Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -546,7 +546,12 @@ HTML_CONTENT = """<!DOCTYPE html>
|
|
| 546 |
|
| 547 |
try {
|
| 548 |
const stream = await navigator.mediaDevices.getUserMedia({
|
| 549 |
-
audio:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 550 |
});
|
| 551 |
setupAudioVisualization(stream);
|
| 552 |
stream.getTracks().forEach(track => {
|
|
@@ -556,7 +561,11 @@ HTML_CONTENT = """<!DOCTYPE html>
|
|
| 556 |
peerConnection.addEventListener('track', (evt) => {
|
| 557 |
if (audioOutput.srcObject !== evt.streams[0]) {
|
| 558 |
audioOutput.srcObject = evt.streams[0];
|
| 559 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 560 |
}
|
| 561 |
});
|
| 562 |
|
|
@@ -677,7 +686,19 @@ HTML_CONTENT = """<!DOCTYPE html>
|
|
| 677 |
}
|
| 678 |
}
|
| 679 |
|
|
|
|
|
|
|
|
|
|
| 680 |
function addMessage(role, content) {
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 681 |
const messageDiv = document.createElement('div');
|
| 682 |
messageDiv.classList.add('message', role);
|
| 683 |
messageDiv.textContent = content;
|
|
@@ -810,6 +831,8 @@ class OpenAIHandler(AsyncStreamHandler):
|
|
| 810 |
self.keep_alive_task = None
|
| 811 |
self.last_activity = datetime.now()
|
| 812 |
self.connection_active = True
|
|
|
|
|
|
|
| 813 |
logger.info(f"Handler created with web_search_enabled={web_search_enabled}, webrtc_id={webrtc_id}")
|
| 814 |
|
| 815 |
def copy(self):
|
|
@@ -895,7 +918,15 @@ class OpenAIHandler(AsyncStreamHandler):
|
|
| 895 |
|
| 896 |
# Define the web search function
|
| 897 |
tools = []
|
| 898 |
-
instructions =
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 899 |
|
| 900 |
if self.web_search_enabled and self.search_client:
|
| 901 |
tools = [{
|
|
@@ -919,7 +950,13 @@ class OpenAIHandler(AsyncStreamHandler):
|
|
| 919 |
|
| 920 |
instructions = (
|
| 921 |
"You are a helpful assistant with web search capabilities. "
|
| 922 |
-
"IMPORTANT
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 923 |
"- Weather (๋ ์จ, ๊ธฐ์จ, ๋น, ๋)\n"
|
| 924 |
"- News (๋ด์ค, ์์)\n"
|
| 925 |
"- Current events (ํ์ฌ, ์ต๊ทผ, ์ค๋, ์ง๊ธ)\n"
|
|
@@ -927,20 +964,27 @@ class OpenAIHandler(AsyncStreamHandler):
|
|
| 927 |
"- Sports scores or results\n"
|
| 928 |
"- Any question about 2024 or 2025\n"
|
| 929 |
"- Any time-sensitive information\n\n"
|
| 930 |
-
"When in doubt, USE web_search.
|
| 931 |
-
"than to guess or use outdated information. Always respond in Korean when the user speaks Korean."
|
| 932 |
)
|
| 933 |
|
| 934 |
try:
|
| 935 |
async with self.client.beta.realtime.connect(
|
| 936 |
model="gpt-4o-mini-realtime-preview-2024-12-17"
|
| 937 |
) as conn:
|
| 938 |
-
# Update session with tools
|
| 939 |
session_update = {
|
| 940 |
-
"turn_detection": {
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 941 |
"instructions": instructions,
|
| 942 |
"tools": tools,
|
| 943 |
-
"tool_choice": "auto" if tools else "none"
|
|
|
|
|
|
|
|
|
|
| 944 |
}
|
| 945 |
|
| 946 |
await conn.session.update(session=session_update)
|
|
@@ -954,14 +998,34 @@ class OpenAIHandler(AsyncStreamHandler):
|
|
| 954 |
async for event in self.connection:
|
| 955 |
self.last_activity = datetime.now()
|
| 956 |
|
| 957 |
-
# Debug logging
|
| 958 |
-
if event.type.
|
| 959 |
-
logger.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 960 |
|
| 961 |
if event.type == "response.audio_transcript.done":
|
| 962 |
await self.output_queue.put(AdditionalOutputs(event))
|
| 963 |
|
| 964 |
elif event.type == "response.audio.delta":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 965 |
await self.output_queue.put(
|
| 966 |
(
|
| 967 |
self.output_sample_rate,
|
|
@@ -1008,7 +1072,7 @@ class OpenAIHandler(AsyncStreamHandler):
|
|
| 1008 |
"output": search_results
|
| 1009 |
}
|
| 1010 |
)
|
| 1011 |
-
|
| 1012 |
|
| 1013 |
except Exception as e:
|
| 1014 |
logger.error(f"Function call error: {e}")
|
|
|
|
| 546 |
|
| 547 |
try {
|
| 548 |
const stream = await navigator.mediaDevices.getUserMedia({
|
| 549 |
+
audio: {
|
| 550 |
+
echoCancellation: true, // ์์ฝ ์ ๊ฑฐ
|
| 551 |
+
noiseSuppression: true, // ๋
ธ์ด์ฆ ์ ๊ฑฐ
|
| 552 |
+
autoGainControl: true, // ์๋ ๊ฒ์ธ ์ ์ด
|
| 553 |
+
sampleRate: 24000
|
| 554 |
+
}
|
| 555 |
});
|
| 556 |
setupAudioVisualization(stream);
|
| 557 |
stream.getTracks().forEach(track => {
|
|
|
|
| 561 |
peerConnection.addEventListener('track', (evt) => {
|
| 562 |
if (audioOutput.srcObject !== evt.streams[0]) {
|
| 563 |
audioOutput.srcObject = evt.streams[0];
|
| 564 |
+
// ์๋ ์ฌ์ ์ ๋ณผ๋ฅจ ์กฐ์ ๋ก ํผ๋๋ฐฑ ๋ฐฉ์ง
|
| 565 |
+
audioOutput.volume = 0.8;
|
| 566 |
+
audioOutput.play().catch(e => {
|
| 567 |
+
console.error('Audio play error:', e);
|
| 568 |
+
});
|
| 569 |
}
|
| 570 |
});
|
| 571 |
|
|
|
|
| 686 |
}
|
| 687 |
}
|
| 688 |
|
| 689 |
+
let lastMessageContent = '';
|
| 690 |
+
let lastMessageTime = 0;
|
| 691 |
+
|
| 692 |
function addMessage(role, content) {
|
| 693 |
+
// ์ค๋ณต ๋ฉ์์ง ๋ฐฉ์ง
|
| 694 |
+
const now = Date.now();
|
| 695 |
+
if (content === lastMessageContent && (now - lastMessageTime) < 1000) {
|
| 696 |
+
return; // 1์ด ์ด๋ด ๊ฐ์ ๋ฉ์์ง ๋ฌด์
|
| 697 |
+
}
|
| 698 |
+
|
| 699 |
+
lastMessageContent = content;
|
| 700 |
+
lastMessageTime = now;
|
| 701 |
+
|
| 702 |
const messageDiv = document.createElement('div');
|
| 703 |
messageDiv.classList.add('message', role);
|
| 704 |
messageDiv.textContent = content;
|
|
|
|
| 831 |
self.keep_alive_task = None
|
| 832 |
self.last_activity = datetime.now()
|
| 833 |
self.connection_active = True
|
| 834 |
+
self.response_in_progress = False # ์๋ต ์ค๋ณต ๋ฐฉ์ง
|
| 835 |
+
self.last_response_time = datetime.now() # ๋ง์ง๋ง ์๋ต ์๊ฐ
|
| 836 |
logger.info(f"Handler created with web_search_enabled={web_search_enabled}, webrtc_id={webrtc_id}")
|
| 837 |
|
| 838 |
def copy(self):
|
|
|
|
| 918 |
|
| 919 |
# Define the web search function
|
| 920 |
tools = []
|
| 921 |
+
instructions = (
|
| 922 |
+
"You are a helpful assistant. Respond in Korean when the user speaks Korean. "
|
| 923 |
+
"IMPORTANT RULES:\n"
|
| 924 |
+
"1. Wait for the user to finish speaking before responding\n"
|
| 925 |
+
"2. Keep responses concise and to the point\n"
|
| 926 |
+
"3. Do not continue speaking if the user interrupts\n"
|
| 927 |
+
"4. Stop immediately when you finish answering the question\n"
|
| 928 |
+
"5. Do not add unnecessary elaboration or follow-up questions unless asked"
|
| 929 |
+
)
|
| 930 |
|
| 931 |
if self.web_search_enabled and self.search_client:
|
| 932 |
tools = [{
|
|
|
|
| 950 |
|
| 951 |
instructions = (
|
| 952 |
"You are a helpful assistant with web search capabilities. "
|
| 953 |
+
"IMPORTANT RULES:\n"
|
| 954 |
+
"1. Wait for the user to finish speaking before responding\n"
|
| 955 |
+
"2. Keep responses concise and to the point\n"
|
| 956 |
+
"3. Do not continue speaking if the user interrupts\n"
|
| 957 |
+
"4. Stop immediately when you finish answering the question\n"
|
| 958 |
+
"5. Do not add unnecessary elaboration or follow-up questions unless asked\n\n"
|
| 959 |
+
"WEB SEARCH RULES: You MUST use the web_search function for ANY of these topics:\n"
|
| 960 |
"- Weather (๋ ์จ, ๊ธฐ์จ, ๋น, ๋)\n"
|
| 961 |
"- News (๋ด์ค, ์์)\n"
|
| 962 |
"- Current events (ํ์ฌ, ์ต๊ทผ, ์ค๋, ์ง๊ธ)\n"
|
|
|
|
| 964 |
"- Sports scores or results\n"
|
| 965 |
"- Any question about 2024 or 2025\n"
|
| 966 |
"- Any time-sensitive information\n\n"
|
| 967 |
+
"When in doubt, USE web_search for accuracy. Always respond in Korean when the user speaks Korean."
|
|
|
|
| 968 |
)
|
| 969 |
|
| 970 |
try:
|
| 971 |
async with self.client.beta.realtime.connect(
|
| 972 |
model="gpt-4o-mini-realtime-preview-2024-12-17"
|
| 973 |
) as conn:
|
| 974 |
+
# Update session with tools - VAD ์ค์ ๊ฐํ
|
| 975 |
session_update = {
|
| 976 |
+
"turn_detection": {
|
| 977 |
+
"type": "server_vad",
|
| 978 |
+
"threshold": 0.5, # ๋ฏผ๊ฐ๋ ์กฐ์
|
| 979 |
+
"prefix_padding_ms": 300, # ์์ฑ ์์ ์ ํจ๋ฉ
|
| 980 |
+
"silence_duration_ms": 500 # ์นจ๋ฌต ๊ฐ์ง ์๊ฐ ์ฆ๊ฐ
|
| 981 |
+
},
|
| 982 |
"instructions": instructions,
|
| 983 |
"tools": tools,
|
| 984 |
+
"tool_choice": "auto" if tools else "none",
|
| 985 |
+
"voice": "echo", # ์์ฑ ์ค์
|
| 986 |
+
"temperature": 0.7, # ์๋ต ๋ค์์ฑ ๊ฐ์
|
| 987 |
+
"max_response_output_tokens": "inf" # ๋ฌดํ ์๋ต ๋ฐฉ์ง
|
| 988 |
}
|
| 989 |
|
| 990 |
await conn.session.update(session=session_update)
|
|
|
|
| 998 |
async for event in self.connection:
|
| 999 |
self.last_activity = datetime.now()
|
| 1000 |
|
| 1001 |
+
# Debug logging
|
| 1002 |
+
if event.type in ["response.audio_transcript.done", "response.done", "response.created", "input_audio_buffer.speech_started", "input_audio_buffer.speech_stopped"]:
|
| 1003 |
+
logger.info(f"Event: {event.type}")
|
| 1004 |
+
|
| 1005 |
+
# ์ฌ์ฉ์ ์์ฑ ๊ฐ์ง
|
| 1006 |
+
if event.type == "input_audio_buffer.speech_started":
|
| 1007 |
+
logger.info("User started speaking")
|
| 1008 |
+
self.response_in_progress = False # ์ฌ์ฉ์๊ฐ ๋งํ๋ฉด ์๋ต ์ค๋จ
|
| 1009 |
+
|
| 1010 |
+
# ์๋ต ์์/์ข
๋ฃ ์ถ์
|
| 1011 |
+
if event.type == "response.created":
|
| 1012 |
+
self.response_in_progress = True
|
| 1013 |
+
self.last_response_time = datetime.now()
|
| 1014 |
+
logger.info("Response started")
|
| 1015 |
+
|
| 1016 |
+
if event.type == "response.done":
|
| 1017 |
+
self.response_in_progress = False
|
| 1018 |
+
logger.info("Response completed")
|
| 1019 |
|
| 1020 |
if event.type == "response.audio_transcript.done":
|
| 1021 |
await self.output_queue.put(AdditionalOutputs(event))
|
| 1022 |
|
| 1023 |
elif event.type == "response.audio.delta":
|
| 1024 |
+
# ์๋ต ์ค๋ณต ๋ฐฉ์ง - ๋๋ฌด ๋น ๋ฅธ ์ฐ์ ์๋ต ์ฐจ๋จ
|
| 1025 |
+
time_since_last = (datetime.now() - self.last_response_time).total_seconds()
|
| 1026 |
+
if time_since_last < 0.1: # 100ms ์ด๋ด ์๋ต ๋ฌด์
|
| 1027 |
+
continue
|
| 1028 |
+
|
| 1029 |
await self.output_queue.put(
|
| 1030 |
(
|
| 1031 |
self.output_sample_rate,
|
|
|
|
| 1072 |
"output": search_results
|
| 1073 |
}
|
| 1074 |
)
|
| 1075 |
+
# response.create() ์ ๊ฑฐ - ์๋์ผ๋ก ์๋ต ์์ฑ๋จ
|
| 1076 |
|
| 1077 |
except Exception as e:
|
| 1078 |
logger.error(f"Function call error: {e}")
|