Spaces:
Sleeping
Sleeping
FauziIsyrinApridal
commited on
Commit
Β·
be1430e
1
Parent(s):
0ef32f0
tambahkan indikator audio dan audio selalu diputar
Browse files- app.py +12 -12
- app/chat.py +33 -17
app.py
CHANGED
@@ -17,15 +17,15 @@ from langchain_community.document_transformers import LongContextReorder
|
|
17 |
|
18 |
load_dotenv()
|
19 |
|
20 |
-
|
21 |
-
#
|
22 |
-
|
23 |
BUCKET_NAME = "pnp-bot-storage-archive"
|
24 |
VECTOR_STORE_PREFIX = "vector_store"
|
25 |
|
26 |
-
|
27 |
-
#
|
28 |
-
|
29 |
def get_latest_data_timestamp_from_files(bucket_name: str) -> float:
|
30 |
"""Get the latest timestamp from files in a Supabase storage bucket."""
|
31 |
files = list_all_files(bucket_name)
|
@@ -76,9 +76,9 @@ def reorder_embedding(docs):
|
|
76 |
return reordering.transform_documents(docs)
|
77 |
|
78 |
|
79 |
-
|
80 |
-
#
|
81 |
-
|
82 |
@traceable(name="Create RAG Conversational Chain")
|
83 |
def create_conversational_chain(vector_store):
|
84 |
"""Create a Conversational Retrieval Chain for RAG."""
|
@@ -102,9 +102,9 @@ def get_rag_chain(vector_store):
|
|
102 |
return create_conversational_chain(vector_store)
|
103 |
|
104 |
|
105 |
-
|
106 |
-
#
|
107 |
-
|
108 |
@traceable(name="Main Chatbot RAG App")
|
109 |
def main():
|
110 |
initialize_session_state()
|
|
|
17 |
|
18 |
load_dotenv()
|
19 |
|
20 |
+
|
21 |
+
# CONFIG
|
22 |
+
|
23 |
BUCKET_NAME = "pnp-bot-storage-archive"
|
24 |
VECTOR_STORE_PREFIX = "vector_store"
|
25 |
|
26 |
+
|
27 |
+
# UTILITY
|
28 |
+
|
29 |
def get_latest_data_timestamp_from_files(bucket_name: str) -> float:
|
30 |
"""Get the latest timestamp from files in a Supabase storage bucket."""
|
31 |
files = list_all_files(bucket_name)
|
|
|
76 |
return reordering.transform_documents(docs)
|
77 |
|
78 |
|
79 |
+
|
80 |
+
# RAG CHAIN
|
81 |
+
|
82 |
@traceable(name="Create RAG Conversational Chain")
|
83 |
def create_conversational_chain(vector_store):
|
84 |
"""Create a Conversational Retrieval Chain for RAG."""
|
|
|
102 |
return create_conversational_chain(vector_store)
|
103 |
|
104 |
|
105 |
+
|
106 |
+
# MAIN FUNCTION
|
107 |
+
|
108 |
@traceable(name="Main Chatbot RAG App")
|
109 |
def main():
|
110 |
initialize_session_state()
|
app/chat.py
CHANGED
@@ -11,20 +11,21 @@ from app.db import supabase
|
|
11 |
import os
|
12 |
import glob
|
13 |
import time
|
|
|
14 |
from dotenv import load_dotenv
|
15 |
|
16 |
load_dotenv()
|
17 |
|
18 |
-
# Bersihkan cache audio lama
|
19 |
def clean_old_cache(tts_dir="cache_tts", max_age_hours=12):
|
20 |
now = time.time()
|
21 |
for f in glob.glob(os.path.join(tts_dir, "*.mp3")):
|
22 |
if os.stat(f).st_mtime < now - max_age_hours * 3600:
|
23 |
os.remove(f)
|
24 |
|
25 |
-
# Jalankan pembersihan saat startup
|
26 |
clean_old_cache()
|
27 |
|
|
|
28 |
def save_feedback_to_supabase(feedback_text):
|
29 |
try:
|
30 |
data = {"message": feedback_text}
|
@@ -34,6 +35,7 @@ def save_feedback_to_supabase(feedback_text):
|
|
34 |
st.error(f"Gagal menyimpan feedback: {e}")
|
35 |
return False
|
36 |
|
|
|
37 |
def initialize_session_state():
|
38 |
if 'history' not in st.session_state:
|
39 |
st.session_state['history'] = []
|
@@ -52,14 +54,14 @@ def initialize_session_state():
|
|
52 |
if 'tts_output' not in st.session_state:
|
53 |
st.session_state['tts_output'] = ""
|
54 |
if 'tts_played' not in st.session_state:
|
55 |
-
st.session_state['tts_played'] = True
|
56 |
|
57 |
-
# edge-tts fallback
|
58 |
async def generate_audio_edge(text, path, voice="id-ID-GadisNeural"):
|
59 |
communicate = edge_tts.Communicate(text, voice=voice)
|
60 |
await communicate.save(path)
|
61 |
|
62 |
-
#
|
63 |
def text_to_speech(text):
|
64 |
cache_dir = "cache_tts"
|
65 |
os.makedirs(cache_dir, exist_ok=True)
|
@@ -85,20 +87,24 @@ def text_to_speech(text):
|
|
85 |
with open(path, "rb") as audio_file:
|
86 |
audio_base64 = base64.b64encode(audio_file.read()).decode()
|
87 |
|
|
|
|
|
88 |
return f"""
|
89 |
<audio autoplay>
|
90 |
-
<source src="data:audio/mp3;base64,{audio_base64}" type="audio/mp3">
|
91 |
</audio>
|
92 |
"""
|
93 |
except Exception as e:
|
94 |
print(f"[Error saat membaca audio] {e}")
|
95 |
return ""
|
96 |
|
|
|
97 |
def conversation_chat(query, chain, history):
|
98 |
result = chain({"question": query, "chat_history": history})
|
99 |
history.append((query, result["answer"]))
|
100 |
return result["answer"]
|
101 |
|
|
|
102 |
def display_chat_history(chain):
|
103 |
reply_container = st.container()
|
104 |
|
@@ -106,16 +112,16 @@ def display_chat_history(chain):
|
|
106 |
|
107 |
col2, col3 = st.columns([1, 1])
|
108 |
|
109 |
-
# Tombol TTS
|
110 |
with col2:
|
111 |
-
if st.button("π Text-to-Speech Aktif" if st.session_state['should_speak'] else "π Text-to-Speech Nonaktif",
|
112 |
-
key="toggle_tts",
|
113 |
-
help="Aktifkan/Nonaktifkan Text-to-Speech",
|
114 |
use_container_width=True):
|
115 |
st.session_state['should_speak'] = not st.session_state['should_speak']
|
116 |
st.experimental_rerun()
|
117 |
|
118 |
-
#
|
119 |
with col3:
|
120 |
stt_text = speech_to_text(
|
121 |
start_prompt="π€ Input Suara",
|
@@ -126,7 +132,6 @@ def display_chat_history(chain):
|
|
126 |
use_container_width=True,
|
127 |
)
|
128 |
|
129 |
-
# Jika ada STT
|
130 |
if stt_text:
|
131 |
st.session_state.input_text = stt_text
|
132 |
st.experimental_rerun()
|
@@ -142,19 +147,30 @@ def display_chat_history(chain):
|
|
142 |
st.session_state['generated'].append(output)
|
143 |
st.session_state.input_text = ""
|
144 |
|
145 |
-
# Reset flag
|
146 |
if st.session_state['should_speak'] and output:
|
147 |
-
st.session_state
|
148 |
-
|
|
|
149 |
|
150 |
-
# Tampilkan
|
151 |
if st.session_state['generated']:
|
152 |
with reply_container:
|
153 |
for i in range(len(st.session_state['generated'])):
|
154 |
message(st.session_state["past"][i], is_user=True, key=str(i) + '_user', avatar_style="no-avatar")
|
155 |
message(st.session_state["generated"][i], key=str(i), avatar_style="no-avatar")
|
156 |
|
157 |
-
# Pemutaran TTS
|
158 |
if st.session_state.get('tts_output') and not st.session_state.get('tts_played'):
|
|
|
|
|
|
|
159 |
st.markdown(text_to_speech(st.session_state['tts_output']), unsafe_allow_html=True)
|
160 |
st.session_state['tts_played'] = True
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
11 |
import os
|
12 |
import glob
|
13 |
import time
|
14 |
+
import threading
|
15 |
from dotenv import load_dotenv
|
16 |
|
17 |
load_dotenv()
|
18 |
|
19 |
+
# Bersihkan cache audio lama
|
20 |
def clean_old_cache(tts_dir="cache_tts", max_age_hours=12):
|
21 |
now = time.time()
|
22 |
for f in glob.glob(os.path.join(tts_dir, "*.mp3")):
|
23 |
if os.stat(f).st_mtime < now - max_age_hours * 3600:
|
24 |
os.remove(f)
|
25 |
|
|
|
26 |
clean_old_cache()
|
27 |
|
28 |
+
# Simpan feedback ke Supabase
|
29 |
def save_feedback_to_supabase(feedback_text):
|
30 |
try:
|
31 |
data = {"message": feedback_text}
|
|
|
35 |
st.error(f"Gagal menyimpan feedback: {e}")
|
36 |
return False
|
37 |
|
38 |
+
# Inisialisasi session state
|
39 |
def initialize_session_state():
|
40 |
if 'history' not in st.session_state:
|
41 |
st.session_state['history'] = []
|
|
|
54 |
if 'tts_output' not in st.session_state:
|
55 |
st.session_state['tts_output'] = ""
|
56 |
if 'tts_played' not in st.session_state:
|
57 |
+
st.session_state['tts_played'] = True
|
58 |
|
59 |
+
# edge-tts fallback
|
60 |
async def generate_audio_edge(text, path, voice="id-ID-GadisNeural"):
|
61 |
communicate = edge_tts.Communicate(text, voice=voice)
|
62 |
await communicate.save(path)
|
63 |
|
64 |
+
# Fungsi utama TTS dengan cache-busting
|
65 |
def text_to_speech(text):
|
66 |
cache_dir = "cache_tts"
|
67 |
os.makedirs(cache_dir, exist_ok=True)
|
|
|
87 |
with open(path, "rb") as audio_file:
|
88 |
audio_base64 = base64.b64encode(audio_file.read()).decode()
|
89 |
|
90 |
+
# πΉ Cache-busting query string
|
91 |
+
ts = int(time.time())
|
92 |
return f"""
|
93 |
<audio autoplay>
|
94 |
+
<source src="data:audio/mp3;base64,{audio_base64}?v={ts}" type="audio/mp3">
|
95 |
</audio>
|
96 |
"""
|
97 |
except Exception as e:
|
98 |
print(f"[Error saat membaca audio] {e}")
|
99 |
return ""
|
100 |
|
101 |
+
# Fungsi chat utama
|
102 |
def conversation_chat(query, chain, history):
|
103 |
result = chain({"question": query, "chat_history": history})
|
104 |
history.append((query, result["answer"]))
|
105 |
return result["answer"]
|
106 |
|
107 |
+
# Tampilan chat + TTS indikator
|
108 |
def display_chat_history(chain):
|
109 |
reply_container = st.container()
|
110 |
|
|
|
112 |
|
113 |
col2, col3 = st.columns([1, 1])
|
114 |
|
115 |
+
# Tombol toggle TTS
|
116 |
with col2:
|
117 |
+
if st.button("π Text-to-Speech Aktif" if st.session_state['should_speak'] else "π Text-to-Speech Nonaktif",
|
118 |
+
key="toggle_tts",
|
119 |
+
help="Aktifkan/Nonaktifkan Text-to-Speech",
|
120 |
use_container_width=True):
|
121 |
st.session_state['should_speak'] = not st.session_state['should_speak']
|
122 |
st.experimental_rerun()
|
123 |
|
124 |
+
# Input suara STT
|
125 |
with col3:
|
126 |
stt_text = speech_to_text(
|
127 |
start_prompt="π€ Input Suara",
|
|
|
132 |
use_container_width=True,
|
133 |
)
|
134 |
|
|
|
135 |
if stt_text:
|
136 |
st.session_state.input_text = stt_text
|
137 |
st.experimental_rerun()
|
|
|
147 |
st.session_state['generated'].append(output)
|
148 |
st.session_state.input_text = ""
|
149 |
|
150 |
+
# Reset flag TTS setiap ada output baru
|
151 |
if st.session_state['should_speak'] and output:
|
152 |
+
if st.session_state.get('tts_output') != output:
|
153 |
+
st.session_state['tts_output'] = output
|
154 |
+
st.session_state['tts_played'] = False
|
155 |
|
156 |
+
# Tampilkan riwayat chat
|
157 |
if st.session_state['generated']:
|
158 |
with reply_container:
|
159 |
for i in range(len(st.session_state['generated'])):
|
160 |
message(st.session_state["past"][i], is_user=True, key=str(i) + '_user', avatar_style="no-avatar")
|
161 |
message(st.session_state["generated"][i], key=str(i), avatar_style="no-avatar")
|
162 |
|
163 |
+
# Pemutaran TTS + indikator
|
164 |
if st.session_state.get('tts_output') and not st.session_state.get('tts_played'):
|
165 |
+
indicator = st.empty()
|
166 |
+
indicator.markdown("π§ **Memutar audio...**")
|
167 |
+
|
168 |
st.markdown(text_to_speech(st.session_state['tts_output']), unsafe_allow_html=True)
|
169 |
st.session_state['tts_played'] = True
|
170 |
+
|
171 |
+
# Hilangkan indikator setelah 3 detik (tidak blok UI)
|
172 |
+
def remove_indicator():
|
173 |
+
time.sleep(3)
|
174 |
+
indicator.empty()
|
175 |
+
|
176 |
+
threading.Thread(target=remove_indicator).start()
|