GithubBackup/BiliNote
1
0
Fork 0
mirror of https://github.com/JefferyHcool/BiliNote.git synced 2026-09-08 00:56:40 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #101 from JefferyHcool/feature/kuaishou

Browse Source 
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:
Jianwu Huang
2025-05-12 15:39:27 +08:00
committed by GitHub
parent 490ee11a85 c1ef98f6d9
commit 1b78fb6417
17 changed files with 744 additions and 353 deletions
Download Patch File Download Diff File Expand all files Collapse all files
BillNote_frontend/package.json
+4
View File
@@ -35,6 +35,10 @@
"lottie-react": "^2.4.1",
"lucide-react": "^0.487.0",
"markdown-navbar": "^1.4.3",
"markmap-common": "^0.18.9",
"markmap-lib": "^0.18.11",
"markmap-toolbar": "^0.18.10",
"markmap-view": "^0.18.10",
"next-themes": "^0.4.6",
"pinyin-match": "^1.2.7",
"react": "^19.0.0",
BillNote_frontend/src/lib/markmap.ts
+8
View File
@@ -0,0 +1,8 @@
import { loadCSS, loadJS } from 'markmap-common'
import { Transformer } from 'markmap-lib'
import * as markmap from 'markmap-view'
export const transformer = new Transformer()
const { scripts, styles } = transformer.getAssets()
loadCSS(styles)
loadJS(scripts, { getMarkmap: () => markmap })
BillNote_frontend/src/pages/HomePage/components/MarkdownHeader.tsx
+21 -1
View File
@@ -1,7 +1,7 @@
'use client'
import { useEffect, useState } from 'react'
import { Copy, Download } from 'lucide-react'
import { Copy, Download, BrainCircuit } from 'lucide-react'
import { Button } from '@/components/ui/button'
import { Select, SelectContent, SelectItem, SelectTrigger } from '@/components/ui/select'
import { Tooltip, TooltipContent, TooltipProvider, TooltipTrigger } from '@/components/ui/tooltip'
@@ -43,6 +43,8 @@ export function MarkdownHeader({
createAt,
showTranscribe,
setShowTranscribe,
viewMode,
setViewMode,
}: NoteHeaderProps) {
const [copied, setCopied] = useState(false)
@@ -122,6 +124,24 @@ export function MarkdownHeader({
{/* 右侧操作按钮 */}
<div className="flex items-center gap-1">
<TooltipProvider>
<Tooltip>
<TooltipTrigger asChild>
<Button
onClick={() => {
setViewMode(viewMode == 'preview' ? 'map' : 'preview')
}}
variant="ghost"
size="sm"
className="h-8 px-2"
>
<BrainCircuit className="mr-1.5 h-4 w-4" />
<span className="text-sm">{viewMode == 'preview' ? '思维导图' : 'markdown'}</span>
</Button>
</TooltipTrigger>
<TooltipContent>思维导图</TooltipContent>
</Tooltip>
</TooltipProvider>
<TooltipProvider>
<Tooltip>
<TooltipTrigger asChild>
BillNote_frontend/src/pages/HomePage/components/MarkdownViewer.tsx
+48 -4
View File
@@ -1,4 +1,4 @@
import { useState, useEffect } from 'react'
import { useState, useEffect, useRef } from 'react'
import ReactMarkdown from 'react-markdown'
import { Button } from '@/components/ui/button.tsx'
import { Copy, Download, ArrowRight, Play, ExternalLink } from 'lucide-react'
@@ -22,6 +22,7 @@ import { useTaskStore } from '@/store/taskStore'
import { noteStyles } from '@/constant/note.ts'
import { MarkdownHeader } from '@/pages/HomePage/components/MarkdownHeader.tsx'
import TranscriptViewer from '@/pages/HomePage/components/transcriptViewer.tsx'
import MarkmapEditor from '@/pages/HomePage/components/MarkmapComponent.tsx'
interface VersionNote {
ver_id: string
@@ -58,6 +59,8 @@ const MarkdownViewer: FC<MarkdownViewerProps> = ({ status }) => {
const retryTask = useTaskStore.getState().retryTask
const isMultiVersion = Array.isArray(currentTask?.markdown)
const [showTranscribe, setShowTranscribe] = useState(false)
const [viewMode, setViewMode] = useState<'map' | 'preview'>('preview')
const svgRef = useRef<SVGSVGElement>(null)
// 多版本内容处理
useEffect(() => {
if (!currentTask) return
@@ -99,7 +102,33 @@ const MarkdownViewer: FC<MarkdownViewerProps> = ({ status }) => {
toast.error('复制失败')
}
}
const alertButton = {
id: 'alert',
title: '测试警告',
content: '⚠️',
onClick: () => alert('你点击了自定义按钮！'),
}
const exportButton = {
id: 'export',
title: '导出思维导图',
content: '⤓',
onClick: () => {
const svgEl = svgRef.current
if (!svgEl) return
// 同上面的序列化逻辑
const serializer = new XMLSerializer()
const source = serializer.serializeToString(svgEl)
const blob = new Blob(['<?xml version="1.0" encoding="UTF-8"?>', source], {
type: 'image/svg+xml;charset=utf-8',
})
const url = URL.createObjectURL(blob)
const a = document.createElement('a')
a.href = url
a.download = 'mindmap.svg'
a.click()
URL.revokeObjectURL(url)
},
}
const handleDownload = () => {
const task = getCurrentTask()
const name = task?.audioMeta.title || 'note'
@@ -167,9 +196,21 @@ const MarkdownViewer: FC<MarkdownViewerProps> = ({ status }) => {
createAt={createTime}
showTranscribe={showTranscribe}
setShowTranscribe={setShowTranscribe}
viewMode={viewMode}
setViewMode={setViewMode}
/>
{/* 中间内容区域：滚动容器 */}
{viewMode === 'map' ? (
<div className="flex w-full flex-1 overflow-hidden bg-white">
<div className={'w-full'}>
<MarkmapEditor
value={selectedContent}
onChange={() => {}}
height="100%" // 根据需求可以设定百分比或固定高度
/>
</div>
</div>
) : (
<div className="flex flex-1 overflow-hidden bg-white py-2">
{selectedContent && selectedContent !== 'loading' && selectedContent !== 'empty' ? (
<>
@@ -289,7 +330,9 @@ const MarkdownViewer: FC<MarkdownViewerProps> = ({ status }) => {
const isFakeHeading = /^(\*\*.+\*\*)$/.test(rawText.trim())
if (isFakeHeading) {
return <div className="text-primary my-4 text-lg font-bold">{children}</div>
return (
<div className="text-primary my-4 text-lg font-bold">{children}</div>
)
}
return (
@@ -431,6 +474,7 @@ const MarkdownViewer: FC<MarkdownViewerProps> = ({ status }) => {
</div>
)}
</div>
)}
</div>
)
}
BillNote_frontend/src/pages/HomePage/components/MarkmapComponent.tsx
+118
View File
@@ -0,0 +1,118 @@
import React, { useEffect, useRef, useState } from 'react'
import { Markmap } from 'markmap-view'
import { transformer } from '@/lib/markmap.ts'
import { Toolbar, ToolbarButton } from 'markmap-toolbar'
import 'markmap-toolbar/dist/style.css'
export interface MarkmapEditorProps {
/** 要渲染的 Markdown 文本 */
value: string
/** 内容变化时的回调 */
onChange: (value: string) => void
/** Toolbar 上要展示的 item id 列表，默认使用 Toolbar.defaultItems */
toolbarItems?: string[]
/** 自定义按钮列表，会依次注册 */
customButtons?: ToolbarButton[]
/** 容器 SVG 的高度，默认为 600px */
height?: string
}
export default function MarkmapEditor({
value,
onChange,
toolbarItems,
customButtons = [],
height = '600px',
}: MarkmapEditorProps) {
const svgRef = useRef<SVGSVGElement>(null)
const mmRef = useRef<Markmap>()
const toolbarRef = useRef<HTMLDivElement>(null)
// 用于跟踪是否处于全屏状态
const [isFullscreen, setIsFullscreen] = useState(false)
// 监听全屏状态变化
useEffect(() => {
const handler = () => {
setIsFullscreen(!!document.fullscreenElement)
}
document.addEventListener('fullscreenchange', handler)
return () => {
document.removeEventListener('fullscreenchange', handler)
}
}, [])
// 进入全屏
const enterFullscreen = () => {
const el = svgRef.current?.parentElement
if (el && el.requestFullscreen) {
el.requestFullscreen()
}
}
// 退出全屏
const exitFullscreen = () => {
if (document.exitFullscreen) {
document.exitFullscreen()
}
}
// 初始化 Markmap 实例 + Toolbar
useEffect(() => {
if (!svgRef.current || mmRef.current) return
const mm = Markmap.create(svgRef.current)
mmRef.current = mm
if (toolbarRef.current) {
toolbarRef.current.innerHTML = ''
const toolbar = new Toolbar()
toolbar.attach(mm)
customButtons.forEach(btn => toolbar.register(btn))
toolbar.setItems(toolbarItems ?? Toolbar.defaultItems)
toolbarRef.current.appendChild(toolbar.render())
}
}, [customButtons, toolbarItems])
// 当 value 变化时，重新渲染数据
useEffect(() => {
const mm = mmRef.current
if (!mm) return
const { root } = transformer.transform(value)
mm.setData(root).then(() => mm.fit())
}, [value])
// 文本输入变化回调（如果你自行添加 textarea 编辑区）
const handleChange = (e: React.ChangeEvent<HTMLTextAreaElement>) => {
onChange(e.target.value)
}
return (
<div className="relative flex h-full flex-col bg-white">
{/* 全屏/退出全屏 按钮 */}
<div className="absolute top-2 right-2 z-20 flex space-x-2">
{isFullscreen ? (
<button
onClick={exitFullscreen}
className="rounded p-1 hover:bg-gray-200"
title="退出全屏"
>
🗗
</button>
) : (
<button onClick={enterFullscreen} className="rounded p-1 hover:bg-gray-200" title="全屏">
🗖
</button>
)}
</div>
{/* 如果需要编辑区，就自己加一个 <textarea> 并把 handleChange 绑上 */}
{/* <textarea value={value} onChange={handleChange} className="mb-2 p-2 border rounded" /> */}
{/* 思维导图区 */}
<svg ref={svgRef} className="w-full flex-1" style={{ height, overflow: 'auto' }} />
{/* 如果你还想保留 markmap-toolbar */}
{/* <div ref={toolbarRef} className="absolute right-2 bottom-2 z-10" /> */}
</div>
)
}
BillNote_frontend/src/pages/HomePage/components/NoteForm.tsx
+30 -6
View File
@@ -150,14 +150,33 @@ const NoteForm = () => {
return
}, [])
useEffect(() => {
const currentTask = getCurrentTask()
const { formData } = currentTask || {}
if (!currentTask) return
const { formData } = currentTask
console.log('currentTask.formData.platform:', formData.platform)
form.reset({
...formData,
extras: formData?.extras || '',
platform: formData.platform || 'bilibili',
video_url: formData.video_url || '',
model_name: formData.model_name || modelList[0]?.model_name || '',
style: formData.style || 'minimal',
quality: formData.quality || 'medium',
extras: formData.extras || '',
screenshot: formData.screenshot ?? false,
link: formData.link ?? false,
video_understanding: formData.video_understanding ?? false,
video_interval: formData.video_interval ?? 4,
grid_size: formData.grid_size ?? [3, 3],
format: formData.format ?? [],
})
}, [currentTaskId])
}, [
// 当下面任意一个变了，就重新 reset
currentTaskId,
// modelList 用来兜底 model_name
modelList.length,
// 还要加上 formData 的各字段，或者直接 currentTask
currentTask?.formData,
])
/* ---- 帮助函数 ---- */
const isGenerating = () => !['SUCCESS', 'FAILED', undefined].includes(getCurrentTask()?.status)
@@ -175,6 +194,7 @@ const NoteForm = () => {
}
const onSubmit = async (values: NoteFormValues) => {
console.log('Not even go here')
const payload: NoteFormValues = {
...values,
provider_id: modelList.find(m => m.model_name === values.model_name)!.provider_id,
@@ -189,6 +209,10 @@ const NoteForm = () => {
const { data } = await generateNote(payload)
addPendingTask(data.task_id, values.platform, payload)
}
const onInvalid = (errors: FieldErrors<NoteFormValues>) => {
console.warn('表单校验失败：', errors)
message.error('请完善所有必填项后再提交')
}
const handleCreateNew = () => {
// 🔁 这里清空当前任务状态
// 比如调用 resetCurrentTask() 或者 navigate 到一个新页面
@@ -222,7 +246,7 @@ const NoteForm = () => {
return (
<div className="h-full w-full">
<Form {...form}>
<form onSubmit={form.handleSubmit(onSubmit)} className="space-y-4">
<form onSubmit={form.handleSubmit(onSubmit, onInvalid)} className="space-y-4">
{/* 顶部按钮 */}
<FormButton></FormButton>
BillNote_frontend/src/utils/request.ts
+20
View File
@@ -3,5 +3,25 @@ const request = axios.create({
baseURL: '/api',
timeout: 10000,
})
function handleErrorResponse(response: any) {
if (!response) return '请求失败，请检查网络连接'
if (typeof response.code !== 'number') return '系统异常'
// 错误码判断
switch (response.code) {
case 1001:
return response.msg || '下载失败，请检查视频链接'
case 1002:
return response.msg || '转写失败，请稍后重试'
case 1003:
return response.msg || '总结失败，可能是模型服务异常'
case 2001:
case 2002:
return Array.isArray(response.data)
? response.data.map(e => `${e.field}: ${e.error}`).join('\n')
: response.msg || '参数错误'
default:
return response.msg || '系统异常'
}
}
export default request
README.md
+1 -1
View File
@@ -3,7 +3,7 @@
<p align="center">
<img src="./doc/icon.svg" alt="BiliNote Banner" width="50" height="50" />
</p>
<h1 align="center" > BiliNote v1.6.1</h1>
<h1 align="center" > BiliNote v1.7.0</h1>
</div>
<p align="center"><i>AI 视频笔记生成工具 让 AI 为你的视频做笔记</i></p>
backend/app/core/__init__.py
View File
backend/app/core/exception_handlers.py
+38
View File
@@ -0,0 +1,38 @@
# app/core/exception_handlers.py
from fastapi import Request, HTTPException
from fastapi.exceptions import RequestValidationError
from fastapi.responses import JSONResponse
from app.utils.logger import get_logger
from app.utils.response import ResponseWrapper
from app.utils.status_code import StatusCode
logger = get_logger(__name__)
def register_exception_handlers(app):
@app.exception_handler(RequestValidationError)
async def validation_exception_handler(request: Request, exc: RequestValidationError):
errors = []
for err in exc.errors():
loc = err.get("loc", [])
field = loc[-1] if loc else "body"
msg = err.get("msg", "参数不合法")
errors.append({"field": field, "error": msg})
return JSONResponse(
status_code=400,
content=ResponseWrapper.error(msg="参数验证失败", code=StatusCode.PARAM_ERROR, data=errors)
)
@app.exception_handler(HTTPException)
async def http_exception_handler(request: Request, exc: HTTPException):
return JSONResponse(
status_code=exc.status_code,
content=ResponseWrapper.error(msg=str(exc.detail), code=StatusCode.FAIL)
)
@app.exception_handler(Exception)
async def global_exception_handler(request: Request, exc: Exception):
logger.exception(f"服务器内部错误: {exc}")
return JSONResponse(
status_code=500,
content=ResponseWrapper.error(msg="服务器内部错误", code=StatusCode.FAIL, data=str(exc))
)
backend/app/db/builtin_providers.json
+8
View File
@@ -38,5 +38,13 @@
"logo": "Gemini",
"api_key": "",
"base_url": "https://generativelanguage.googleapis.com/v1beta/openai/"
},
{
"id": "groq",
"name": "Groq",
"type": "built-in",
"logo": "Groq",
"api_key": "",
"base_url": "https://api.groq.com/openai/v1"
}
]
backend/app/gpt/provider/OpenAI_compatible_provider.py
+2
View File
@@ -13,8 +13,10 @@ class OpenAICompatibleProvider:
@staticmethod
def test_connection(api_key: str, base_url: str) -> bool:
print(api_key)
try:
client = OpenAI(api_key=api_key, base_url=base_url)
client.models.list()
return True
except Exception as e:
backend/app/services/note.py
+64 -14
View File
@@ -1,6 +1,8 @@
import json
from dataclasses import asdict
from fastapi import HTTPException
from app.downloaders.local_downloader import LocalDownloader
from app.enmus.task_status_enums import TaskStatus
import os
@@ -33,6 +35,7 @@ from app.transcriber.whisper import WhisperTranscriber
import re
from app.utils.note_helper import replace_content_markers
from app.utils.status_code import StatusCode
from app.utils.video_helper import generate_screenshot
# from app.services.whisperer import transcribe_audio
@@ -143,7 +146,15 @@ class NoteGenerator:
return new_markdown
except Exception as e:
logger.error(f"截图生成失败：{e}")
raise e
raise HTTPException(
status_code=500,
detail={
"code": StatusCode.DOWNLOAD_ERROR,
"msg": f"截图生成失败",
"error": str(e)
}
)
@staticmethod
def delete_note(video_id: str, platform: str):
@@ -226,8 +237,16 @@ class NoteGenerator:
save_quality=90,
).run()
except Exception as e:
logger.error(f"❌ 下载视频失败，task_id={task_id}，错误信息：{e}")
logger.error(f"Error 下载视频失败，task_id={task_id}，错误信息：{e}")
self.update_task_status(task_id, TaskStatus.FAILED, message=f"下载音频失败：{e}")
raise HTTPException(
status_code=500,
detail={
"code": StatusCode.DOWNLOAD_ERROR,
"msg": f"下载视频失败，task_id={task_id}",
"error": str(e)
}
)
# 没有音频缓存就下载音频（可能同时也带上视频）
if audio is None:
@@ -241,9 +260,17 @@ class NoteGenerator:
json.dump(asdict(audio), f, ensure_ascii=False, indent=2)
logger.info(f"音频下载并缓存成功，task_id={task_id}")
except Exception as e:
logger.error(f"❌ 下载音频失败，task_id={task_id}，错误信息：{e}")
logger.error(f"Error 下载音频失败，task_id={task_id}，错误信息：{e}")
self.update_task_status(task_id, TaskStatus.FAILED, message=f"下载音频失败：{e}")
raise e
raise HTTPException(
status_code=500,
detail={
"code": StatusCode.DOWNLOAD_ERROR,
"msg": f"下载音频失败，task_id={task_id}",
"error": str(e)
}
)
# -------- 2. 转写文字 --------
try:
@@ -259,7 +286,7 @@ class NoteGenerator:
segments=[TranscriptSegment(**seg) for seg in transcript_data["segments"]]
)
except (json.JSONDecodeError, KeyError) as e:
logger.warning(f"⚠️ 读取转录缓存失败，重新转录，task_id={task_id}，错误信息：{e}")
logger.warning(f"Warning 读取转录缓存失败，重新转录，task_id={task_id}，错误信息：{e}")
transcript: TranscriptResult = self.transcriber.transcript(file_path=audio.file_path)
with open(transcript_cache_path, "w", encoding="utf-8") as f:
json.dump(asdict(transcript), f, ensure_ascii=False, indent=2)
@@ -269,9 +296,16 @@ class NoteGenerator:
json.dump(asdict(transcript), f, ensure_ascii=False, indent=2)
logger.info(f"文字转写并缓存成功，task_id={task_id}")
except Exception as e:
logger.error(f"❌ 转写文字失败，task_id={task_id}，错误信息：{e}")
logger.error(f"Error 转写文字失败，task_id={task_id}，错误信息：{e}")
self.update_task_status(task_id, TaskStatus.FAILED, message=f"转写文字失败：{e}")
raise e
raise HTTPException(
status_code=500,
detail={
"code": StatusCode.GENERATE_ERROR, # =1003
"msg": f"转写文字失败，task_id={task_id}",
"error": str(e)
}
)
# -------- 3. 总结内容 --------
try:
@@ -298,9 +332,16 @@ class NoteGenerator:
f.write(markdown)
logger.info(f"GPT总结并缓存成功，task_id={task_id}")
except Exception as e:
logger.error(f"❌ 总结内容失败，task_id={task_id}，错误信息：{e}")
logger.error(f"Error 总结内容失败，task_id={task_id}，错误信息：{e}")
self.update_task_status(task_id, TaskStatus.FAILED, message=f"总结内容失败：{e}")
raise e
raise HTTPException(
status_code=500,
detail={
"code": StatusCode.GENERATE_ERROR, # =1003
"msg": f"总结内容失败，task_id={task_id}",
"error": str(e)
}
)
# -------- 4. 插入截图 --------
if _format and 'screenshot' in _format:
@@ -308,12 +349,12 @@ class NoteGenerator:
markdown = self.insert_screenshots_into_markdown(markdown, self.video_path, image_base_url,
output_dir, _format)
except Exception as e:
logger.warning(f"⚠️ 插入截图失败，跳过处理，task_id={task_id}，错误信息：{e}")
logger.warning(f"Warning 插入截图失败，跳过处理，task_id={task_id}，错误信息：{e}")
if _format and 'link' in _format:
try:
markdown = replace_content_markers(markdown, video_id=audio.video_id, platform=platform)
except Exception as e:
logger.warning(f"⚠️ 插入链接失败，跳过处理，task_id={task_id}，错误信息：{e}")
logger.warning(f"Warning 插入链接失败，跳过处理，task_id={task_id}，错误信息：{e}")
# 注意：截图失败不终止整体流程
# -------- 5. 保存数据库记录 --------
@@ -322,7 +363,7 @@ class NoteGenerator:
# -------- 6. 完成 --------
self.update_task_status(task_id, TaskStatus.SUCCESS)
logger.info(f"✅ 笔记生成成功，task_id={task_id}")
logger.info(f"succeed 笔记生成成功，task_id={task_id}")
# TODO :改为前端一键清除缓存
# if platform != 'local':
# transcription_finished.send({
@@ -335,6 +376,15 @@ class NoteGenerator:
)
except Exception as e:
logger.error(f"❌ 笔记生成流程异常终止，task_id={task_id}，错误信息：{e}")
logger.error(f"Error 笔记生成流程异常终止，task_id={task_id}，错误信息：{e}")
self.update_task_status(task_id, TaskStatus.FAILED, message=str(e))
raise f'❌ 笔记生成流程异常终止，task_id={task_id}，错误信息：{e}'
# 返回结构化错误信息给前端（可以用于日志 + 显示 + 错误定位）
raise HTTPException(
status_code=500,
detail={
"code": StatusCode.FAIL,
"msg": f"笔记生成流程异常终止，task_id={task_id}",
"error": str(e)
}
)
backend/app/transcriber/groq.py
+52
View 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
backend/app/transcriber/transcriber_provider.py
+66 -64
View File
@@ -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)
backend/main.py
+3 -2
View File
@@ -4,6 +4,7 @@ import uvicorn
from starlette.staticfiles import StaticFiles
from dotenv import load_dotenv
from app.core.exception_handlers import register_exception_handlers
from app.db.model_dao import init_model_table
from app.db.provider_dao import init_provider_table
from app.utils.logger import get_logger
@@ -34,12 +35,12 @@ if not os.path.exists(out_dir):
app = create_app()
app.mount(static_path, StaticFiles(directory=static_dir), name="static")
app.mount("/uploads", StaticFiles(directory=uploads_dir), name="uploads")
async def startup_event():
register_handler()
@app.on_event("startup")
async def startup_event():
register_exception_handlers(app)
register_handler()
ensure_ffmpeg_or_raise()
register_handler()
get_transcriber(transcriber_type=os.getenv("TRANSCRIBER_TYPE","fast-whisper"))
init_video_task_table()
init_provider_table()
backend/requirements.txt
BIN
View File
Binary file not shown.