dale_yxc/ragflow_python
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

增加加了博物馆展品清单数据库及对前端获取展品清单、展品详细的接口，增加了QWenOmni多模态大模型的支 持（主要为了测试）,增加了本地部署大模型支持（主要为了测试，在autoDL上），修正了TTS生成和返回前端的逻辑与参数，增加了判断用户问题有没有在知识库中检索到相关片段、如果没有则直接返回并提示未包含

Browse Source
This commit is contained in:
qcloud
2025-04-08 08:41:07 +08:00
parent a5e83f4d3b
commit 330976812d
12 changed files with 593 additions and 158 deletions
Download Patch File Download Diff File Expand all files Collapse all files

47
rag/llm/tts_model.py
View File

@@ -133,7 +133,8 @@ class QwenTTS(Base):
if parts[0] == 'cosyvoice-v1':
self.is_cosyvoice = True
self.voice = parts[1]
def tts(self, text):
# 参数stream_format 为产生的tts 音频数据格式, mp3 wav pcm
def tts(self, text, sample_rate=8000,stream_format="mp3"):
from dashscope.api_entities.dashscope_response import SpeechSynthesisResponse
if self.is_cosyvoice is False:
from dashscope.audio.tts import ResultCallback, SpeechSynthesizer, SpeechSynthesisResult
@@ -143,7 +144,7 @@ class QwenTTS(Base):
from dashscope.audio.tts_v2 import ResultCallback, SpeechSynthesizer, AudioFormat #, SpeechSynthesisResult
from dashscope.audio.tts import SpeechSynthesisResult
from collections import deque
print(f"--QwenTTS--tts_stream begin-- {text}") # cyx
# print(f"--QwenTTS--tts_stream begin-- {text}") # cyx
class Callback(ResultCallback):
def __init__(self) -> None:
self.dque = deque()
@@ -237,12 +238,52 @@ class QwenTTS(Base):
format="mp3")
else:
self.callback = Callback_v2()
print(f"--tts {sample_rate} {stream_format}")
if sample_rate == 8000:
if stream_format == 'mp3':
format = AudioFormat.MP3_8000HZ_MONO_128KBPS
elif stream_format == 'pcm':
format = AudioFormat.PCM_8000HZ_MONO_16BIT
elif stream_format == 'wav':
format = AudioFormat.WAV_8000HZ_MONO_16BIT
else:
format = AudioFormat.MP3_8000HZ_MONO_128KBPS
elif sample_rate == 16000:
if stream_format == 'mp3':
format = AudioFormat.MP3_16000HZ_MONO_128KBPS
elif stream_format == 'pcm':
format = AudioFormat.PCM_16000HZ_MONO_16BIT
elif stream_format == 'wav':
format = AudioFormat.WAV_16000HZ_MONO_16BIT
else:
format = AudioFormat.MP3_16000HZ_MONO_128KBPS
elif sample_rate == 22050:
if stream_format == 'mp3':
format = AudioFormat.MP3_22050HZ_MONO_256KBPS
elif stream_format == 'pcm':
format = AudioFormat.PCM_22050HZ_MONO_16BIT
elif stream_format == 'wav':
format = AudioFormat.WAV_22050HZ_MONO_16BIT
else:
format = AudioFormat.MP3_22050HZ_MONO_256KBPS
elif sample_rate == 44100:
if stream_format == 'mp3':
format = AudioFormat.MP3_44100HZ_MONO_256KBPS
elif stream_format == 'pcm':
format = AudioFormat.PCM_44100HZ_MONO_16BIT
elif stream_format == 'wav':
format = AudioFormat.WAV_44100HZ_MONO_16BIT
else:
format = AudioFormat.MP3_44100HZ_MONO_256KBPS
# format=AudioFormat.MP3_44100HZ_MONO_256KBPS
else:
format = AudioFormat.MP3_44100HZ_MONO_256KBPS
self.synthesizer = SpeechSynthesizer(
model='cosyvoice-v1',
# voice="longyuan", #"longfei",
voice = self.voice,
callback=self.callback,
format=AudioFormat.MP3_44100HZ_MONO_256KBPS
format=format
)
self.synthesizer.call(text)