{ "text": "Hello,大家好,我是周日金。AI這兩年離咱們普通用戶越來越近。現在甚至不需要單獨買一台服務器,我們就可以直接把手頭閒置在Mac,變成一台屬於自己的本地AI服務器。\n這次視頻我會從零開始帶大家把Olama、OpenWeb UI、REG、OCR、Whisper.cpp和N8N一步步串起來。這套方案的重點就是,模型和大部分數據都運行在自己本地的Mac上,也不要求特別高的配置。\n如果大家对这类玩法感兴趣\n后面还可以继续带大家折腾本地图片和视频生成\n本地AI编程以及把AI介入Home Assistant\n通过自然语言控制家里的智能设备\n视频正式开始前先求一波一进三连加关注支持一下\n这样以后有新的教程更新也能第一时间受到推送\n非常感谢\n接下来我们就从我的这台Mac mini开始\n在开始下载模型之前\n之前我们需要先在mac上安装olama\n你可以把它理解为让普通用户可以在本地运行AI大模型的运行环境\n直接前往olama的官网下载对应的安装包\n然后拖动到macOS用程式文件夹就可以了\n平时使用chatgpt cloud这类AI服务时\n模型都是运行在云端服务器上\n而olama要做的事情就是把像deepseek签问lama这类开源模型下载安装到你的mac上\n让他们直接利用本机的cpu和gpu运行\n安装好Olama并打开之后\n下一步就是下载模型\n不过这里需要注意啊\n本地AI和ChatGPT这类云竿的模型\n最大的区别\n就是模型大小会直接影响运行体验\n所以不能盲目追求越大的模型就越好\n我这里使用的是一台M4入门款的Mac mini\n采用16GB的同一列存加256GB的固态硬盘\n确认Olama的服务在正常运行后\n我们在MacOS的中段App中\n输入这三条命令\n它们分别负责不同的功能\n第一个是大约模型\n也就是本地AI的大脑\n负责日常聊天\n内容生成\n总业资料等任务\n其中4B代表40亿参数\nKill4KM代表经过4bit量化\n可以在降低模型状的同时\n保持不错的运行效果\n是目前比较适合普通Mac的选择\n第二个是向量模型\n主要用于知识库检索\n如果你以后想让AI读取自己的PDF\n文档视频稿件\n然后根据这些资料回答问题\n就需要通过Embedding模型\n先帮AI找到相关内容\n第三个是欧塞亚文字试园模型\n负责处理图片\n扫描版PDF等内容\n把图片里的文字提许出来\n再交给语言模型进行分析\n这里之所以选择千问的AR模型\n是因为千问是目前比较成熟的\n开源大模型之一\n对中文理解和中文生成比较友好\n其次是它提供的模型规格丰富\n从4B到更大的参数版本\n可以适配不同配置的Mac\n还有一个是\n它在欧拉玛里的支持比较完善\n安装和电用也都比较简单\n当然除了千问\nDeepSeq,Lama等模型也都可以运行\n大家可以根据自己的需求进行选择\n那至于不同的Mac应该选择多大的模型\n主要看新变性能和统一内存的大小\n8GB内存的Mac建议先从4B级别模型开始\n16GB内存运行4B会相对轻松\n也可以在实际内存压力云雪情况下尝试9B\n而24GB,32GB以及更高内存的Mac\n则可以继续尝试更大的规格\n但具体体验还会受到上下文长度和后台任务的影响\n这里需要注意啊\n模型标签不能随便改数字\n如果继续使用千万3.5\n目前Olama还提供9B\n27B 35B等更大规格\n具体名称一定以Olama模型库\n实际存在的标签为准\n那咱们还是先把本地AI服务器跑起来\n后续再根据实际需求升级模型\n接下来我们在Microsoft中的App中\n输入以下命令\nOlama List\n查看是否已经全部下载完成\n如果前面下载的模型都正常列出来\n就可以来简单试一下\n模型能不能在本地正常运行\n输入制传命令\nolama run\n千万3.5\n4b-q4km\n然后随便发一句话\n比如你好\n如果能正常给出回复就没问题了\n随后输入制传命令结束对话\nBye\n之后我们就可以停止运行\n以释放资源\n输入制传命令\nolama stop\n千万3.5\n4b-q4km\n然后再输入制传命令\nolama ps\n查看这个AR大模型是否人在运行\n此时应该为空\n那么就说明我们的大模型已经没有在运行了\n接下来咱们就要开始准备部署\n家庭局网AI对话界面了\n前面我们已经通过Olama把AI模型运行在Mac上\n但是目前还需要通过中端的进行操作\n对于日常使用来说不太方便\n所以接下来我们就要给它搭建一个\n类似ChatGP的网页界面\n这里使用OpenWebUI\n也就是让Olama负责运行AI模型\nOpenWebUI负责提供交互界面\n那部署完成后\n家里的手机平板电脑都可以通过浏览器\n访问这台Mac mini\n直接使用自己的本地AI\n能考虑到中国大陆用户的网络环境\n这里直接通过Python安装OpenWeb UI\n同时Olama继续只监听本机地址\n不会直接暴露到网络中\n本教程固定使用OpenWeb UI 0.11.0版本\n不使用latest或滚动更新版本\n避免版本变化导致操作流程不一炸\n首先准备Python环境\n如果你的Mac已经安装Homebrew\n则可以输入brew install python at 3.11\n如果没有安装Homebrew\n也可以直接向Python官方网站下载安装Python 3.1对应的MicroOS安装包\n后面如果会继续折腾这类工具\n建议提前装好Homebrew\n管理和升级会方便不少\n接下来创建OpenWeb UI独立运行环境\n避免影响系统里的其他Python项目\n后续相关代码我不再一一进行解释\n如果有看不懂可以查看代码上方的文本注释\n然后通过轻滑镜像安装OpenWeb UI\n安装完成后\n第一次启动建议只允许本机访问\n先完成账号和安全设置\n等命令跑起来以后\n不要关闭这个中端窗口\n因为OpenWeb UI的服务就运行在这里\n然后流量器打开\n127.0.0.1\n网号3000\n第一次进入需要创建账号\n第一个账号就会自动成为管理员\n进入管理后台后\n前往设置身份验证\n把newsign apps关闭\n也就是不允许新用户自由注册\n同时暂时不要安装来源不明的\n第三方Tools,Functions或Pipelines\n接着就可以在流量器打开\n127.0.1\n网号11434\n-API-version\n确认Olama本地API正常响应\n回到OpenWeb UI\n再检查之前下载的\n千万3.5 4B-TU4KM\n是否已经出现在母型列表当中\n确认没有问题后\n在中端APP中按Ctrl-C\n停止OpenWeb UI\n接下来开放家庭局域网\n实际在系统设置网络中\n查看Mac mini的局域网IP\n实际操作需要替完成你自己的IP\n重新启动OpenWeb UI\n现在Mac mini本机可以继续访问127.0.0.1\n网号3000\n并使用刚才创建的管理员账号进行登录\n通过家庭网络里的手机\n电脑则可以通过Mac mini.ipd设计\n加段稿号3000来访问\n同样使用刚才的账号来登录\n这时候可以随便问你问题\n看看局域网访问和模型交用是否正常\n如果其他设备都无法打开\n依次可以清查几个地方\n最后提醒一下\n这里搭建的家庭局域网AI服务器\n不会把服务暴露到公网\n因此不需要进行路由器短口转发\n也不要直接把OpenWeb AI开放到互联网环境\n而这样配置后保持终端App不关闭\nAI模式运行在自己的Mac上\n数据只在家庭网络里进行流转\n同时手机平板和电脑都可以共享这一台本地AI服务器\n前面OpenWeb UI已经可以这张访问了,接下来我们再给整个家庭AI服务器建立一个统一的API接口。\n这样以后不管是N8N手机快捷指令\n还是其他支持OpenAI API的程序\n都统一连接OpenWeb UI\n不需要直接访问底层的奥拉玛\n这样的好处是\n模型、照顽和权限都可以在OpenWeb UI这一层统一管理\n后面更换模型时\n也不用重新修改每一个应用\n实际上咱们回到OpenWeb UI\n点击左下角头像依次进入设置\n身份验证找到API密钥并打开\n如果界面提供API密钥端点限制\n也一并开启\n然后填写这个\n这样生成的key就只允许访问模型列表和聊天接口\n接下来进入账号API MIAO\n点击右侧的显示创建新的API MIAO\n生成MIAO通常会以SK开头\n这个MIAO相当于访问我们家庭iF5系的密码\n不要直接公开或者分享出去\n如果后面准备长期让N8N这类的自动化工具使用\n建议专门创建一个普通的本地账号\n再用这个账号生成API MIAO\n不要长期使用管理员账号的MIAO\n因为MIAO会继续创建这个账号的权限\n而且一个账号通常只保留一个API密钥\n接着我们来测试接口是否正常\n打开一个新的终端APP窗口\n把下面的命令中的SK换成刚才生成的密钥\n如果接口正常\n终端会访问当前OpenWeb UI可以调用的模型列表\n这里可以顺便记一下返回结果里模型队的真实ID\n因为后面其他地方对用AI时就需要填写这个名称\n接下来再来测试聊天接口\n同样的记得把SK开头的这串换成刚才生成的密钥\n而Model后面需要填写刚才模型列表里面\n实际反规的ID\n如果你下载的是其他签问或者其他模型\n就替完成对应的模型\n如果最后反规Json你能够看到API正常\n就说明这一套接口已经可以正常使用了\n签问3.5在开启思考时\n反规内容里可能还会出现\nReasoning Content这一类字段\n这是正常现象\n到这里我们就相当于给整个家庭\nAI服务器建了一个统一的入口\n最底层人是Olama负责运行模型\n上面的OpenWeb UI\n统一管理模型和API\n在网上则可以接N8N手机快点指令\n或者其他应用\n后面不管增加多少个客户端\n都尽量连接OpenWeb UI\n而不是让每台设备直接访问欧拉玛\n整个架构会更容易管理\n前面我们已经把本地AI和聊天的API都搭好了\n接下来就再补上一个非常熟能功能\nCell文档知识库\n也就是REG\nREG可以列成\n先从你的资料里找内容\n再交给AI声称回答\n这样以后不管是自己的PDF\n说明书\n工作文档\n还是视频稿\n都可以放进知识库\n让AI尽量根据这些资料来回答\n而不是只靠模型本身的记忆\n首先进入OpenWeb UI试试里的文档页面\n轻入模型引擎选择Olama\nAPI MIO可以留空\n地址填写127.0.0.1\n冒号11434\n轻入模型输入\n我们在视频一开始就下载好的\n轻入3-embedding0.6b\n而这里顺便解释一下\n这个模型不负责聊天\n它的工作是把文档\n转完成便于检索的向量数据\n只要你提问以后\n系统才能快速找到\n和问题最相关的资料\n其他参数可以先按照这一套来设置\n大部分是我认的\n这些参数主要是在减速效果\n和上下文本占用之间\n做一个比较稳妥的平衡\n比如快大小可以列列成\n每次把文档切成多大片段\n快重叠则是在相邻片段之间\n保留的部分重复内容\n避免一句话刚好被重点切断\nTableK代表每次提问\n最多取回多少个相关片段\n这里先设置为3\n并不是越大越好\n因为塞进太多资料\n反而会占模型有限的上下文空间\n接下来我们创建一个非常简单的测试文件\n方便确认RIG是否真的正常工作\n用任意文本编定器新建一个TXT文本\n如果你用的是MicroOS自带的文本编辑App\n记得在菜单里选择制作手文本\n比如里面写上这些内容\n然后保存为家庭AI测试资料.txt\n回到OpenWeb UI\n找到左侧边来的工作空间\n再点击顶部的知识库\n最后点击右上角的创建\n名称填写家庭AI测试库\n目标可以简单写真\n用于测试本地RIG文档检索和应用功能\n知识库创建完之后\n再点击右上角的加号选择上传文件\n把刚才的TXD文件上传进去\n后面也可以上传带有文字层的PDF\nDOCX或Markdown文档\n上传完之后\n在新兑化的聊天输入框里\n通过警号手动附加这个知识库\n而我这里为了教程演示方便\n先手动选择\n因为小参数模型主动判断\n什么时候动用知识库有时候不够稳定\n然后我们可以问一个问题\n只依据知识库回答\n项目代号和设备维护时间分配是什么\n请给出引用\n正常情况下\n耐心等待一会儿\n他应该回答项目代号是70\n设备维护时间是每月第一周六上午10点\n而且回答旁边能够看到对应的资料引用\n那么这套最基础的REG知识库就算大奖成功了\n不过这样配置下来\n你可能会发现还有一个地方容易踩坑\n如果你的PDF是扫描件\n本身没有可以复制的文字层\n就不要拿它来测试REG了\n因为知识库首先需要能够读取里面的文字\n才能进行后续的处理\n所以这种文件就需要先经过OCR\n把图片里的文字设备成TXC或者Markdown\n再上传到知识库来进行处理\n接下来我们就继续把本地OCR的功能也接进来\n前面我们已经让AI能够读取自的文档\n接下来再补上图片理解和OCR\n也就是让这台本地AI不只能读文字\n还可以看图片\n那这里先区分两个概念\n普通图片理解更适合看一张图里有什么\n界面再显示什么\n再顺便识别其中的文字\nOCR则更专注于把图片里的文字\n表格或者图表内容尽可能提取出来\n先在OpenWeb UI里选择我们前面使用的\n切纹3.5 4B-KL4KM\n如果你的Mac配置更高\n也可以用9B的版本\n上传一张同时包含物体和文字图片\n然后输入\n先描述画面内容\n再逐项提取可见文字\n不确定的部分标注为不确定\n不需要自信补圈\n这里可以分别拿产品包装\n软件界面截图\n中文表格\n或者已经做好隐私透明的发票和数据照片来测试一下\n这样可以比较直观的判断出\n模型到底是理解了图片内容\n还是只是把里面的文字给识别出来\n如果你出现类似我这样的提示\n就前往OpenWeb UI的设置\n模型\n找到刚才你前面使用的模型\n进入编辑界面\n找到高级参数下的NUM CTX OLAMA\n针对我16GB的Mac mini\n将数字更改为8192\n然后重复上的操作就可以了\n如果我们的目标不是理解图片\n而是尽可能把图片里的文字给提取出来\n就可以使用我们前面已经下载好的GLM OCR\nOCR模型运行前\n如果较大的千万模型还占着同一印寸\n先可以把它停掉\n接下来我们准备好需要识别图片\n放到一个统一的文件夹中\n比如我这个目录下\n对于我们Mac仿达中的用户文件夹\nAI-TEST\n这里手动新建一个TEST的文件夹就可以了\n对于需要识别不同的内容图片\n比如普通收据表格图表\n都可以使用不同的命令\n注意命令结尾的文件名\n需要换成你自己图片的文件名\n第一次测试时\n建议每次只处理一张图片\n先确认整套工作流能够稳定运行\n如果遇到扫描PDF\n也不用一开始就折腾完整的PDF解析工具\n最简单办法就是\n先用MicOS自带的预览打开PDF\n把其中页导成PN器\n再交给GLM OCR进行处理\n等确认单页OCR没问题后\n再考虑批量处理整份文档\n识别出来内容还可以继续利用\n保存成Tek-C或者Markdown\n再上传到我们上一节创建的知识库\n然后交给签问做摘要\n自端整理或者检索问答\n那么整个流程就传起来了\n如果OCR出现空白重复或者乱码\n可以先更新OLAMA\n然后每次只处理一张图片\n图片过大时\n可以把长辨锁到大约1600至2200像素\n并优先使用PNG或者GPG\n同时检查原图有没有明显模糊\n倾斜反光\n这些都会直接影响识别效果\n好\n那除了文字和图片之外,本DI还有一个非常实用的场景,就是处理视频和录音。比如把会议录音转成文字,把视频自动生成字幕,或者把自己的素材库快速整理成可搜索的文本。\n这里我们使用Whisper.cpp\n它是Whisper模型C++实现的\n针对Apple Silicon提供Mental加速\n可以直接利用Mac的GPU进行本地转写\n首先安装需要用到的工具\nFMMP1G负责处理音频格式\nWhisper.cpp负责语音识别\n如果你的Mac已经安装HomeBrew\n可以直接输入代码来进行安装\n如果你仍然没有安装HomeBrew\n安装Whisper.cpp可以走原码编译路线\n不过对于普通用户来说\n这一步会比HomeBrew更复杂一些\n还需要提前安装Xcode command line tools\n这里就不展开了\n安装完整后\n可以通过下面两个命令来确认是否正常\n接下来下载Whisper模型\n我这里选择的是这个\n这是一个支持中文和多语言的模型\n相比完整LARGE模型体积更小\n同时保留了比较好的转眼效果\n更适合Mac本地运行\n我们先创建存放模型和临时文件的目录\n然后下载模型\n下载完整后再计算一次SHA-1的值\n方便和Whisper.cpp模型文档公开的接应值进行核对\n模型准备好后就可以开始处理视频\n假设你的视频文件位于这个文件夹\n也就是MicroOS用户目录的影片文件夹当中\n大家根据自己的实际文件存储位置进行修改\n首先使用FMMP1G把视频里的音频转换成Whisper更适合处理的格式\n这些几个参数简单介绍一下\nVN表示去掉视频只保留音频\nAR16表示转换成16000Hz的采样率\nAC1表示单身道\nPCM-S16LE则是转换成Whisper更容易处理的WAV格式\n接下来运行Whisper同时生成文字稿和字幕文件\n-m用来指定我们前面下载的Whisper模型\n-f指定需要转写的音频文件\n-lzh告诉我们模型这段文字主要内容是中文\n-prompt则可以提前提供一些容易识别错误的中文名词\n比如Mac mini,Apple Silicon,Olama和签问\n对于科技类视频有其有用\n比如这里我就加上了我的Lookup Weekly\n后面的-otxt-oosrt和-ovtt分别让它同时生成TXT文字稿\nSRT字幕和VTT字幕\n最后通过-off指定输出文件的位置和名称\n运行完成后会生成这些文件\n其中TXT是纯文字稿\n适合后续交给AI总结或者整理\nSRT是常见的字幕格式\n可以直接导入紧接软件\nVTT是主要用于网页播放器的字幕\n如果是8GB的Mac\n也可以先尝试这里的Large-V3-Turbo Q50\n电话模型\n如果实际转写时内存压力较大或者速度不理想\n再换成体积更小的small多语言模型\n不要下载来带EAN后制的英文专用模型\n因为它只针对英文优化\n如果需要中文转写\n也不要开启翻译参数\n否则模型会把内容翻译成英文\n而不是保留原始的语言\n那到这里呢\n整套本地i工作流基本已经闭环\n视频和录音可以通过whisper\n转写文字\n再交给前文做总结\n整理\n最后也可以存入前面搭建的知识库\n实现自己本地的内容管理系统\n前面我们已经分别完成了本地云转写\nOpenWebUI API和签问调用\n这就是使用N8N搭建一条自动化工作流\n上传音频或者视频后\nN8N自动调用Whisper来转写\n再把转写稿交给签问来整理\n最后返回摘要提纲标题和症结建议\n不需要每次手动复制粘贴\n首先把前面安装好的Whisper.cpp\n变成一个可以被其他程序调用的本地API\n先尝试允许Mac本机访问\n保持这个中端窗口开启\n再新开一个中端APP窗口进行测试\n如果能够正常返回转写结果\nWhisper API就已经运行成功\n后面N8N会运行在Darker容器里\n如果出现无法连接Whisper的情况\n可以按Ctrl加C停止服务\n再这样改\n这里需要特别注意\nWhisper server本身没有用户登陆和全线验证的功能\n所以使用0.0.0.0时\n只建议放在可信的家庭局网内\n不要进行路由器端口转发\n也不要把8178断口直接暴露到公网\n接下来安装N8N\n这里使用Docker Desktop来运行\n并固定N8N版本\n避免后续更新导致教程界面或者节点配置发生变化\n先从Docker官网下载安装Docker Desktop\n拖动到应用程序文件夹并启动\n确认Docker正常运行后\n再打开中端App输入下面的命令\n接下来生成一段随机的加密密钥\n把生成的结果替换到下面的XXX\n然后我们访问127.0.0.1\n冒号5678\n第一次进入需要创建N8N的本地管理员账号\n接下来我们来新建一条workflow工作流\n这套流程一共只有四个节点\nWebhook\nhttprequest whisper转写\nhttprequest签文整理\nrespond to webhook\n第一个是webhook节点\n它相当于整条工作流的入口\n以后只要把视频post到这个地址\n工作流就会自动开始\n外部hook的具体设置如下\n没提到的设置保持不变即可\n第二个节点添加http request\n名称可以改成Whisper转写\n它负责把刚刚上传的音视频文件\n交给Mac上的Whisper server\n第二个节点HTTPRequest的具体设置如下\n没提到的设置保持不变即可\n值得注意的是N8MBinaryFile这一类型中\nInputDataFieldName需要填写WebHook节点\n实际收到的二进制字段名称\n点击前一个WebHook节点\n在Output的Binary中进行查看\n如果显示为Data那就填写为Data\n如果显示为Data0那就填写为Data0\n这个名称必须完全一致\n否则会出现TheItem has no binary field的报错\n第三个节点继续添加HTTPRequest\n名称改成前文整理\n这个节点负责把Whisper返回完整转写稿\n交给前面搭建好的OpenWebUI API\n第三个节点HTTPRequest\n具体设置如下\n没提到的设置保持不便即可\n接下来回到前面的JSON输入框\n把整个字段切换到Expression模式\n再粘贴下面的表达式\n它的逻辑很简单\nN8N自动读取Whisper转写节点里面的text\n再连同预设提示词一起发给亲吻\n最后添加Respond to Webhook节点\n它负责把整条工作流结果返回给定用者\nRespond with选择Json\n然后把Respond body切换成Expression模式\n粘贴下面的表达式\nRespond code保持200即可\n工作流设置完成以后\n先不要急着正式启用\n在N8N里点击Listen for Test Event\n让Webhook等待一次测试请求\n然后回到终端执行下面的命令\n注意要把XAI Token后面的站位内容\n替换成你前面实际设置的随机自书串\n如果整个流程正常\n你最终应该会得到类似这样的结果\n测试成功以后回到N8N的工作流程页面\n点击右上角的Publish发布\n发布后就不需要每次先点击Lessens for Test Event了\n正式Webhook地址也会从Webhook Test AI Video Notes\n变成Webhook AI Video Notes\n以后只要把视频传到这个接口\n整套流程就会自动变成Whisper\n在本地把视频转成文字\n亲吻再在本地完成摘要提纲标题和章节建议\n最后把转写稿和整理结果一起返回\n整个过程中的语音和文本内容\n都可以在自己的Mac上完成处理\n到这里整套工作流已经跑通了\n不过直接使用Curl调用正式WebHook时\n终端返回的是标准的Json\n所以会看到大量的写杠N\n引号和转译字符\n比如写杠N其实就是换\n并不是烂码\n只是Json保存Markdown和多项文字时的正常格式\n为了让最终结果更加适合阅读\n我们可以再用JQ\n把Json里的转写稿和千万之立结果提取出来\n如果已经安装Homebrew\n则可以执行Brew Install JQ\n安装完整后\n把刚才电用正式工作流命令稍微改一下\n哦对了如果你的Mac没有安装Homebrew\n也不一定非要额外安装JQ\n前面已经准备了Python3.11\n可以直接用Python自带的Json模块完成同样的处理\n视频最后在补一个实际使用时一定会碰到的问题\nMac mini重启以后哪些东西需要重新打开\n我们前面搭载的模型和数据都保存在硬盘里\n不需要重新安装或者重新下载\n但Olama,OpenWebUI,WhisperServer和N8N这些运行中的服务需要重新启动\n建议按照下面这个顺序来\n首先启动最底层的Olama\n如果安装的是MicOS版的Olama\n直接打开OlamaApp即可\n也可以在中端里执行Open-AOlama\n然后可以检查Olama是否已经恢复\n如果能够正常返回Olama的版本信息\n就说明底层模型服务已经正常\n接下来再来启动OpenWebUI\n冲击后可以先确认MacMe的局域网IP地址有没有发生变化\n因为我们前面的启动命令里填写过这个地址\n如果地址没有发生变化\n就继续使用之前的IP\n如果变化了\n下面命令行里的也要一起替换\n这个中轮装口需要保持运行\n然后617访问\n127.0.0.1网号3000\n居于网中的其他设备\n仍然访问Mac mini的IP地址\n网号3000\n长期使用的话\n也可以在路由器里\n给Mac mini固定一个局网地址\n选择每次重建后来进行检查\n如果只是聊天\n知识库或者OpenWeb UI\n其实不用一直启动Whisper server\n只有需要运行视频转写\n或者后面的N8n视频工作流时\n再打开即可\n同样这个中端窗口需要保持运行\n可以打开另一个中端测试\n只要服务器能够这样响应\n就可以继续交给N8n来调用了\n接着打开Docker Desktop App\n因为我们创建N8n容器时\n已经设置了Restart Unless Stoped\n所以正常情况下\n只要Docker App启动\n之前的N8n容器也会自动回复运行\n不需要重新执行那一串Docker Run\n可以先检查Docker PS\n如果列表里能够看到N8n\n就说明N8N已经自动恢复\n如果没有启动\n可以手动执行Docker StartN8N\n然后浏览器打开\n127.0.0.1-5678\n如果能够进入N8N\n就说明自动化服务已经恢复正常\n到这里我们就从一台普通的Mac mini开始\n完成一套相对完整的家庭本地AI服务\nOlama负责运行AI模型\nOpenWeb UI提供对话界面和统一API\nEmbedding和IAG负责自己的文档支援库\nGLM OCR处理图片和扫描件\nwhisper.cpp负责视频录音转写\n再通过N8N把这些能力串成一套\n简单的自动化工作流\n这套方案其实不只是在Mac上\n跑了一个AI模型\n当对话文档图片\n音视频和自动化都连起来之后\n这台Mac mini才开始从一台\n跑模型的电脑\n变成一台可以长期放在家里\n由不同设备随时调用的AI服务器\n当然这只是一个非常技术的框架\n后面还可以继续加入更大的模型\n自己的资料库\n手机快据指令\n甚至更多自动化流程\n大家完全可以根据自己的Mac配置\n和实际需求继续往上搭\n先把这套技术环境稳定跑起来\n再逐步增加功能\n更适合第一次接触本地AI的各位\n好了\n那么以上就是本期视频的全部内容了\n如果觉得本期视频有所帮助的话\n还请不要忘记\n剪辑赛链加公主支持一下\n支持旅金\n我们下期视频再见\n拜拜", "language": "zh", "duration": 1666.751563, "segments": [ { "start": 0.0, "end": 11.0, "text": "Hello,大家好,我是周日金。AI這兩年離咱們普通用戶越來越近。現在甚至不需要單獨買一台服務器,我們就可以直接把手頭閒置在Mac,變成一台屬於自己的本地AI服務器。", "chunk": 0, "language": "zh", "avg_logprob": -0.20352762064356714, "no_speech_prob": 9.806102110876402e-12, "compression_ratio": 1.1909814323607426 }, { "start": 11.0, "end": 27.0, "text": "這次視頻我會從零開始帶大家把Olama、OpenWeb UI、REG、OCR、Whisper.cpp和N8N一步步串起來。這套方案的重點就是,模型和大部分數據都運行在自己本地的Mac上,也不要求特別高的配置。", "chunk": 0, "language": "zh", "avg_logprob": -0.20352762064356714, "no_speech_prob": 9.806102110876402e-12, "compression_ratio": 1.1909814323607426 }, { "start": 27.0, "end": 28.8, "text": "如果大家对这类玩法感兴趣", "chunk": 0, "language": "zh", "avg_logprob": -0.16873610178629558, "no_speech_prob": 2.3541019827932885e-11, "compression_ratio": 1.2023809523809523 }, { "start": 28.8, "end": 32.64, "text": "后面还可以继续带大家折腾本地图片和视频生成", "chunk": 0, "language": "zh", "avg_logprob": -0.16873610178629558, "no_speech_prob": 2.3541019827932885e-11, "compression_ratio": 1.2023809523809523 }, { "start": 32.64, "end": 35.96, "text": "本地AI编程以及把AI介入Home Assistant", "chunk": 0, "language": "zh", "avg_logprob": -0.16873610178629558, "no_speech_prob": 2.3541019827932885e-11, "compression_ratio": 1.2023809523809523 }, { "start": 35.96, "end": 38.52, "text": "通过自然语言控制家里的智能设备", "chunk": 0, "language": "zh", "avg_logprob": -0.16873610178629558, "no_speech_prob": 2.3541019827932885e-11, "compression_ratio": 1.2023809523809523 }, { "start": 38.52, "end": 41.84, "text": "视频正式开始前先求一波一进三连加关注支持一下", "chunk": 0, "language": "zh", "avg_logprob": -0.16873610178629558, "no_speech_prob": 2.3541019827932885e-11, "compression_ratio": 1.2023809523809523 }, { "start": 41.84, "end": 44.92, "text": "这样以后有新的教程更新也能第一时间受到推送", "chunk": 0, "language": "zh", "avg_logprob": -0.16873610178629558, "no_speech_prob": 2.3541019827932885e-11, "compression_ratio": 1.2023809523809523 }, { "start": 44.92, "end": 45.94, "text": "非常感谢", "chunk": 0, "language": "zh", "avg_logprob": -0.16873610178629558, "no_speech_prob": 2.3541019827932885e-11, "compression_ratio": 1.2023809523809523 }, { "start": 45.94, "end": 49.28, "text": "接下来我们就从我的这台Mac mini开始", "chunk": 0, "language": "zh", "avg_logprob": -0.16873610178629558, "no_speech_prob": 2.3541019827932885e-11, "compression_ratio": 1.2023809523809523 }, { "start": 54.4, "end": 55.92, "text": "在开始下载模型之前", "chunk": 0, "language": "zh", "avg_logprob": -0.16873610178629558, "no_speech_prob": 2.3541019827932885e-11, "compression_ratio": 1.2023809523809523 }, { "start": 55.92, "end": 58.92, "text": "之前我们需要先在mac上安装olama", "chunk": 0, "language": "zh", "avg_logprob": -0.1639993732625788, "no_speech_prob": 3.601727596724835e-11, "compression_ratio": 1.3072463768115943 }, { "start": 58.92, "end": 63.92, "text": "你可以把它理解为让普通用户可以在本地运行AI大模型的运行环境", "chunk": 0, "language": "zh", "avg_logprob": -0.1639993732625788, "no_speech_prob": 3.601727596724835e-11, "compression_ratio": 1.3072463768115943 }, { "start": 63.92, "end": 66.92, "text": "直接前往olama的官网下载对应的安装包", "chunk": 0, "language": "zh", "avg_logprob": -0.1639993732625788, "no_speech_prob": 3.601727596724835e-11, "compression_ratio": 1.3072463768115943 }, { "start": 66.92, "end": 69.92, "text": "然后拖动到macOS用程式文件夹就可以了", "chunk": 0, "language": "zh", "avg_logprob": -0.1639993732625788, "no_speech_prob": 3.601727596724835e-11, "compression_ratio": 1.3072463768115943 }, { "start": 69.92, "end": 72.92, "text": "平时使用chatgpt cloud这类AI服务时", "chunk": 0, "language": "zh", "avg_logprob": -0.1639993732625788, "no_speech_prob": 3.601727596724835e-11, "compression_ratio": 1.3072463768115943 }, { "start": 72.92, "end": 74.92, "text": "模型都是运行在云端服务器上", "chunk": 0, "language": "zh", "avg_logprob": -0.1639993732625788, "no_speech_prob": 3.601727596724835e-11, "compression_ratio": 1.3072463768115943 }, { "start": 74.92, "end": 81.92, "text": "而olama要做的事情就是把像deepseek签问lama这类开源模型下载安装到你的mac上", "chunk": 0, "language": "zh", "avg_logprob": -0.1639993732625788, "no_speech_prob": 3.601727596724835e-11, "compression_ratio": 1.3072463768115943 }, { "start": 81.92, "end": 84.92, "text": "让他们直接利用本机的cpu和gpu运行", "chunk": 0, "language": "zh", "avg_logprob": -0.1639993732625788, "no_speech_prob": 3.601727596724835e-11, "compression_ratio": 1.3072463768115943 }, { "start": 84.92, "end": 87.16, "text": "安装好Olama并打开之后", "chunk": 0, "language": "zh", "avg_logprob": -0.12616550332248802, "no_speech_prob": 4.661992666910031e-11, "compression_ratio": 1.2688172043010753 }, { "start": 87.16, "end": 88.52, "text": "下一步就是下载模型", "chunk": 0, "language": "zh", "avg_logprob": -0.12616550332248802, "no_speech_prob": 4.661992666910031e-11, "compression_ratio": 1.2688172043010753 }, { "start": 88.52, "end": 90.0, "text": "不过这里需要注意啊", "chunk": 0, "language": "zh", "avg_logprob": -0.12616550332248802, "no_speech_prob": 4.661992666910031e-11, "compression_ratio": 1.2688172043010753 }, { "start": 90.0, "end": 92.58, "text": "本地AI和ChatGPT这类云竿的模型", "chunk": 0, "language": "zh", "avg_logprob": -0.12616550332248802, "no_speech_prob": 4.661992666910031e-11, "compression_ratio": 1.2688172043010753 }, { "start": 92.58, "end": 93.34, "text": "最大的区别", "chunk": 0, "language": "zh", "avg_logprob": -0.12616550332248802, "no_speech_prob": 4.661992666910031e-11, "compression_ratio": 1.2688172043010753 }, { "start": 93.34, "end": 95.8, "text": "就是模型大小会直接影响运行体验", "chunk": 0, "language": "zh", "avg_logprob": -0.12616550332248802, "no_speech_prob": 4.661992666910031e-11, "compression_ratio": 1.2688172043010753 }, { "start": 95.8, "end": 98.62, "text": "所以不能盲目追求越大的模型就越好", "chunk": 0, "language": "zh", "avg_logprob": -0.12616550332248802, "no_speech_prob": 4.661992666910031e-11, "compression_ratio": 1.2688172043010753 }, { "start": 98.62, "end": 101.54, "text": "我这里使用的是一台M4入门款的Mac mini", "chunk": 0, "language": "zh", "avg_logprob": -0.12616550332248802, "no_speech_prob": 4.661992666910031e-11, "compression_ratio": 1.2688172043010753 }, { "start": 101.54, "end": 104.92, "text": "采用16GB的同一列存加256GB的固态硬盘", "chunk": 0, "language": "zh", "avg_logprob": -0.12616550332248802, "no_speech_prob": 4.661992666910031e-11, "compression_ratio": 1.2688172043010753 }, { "start": 104.92, "end": 107.3, "text": "确认Olama的服务在正常运行后", "chunk": 0, "language": "zh", "avg_logprob": -0.12616550332248802, "no_speech_prob": 4.661992666910031e-11, "compression_ratio": 1.2688172043010753 }, { "start": 107.3, "end": 109.12, "text": "我们在MacOS的中段App中", "chunk": 0, "language": "zh", "avg_logprob": -0.12616550332248802, "no_speech_prob": 4.661992666910031e-11, "compression_ratio": 1.2688172043010753 }, { "start": 109.12, "end": 110.42, "text": "输入这三条命令", "chunk": 0, "language": "zh", "avg_logprob": -0.12616550332248802, "no_speech_prob": 4.661992666910031e-11, "compression_ratio": 1.2688172043010753 }, { "start": 110.42, "end": 112.26, "text": "它们分别负责不同的功能", "chunk": 0, "language": "zh", "avg_logprob": -0.12616550332248802, "no_speech_prob": 4.661992666910031e-11, "compression_ratio": 1.2688172043010753 }, { "start": 112.26, "end": 113.88, "text": "第一个是大约模型", "chunk": 0, "language": "zh", "avg_logprob": -0.12616550332248802, "no_speech_prob": 4.661992666910031e-11, "compression_ratio": 1.2688172043010753 }, { "start": 113.88, "end": 115.7, "text": "也就是本地AI的大脑", "chunk": 0, "language": "zh", "avg_logprob": -0.11466631889343262, "no_speech_prob": 1.374132665293848e-11, "compression_ratio": 1.1644204851752022 }, { "start": 115.7, "end": 116.76, "text": "负责日常聊天", "chunk": 0, "language": "zh", "avg_logprob": -0.11466631889343262, "no_speech_prob": 1.374132665293848e-11, "compression_ratio": 1.1644204851752022 }, { "start": 116.76, "end": 117.5, "text": "内容生成", "chunk": 0, "language": "zh", "avg_logprob": -0.11466631889343262, "no_speech_prob": 1.374132665293848e-11, "compression_ratio": 1.1644204851752022 }, { "start": 117.5, "end": 118.74, "text": "总业资料等任务", "chunk": 0, "language": "zh", "avg_logprob": -0.11466631889343262, "no_speech_prob": 1.374132665293848e-11, "compression_ratio": 1.1644204851752022 }, { "start": 118.74, "end": 120.96, "text": "其中4B代表40亿参数", "chunk": 0, "language": "zh", "avg_logprob": -0.11466631889343262, "no_speech_prob": 1.374132665293848e-11, "compression_ratio": 1.1644204851752022 }, { "start": 120.96, "end": 123.68, "text": "Kill4KM代表经过4bit量化", "chunk": 0, "language": "zh", "avg_logprob": -0.11466631889343262, "no_speech_prob": 1.374132665293848e-11, "compression_ratio": 1.1644204851752022 }, { "start": 123.68, "end": 125.64, "text": "可以在降低模型状的同时", "chunk": 0, "language": "zh", "avg_logprob": -0.11466631889343262, "no_speech_prob": 1.374132665293848e-11, "compression_ratio": 1.1644204851752022 }, { "start": 125.64, "end": 127.18, "text": "保持不错的运行效果", "chunk": 0, "language": "zh", "avg_logprob": -0.11466631889343262, "no_speech_prob": 1.374132665293848e-11, "compression_ratio": 1.1644204851752022 }, { "start": 127.18, "end": 129.68, "text": "是目前比较适合普通Mac的选择", "chunk": 0, "language": "zh", "avg_logprob": -0.11466631889343262, "no_speech_prob": 1.374132665293848e-11, "compression_ratio": 1.1644204851752022 }, { "start": 129.68, "end": 131.44, "text": "第二个是向量模型", "chunk": 0, "language": "zh", "avg_logprob": -0.11466631889343262, "no_speech_prob": 1.374132665293848e-11, "compression_ratio": 1.1644204851752022 }, { "start": 131.44, "end": 132.84, "text": "主要用于知识库检索", "chunk": 0, "language": "zh", "avg_logprob": -0.11466631889343262, "no_speech_prob": 1.374132665293848e-11, "compression_ratio": 1.1644204851752022 }, { "start": 132.84, "end": 135.26, "text": "如果你以后想让AI读取自己的PDF", "chunk": 0, "language": "zh", "avg_logprob": -0.11466631889343262, "no_speech_prob": 1.374132665293848e-11, "compression_ratio": 1.1644204851752022 }, { "start": 135.26, "end": 136.58, "text": "文档视频稿件", "chunk": 0, "language": "zh", "avg_logprob": -0.11466631889343262, "no_speech_prob": 1.374132665293848e-11, "compression_ratio": 1.1644204851752022 }, { "start": 136.58, "end": 138.44, "text": "然后根据这些资料回答问题", "chunk": 0, "language": "zh", "avg_logprob": -0.11466631889343262, "no_speech_prob": 1.374132665293848e-11, "compression_ratio": 1.1644204851752022 }, { "start": 138.44, "end": 140.06, "text": "就需要通过Embedding模型", "chunk": 0, "language": "zh", "avg_logprob": -0.11466631889343262, "no_speech_prob": 1.374132665293848e-11, "compression_ratio": 1.1644204851752022 }, { "start": 140.06, "end": 142.18, "text": "先帮AI找到相关内容", "chunk": 0, "language": "zh", "avg_logprob": -0.11466631889343262, "no_speech_prob": 1.374132665293848e-11, "compression_ratio": 1.1644204851752022 }, { "start": 142.18, "end": 144.8, "text": "第三个是欧塞亚文字试园模型", "chunk": 0, "language": "zh", "avg_logprob": -0.09929898463257956, "no_speech_prob": 1.703572873046344e-11, "compression_ratio": 1.3603351955307263 }, { "start": 144.8, "end": 145.92, "text": "负责处理图片", "chunk": 0, "language": "zh", "avg_logprob": -0.09929898463257956, "no_speech_prob": 1.703572873046344e-11, "compression_ratio": 1.3603351955307263 }, { "start": 145.92, "end": 147.48, "text": "扫描版PDF等内容", "chunk": 0, "language": "zh", "avg_logprob": -0.09929898463257956, "no_speech_prob": 1.703572873046344e-11, "compression_ratio": 1.3603351955307263 }, { "start": 147.48, "end": 149.34, "text": "把图片里的文字提许出来", "chunk": 0, "language": "zh", "avg_logprob": -0.09929898463257956, "no_speech_prob": 1.703572873046344e-11, "compression_ratio": 1.3603351955307263 }, { "start": 149.34, "end": 151.32, "text": "再交给语言模型进行分析", "chunk": 0, "language": "zh", "avg_logprob": -0.09929898463257956, "no_speech_prob": 1.703572873046344e-11, "compression_ratio": 1.3603351955307263 }, { "start": 151.32, "end": 153.62, "text": "这里之所以选择千问的AR模型", "chunk": 0, "language": "zh", "avg_logprob": -0.09929898463257956, "no_speech_prob": 1.703572873046344e-11, "compression_ratio": 1.3603351955307263 }, { "start": 153.62, "end": 155.72, "text": "是因为千问是目前比较成熟的", "chunk": 0, "language": "zh", "avg_logprob": -0.09929898463257956, "no_speech_prob": 1.703572873046344e-11, "compression_ratio": 1.3603351955307263 }, { "start": 155.72, "end": 156.76, "text": "开源大模型之一", "chunk": 0, "language": "zh", "avg_logprob": -0.09929898463257956, "no_speech_prob": 1.703572873046344e-11, "compression_ratio": 1.3603351955307263 }, { "start": 156.76, "end": 159.36, "text": "对中文理解和中文生成比较友好", "chunk": 0, "language": "zh", "avg_logprob": -0.09929898463257956, "no_speech_prob": 1.703572873046344e-11, "compression_ratio": 1.3603351955307263 }, { "start": 159.36, "end": 161.42, "text": "其次是它提供的模型规格丰富", "chunk": 0, "language": "zh", "avg_logprob": -0.09929898463257956, "no_speech_prob": 1.703572873046344e-11, "compression_ratio": 1.3603351955307263 }, { "start": 161.42, "end": 163.7, "text": "从4B到更大的参数版本", "chunk": 0, "language": "zh", "avg_logprob": -0.09929898463257956, "no_speech_prob": 1.703572873046344e-11, "compression_ratio": 1.3603351955307263 }, { "start": 163.7, "end": 165.38, "text": "可以适配不同配置的Mac", "chunk": 0, "language": "zh", "avg_logprob": -0.09929898463257956, "no_speech_prob": 1.703572873046344e-11, "compression_ratio": 1.3603351955307263 }, { "start": 165.38, "end": 166.16, "text": "还有一个是", "chunk": 0, "language": "zh", "avg_logprob": -0.09929898463257956, "no_speech_prob": 1.703572873046344e-11, "compression_ratio": 1.3603351955307263 }, { "start": 166.16, "end": 168.2, "text": "它在欧拉玛里的支持比较完善", "chunk": 0, "language": "zh", "avg_logprob": -0.09929898463257956, "no_speech_prob": 1.703572873046344e-11, "compression_ratio": 1.3603351955307263 }, { "start": 168.2, "end": 170.02, "text": "安装和电用也都比较简单", "chunk": 0, "language": "zh", "avg_logprob": -0.09929898463257956, "no_speech_prob": 1.703572873046344e-11, "compression_ratio": 1.3603351955307263 }, { "start": 170.02, "end": 171.28, "text": "当然除了千问", "chunk": 0, "language": "zh", "avg_logprob": -0.09929898463257956, "no_speech_prob": 1.703572873046344e-11, "compression_ratio": 1.3603351955307263 }, { "start": 171.28, "end": 173.82, "text": "DeepSeq,Lama等模型也都可以运行", "chunk": 0, "language": "zh", "avg_logprob": -0.09681785779121595, "no_speech_prob": 2.0698926947448903e-11, "compression_ratio": 1.2873563218390804 }, { "start": 173.82, "end": 176.1, "text": "大家可以根据自己的需求进行选择", "chunk": 0, "language": "zh", "avg_logprob": -0.09681785779121595, "no_speech_prob": 2.0698926947448903e-11, "compression_ratio": 1.2873563218390804 }, { "start": 176.1, "end": 179.02, "text": "那至于不同的Mac应该选择多大的模型", "chunk": 0, "language": "zh", "avg_logprob": -0.09681785779121595, "no_speech_prob": 2.0698926947448903e-11, "compression_ratio": 1.2873563218390804 }, { "start": 179.02, "end": 181.26, "text": "主要看新变性能和统一内存的大小", "chunk": 0, "language": "zh", "avg_logprob": -0.09681785779121595, "no_speech_prob": 2.0698926947448903e-11, "compression_ratio": 1.2873563218390804 }, { "start": 181.26, "end": 184.46, "text": "8GB内存的Mac建议先从4B级别模型开始", "chunk": 0, "language": "zh", "avg_logprob": -0.09681785779121595, "no_speech_prob": 2.0698926947448903e-11, "compression_ratio": 1.2873563218390804 }, { "start": 184.46, "end": 187.86, "text": "16GB内存运行4B会相对轻松", "chunk": 0, "language": "zh", "avg_logprob": -0.09681785779121595, "no_speech_prob": 2.0698926947448903e-11, "compression_ratio": 1.2873563218390804 }, { "start": 187.86, "end": 191.34, "text": "也可以在实际内存压力云雪情况下尝试9B", "chunk": 0, "language": "zh", "avg_logprob": -0.09681785779121595, "no_speech_prob": 2.0698926947448903e-11, "compression_ratio": 1.2873563218390804 }, { "start": 191.34, "end": 194.64, "text": "而24GB,32GB以及更高内存的Mac", "chunk": 0, "language": "zh", "avg_logprob": -0.09681785779121595, "no_speech_prob": 2.0698926947448903e-11, "compression_ratio": 1.2873563218390804 }, { "start": 194.64, "end": 196.48, "text": "则可以继续尝试更大的规格", "chunk": 0, "language": "zh", "avg_logprob": -0.09681785779121595, "no_speech_prob": 2.0698926947448903e-11, "compression_ratio": 1.2873563218390804 }, { "start": 196.48, "end": 200.18, "text": "但具体体验还会受到上下文长度和后台任务的影响", "chunk": 0, "language": "zh", "avg_logprob": -0.09681785779121595, "no_speech_prob": 2.0698926947448903e-11, "compression_ratio": 1.2873563218390804 }, { "start": 200.18, "end": 201.52, "text": "这里需要注意啊", "chunk": 0, "language": "zh", "avg_logprob": -0.12297219354749302, "no_speech_prob": 1.600298192572236e-11, "compression_ratio": 1.2474226804123711 }, { "start": 201.52, "end": 203.4, "text": "模型标签不能随便改数字", "chunk": 0, "language": "zh", "avg_logprob": -0.12297219354749302, "no_speech_prob": 1.600298192572236e-11, "compression_ratio": 1.2474226804123711 }, { "start": 203.4, "end": 205.26, "text": "如果继续使用千万3.5", "chunk": 0, "language": "zh", "avg_logprob": -0.12297219354749302, "no_speech_prob": 1.600298192572236e-11, "compression_ratio": 1.2474226804123711 }, { "start": 205.26, "end": 206.94, "text": "目前Olama还提供9B", "chunk": 0, "language": "zh", "avg_logprob": -0.12297219354749302, "no_speech_prob": 1.600298192572236e-11, "compression_ratio": 1.2474226804123711 }, { "start": 206.94, "end": 209.08, "text": "27B 35B等更大规格", "chunk": 0, "language": "zh", "avg_logprob": -0.12297219354749302, "no_speech_prob": 1.600298192572236e-11, "compression_ratio": 1.2474226804123711 }, { "start": 209.08, "end": 211.12, "text": "具体名称一定以Olama模型库", "chunk": 0, "language": "zh", "avg_logprob": -0.12297219354749302, "no_speech_prob": 1.600298192572236e-11, "compression_ratio": 1.2474226804123711 }, { "start": 211.12, "end": 212.5, "text": "实际存在的标签为准", "chunk": 0, "language": "zh", "avg_logprob": -0.12297219354749302, "no_speech_prob": 1.600298192572236e-11, "compression_ratio": 1.2474226804123711 }, { "start": 212.5, "end": 215.16, "text": "那咱们还是先把本地AI服务器跑起来", "chunk": 0, "language": "zh", "avg_logprob": -0.12297219354749302, "no_speech_prob": 1.600298192572236e-11, "compression_ratio": 1.2474226804123711 }, { "start": 215.16, "end": 217.42, "text": "后续再根据实际需求升级模型", "chunk": 0, "language": "zh", "avg_logprob": -0.12297219354749302, "no_speech_prob": 1.600298192572236e-11, "compression_ratio": 1.2474226804123711 }, { "start": 217.42, "end": 219.68, "text": "接下来我们在Microsoft中的App中", "chunk": 0, "language": "zh", "avg_logprob": -0.12297219354749302, "no_speech_prob": 1.600298192572236e-11, "compression_ratio": 1.2474226804123711 }, { "start": 219.68, "end": 220.84, "text": "输入以下命令", "chunk": 0, "language": "zh", "avg_logprob": -0.12297219354749302, "no_speech_prob": 1.600298192572236e-11, "compression_ratio": 1.2474226804123711 }, { "start": 220.84, "end": 221.84, "text": "Olama List", "chunk": 0, "language": "zh", "avg_logprob": -0.12297219354749302, "no_speech_prob": 1.600298192572236e-11, "compression_ratio": 1.2474226804123711 }, { "start": 221.84, "end": 224.12, "text": "查看是否已经全部下载完成", "chunk": 0, "language": "zh", "avg_logprob": -0.12297219354749302, "no_speech_prob": 1.600298192572236e-11, "compression_ratio": 1.2474226804123711 }, { "start": 224.12, "end": 226.64, "text": "如果前面下载的模型都正常列出来", "chunk": 0, "language": "zh", "avg_logprob": -0.12297219354749302, "no_speech_prob": 1.600298192572236e-11, "compression_ratio": 1.2474226804123711 }, { "start": 226.64, "end": 227.96, "text": "就可以来简单试一下", "chunk": 0, "language": "zh", "avg_logprob": -0.12297219354749302, "no_speech_prob": 1.600298192572236e-11, "compression_ratio": 1.2474226804123711 }, { "start": 227.96, "end": 229.9, "text": "模型能不能在本地正常运行", "chunk": 0, "language": "zh", "avg_logprob": -0.12297219354749302, "no_speech_prob": 1.600298192572236e-11, "compression_ratio": 1.2474226804123711 }, { "start": 229.9, "end": 231.08, "text": "输入制传命令", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 231.08, "end": 232.04, "text": "olama run", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 232.04, "end": 232.9, "text": "千万3.5", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 232.9, "end": 234.54, "text": "4b-q4km", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 234.54, "end": 235.9, "text": "然后随便发一句话", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 235.9, "end": 236.64, "text": "比如你好", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 236.64, "end": 239.08, "text": "如果能正常给出回复就没问题了", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 239.08, "end": 241.06, "text": "随后输入制传命令结束对话", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 241.06, "end": 241.54, "text": "Bye", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 241.54, "end": 243.66, "text": "之后我们就可以停止运行", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 243.66, "end": 244.62, "text": "以释放资源", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 244.62, "end": 245.78, "text": "输入制传命令", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 245.78, "end": 246.82, "text": "olama stop", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 246.82, "end": 247.64, "text": "千万3.5", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 247.64, "end": 249.08, "text": "4b-q4km", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 249.08, "end": 250.68, "text": "然后再输入制传命令", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 250.68, "end": 251.48, "text": "olama ps", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 251.48, "end": 254.14, "text": "查看这个AR大模型是否人在运行", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 254.14, "end": 255.46, "text": "此时应该为空", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 255.46, "end": 258.2, "text": "那么就说明我们的大模型已经没有在运行了", "chunk": 0, "language": "zh", "avg_logprob": -0.15430423191615514, "no_speech_prob": 1.884373906912984e-11, "compression_ratio": 1.3661016949152542 }, { "start": 258.2, "end": 260.32, "text": "接下来咱们就要开始准备部署", "chunk": 0, "language": "zh", "avg_logprob": -0.12013424010503859, "no_speech_prob": 1.3789337727221351e-11, "compression_ratio": 1.2597402597402598 }, { "start": 260.32, "end": 262.36, "text": "家庭局网AI对话界面了", "chunk": 0, "language": "zh", "avg_logprob": -0.12013424010503859, "no_speech_prob": 1.3789337727221351e-11, "compression_ratio": 1.2597402597402598 }, { "start": 262.36, "end": 270.98, "text": "前面我们已经通过Olama把AI模型运行在Mac上", "chunk": 0, "language": "zh", "avg_logprob": -0.12013424010503859, "no_speech_prob": 1.3789337727221351e-11, "compression_ratio": 1.2597402597402598 }, { "start": 270.98, "end": 273.5, "text": "但是目前还需要通过中端的进行操作", "chunk": 0, "language": "zh", "avg_logprob": -0.12013424010503859, "no_speech_prob": 1.3789337727221351e-11, "compression_ratio": 1.2597402597402598 }, { "start": 273.5, "end": 275.28, "text": "对于日常使用来说不太方便", "chunk": 0, "language": "zh", "avg_logprob": -0.12013424010503859, "no_speech_prob": 1.3789337727221351e-11, "compression_ratio": 1.2597402597402598 }, { "start": 275.28, "end": 277.2, "text": "所以接下来我们就要给它搭建一个", "chunk": 0, "language": "zh", "avg_logprob": -0.12013424010503859, "no_speech_prob": 1.3789337727221351e-11, "compression_ratio": 1.2597402597402598 }, { "start": 277.2, "end": 279.06, "text": "类似ChatGP的网页界面", "chunk": 0, "language": "zh", "avg_logprob": -0.12013424010503859, "no_speech_prob": 1.3789337727221351e-11, "compression_ratio": 1.2597402597402598 }, { "start": 279.06, "end": 280.72, "text": "这里使用OpenWebUI", "chunk": 0, "language": "zh", "avg_logprob": -0.12013424010503859, "no_speech_prob": 1.3789337727221351e-11, "compression_ratio": 1.2597402597402598 }, { "start": 280.72, "end": 283.24, "text": "也就是让Olama负责运行AI模型", "chunk": 0, "language": "zh", "avg_logprob": -0.12013424010503859, "no_speech_prob": 1.3789337727221351e-11, "compression_ratio": 1.2597402597402598 }, { "start": 283.24, "end": 285.62, "text": "OpenWebUI负责提供交互界面", "chunk": 0, "language": "zh", "avg_logprob": -0.12013424010503859, "no_speech_prob": 1.3789337727221351e-11, "compression_ratio": 1.2597402597402598 }, { "start": 285.62, "end": 286.82, "text": "那部署完成后", "chunk": 0, "language": "zh", "avg_logprob": -0.12013424010503859, "no_speech_prob": 1.3789337727221351e-11, "compression_ratio": 1.2597402597402598 }, { "start": 286.82, "end": 289.3, "text": "家里的手机平板电脑都可以通过浏览器", "chunk": 0, "language": "zh", "avg_logprob": -0.10979039201112552, "no_speech_prob": 2.5058932359711683e-11, "compression_ratio": 1.2506527415143602 }, { "start": 289.3, "end": 290.34, "text": "访问这台Mac mini", "chunk": 0, "language": "zh", "avg_logprob": -0.10979039201112552, "no_speech_prob": 2.5058932359711683e-11, "compression_ratio": 1.2506527415143602 }, { "start": 290.34, "end": 292.3, "text": "直接使用自己的本地AI", "chunk": 0, "language": "zh", "avg_logprob": -0.10979039201112552, "no_speech_prob": 2.5058932359711683e-11, "compression_ratio": 1.2506527415143602 }, { "start": 292.3, "end": 294.6, "text": "能考虑到中国大陆用户的网络环境", "chunk": 0, "language": "zh", "avg_logprob": -0.10979039201112552, "no_speech_prob": 2.5058932359711683e-11, "compression_ratio": 1.2506527415143602 }, { "start": 294.6, "end": 296.92, "text": "这里直接通过Python安装OpenWeb UI", "chunk": 0, "language": "zh", "avg_logprob": -0.10979039201112552, "no_speech_prob": 2.5058932359711683e-11, "compression_ratio": 1.2506527415143602 }, { "start": 296.92, "end": 299.44, "text": "同时Olama继续只监听本机地址", "chunk": 0, "language": "zh", "avg_logprob": -0.10979039201112552, "no_speech_prob": 2.5058932359711683e-11, "compression_ratio": 1.2506527415143602 }, { "start": 299.44, "end": 300.92, "text": "不会直接暴露到网络中", "chunk": 0, "language": "zh", "avg_logprob": -0.10979039201112552, "no_speech_prob": 2.5058932359711683e-11, "compression_ratio": 1.2506527415143602 }, { "start": 300.92, "end": 304.16, "text": "本教程固定使用OpenWeb UI 0.11.0版本", "chunk": 0, "language": "zh", "avg_logprob": -0.10979039201112552, "no_speech_prob": 2.5058932359711683e-11, "compression_ratio": 1.2506527415143602 }, { "start": 304.16, "end": 306.58, "text": "不使用latest或滚动更新版本", "chunk": 0, "language": "zh", "avg_logprob": -0.10979039201112552, "no_speech_prob": 2.5058932359711683e-11, "compression_ratio": 1.2506527415143602 }, { "start": 306.58, "end": 309.38, "text": "避免版本变化导致操作流程不一炸", "chunk": 0, "language": "zh", "avg_logprob": -0.10979039201112552, "no_speech_prob": 2.5058932359711683e-11, "compression_ratio": 1.2506527415143602 }, { "start": 309.38, "end": 311.0, "text": "首先准备Python环境", "chunk": 0, "language": "zh", "avg_logprob": -0.10979039201112552, "no_speech_prob": 2.5058932359711683e-11, "compression_ratio": 1.2506527415143602 }, { "start": 311.0, "end": 313.02, "text": "如果你的Mac已经安装Homebrew", "chunk": 0, "language": "zh", "avg_logprob": -0.10979039201112552, "no_speech_prob": 2.5058932359711683e-11, "compression_ratio": 1.2506527415143602 }, { "start": 313.02, "end": 316.12, "text": "则可以输入brew install python at 3.11", "chunk": 0, "language": "zh", "avg_logprob": -0.10979039201112552, "no_speech_prob": 2.5058932359711683e-11, "compression_ratio": 1.2506527415143602 }, { "start": 316.12, "end": 317.54, "text": "如果没有安装Homebrew", "chunk": 0, "language": "zh", "avg_logprob": -0.09014611408628266, "no_speech_prob": 1.4056430498732286e-11, "compression_ratio": 1.1920731707317074 }, { "start": 317.54, "end": 322.2, "text": "也可以直接向Python官方网站下载安装Python 3.1对应的MicroOS安装包", "chunk": 0, "language": "zh", "avg_logprob": -0.09014611408628266, "no_speech_prob": 1.4056430498732286e-11, "compression_ratio": 1.1920731707317074 }, { "start": 322.2, "end": 324.66, "text": "后面如果会继续折腾这类工具", "chunk": 0, "language": "zh", "avg_logprob": -0.09014611408628266, "no_speech_prob": 1.4056430498732286e-11, "compression_ratio": 1.1920731707317074 }, { "start": 324.66, "end": 326.12, "text": "建议提前装好Homebrew", "chunk": 0, "language": "zh", "avg_logprob": -0.09014611408628266, "no_speech_prob": 1.4056430498732286e-11, "compression_ratio": 1.1920731707317074 }, { "start": 326.12, "end": 327.84, "text": "管理和升级会方便不少", "chunk": 0, "language": "zh", "avg_logprob": -0.09014611408628266, "no_speech_prob": 1.4056430498732286e-11, "compression_ratio": 1.1920731707317074 }, { "start": 327.84, "end": 330.54, "text": "接下来创建OpenWeb UI独立运行环境", "chunk": 0, "language": "zh", "avg_logprob": -0.09014611408628266, "no_speech_prob": 1.4056430498732286e-11, "compression_ratio": 1.1920731707317074 }, { "start": 330.54, "end": 332.84, "text": "避免影响系统里的其他Python项目", "chunk": 0, "language": "zh", "avg_logprob": -0.09014611408628266, "no_speech_prob": 1.4056430498732286e-11, "compression_ratio": 1.1920731707317074 }, { "start": 332.84, "end": 335.02, "text": "后续相关代码我不再一一进行解释", "chunk": 0, "language": "zh", "avg_logprob": -0.09014611408628266, "no_speech_prob": 1.4056430498732286e-11, "compression_ratio": 1.1920731707317074 }, { "start": 335.02, "end": 338.06, "text": "如果有看不懂可以查看代码上方的文本注释", "chunk": 0, "language": "zh", "avg_logprob": -0.09014611408628266, "no_speech_prob": 1.4056430498732286e-11, "compression_ratio": 1.1920731707317074 }, { "start": 346.12, "end": 352.64, "text": "然后通过轻滑镜像安装OpenWeb UI", "chunk": 0, "language": "zh", "avg_logprob": -0.09514867782592773, "no_speech_prob": 1.1638057705043447e-11, "compression_ratio": 1.0980392156862746 }, { "start": 352.64, "end": 363.8, "text": "安装完成后", "chunk": 0, "language": "zh", "avg_logprob": -0.09514867782592773, "no_speech_prob": 1.1638057705043447e-11, "compression_ratio": 1.0980392156862746 }, { "start": 363.8, "end": 366.04, "text": "第一次启动建议只允许本机访问", "chunk": 0, "language": "zh", "avg_logprob": -0.09514867782592773, "no_speech_prob": 1.1638057705043447e-11, "compression_ratio": 1.0980392156862746 }, { "start": 366.04, "end": 367.94, "text": "先完成账号和安全设置", "chunk": 0, "language": "zh", "avg_logprob": -0.09514867782592773, "no_speech_prob": 1.1638057705043447e-11, "compression_ratio": 1.0980392156862746 }, { "start": 367.94, "end": 369.3, "text": "等命令跑起来以后", "chunk": 0, "language": "zh", "avg_logprob": -0.09514867782592773, "no_speech_prob": 1.1638057705043447e-11, "compression_ratio": 1.0980392156862746 }, { "start": 369.3, "end": 370.88, "text": "不要关闭这个中端窗口", "chunk": 0, "language": "zh", "avg_logprob": -0.09514867782592773, "no_speech_prob": 1.1638057705043447e-11, "compression_ratio": 1.0980392156862746 }, { "start": 370.88, "end": 373.4, "text": "因为OpenWeb UI的服务就运行在这里", "chunk": 0, "language": "zh", "avg_logprob": -0.09514867782592773, "no_speech_prob": 1.1638057705043447e-11, "compression_ratio": 1.0980392156862746 }, { "start": 373.4, "end": 377.24, "text": "然后流量器打开", "chunk": 0, "language": "zh", "avg_logprob": -0.14350600301483532, "no_speech_prob": 1.1033456266684727e-11, "compression_ratio": 1.1768707482993197 }, { "start": 377.24, "end": 378.76, "text": "127.0.0.1", "chunk": 0, "language": "zh", "avg_logprob": -0.14350600301483532, "no_speech_prob": 1.1033456266684727e-11, "compression_ratio": 1.1768707482993197 }, { "start": 378.76, "end": 379.58, "text": "网号3000", "chunk": 0, "language": "zh", "avg_logprob": -0.14350600301483532, "no_speech_prob": 1.1033456266684727e-11, "compression_ratio": 1.1768707482993197 }, { "start": 379.58, "end": 381.7, "text": "第一次进入需要创建账号", "chunk": 0, "language": "zh", "avg_logprob": -0.14350600301483532, "no_speech_prob": 1.1033456266684727e-11, "compression_ratio": 1.1768707482993197 }, { "start": 381.7, "end": 383.84, "text": "第一个账号就会自动成为管理员", "chunk": 0, "language": "zh", "avg_logprob": -0.14350600301483532, "no_speech_prob": 1.1033456266684727e-11, "compression_ratio": 1.1768707482993197 }, { "start": 383.84, "end": 388.46, "text": "进入管理后台后", "chunk": 0, "language": "zh", "avg_logprob": -0.14350600301483532, "no_speech_prob": 1.1033456266684727e-11, "compression_ratio": 1.1768707482993197 }, { "start": 388.46, "end": 389.94, "text": "前往设置身份验证", "chunk": 0, "language": "zh", "avg_logprob": -0.14350600301483532, "no_speech_prob": 1.1033456266684727e-11, "compression_ratio": 1.1768707482993197 }, { "start": 389.94, "end": 391.46, "text": "把newsign apps关闭", "chunk": 0, "language": "zh", "avg_logprob": -0.14350600301483532, "no_speech_prob": 1.1033456266684727e-11, "compression_ratio": 1.1768707482993197 }, { "start": 391.46, "end": 393.46, "text": "也就是不允许新用户自由注册", "chunk": 0, "language": "zh", "avg_logprob": -0.14350600301483532, "no_speech_prob": 1.1033456266684727e-11, "compression_ratio": 1.1768707482993197 }, { "start": 393.46, "end": 395.8, "text": "同时暂时不要安装来源不明的", "chunk": 0, "language": "zh", "avg_logprob": -0.14350600301483532, "no_speech_prob": 1.1033456266684727e-11, "compression_ratio": 1.1768707482993197 }, { "start": 395.8, "end": 397.98, "text": "第三方Tools,Functions或Pipelines", "chunk": 0, "language": "zh", "avg_logprob": -0.14350600301483532, "no_speech_prob": 1.1033456266684727e-11, "compression_ratio": 1.1768707482993197 }, { "start": 397.98, "end": 399.66, "text": "接着就可以在流量器打开", "chunk": 0, "language": "zh", "avg_logprob": -0.14350600301483532, "no_speech_prob": 1.1033456266684727e-11, "compression_ratio": 1.1768707482993197 }, { "start": 399.66, "end": 400.98, "text": "127.0.1", "chunk": 0, "language": "zh", "avg_logprob": -0.14350600301483532, "no_speech_prob": 1.1033456266684727e-11, "compression_ratio": 1.1768707482993197 }, { "start": 400.98, "end": 402.44, "text": "网号11434", "chunk": 0, "language": "zh", "avg_logprob": -0.14350600301483532, "no_speech_prob": 1.1033456266684727e-11, "compression_ratio": 1.1768707482993197 }, { "start": 402.44, "end": 404.48, "text": "-API-version", "chunk": 0, "language": "zh", "avg_logprob": -0.1524330030168806, "no_speech_prob": 2.0539733108781988e-11, "compression_ratio": 1.135048231511254 }, { "start": 404.48, "end": 406.88, "text": "确认Olama本地API正常响应", "chunk": 0, "language": "zh", "avg_logprob": -0.1524330030168806, "no_speech_prob": 2.0539733108781988e-11, "compression_ratio": 1.135048231511254 }, { "start": 406.88, "end": 408.12, "text": "回到OpenWeb UI", "chunk": 0, "language": "zh", "avg_logprob": -0.1524330030168806, "no_speech_prob": 2.0539733108781988e-11, "compression_ratio": 1.135048231511254 }, { "start": 408.12, "end": 409.64, "text": "再检查之前下载的", "chunk": 0, "language": "zh", "avg_logprob": -0.1524330030168806, "no_speech_prob": 2.0539733108781988e-11, "compression_ratio": 1.135048231511254 }, { "start": 409.64, "end": 411.76, "text": "千万3.5 4B-TU4KM", "chunk": 0, "language": "zh", "avg_logprob": -0.1524330030168806, "no_speech_prob": 2.0539733108781988e-11, "compression_ratio": 1.135048231511254 }, { "start": 411.76, "end": 413.78, "text": "是否已经出现在母型列表当中", "chunk": 0, "language": "zh", "avg_logprob": -0.1524330030168806, "no_speech_prob": 2.0539733108781988e-11, "compression_ratio": 1.135048231511254 }, { "start": 413.78, "end": 414.88, "text": "确认没有问题后", "chunk": 0, "language": "zh", "avg_logprob": -0.1524330030168806, "no_speech_prob": 2.0539733108781988e-11, "compression_ratio": 1.135048231511254 }, { "start": 414.88, "end": 417.06, "text": "在中端APP中按Ctrl-C", "chunk": 0, "language": "zh", "avg_logprob": -0.1524330030168806, "no_speech_prob": 2.0539733108781988e-11, "compression_ratio": 1.135048231511254 }, { "start": 417.06, "end": 418.42, "text": "停止OpenWeb UI", "chunk": 0, "language": "zh", "avg_logprob": -0.1524330030168806, "no_speech_prob": 2.0539733108781988e-11, "compression_ratio": 1.135048231511254 }, { "start": 418.42, "end": 420.16, "text": "接下来开放家庭局域网", "chunk": 0, "language": "zh", "avg_logprob": -0.1524330030168806, "no_speech_prob": 2.0539733108781988e-11, "compression_ratio": 1.135048231511254 }, { "start": 420.16, "end": 422.02, "text": "实际在系统设置网络中", "chunk": 0, "language": "zh", "avg_logprob": -0.1524330030168806, "no_speech_prob": 2.0539733108781988e-11, "compression_ratio": 1.135048231511254 }, { "start": 422.02, "end": 423.72, "text": "查看Mac mini的局域网IP", "chunk": 0, "language": "zh", "avg_logprob": -0.1524330030168806, "no_speech_prob": 2.0539733108781988e-11, "compression_ratio": 1.135048231511254 }, { "start": 423.72, "end": 426.04, "text": "实际操作需要替完成你自己的IP", "chunk": 0, "language": "zh", "avg_logprob": -0.1524330030168806, "no_speech_prob": 2.0539733108781988e-11, "compression_ratio": 1.135048231511254 }, { "start": 426.04, "end": 427.86, "text": "重新启动OpenWeb UI", "chunk": 0, "language": "zh", "avg_logprob": -0.1524330030168806, "no_speech_prob": 2.0539733108781988e-11, "compression_ratio": 1.135048231511254 }, { "start": 427.86, "end": 434.34, "text": "现在Mac mini本机可以继续访问127.0.0.1", "chunk": 0, "language": "zh", "avg_logprob": -0.13215751118130153, "no_speech_prob": 1.9985006705081076e-11, "compression_ratio": 1.319327731092437 }, { "start": 434.34, "end": 435.08, "text": "网号3000", "chunk": 0, "language": "zh", "avg_logprob": -0.13215751118130153, "no_speech_prob": 1.9985006705081076e-11, "compression_ratio": 1.319327731092437 }, { "start": 435.08, "end": 437.96, "text": "并使用刚才创建的管理员账号进行登录", "chunk": 0, "language": "zh", "avg_logprob": -0.13215751118130153, "no_speech_prob": 1.9985006705081076e-11, "compression_ratio": 1.319327731092437 }, { "start": 437.96, "end": 439.7, "text": "通过家庭网络里的手机", "chunk": 0, "language": "zh", "avg_logprob": -0.13215751118130153, "no_speech_prob": 1.9985006705081076e-11, "compression_ratio": 1.319327731092437 }, { "start": 439.7, "end": 441.88, "text": "电脑则可以通过Mac mini.ipd设计", "chunk": 0, "language": "zh", "avg_logprob": -0.13215751118130153, "no_speech_prob": 1.9985006705081076e-11, "compression_ratio": 1.319327731092437 }, { "start": 441.88, "end": 443.54, "text": "加段稿号3000来访问", "chunk": 0, "language": "zh", "avg_logprob": -0.13215751118130153, "no_speech_prob": 1.9985006705081076e-11, "compression_ratio": 1.319327731092437 }, { "start": 443.54, "end": 445.5, "text": "同样使用刚才的账号来登录", "chunk": 0, "language": "zh", "avg_logprob": -0.13215751118130153, "no_speech_prob": 1.9985006705081076e-11, "compression_ratio": 1.319327731092437 }, { "start": 445.5, "end": 446.9, "text": "这时候可以随便问你问题", "chunk": 0, "language": "zh", "avg_logprob": -0.13215751118130153, "no_speech_prob": 1.9985006705081076e-11, "compression_ratio": 1.319327731092437 }, { "start": 446.9, "end": 449.38, "text": "看看局域网访问和模型交用是否正常", "chunk": 0, "language": "zh", "avg_logprob": -0.13215751118130153, "no_speech_prob": 1.9985006705081076e-11, "compression_ratio": 1.319327731092437 }, { "start": 449.38, "end": 451.04, "text": "如果其他设备都无法打开", "chunk": 0, "language": "zh", "avg_logprob": -0.13215751118130153, "no_speech_prob": 1.9985006705081076e-11, "compression_ratio": 1.319327731092437 }, { "start": 451.04, "end": 452.36, "text": "依次可以清查几个地方", "chunk": 0, "language": "zh", "avg_logprob": -0.13215751118130153, "no_speech_prob": 1.9985006705081076e-11, "compression_ratio": 1.319327731092437 }, { "start": 452.36, "end": 453.6, "text": "最后提醒一下", "chunk": 0, "language": "zh", "avg_logprob": -0.13215751118130153, "no_speech_prob": 1.9985006705081076e-11, "compression_ratio": 1.319327731092437 }, { "start": 453.6, "end": 455.64, "text": "这里搭建的家庭局域网AI服务器", "chunk": 0, "language": "zh", "avg_logprob": -0.13215751118130153, "no_speech_prob": 1.9985006705081076e-11, "compression_ratio": 1.319327731092437 }, { "start": 455.64, "end": 457.4, "text": "不会把服务暴露到公网", "chunk": 0, "language": "zh", "avg_logprob": -0.13215751118130153, "no_speech_prob": 1.9985006705081076e-11, "compression_ratio": 1.319327731092437 }, { "start": 457.4, "end": 459.6, "text": "因此不需要进行路由器短口转发", "chunk": 0, "language": "zh", "avg_logprob": -0.09548377990722656, "no_speech_prob": 1.2406990365643189e-11, "compression_ratio": 1.0912698412698412 }, { "start": 459.6, "end": 462.54, "text": "也不要直接把OpenWeb AI开放到互联网环境", "chunk": 0, "language": "zh", "avg_logprob": -0.09548377990722656, "no_speech_prob": 1.2406990365643189e-11, "compression_ratio": 1.0912698412698412 }, { "start": 462.54, "end": 465.0, "text": "而这样配置后保持终端App不关闭", "chunk": 0, "language": "zh", "avg_logprob": -0.09548377990722656, "no_speech_prob": 1.2406990365643189e-11, "compression_ratio": 1.0912698412698412 }, { "start": 465.0, "end": 466.7, "text": "AI模式运行在自己的Mac上", "chunk": 0, "language": "zh", "avg_logprob": -0.09548377990722656, "no_speech_prob": 1.2406990365643189e-11, "compression_ratio": 1.0912698412698412 }, { "start": 466.7, "end": 468.94, "text": "数据只在家庭网络里进行流转", "chunk": 0, "language": "zh", "avg_logprob": -0.09548377990722656, "no_speech_prob": 1.2406990365643189e-11, "compression_ratio": 1.0912698412698412 }, { "start": 468.94, "end": 472.86, "text": "同时手机平板和电脑都可以共享这一台本地AI服务器", "chunk": 0, "language": "zh", "avg_logprob": -0.09548377990722656, "no_speech_prob": 1.2406990365643189e-11, "compression_ratio": 1.0912698412698412 }, { "start": 478.028, "end": 484.368, "text": "前面OpenWeb UI已经可以这张访问了,接下来我们再给整个家庭AI服务器建立一个统一的API接口。", "chunk": 1, "language": "zh", "avg_logprob": -0.15065390674780446, "no_speech_prob": 9.781147246312738e-12, "compression_ratio": 1.18 }, { "start": 484.368, "end": 486.948, "text": "这样以后不管是N8N手机快捷指令", "chunk": 1, "language": "zh", "avg_logprob": -0.12399663374974178, "no_speech_prob": 1.9431465117514257e-11, "compression_ratio": 1.2932960893854748 }, { "start": 486.948, "end": 489.348, "text": "还是其他支持OpenAI API的程序", "chunk": 1, "language": "zh", "avg_logprob": -0.12399663374974178, "no_speech_prob": 1.9431465117514257e-11, "compression_ratio": 1.2932960893854748 }, { "start": 489.348, "end": 491.208, "text": "都统一连接OpenWeb UI", "chunk": 1, "language": "zh", "avg_logprob": -0.12399663374974178, "no_speech_prob": 1.9431465117514257e-11, "compression_ratio": 1.2932960893854748 }, { "start": 491.208, "end": 493.108, "text": "不需要直接访问底层的奥拉玛", "chunk": 1, "language": "zh", "avg_logprob": -0.12399663374974178, "no_speech_prob": 1.9431465117514257e-11, "compression_ratio": 1.2932960893854748 }, { "start": 493.108, "end": 494.148, "text": "这样的好处是", "chunk": 1, "language": "zh", "avg_logprob": -0.12399663374974178, "no_speech_prob": 1.9431465117514257e-11, "compression_ratio": 1.2932960893854748 }, { "start": 494.148, "end": 497.808, "text": "模型、照顽和权限都可以在OpenWeb UI这一层统一管理", "chunk": 1, "language": "zh", "avg_logprob": -0.12399663374974178, "no_speech_prob": 1.9431465117514257e-11, "compression_ratio": 1.2932960893854748 }, { "start": 497.808, "end": 499.048, "text": "后面更换模型时", "chunk": 1, "language": "zh", "avg_logprob": -0.12399663374974178, "no_speech_prob": 1.9431465117514257e-11, "compression_ratio": 1.2932960893854748 }, { "start": 499.048, "end": 501.088, "text": "也不用重新修改每一个应用", "chunk": 1, "language": "zh", "avg_logprob": -0.12399663374974178, "no_speech_prob": 1.9431465117514257e-11, "compression_ratio": 1.2932960893854748 }, { "start": 501.088, "end": 502.608, "text": "实际上咱们回到OpenWeb UI", "chunk": 1, "language": "zh", "avg_logprob": -0.12399663374974178, "no_speech_prob": 1.9431465117514257e-11, "compression_ratio": 1.2932960893854748 }, { "start": 502.608, "end": 505.068, "text": "点击左下角头像依次进入设置", "chunk": 1, "language": "zh", "avg_logprob": -0.12399663374974178, "no_speech_prob": 1.9431465117514257e-11, "compression_ratio": 1.2932960893854748 }, { "start": 505.068, "end": 507.568, "text": "身份验证找到API密钥并打开", "chunk": 1, "language": "zh", "avg_logprob": -0.12399663374974178, "no_speech_prob": 1.9431465117514257e-11, "compression_ratio": 1.2932960893854748 }, { "start": 507.568, "end": 510.188, "text": "如果界面提供API密钥端点限制", "chunk": 1, "language": "zh", "avg_logprob": -0.12399663374974178, "no_speech_prob": 1.9431465117514257e-11, "compression_ratio": 1.2932960893854748 }, { "start": 510.188, "end": 511.128, "text": "也一并开启", "chunk": 1, "language": "zh", "avg_logprob": -0.12399663374974178, "no_speech_prob": 1.9431465117514257e-11, "compression_ratio": 1.2932960893854748 }, { "start": 511.128, "end": 512.488, "text": "然后填写这个", "chunk": 1, "language": "zh", "avg_logprob": -0.12399663374974178, "no_speech_prob": 1.9431465117514257e-11, "compression_ratio": 1.2932960893854748 }, { "start": 512.488, "end": 517.368, "text": "这样生成的key就只允许访问模型列表和聊天接口", "chunk": 1, "language": "zh", "avg_logprob": -0.12123180995477695, "no_speech_prob": 1.783622902151727e-11, "compression_ratio": 1.3352112676056338 }, { "start": 517.368, "end": 519.328, "text": "接下来进入账号API MIAO", "chunk": 1, "language": "zh", "avg_logprob": -0.12123180995477695, "no_speech_prob": 1.783622902151727e-11, "compression_ratio": 1.3352112676056338 }, { "start": 519.328, "end": 522.048, "text": "点击右侧的显示创建新的API MIAO", "chunk": 1, "language": "zh", "avg_logprob": -0.12123180995477695, "no_speech_prob": 1.783622902151727e-11, "compression_ratio": 1.3352112676056338 }, { "start": 522.048, "end": 524.368, "text": "生成MIAO通常会以SK开头", "chunk": 1, "language": "zh", "avg_logprob": -0.12123180995477695, "no_speech_prob": 1.783622902151727e-11, "compression_ratio": 1.3352112676056338 }, { "start": 524.368, "end": 527.348, "text": "这个MIAO相当于访问我们家庭iF5系的密码", "chunk": 1, "language": "zh", "avg_logprob": -0.12123180995477695, "no_speech_prob": 1.783622902151727e-11, "compression_ratio": 1.3352112676056338 }, { "start": 527.348, "end": 529.328, "text": "不要直接公开或者分享出去", "chunk": 1, "language": "zh", "avg_logprob": -0.12123180995477695, "no_speech_prob": 1.783622902151727e-11, "compression_ratio": 1.3352112676056338 }, { "start": 529.328, "end": 532.648, "text": "如果后面准备长期让N8N这类的自动化工具使用", "chunk": 1, "language": "zh", "avg_logprob": -0.12123180995477695, "no_speech_prob": 1.783622902151727e-11, "compression_ratio": 1.3352112676056338 }, { "start": 532.648, "end": 535.148, "text": "建议专门创建一个普通的本地账号", "chunk": 1, "language": "zh", "avg_logprob": -0.12123180995477695, "no_speech_prob": 1.783622902151727e-11, "compression_ratio": 1.3352112676056338 }, { "start": 535.148, "end": 537.208, "text": "再用这个账号生成API MIAO", "chunk": 1, "language": "zh", "avg_logprob": -0.12123180995477695, "no_speech_prob": 1.783622902151727e-11, "compression_ratio": 1.3352112676056338 }, { "start": 537.208, "end": 539.388, "text": "不要长期使用管理员账号的MIAO", "chunk": 1, "language": "zh", "avg_logprob": -0.12123180995477695, "no_speech_prob": 1.783622902151727e-11, "compression_ratio": 1.3352112676056338 }, { "start": 539.388, "end": 542.048, "text": "因为MIAO会继续创建这个账号的权限", "chunk": 1, "language": "zh", "avg_logprob": -0.12123180995477695, "no_speech_prob": 1.783622902151727e-11, "compression_ratio": 1.3352112676056338 }, { "start": 542.048, "end": 544.868, "text": "而且一个账号通常只保留一个API密钥", "chunk": 1, "language": "zh", "avg_logprob": -0.09382172350613575, "no_speech_prob": 2.020636089006267e-11, "compression_ratio": 1.39247311827957 }, { "start": 544.868, "end": 546.968, "text": "接着我们来测试接口是否正常", "chunk": 1, "language": "zh", "avg_logprob": -0.09382172350613575, "no_speech_prob": 2.020636089006267e-11, "compression_ratio": 1.39247311827957 }, { "start": 546.968, "end": 549.008, "text": "打开一个新的终端APP窗口", "chunk": 1, "language": "zh", "avg_logprob": -0.09382172350613575, "no_speech_prob": 2.020636089006267e-11, "compression_ratio": 1.39247311827957 }, { "start": 549.008, "end": 552.328, "text": "把下面的命令中的SK换成刚才生成的密钥", "chunk": 1, "language": "zh", "avg_logprob": -0.09382172350613575, "no_speech_prob": 2.020636089006267e-11, "compression_ratio": 1.39247311827957 }, { "start": 552.328, "end": 553.588, "text": "如果接口正常", "chunk": 1, "language": "zh", "avg_logprob": -0.09382172350613575, "no_speech_prob": 2.020636089006267e-11, "compression_ratio": 1.39247311827957 }, { "start": 553.588, "end": 556.888, "text": "终端会访问当前OpenWeb UI可以调用的模型列表", "chunk": 1, "language": "zh", "avg_logprob": -0.09382172350613575, "no_speech_prob": 2.020636089006267e-11, "compression_ratio": 1.39247311827957 }, { "start": 556.888, "end": 560.308, "text": "这里可以顺便记一下返回结果里模型队的真实ID", "chunk": 1, "language": "zh", "avg_logprob": -0.09382172350613575, "no_speech_prob": 2.020636089006267e-11, "compression_ratio": 1.39247311827957 }, { "start": 560.308, "end": 563.488, "text": "因为后面其他地方对用AI时就需要填写这个名称", "chunk": 1, "language": "zh", "avg_logprob": -0.09382172350613575, "no_speech_prob": 2.020636089006267e-11, "compression_ratio": 1.39247311827957 }, { "start": 563.488, "end": 565.528, "text": "接下来再来测试聊天接口", "chunk": 1, "language": "zh", "avg_logprob": -0.09382172350613575, "no_speech_prob": 2.020636089006267e-11, "compression_ratio": 1.39247311827957 }, { "start": 565.528, "end": 568.908, "text": "同样的记得把SK开头的这串换成刚才生成的密钥", "chunk": 1, "language": "zh", "avg_logprob": -0.09382172350613575, "no_speech_prob": 2.020636089006267e-11, "compression_ratio": 1.39247311827957 }, { "start": 568.908, "end": 571.508, "text": "而Model后面需要填写刚才模型列表里面", "chunk": 1, "language": "zh", "avg_logprob": -0.09382172350613575, "no_speech_prob": 2.020636089006267e-11, "compression_ratio": 1.39247311827957 }, { "start": 571.508, "end": 572.648, "text": "实际反规的ID", "chunk": 1, "language": "zh", "avg_logprob": -0.2089262411627971, "no_speech_prob": 3.7000107838691676e-11, "compression_ratio": 1.25 }, { "start": 572.648, "end": 575.108, "text": "如果你下载的是其他签问或者其他模型", "chunk": 1, "language": "zh", "avg_logprob": -0.2089262411627971, "no_speech_prob": 3.7000107838691676e-11, "compression_ratio": 1.25 }, { "start": 575.108, "end": 576.508, "text": "就替完成对应的模型", "chunk": 1, "language": "zh", "avg_logprob": -0.2089262411627971, "no_speech_prob": 3.7000107838691676e-11, "compression_ratio": 1.25 }, { "start": 576.508, "end": 579.168, "text": "如果最后反规Json你能够看到API正常", "chunk": 1, "language": "zh", "avg_logprob": -0.2089262411627971, "no_speech_prob": 3.7000107838691676e-11, "compression_ratio": 1.25 }, { "start": 579.168, "end": 581.748, "text": "就说明这一套接口已经可以正常使用了", "chunk": 1, "language": "zh", "avg_logprob": -0.2089262411627971, "no_speech_prob": 3.7000107838691676e-11, "compression_ratio": 1.25 }, { "start": 582.008, "end": 583.848, "text": "签问3.5在开启思考时", "chunk": 1, "language": "zh", "avg_logprob": -0.2089262411627971, "no_speech_prob": 3.7000107838691676e-11, "compression_ratio": 1.25 }, { "start": 583.848, "end": 585.248, "text": "反规内容里可能还会出现", "chunk": 1, "language": "zh", "avg_logprob": -0.2089262411627971, "no_speech_prob": 3.7000107838691676e-11, "compression_ratio": 1.25 }, { "start": 585.248, "end": 587.108, "text": "Reasoning Content这一类字段", "chunk": 1, "language": "zh", "avg_logprob": -0.2089262411627971, "no_speech_prob": 3.7000107838691676e-11, "compression_ratio": 1.25 }, { "start": 587.108, "end": 588.168, "text": "这是正常现象", "chunk": 1, "language": "zh", "avg_logprob": -0.2089262411627971, "no_speech_prob": 3.7000107838691676e-11, "compression_ratio": 1.25 }, { "start": 588.408, "end": 590.608, "text": "到这里我们就相当于给整个家庭", "chunk": 1, "language": "zh", "avg_logprob": -0.2089262411627971, "no_speech_prob": 3.7000107838691676e-11, "compression_ratio": 1.25 }, { "start": 590.608, "end": 592.448, "text": "AI服务器建了一个统一的入口", "chunk": 1, "language": "zh", "avg_logprob": -0.2089262411627971, "no_speech_prob": 3.7000107838691676e-11, "compression_ratio": 1.25 }, { "start": 592.748, "end": 595.008, "text": "最底层人是Olama负责运行模型", "chunk": 1, "language": "zh", "avg_logprob": -0.2089262411627971, "no_speech_prob": 3.7000107838691676e-11, "compression_ratio": 1.25 }, { "start": 595.168, "end": 596.508, "text": "上面的OpenWeb UI", "chunk": 1, "language": "zh", "avg_logprob": -0.2089262411627971, "no_speech_prob": 3.7000107838691676e-11, "compression_ratio": 1.25 }, { "start": 596.508, "end": 598.208, "text": "统一管理模型和API", "chunk": 1, "language": "zh", "avg_logprob": -0.2089262411627971, "no_speech_prob": 3.7000107838691676e-11, "compression_ratio": 1.25 }, { "start": 598.348, "end": 600.948, "text": "在网上则可以接N8N手机快点指令", "chunk": 1, "language": "zh", "avg_logprob": -0.2089262411627971, "no_speech_prob": 3.7000107838691676e-11, "compression_ratio": 1.25 }, { "start": 600.948, "end": 601.988, "text": "或者其他应用", "chunk": 1, "language": "zh", "avg_logprob": -0.1882660736908784, "no_speech_prob": 2.5263190844837524e-11, "compression_ratio": 1.175 }, { "start": 601.988, "end": 603.668, "text": "后面不管增加多少个客户端", "chunk": 1, "language": "zh", "avg_logprob": -0.1882660736908784, "no_speech_prob": 2.5263190844837524e-11, "compression_ratio": 1.175 }, { "start": 603.668, "end": 605.468, "text": "都尽量连接OpenWeb UI", "chunk": 1, "language": "zh", "avg_logprob": -0.1882660736908784, "no_speech_prob": 2.5263190844837524e-11, "compression_ratio": 1.175 }, { "start": 605.468, "end": 607.828, "text": "而不是让每台设备直接访问欧拉玛", "chunk": 1, "language": "zh", "avg_logprob": -0.1882660736908784, "no_speech_prob": 2.5263190844837524e-11, "compression_ratio": 1.175 }, { "start": 607.828, "end": 609.628, "text": "整个架构会更容易管理", "chunk": 1, "language": "zh", "avg_logprob": -0.1882660736908784, "no_speech_prob": 2.5263190844837524e-11, "compression_ratio": 1.175 }, { "start": 614.628, "end": 617.588, "text": "前面我们已经把本地AI和聊天的API都搭好了", "chunk": 1, "language": "zh", "avg_logprob": -0.1882660736908784, "no_speech_prob": 2.5263190844837524e-11, "compression_ratio": 1.175 }, { "start": 617.588, "end": 620.188, "text": "接下来就再补上一个非常熟能功能", "chunk": 1, "language": "zh", "avg_logprob": -0.1882660736908784, "no_speech_prob": 2.5263190844837524e-11, "compression_ratio": 1.175 }, { "start": 620.188, "end": 621.428, "text": "Cell文档知识库", "chunk": 1, "language": "zh", "avg_logprob": -0.1882660736908784, "no_speech_prob": 2.5263190844837524e-11, "compression_ratio": 1.175 }, { "start": 621.428, "end": 622.748, "text": "也就是REG", "chunk": 1, "language": "zh", "avg_logprob": -0.1882660736908784, "no_speech_prob": 2.5263190844837524e-11, "compression_ratio": 1.175 }, { "start": 622.748, "end": 623.948, "text": "REG可以列成", "chunk": 1, "language": "zh", "avg_logprob": -0.1882660736908784, "no_speech_prob": 2.5263190844837524e-11, "compression_ratio": 1.175 }, { "start": 623.948, "end": 625.508, "text": "先从你的资料里找内容", "chunk": 1, "language": "zh", "avg_logprob": -0.1882660736908784, "no_speech_prob": 2.5263190844837524e-11, "compression_ratio": 1.175 }, { "start": 625.508, "end": 627.068, "text": "再交给AI声称回答", "chunk": 1, "language": "zh", "avg_logprob": -0.1882660736908784, "no_speech_prob": 2.5263190844837524e-11, "compression_ratio": 1.175 }, { "start": 627.068, "end": 628.748, "text": "这样以后不管是自己的PDF", "chunk": 1, "language": "zh", "avg_logprob": -0.1882660736908784, "no_speech_prob": 2.5263190844837524e-11, "compression_ratio": 1.175 }, { "start": 628.748, "end": 629.468, "text": "说明书", "chunk": 1, "language": "zh", "avg_logprob": -0.1882660736908784, "no_speech_prob": 2.5263190844837524e-11, "compression_ratio": 1.175 }, { "start": 629.468, "end": 630.108, "text": "工作文档", "chunk": 1, "language": "zh", "avg_logprob": -0.1882660736908784, "no_speech_prob": 2.5263190844837524e-11, "compression_ratio": 1.175 }, { "start": 630.108, "end": 630.908, "text": "还是视频稿", "chunk": 1, "language": "zh", "avg_logprob": -0.1882660736908784, "no_speech_prob": 2.5263190844837524e-11, "compression_ratio": 1.175 }, { "start": 630.908, "end": 632.028, "text": "都可以放进知识库", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 632.028, "end": 634.268, "text": "让AI尽量根据这些资料来回答", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 634.268, "end": 636.248, "text": "而不是只靠模型本身的记忆", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 636.248, "end": 638.988, "text": "首先进入OpenWeb UI试试里的文档页面", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 638.988, "end": 640.888, "text": "轻入模型引擎选择Olama", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 640.888, "end": 642.248, "text": "API MIO可以留空", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 642.248, "end": 644.668, "text": "地址填写127.0.0.1", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 644.668, "end": 646.028, "text": "冒号11434", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 646.028, "end": 647.108, "text": "轻入模型输入", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 647.108, "end": 648.828, "text": "我们在视频一开始就下载好的", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 648.828, "end": 650.888, "text": "轻入3-embedding0.6b", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 650.888, "end": 652.448, "text": "而这里顺便解释一下", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 652.448, "end": 653.868, "text": "这个模型不负责聊天", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 653.868, "end": 655.288, "text": "它的工作是把文档", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 655.288, "end": 657.308, "text": "转完成便于检索的向量数据", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 657.308, "end": 658.508, "text": "只要你提问以后", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 658.508, "end": 659.728, "text": "系统才能快速找到", "chunk": 1, "language": "zh", "avg_logprob": -0.15276504311326372, "no_speech_prob": 2.6165573180358948e-11, "compression_ratio": 1.1881443298969072 }, { "start": 659.728, "end": 661.148, "text": "和问题最相关的资料", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 661.148, "end": 663.288, "text": "其他参数可以先按照这一套来设置", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 663.288, "end": 664.628, "text": "大部分是我认的", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 664.628, "end": 666.388, "text": "这些参数主要是在减速效果", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 666.388, "end": 667.668, "text": "和上下文本占用之间", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 667.668, "end": 669.208, "text": "做一个比较稳妥的平衡", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 669.208, "end": 670.748, "text": "比如快大小可以列列成", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 670.748, "end": 672.668, "text": "每次把文档切成多大片段", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 672.668, "end": 674.808, "text": "快重叠则是在相邻片段之间", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 674.808, "end": 676.008, "text": "保留的部分重复内容", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 676.008, "end": 677.848, "text": "避免一句话刚好被重点切断", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 677.848, "end": 679.448, "text": "TableK代表每次提问", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 679.448, "end": 681.508, "text": "最多取回多少个相关片段", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 681.508, "end": 682.968, "text": "这里先设置为3", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 682.968, "end": 684.188, "text": "并不是越大越好", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 684.188, "end": 685.568, "text": "因为塞进太多资料", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 685.568, "end": 687.988, "text": "反而会占模型有限的上下文空间", "chunk": 1, "language": "zh", "avg_logprob": -0.09741358439127604, "no_speech_prob": 3.106967461086185e-11, "compression_ratio": 1.3591731266149871 }, { "start": 687.988, "end": 690.268, "text": "接下来我们创建一个非常简单的测试文件", "chunk": 1, "language": "zh", "avg_logprob": -0.08299861833887193, "no_speech_prob": 1.728324428018624e-11, "compression_ratio": 1.2681564245810055 }, { "start": 690.268, "end": 692.728, "text": "方便确认RIG是否真的正常工作", "chunk": 1, "language": "zh", "avg_logprob": -0.08299861833887193, "no_speech_prob": 1.728324428018624e-11, "compression_ratio": 1.2681564245810055 }, { "start": 692.728, "end": 695.848, "text": "用任意文本编定器新建一个TXT文本", "chunk": 1, "language": "zh", "avg_logprob": -0.08299861833887193, "no_speech_prob": 1.728324428018624e-11, "compression_ratio": 1.2681564245810055 }, { "start": 695.848, "end": 698.688, "text": "如果你用的是MicroOS自带的文本编辑App", "chunk": 1, "language": "zh", "avg_logprob": -0.08299861833887193, "no_speech_prob": 1.728324428018624e-11, "compression_ratio": 1.2681564245810055 }, { "start": 698.688, "end": 700.828, "text": "记得在菜单里选择制作手文本", "chunk": 1, "language": "zh", "avg_logprob": -0.08299861833887193, "no_speech_prob": 1.728324428018624e-11, "compression_ratio": 1.2681564245810055 }, { "start": 700.828, "end": 702.588, "text": "比如里面写上这些内容", "chunk": 1, "language": "zh", "avg_logprob": -0.08299861833887193, "no_speech_prob": 1.728324428018624e-11, "compression_ratio": 1.2681564245810055 }, { "start": 702.588, "end": 705.448, "text": "然后保存为家庭AI测试资料.txt", "chunk": 1, "language": "zh", "avg_logprob": -0.08299861833887193, "no_speech_prob": 1.728324428018624e-11, "compression_ratio": 1.2681564245810055 }, { "start": 705.448, "end": 706.628, "text": "回到OpenWeb UI", "chunk": 1, "language": "zh", "avg_logprob": -0.08299861833887193, "no_speech_prob": 1.728324428018624e-11, "compression_ratio": 1.2681564245810055 }, { "start": 706.628, "end": 708.528, "text": "找到左侧边来的工作空间", "chunk": 1, "language": "zh", "avg_logprob": -0.08299861833887193, "no_speech_prob": 1.728324428018624e-11, "compression_ratio": 1.2681564245810055 }, { "start": 708.528, "end": 710.068, "text": "再点击顶部的知识库", "chunk": 1, "language": "zh", "avg_logprob": -0.08299861833887193, "no_speech_prob": 1.728324428018624e-11, "compression_ratio": 1.2681564245810055 }, { "start": 710.068, "end": 711.948, "text": "最后点击右上角的创建", "chunk": 1, "language": "zh", "avg_logprob": -0.08299861833887193, "no_speech_prob": 1.728324428018624e-11, "compression_ratio": 1.2681564245810055 }, { "start": 711.948, "end": 714.148, "text": "名称填写家庭AI测试库", "chunk": 1, "language": "zh", "avg_logprob": -0.08299861833887193, "no_speech_prob": 1.728324428018624e-11, "compression_ratio": 1.2681564245810055 }, { "start": 714.148, "end": 715.408, "text": "目标可以简单写真", "chunk": 1, "language": "zh", "avg_logprob": -0.08299861833887193, "no_speech_prob": 1.728324428018624e-11, "compression_ratio": 1.2681564245810055 }, { "start": 715.408, "end": 718.648, "text": "用于测试本地RIG文档检索和应用功能", "chunk": 1, "language": "zh", "avg_logprob": -0.10378520829336983, "no_speech_prob": 2.0449478915773867e-11, "compression_ratio": 1.3074866310160427 }, { "start": 718.648, "end": 720.048, "text": "知识库创建完之后", "chunk": 1, "language": "zh", "avg_logprob": -0.10378520829336983, "no_speech_prob": 2.0449478915773867e-11, "compression_ratio": 1.3074866310160427 }, { "start": 720.048, "end": 722.428, "text": "再点击右上角的加号选择上传文件", "chunk": 1, "language": "zh", "avg_logprob": -0.10378520829336983, "no_speech_prob": 2.0449478915773867e-11, "compression_ratio": 1.3074866310160427 }, { "start": 722.428, "end": 724.948, "text": "把刚才的TXD文件上传进去", "chunk": 1, "language": "zh", "avg_logprob": -0.10378520829336983, "no_speech_prob": 2.0449478915773867e-11, "compression_ratio": 1.3074866310160427 }, { "start": 724.948, "end": 727.768, "text": "后面也可以上传带有文字层的PDF", "chunk": 1, "language": "zh", "avg_logprob": -0.10378520829336983, "no_speech_prob": 2.0449478915773867e-11, "compression_ratio": 1.3074866310160427 }, { "start": 727.768, "end": 729.688, "text": "DOCX或Markdown文档", "chunk": 1, "language": "zh", "avg_logprob": -0.10378520829336983, "no_speech_prob": 2.0449478915773867e-11, "compression_ratio": 1.3074866310160427 }, { "start": 729.688, "end": 730.608, "text": "上传完之后", "chunk": 1, "language": "zh", "avg_logprob": -0.10378520829336983, "no_speech_prob": 2.0449478915773867e-11, "compression_ratio": 1.3074866310160427 }, { "start": 730.608, "end": 732.328, "text": "在新兑化的聊天输入框里", "chunk": 1, "language": "zh", "avg_logprob": -0.10378520829336983, "no_speech_prob": 2.0449478915773867e-11, "compression_ratio": 1.3074866310160427 }, { "start": 732.328, "end": 734.688, "text": "通过警号手动附加这个知识库", "chunk": 1, "language": "zh", "avg_logprob": -0.10378520829336983, "no_speech_prob": 2.0449478915773867e-11, "compression_ratio": 1.3074866310160427 }, { "start": 734.688, "end": 736.728, "text": "而我这里为了教程演示方便", "chunk": 1, "language": "zh", "avg_logprob": -0.10378520829336983, "no_speech_prob": 2.0449478915773867e-11, "compression_ratio": 1.3074866310160427 }, { "start": 736.728, "end": 737.668, "text": "先手动选择", "chunk": 1, "language": "zh", "avg_logprob": -0.10378520829336983, "no_speech_prob": 2.0449478915773867e-11, "compression_ratio": 1.3074866310160427 }, { "start": 737.668, "end": 739.708, "text": "因为小参数模型主动判断", "chunk": 1, "language": "zh", "avg_logprob": -0.10378520829336983, "no_speech_prob": 2.0449478915773867e-11, "compression_ratio": 1.3074866310160427 }, { "start": 739.708, "end": 741.868, "text": "什么时候动用知识库有时候不够稳定", "chunk": 1, "language": "zh", "avg_logprob": -0.10378520829336983, "no_speech_prob": 2.0449478915773867e-11, "compression_ratio": 1.3074866310160427 }, { "start": 741.868, "end": 743.668, "text": "然后我们可以问一个问题", "chunk": 1, "language": "zh", "avg_logprob": -0.10378520829336983, "no_speech_prob": 2.0449478915773867e-11, "compression_ratio": 1.3074866310160427 }, { "start": 743.668, "end": 745.088, "text": "只依据知识库回答", "chunk": 1, "language": "zh", "avg_logprob": -0.10378520829336983, "no_speech_prob": 2.0449478915773867e-11, "compression_ratio": 1.3074866310160427 }, { "start": 745.088, "end": 747.648, "text": "项目代号和设备维护时间分配是什么", "chunk": 1, "language": "zh", "avg_logprob": -0.07785890687186763, "no_speech_prob": 1.76431508291941e-11, "compression_ratio": 1.3263707571801566 }, { "start": 747.648, "end": 748.808, "text": "请给出引用", "chunk": 1, "language": "zh", "avg_logprob": -0.07785890687186763, "no_speech_prob": 1.76431508291941e-11, "compression_ratio": 1.3263707571801566 }, { "start": 748.808, "end": 750.048, "text": "正常情况下", "chunk": 1, "language": "zh", "avg_logprob": -0.07785890687186763, "no_speech_prob": 1.76431508291941e-11, "compression_ratio": 1.3263707571801566 }, { "start": 750.048, "end": 750.968, "text": "耐心等待一会儿", "chunk": 1, "language": "zh", "avg_logprob": -0.07785890687186763, "no_speech_prob": 1.76431508291941e-11, "compression_ratio": 1.3263707571801566 }, { "start": 750.968, "end": 753.448, "text": "他应该回答项目代号是70", "chunk": 1, "language": "zh", "avg_logprob": -0.07785890687186763, "no_speech_prob": 1.76431508291941e-11, "compression_ratio": 1.3263707571801566 }, { "start": 753.448, "end": 756.368, "text": "设备维护时间是每月第一周六上午10点", "chunk": 1, "language": "zh", "avg_logprob": -0.07785890687186763, "no_speech_prob": 1.76431508291941e-11, "compression_ratio": 1.3263707571801566 }, { "start": 756.368, "end": 759.068, "text": "而且回答旁边能够看到对应的资料引用", "chunk": 1, "language": "zh", "avg_logprob": -0.07785890687186763, "no_speech_prob": 1.76431508291941e-11, "compression_ratio": 1.3263707571801566 }, { "start": 759.068, "end": 762.888, "text": "那么这套最基础的REG知识库就算大奖成功了", "chunk": 1, "language": "zh", "avg_logprob": -0.07785890687186763, "no_speech_prob": 1.76431508291941e-11, "compression_ratio": 1.3263707571801566 }, { "start": 762.888, "end": 763.988, "text": "不过这样配置下来", "chunk": 1, "language": "zh", "avg_logprob": -0.07785890687186763, "no_speech_prob": 1.76431508291941e-11, "compression_ratio": 1.3263707571801566 }, { "start": 763.988, "end": 766.088, "text": "你可能会发现还有一个地方容易踩坑", "chunk": 1, "language": "zh", "avg_logprob": -0.07785890687186763, "no_speech_prob": 1.76431508291941e-11, "compression_ratio": 1.3263707571801566 }, { "start": 766.088, "end": 767.868, "text": "如果你的PDF是扫描件", "chunk": 1, "language": "zh", "avg_logprob": -0.07785890687186763, "no_speech_prob": 1.76431508291941e-11, "compression_ratio": 1.3263707571801566 }, { "start": 767.868, "end": 770.048, "text": "本身没有可以复制的文字层", "chunk": 1, "language": "zh", "avg_logprob": -0.07785890687186763, "no_speech_prob": 1.76431508291941e-11, "compression_ratio": 1.3263707571801566 }, { "start": 770.048, "end": 772.008, "text": "就不要拿它来测试REG了", "chunk": 1, "language": "zh", "avg_logprob": -0.07785890687186763, "no_speech_prob": 1.76431508291941e-11, "compression_ratio": 1.3263707571801566 }, { "start": 772.008, "end": 774.788, "text": "因为知识库首先需要能够读取里面的文字", "chunk": 1, "language": "zh", "avg_logprob": -0.07785890687186763, "no_speech_prob": 1.76431508291941e-11, "compression_ratio": 1.3263707571801566 }, { "start": 774.788, "end": 776.048, "text": "才能进行后续的处理", "chunk": 1, "language": "zh", "avg_logprob": -0.08606964785878252, "no_speech_prob": 1.478953938482963e-11, "compression_ratio": 1.274247491638796 }, { "start": 776.048, "end": 778.308, "text": "所以这种文件就需要先经过OCR", "chunk": 1, "language": "zh", "avg_logprob": -0.08606964785878252, "no_speech_prob": 1.478953938482963e-11, "compression_ratio": 1.274247491638796 }, { "start": 778.308, "end": 781.328, "text": "把图片里的文字设备成TXC或者Markdown", "chunk": 1, "language": "zh", "avg_logprob": -0.08606964785878252, "no_speech_prob": 1.478953938482963e-11, "compression_ratio": 1.274247491638796 }, { "start": 781.328, "end": 783.268, "text": "再上传到知识库来进行处理", "chunk": 1, "language": "zh", "avg_logprob": -0.08606964785878252, "no_speech_prob": 1.478953938482963e-11, "compression_ratio": 1.274247491638796 }, { "start": 783.268, "end": 787.388, "text": "接下来我们就继续把本地OCR的功能也接进来", "chunk": 1, "language": "zh", "avg_logprob": -0.08606964785878252, "no_speech_prob": 1.478953938482963e-11, "compression_ratio": 1.274247491638796 }, { "start": 787.388, "end": 794.788, "text": "前面我们已经让AI能够读取自的文档", "chunk": 1, "language": "zh", "avg_logprob": -0.08606964785878252, "no_speech_prob": 1.478953938482963e-11, "compression_ratio": 1.274247491638796 }, { "start": 794.788, "end": 796.988, "text": "接下来再补上图片理解和OCR", "chunk": 1, "language": "zh", "avg_logprob": -0.08606964785878252, "no_speech_prob": 1.478953938482963e-11, "compression_ratio": 1.274247491638796 }, { "start": 796.988, "end": 799.808, "text": "也就是让这台本地AI不只能读文字", "chunk": 1, "language": "zh", "avg_logprob": -0.08606964785878252, "no_speech_prob": 1.478953938482963e-11, "compression_ratio": 1.274247491638796 }, { "start": 799.808, "end": 800.948, "text": "还可以看图片", "chunk": 1, "language": "zh", "avg_logprob": -0.08606964785878252, "no_speech_prob": 1.478953938482963e-11, "compression_ratio": 1.274247491638796 }, { "start": 800.948, "end": 802.868, "text": "那这里先区分两个概念", "chunk": 1, "language": "zh", "avg_logprob": -0.08606964785878252, "no_speech_prob": 1.478953938482963e-11, "compression_ratio": 1.274247491638796 }, { "start": 802.868, "end": 805.448, "text": "普通图片理解更适合看一张图里有什么", "chunk": 1, "language": "zh", "avg_logprob": -0.10298146747407459, "no_speech_prob": 2.01446116732118e-11, "compression_ratio": 1.2520325203252032 }, { "start": 805.448, "end": 806.708, "text": "界面再显示什么", "chunk": 1, "language": "zh", "avg_logprob": -0.10298146747407459, "no_speech_prob": 2.01446116732118e-11, "compression_ratio": 1.2520325203252032 }, { "start": 806.708, "end": 808.488, "text": "再顺便识别其中的文字", "chunk": 1, "language": "zh", "avg_logprob": -0.10298146747407459, "no_speech_prob": 2.01446116732118e-11, "compression_ratio": 1.2520325203252032 }, { "start": 808.488, "end": 811.488, "text": "OCR则更专注于把图片里的文字", "chunk": 1, "language": "zh", "avg_logprob": -0.10298146747407459, "no_speech_prob": 2.01446116732118e-11, "compression_ratio": 1.2520325203252032 }, { "start": 811.488, "end": 814.128, "text": "表格或者图表内容尽可能提取出来", "chunk": 1, "language": "zh", "avg_logprob": -0.10298146747407459, "no_speech_prob": 2.01446116732118e-11, "compression_ratio": 1.2520325203252032 }, { "start": 814.128, "end": 816.968, "text": "先在OpenWeb UI里选择我们前面使用的", "chunk": 1, "language": "zh", "avg_logprob": -0.10298146747407459, "no_speech_prob": 2.01446116732118e-11, "compression_ratio": 1.2520325203252032 }, { "start": 816.968, "end": 819.348, "text": "切纹3.5 4B-KL4KM", "chunk": 1, "language": "zh", "avg_logprob": -0.10298146747407459, "no_speech_prob": 2.01446116732118e-11, "compression_ratio": 1.2520325203252032 }, { "start": 819.348, "end": 820.808, "text": "如果你的Mac配置更高", "chunk": 1, "language": "zh", "avg_logprob": -0.10298146747407459, "no_speech_prob": 2.01446116732118e-11, "compression_ratio": 1.2520325203252032 }, { "start": 820.808, "end": 822.068, "text": "也可以用9B的版本", "chunk": 1, "language": "zh", "avg_logprob": -0.10298146747407459, "no_speech_prob": 2.01446116732118e-11, "compression_ratio": 1.2520325203252032 }, { "start": 822.068, "end": 824.908, "text": "上传一张同时包含物体和文字图片", "chunk": 1, "language": "zh", "avg_logprob": -0.10298146747407459, "no_speech_prob": 2.01446116732118e-11, "compression_ratio": 1.2520325203252032 }, { "start": 824.908, "end": 825.588, "text": "然后输入", "chunk": 1, "language": "zh", "avg_logprob": -0.10298146747407459, "no_speech_prob": 2.01446116732118e-11, "compression_ratio": 1.2520325203252032 }, { "start": 825.588, "end": 826.788, "text": "先描述画面内容", "chunk": 1, "language": "zh", "avg_logprob": -0.10298146747407459, "no_speech_prob": 2.01446116732118e-11, "compression_ratio": 1.2520325203252032 }, { "start": 826.788, "end": 828.528, "text": "再逐项提取可见文字", "chunk": 1, "language": "zh", "avg_logprob": -0.10298146747407459, "no_speech_prob": 2.01446116732118e-11, "compression_ratio": 1.2520325203252032 }, { "start": 828.528, "end": 830.488, "text": "不确定的部分标注为不确定", "chunk": 1, "language": "zh", "avg_logprob": -0.10298146747407459, "no_speech_prob": 2.01446116732118e-11, "compression_ratio": 1.2520325203252032 }, { "start": 830.488, "end": 831.588, "text": "不需要自信补圈", "chunk": 1, "language": "zh", "avg_logprob": -0.10298146747407459, "no_speech_prob": 2.01446116732118e-11, "compression_ratio": 1.2520325203252032 }, { "start": 831.588, "end": 833.728, "text": "这里可以分别拿产品包装", "chunk": 1, "language": "zh", "avg_logprob": -0.09114494323730468, "no_speech_prob": 1.9151418298446465e-11, "compression_ratio": 1.2102689486552567 }, { "start": 833.728, "end": 834.708, "text": "软件界面截图", "chunk": 1, "language": "zh", "avg_logprob": -0.09114494323730468, "no_speech_prob": 1.9151418298446465e-11, "compression_ratio": 1.2102689486552567 }, { "start": 834.708, "end": 835.468, "text": "中文表格", "chunk": 1, "language": "zh", "avg_logprob": -0.09114494323730468, "no_speech_prob": 1.9151418298446465e-11, "compression_ratio": 1.2102689486552567 }, { "start": 835.468, "end": 839.168, "text": "或者已经做好隐私透明的发票和数据照片来测试一下", "chunk": 1, "language": "zh", "avg_logprob": -0.09114494323730468, "no_speech_prob": 1.9151418298446465e-11, "compression_ratio": 1.2102689486552567 }, { "start": 839.168, "end": 841.128, "text": "这样可以比较直观的判断出", "chunk": 1, "language": "zh", "avg_logprob": -0.09114494323730468, "no_speech_prob": 1.9151418298446465e-11, "compression_ratio": 1.2102689486552567 }, { "start": 841.128, "end": 842.948, "text": "模型到底是理解了图片内容", "chunk": 1, "language": "zh", "avg_logprob": -0.09114494323730468, "no_speech_prob": 1.9151418298446465e-11, "compression_ratio": 1.2102689486552567 }, { "start": 842.948, "end": 845.388, "text": "还是只是把里面的文字给识别出来", "chunk": 1, "language": "zh", "avg_logprob": -0.09114494323730468, "no_speech_prob": 1.9151418298446465e-11, "compression_ratio": 1.2102689486552567 }, { "start": 845.388, "end": 847.468, "text": "如果你出现类似我这样的提示", "chunk": 1, "language": "zh", "avg_logprob": -0.09114494323730468, "no_speech_prob": 1.9151418298446465e-11, "compression_ratio": 1.2102689486552567 }, { "start": 847.468, "end": 849.148, "text": "就前往OpenWeb UI的设置", "chunk": 1, "language": "zh", "avg_logprob": -0.09114494323730468, "no_speech_prob": 1.9151418298446465e-11, "compression_ratio": 1.2102689486552567 }, { "start": 849.148, "end": 849.708, "text": "模型", "chunk": 1, "language": "zh", "avg_logprob": -0.09114494323730468, "no_speech_prob": 1.9151418298446465e-11, "compression_ratio": 1.2102689486552567 }, { "start": 849.708, "end": 851.548, "text": "找到刚才你前面使用的模型", "chunk": 1, "language": "zh", "avg_logprob": -0.09114494323730468, "no_speech_prob": 1.9151418298446465e-11, "compression_ratio": 1.2102689486552567 }, { "start": 851.548, "end": 852.668, "text": "进入编辑界面", "chunk": 1, "language": "zh", "avg_logprob": -0.09114494323730468, "no_speech_prob": 1.9151418298446465e-11, "compression_ratio": 1.2102689486552567 }, { "start": 852.668, "end": 855.728, "text": "找到高级参数下的NUM CTX OLAMA", "chunk": 1, "language": "zh", "avg_logprob": -0.09114494323730468, "no_speech_prob": 1.9151418298446465e-11, "compression_ratio": 1.2102689486552567 }, { "start": 855.728, "end": 857.468, "text": "针对我16GB的Mac mini", "chunk": 1, "language": "zh", "avg_logprob": -0.09114494323730468, "no_speech_prob": 1.9151418298446465e-11, "compression_ratio": 1.2102689486552567 }, { "start": 857.468, "end": 859.088, "text": "将数字更改为8192", "chunk": 1, "language": "zh", "avg_logprob": -0.09114494323730468, "no_speech_prob": 1.9151418298446465e-11, "compression_ratio": 1.2102689486552567 }, { "start": 859.088, "end": 861.168, "text": "然后重复上的操作就可以了", "chunk": 1, "language": "zh", "avg_logprob": -0.09114494323730468, "no_speech_prob": 1.9151418298446465e-11, "compression_ratio": 1.2102689486552567 }, { "start": 861.168, "end": 862.908, "text": "如果我们的目标不是理解图片", "chunk": 1, "language": "zh", "avg_logprob": -0.09943887263393858, "no_speech_prob": 1.9523983124658528e-11, "compression_ratio": 1.4585635359116023 }, { "start": 862.908, "end": 865.368, "text": "而是尽可能把图片里的文字给提取出来", "chunk": 1, "language": "zh", "avg_logprob": -0.09943887263393858, "no_speech_prob": 1.9523983124658528e-11, "compression_ratio": 1.4585635359116023 }, { "start": 865.368, "end": 868.528, "text": "就可以使用我们前面已经下载好的GLM OCR", "chunk": 1, "language": "zh", "avg_logprob": -0.09943887263393858, "no_speech_prob": 1.9523983124658528e-11, "compression_ratio": 1.4585635359116023 }, { "start": 868.528, "end": 870.048, "text": "OCR模型运行前", "chunk": 1, "language": "zh", "avg_logprob": -0.09943887263393858, "no_speech_prob": 1.9523983124658528e-11, "compression_ratio": 1.4585635359116023 }, { "start": 870.048, "end": 872.288, "text": "如果较大的千万模型还占着同一印寸", "chunk": 1, "language": "zh", "avg_logprob": -0.09943887263393858, "no_speech_prob": 1.9523983124658528e-11, "compression_ratio": 1.4585635359116023 }, { "start": 872.288, "end": 873.648, "text": "先可以把它停掉", "chunk": 1, "language": "zh", "avg_logprob": -0.09943887263393858, "no_speech_prob": 1.9523983124658528e-11, "compression_ratio": 1.4585635359116023 }, { "start": 873.648, "end": 875.848, "text": "接下来我们准备好需要识别图片", "chunk": 1, "language": "zh", "avg_logprob": -0.09943887263393858, "no_speech_prob": 1.9523983124658528e-11, "compression_ratio": 1.4585635359116023 }, { "start": 875.848, "end": 877.468, "text": "放到一个统一的文件夹中", "chunk": 1, "language": "zh", "avg_logprob": -0.09943887263393858, "no_speech_prob": 1.9523983124658528e-11, "compression_ratio": 1.4585635359116023 }, { "start": 877.468, "end": 878.988, "text": "比如我这个目录下", "chunk": 1, "language": "zh", "avg_logprob": -0.09943887263393858, "no_speech_prob": 1.9523983124658528e-11, "compression_ratio": 1.4585635359116023 }, { "start": 878.988, "end": 881.248, "text": "对于我们Mac仿达中的用户文件夹", "chunk": 1, "language": "zh", "avg_logprob": -0.09943887263393858, "no_speech_prob": 1.9523983124658528e-11, "compression_ratio": 1.4585635359116023 }, { "start": 881.248, "end": 882.228, "text": "AI-TEST", "chunk": 1, "language": "zh", "avg_logprob": -0.09943887263393858, "no_speech_prob": 1.9523983124658528e-11, "compression_ratio": 1.4585635359116023 }, { "start": 882.228, "end": 884.628, "text": "这里手动新建一个TEST的文件夹就可以了", "chunk": 1, "language": "zh", "avg_logprob": -0.09943887263393858, "no_speech_prob": 1.9523983124658528e-11, "compression_ratio": 1.4585635359116023 }, { "start": 884.628, "end": 886.628, "text": "对于需要识别不同的内容图片", "chunk": 1, "language": "zh", "avg_logprob": -0.09943887263393858, "no_speech_prob": 1.9523983124658528e-11, "compression_ratio": 1.4585635359116023 }, { "start": 886.628, "end": 888.668, "text": "比如普通收据表格图表", "chunk": 1, "language": "zh", "avg_logprob": -0.09943887263393858, "no_speech_prob": 1.9523983124658528e-11, "compression_ratio": 1.4585635359116023 }, { "start": 888.668, "end": 890.088, "text": "都可以使用不同的命令", "chunk": 1, "language": "zh", "avg_logprob": -0.09943887263393858, "no_speech_prob": 1.9523983124658528e-11, "compression_ratio": 1.4585635359116023 }, { "start": 890.088, "end": 891.768, "text": "注意命令结尾的文件名", "chunk": 1, "language": "zh", "avg_logprob": -0.08620607111442602, "no_speech_prob": 1.6574526473522866e-11, "compression_ratio": 1.2514124293785311 }, { "start": 891.768, "end": 893.768, "text": "需要换成你自己图片的文件名", "chunk": 1, "language": "zh", "avg_logprob": -0.08620607111442602, "no_speech_prob": 1.6574526473522866e-11, "compression_ratio": 1.2514124293785311 }, { "start": 893.768, "end": 894.868, "text": "第一次测试时", "chunk": 1, "language": "zh", "avg_logprob": -0.08620607111442602, "no_speech_prob": 1.6574526473522866e-11, "compression_ratio": 1.2514124293785311 }, { "start": 894.868, "end": 896.568, "text": "建议每次只处理一张图片", "chunk": 1, "language": "zh", "avg_logprob": -0.08620607111442602, "no_speech_prob": 1.6574526473522866e-11, "compression_ratio": 1.2514124293785311 }, { "start": 896.568, "end": 898.928, "text": "先确认整套工作流能够稳定运行", "chunk": 1, "language": "zh", "avg_logprob": -0.08620607111442602, "no_speech_prob": 1.6574526473522866e-11, "compression_ratio": 1.2514124293785311 }, { "start": 898.928, "end": 900.948, "text": "如果遇到扫描PDF", "chunk": 1, "language": "zh", "avg_logprob": -0.08620607111442602, "no_speech_prob": 1.6574526473522866e-11, "compression_ratio": 1.2514124293785311 }, { "start": 900.948, "end": 903.888, "text": "也不用一开始就折腾完整的PDF解析工具", "chunk": 1, "language": "zh", "avg_logprob": -0.08620607111442602, "no_speech_prob": 1.6574526473522866e-11, "compression_ratio": 1.2514124293785311 }, { "start": 903.888, "end": 905.288, "text": "最简单办法就是", "chunk": 1, "language": "zh", "avg_logprob": -0.08620607111442602, "no_speech_prob": 1.6574526473522866e-11, "compression_ratio": 1.2514124293785311 }, { "start": 905.288, "end": 907.468, "text": "先用MicOS自带的预览打开PDF", "chunk": 1, "language": "zh", "avg_logprob": -0.08620607111442602, "no_speech_prob": 1.6574526473522866e-11, "compression_ratio": 1.2514124293785311 }, { "start": 907.468, "end": 909.188, "text": "把其中页导成PN器", "chunk": 1, "language": "zh", "avg_logprob": -0.08620607111442602, "no_speech_prob": 1.6574526473522866e-11, "compression_ratio": 1.2514124293785311 }, { "start": 909.188, "end": 911.488, "text": "再交给GLM OCR进行处理", "chunk": 1, "language": "zh", "avg_logprob": -0.08620607111442602, "no_speech_prob": 1.6574526473522866e-11, "compression_ratio": 1.2514124293785311 }, { "start": 911.488, "end": 913.668, "text": "等确认单页OCR没问题后", "chunk": 1, "language": "zh", "avg_logprob": -0.08620607111442602, "no_speech_prob": 1.6574526473522866e-11, "compression_ratio": 1.2514124293785311 }, { "start": 913.668, "end": 916.208, "text": "再考虑批量处理整份文档", "chunk": 1, "language": "zh", "avg_logprob": -0.08620607111442602, "no_speech_prob": 1.6574526473522866e-11, "compression_ratio": 1.2514124293785311 }, { "start": 916.208, "end": 917.988, "text": "识别出来内容还可以继续利用", "chunk": 1, "language": "zh", "avg_logprob": -0.08620607111442602, "no_speech_prob": 1.6574526473522866e-11, "compression_ratio": 1.2514124293785311 }, { "start": 917.988, "end": 919.968, "text": "保存成Tek-C或者Markdown", "chunk": 1, "language": "zh", "avg_logprob": -0.141378911336263, "no_speech_prob": 2.7425247836609934e-11, "compression_ratio": 1.1726027397260275 }, { "start": 919.968, "end": 922.568, "text": "再上传到我们上一节创建的知识库", "chunk": 1, "language": "zh", "avg_logprob": -0.141378911336263, "no_speech_prob": 2.7425247836609934e-11, "compression_ratio": 1.1726027397260275 }, { "start": 922.568, "end": 924.228, "text": "然后交给签问做摘要", "chunk": 1, "language": "zh", "avg_logprob": -0.141378911336263, "no_speech_prob": 2.7425247836609934e-11, "compression_ratio": 1.1726027397260275 }, { "start": 924.228, "end": 926.128, "text": "自端整理或者检索问答", "chunk": 1, "language": "zh", "avg_logprob": -0.141378911336263, "no_speech_prob": 2.7425247836609934e-11, "compression_ratio": 1.1726027397260275 }, { "start": 926.128, "end": 928.048, "text": "那么整个流程就传起来了", "chunk": 1, "language": "zh", "avg_logprob": -0.141378911336263, "no_speech_prob": 2.7425247836609934e-11, "compression_ratio": 1.1726027397260275 }, { "start": 928.048, "end": 930.988, "text": "如果OCR出现空白重复或者乱码", "chunk": 1, "language": "zh", "avg_logprob": -0.141378911336263, "no_speech_prob": 2.7425247836609934e-11, "compression_ratio": 1.1726027397260275 }, { "start": 930.988, "end": 932.288, "text": "可以先更新OLAMA", "chunk": 1, "language": "zh", "avg_logprob": -0.141378911336263, "no_speech_prob": 2.7425247836609934e-11, "compression_ratio": 1.1726027397260275 }, { "start": 932.288, "end": 934.268, "text": "然后每次只处理一张图片", "chunk": 1, "language": "zh", "avg_logprob": -0.141378911336263, "no_speech_prob": 2.7425247836609934e-11, "compression_ratio": 1.1726027397260275 }, { "start": 934.268, "end": 935.428, "text": "图片过大时", "chunk": 1, "language": "zh", "avg_logprob": -0.141378911336263, "no_speech_prob": 2.7425247836609934e-11, "compression_ratio": 1.1726027397260275 }, { "start": 935.428, "end": 939.048, "text": "可以把长辨锁到大约1600至2200像素", "chunk": 1, "language": "zh", "avg_logprob": -0.141378911336263, "no_speech_prob": 2.7425247836609934e-11, "compression_ratio": 1.1726027397260275 }, { "start": 939.048, "end": 941.148, "text": "并优先使用PNG或者GPG", "chunk": 1, "language": "zh", "avg_logprob": -0.141378911336263, "no_speech_prob": 2.7425247836609934e-11, "compression_ratio": 1.1726027397260275 }, { "start": 941.148, "end": 943.968, "text": "同时检查原图有没有明显模糊", "chunk": 1, "language": "zh", "avg_logprob": -0.141378911336263, "no_speech_prob": 2.7425247836609934e-11, "compression_ratio": 1.1726027397260275 }, { "start": 943.968, "end": 945.068, "text": "倾斜反光", "chunk": 1, "language": "zh", "avg_logprob": -0.141378911336263, "no_speech_prob": 2.7425247836609934e-11, "compression_ratio": 1.1726027397260275 }, { "start": 945.068, "end": 947.048, "text": "这些都会直接影响识别效果", "chunk": 1, "language": "zh", "avg_logprob": -0.141378911336263, "no_speech_prob": 2.7425247836609934e-11, "compression_ratio": 1.1726027397260275 }, { "start": 947.048, "end": 949.048, "text": "好", "chunk": 1, "language": "zh", "avg_logprob": -1.2935338020324707, "no_speech_prob": 1.602794286181819e-10, "compression_ratio": 0.2727272727272727 }, { "start": 952.515, "end": 964.255, "text": "那除了文字和图片之外,本DI还有一个非常实用的场景,就是处理视频和录音。比如把会议录音转成文字,把视频自动生成字幕,或者把自己的素材库快速整理成可搜索的文本。", "chunk": 2, "language": "zh", "avg_logprob": -0.1366638992771958, "no_speech_prob": 8.964126316235443e-12, "compression_ratio": 1.2261904761904763 }, { "start": 964.255, "end": 966.175, "text": "这里我们使用Whisper.cpp", "chunk": 2, "language": "zh", "avg_logprob": -0.08014818701413598, "no_speech_prob": 1.929389634169887e-11, "compression_ratio": 1.2780898876404494 }, { "start": 966.175, "end": 968.615, "text": "它是Whisper模型C++实现的", "chunk": 2, "language": "zh", "avg_logprob": -0.08014818701413598, "no_speech_prob": 1.929389634169887e-11, "compression_ratio": 1.2780898876404494 }, { "start": 968.615, "end": 970.615, "text": "针对Apple Silicon提供Mental加速", "chunk": 2, "language": "zh", "avg_logprob": -0.08014818701413598, "no_speech_prob": 1.929389634169887e-11, "compression_ratio": 1.2780898876404494 }, { "start": 970.615, "end": 973.355, "text": "可以直接利用Mac的GPU进行本地转写", "chunk": 2, "language": "zh", "avg_logprob": -0.08014818701413598, "no_speech_prob": 1.929389634169887e-11, "compression_ratio": 1.2780898876404494 }, { "start": 973.355, "end": 975.235, "text": "首先安装需要用到的工具", "chunk": 2, "language": "zh", "avg_logprob": -0.08014818701413598, "no_speech_prob": 1.929389634169887e-11, "compression_ratio": 1.2780898876404494 }, { "start": 975.235, "end": 977.835, "text": "FMMP1G负责处理音频格式", "chunk": 2, "language": "zh", "avg_logprob": -0.08014818701413598, "no_speech_prob": 1.929389634169887e-11, "compression_ratio": 1.2780898876404494 }, { "start": 977.835, "end": 980.155, "text": "Whisper.cpp负责语音识别", "chunk": 2, "language": "zh", "avg_logprob": -0.08014818701413598, "no_speech_prob": 1.929389634169887e-11, "compression_ratio": 1.2780898876404494 }, { "start": 980.155, "end": 982.095, "text": "如果你的Mac已经安装HomeBrew", "chunk": 2, "language": "zh", "avg_logprob": -0.08014818701413598, "no_speech_prob": 1.929389634169887e-11, "compression_ratio": 1.2780898876404494 }, { "start": 982.095, "end": 983.935, "text": "可以直接输入代码来进行安装", "chunk": 2, "language": "zh", "avg_logprob": -0.08014818701413598, "no_speech_prob": 1.929389634169887e-11, "compression_ratio": 1.2780898876404494 }, { "start": 983.935, "end": 985.675, "text": "如果你仍然没有安装HomeBrew", "chunk": 2, "language": "zh", "avg_logprob": -0.08014818701413598, "no_speech_prob": 1.929389634169887e-11, "compression_ratio": 1.2780898876404494 }, { "start": 985.675, "end": 988.295, "text": "安装Whisper.cpp可以走原码编译路线", "chunk": 2, "language": "zh", "avg_logprob": -0.08014818701413598, "no_speech_prob": 1.929389634169887e-11, "compression_ratio": 1.2780898876404494 }, { "start": 988.295, "end": 989.795, "text": "不过对于普通用户来说", "chunk": 2, "language": "zh", "avg_logprob": -0.08014818701413598, "no_speech_prob": 1.929389634169887e-11, "compression_ratio": 1.2780898876404494 }, { "start": 989.795, "end": 991.795, "text": "这一步会比HomeBrew更复杂一些", "chunk": 2, "language": "zh", "avg_logprob": -0.08014818701413598, "no_speech_prob": 1.929389634169887e-11, "compression_ratio": 1.2780898876404494 }, { "start": 991.795, "end": 994.355, "text": "还需要提前安装Xcode command line tools", "chunk": 2, "language": "zh", "avg_logprob": -0.10192013703859769, "no_speech_prob": 2.2023164542028795e-11, "compression_ratio": 1.2613065326633166 }, { "start": 994.355, "end": 995.695, "text": "这里就不展开了", "chunk": 2, "language": "zh", "avg_logprob": -0.10192013703859769, "no_speech_prob": 2.2023164542028795e-11, "compression_ratio": 1.2613065326633166 }, { "start": 995.695, "end": 996.615, "text": "安装完整后", "chunk": 2, "language": "zh", "avg_logprob": -0.10192013703859769, "no_speech_prob": 2.2023164542028795e-11, "compression_ratio": 1.2613065326633166 }, { "start": 996.615, "end": 999.135, "text": "可以通过下面两个命令来确认是否正常", "chunk": 2, "language": "zh", "avg_logprob": -0.10192013703859769, "no_speech_prob": 2.2023164542028795e-11, "compression_ratio": 1.2613065326633166 }, { "start": 999.135, "end": 1000.735, "text": "接下来下载Whisper模型", "chunk": 2, "language": "zh", "avg_logprob": -0.10192013703859769, "no_speech_prob": 2.2023164542028795e-11, "compression_ratio": 1.2613065326633166 }, { "start": 1000.735, "end": 1002.195, "text": "我这里选择的是这个", "chunk": 2, "language": "zh", "avg_logprob": -0.10192013703859769, "no_speech_prob": 2.2023164542028795e-11, "compression_ratio": 1.2613065326633166 }, { "start": 1002.195, "end": 1004.555, "text": "这是一个支持中文和多语言的模型", "chunk": 2, "language": "zh", "avg_logprob": -0.10192013703859769, "no_speech_prob": 2.2023164542028795e-11, "compression_ratio": 1.2613065326633166 }, { "start": 1004.555, "end": 1006.815, "text": "相比完整LARGE模型体积更小", "chunk": 2, "language": "zh", "avg_logprob": -0.10192013703859769, "no_speech_prob": 2.2023164542028795e-11, "compression_ratio": 1.2613065326633166 }, { "start": 1006.815, "end": 1008.715, "text": "同时保留了比较好的转眼效果", "chunk": 2, "language": "zh", "avg_logprob": -0.10192013703859769, "no_speech_prob": 2.2023164542028795e-11, "compression_ratio": 1.2613065326633166 }, { "start": 1008.715, "end": 1010.155, "text": "更适合Mac本地运行", "chunk": 2, "language": "zh", "avg_logprob": -0.10192013703859769, "no_speech_prob": 2.2023164542028795e-11, "compression_ratio": 1.2613065326633166 }, { "start": 1010.155, "end": 1013.155, "text": "我们先创建存放模型和临时文件的目录", "chunk": 2, "language": "zh", "avg_logprob": -0.10192013703859769, "no_speech_prob": 2.2023164542028795e-11, "compression_ratio": 1.2613065326633166 }, { "start": 1013.155, "end": 1014.195, "text": "然后下载模型", "chunk": 2, "language": "zh", "avg_logprob": -0.10192013703859769, "no_speech_prob": 2.2023164542028795e-11, "compression_ratio": 1.2613065326633166 }, { "start": 1014.195, "end": 1017.235, "text": "下载完整后再计算一次SHA-1的值", "chunk": 2, "language": "zh", "avg_logprob": -0.10192013703859769, "no_speech_prob": 2.2023164542028795e-11, "compression_ratio": 1.2613065326633166 }, { "start": 1017.235, "end": 1021.135, "text": "方便和Whisper.cpp模型文档公开的接应值进行核对", "chunk": 2, "language": "zh", "avg_logprob": -0.10192013703859769, "no_speech_prob": 2.2023164542028795e-11, "compression_ratio": 1.2613065326633166 }, { "start": 1021.135, "end": 1023.535, "text": "模型准备好后就可以开始处理视频", "chunk": 2, "language": "zh", "avg_logprob": -0.10772609236228525, "no_speech_prob": 1.9222822986164623e-11, "compression_ratio": 1.2932960893854748 }, { "start": 1023.535, "end": 1026.255, "text": "假设你的视频文件位于这个文件夹", "chunk": 2, "language": "zh", "avg_logprob": -0.10772609236228525, "no_speech_prob": 1.9222822986164623e-11, "compression_ratio": 1.2932960893854748 }, { "start": 1026.255, "end": 1029.335, "text": "也就是MicroOS用户目录的影片文件夹当中", "chunk": 2, "language": "zh", "avg_logprob": -0.10772609236228525, "no_speech_prob": 1.9222822986164623e-11, "compression_ratio": 1.2932960893854748 }, { "start": 1029.335, "end": 1032.075, "text": "大家根据自己的实际文件存储位置进行修改", "chunk": 2, "language": "zh", "avg_logprob": -0.10772609236228525, "no_speech_prob": 1.9222822986164623e-11, "compression_ratio": 1.2932960893854748 }, { "start": 1032.075, "end": 1037.055, "text": "首先使用FMMP1G把视频里的音频转换成Whisper更适合处理的格式", "chunk": 2, "language": "zh", "avg_logprob": -0.10772609236228525, "no_speech_prob": 1.9222822986164623e-11, "compression_ratio": 1.2932960893854748 }, { "start": 1037.055, "end": 1038.895, "text": "这些几个参数简单介绍一下", "chunk": 2, "language": "zh", "avg_logprob": -0.10772609236228525, "no_speech_prob": 1.9222822986164623e-11, "compression_ratio": 1.2932960893854748 }, { "start": 1038.895, "end": 1041.255, "text": "VN表示去掉视频只保留音频", "chunk": 2, "language": "zh", "avg_logprob": -0.10772609236228525, "no_speech_prob": 1.9222822986164623e-11, "compression_ratio": 1.2932960893854748 }, { "start": 1041.255, "end": 1044.635, "text": "AR16表示转换成16000Hz的采样率", "chunk": 2, "language": "zh", "avg_logprob": -0.10772609236228525, "no_speech_prob": 1.9222822986164623e-11, "compression_ratio": 1.2932960893854748 }, { "start": 1044.635, "end": 1046.315, "text": "AC1表示单身道", "chunk": 2, "language": "zh", "avg_logprob": -0.10772609236228525, "no_speech_prob": 1.9222822986164623e-11, "compression_ratio": 1.2932960893854748 }, { "start": 1046.315, "end": 1051.075, "text": "PCM-S16LE则是转换成Whisper更容易处理的WAV格式", "chunk": 2, "language": "zh", "avg_logprob": -0.10772609236228525, "no_speech_prob": 1.9222822986164623e-11, "compression_ratio": 1.2932960893854748 }, { "start": 1051.075, "end": 1054.355, "text": "接下来运行Whisper同时生成文字稿和字幕文件", "chunk": 2, "language": "zh", "avg_logprob": -0.12828174612140128, "no_speech_prob": 2.3549591096627687e-11, "compression_ratio": 1.2301369863013698 }, { "start": 1054.355, "end": 1057.255, "text": "-m用来指定我们前面下载的Whisper模型", "chunk": 2, "language": "zh", "avg_logprob": -0.12828174612140128, "no_speech_prob": 2.3549591096627687e-11, "compression_ratio": 1.2301369863013698 }, { "start": 1057.255, "end": 1059.915, "text": "-f指定需要转写的音频文件", "chunk": 2, "language": "zh", "avg_logprob": -0.12828174612140128, "no_speech_prob": 2.3549591096627687e-11, "compression_ratio": 1.2301369863013698 }, { "start": 1059.915, "end": 1063.495, "text": "-lzh告诉我们模型这段文字主要内容是中文", "chunk": 2, "language": "zh", "avg_logprob": -0.12828174612140128, "no_speech_prob": 2.3549591096627687e-11, "compression_ratio": 1.2301369863013698 }, { "start": 1063.495, "end": 1067.355, "text": "-prompt则可以提前提供一些容易识别错误的中文名词", "chunk": 2, "language": "zh", "avg_logprob": -0.12828174612140128, "no_speech_prob": 2.3549591096627687e-11, "compression_ratio": 1.2301369863013698 }, { "start": 1067.355, "end": 1070.395, "text": "比如Mac mini,Apple Silicon,Olama和签问", "chunk": 2, "language": "zh", "avg_logprob": -0.12828174612140128, "no_speech_prob": 2.3549591096627687e-11, "compression_ratio": 1.2301369863013698 }, { "start": 1070.395, "end": 1071.935, "text": "对于科技类视频有其有用", "chunk": 2, "language": "zh", "avg_logprob": -0.12828174612140128, "no_speech_prob": 2.3549591096627687e-11, "compression_ratio": 1.2301369863013698 }, { "start": 1071.935, "end": 1074.175, "text": "比如这里我就加上了我的Lookup Weekly", "chunk": 2, "language": "zh", "avg_logprob": -0.12828174612140128, "no_speech_prob": 2.3549591096627687e-11, "compression_ratio": 1.2301369863013698 }, { "start": 1074.175, "end": 1080.955, "text": "后面的-otxt-oosrt和-ovtt分别让它同时生成TXT文字稿", "chunk": 2, "language": "zh", "avg_logprob": -0.12828174612140128, "no_speech_prob": 2.3549591096627687e-11, "compression_ratio": 1.2301369863013698 }, { "start": 1080.955, "end": 1083.295, "text": "SRT字幕和VTT字幕", "chunk": 2, "language": "zh", "avg_logprob": -0.08696064304661107, "no_speech_prob": 1.8131529264109325e-11, "compression_ratio": 1.1842105263157894 }, { "start": 1083.295, "end": 1086.775, "text": "最后通过-off指定输出文件的位置和名称", "chunk": 2, "language": "zh", "avg_logprob": -0.08696064304661107, "no_speech_prob": 1.8131529264109325e-11, "compression_ratio": 1.1842105263157894 }, { "start": 1086.775, "end": 1089.035, "text": "运行完成后会生成这些文件", "chunk": 2, "language": "zh", "avg_logprob": -0.08696064304661107, "no_speech_prob": 1.8131529264109325e-11, "compression_ratio": 1.1842105263157894 }, { "start": 1089.035, "end": 1090.895, "text": "其中TXT是纯文字稿", "chunk": 2, "language": "zh", "avg_logprob": -0.08696064304661107, "no_speech_prob": 1.8131529264109325e-11, "compression_ratio": 1.1842105263157894 }, { "start": 1090.895, "end": 1093.435, "text": "适合后续交给AI总结或者整理", "chunk": 2, "language": "zh", "avg_logprob": -0.08696064304661107, "no_speech_prob": 1.8131529264109325e-11, "compression_ratio": 1.1842105263157894 }, { "start": 1093.435, "end": 1095.335, "text": "SRT是常见的字幕格式", "chunk": 2, "language": "zh", "avg_logprob": -0.08696064304661107, "no_speech_prob": 1.8131529264109325e-11, "compression_ratio": 1.1842105263157894 }, { "start": 1095.335, "end": 1096.815, "text": "可以直接导入紧接软件", "chunk": 2, "language": "zh", "avg_logprob": -0.08696064304661107, "no_speech_prob": 1.8131529264109325e-11, "compression_ratio": 1.1842105263157894 }, { "start": 1096.815, "end": 1099.735, "text": "VTT是主要用于网页播放器的字幕", "chunk": 2, "language": "zh", "avg_logprob": -0.08696064304661107, "no_speech_prob": 1.8131529264109325e-11, "compression_ratio": 1.1842105263157894 }, { "start": 1099.735, "end": 1101.035, "text": "如果是8GB的Mac", "chunk": 2, "language": "zh", "avg_logprob": -0.08696064304661107, "no_speech_prob": 1.8131529264109325e-11, "compression_ratio": 1.1842105263157894 }, { "start": 1101.035, "end": 1104.795, "text": "也可以先尝试这里的Large-V3-Turbo Q50", "chunk": 2, "language": "zh", "avg_logprob": -0.08696064304661107, "no_speech_prob": 1.8131529264109325e-11, "compression_ratio": 1.1842105263157894 }, { "start": 1104.795, "end": 1105.735, "text": "电话模型", "chunk": 2, "language": "zh", "avg_logprob": -0.08696064304661107, "no_speech_prob": 1.8131529264109325e-11, "compression_ratio": 1.1842105263157894 }, { "start": 1105.735, "end": 1108.675, "text": "如果实际转写时内存压力较大或者速度不理想", "chunk": 2, "language": "zh", "avg_logprob": -0.08696064304661107, "no_speech_prob": 1.8131529264109325e-11, "compression_ratio": 1.1842105263157894 }, { "start": 1108.675, "end": 1110.955, "text": "再换成体积更小的small多语言模型", "chunk": 2, "language": "zh", "avg_logprob": -0.096618633174417, "no_speech_prob": 1.968196265689226e-11, "compression_ratio": 1.2946175637393769 }, { "start": 1110.955, "end": 1113.435, "text": "不要下载来带EAN后制的英文专用模型", "chunk": 2, "language": "zh", "avg_logprob": -0.096618633174417, "no_speech_prob": 1.968196265689226e-11, "compression_ratio": 1.2946175637393769 }, { "start": 1113.435, "end": 1115.095, "text": "因为它只针对英文优化", "chunk": 2, "language": "zh", "avg_logprob": -0.096618633174417, "no_speech_prob": 1.968196265689226e-11, "compression_ratio": 1.2946175637393769 }, { "start": 1115.095, "end": 1116.415, "text": "如果需要中文转写", "chunk": 2, "language": "zh", "avg_logprob": -0.096618633174417, "no_speech_prob": 1.968196265689226e-11, "compression_ratio": 1.2946175637393769 }, { "start": 1116.415, "end": 1117.695, "text": "也不要开启翻译参数", "chunk": 2, "language": "zh", "avg_logprob": -0.096618633174417, "no_speech_prob": 1.968196265689226e-11, "compression_ratio": 1.2946175637393769 }, { "start": 1117.695, "end": 1119.715, "text": "否则模型会把内容翻译成英文", "chunk": 2, "language": "zh", "avg_logprob": -0.096618633174417, "no_speech_prob": 1.968196265689226e-11, "compression_ratio": 1.2946175637393769 }, { "start": 1119.715, "end": 1121.315, "text": "而不是保留原始的语言", "chunk": 2, "language": "zh", "avg_logprob": -0.096618633174417, "no_speech_prob": 1.968196265689226e-11, "compression_ratio": 1.2946175637393769 }, { "start": 1121.315, "end": 1122.255, "text": "那到这里呢", "chunk": 2, "language": "zh", "avg_logprob": -0.096618633174417, "no_speech_prob": 1.968196265689226e-11, "compression_ratio": 1.2946175637393769 }, { "start": 1122.255, "end": 1124.415, "text": "整套本地i工作流基本已经闭环", "chunk": 2, "language": "zh", "avg_logprob": -0.096618633174417, "no_speech_prob": 1.968196265689226e-11, "compression_ratio": 1.2946175637393769 }, { "start": 1124.415, "end": 1126.015, "text": "视频和录音可以通过whisper", "chunk": 2, "language": "zh", "avg_logprob": -0.096618633174417, "no_speech_prob": 1.968196265689226e-11, "compression_ratio": 1.2946175637393769 }, { "start": 1126.015, "end": 1126.755, "text": "转写文字", "chunk": 2, "language": "zh", "avg_logprob": -0.096618633174417, "no_speech_prob": 1.968196265689226e-11, "compression_ratio": 1.2946175637393769 }, { "start": 1126.755, "end": 1128.215, "text": "再交给前文做总结", "chunk": 2, "language": "zh", "avg_logprob": -0.096618633174417, "no_speech_prob": 1.968196265689226e-11, "compression_ratio": 1.2946175637393769 }, { "start": 1128.215, "end": 1128.655, "text": "整理", "chunk": 2, "language": "zh", "avg_logprob": -0.096618633174417, "no_speech_prob": 1.968196265689226e-11, "compression_ratio": 1.2946175637393769 }, { "start": 1128.655, "end": 1130.855, "text": "最后也可以存入前面搭建的知识库", "chunk": 2, "language": "zh", "avg_logprob": -0.096618633174417, "no_speech_prob": 1.968196265689226e-11, "compression_ratio": 1.2946175637393769 }, { "start": 1130.855, "end": 1133.155, "text": "实现自己本地的内容管理系统", "chunk": 2, "language": "zh", "avg_logprob": -0.096618633174417, "no_speech_prob": 1.968196265689226e-11, "compression_ratio": 1.2946175637393769 }, { "start": 1133.155, "end": 1140.935, "text": "前面我们已经分别完成了本地云转写", "chunk": 2, "language": "zh", "avg_logprob": -0.10297163827078683, "no_speech_prob": 1.7635823357231573e-11, "compression_ratio": 1.1993769470404985 }, { "start": 1140.935, "end": 1143.275, "text": "OpenWebUI API和签问调用", "chunk": 2, "language": "zh", "avg_logprob": -0.10297163827078683, "no_speech_prob": 1.7635823357231573e-11, "compression_ratio": 1.1993769470404985 }, { "start": 1143.275, "end": 1145.855, "text": "这就是使用N8N搭建一条自动化工作流", "chunk": 2, "language": "zh", "avg_logprob": -0.10297163827078683, "no_speech_prob": 1.7635823357231573e-11, "compression_ratio": 1.1993769470404985 }, { "start": 1145.855, "end": 1147.515, "text": "上传音频或者视频后", "chunk": 2, "language": "zh", "avg_logprob": -0.10297163827078683, "no_speech_prob": 1.7635823357231573e-11, "compression_ratio": 1.1993769470404985 }, { "start": 1147.515, "end": 1149.415, "text": "N8N自动调用Whisper来转写", "chunk": 2, "language": "zh", "avg_logprob": -0.10297163827078683, "no_speech_prob": 1.7635823357231573e-11, "compression_ratio": 1.1993769470404985 }, { "start": 1149.415, "end": 1151.755, "text": "再把转写稿交给签问来整理", "chunk": 2, "language": "zh", "avg_logprob": -0.10297163827078683, "no_speech_prob": 1.7635823357231573e-11, "compression_ratio": 1.1993769470404985 }, { "start": 1151.755, "end": 1155.075, "text": "最后返回摘要提纲标题和症结建议", "chunk": 2, "language": "zh", "avg_logprob": -0.10297163827078683, "no_speech_prob": 1.7635823357231573e-11, "compression_ratio": 1.1993769470404985 }, { "start": 1155.075, "end": 1156.835, "text": "不需要每次手动复制粘贴", "chunk": 2, "language": "zh", "avg_logprob": -0.10297163827078683, "no_speech_prob": 1.7635823357231573e-11, "compression_ratio": 1.1993769470404985 }, { "start": 1156.835, "end": 1159.115, "text": "首先把前面安装好的Whisper.cpp", "chunk": 2, "language": "zh", "avg_logprob": -0.10297163827078683, "no_speech_prob": 1.7635823357231573e-11, "compression_ratio": 1.1993769470404985 }, { "start": 1159.115, "end": 1161.735, "text": "变成一个可以被其他程序调用的本地API", "chunk": 2, "language": "zh", "avg_logprob": -0.10297163827078683, "no_speech_prob": 1.7635823357231573e-11, "compression_ratio": 1.1993769470404985 }, { "start": 1161.735, "end": 1164.215, "text": "先尝试允许Mac本机访问", "chunk": 2, "language": "zh", "avg_logprob": -0.07835123324281232, "no_speech_prob": 2.353563698098693e-11, "compression_ratio": 1.2193211488250653 }, { "start": 1164.215, "end": 1165.815, "text": "保持这个中端窗口开启", "chunk": 2, "language": "zh", "avg_logprob": -0.07835123324281232, "no_speech_prob": 2.353563698098693e-11, "compression_ratio": 1.2193211488250653 }, { "start": 1165.815, "end": 1168.135, "text": "再新开一个中端APP窗口进行测试", "chunk": 2, "language": "zh", "avg_logprob": -0.07835123324281232, "no_speech_prob": 2.353563698098693e-11, "compression_ratio": 1.2193211488250653 }, { "start": 1168.135, "end": 1170.195, "text": "如果能够正常返回转写结果", "chunk": 2, "language": "zh", "avg_logprob": -0.07835123324281232, "no_speech_prob": 2.353563698098693e-11, "compression_ratio": 1.2193211488250653 }, { "start": 1170.195, "end": 1172.055, "text": "Whisper API就已经运行成功", "chunk": 2, "language": "zh", "avg_logprob": -0.07835123324281232, "no_speech_prob": 2.353563698098693e-11, "compression_ratio": 1.2193211488250653 }, { "start": 1172.055, "end": 1174.675, "text": "后面N8N会运行在Darker容器里", "chunk": 2, "language": "zh", "avg_logprob": -0.07835123324281232, "no_speech_prob": 2.353563698098693e-11, "compression_ratio": 1.2193211488250653 }, { "start": 1174.675, "end": 1176.675, "text": "如果出现无法连接Whisper的情况", "chunk": 2, "language": "zh", "avg_logprob": -0.07835123324281232, "no_speech_prob": 2.353563698098693e-11, "compression_ratio": 1.2193211488250653 }, { "start": 1176.675, "end": 1178.475, "text": "可以按Ctrl加C停止服务", "chunk": 2, "language": "zh", "avg_logprob": -0.07835123324281232, "no_speech_prob": 2.353563698098693e-11, "compression_ratio": 1.2193211488250653 }, { "start": 1178.475, "end": 1179.395, "text": "再这样改", "chunk": 2, "language": "zh", "avg_logprob": -0.07835123324281232, "no_speech_prob": 2.353563698098693e-11, "compression_ratio": 1.2193211488250653 }, { "start": 1179.395, "end": 1180.515, "text": "这里需要特别注意", "chunk": 2, "language": "zh", "avg_logprob": -0.07835123324281232, "no_speech_prob": 2.353563698098693e-11, "compression_ratio": 1.2193211488250653 }, { "start": 1180.515, "end": 1183.955, "text": "Whisper server本身没有用户登陆和全线验证的功能", "chunk": 2, "language": "zh", "avg_logprob": -0.07835123324281232, "no_speech_prob": 2.353563698098693e-11, "compression_ratio": 1.2193211488250653 }, { "start": 1183.955, "end": 1186.035, "text": "所以使用0.0.0.0时", "chunk": 2, "language": "zh", "avg_logprob": -0.07835123324281232, "no_speech_prob": 2.353563698098693e-11, "compression_ratio": 1.2193211488250653 }, { "start": 1186.035, "end": 1188.375, "text": "只建议放在可信的家庭局网内", "chunk": 2, "language": "zh", "avg_logprob": -0.07835123324281232, "no_speech_prob": 2.353563698098693e-11, "compression_ratio": 1.2193211488250653 }, { "start": 1188.375, "end": 1189.955, "text": "不要进行路由器端口转发", "chunk": 2, "language": "zh", "avg_logprob": -0.07835123324281232, "no_speech_prob": 2.353563698098693e-11, "compression_ratio": 1.2193211488250653 }, { "start": 1189.955, "end": 1192.555, "text": "也不要把8178断口直接暴露到公网", "chunk": 2, "language": "zh", "avg_logprob": -0.10812189362265846, "no_speech_prob": 1.235354960688051e-11, "compression_ratio": 1.2163934426229508 }, { "start": 1192.555, "end": 1193.875, "text": "接下来安装N8N", "chunk": 2, "language": "zh", "avg_logprob": -0.10812189362265846, "no_speech_prob": 1.235354960688051e-11, "compression_ratio": 1.2163934426229508 }, { "start": 1193.875, "end": 1196.015, "text": "这里使用Docker Desktop来运行", "chunk": 2, "language": "zh", "avg_logprob": -0.10812189362265846, "no_speech_prob": 1.235354960688051e-11, "compression_ratio": 1.2163934426229508 }, { "start": 1196.015, "end": 1197.455, "text": "并固定N8N版本", "chunk": 2, "language": "zh", "avg_logprob": -0.10812189362265846, "no_speech_prob": 1.235354960688051e-11, "compression_ratio": 1.2163934426229508 }, { "start": 1197.455, "end": 1201.155, "text": "避免后续更新导致教程界面或者节点配置发生变化", "chunk": 2, "language": "zh", "avg_logprob": -0.10812189362265846, "no_speech_prob": 1.235354960688051e-11, "compression_ratio": 1.2163934426229508 }, { "start": 1201.155, "end": 1203.855, "text": "先从Docker官网下载安装Docker Desktop", "chunk": 2, "language": "zh", "avg_logprob": -0.10812189362265846, "no_speech_prob": 1.235354960688051e-11, "compression_ratio": 1.2163934426229508 }, { "start": 1203.855, "end": 1205.995, "text": "拖动到应用程序文件夹并启动", "chunk": 2, "language": "zh", "avg_logprob": -0.10812189362265846, "no_speech_prob": 1.235354960688051e-11, "compression_ratio": 1.2163934426229508 }, { "start": 1205.995, "end": 1207.355, "text": "确认Docker正常运行后", "chunk": 2, "language": "zh", "avg_logprob": -0.10812189362265846, "no_speech_prob": 1.235354960688051e-11, "compression_ratio": 1.2163934426229508 }, { "start": 1207.355, "end": 1209.535, "text": "再打开中端App输入下面的命令", "chunk": 2, "language": "zh", "avg_logprob": -0.10812189362265846, "no_speech_prob": 1.235354960688051e-11, "compression_ratio": 1.2163934426229508 }, { "start": 1209.535, "end": 1219.175, "text": "接下来生成一段随机的加密密钥", "chunk": 2, "language": "zh", "avg_logprob": -0.10812189362265846, "no_speech_prob": 1.235354960688051e-11, "compression_ratio": 1.2163934426229508 }, { "start": 1219.175, "end": 1225.175, "text": "把生成的结果替换到下面的XXX", "chunk": 2, "language": "zh", "avg_logprob": -0.1402413677161848, "no_speech_prob": 9.585757534957828e-12, "compression_ratio": 1.1227436823104693 }, { "start": 1225.175, "end": 1231.835, "text": "然后我们访问127.0.0.1", "chunk": 2, "language": "zh", "avg_logprob": -0.1402413677161848, "no_speech_prob": 9.585757534957828e-12, "compression_ratio": 1.1227436823104693 }, { "start": 1231.835, "end": 1232.895, "text": "冒号5678", "chunk": 2, "language": "zh", "avg_logprob": -0.1402413677161848, "no_speech_prob": 9.585757534957828e-12, "compression_ratio": 1.1227436823104693 }, { "start": 1232.895, "end": 1236.115, "text": "第一次进入需要创建N8N的本地管理员账号", "chunk": 2, "language": "zh", "avg_logprob": -0.1402413677161848, "no_speech_prob": 9.585757534957828e-12, "compression_ratio": 1.1227436823104693 }, { "start": 1236.115, "end": 1238.555, "text": "接下来我们来新建一条workflow工作流", "chunk": 2, "language": "zh", "avg_logprob": -0.1402413677161848, "no_speech_prob": 9.585757534957828e-12, "compression_ratio": 1.1227436823104693 }, { "start": 1238.555, "end": 1240.475, "text": "这套流程一共只有四个节点", "chunk": 2, "language": "zh", "avg_logprob": -0.1402413677161848, "no_speech_prob": 9.585757534957828e-12, "compression_ratio": 1.1227436823104693 }, { "start": 1240.475, "end": 1241.315, "text": "Webhook", "chunk": 2, "language": "zh", "avg_logprob": -0.1402413677161848, "no_speech_prob": 9.585757534957828e-12, "compression_ratio": 1.1227436823104693 }, { "start": 1241.315, "end": 1243.575, "text": "httprequest whisper转写", "chunk": 2, "language": "zh", "avg_logprob": -0.1402413677161848, "no_speech_prob": 9.585757534957828e-12, "compression_ratio": 1.1227436823104693 }, { "start": 1243.575, "end": 1245.675, "text": "httprequest签文整理", "chunk": 2, "language": "zh", "avg_logprob": -0.1402413677161848, "no_speech_prob": 9.585757534957828e-12, "compression_ratio": 1.1227436823104693 }, { "start": 1245.675, "end": 1247.215, "text": "respond to webhook", "chunk": 2, "language": "zh", "avg_logprob": -0.1402413677161848, "no_speech_prob": 9.585757534957828e-12, "compression_ratio": 1.1227436823104693 }, { "start": 1247.215, "end": 1248.875, "text": "第一个是webhook节点", "chunk": 2, "language": "zh", "avg_logprob": -0.1402413677161848, "no_speech_prob": 9.585757534957828e-12, "compression_ratio": 1.1227436823104693 }, { "start": 1248.875, "end": 1250.595, "text": "它相当于整条工作流的入口", "chunk": 2, "language": "zh", "avg_logprob": -0.11202583162803349, "no_speech_prob": 1.0838262579082603e-11, "compression_ratio": 1.150943396226415 }, { "start": 1250.595, "end": 1252.895, "text": "以后只要把视频post到这个地址", "chunk": 2, "language": "zh", "avg_logprob": -0.11202583162803349, "no_speech_prob": 1.0838262579082603e-11, "compression_ratio": 1.150943396226415 }, { "start": 1252.895, "end": 1254.355, "text": "工作流就会自动开始", "chunk": 2, "language": "zh", "avg_logprob": -0.11202583162803349, "no_speech_prob": 1.0838262579082603e-11, "compression_ratio": 1.150943396226415 }, { "start": 1254.355, "end": 1256.055, "text": "外部hook的具体设置如下", "chunk": 2, "language": "zh", "avg_logprob": -0.11202583162803349, "no_speech_prob": 1.0838262579082603e-11, "compression_ratio": 1.150943396226415 }, { "start": 1256.055, "end": 1258.055, "text": "没提到的设置保持不变即可", "chunk": 2, "language": "zh", "avg_logprob": -0.11202583162803349, "no_speech_prob": 1.0838262579082603e-11, "compression_ratio": 1.150943396226415 }, { "start": 1258.055, "end": 1270.295, "text": "第二个节点添加http request", "chunk": 2, "language": "zh", "avg_logprob": -0.11202583162803349, "no_speech_prob": 1.0838262579082603e-11, "compression_ratio": 1.150943396226415 }, { "start": 1270.295, "end": 1272.415, "text": "名称可以改成Whisper转写", "chunk": 2, "language": "zh", "avg_logprob": -0.11202583162803349, "no_speech_prob": 1.0838262579082603e-11, "compression_ratio": 1.150943396226415 }, { "start": 1272.415, "end": 1274.735, "text": "它负责把刚刚上传的音视频文件", "chunk": 2, "language": "zh", "avg_logprob": -0.11202583162803349, "no_speech_prob": 1.0838262579082603e-11, "compression_ratio": 1.150943396226415 }, { "start": 1274.735, "end": 1276.395, "text": "交给Mac上的Whisper server", "chunk": 2, "language": "zh", "avg_logprob": -0.11202583162803349, "no_speech_prob": 1.0838262579082603e-11, "compression_ratio": 1.150943396226415 }, { "start": 1276.395, "end": 1279.115, "text": "第二个节点HTTPRequest的具体设置如下", "chunk": 2, "language": "zh", "avg_logprob": -0.12674848706114525, "no_speech_prob": 2.0112982460074313e-11, "compression_ratio": 1.3089887640449438 }, { "start": 1279.115, "end": 1280.995, "text": "没提到的设置保持不变即可", "chunk": 2, "language": "zh", "avg_logprob": -0.12674848706114525, "no_speech_prob": 2.0112982460074313e-11, "compression_ratio": 1.3089887640449438 }, { "start": 1280.995, "end": 1284.195, "text": "值得注意的是N8MBinaryFile这一类型中", "chunk": 2, "language": "zh", "avg_logprob": -0.12674848706114525, "no_speech_prob": 2.0112982460074313e-11, "compression_ratio": 1.3089887640449438 }, { "start": 1284.195, "end": 1287.255, "text": "InputDataFieldName需要填写WebHook节点", "chunk": 2, "language": "zh", "avg_logprob": -0.12674848706114525, "no_speech_prob": 2.0112982460074313e-11, "compression_ratio": 1.3089887640449438 }, { "start": 1287.255, "end": 1289.415, "text": "实际收到的二进制字段名称", "chunk": 2, "language": "zh", "avg_logprob": -0.12674848706114525, "no_speech_prob": 2.0112982460074313e-11, "compression_ratio": 1.3089887640449438 }, { "start": 1289.415, "end": 1291.075, "text": "点击前一个WebHook节点", "chunk": 2, "language": "zh", "avg_logprob": -0.12674848706114525, "no_speech_prob": 2.0112982460074313e-11, "compression_ratio": 1.3089887640449438 }, { "start": 1291.075, "end": 1293.555, "text": "在Output的Binary中进行查看", "chunk": 2, "language": "zh", "avg_logprob": -0.12674848706114525, "no_speech_prob": 2.0112982460074313e-11, "compression_ratio": 1.3089887640449438 }, { "start": 1293.555, "end": 1295.795, "text": "如果显示为Data那就填写为Data", "chunk": 2, "language": "zh", "avg_logprob": -0.12674848706114525, "no_speech_prob": 2.0112982460074313e-11, "compression_ratio": 1.3089887640449438 }, { "start": 1295.795, "end": 1298.355, "text": "如果显示为Data0那就填写为Data0", "chunk": 2, "language": "zh", "avg_logprob": -0.12674848706114525, "no_speech_prob": 2.0112982460074313e-11, "compression_ratio": 1.3089887640449438 }, { "start": 1298.355, "end": 1299.855, "text": "这个名称必须完全一致", "chunk": 2, "language": "zh", "avg_logprob": -0.12674848706114525, "no_speech_prob": 2.0112982460074313e-11, "compression_ratio": 1.3089887640449438 }, { "start": 1299.855, "end": 1303.335, "text": "否则会出现TheItem has no binary field的报错", "chunk": 2, "language": "zh", "avg_logprob": -0.12674848706114525, "no_speech_prob": 2.0112982460074313e-11, "compression_ratio": 1.3089887640449438 }, { "start": 1303.335, "end": 1305.895, "text": "第三个节点继续添加HTTPRequest", "chunk": 2, "language": "zh", "avg_logprob": -0.12674848706114525, "no_speech_prob": 2.0112982460074313e-11, "compression_ratio": 1.3089887640449438 }, { "start": 1305.895, "end": 1307.495, "text": "名称改成前文整理", "chunk": 2, "language": "zh", "avg_logprob": -0.13365476661258274, "no_speech_prob": 2.0916516782487626e-11, "compression_ratio": 1.1209964412811388 }, { "start": 1307.495, "end": 1310.615, "text": "这个节点负责把Whisper返回完整转写稿", "chunk": 2, "language": "zh", "avg_logprob": -0.13365476661258274, "no_speech_prob": 2.0916516782487626e-11, "compression_ratio": 1.1209964412811388 }, { "start": 1310.615, "end": 1313.115, "text": "交给前面搭建好的OpenWebUI API", "chunk": 2, "language": "zh", "avg_logprob": -0.13365476661258274, "no_speech_prob": 2.0916516782487626e-11, "compression_ratio": 1.1209964412811388 }, { "start": 1313.115, "end": 1314.995, "text": "第三个节点HTTPRequest", "chunk": 2, "language": "zh", "avg_logprob": -0.13365476661258274, "no_speech_prob": 2.0916516782487626e-11, "compression_ratio": 1.1209964412811388 }, { "start": 1314.995, "end": 1315.915, "text": "具体设置如下", "chunk": 2, "language": "zh", "avg_logprob": -0.13365476661258274, "no_speech_prob": 2.0916516782487626e-11, "compression_ratio": 1.1209964412811388 }, { "start": 1315.915, "end": 1317.995, "text": "没提到的设置保持不便即可", "chunk": 2, "language": "zh", "avg_logprob": -0.13365476661258274, "no_speech_prob": 2.0916516782487626e-11, "compression_ratio": 1.1209964412811388 }, { "start": 1317.995, "end": 1330.095, "text": "接下来回到前面的JSON输入框", "chunk": 2, "language": "zh", "avg_logprob": -0.13365476661258274, "no_speech_prob": 2.0916516782487626e-11, "compression_ratio": 1.1209964412811388 }, { "start": 1330.095, "end": 1332.295, "text": "把整个字段切换到Expression模式", "chunk": 2, "language": "zh", "avg_logprob": -0.13365476661258274, "no_speech_prob": 2.0916516782487626e-11, "compression_ratio": 1.1209964412811388 }, { "start": 1332.295, "end": 1333.895, "text": "再粘贴下面的表达式", "chunk": 2, "language": "zh", "avg_logprob": -0.13365476661258274, "no_speech_prob": 2.0916516782487626e-11, "compression_ratio": 1.1209964412811388 }, { "start": 1333.895, "end": 1335.135, "text": "它的逻辑很简单", "chunk": 2, "language": "zh", "avg_logprob": -0.13365476661258274, "no_speech_prob": 2.0916516782487626e-11, "compression_ratio": 1.1209964412811388 }, { "start": 1335.135, "end": 1338.195, "text": "N8N自动读取Whisper转写节点里面的text", "chunk": 2, "language": "zh", "avg_logprob": -0.1252895154451069, "no_speech_prob": 1.0294205592120687e-11, "compression_ratio": 1.088560885608856 }, { "start": 1338.195, "end": 1340.795, "text": "再连同预设提示词一起发给亲吻", "chunk": 2, "language": "zh", "avg_logprob": -0.1252895154451069, "no_speech_prob": 1.0294205592120687e-11, "compression_ratio": 1.088560885608856 }, { "start": 1340.795, "end": 1346.035, "text": "最后添加Respond to Webhook节点", "chunk": 2, "language": "zh", "avg_logprob": -0.1252895154451069, "no_speech_prob": 1.0294205592120687e-11, "compression_ratio": 1.088560885608856 }, { "start": 1346.035, "end": 1349.015, "text": "它负责把整条工作流结果返回给定用者", "chunk": 2, "language": "zh", "avg_logprob": -0.1252895154451069, "no_speech_prob": 1.0294205592120687e-11, "compression_ratio": 1.088560885608856 }, { "start": 1349.015, "end": 1350.755, "text": "Respond with选择Json", "chunk": 2, "language": "zh", "avg_logprob": -0.1252895154451069, "no_speech_prob": 1.0294205592120687e-11, "compression_ratio": 1.088560885608856 }, { "start": 1350.755, "end": 1353.355, "text": "然后把Respond body切换成Expression模式", "chunk": 2, "language": "zh", "avg_logprob": -0.1252895154451069, "no_speech_prob": 1.0294205592120687e-11, "compression_ratio": 1.088560885608856 }, { "start": 1353.355, "end": 1354.555, "text": "粘贴下面的表达式", "chunk": 2, "language": "zh", "avg_logprob": -0.1252895154451069, "no_speech_prob": 1.0294205592120687e-11, "compression_ratio": 1.088560885608856 }, { "start": 1354.555, "end": 1356.515, "text": "Respond code保持200即可", "chunk": 2, "language": "zh", "avg_logprob": -0.1252895154451069, "no_speech_prob": 1.0294205592120687e-11, "compression_ratio": 1.088560885608856 }, { "start": 1365.135, "end": 1367.495, "text": "工作流设置完成以后", "chunk": 2, "language": "zh", "avg_logprob": -0.10367698254792587, "no_speech_prob": 1.4041091206395961e-11, "compression_ratio": 1.098976109215017 }, { "start": 1367.495, "end": 1368.895, "text": "先不要急着正式启用", "chunk": 2, "language": "zh", "avg_logprob": -0.10367698254792587, "no_speech_prob": 1.4041091206395961e-11, "compression_ratio": 1.098976109215017 }, { "start": 1368.895, "end": 1371.735, "text": "在N8N里点击Listen for Test Event", "chunk": 2, "language": "zh", "avg_logprob": -0.10367698254792587, "no_speech_prob": 1.4041091206395961e-11, "compression_ratio": 1.098976109215017 }, { "start": 1371.735, "end": 1374.095, "text": "让Webhook等待一次测试请求", "chunk": 2, "language": "zh", "avg_logprob": -0.10367698254792587, "no_speech_prob": 1.4041091206395961e-11, "compression_ratio": 1.098976109215017 }, { "start": 1374.095, "end": 1376.115, "text": "然后回到终端执行下面的命令", "chunk": 2, "language": "zh", "avg_logprob": -0.10367698254792587, "no_speech_prob": 1.4041091206395961e-11, "compression_ratio": 1.098976109215017 }, { "start": 1376.115, "end": 1378.815, "text": "注意要把XAI Token后面的站位内容", "chunk": 2, "language": "zh", "avg_logprob": -0.10367698254792587, "no_speech_prob": 1.4041091206395961e-11, "compression_ratio": 1.098976109215017 }, { "start": 1378.815, "end": 1381.455, "text": "替换成你前面实际设置的随机自书串", "chunk": 2, "language": "zh", "avg_logprob": -0.10367698254792587, "no_speech_prob": 1.4041091206395961e-11, "compression_ratio": 1.098976109215017 }, { "start": 1381.455, "end": 1388.815, "text": "如果整个流程正常", "chunk": 2, "language": "zh", "avg_logprob": -0.10367698254792587, "no_speech_prob": 1.4041091206395961e-11, "compression_ratio": 1.098976109215017 }, { "start": 1388.815, "end": 1391.155, "text": "你最终应该会得到类似这样的结果", "chunk": 2, "language": "zh", "avg_logprob": -0.10367698254792587, "no_speech_prob": 1.4041091206395961e-11, "compression_ratio": 1.098976109215017 }, { "start": 1391.155, "end": 1397.335, "text": "测试成功以后回到N8N的工作流程页面", "chunk": 2, "language": "zh", "avg_logprob": -0.11242360603518603, "no_speech_prob": 1.5906335276705263e-11, "compression_ratio": 1.2371794871794872 }, { "start": 1397.335, "end": 1399.295, "text": "点击右上角的Publish发布", "chunk": 2, "language": "zh", "avg_logprob": -0.11242360603518603, "no_speech_prob": 1.5906335276705263e-11, "compression_ratio": 1.2371794871794872 }, { "start": 1399.295, "end": 1403.095, "text": "发布后就不需要每次先点击Lessens for Test Event了", "chunk": 2, "language": "zh", "avg_logprob": -0.11242360603518603, "no_speech_prob": 1.5906335276705263e-11, "compression_ratio": 1.2371794871794872 }, { "start": 1403.095, "end": 1406.595, "text": "正式Webhook地址也会从Webhook Test AI Video Notes", "chunk": 2, "language": "zh", "avg_logprob": -0.11242360603518603, "no_speech_prob": 1.5906335276705263e-11, "compression_ratio": 1.2371794871794872 }, { "start": 1406.595, "end": 1408.755, "text": "变成Webhook AI Video Notes", "chunk": 2, "language": "zh", "avg_logprob": -0.11242360603518603, "no_speech_prob": 1.5906335276705263e-11, "compression_ratio": 1.2371794871794872 }, { "start": 1408.755, "end": 1410.755, "text": "以后只要把视频传到这个接口", "chunk": 2, "language": "zh", "avg_logprob": -0.11242360603518603, "no_speech_prob": 1.5906335276705263e-11, "compression_ratio": 1.2371794871794872 }, { "start": 1410.755, "end": 1413.075, "text": "整套流程就会自动变成Whisper", "chunk": 2, "language": "zh", "avg_logprob": -0.11242360603518603, "no_speech_prob": 1.5906335276705263e-11, "compression_ratio": 1.2371794871794872 }, { "start": 1413.075, "end": 1415.075, "text": "在本地把视频转成文字", "chunk": 2, "language": "zh", "avg_logprob": -0.11242360603518603, "no_speech_prob": 1.5906335276705263e-11, "compression_ratio": 1.2371794871794872 }, { "start": 1415.075, "end": 1418.975, "text": "亲吻再在本地完成摘要提纲标题和章节建议", "chunk": 2, "language": "zh", "avg_logprob": -0.11242360603518603, "no_speech_prob": 1.5906335276705263e-11, "compression_ratio": 1.2371794871794872 }, { "start": 1418.975, "end": 1421.695, "text": "最后把转写稿和整理结果一起返回", "chunk": 2, "language": "zh", "avg_logprob": -0.12499096196725829, "no_speech_prob": 2.4016337529797482e-11, "compression_ratio": 1.2734177215189872 }, { "start": 1421.695, "end": 1423.655, "text": "整个过程中的语音和文本内容", "chunk": 2, "language": "zh", "avg_logprob": -0.12499096196725829, "no_speech_prob": 2.4016337529797482e-11, "compression_ratio": 1.2734177215189872 }, { "start": 1423.655, "end": 1425.695, "text": "都可以在自己的Mac上完成处理", "chunk": 2, "language": "zh", "avg_logprob": -0.12499096196725829, "no_speech_prob": 2.4016337529797482e-11, "compression_ratio": 1.2734177215189872 }, { "start": 1425.695, "end": 1428.075, "text": "到这里整套工作流已经跑通了", "chunk": 2, "language": "zh", "avg_logprob": -0.12499096196725829, "no_speech_prob": 2.4016337529797482e-11, "compression_ratio": 1.2734177215189872 }, { "start": 1428.075, "end": 1430.595, "text": "不过直接使用Curl调用正式WebHook时", "chunk": 2, "language": "zh", "avg_logprob": -0.12499096196725829, "no_speech_prob": 2.4016337529797482e-11, "compression_ratio": 1.2734177215189872 }, { "start": 1430.595, "end": 1432.555, "text": "终端返回的是标准的Json", "chunk": 2, "language": "zh", "avg_logprob": -0.12499096196725829, "no_speech_prob": 2.4016337529797482e-11, "compression_ratio": 1.2734177215189872 }, { "start": 1432.555, "end": 1434.515, "text": "所以会看到大量的写杠N", "chunk": 2, "language": "zh", "avg_logprob": -0.12499096196725829, "no_speech_prob": 2.4016337529797482e-11, "compression_ratio": 1.2734177215189872 }, { "start": 1434.515, "end": 1436.115, "text": "引号和转译字符", "chunk": 2, "language": "zh", "avg_logprob": -0.12499096196725829, "no_speech_prob": 2.4016337529797482e-11, "compression_ratio": 1.2734177215189872 }, { "start": 1436.115, "end": 1437.795, "text": "比如写杠N其实就是换", "chunk": 2, "language": "zh", "avg_logprob": -0.12499096196725829, "no_speech_prob": 2.4016337529797482e-11, "compression_ratio": 1.2734177215189872 }, { "start": 1437.795, "end": 1438.775, "text": "并不是烂码", "chunk": 2, "language": "zh", "avg_logprob": -0.12499096196725829, "no_speech_prob": 2.4016337529797482e-11, "compression_ratio": 1.2734177215189872 }, { "start": 1438.775, "end": 1442.495, "text": "只是Json保存Markdown和多项文字时的正常格式", "chunk": 2, "language": "zh", "avg_logprob": -0.12499096196725829, "no_speech_prob": 2.4016337529797482e-11, "compression_ratio": 1.2734177215189872 }, { "start": 1442.495, "end": 1444.655, "text": "为了让最终结果更加适合阅读", "chunk": 2, "language": "zh", "avg_logprob": -0.12499096196725829, "no_speech_prob": 2.4016337529797482e-11, "compression_ratio": 1.2734177215189872 }, { "start": 1444.655, "end": 1445.815, "text": "我们可以再用JQ", "chunk": 2, "language": "zh", "avg_logprob": -0.12499096196725829, "no_speech_prob": 2.4016337529797482e-11, "compression_ratio": 1.2734177215189872 }, { "start": 1445.815, "end": 1448.915, "text": "把Json里的转写稿和千万之立结果提取出来", "chunk": 2, "language": "zh", "avg_logprob": -0.12499096196725829, "no_speech_prob": 2.4016337529797482e-11, "compression_ratio": 1.2734177215189872 }, { "start": 1448.915, "end": 1450.275, "text": "如果已经安装Homebrew", "chunk": 2, "language": "zh", "avg_logprob": -0.1930213165283203, "no_speech_prob": 7.630149984061418e-12, "compression_ratio": 0.9172932330827067 }, { "start": 1450.275, "end": 1452.315, "text": "则可以执行Brew Install JQ", "chunk": 2, "language": "zh", "avg_logprob": -0.1930213165283203, "no_speech_prob": 7.630149984061418e-12, "compression_ratio": 0.9172932330827067 }, { "start": 1452.315, "end": 1453.415, "text": "安装完整后", "chunk": 2, "language": "zh", "avg_logprob": -0.1930213165283203, "no_speech_prob": 7.630149984061418e-12, "compression_ratio": 0.9172932330827067 }, { "start": 1453.415, "end": 1456.115, "text": "把刚才电用正式工作流命令稍微改一下", "chunk": 2, "language": "zh", "avg_logprob": -0.1930213165283203, "no_speech_prob": 7.630149984061418e-12, "compression_ratio": 0.9172932330827067 }, { "start": 1474.207, "end": 1476.407, "text": "哦对了如果你的Mac没有安装Homebrew", "chunk": 3, "language": "zh", "avg_logprob": -0.14379198927628367, "no_speech_prob": 1.804685220707647e-11, "compression_ratio": 1.1855345911949686 }, { "start": 1476.407, "end": 1478.607, "text": "也不一定非要额外安装JQ", "chunk": 3, "language": "zh", "avg_logprob": -0.14379198927628367, "no_speech_prob": 1.804685220707647e-11, "compression_ratio": 1.1855345911949686 }, { "start": 1478.607, "end": 1480.607, "text": "前面已经准备了Python3.11", "chunk": 3, "language": "zh", "avg_logprob": -0.14379198927628367, "no_speech_prob": 1.804685220707647e-11, "compression_ratio": 1.1855345911949686 }, { "start": 1480.607, "end": 1484.007, "text": "可以直接用Python自带的Json模块完成同样的处理", "chunk": 3, "language": "zh", "avg_logprob": -0.14379198927628367, "no_speech_prob": 1.804685220707647e-11, "compression_ratio": 1.1855345911949686 }, { "start": 1489.207, "end": 1492.207, "text": "视频最后在补一个实际使用时一定会碰到的问题", "chunk": 3, "language": "zh", "avg_logprob": -0.14379198927628367, "no_speech_prob": 1.804685220707647e-11, "compression_ratio": 1.1855345911949686 }, { "start": 1492.207, "end": 1495.607, "text": "Mac mini重启以后哪些东西需要重新打开", "chunk": 3, "language": "zh", "avg_logprob": -0.14379198927628367, "no_speech_prob": 1.804685220707647e-11, "compression_ratio": 1.1855345911949686 }, { "start": 1495.607, "end": 1498.407, "text": "我们前面搭载的模型和数据都保存在硬盘里", "chunk": 3, "language": "zh", "avg_logprob": -0.14379198927628367, "no_speech_prob": 1.804685220707647e-11, "compression_ratio": 1.1855345911949686 }, { "start": 1498.407, "end": 1500.407, "text": "不需要重新安装或者重新下载", "chunk": 3, "language": "zh", "avg_logprob": -0.14379198927628367, "no_speech_prob": 1.804685220707647e-11, "compression_ratio": 1.1855345911949686 }, { "start": 1500.407, "end": 1505.647, "text": "但Olama,OpenWebUI,WhisperServer和N8N这些运行中的服务需要重新启动", "chunk": 3, "language": "zh", "avg_logprob": -0.19320030212402345, "no_speech_prob": 2.2113401387802156e-11, "compression_ratio": 1.2853185595567866 }, { "start": 1505.647, "end": 1507.307, "text": "建议按照下面这个顺序来", "chunk": 3, "language": "zh", "avg_logprob": -0.19320030212402345, "no_speech_prob": 2.2113401387802156e-11, "compression_ratio": 1.2853185595567866 }, { "start": 1507.607, "end": 1509.247, "text": "首先启动最底层的Olama", "chunk": 3, "language": "zh", "avg_logprob": -0.19320030212402345, "no_speech_prob": 2.2113401387802156e-11, "compression_ratio": 1.2853185595567866 }, { "start": 1509.247, "end": 1511.147, "text": "如果安装的是MicOS版的Olama", "chunk": 3, "language": "zh", "avg_logprob": -0.19320030212402345, "no_speech_prob": 2.2113401387802156e-11, "compression_ratio": 1.2853185595567866 }, { "start": 1511.147, "end": 1512.847, "text": "直接打开OlamaApp即可", "chunk": 3, "language": "zh", "avg_logprob": -0.19320030212402345, "no_speech_prob": 2.2113401387802156e-11, "compression_ratio": 1.2853185595567866 }, { "start": 1512.847, "end": 1515.747, "text": "也可以在中端里执行Open-AOlama", "chunk": 3, "language": "zh", "avg_logprob": -0.19320030212402345, "no_speech_prob": 2.2113401387802156e-11, "compression_ratio": 1.2853185595567866 }, { "start": 1515.747, "end": 1518.267, "text": "然后可以检查Olama是否已经恢复", "chunk": 3, "language": "zh", "avg_logprob": -0.19320030212402345, "no_speech_prob": 2.2113401387802156e-11, "compression_ratio": 1.2853185595567866 }, { "start": 1518.567, "end": 1520.967, "text": "如果能够正常返回Olama的版本信息", "chunk": 3, "language": "zh", "avg_logprob": -0.19320030212402345, "no_speech_prob": 2.2113401387802156e-11, "compression_ratio": 1.2853185595567866 }, { "start": 1520.967, "end": 1523.267, "text": "就说明底层模型服务已经正常", "chunk": 3, "language": "zh", "avg_logprob": -0.19320030212402345, "no_speech_prob": 2.2113401387802156e-11, "compression_ratio": 1.2853185595567866 }, { "start": 1523.267, "end": 1525.207, "text": "接下来再来启动OpenWebUI", "chunk": 3, "language": "zh", "avg_logprob": -0.19320030212402345, "no_speech_prob": 2.2113401387802156e-11, "compression_ratio": 1.2853185595567866 }, { "start": 1525.207, "end": 1529.067, "text": "冲击后可以先确认MacMe的局域网IP地址有没有发生变化", "chunk": 3, "language": "zh", "avg_logprob": -0.19320030212402345, "no_speech_prob": 2.2113401387802156e-11, "compression_ratio": 1.2853185595567866 }, { "start": 1529.067, "end": 1531.867, "text": "因为我们前面的启动命令里填写过这个地址", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1531.867, "end": 1533.547, "text": "如果地址没有发生变化", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1533.547, "end": 1534.767, "text": "就继续使用之前的IP", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1534.767, "end": 1535.867, "text": "如果变化了", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1535.867, "end": 1537.787, "text": "下面命令行里的也要一起替换", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1537.787, "end": 1539.647, "text": "这个中轮装口需要保持运行", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1539.647, "end": 1540.887, "text": "然后617访问", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1540.887, "end": 1542.847, "text": "127.0.0.1网号3000", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1542.847, "end": 1544.207, "text": "居于网中的其他设备", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1544.207, "end": 1545.767, "text": "仍然访问Mac mini的IP地址", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1545.767, "end": 1546.487, "text": "网号3000", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1546.487, "end": 1547.587, "text": "长期使用的话", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1547.587, "end": 1548.687, "text": "也可以在路由器里", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1548.687, "end": 1550.667, "text": "给Mac mini固定一个局网地址", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1550.667, "end": 1552.767, "text": "选择每次重建后来进行检查", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1552.767, "end": 1553.887, "text": "如果只是聊天", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1553.887, "end": 1555.467, "text": "知识库或者OpenWeb UI", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1555.467, "end": 1557.447, "text": "其实不用一直启动Whisper server", "chunk": 3, "language": "zh", "avg_logprob": -0.137040545723655, "no_speech_prob": 2.9758435299287456e-11, "compression_ratio": 1.2729591836734695 }, { "start": 1557.447, "end": 1559.027, "text": "只有需要运行视频转写", "chunk": 3, "language": "zh", "avg_logprob": -0.0922096911991868, "no_speech_prob": 1.7503587121381337e-11, "compression_ratio": 1.3157894736842106 }, { "start": 1559.027, "end": 1561.027, "text": "或者后面的N8n视频工作流时", "chunk": 3, "language": "zh", "avg_logprob": -0.0922096911991868, "no_speech_prob": 1.7503587121381337e-11, "compression_ratio": 1.3157894736842106 }, { "start": 1561.027, "end": 1561.967, "text": "再打开即可", "chunk": 3, "language": "zh", "avg_logprob": -0.0922096911991868, "no_speech_prob": 1.7503587121381337e-11, "compression_ratio": 1.3157894736842106 }, { "start": 1561.967, "end": 1563.987, "text": "同样这个中端窗口需要保持运行", "chunk": 3, "language": "zh", "avg_logprob": -0.0922096911991868, "no_speech_prob": 1.7503587121381337e-11, "compression_ratio": 1.3157894736842106 }, { "start": 1563.987, "end": 1565.707, "text": "可以打开另一个中端测试", "chunk": 3, "language": "zh", "avg_logprob": -0.0922096911991868, "no_speech_prob": 1.7503587121381337e-11, "compression_ratio": 1.3157894736842106 }, { "start": 1565.707, "end": 1567.487, "text": "只要服务器能够这样响应", "chunk": 3, "language": "zh", "avg_logprob": -0.0922096911991868, "no_speech_prob": 1.7503587121381337e-11, "compression_ratio": 1.3157894736842106 }, { "start": 1567.487, "end": 1569.707, "text": "就可以继续交给N8n来调用了", "chunk": 3, "language": "zh", "avg_logprob": -0.0922096911991868, "no_speech_prob": 1.7503587121381337e-11, "compression_ratio": 1.3157894736842106 }, { "start": 1569.707, "end": 1571.747, "text": "接着打开Docker Desktop App", "chunk": 3, "language": "zh", "avg_logprob": -0.0922096911991868, "no_speech_prob": 1.7503587121381337e-11, "compression_ratio": 1.3157894736842106 }, { "start": 1571.747, "end": 1573.607, "text": "因为我们创建N8n容器时", "chunk": 3, "language": "zh", "avg_logprob": -0.0922096911991868, "no_speech_prob": 1.7503587121381337e-11, "compression_ratio": 1.3157894736842106 }, { "start": 1573.607, "end": 1576.087, "text": "已经设置了Restart Unless Stoped", "chunk": 3, "language": "zh", "avg_logprob": -0.0922096911991868, "no_speech_prob": 1.7503587121381337e-11, "compression_ratio": 1.3157894736842106 }, { "start": 1576.087, "end": 1577.127, "text": "所以正常情况下", "chunk": 3, "language": "zh", "avg_logprob": -0.0922096911991868, "no_speech_prob": 1.7503587121381337e-11, "compression_ratio": 1.3157894736842106 }, { "start": 1577.127, "end": 1578.487, "text": "只要Docker App启动", "chunk": 3, "language": "zh", "avg_logprob": -0.0922096911991868, "no_speech_prob": 1.7503587121381337e-11, "compression_ratio": 1.3157894736842106 }, { "start": 1578.487, "end": 1580.627, "text": "之前的N8n容器也会自动回复运行", "chunk": 3, "language": "zh", "avg_logprob": -0.0922096911991868, "no_speech_prob": 1.7503587121381337e-11, "compression_ratio": 1.3157894736842106 }, { "start": 1580.627, "end": 1582.807, "text": "不需要重新执行那一串Docker Run", "chunk": 3, "language": "zh", "avg_logprob": -0.0922096911991868, "no_speech_prob": 1.7503587121381337e-11, "compression_ratio": 1.3157894736842106 }, { "start": 1582.807, "end": 1584.727, "text": "可以先检查Docker PS", "chunk": 3, "language": "zh", "avg_logprob": -0.0922096911991868, "no_speech_prob": 1.7503587121381337e-11, "compression_ratio": 1.3157894736842106 }, { "start": 1584.727, "end": 1586.487, "text": "如果列表里能够看到N8n", "chunk": 3, "language": "zh", "avg_logprob": -0.0922096911991868, "no_speech_prob": 1.7503587121381337e-11, "compression_ratio": 1.3157894736842106 }, { "start": 1586.487, "end": 1588.347, "text": "就说明N8N已经自动恢复", "chunk": 3, "language": "zh", "avg_logprob": -0.10418566703796386, "no_speech_prob": 2.1596158888970152e-11, "compression_ratio": 1.1615598885793872 }, { "start": 1588.347, "end": 1589.387, "text": "如果没有启动", "chunk": 3, "language": "zh", "avg_logprob": -0.10418566703796386, "no_speech_prob": 2.1596158888970152e-11, "compression_ratio": 1.1615598885793872 }, { "start": 1589.387, "end": 1591.687, "text": "可以手动执行Docker StartN8N", "chunk": 3, "language": "zh", "avg_logprob": -0.10418566703796386, "no_speech_prob": 2.1596158888970152e-11, "compression_ratio": 1.1615598885793872 }, { "start": 1591.687, "end": 1592.947, "text": "然后浏览器打开", "chunk": 3, "language": "zh", "avg_logprob": -0.10418566703796386, "no_speech_prob": 2.1596158888970152e-11, "compression_ratio": 1.1615598885793872 }, { "start": 1592.947, "end": 1595.207, "text": "127.0.0.1-5678", "chunk": 3, "language": "zh", "avg_logprob": -0.10418566703796386, "no_speech_prob": 2.1596158888970152e-11, "compression_ratio": 1.1615598885793872 }, { "start": 1595.207, "end": 1596.947, "text": "如果能够进入N8N", "chunk": 3, "language": "zh", "avg_logprob": -0.10418566703796386, "no_speech_prob": 2.1596158888970152e-11, "compression_ratio": 1.1615598885793872 }, { "start": 1596.947, "end": 1599.047, "text": "就说明自动化服务已经恢复正常", "chunk": 3, "language": "zh", "avg_logprob": -0.10418566703796386, "no_speech_prob": 2.1596158888970152e-11, "compression_ratio": 1.1615598885793872 }, { "start": 1599.047, "end": 1602.167, "text": "到这里我们就从一台普通的Mac mini开始", "chunk": 3, "language": "zh", "avg_logprob": -0.10418566703796386, "no_speech_prob": 2.1596158888970152e-11, "compression_ratio": 1.1615598885793872 }, { "start": 1602.167, "end": 1605.647, "text": "完成一套相对完整的家庭本地AI服务", "chunk": 3, "language": "zh", "avg_logprob": -0.10418566703796386, "no_speech_prob": 2.1596158888970152e-11, "compression_ratio": 1.1615598885793872 }, { "start": 1605.647, "end": 1607.467, "text": "Olama负责运行AI模型", "chunk": 3, "language": "zh", "avg_logprob": -0.10418566703796386, "no_speech_prob": 2.1596158888970152e-11, "compression_ratio": 1.1615598885793872 }, { "start": 1607.467, "end": 1610.447, "text": "OpenWeb UI提供对话界面和统一API", "chunk": 3, "language": "zh", "avg_logprob": -0.10418566703796386, "no_speech_prob": 2.1596158888970152e-11, "compression_ratio": 1.1615598885793872 }, { "start": 1610.447, "end": 1613.327, "text": "Embedding和IAG负责自己的文档支援库", "chunk": 3, "language": "zh", "avg_logprob": -0.10418566703796386, "no_speech_prob": 2.1596158888970152e-11, "compression_ratio": 1.1615598885793872 }, { "start": 1613.327, "end": 1616.187, "text": "GLM OCR处理图片和扫描件", "chunk": 3, "language": "zh", "avg_logprob": -0.10418566703796386, "no_speech_prob": 2.1596158888970152e-11, "compression_ratio": 1.1615598885793872 }, { "start": 1616.187, "end": 1618.827, "text": "whisper.cpp负责视频录音转写", "chunk": 3, "language": "zh", "avg_logprob": -0.09457722555827625, "no_speech_prob": 1.6106957781425457e-11, "compression_ratio": 1.2406417112299466 }, { "start": 1618.827, "end": 1621.387, "text": "再通过N8N把这些能力串成一套", "chunk": 3, "language": "zh", "avg_logprob": -0.09457722555827625, "no_speech_prob": 1.6106957781425457e-11, "compression_ratio": 1.2406417112299466 }, { "start": 1621.387, "end": 1622.987, "text": "简单的自动化工作流", "chunk": 3, "language": "zh", "avg_logprob": -0.09457722555827625, "no_speech_prob": 1.6106957781425457e-11, "compression_ratio": 1.2406417112299466 }, { "start": 1622.987, "end": 1625.287, "text": "这套方案其实不只是在Mac上", "chunk": 3, "language": "zh", "avg_logprob": -0.09457722555827625, "no_speech_prob": 1.6106957781425457e-11, "compression_ratio": 1.2406417112299466 }, { "start": 1625.287, "end": 1626.507, "text": "跑了一个AI模型", "chunk": 3, "language": "zh", "avg_logprob": -0.09457722555827625, "no_speech_prob": 1.6106957781425457e-11, "compression_ratio": 1.2406417112299466 }, { "start": 1626.507, "end": 1628.507, "text": "当对话文档图片", "chunk": 3, "language": "zh", "avg_logprob": -0.09457722555827625, "no_speech_prob": 1.6106957781425457e-11, "compression_ratio": 1.2406417112299466 }, { "start": 1628.507, "end": 1631.027, "text": "音视频和自动化都连起来之后", "chunk": 3, "language": "zh", "avg_logprob": -0.09457722555827625, "no_speech_prob": 1.6106957781425457e-11, "compression_ratio": 1.2406417112299466 }, { "start": 1631.027, "end": 1632.927, "text": "这台Mac mini才开始从一台", "chunk": 3, "language": "zh", "avg_logprob": -0.09457722555827625, "no_speech_prob": 1.6106957781425457e-11, "compression_ratio": 1.2406417112299466 }, { "start": 1632.927, "end": 1633.887, "text": "跑模型的电脑", "chunk": 3, "language": "zh", "avg_logprob": -0.09457722555827625, "no_speech_prob": 1.6106957781425457e-11, "compression_ratio": 1.2406417112299466 }, { "start": 1633.887, "end": 1636.127, "text": "变成一台可以长期放在家里", "chunk": 3, "language": "zh", "avg_logprob": -0.09457722555827625, "no_speech_prob": 1.6106957781425457e-11, "compression_ratio": 1.2406417112299466 }, { "start": 1636.127, "end": 1638.767, "text": "由不同设备随时调用的AI服务器", "chunk": 3, "language": "zh", "avg_logprob": -0.09457722555827625, "no_speech_prob": 1.6106957781425457e-11, "compression_ratio": 1.2406417112299466 }, { "start": 1638.767, "end": 1641.507, "text": "当然这只是一个非常技术的框架", "chunk": 3, "language": "zh", "avg_logprob": -0.09457722555827625, "no_speech_prob": 1.6106957781425457e-11, "compression_ratio": 1.2406417112299466 }, { "start": 1641.507, "end": 1643.527, "text": "后面还可以继续加入更大的模型", "chunk": 3, "language": "zh", "avg_logprob": -0.09457722555827625, "no_speech_prob": 1.6106957781425457e-11, "compression_ratio": 1.2406417112299466 }, { "start": 1643.527, "end": 1644.367, "text": "自己的资料库", "chunk": 3, "language": "zh", "avg_logprob": -0.09457722555827625, "no_speech_prob": 1.6106957781425457e-11, "compression_ratio": 1.2406417112299466 }, { "start": 1644.367, "end": 1645.327, "text": "手机快据指令", "chunk": 3, "language": "zh", "avg_logprob": -0.09457722555827625, "no_speech_prob": 1.6106957781425457e-11, "compression_ratio": 1.2406417112299466 }, { "start": 1645.327, "end": 1646.927, "text": "甚至更多自动化流程", "chunk": 3, "language": "zh", "avg_logprob": -0.17684212213830103, "no_speech_prob": 2.1418217893698355e-11, "compression_ratio": 1.2467948717948718 }, { "start": 1646.927, "end": 1649.227, "text": "大家完全可以根据自己的Mac配置", "chunk": 3, "language": "zh", "avg_logprob": -0.17684212213830103, "no_speech_prob": 2.1418217893698355e-11, "compression_ratio": 1.2467948717948718 }, { "start": 1649.227, "end": 1651.007, "text": "和实际需求继续往上搭", "chunk": 3, "language": "zh", "avg_logprob": -0.17684212213830103, "no_speech_prob": 2.1418217893698355e-11, "compression_ratio": 1.2467948717948718 }, { "start": 1651.007, "end": 1653.367, "text": "先把这套技术环境稳定跑起来", "chunk": 3, "language": "zh", "avg_logprob": -0.17684212213830103, "no_speech_prob": 2.1418217893698355e-11, "compression_ratio": 1.2467948717948718 }, { "start": 1653.367, "end": 1654.667, "text": "再逐步增加功能", "chunk": 3, "language": "zh", "avg_logprob": -0.17684212213830103, "no_speech_prob": 2.1418217893698355e-11, "compression_ratio": 1.2467948717948718 }, { "start": 1654.667, "end": 1657.167, "text": "更适合第一次接触本地AI的各位", "chunk": 3, "language": "zh", "avg_logprob": -0.17684212213830103, "no_speech_prob": 2.1418217893698355e-11, "compression_ratio": 1.2467948717948718 }, { "start": 1657.167, "end": 1657.627, "text": "好了", "chunk": 3, "language": "zh", "avg_logprob": -0.17684212213830103, "no_speech_prob": 2.1418217893698355e-11, "compression_ratio": 1.2467948717948718 }, { "start": 1657.627, "end": 1659.607, "text": "那么以上就是本期视频的全部内容了", "chunk": 3, "language": "zh", "avg_logprob": -0.17684212213830103, "no_speech_prob": 2.1418217893698355e-11, "compression_ratio": 1.2467948717948718 }, { "start": 1659.607, "end": 1661.367, "text": "如果觉得本期视频有所帮助的话", "chunk": 3, "language": "zh", "avg_logprob": -0.17684212213830103, "no_speech_prob": 2.1418217893698355e-11, "compression_ratio": 1.2467948717948718 }, { "start": 1661.367, "end": 1662.227, "text": "还请不要忘记", "chunk": 3, "language": "zh", "avg_logprob": -0.17684212213830103, "no_speech_prob": 2.1418217893698355e-11, "compression_ratio": 1.2467948717948718 }, { "start": 1662.227, "end": 1663.647, "text": "剪辑赛链加公主支持一下", "chunk": 3, "language": "zh", "avg_logprob": -0.17684212213830103, "no_speech_prob": 2.1418217893698355e-11, "compression_ratio": 1.2467948717948718 }, { "start": 1663.647, "end": 1664.167, "text": "支持旅金", "chunk": 3, "language": "zh", "avg_logprob": -0.17684212213830103, "no_speech_prob": 2.1418217893698355e-11, "compression_ratio": 1.2467948717948718 }, { "start": 1664.167, "end": 1665.547, "text": "我们下期视频再见", "chunk": 3, "language": "zh", "avg_logprob": -0.17684212213830103, "no_speech_prob": 2.1418217893698355e-11, "compression_ratio": 1.2467948717948718 }, { "start": 1665.547, "end": 1666.087, "text": "拜拜", "chunk": 3, "language": "zh", "avg_logprob": -0.17684212213830103, "no_speech_prob": 2.1418217893698355e-11, "compression_ratio": 1.2467948717948718 } ], "segmented_transcription": { "status": "SEGMENTED_TRANSCRIPT_OK", "created_at": "2026-10-01T08:27:35", "manifest": "/Users/sagawa/AI_WORK/video_notes/url/20261001_082635__Mac_mini_变身家庭_AI_服务器_从零搭建完整本地_AI_工作流/segmented_work/chunk_manifest.json", "audio": "/Users/sagawa/AI_WORK/video_notes/url/20261001_082635__Mac_mini_变身家庭_AI_服务器_从零搭建完整本地_AI_工作流/audio_16k_mono.wav", "audio_sha256": "552e76d8fd45b1062d70b690a9057d5f74f2ebcd1bd79c7137a7b27b07e78ea8", "audio_duration": 1666.751563, "model": "/Users/sagawa/AI_WORK/cache/huggingface/hub/models--mlx-community--whisper-large-v3-turbo/snapshots/a4aaeec0636e6fef84abdcbe3544cb2bf7e9f6fb", "language_hint": "zh", "condition_on_previous_text": false, "chunk_count": 4, "resumed_chunks": 0, "merged_segment_count": 717, "cleaned_segment_count": 717, "removed_hallucination_segment_count": 0, "final_segment_end": 1666.087, "global_repeat_run": 1, "global_repeat_text": "Hello,大家好,我是周日金。AI這兩年離咱們普通用戶越來越近。現在甚至不需要單獨買一台服務器,我們就可以直接把手頭閒置在Mac,變成一台屬於自己的本地AI服務器。", "tail_repeat_run": 1, "tail_repeat_text": "GLM OCR处理图片和扫描件", "hallucination_region_count": 0, "hallucination_regions": [], "qc_ok": true, "qc_errors": [] } }