech(text, token=None, accent="chinese", gender="female", age=3
         print(f"TTS error: {e}")
         return None
 
-def generate_response(message, history, model_name, profile, native_lang, token=None):
+def user_message(message, history):
+    """Append user message to history and clear input."""
     if not message.strip():
-        yield "Please enter a message."
-        return
+        return history, ""
+    history.append((message, None))
+    return history, ""
 
+def bot_response(history, model_name, profile, native_lang, token):
+    """Stream assistant response and update history."""
+    if not history or history[-1][1] is not None:
+        yield history
+        return
+    user_msg = history[-1][0]
     client = get_client(token)
     system_prompt = SYSTEM_PROMPT_TEMPLATE.format(**profile)
     messages = [{"role": "system", "content": system_prompt}]
-    for user_msg, assistant_msg in history:
-        if user_msg:
-            messages.append({"role": "user", "content": user_msg})
-        if assistant_msg:
-            messages.append({"role": "assistant", "content": assistant_msg})
-    messages.append({"role": "user", "content": message})
-
+    for u, a in history[:-1]:
+        if u:
+            messages.append({"role": "user", "content": u})
+        if a:
+            messages.append({"role": "assistant", "content": a})
+    messages.append({"role": "user", "content": user_msg})
     model_id = MODEL_MAP.get(model_name, MODEL_MAP[DEFAULT_MODEL])
     try:
         stream = client.chat_completion(
@@ -145,14 +144,19 @@ def generate_response(message, history, model_name, profile, native_lang, token=
         for chunk in stream:
             if chunk.choices and chunk.choices[0].delta.content:
                 partial += chunk.choices[0].delta.content
-                yield partial
+                new_history = history.copy()
+                new_history[-1] = (user_msg, partial)
+                yield new_history
     except Exception as e:
         error_msg = f"Error: {str(e)}"
         if "401" in str(e) or "Authorization" in str(e):
             error_msg = "⚠️ Authentication failed. Please provide a valid Hugging Face token."
-        yield error_msg
+        new_history = history.copy()
+        new_history[-1] = (user_msg, error_msg)
+        yield new_history
 
-def voice_input(audio, history, token=None):
+def voice_input(audio, history, token):
+    """Transcribe audio and return updated history + transcript."""
     if audio is None:
         return history, "", None
     transcript = transcribe_audio(audio, token)
@@ -161,14 +165,45 @@ def voice_input(audio, history, token=None):
         return history, "", None
     return history, transcript, audio
 
-def send_message(message, history):
-    if not message.strip():
-        return history, ""
-    history.append((message, None))
-    return history, ""
-
-def quick_action_clicked(action, history):
-    return action, history
+def quick_action_send(action, history, model, profile, lang, token):
+    """Handle quick action button: append message and stream response."""
+    if not action.strip():
+        yield history
+        return
+    history.append((action, None))
+    yield history  # show user message immediately
+    # Then stream bot response
+    client = get_client(token)
+    system_prompt = SYSTEM_PROMPT_TEMPLATE.format(**profile)
+    messages = [{"role": "system", "content": system_prompt}]
+    for u, a in history[:-1]:
+        if u:
+            messages.append({"role": "user", "content": u})
+        if a:
+            messages.append({"role": "assistant", "content": a})
+    messages.append({"role": "user", "content": action})
+    model_id = MODEL_MAP.get(model, MODEL_MAP[DEFAULT_MODEL])
+    try:
+        stream = client.chat_completion(
+            messages=messages,
+            model=model_id,
+            max_tokens=1024,
+            stream=True,
+        )
+        partial = ""
+        for chunk in stream:
+            if chunk.choices and chunk.choices[0].delta.content:
+                partial += chunk.choices[0].delta.content
+                new_history = history.copy()
+                new_history[-1] = (action, partial)
+                yield new_history
+    except Exception as e:
+        error_msg = f"Error: {str(e)}"
+        if "401" in str(e) or "Authorization" in str(e):
+            error_msg = "⚠️ Authentication failed. Please provide a valid Hugging Face token."
+        new_history = history.copy()
+        new_history[-1] = (action, error_msg)
+        yield new_history
 
 def update_profile(name, native_lang, target_job, skills, location):
     return {
@@ -192,10 +227,16 @@ def toggle_ui_language(lang):
 def clear_chat():
     return [], None
 
+def speak_last_response(history, token, accent, gender, age):
+    if not history or not history[-1][1]:
+        return None
+    text = history[-1][1]
+    return synthesize_speech(text, token, accent, gender, age)
+
 # ----------------------------------------------------------------------
 # Gradio Interface
 # ----------------------------------------------------------------------
-with gr.Blocks(title="Mandarin Job Application Assistant", theme=gr.themes.Soft()) as demo:
+with gr.Blocks(title="Mandarin Job Application Assistant") as demo:
     gr.Markdown("""
     # 🗣️ Mandarin Job Application Assistant
     **For non‑Mandarin speakers applying to jobs that require Mandarin.**  
@@ -250,14 +291,13 @@ with gr.Blocks(title="Mandarin Job Application Assistant", theme=gr.themes.Soft(
                 interactive=True
             )
 
-            # Voice Configuration (inspired by React components)
+            # Voice Configuration
             gr.Markdown("### 🎙️ Voice Settings (for spoken responses)")
             with gr.Row():
                 accent = gr.Dropdown(
                     choices=list(ACCENT_EMOJI.keys()),
                     value="chinese",
-                    label="Accent",
-                    info="Choose the accent for the spoken response"
+                    label="Accent"
                 )
                 gender = gr.Radio(
                     choices=["male", "female", "non-binary", "other"],
@@ -313,10 +353,10 @@ with gr.Blocks(title="Mandarin Job Application Assistant", theme=gr.themes.Soft(
 
     # Send message (text input)
     send_btn.click(
-        send_message, inputs=[msg, chat_state], outputs=[chatbot, msg]
+        user_message, inputs=[msg, chat_state], outputs=[chatbot, msg]
     ).then(
-        generate_response,
-        inputs=[msg, chat_state, model, profile_state, native_lang_input, token_state],
+        bot_response,
+        inputs=[chat_state, model, profile_state, native_lang_input, token_state],
         outputs=chatbot
     )
 
@@ -324,8 +364,10 @@ with gr.Blocks(title="Mandarin Job Application Assistant", theme=gr.themes.Soft(
     voice_btn.click(
         voice_input, inputs=[audio_input, chat_state, token_state], outputs=[chatbot, msg, audio_input]
     ).then(
-        generate_response,
-        inputs=[msg, chat_state, model, profile_state, native_lang_input, token_state],
+        user_message, inputs=[msg, chat_state], outputs=[chatbot, msg]
+    ).then(
+        bot_response,
+        inputs=[chat_state, model, profile_state, native_lang_input, token_state],
         outputs=chatbot
     )
 
@@ -333,11 +375,11 @@ with gr.Blocks(title="Mandarin Job Application Assistant", theme=gr.themes.Soft(
     all_quick_btns = [btn_en1, btn_en2, btn_en3, btn_en4, btn_ms1, btn_ms2, btn_ms3, btn_ms4]
     for btn in all_quick_btns:
         btn.click(
-            quick_action_clicked, inputs=[btn, chat_state], outputs=[msg, chat_state]
-        ).then(
-            generate_response,
-            inputs=[msg, chat_state, model, profile_state, native_lang_input, token_state],
+            quick_action_send,
+            inputs=[btn, chat_state, model, profile_state, native_lang_input, token_state],
             outputs=chatbot
+        ).then(
+            lambda: "", None, msg  # clear message box
         )
 
     # UI language toggle
@@ -347,13 +389,7 @@ with gr.Blocks(title="Mandarin Job Application Assistant", theme=gr.themes.Soft(
         outputs=[btn_en1, btn_en2, btn_en3, btn_en4, btn_ms1, btn_ms2, btn_ms3, btn_ms4]
     )
 
-    # TTS for last response (with voice settings)
-    def speak_last_response(history, token, accent, gender, age):
-        if not history or not history[-1][1]:
-            return None
-        text = history[-1][1]
-        return synthesize_speech(text, token, accent, gender, age)
-
+    # TTS for last response
     tts_btn.click(
         speak_last_response,
         inputs=[chat_state, token_state, accent, gender, age],
@@ -365,5 +401,8 @@ with gr.Blocks(title="Mandarin Job Application Assistant", theme=gr.themes.Soft(
     # Clear chat
     clear_btn.click(clear_chat, outputs=[chatbot, audio_output])
 
+# ----------------------------------------------------------------------
+# Launch
+# ----------------------------------------------------------------------
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch(theme=gr.themes.Soft())