diff --git a/app/controllers/chat_controller.py b/app/controllers/chat_controller.py index baf68e5..9d31bbb 100644 --- a/app/controllers/chat_controller.py +++ b/app/controllers/chat_controller.py @@ -148,7 +148,8 @@ async def create_chat_handler(chat: ChatCreateRequest, db, redis): "speaker_id":db_character.voice_id, "noise_scale": 0.1, "noise_scale_w":0.668, - "length_scale": 1.2 + "length_scale": 1.2, + "speend":1 } llm_info = { "model": "abab5.5-chat", @@ -318,12 +319,12 @@ async def sct_llm_handler(ws,session_id,response_type,llm_info,tts_info,db,redis audio,sr = tts._base_tts(text=sentence, noise_scale=tts_info["noise_scale"], noise_scale_w=tts_info["noise_scale_w"], - speed=tts_info["length_scale"]) + speed=tts_info["speed"]) else: audio,sr = tts.synthesize(text=sentence, noise_scale=tts_info["noise_scale"], noise_scale_w=tts_info["noise_scale_w"], - speed=tts_info["length_scale"], + speed=tts_info["speed"], target_se=target_se) response_message = {"type": "text", "code":200, "msg": sentence} await ws.send_bytes(audio) #返回音频数据 @@ -499,12 +500,12 @@ async def scl_llm_handler(ws,session_id,response_type,llm_info,tts_info,db,redis audio,sr = tts._base_tts(text=sentence, noise_scale=tts_info["noise_scale"], noise_scale_w=tts_info["noise_scale_w"], - speed=tts_info["length_scale"]) + speed=tts_info["speed"]) else: audio,sr = tts.synthesize(text=sentence, noise_scale=tts_info["noise_scale"], noise_scale_w=tts_info["noise_scale_w"], - speed=tts_info["length_scale"], + speed=tts_info["speed"], target_se=target_se) response_message = {"type": "text", "code":200, "msg": sentence} await ws.send_bytes(audio) @@ -683,17 +684,17 @@ async def voice_call_llm_handler(ws,session_id,llm_info,tts_info,db,redis,asr_re audio,sr = tts._base_tts(text=sentence, noise_scale=tts_info["noise_scale"], noise_scale_w=tts_info["noise_scale_w"], - speed=tts_info["length_scale"]) + speed=tts_info["speed"]) else: audio,sr = tts.synthesize(text=sentence, noise_scale=tts_info["noise_scale"], noise_scale_w=tts_info["noise_scale_w"], - speed=tts_info["length_scale"], + speed=tts_info["speed"], target_se=target_se) text_response = {"type": "llm_text", "code": 200, "msg": sentence} await ws.send_bytes(audio) #返回音频二进制流数据 await ws.send_text(json.dumps(text_response, ensure_ascii=False)) #返回文本数据 - logger.debug(f"llm返回结果: {sentence}") + logger.debug(f"websocket返回: {sentence}") if is_end: logger.debug(f"llm返回结果: {llm_response}") await ws.send_text(json.dumps({"type": "end", "code": 200, "msg": ""}, ensure_ascii=False)) diff --git a/tests/unit_test/chat_test.py b/tests/unit_test/chat_test.py index 7ca335d..97225de 100644 --- a/tests/unit_test/chat_test.py +++ b/tests/unit_test/chat_test.py @@ -10,7 +10,7 @@ import websockets class ChatServiceTest: - def __init__(self,socket="http://127.0.0.1:8001"): + def __init__(self,socket="http://127.0.0.1:7878"): self.socket = socket @@ -119,7 +119,7 @@ class ChatServiceTest: "user_id": self.user_id, "messages": "[{\"role\": \"system\", \"content\": \"我们正在角色扮演对话游戏中，你需要始终保持角色扮演并待在角色设定的情景中，你扮演的角色信息如下：\\n角色名称: 海绵宝宝。\\n角色背景: 厨师，做汉堡\\n角色所处环境: 海绵宝宝住在深海的大菠萝里面\\n角色的常用问候语: 你好啊，海绵宝宝。\\n\\n你需要用简单、通俗易懂的口语化方式进行对话，在没有经过允许的情况下，你需要保持上述角色，不得擅自跳出角色设定。\\n\"}]", "user_info": "{\"character\": \"\", \"events\": [] }", - "tts_info": "{\"language\": 0, \"speaker_id\": 97, \"noise_scale\": 0.1, \"noise_scale_w\": 0.668, \"length_scale\": 1.2}", + "tts_info": "{\"language\": 0, \"speaker_id\": 97, \"noise_scale\": 0.1, \"noise_scale_w\": 0.668, \"length_scale\": 1.2, \"speed\":1.0}", "llm_info": "{\"model\": \"abab5.5-chat\", \"temperature\": 1, \"top_p\": 0.9}", "token": 0} ) @@ -166,7 +166,7 @@ class ChatServiceTest: await websocket.send(message) - async with websockets.connect(f'ws://127.0.0.1:8001/chat/streaming/temporary') as websocket: + async with websockets.connect(f'ws://127.0.0.1:7878/chat/streaming/temporary') as websocket: chunks = read_wav_file_in_chunks(2048) # 读取PCM文件并生成数据块 for chunk in chunks: await send_audio_chunk(websocket, chunk) @@ -222,7 +222,7 @@ class ChatServiceTest: message = json.dumps(data) await websocket.send(message) - async with websockets.connect(f'ws://127.0.0.1:8001/chat/streaming/lasting') as websocket: + async with websockets.connect(f'ws://127.0.0.1:7878/chat/streaming/lasting') as websocket: #发送第一次 chunks = read_wav_file_in_chunks(2048) for chunk in chunks: @@ -272,7 +272,7 @@ class ChatServiceTest: current_dir = os.path.dirname(current_file_path) tests_dir = os.path.dirname(current_dir) file_path = os.path.join(tests_dir, 'assets', 'voice_call.wav') - url = f"ws://127.0.0.1:8001/chat/voice_call" + url = f"ws://127.0.0.1:7878/chat/voice_call" #发送格式 ws_data = { "audio" : "", @@ -341,8 +341,8 @@ def chat_test(): chat_service_test.test_session_content_query() chat_service_test.test_session_update() asyncio.run(chat_service_test.test_chat_temporary()) - # asyncio.run(chat_service_test.test_chat_lasting()) - # asyncio.run(chat_service_test.test_voice_call()) + asyncio.run(chat_service_test.test_chat_lasting()) + asyncio.run(chat_service_test.test_voice_call()) chat_service_test.test_chat_delete()