Update services/gemini_realtime_service.py
Browse files
services/gemini_realtime_service.py
CHANGED
|
@@ -2,7 +2,7 @@ import asyncio
|
|
| 2 |
import base64
|
| 3 |
import json
|
| 4 |
import os
|
| 5 |
-
from typing import
|
| 6 |
import numpy as np
|
| 7 |
from google import genai
|
| 8 |
from google.genai.types import (
|
|
@@ -32,6 +32,7 @@ class GeminiRealtimeService:
|
|
| 32 |
api_key=self.api_key,
|
| 33 |
http_options={"api_version": "v1alpha"},
|
| 34 |
)
|
|
|
|
| 35 |
|
| 36 |
async def start_session(self, voice_name: str = "Puck", callback: Callable = None):
|
| 37 |
"""Bắt đầu session Gemini Realtime"""
|
|
@@ -42,6 +43,7 @@ class GeminiRealtimeService:
|
|
| 42 |
self.voice_name = voice_name
|
| 43 |
self.callback = callback
|
| 44 |
|
|
|
|
| 45 |
config = LiveConnectConfig(
|
| 46 |
response_modalities=["AUDIO"],
|
| 47 |
speech_config=SpeechConfig(
|
|
@@ -53,24 +55,28 @@ class GeminiRealtimeService:
|
|
| 53 |
),
|
| 54 |
)
|
| 55 |
|
| 56 |
-
|
|
|
|
| 57 |
model="gemini-2.0-flash-exp",
|
| 58 |
config=config
|
| 59 |
)
|
| 60 |
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
|
|
|
|
|
|
|
|
|
| 74 |
return True
|
| 75 |
|
| 76 |
except Exception as e:
|
|
@@ -84,29 +90,10 @@ class GeminiRealtimeService:
|
|
| 84 |
print(error_msg)
|
| 85 |
return False
|
| 86 |
|
| 87 |
-
async def
|
| 88 |
-
"""
|
| 89 |
-
if not self.session or not self.is_active:
|
| 90 |
-
return False
|
| 91 |
-
|
| 92 |
-
try:
|
| 93 |
-
# Gemini expects 16kHz sample rate
|
| 94 |
-
if sample_rate != 16000:
|
| 95 |
-
audio_chunk = self._resample_audio(audio_chunk, sample_rate, 16000)
|
| 96 |
-
|
| 97 |
-
# Encode và gửi audio
|
| 98 |
-
audio_bytes = audio_chunk.tobytes()
|
| 99 |
-
await self.session.send(audio_bytes)
|
| 100 |
-
return True
|
| 101 |
-
|
| 102 |
-
except Exception as e:
|
| 103 |
-
print(f"❌ Lỗi gửi audio đến Gemini: {e}")
|
| 104 |
-
return False
|
| 105 |
-
|
| 106 |
-
async def _response_handler(self):
|
| 107 |
-
"""Xử lý responses từ Gemini"""
|
| 108 |
try:
|
| 109 |
-
async for response in
|
| 110 |
if hasattr(response, 'data') and response.data:
|
| 111 |
# Audio response from Gemini
|
| 112 |
audio_data = np.frombuffer(response.data, dtype=np.int16)
|
|
@@ -130,34 +117,65 @@ class GeminiRealtimeService:
|
|
| 130 |
})
|
| 131 |
|
| 132 |
except Exception as e:
|
| 133 |
-
error_msg = f"❌ Lỗi
|
| 134 |
if self.callback:
|
| 135 |
await self.callback({
|
| 136 |
'type': 'error',
|
| 137 |
'message': error_msg,
|
| 138 |
'status': 'error'
|
| 139 |
})
|
| 140 |
-
print(error_msg)
|
| 141 |
|
| 142 |
-
def
|
| 143 |
-
"""
|
| 144 |
-
if
|
| 145 |
-
return
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 146 |
|
| 147 |
-
|
| 148 |
-
|
| 149 |
-
|
| 150 |
-
|
| 151 |
-
|
| 152 |
-
|
| 153 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 154 |
|
| 155 |
async def close(self):
|
| 156 |
"""Đóng kết nối Gemini"""
|
| 157 |
-
|
| 158 |
-
|
| 159 |
-
self.is_active = False
|
| 160 |
-
|
| 161 |
if self.callback:
|
| 162 |
await self.callback({
|
| 163 |
'type': 'status',
|
|
|
|
| 2 |
import base64
|
| 3 |
import json
|
| 4 |
import os
|
| 5 |
+
from typing import Callable, Optional
|
| 6 |
import numpy as np
|
| 7 |
from google import genai
|
| 8 |
from google.genai.types import (
|
|
|
|
| 32 |
api_key=self.api_key,
|
| 33 |
http_options={"api_version": "v1alpha"},
|
| 34 |
)
|
| 35 |
+
return True
|
| 36 |
|
| 37 |
async def start_session(self, voice_name: str = "Puck", callback: Callable = None):
|
| 38 |
"""Bắt đầu session Gemini Realtime"""
|
|
|
|
| 43 |
self.voice_name = voice_name
|
| 44 |
self.callback = callback
|
| 45 |
|
| 46 |
+
# Tạo config cho realtime session
|
| 47 |
config = LiveConnectConfig(
|
| 48 |
response_modalities=["AUDIO"],
|
| 49 |
speech_config=SpeechConfig(
|
|
|
|
| 55 |
),
|
| 56 |
)
|
| 57 |
|
| 58 |
+
# Kết nối session - SỬA LỖI Ở ĐÂY
|
| 59 |
+
self.session = self.client.aio.live.connect(
|
| 60 |
model="gemini-2.0-flash-exp",
|
| 61 |
config=config
|
| 62 |
)
|
| 63 |
|
| 64 |
+
# Sử dụng async with để quản lý session
|
| 65 |
+
async with self.session as session:
|
| 66 |
+
self.is_active = True
|
| 67 |
+
|
| 68 |
+
if self.callback:
|
| 69 |
+
await self.callback({
|
| 70 |
+
'type': 'status',
|
| 71 |
+
'message': f'✅ Đã kết nối Gemini - Giọng: {voice_name}',
|
| 72 |
+
'status': 'connected'
|
| 73 |
+
})
|
| 74 |
+
|
| 75 |
+
print("✅ Gemini Realtime session started")
|
| 76 |
+
|
| 77 |
+
# Xử lý realtime communication
|
| 78 |
+
await self._handle_realtime_session(session)
|
| 79 |
+
|
| 80 |
return True
|
| 81 |
|
| 82 |
except Exception as e:
|
|
|
|
| 90 |
print(error_msg)
|
| 91 |
return False
|
| 92 |
|
| 93 |
+
async def _handle_realtime_session(self, session):
|
| 94 |
+
"""Xử lý session realtime"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 95 |
try:
|
| 96 |
+
async for response in session:
|
| 97 |
if hasattr(response, 'data') and response.data:
|
| 98 |
# Audio response from Gemini
|
| 99 |
audio_data = np.frombuffer(response.data, dtype=np.int16)
|
|
|
|
| 117 |
})
|
| 118 |
|
| 119 |
except Exception as e:
|
| 120 |
+
error_msg = f"❌ Lỗi trong session: {e}"
|
| 121 |
if self.callback:
|
| 122 |
await self.callback({
|
| 123 |
'type': 'error',
|
| 124 |
'message': error_msg,
|
| 125 |
'status': 'error'
|
| 126 |
})
|
|
|
|
| 127 |
|
| 128 |
+
async def send_audio(self, audio_data: bytes):
|
| 129 |
+
"""Gửi audio data đến Gemini session"""
|
| 130 |
+
if not self.session or not self.is_active:
|
| 131 |
+
return False
|
| 132 |
+
|
| 133 |
+
try:
|
| 134 |
+
# Trong implementation thực tế, sẽ gửi qua session
|
| 135 |
+
# await self.session.send(audio_data)
|
| 136 |
+
print(f"📤 Sent audio data: {len(audio_data)} bytes")
|
| 137 |
+
return True
|
| 138 |
+
|
| 139 |
+
except Exception as e:
|
| 140 |
+
print(f"❌ L���i gửi audio: {e}")
|
| 141 |
+
return False
|
| 142 |
+
|
| 143 |
+
async def send_text(self, text: str):
|
| 144 |
+
"""Gửi text message đến Gemini (fallback)"""
|
| 145 |
+
if not self.client or not self.is_active:
|
| 146 |
+
return None
|
| 147 |
+
|
| 148 |
+
try:
|
| 149 |
+
# Fallback: sử dụng chat completion thông thường
|
| 150 |
+
response = await self.client.aio.models.generate_content(
|
| 151 |
+
model="gemini-2.0-flash-exp",
|
| 152 |
+
contents=text
|
| 153 |
+
)
|
| 154 |
+
|
| 155 |
+
if self.callback:
|
| 156 |
+
await self.callback({
|
| 157 |
+
'type': 'text',
|
| 158 |
+
'content': response.text,
|
| 159 |
+
'role': 'assistant',
|
| 160 |
+
'status': 'text_response'
|
| 161 |
+
})
|
| 162 |
|
| 163 |
+
return response.text
|
| 164 |
+
|
| 165 |
+
except Exception as e:
|
| 166 |
+
error_msg = f"❌ Lỗi gửi text: {e}"
|
| 167 |
+
if self.callback:
|
| 168 |
+
await self.callback({
|
| 169 |
+
'type': 'error',
|
| 170 |
+
'message': error_msg,
|
| 171 |
+
'status': 'error'
|
| 172 |
+
})
|
| 173 |
+
return None
|
| 174 |
|
| 175 |
async def close(self):
|
| 176 |
"""Đóng kết nối Gemini"""
|
| 177 |
+
self.is_active = False
|
| 178 |
+
|
|
|
|
|
|
|
| 179 |
if self.callback:
|
| 180 |
await self.callback({
|
| 181 |
'type': 'status',
|