..

app/chat.py

@@ -6,28 +6,25 @@ import asyncio
 import edge_tts
 import gtts
 from hashlib import md5
+from io import BytesIO
+from app.db import supabase
 import os
 import glob
 import time
 from dotenv import load_dotenv
-from app.db import supabase
 
 load_dotenv()
 
-# ----------------------------
-# Pembersihan cache audio lama
-# ----------------------------
+# Bersihkan cache audio lama (opsional)
 def clean_old_cache(tts_dir="cache_tts", max_age_hours=12):
     now = time.time()
     for f in glob.glob(os.path.join(tts_dir, "*.mp3")):
         if os.stat(f).st_mtime < now - max_age_hours * 3600:
             os.remove(f)
 
+# Jalankan pembersihan saat startup
 clean_old_cache()
 
-# ----------------------------
-# Simpan feedback ke Supabase
-# ----------------------------
 def save_feedback_to_supabase(feedback_text):
     try:
         data = {"message": feedback_text}
@@ -37,9 +34,6 @@ def save_feedback_to_supabase(feedback_text):
         st.error(f"Gagal menyimpan feedback: {e}")
         return False
 
-# ----------------------------
-# State awal
-# ----------------------------
 def initialize_session_state():
     if 'history' not in st.session_state:
         st.session_state['history'] = []
@@ -47,59 +41,59 @@ def initialize_session_state():
         st.session_state['generated'] = ["Halo! Saya bisa membantu anda menjawab pertanyaan seputar Politeknik Negeri Padang!"]
     if 'past' not in st.session_state:
         st.session_state['past'] = ["Hai! 👋"]
+    if 'data_len' not in st.session_state:
+        st.session_state['data_len'] = 0
+    if 'vector_store' not in st.session_state:
+        st.session_state['vector_store'] = None
     if 'should_speak' not in st.session_state:
         st.session_state['should_speak'] = True
     if 'input_text' not in st.session_state:
         st.session_state['input_text'] = ""
-    if 'tts_cache' not in st.session_state:
-        st.session_state['tts_cache'] = {}
-    if 'last_tts' not in st.session_state:
-        st.session_state['last_tts'] = None
-
-# ----------------------------
-# TTS gTTS → Edge fallback
-# ----------------------------
+    if 'tts_output' not in st.session_state:
+        st.session_state['tts_output'] = ""
+    if 'tts_played' not in st.session_state:
+        st.session_state['tts_played'] = True  # default True supaya tidak main saat awal
+
+# edge-tts fallback (cadangan)
 async def generate_audio_edge(text, path, voice="id-ID-GadisNeural"):
     communicate = edge_tts.Communicate(text, voice=voice)
     await communicate.save(path)
 
-def get_tts_path(text):
+# fungsi utama TTS dengan fallback
+def text_to_speech(text):
     cache_dir = "cache_tts"
     os.makedirs(cache_dir, exist_ok=True)
     filename = f"{md5(text.encode()).hexdigest()}.mp3"
-    return os.path.join(cache_dir, filename)
+    path = os.path.join(cache_dir, filename)
 
-def text_to_speech(text):
-    path = get_tts_path(text)
     if not os.path.exists(path):
         try:
+            # ✅ Utama: gTTS
             tts = gtts.gTTS(text, lang="id")
             tts.save(path)
         except Exception as e:
             print(f"[gTTS gagal] {e}")
             try:
+                # ✅ Cadangan: edge-tts
                 asyncio.run(generate_audio_edge(text, path))
             except Exception as e2:
-                print(f"[Edge-TTS gagal] {e2}")
-                return None
-    return path
-
-def play_audio_ui(path, autoplay=False):
-    with open(path, "rb") as audio_file:
-        audio_base64 = base64.b64encode(audio_file.read()).decode()
-    autoplay_attr = "autoplay" if autoplay else ""
-    st.markdown(
-        f"""
-        <audio {autoplay_attr} style="display:none;">
+                print(f"[Edge-TTS juga gagal] {e2}")
+                st.warning("🔇 Gagal membuat audio TTS.")
+                return ""
+
+    try:
+        with open(path, "rb") as audio_file:
+            audio_base64 = base64.b64encode(audio_file.read()).decode()
+
+        return f"""
+        <audio autoplay>
             <source src="data:audio/mp3;base64,{audio_base64}" type="audio/mp3">
         </audio>
-        """,
-        unsafe_allow_html=True
-    )
+        """
+    except Exception as e:
+        print(f"[Error saat membaca audio] {e}")
+        return ""
 
-# ----------------------------
-# Chat & TTS Bubble
-# ----------------------------
 def conversation_chat(query, chain, history):
     result = chain({"question": query, "chat_history": history})
     history.append((query, result["answer"]))
@@ -107,15 +101,21 @@ def conversation_chat(query, chain, history):
 
 def display_chat_history(chain):
     reply_container = st.container()
+
     user_input_obj = st.chat_input("Masukkan pertanyaan", key="chat_input_field")
 
     col2, col3 = st.columns([1, 1])
+
+    # Tombol TTS Aktif / Nonaktif
     with col2:
-        if st.button("🔊 Aktifkan" if st.session_state['should_speak'] else "🔇 Nonaktifkan",
-                     key="toggle_tts", use_container_width=True):
+        if st.button("🔊 Text-to-Speech Aktif" if st.session_state['should_speak'] else "🔇 Text-to-Speech Nonaktif", 
+                     key="toggle_tts", 
+                     help="Aktifkan/Nonaktifkan Text-to-Speech", 
+                     use_container_width=True):
             st.session_state['should_speak'] = not st.session_state['should_speak']
             st.experimental_rerun()
 
+    # Tombol Input Suara
     with col3:
         stt_text = speech_to_text(
             start_prompt="🎤 Input Suara",
@@ -126,10 +126,12 @@ def display_chat_history(chain):
             use_container_width=True,
         )
 
+    # Jika ada STT
     if stt_text:
         st.session_state.input_text = stt_text
         st.experimental_rerun()
 
+    # Ambil input user
     user_input = user_input_obj or st.session_state.get("input_text", "")
 
     if user_input:
@@ -140,48 +142,19 @@ def display_chat_history(chain):
         st.session_state['generated'].append(output)
         st.session_state.input_text = ""
 
-        # Cache TTS
-        tts_path = text_to_speech(output)
-        if tts_path:
-            st.session_state['tts_cache'][output] = tts_path
-            if st.session_state['should_speak']:
-                st.session_state['last_tts'] = output
+        # Reset flag supaya TTS siap memutar lagi
+        if st.session_state['should_speak'] and output:
+            st.session_state['tts_output'] = output
+            st.session_state['tts_played'] = False
 
-    # Tampilkan chat + tombol icon kecil
+    # Tampilkan Riwayat Chat
     if st.session_state['generated']:
         with reply_container:
             for i in range(len(st.session_state['generated'])):
-                # Bubble user
-                message(st.session_state["past"][i], is_user=True, key=f"{i}_user", avatar_style="no-avatar")
-
-                # Bubble bot + icon play kecil
-                bot_text = st.session_state["generated"][i]
-                message(bot_text, key=f"{i}", avatar_style="no-avatar")
-
-                if bot_text in st.session_state['tts_cache']:
-                    audio_id = f"audio_{i}"
-                    audio_path = st.session_state['tts_cache'][bot_text]
-                    with open(audio_path, "rb") as audio_file:
-                        audio_base64 = base64.b64encode(audio_file.read()).decode()
-
-                    st.markdown(
-                        f"""
-                        <div style="text-align:right; margin-top:-30px; margin-bottom:10px;">
-                            <button onclick="var audio = document.getElementById('{audio_id}'); audio.play();" 
-                                    style="background:none; border:none; cursor:pointer; font-size:18px;">
-                                🔊
-                            </button>
-                            <audio id="{audio_id}">
-                                <source src="data:audio/mp3;base64,{audio_base64}" type="audio/mp3">
-                            </audio>
-                        </div>
-                        """,
-                        unsafe_allow_html=True
-                    )
-
-    # Auto-play untuk jawaban baru
-    if st.session_state['last_tts']:
-        last_text = st.session_state['last_tts']
-        if last_text in st.session_state['tts_cache']:
-            play_audio_ui(st.session_state['tts_cache'][last_text], autoplay=True)
-        st.session_state['last_tts'] = None
+                message(st.session_state["past"][i], is_user=True, key=str(i) + '_user', avatar_style="no-avatar")
+                message(st.session_state["generated"][i], key=str(i), avatar_style="no-avatar")
+
+    # Pemutaran TTS
+    if st.session_state.get('tts_output') and not st.session_state.get('tts_played'):
+        st.markdown(text_to_speech(st.session_state['tts_output']), unsafe_allow_html=True)
+        st.session_state['tts_played'] = True