mirror of
https://github.com/JefferyHcool/BiliNote.git
synced 2026-07-07 15:41:22 +08:00
feat(backend): 添加 Groq供应商支持并优化笔记生成流程- 在 builtin_providers.json 中添加 Groq 供应商信息
- 实现 GroqTranscriber 类以支持 Groq 语音转录服务 - 新增异常处理中间件以提高系统稳定性 - 优化笔记生成流程，增加错误处理和日志记录 - 添加思维导图功能和相关组件 -重构 Markdown 查看器以支持切换视图模式
This commit is contained in:
黄建武
52
backend/app/transcriber/groq.py
Normal file
52
backend/app/transcriber/groq.py
Normal file
@@ -0,0 +1,52 @@
|
||||
from abc import ABC
|
||||
import os
|
||||
|
||||
from app.decorators.timeit import timeit
|
||||
from app.models.transcriber_model import TranscriptResult, TranscriptSegment
|
||||
from app.services.provider import ProviderService
|
||||
from app.transcriber.base import Transcriber
|
||||
from openai import OpenAI
|
||||
from dotenv import load_dotenv
|
||||
load_dotenv()
|
||||
|
||||
class GroqTranscriber(Transcriber, ABC):
|
||||
|
||||
|
||||
@timeit
|
||||
def transcript(self, file_path: str) -> TranscriptResult:
|
||||
provider = ProviderService.get_provider_by_id('groq')
|
||||
if not provider:
|
||||
raise Exception("Groq 供应商未配置,请配置以后使用。")
|
||||
client = OpenAI(
|
||||
api_key=provider.get('api_key'),
|
||||
base_url=provider.get('base_url')
|
||||
)
|
||||
filename = file_path
|
||||
|
||||
with open(filename, "rb") as file:
|
||||
transcription = client.audio.transcriptions.create(
|
||||
file=(filename, file.read()),
|
||||
model=os.getenv('GROQ_TRANSCRIBER_MODEL'),
|
||||
response_format="verbose_json",
|
||||
)
|
||||
print(transcription.text)
|
||||
print(transcription)
|
||||
segments = []
|
||||
full_text = ""
|
||||
|
||||
for seg in transcription.segments:
|
||||
text = seg.text.strip()
|
||||
full_text += text + " "
|
||||
segments.append(TranscriptSegment(
|
||||
start=seg.start,
|
||||
end=seg.end,
|
||||
text=text
|
||||
))
|
||||
|
||||
result = TranscriptResult(
|
||||
language=transcription.language,
|
||||
full_text=full_text.strip(),
|
||||
segments=segments,
|
||||
raw=transcription.to_dict()
|
||||
)
|
||||
return result
|
||||
@@ -1,113 +1,115 @@
|
||||
import os
|
||||
import platform
|
||||
from enum import Enum
|
||||
|
||||
from app.transcriber.groq import GroqTranscriber
|
||||
from app.transcriber.whisper import WhisperTranscriber
|
||||
from app.transcriber.bcut import BcutTranscriber
|
||||
from app.transcriber.kuaishou import KuaishouTranscriber
|
||||
from app.utils.logger import get_logger
|
||||
|
||||
logger = get_logger(__name__)
|
||||
|
||||
# 只在Apple平台且设置了环境变量时才导入MLX Whisper
|
||||
class TranscriberType(str, Enum):
|
||||
FAST_WHISPER = "fast-whisper"
|
||||
MLX_WHISPER = "mlx-whisper"
|
||||
BCUT = "bcut"
|
||||
KUAISHOU = "kuaishou"
|
||||
GROQ = "groq"
|
||||
|
||||
# 仅在 Apple 平台启用 MLX Whisper
|
||||
MLX_WHISPER_AVAILABLE = False
|
||||
if platform.system() == "Darwin" and os.environ.get("TRANSCRIBER_TYPE") == "mlx-whisper":
|
||||
try:
|
||||
from app.transcriber.mlx_whisper_transcriber import MLXWhisperTranscriber
|
||||
MLX_WHISPER_AVAILABLE = True
|
||||
logger.info("MLX Whisper 可用，已导入")
|
||||
except ImportError:
|
||||
MLX_WHISPER_AVAILABLE = False
|
||||
logger.warning("MLX Whisper 导入失败，可能未安装或平台不支持")
|
||||
else:
|
||||
MLX_WHISPER_AVAILABLE = False
|
||||
|
||||
logger.info('初始化转录服务提供器')
|
||||
|
||||
# 维护各种转录器的单例实例
|
||||
# 转录器单例缓存
|
||||
_transcribers = {
|
||||
'bcut': None,
|
||||
'kuaishou': None,
|
||||
'mlx-whisper': None,
|
||||
'fast-whisper':None
|
||||
TranscriberType.FAST_WHISPER: None,
|
||||
TranscriberType.MLX_WHISPER: None,
|
||||
TranscriberType.BCUT: None,
|
||||
TranscriberType.KUAISHOU: None,
|
||||
TranscriberType.GROQ: None,
|
||||
}
|
||||
|
||||
def get_whisper_transcriber(model_size="base", device="cuda"):
|
||||
"""获取 Whisper 转录器实例"""
|
||||
if _transcribers['fast-whisper'] is None:
|
||||
logger.info(f'创建 Whisper 转录器实例，参数：{model_size}, {device}')
|
||||
# 公共实例初始化函数
|
||||
def _init_transcriber(key: TranscriberType, cls, *args, **kwargs):
|
||||
if _transcribers[key] is None:
|
||||
logger.info(f'创建 {cls.__name__} 实例: {key}')
|
||||
try:
|
||||
_transcribers['whisper'] = WhisperTranscriber(model_size=model_size, device=device)
|
||||
logger.info('Whisper 转录器创建成功')
|
||||
_transcribers[key] = cls(*args, **kwargs)
|
||||
logger.info(f'{cls.__name__} 创建成功')
|
||||
except Exception as e:
|
||||
logger.error(f"Whisper 转录器创建失败: {e}")
|
||||
logger.error(f"{cls.__name__} 创建失败: {e}")
|
||||
raise
|
||||
return _transcribers['whisper']
|
||||
return _transcribers[key]
|
||||
|
||||
# 各类型获取方法
|
||||
def get_groq_transcriber():
|
||||
return _init_transcriber(TranscriberType.GROQ, GroqTranscriber)
|
||||
|
||||
def get_whisper_transcriber(model_size="base", device="cuda"):
|
||||
return _init_transcriber(TranscriberType.FAST_WHISPER, WhisperTranscriber, model_size=model_size, device=device)
|
||||
|
||||
def get_bcut_transcriber():
|
||||
"""获取 Bcut 转录器实例"""
|
||||
if _transcribers['bcut'] is None:
|
||||
logger.info('创建 Bcut 转录器实例')
|
||||
try:
|
||||
_transcribers['bcut'] = BcutTranscriber()
|
||||
logger.info('Bcut 转录器创建成功')
|
||||
except Exception as e:
|
||||
logger.error(f"Bcut 转录器创建失败: {e}")
|
||||
raise
|
||||
return _transcribers['bcut']
|
||||
return _init_transcriber(TranscriberType.BCUT, BcutTranscriber)
|
||||
|
||||
def get_kuaishou_transcriber():
|
||||
"""获取快手转录器实例"""
|
||||
if _transcribers['kuaishou'] is None:
|
||||
logger.info('创建快手转录器实例')
|
||||
try:
|
||||
_transcribers['kuaishou'] = KuaishouTranscriber()
|
||||
logger.info('快手转录器创建成功')
|
||||
except Exception as e:
|
||||
logger.error(f"快手转录器创建失败: {e}")
|
||||
raise
|
||||
return _transcribers['kuaishou']
|
||||
return _init_transcriber(TranscriberType.KUAISHOU, KuaishouTranscriber)
|
||||
|
||||
def get_mlx_whisper_transcriber(model_size="base"):
|
||||
"""获取 MLX Whisper 转录器实例"""
|
||||
if not MLX_WHISPER_AVAILABLE:
|
||||
logger.warning("MLX Whisper 不可用，请确保在Apple平台且已安装mlx_whisper")
|
||||
raise ImportError("MLX Whisper 不可用，请确保在Apple平台且已安装mlx_whisper")
|
||||
|
||||
if _transcribers['mlx-whisper'] is None:
|
||||
logger.info(f'创建 MLX Whisper 转录器实例，参数：{model_size}')
|
||||
try:
|
||||
_transcribers['mlx-whisper'] = MLXWhisperTranscriber(model_size=model_size)
|
||||
logger.info('MLX Whisper 转录器创建成功')
|
||||
except Exception as e:
|
||||
logger.error(f"MLX Whisper 转录器创建失败: {e}")
|
||||
raise
|
||||
return _transcribers['mlx-whisper']
|
||||
logger.warning("MLX Whisper 不可用，请确保在 Apple 平台且已安装 mlx_whisper")
|
||||
raise ImportError("MLX Whisper 不可用")
|
||||
return _init_transcriber(TranscriberType.MLX_WHISPER, MLXWhisperTranscriber, model_size=model_size)
|
||||
|
||||
# 通用入口
|
||||
def get_transcriber(transcriber_type="fast-whisper", model_size="base", device="cuda"):
|
||||
"""
|
||||
获取指定类型的转录器实例
|
||||
|
||||
|
||||
参数:
|
||||
transcriber_type: 转录器类型，支持 "fast-whisper", "bcut", "kuaishou", "mlx-whisper"(仅Apple平台)
|
||||
model_size: 模型大小，whisper 和 mlx-whisper 特有参数
|
||||
device: 设备类型，whisper 特有参数
|
||||
|
||||
transcriber_type: 支持 "fast-whisper", "mlx-whisper", "bcut", "kuaishou", "groq"
|
||||
model_size: 模型大小，适用于 whisper 类
|
||||
device: 设备类型（如 cuda / cpu），仅 whisper 使用
|
||||
|
||||
返回:
|
||||
对应类型的转录器实例
|
||||
"""
|
||||
logger.info(f'获取转录器，类型: {transcriber_type}')
|
||||
if transcriber_type == "fast-whisper":
|
||||
whisper_model_size = os.environ.get("WHISPER_MODEL_SIZE",model_size)
|
||||
logger.info(f'请求转录器类型: {transcriber_type}')
|
||||
|
||||
try:
|
||||
transcriber_enum = TranscriberType(transcriber_type)
|
||||
except ValueError:
|
||||
logger.warning(f'未知转录器类型 "{transcriber_type}"，默认使用 fast-whisper')
|
||||
transcriber_enum = TranscriberType.FAST_WHISPER
|
||||
|
||||
whisper_model_size = os.environ.get("WHISPER_MODEL_SIZE", model_size)
|
||||
|
||||
if transcriber_enum == TranscriberType.FAST_WHISPER:
|
||||
return get_whisper_transcriber(whisper_model_size, device=device)
|
||||
elif transcriber_type == "mlx-whisper":
|
||||
whisper_model_size = os.environ.get("WHISPER_MODEL_SIZE",model_size)
|
||||
|
||||
elif transcriber_enum == TranscriberType.MLX_WHISPER:
|
||||
if not MLX_WHISPER_AVAILABLE:
|
||||
logger.warning("MLX Whisper 不可用，回退到 fast-whisper")
|
||||
return get_whisper_transcriber(whisper_model_size, device=device)
|
||||
return get_mlx_whisper_transcriber(whisper_model_size)
|
||||
elif transcriber_type == "bcut":
|
||||
|
||||
elif transcriber_enum == TranscriberType.BCUT:
|
||||
return get_bcut_transcriber()
|
||||
elif transcriber_type == "kuaishou":
|
||||
|
||||
elif transcriber_enum == TranscriberType.KUAISHOU:
|
||||
return get_kuaishou_transcriber()
|
||||
else:
|
||||
logger.warning(f'未知转录器类型 "{transcriber_type}"，使用默认 whisper')
|
||||
whisper_model_size = os.environ.get("WHISPER_MODEL_SIZE",model_size)
|
||||
return get_whisper_transcriber(whisper_model_size, device)
|
||||
|
||||
elif transcriber_enum == TranscriberType.GROQ:
|
||||
return get_groq_transcriber()
|
||||
|
||||
# fallback
|
||||
logger.warning(f'未识别转录器类型 "{transcriber_type}"，使用 fast-whisper 作为默认')
|
||||
return get_whisper_transcriber(whisper_model_size, device=device)
|
||||
|
||||
Reference in New Issue
Block a user