{ "text": "Agent 将彻底改变了。一条命令，十个核心玩法。我的 Deepseek Harness 已经变成了另外一个 Codex。一切皆插件，模型、工具、界面、状态栏全是插件。究竟它有怎样的魅力？这期视频当中所有的插件全部都免费开源。\n你能够用的安装说明。我们先来看安装。\n关于安装，官方提供了一个 NPX 命令，可以一键启动我们的 Deepseek Harness。前提是你要安装好 Node.js。\n如果你使用的是 Windows 系统，它可能会出错。遇到这种情况，你可以先用 NPM 全局安装它，然后再用 DSH 命令来启动。\n这样就可以解决了。如果你用的是官方的默认版本，大概率觉得只不过是又一个玩具，很粗糙，很多功能都没有。\n你先去看看官方的 GitHub，已经有 100 多 K 的 STAR。如果你去搜索插件，仓库已经有 300 个。GitHub 的仓库散落在各个地方，琳琅满目，无从下手。\n所以我们首先要装一个插件市场。其实非常简单，直接打开命令行，用 DSH 的插件命令安装几个。\n安装完以后，我们重启一下。在设置里面，我们就可以找到插件。安装完插件以后，界面会多出一个插件市场的 Tab 栏。\n在这里面，我们就可以搜索你任意想要的任何插件的关键字进行查询。查询的结果，我们可以看到 STAR。你可以点具体某一个仓库，就会跳到 GitHub 上面去看它的详情。\n重要的是，它可以直接安装。安装完以后，再重启一下就可以。在使用 Agent 的时候，其实大家平时用的最多的就是上下文。\n你比如说要给它看一个文件或者一个目录，你需要去 Ata 去指定。但它默认是不支持。所以我们首先要在刚刚的插件市场里面去找到一个 Add File 这个插件。\n然后直接安装它。安装好以后，我们重启。当然也可以用命令来安装。接着我们在输入 Ata 的时候，它就能够选择我们当前工作目录下的文件和文件夹了。\n接下来我安装了一个任务看板。无论是 Codex 还是 Workbox 都是没有的。但我特别喜欢这个功能。我们直接用命令安装它，直接用 GitHub 的形式来安装。安装完以后我们也要重启。\n我们在左边任意点击一个项目里面的一个项目。我们看到右上角就会有一个任务看板。点开就可以看到它的样子。\n里面我们可以切换不同的项目。切换以后我们可以直接在这里发送需求。然后等审核以后拉到待办区里面。Agent 就可以自动工作了。而且你还可以设定定时任务。\n运行的时候相应的任务也会显示在左边的目录里面。任务是广注了。但还有一个硬伤。Deepseek 是文字模型。它不是多模态。我们用两种方式来解决。\n第一种就是网上用的最多的就是 Multimodal。我们直接一个命令就可以安装好这个插件。重启以后，如果你是选择 Deepseek Flush，系统还是会提示你模型不支持这个图像。\n你需要选择。你需要选择我们刚刚安装好这个模型以后，会多出来的两个模型。Deepseek Plus 和 Pro 后面都多了一个 Multimodal。你选择这两个模型以后，那它就会自动识别这个图形。就不用再会提示你不支持图像了。\n第二种方式，我们装另外一个插件。它可以指定你图像识别的模型。这样的话，你可以指定任意你想要的图像模型来识别。你只需要安装它的 SDK 就可以了。\n同样的，安装好以后，我们在设置里面，你就可以看到它多了一个菜单。这里面你就可以选择或者是更换你的图像识别的模型。\n这样你在即便是选择 Deepseek Plus 的时候，它也能够自动调取图像识别工具，进行图像识别了。\n如果你想要发送文件，或者是拖出图片进行聊天，却发现无法拖动。这里我们需要安装一个文件拖动的功能。\n我们直接用插件命令进行安装。安装好以后，我们就可以在电脑的任何地方把你想要拉入的文件或者是图片拉入进来。\n拖入以后，它会自动翻转输入框，并且带有你的文件路径，非常方便。\n我们发现，在输出的对话框里面的这些文件，我们点击它是没有反应的，无法打开。\n实际上这个时候我们像 Codex 或者 Workbox 一样需要一个类似的侧边栏。\n我们可以在右侧的侧边栏，点击文件就可以进行预览，或者是查看目录。\n安装好以后，我们就可以在右侧看到工作目录下的所有文件。\n点击文件以后，像 MD 或者 HTML 这样的一些文件，我们就可以在右边进行浏览了。\n除了 MD 和浏览以外，它还提供了终端和源代码的管理功能。\n而且还能看到多个 Agent 的列表，点击 Agent 还能看到它具体的日志，非常方便。\n但它的这个终端，我们要给它禁掉，因为它有一点问题。\n我们待会儿重新装一个新的终端。\n如果你要使用 MCP，那也需要一个可视化的 MCP 管理。\n我们这里就装一个 MCP 的插件。这样的话，你就可以在设置的菜单里面找到 MCP 的菜单。\n你就可以手动在页面上进行管理你的 MCP 了。不用再去街市里面修改。\n接下来我们看到在整个过程里面，整个页面显得不是那么灵动。\n状态栏的内容是固定的。我们需要安装一个自定义的状态栏。\n一个命令就可以给它装好。\n然后我们看到这里的词语，它会不断的变化。\n这些词语你可以进行自定义,你可以进行修改。\n看起来是不是更加的接近于Codex和Cloud Code了呢?\n现在状态是好看的。但是文本输出还是千篇一律全是文字。\n对于Markdown这样的东西,没有办法输出相应的表格,也不够美化。\n我们需要装一个副文本的显示插件。\n这样我们就可以像Mermud这样的流程图,进行可视化了。\n同时表格文字它还给你加一些底色,高亮,排版美化等等。\n更重要的是它还可以输出图表一样的内容。\n我们直接让它分析一个内容,它就可以直接用图表的形式给你呈现。\n这种方式我太喜欢了。\n最后一个我们把终端给它集成进来。\n刚刚特别难自带的终端有一些问题。\n我们独自安装一个终端,让这个终端也是一个命令就可以安装好。\n而且安装好以后它就会出现在我们页面下半。\n默认状态下它是收起来的。\n点开我们就可以看到,非常方便。\n在上面你可以写命令,跑脚本,都是没有问题的。\n上面的这些插件安装完以后,你会发现一件事情。\nDeepseekHarness它不是一个软件。\n它是一个像落高的底座,插件就是它的积木。\n现在AI已经从谁的大脑更聪明,变成了谁的身体更好用。\n模型决定了Agent多聪明,插件决定了Agent也多能干。\n当然还有很多好玩的插件,也可以自己去专一行。\n如果你喜欢,下期我教你如何写一个自己的DeepseekHarness插件。\n点个关注,别错过了。", "language": "zh", "duration": 461.450188, "segments": [ { "id": 0, "seek": 0, "start": 0.0, "end": 21.34, "text": "Agent 将彻底改变了。一条命令，十个核心玩法。我的 Deepseek Harness 已经变成了另外一个 Codex。一切皆插件，模型、工具、界面、状态栏全是插件。究竟它有怎样的魅力？这期视频当中所有的插件全部都免费开源。", "tokens": [ 50365, 32, 6930, 220, 45456, 7391, 119, 22816, 34490, 2129, 246, 2289, 1543, 2257, 2392, 104, 25236, 49061, 11, 20145, 7549, 8987, 116, 7945, 19912, 11148, 1543, 14200, 11089, 595, 405, 916, 389, 16937, 49161, 2129, 246, 11336, 2289, 26202, 20182, 34, 1429, 87, 1543, 2257, 23632, 14468, 11673, 240, 20485, 11, 41908, 39823, 1231, 23323, 39806, 1231, 21185, 8833, 1231, 35276, 114, 3757, 223, 8987, 237, 11319, 1541, 11673, 240, 20485, 1543, 44704, 11957, 253, 11284, 2412, 7627, 14496, 1546, 20149, 227, 13486, 30, 5562, 16786, 40656, 39752, 16233, 5975, 39300, 1546, 11673, 240, 20485, 38714, 7182, 2347, 235, 18464, 117, 18937, 47402, 1543, 51432 ], "temperature": 0, "avg_logprob": -0.2151776772958261, "compression_ratio": 1.0931174089068827, "no_speech_prob": 8.666719251981814e-12, "source": "whisper", "text_original": "Agent 将彻底改变了。一含命令,十个核心玩法。我的Deepseek Harness已经变成了另外一个Codex。一切接插件,模型、工具、界面、状态栏全是插件。究竟它有怎样的魅力?这期视频当中所有的插件全部都免费开源。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 1, "seek": 2134, "start": 21.34, "end": 30.7, "text": "你能够用的安装说明。我们先来看安装。", "tokens": [ 50365, 2166, 8225, 1787, 253, 9254, 1546, 16206, 8083, 227, 8090, 11100, 1543, 2166, 8225, 1787, 253, 9254, 1546, 16206, 8083, 227, 8090, 11100, 1543, 15003, 10108, 6912, 4200, 16206, 8083, 227, 1543, 50833 ], "temperature": 0, "avg_logprob": -0.402961341133953, "compression_ratio": 1.2717391304347827, "no_speech_prob": 8.909898860376408e-12, "source": "whisper", "text_original": "你能够用的安装说明。你能够用的安装说明。我们先来看安装。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 2, "seek": 2134, "start": 30.7, "end": 40.1, "text": "关于安装，官方提供了一个 NPX 命令，可以一键启动我们的 Deepseek Harness。前提是你要安装好 Node.js。", "tokens": [ 50833, 28053, 37732, 16206, 8083, 227, 11, 31929, 9249, 20949, 3254, 249, 2289, 20182, 45, 47, 55, 220, 25236, 49061, 11, 6723, 2257, 23049, 106, 2392, 107, 34961, 15003, 1546, 11089, 595, 405, 916, 389, 16937, 1543, 8945, 20949, 1541, 32085, 16206, 8083, 227, 2131, 49, 378, 13, 25530, 1543, 51303 ], "temperature": 0, "avg_logprob": -0.402961341133953, "compression_ratio": 1.2717391304347827, "no_speech_prob": 8.909898860376408e-12, "source": "whisper", "text_original": "关于安装,官方提供了一个NPX 命令,可以一键启动我们的Deepseek Harness。前提是你要安装好Rod.js。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 3, "seek": 2134, "start": 40.1, "end": 50.5, "text": "如果你使用的是 Windows 系统，它可能会出错。遇到这种情况，你可以先用 NPM 全局安装它，然后再用 DSH 命令来启动。", "tokens": [ 51303, 45669, 22982, 9254, 24620, 45813, 1509, 25368, 10115, 253, 11, 11284, 16657, 12949, 7781, 29900, 1543, 3330, 229, 4511, 5562, 39810, 46514, 11, 42766, 10108, 9254, 45, 18819, 11319, 34703, 16206, 8083, 227, 11284, 11, 26636, 8623, 9254, 35, 17308, 220, 25236, 49061, 6912, 2392, 107, 34961, 1543, 51823 ], "temperature": 0, "avg_logprob": -0.402961341133953, "compression_ratio": 1.2717391304347827, "no_speech_prob": 8.909898860376408e-12, "source": "whisper", "text_original": "如果你使用的是Windows系统,它可能会出错。遇到这种情况,你可以先用NPM全局安装它,然后再用DSH 命令来启动。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 4, "seek": 5050, "start": 50.5, "end": 60.42, "text": "这样就可以解决了。如果你用的是官方的默认版本，大概率觉得只不过是又一个玩具，很粗糙，很多功能都没有。", "tokens": [ 50365, 21209, 36781, 17278, 5676, 111, 2289, 1543, 45669, 9254, 24620, 31929, 9249, 1546, 6173, 246, 7422, 97, 42096, 8802, 11, 32044, 44866, 29992, 14003, 1960, 16866, 1541, 17047, 20182, 19912, 39806, 11, 4563, 18393, 245, 15239, 247, 11, 20778, 32311, 8225, 7182, 17944, 1543, 50861 ], "temperature": 0, "avg_logprob": -0.16052622714285122, "compression_ratio": 1.1977186311787071, "no_speech_prob": 1.3086955998053984e-11, "source": "whisper", "text_original": "这样就可以解决了。如果你用的是官方的默认版本,大概率觉得只不过是又一个玩具,很粗糙,很多功能都没有。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 5, "seek": 5050, "start": 62.0, "end": 74.98, "text": "你先去看看官方的 GitHub，已经有 100 多 K 的 STAR。如果你去搜索插件，仓库已经有 300 个。GitHub 的仓库散落在各个地方，琳琅满目，无从下手。", "tokens": [ 50940, 2166, 10108, 6734, 22884, 31929, 9249, 1546, 38, 270, 21150, 11, 49161, 2412, 6879, 6392, 42, 1546, 6840, 1899, 1543, 45669, 6734, 162, 14987, 7732, 95, 11673, 240, 20485, 11, 1550, 241, 6346, 241, 49161, 2412, 12566, 7549, 1543, 38, 270, 21150, 1546, 1550, 241, 6346, 241, 8259, 103, 30848, 3581, 17516, 7549, 30146, 11, 5419, 230, 8987, 237, 15868, 94, 11386, 11, 31364, 35630, 4438, 11389, 1543, 51589 ], "temperature": 0, "avg_logprob": -0.16052622714285122, "compression_ratio": 1.1977186311787071, "no_speech_prob": 1.3086955998053984e-11, "source": "whisper", "text_original": "你先去看看官方的GitHub,已经有100多K的STAR。如果你去搜索插件,仓库已经有300个。GitHub的仓库闪落在各个地方,盈栏满目,无从下手。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 6, "seek": 7498, "start": 74.98, "end": 82.9, "text": "所以我们首先要装一个插件市场。其实非常简单，直接打开命令行，用 DSH 的插件命令安装几个。", "tokens": [ 50365, 7239, 15003, 36490, 4275, 8083, 227, 20182, 11673, 240, 20485, 27261, 50255, 1543, 41646, 14392, 11249, 222, 47446, 11, 43297, 12467, 18937, 25236, 49061, 8082, 11, 9254, 35, 17308, 27949, 11673, 240, 20485, 25236, 49061, 16206, 8083, 227, 6336, 254, 7549, 1543, 50761 ], "temperature": 0, "avg_logprob": -0.16530521243226295, "compression_ratio": 1.3033175355450237, "no_speech_prob": 1.2839694585320416e-11, "source": "whisper", "text_original": "所以我们首先要装一个插件市场。其实非常简单,直接打开命令行,用DSH 的插件命令安装几个。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 7, "seek": 7498, "start": 82.9, "end": 90.98, "text": "安装完以后，我们重启一下。在设置里面，我们就可以找到插件。安装完插件以后，界面会多出一个插件市场的 Tab 栏。", "tokens": [ 50761, 16206, 8083, 227, 14128, 3588, 13547, 11, 15003, 12624, 2392, 107, 8861, 1543, 3581, 7422, 122, 34719, 15759, 8833, 11, 15003, 36781, 25085, 4511, 11673, 240, 20485, 1543, 16206, 8083, 227, 14128, 11673, 240, 20485, 3588, 13547, 11, 21185, 8833, 12949, 6392, 7781, 20182, 11673, 240, 20485, 27261, 50255, 1546, 51, 712, 294, 1543, 51165 ], "temperature": 0, "avg_logprob": -0.16530521243226295, "compression_ratio": 1.3033175355450237, "no_speech_prob": 1.2839694585320416e-11, "source": "whisper", "text_original": "安装完以后,我们重启一下。在设置里面,我们就可以找到插件。安装完插件以后,界面会多出一个插件市场的Table in。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 8, "seek": 9098, "start": 90.98, "end": 104.98, "text": "在这里面，我们就可以搜索你任意想要的任何插件的关键字进行查询。查询的结果，我们可以看到 STAR。你可以点具体某一个仓库，就会跳到 GitHub 上面去看它的详情。", "tokens": [ 50365, 3581, 35102, 8833, 11, 15003, 36781, 162, 14987, 7732, 95, 2166, 26443, 9042, 7093, 4275, 1546, 26443, 11906, 11673, 240, 20485, 1546, 28053, 23049, 106, 22381, 36700, 8082, 42623, 5233, 95, 1543, 42623, 5233, 95, 1546, 45641, 9319, 11, 15003, 6723, 18032, 11, 14852, 18032, 6840, 1899, 1543, 42766, 12579, 39806, 29485, 17238, 238, 20182, 1550, 241, 6346, 241, 11, 3111, 12949, 29366, 4511, 38, 270, 21150, 49750, 6734, 4200, 45224, 5233, 99, 10570, 1543, 51065 ], "temperature": 0, "avg_logprob": -0.09634552726262732, "compression_ratio": 1.1391752577319587, "no_speech_prob": 1.3859691172513067e-11, "source": "whisper", "text_original": "在这里面,我们就可以搜索你任意想要的任何插件的关键字进行查询。查询的结果,我们可以看到,还看到STAR。你可以点具体某一个仓库,就会跳到GitHub上面去看它的详情。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 9, "seek": 10498, "start": 104.98, "end": 113.98, "text": "重要的是，它可以直接安装。安装完以后，再重启一下就可以。在使用 Agent 的时候，其实大家平时用的最多的就是上下文。", "tokens": [ 50365, 24928, 24620, 11, 11284, 6723, 43297, 16206, 8083, 227, 1543, 16206, 8083, 227, 14128, 3588, 13547, 11, 8623, 12624, 2392, 107, 8861, 36781, 1543, 3581, 22982, 9254, 32, 6930, 27949, 29111, 11, 41646, 6868, 16716, 15729, 9254, 1546, 8661, 6392, 1546, 5620, 5708, 4438, 17174, 1543, 50815 ], "temperature": 0, "avg_logprob": -0.15549900220788043, "compression_ratio": 1.2535211267605635, "no_speech_prob": 1.2941155959345085e-11, "source": "whisper", "text_original": "重要的是,它可以直接安装。安装完以后,再重启一下就可以。在使用Agent 的时候,其实大家平时用的最多的就是上下文。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 10, "seek": 10498, "start": 113.98, "end": 124.98, "text": "你比如说要给它看一个文件或者一个目录，你需要去 Ata 去指定。但它默认是不支持。所以我们首先要在刚刚的插件市场里面去找到一个 Add File 这个插件。", "tokens": [ 50815, 2166, 36757, 8090, 4275, 23197, 11284, 4200, 20182, 17174, 20485, 31148, 20182, 11386, 7391, 243, 11, 2166, 35748, 6734, 32, 1328, 220, 6734, 25922, 12088, 1543, 8395, 11284, 6173, 246, 7422, 97, 1541, 1960, 35488, 1543, 7239, 15003, 36490, 4275, 3581, 49160, 49160, 1546, 11673, 240, 20485, 27261, 50255, 15759, 8833, 6734, 25085, 4511, 20182, 36189, 25315, 220, 15368, 11673, 240, 20485, 1543, 51365 ], "temperature": 0, "avg_logprob": -0.15549900220788043, "compression_ratio": 1.2535211267605635, "no_speech_prob": 1.2941155959345085e-11, "source": "whisper", "text_original": "你比如说要给它看一个文件或者一个目录,你需要去Ata 去指定。但它默认是不支持。所以我们首先要在刚刚的插件市场里面去找到一个Add Fill 这个插件。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 11, "seek": 12498, "start": 124.98, "end": 136.98000000000002, "text": "然后直接安装它。安装好以后，我们重启。当然也可以用命令来安装。接着我们在输入 Ata 的时候，它就能够选择我们当前工作目录下的文件和文件夹了。", "tokens": [ 50365, 26636, 43297, 16206, 8083, 227, 11284, 1543, 16206, 8083, 227, 2131, 3588, 13547, 11, 15003, 12624, 2392, 107, 1543, 40486, 47935, 9254, 25236, 49061, 6912, 16206, 8083, 227, 1543, 14468, 20708, 15003, 3581, 9830, 241, 14028, 32, 1328, 27949, 29111, 11, 11284, 3111, 8225, 1787, 253, 2215, 231, 6852, 102, 15003, 47446, 8945, 41315, 11386, 7391, 243, 4438, 1546, 17174, 20485, 12565, 17174, 20485, 1787, 117, 2289, 1543, 50965 ], "temperature": 0, "avg_logprob": -0.12224957766948931, "compression_ratio": 1.3778501628664495, "no_speech_prob": 1.6186307502663588e-11, "source": "whisper", "text_original": "然后直接安装它。安装好以后,我们重启。当然也可以用命令来安装。接着我们在输入Ata 的时候,它就能够选择我们单前工作目录下的文件和文件夹了。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 12, "seek": 12498, "start": 137.98000000000002, "end": 149.98000000000002, "text": "接下来我安装了一个任务看板。无论是 Codex 还是 Workbox 都是没有的。但我特别喜欢这个功能。我们直接用命令安装它，直接用 GitHub 的形式来安装。安装完以后我们也要重启。", "tokens": [ 51015, 14468, 4438, 6912, 1654, 16206, 8083, 227, 2289, 20182, 26443, 5087, 94, 4200, 43664, 1543, 31364, 7422, 118, 1541, 34, 1429, 87, 220, 45726, 28846, 1067, 22796, 17944, 1546, 1543, 8395, 1654, 17682, 18453, 41606, 15368, 32311, 8225, 1543, 15003, 43297, 9254, 25236, 49061, 16206, 8083, 227, 11284, 11, 43297, 9254, 38, 270, 21150, 1546, 30900, 27584, 6912, 16206, 8083, 227, 1543, 16206, 8083, 227, 14128, 3588, 13547, 15003, 6404, 4275, 12624, 2392, 107, 1543, 51615 ], "temperature": 0, "avg_logprob": -0.12224957766948931, "compression_ratio": 1.3778501628664495, "no_speech_prob": 1.6186307502663588e-11, "source": "whisper", "text_original": "接下来我安装了一个任务看板。无论是Codex 还是Workbody都是没有的。但我特别喜欢这个功能。我们直接用命令安装它,直接用GitHub的形式来安装。安装完以后我们也要重启。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 13, "seek": 14998, "start": 149.98, "end": 160.98, "text": "我们在左边任意点击一个项目里面的一个项目。我们看到右上角就会有一个任务看板。点开就可以看到它的样子。", "tokens": [ 50365, 15003, 3581, 21975, 40503, 26443, 9042, 12579, 6336, 119, 20182, 10178, 117, 11386, 15759, 8833, 1546, 20182, 10115, 246, 27126, 1543, 15003, 18032, 22060, 5708, 29389, 3111, 12949, 2412, 20182, 26443, 5087, 94, 4200, 43664, 1543, 12579, 18937, 36781, 18032, 45224, 14496, 7626, 1543, 50915 ], "temperature": 0, "avg_logprob": -0.09876503663904526, "compression_ratio": 1.396153846153846, "no_speech_prob": 1.3930964888608788e-11, "source": "whisper", "text_original": "我们在左边任意点击一个项目里面的一个绘画。我们看到右上角就会有一个任务看板。点开就可以看到它的样子。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 14, "seek": 14998, "start": 160.98, "end": 175.98, "text": "里面我们可以切换不同的项目。切换以后我们可以直接在这里发送需求。然后等审核以后拉到待办区里面。Agent 就可以自动工作了。而且你还可以设定定时任务。", "tokens": [ 50915, 15759, 8833, 15003, 6723, 23632, 26075, 95, 47123, 1546, 10178, 117, 11386, 1543, 23632, 26075, 95, 3588, 13547, 15003, 6723, 43297, 3581, 35102, 28926, 29309, 32535, 32718, 1543, 26636, 10187, 2415, 94, 8987, 116, 3588, 13547, 29717, 4511, 4845, 99, 12501, 228, 9937, 118, 15759, 8833, 1543, 32, 6930, 32609, 6723, 9722, 34961, 41315, 2289, 1543, 22942, 2166, 14852, 6723, 7422, 122, 12088, 12088, 15729, 26443, 5087, 94, 1543, 51665 ], "temperature": 0, "avg_logprob": -0.09876503663904526, "compression_ratio": 1.396153846153846, "no_speech_prob": 1.3930964888608788e-11, "source": "whisper", "text_original": "里面我们可以切换不同的项目。切换以后我们可以直接在这里发送需求。然后等审核以后拉到带领区里面。Agent 就可以自动工作了。而且你还可以设定定时任务。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 15, "seek": 17598, "start": 175.98, "end": 187.98, "text": "运行的时候相应的任务也会显示在左边的目录里面。任务是广注了。但还有一个硬伤。Deepseek 是文字模型。它不是多模态。我们用两种方式来解决。", "tokens": [ 50365, 3316, 238, 8082, 49873, 15106, 44297, 1546, 26443, 5087, 94, 6404, 12949, 1431, 122, 25696, 3581, 21975, 40503, 1546, 11386, 7391, 243, 15759, 8833, 1543, 26443, 5087, 94, 1541, 3509, 123, 26432, 2289, 1543, 8395, 35091, 20182, 38114, 105, 45581, 1543, 11089, 595, 405, 916, 11947, 17174, 22381, 41908, 39823, 1543, 11284, 7296, 6392, 6656, 6847, 108, 1543, 15003, 9254, 36257, 39810, 9249, 27584, 6912, 17278, 5676, 111, 1543, 50965 ], "temperature": 0, "avg_logprob": -0.17400428131743745, "compression_ratio": 1.2523076923076923, "no_speech_prob": 1.3849948965471981e-11, "source": "whisper", "text_original": "运行的时候相应的任务也会显示在左边的目录里面。任务是广注了。但还有一个硬商。Deepseek 是文字模型。它不是多么泰。我们用两种方式来解决。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 16, "seek": 17598, "start": 187.98, "end": 199.98, "text": "第一种就是网上用的最多的就是 Multimodal。我们直接一个命令就可以安装好这个插件。重启以后，如果你是选择 Deepseek Flush，系统还是会提示你模型不支持这个图像。", "tokens": [ 50965, 18049, 39810, 5620, 16469, 239, 5708, 9254, 1546, 8661, 6392, 1546, 5620, 44, 378, 1248, 719, 1543, 15003, 43297, 20182, 25236, 49061, 36781, 16206, 8083, 227, 2131, 15368, 11673, 240, 20485, 1543, 12624, 2392, 107, 3588, 13547, 11, 13119, 32526, 2215, 231, 6852, 102, 11089, 595, 405, 916, 3235, 1498, 11, 25368, 10115, 253, 45726, 12949, 20949, 25696, 2166, 41908, 39823, 1960, 35488, 15368, 3919, 122, 12760, 1543, 51565 ], "temperature": 0, "avg_logprob": -0.17400428131743745, "compression_ratio": 1.2523076923076923, "no_speech_prob": 1.3849948965471981e-11, "source": "whisper", "text_original": "第一种就是网上用的最多的就是Modernance。我们直接一个命令就可以安装好这个插件。重启以后,如果你是选择Deepseek Flush,系统还是会提示你模型不支持这个图像。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 17, "seek": 19998, "start": 199.98, "end": 217.98, "text": "你需要选择。你需要选择我们刚刚安装好这个模型以后，会多出来的两个模型。Deepseek Plus 和 Pro 后面都多了一个 Multimodal。你选择这两个模型以后，那它就会自动识别这个图形。就不用再会提示你不支持图像了。", "tokens": [ 50365, 2166, 35748, 2215, 231, 6852, 102, 1543, 2166, 35748, 2215, 231, 6852, 102, 15003, 49160, 49160, 16206, 8083, 227, 2131, 15368, 41908, 39823, 3588, 13547, 11, 12949, 6392, 44561, 1546, 36257, 7549, 41908, 39823, 1543, 11089, 595, 405, 916, 3235, 1498, 12565, 12681, 13547, 8833, 7182, 6392, 2289, 20182, 44, 378, 1248, 719, 1543, 2166, 2215, 231, 6852, 102, 5562, 36257, 7549, 41908, 39823, 3588, 13547, 11, 4184, 11284, 3111, 12949, 9722, 34961, 5233, 228, 18453, 15368, 3919, 122, 30900, 1543, 3111, 24384, 8623, 12949, 20949, 25696, 2166, 1960, 35488, 3919, 122, 12760, 2289, 1543, 51265 ], "temperature": 0, "avg_logprob": -0.10922387633660827, "compression_ratio": 1.2035398230088497, "no_speech_prob": 1.7945707420086165e-11, "source": "whisper", "text_original": "你需要选择。你需要选择我们刚刚安装好这个模型以后,会多出来的两个模型。Deepseek Flush和Pro后面都多了一个Modernance。你选择这两个模型以后,那它就会自动识别这个图形。就不用再会提示你不支持图像了。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 18, "seek": 21798, "start": 217.98, "end": 230.98, "text": "第二种方式，我们装另外一个插件。它可以指定你图像识别的模型。这样的话，你可以指定任意你想要的图像模型来识别。你只需要安装它的 SDK 就可以了。", "tokens": [ 50365, 19693, 39810, 9249, 27584, 11, 15003, 25153, 119, 26202, 20182, 11673, 240, 20485, 1543, 11284, 6723, 25922, 12088, 2166, 3919, 122, 12760, 5233, 228, 18453, 1546, 41908, 39823, 1543, 21209, 44575, 11, 42766, 25922, 12088, 26443, 9042, 38386, 4275, 1546, 3919, 122, 12760, 41908, 39823, 6912, 5233, 228, 18453, 1543, 2166, 14003, 35748, 16206, 8083, 227, 45224, 42, 2777, 36781, 2289, 1543, 51015 ], "temperature": 0, "avg_logprob": -0.10318417982621626, "compression_ratio": 1.4775510204081632, "no_speech_prob": 1.4012991288170351e-11, "source": "whisper", "text_original": "第二种方式,我们钻另外一个插件。它可以指定你图像识别的模型。这样的话,你可以指定任意你想要的图像模型来识别。你只需要安装它的KID就可以了。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 19, "seek": 21798, "start": 230.98, "end": 240.98, "text": "同样的，安装好以后，我们在设置里面，你就可以看到它多了一个菜单。这里面你就可以选择或者是更换你的图像识别的模型。", "tokens": [ 51015, 13089, 14496, 1546, 11, 16206, 8083, 227, 2131, 3588, 13547, 11, 15003, 3581, 7422, 122, 34719, 15759, 8833, 11, 2166, 36781, 18032, 11284, 6392, 2289, 20182, 33025, 47446, 1543, 35102, 8833, 2166, 36781, 2215, 231, 6852, 102, 31148, 1541, 19002, 26075, 95, 18961, 3919, 122, 12760, 5233, 228, 18453, 1546, 41908, 39823, 1543, 51515 ], "temperature": 0, "avg_logprob": -0.10318417982621626, "compression_ratio": 1.4775510204081632, "no_speech_prob": 1.4012991288170351e-11, "source": "whisper", "text_original": "同样的,安装好以后,我们在设置里面,你就可以看到它多了一个菜单。这里面你就可以选择或者是更换你的图像识别的模型。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 20, "seek": 24098, "start": 240.98, "end": 250.98, "text": "这样你在即便是选择 Deepseek Plus 的时候，它也能够自动调取图像识别工具，进行图像识别了。", "tokens": [ 50365, 21209, 41410, 39127, 27364, 1541, 2215, 231, 6852, 102, 11089, 595, 405, 916, 3235, 1498, 49873, 11, 11284, 6404, 8225, 1787, 253, 9722, 34961, 8897, 225, 29436, 3919, 122, 12760, 5233, 228, 18453, 23323, 39806, 11, 36700, 8082, 3919, 122, 12760, 5233, 228, 18453, 2289, 1543, 50865 ], "temperature": 0, "avg_logprob": -0.10119637576016513, "compression_ratio": 1.131578947368421, "no_speech_prob": 1.4476928336670802e-11, "source": "whisper", "text_original": "这样你在即便是选择Deepseek Flush的时候,它也能够自动调取图像识别工具,进行图像识别了。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 21, "seek": 24098, "start": 251.98, "end": 259.98, "text": "如果你想要发送文件，或者是拖出图片进行聊天，却发现无法拖动。这里我们需要安装一个文件拖动的功能。", "tokens": [ 50915, 13119, 38386, 4275, 28926, 29309, 17174, 20485, 11, 31148, 1541, 6852, 244, 7781, 3919, 122, 16668, 36700, 8082, 40096, 6135, 11, 5322, 112, 28926, 20204, 31364, 11148, 6852, 244, 34961, 1543, 35102, 15003, 35748, 16206, 8083, 227, 20182, 17174, 20485, 6852, 244, 34961, 1546, 32311, 8225, 1543, 51315 ], "temperature": 0, "avg_logprob": -0.10119637576016513, "compression_ratio": 1.131578947368421, "no_speech_prob": 1.4476928336670802e-11, "source": "whisper", "text_original": "如果你想要发送文件,或者是拖出图片进行聊天,却发现无法拖动。这里我们需要安装一个文件拖动的功能。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 22, "seek": 25998, "start": 259.98, "end": 269.98, "text": "我们直接用插件命令进行安装。安装好以后，我们就可以在电脑的任何地方把你想要拉入的文件或者是图片拉入进来。", "tokens": [ 50365, 15003, 43297, 9254, 11673, 240, 20485, 25236, 49061, 36700, 8082, 16206, 8083, 227, 1543, 16206, 8083, 227, 2131, 3588, 13547, 11, 15003, 36781, 3581, 42182, 27067, 239, 1546, 26443, 11906, 30146, 16075, 38386, 4275, 29717, 14028, 1546, 17174, 20485, 31148, 1541, 3919, 122, 16668, 29717, 14028, 36700, 6912, 1543, 50865 ], "temperature": 0, "avg_logprob": -0.1161343138894917, "compression_ratio": 1.3033033033033032, "no_speech_prob": 1.453310822380205e-11, "source": "whisper", "text_original": "我们直接用插件命令进行安装。安装好以后,我们就可以在电脑的任何地方把你想要拉入的文件或者是图片拉入进来。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 23, "seek": 25998, "start": 269.98, "end": 275.98, "text": "拖入以后，它会自动翻转输入框，并且带有你的文件路径，非常方便。", "tokens": [ 50865, 6852, 244, 14028, 3588, 13547, 11, 11284, 12949, 9722, 34961, 42716, 17819, 105, 9830, 241, 14028, 21416, 228, 11, 3509, 114, 20334, 4845, 99, 2412, 18961, 17174, 20485, 24658, 2172, 226, 11, 14392, 9249, 27364, 1543, 51165 ], "temperature": 0, "avg_logprob": -0.1161343138894917, "compression_ratio": 1.3033033033033032, "no_speech_prob": 1.453310822380205e-11, "source": "whisper", "text_original": "拖入以后,它会自动翻转输入框,并且带有你的文件路径,非常方便。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 24, "seek": 25998, "start": 275.98, "end": 281.98, "text": "我们发现，在输出的对话框里面的这些文件，我们点击它是没有反应的，无法打开。", "tokens": [ 51165, 15003, 28926, 20204, 11, 3581, 9830, 241, 7781, 1546, 8713, 21596, 21416, 228, 15759, 8833, 1546, 5562, 13824, 17174, 20485, 11, 15003, 12579, 6336, 119, 11284, 1541, 17944, 22138, 44297, 1546, 11, 31364, 11148, 12467, 18937, 1543, 51465 ], "temperature": 0, "avg_logprob": -0.1161343138894917, "compression_ratio": 1.3033033033033032, "no_speech_prob": 1.453310822380205e-11, "source": "whisper", "text_original": "我们发现,在输出的对话框里面的这些文件,我们点击它是没有反应的,无法打开。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 25, "seek": 25998, "start": 281.98, "end": 287.98, "text": "实际上这个时候我们像 Codex 或者 Workbox 一样需要一个类似的侧边栏。", "tokens": [ 51465, 24726, 8842, 227, 5708, 15368, 29111, 15003, 12760, 34, 1429, 87, 31148, 28846, 1067, 2257, 14496, 35748, 20182, 22113, 119, 7384, 120, 1546, 3254, 100, 38109, 244, 8987, 237, 1543, 51765 ], "temperature": 0, "avg_logprob": -0.1161343138894917, "compression_ratio": 1.3033033033033032, "no_speech_prob": 1.453310822380205e-11, "source": "whisper", "text_original": "实际上这个时候我们像Codex或者Workbody一样需要一个类似的侧编栏。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 26, "seek": 28798, "start": 287.98, "end": 293.98, "text": "我们可以在右侧的侧边栏，点击文件就可以进行预览，或者是查看目录。", "tokens": [ 50365, 15003, 6723, 3581, 22060, 3254, 100, 1546, 3254, 100, 40503, 8987, 237, 11, 12579, 6336, 119, 17174, 20485, 36781, 36700, 8082, 12501, 226, 6758, 230, 11, 31148, 1541, 42623, 4200, 11386, 7391, 243, 1543, 50665 ], "temperature": 0, "avg_logprob": -0.1635475582546658, "compression_ratio": 1.4040816326530612, "no_speech_prob": 1.7023467704935236e-11, "source": "whisper", "text_original": "我们可以在右侧的侧边栏,点击文件就可以进行预览,或者是查看目录。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 27, "seek": 28798, "start": 293.98, "end": 298.98, "text": "安装好以后，我们就可以在右侧看到工作目录下的所有文件。", "tokens": [ 50665, 16206, 8083, 227, 2131, 3588, 13547, 11, 15003, 36781, 3581, 22060, 3254, 100, 18032, 41315, 11386, 7391, 243, 4438, 1546, 39300, 17174, 20485, 1543, 50915 ], "temperature": 0, "avg_logprob": -0.1635475582546658, "compression_ratio": 1.4040816326530612, "no_speech_prob": 1.7023467704935236e-11, "source": "whisper", "text_original": "安装好以后,我们就可以在右侧看到工作目录下的所有文件。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 28, "seek": 28798, "start": 298.98, "end": 305.98, "text": "点击文件以后，像 MD 或者 HTML 这样的一些文件，我们就可以在右边进行浏览了。", "tokens": [ 50915, 12579, 6336, 119, 17174, 20485, 3588, 13547, 11, 12760, 44, 35, 31148, 39, 51, 12683, 21209, 1546, 38515, 17174, 20485, 11, 15003, 36781, 3581, 22060, 40503, 36700, 8082, 11038, 237, 6758, 230, 2289, 1543, 51265 ], "temperature": 0, "avg_logprob": -0.1635475582546658, "compression_ratio": 1.4040816326530612, "no_speech_prob": 1.7023467704935236e-11, "source": "whisper", "text_original": "点击文件以后,像MD或者HTML这样的一些文件,我们就可以在右边进行浏览了。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 29, "seek": 28798, "start": 305.98, "end": 311.98, "text": "除了 MD 和浏览以外，它还提供了终端和源代码的管理功能。", "tokens": [ 51265, 32999, 2289, 44, 35, 12565, 11038, 237, 6758, 230, 3588, 12022, 11, 11284, 14852, 20949, 3254, 249, 2289, 10115, 230, 11957, 107, 12565, 19683, 19105, 23230, 223, 1546, 23131, 13876, 32311, 8225, 1543, 51565 ], "temperature": 0, "avg_logprob": -0.1635475582546658, "compression_ratio": 1.4040816326530612, "no_speech_prob": 1.7023467704935236e-11, "source": "whisper", "text_original": "除了MD和浏览以外,它还提供了终端和原代码的管理功能。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 30, "seek": 31198, "start": 311.98, "end": 319.98, "text": "而且还能看到多个 Agent 的列表，点击 Agent 还能看到它具体的日志，非常方便。", "tokens": [ 50365, 22942, 14852, 8225, 18032, 6392, 7549, 32, 6930, 1546, 43338, 17571, 11, 12579, 6336, 119, 32, 6930, 14852, 8225, 18032, 11284, 39806, 29485, 1546, 6890, 7626, 11, 14392, 9249, 27364, 1543, 50765 ], "temperature": 0, "avg_logprob": -0.11553039936104206, "compression_ratio": 1.2, "no_speech_prob": 1.4684833210543147e-11, "source": "whisper", "text_original": "而且还能看到多个Agent的列表,点击Agent还能看到它具体的日子,非常方便。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 31, "seek": 31198, "start": 319.98, "end": 323.98, "text": "但它的这个终端，我们要给它禁掉，因为它有一点问题。", "tokens": [ 50765, 8395, 45224, 15368, 10115, 230, 11957, 107, 11, 15003, 4275, 23197, 11284, 18614, 223, 29327, 11, 34627, 11284, 32241, 12579, 34069, 1543, 50965 ], "temperature": 0, "avg_logprob": -0.11553039936104206, "compression_ratio": 1.2, "no_speech_prob": 1.4684833210543147e-11, "source": "whisper", "text_original": "但它的这个终端,我们要给它禁掉,因为它有一点问题。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 32, "seek": 31198, "start": 323.98, "end": 326.98, "text": "我们待会儿重新装一个新的终端。", "tokens": [ 50965, 15003, 18390, 12949, 22933, 12624, 12560, 8083, 227, 20182, 12560, 1546, 10115, 230, 11957, 107, 1543, 51115 ], "temperature": 0, "avg_logprob": -0.11553039936104206, "compression_ratio": 1.2, "no_speech_prob": 1.4684833210543147e-11, "source": "whisper" }, { "id": 33, "seek": 31198, "start": 326.98, "end": 331.98, "text": "如果你要使用 MCP，那也需要一个可视化的 MCP 管理。", "tokens": [ 51115, 13119, 32085, 22982, 9254, 44, 20049, 11, 4184, 6404, 35748, 20182, 4429, 40656, 23756, 1546, 44, 20049, 23131, 13876, 1543, 51365 ], "temperature": 0, "avg_logprob": -0.11553039936104206, "compression_ratio": 1.2, "no_speech_prob": 1.4684833210543147e-11, "source": "whisper", "text_original": "如果你要使用MCP,那也需要一个可视化的MCP管理。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 34, "seek": 33198, "start": 331.98, "end": 339.98, "text": "我们这里就装一个 MCP 的插件。这样的话，你就可以在设置的菜单里面找到 MCP 的菜单。", "tokens": [ 50365, 15003, 35102, 3111, 8083, 227, 20182, 44, 20049, 1546, 11673, 240, 20485, 1543, 21209, 44575, 11, 2166, 36781, 3581, 7422, 122, 34719, 1546, 33025, 47446, 15759, 8833, 25085, 4511, 44, 20049, 1546, 33025, 47446, 1543, 50765 ], "temperature": 0, "avg_logprob": -0.09470260959782012, "compression_ratio": 1.4214285714285715, "no_speech_prob": 1.3449632900452979e-11, "source": "whisper", "text_original": "我们这里就装一个MCP的插件。这样的话,你就可以在设置的菜单里面找到MCP的菜单。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 35, "seek": 33198, "start": 339.98, "end": 344.98, "text": "你就可以手动在页面上进行管理你的 MCP 了。不用再去街市里面修改。", "tokens": [ 50765, 2166, 36781, 11389, 34961, 3581, 10178, 113, 8833, 5708, 36700, 8082, 23131, 13876, 18961, 44, 20049, 2289, 1543, 24384, 8623, 6734, 42529, 27261, 15759, 8833, 7792, 106, 34490, 1543, 51015 ], "temperature": 0, "avg_logprob": -0.09470260959782012, "compression_ratio": 1.4214285714285715, "no_speech_prob": 1.3449632900452979e-11, "source": "whisper", "text_original": "你就可以手动在页面上进行管理你的MCP了。不用再去街市里面修改。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 36, "seek": 33198, "start": 344.98, "end": 350.98, "text": "接下来我们看到在整个过程里面，整个页面显得不是那么灵动。", "tokens": [ 51015, 14468, 4438, 6912, 15003, 18032, 3581, 27662, 7549, 16866, 29649, 15759, 8833, 11, 27662, 7549, 10178, 113, 8833, 1431, 122, 5916, 7296, 35693, 13290, 113, 34961, 1543, 51315 ], "temperature": 0, "avg_logprob": -0.09470260959782012, "compression_ratio": 1.4214285714285715, "no_speech_prob": 1.3449632900452979e-11, "source": "whisper", "text_original": "接下来我们看到在整个过程里面,整个页面显得不是那么灵动。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 37, "seek": 33198, "start": 350.98, "end": 354.98, "text": "状态栏的内容是固定的。我们需要安装一个自定义的状态栏。", "tokens": [ 51315, 35276, 114, 3757, 223, 8987, 237, 1546, 34742, 25750, 1541, 3919, 118, 12088, 1546, 1543, 15003, 35748, 16206, 8083, 227, 20182, 9722, 12088, 2930, 231, 1546, 35276, 114, 3757, 223, 8987, 237, 1543, 51515 ], "temperature": 0, "avg_logprob": -0.09470260959782012, "compression_ratio": 1.4214285714285715, "no_speech_prob": 1.3449632900452979e-11, "source": "whisper" }, { "id": 38, "seek": 33198, "start": 354.98, "end": 356.98, "text": "一个命令就可以给它装好。", "tokens": [ 51515, 20182, 25236, 49061, 36781, 23197, 11284, 8083, 227, 2131, 1543, 51615 ], "temperature": 0, "avg_logprob": -0.09470260959782012, "compression_ratio": 1.4214285714285715, "no_speech_prob": 1.3449632900452979e-11, "source": "whisper" }, { "id": 39, "seek": 35698, "start": 356.98, "end": 363.98, "text": "然后我们看到这里的词语，它会不断的变化。", "tokens": [ 50365, 26636, 15003, 18032, 35102, 1546, 5233, 235, 5233, 255, 11, 11284, 12949, 1960, 4307, 255, 1546, 2129, 246, 23756, 1543, 50715 ], "temperature": 0, "avg_logprob": -0.13110137472347339, "compression_ratio": 1.2396166134185302, "no_speech_prob": 1.6149904330520215e-11, "source": "whisper", "text_original": "然后我们看到这里的词语,它会不断的变化。", "correction_applied": true, "correction_source": "llm_correct_a_presegmentation" }, { "id": 40, "seek": 35698, "start": 363.98, "end": 366.98, "text": "这些词语你可以进行自定义,你可以进行修改。", "tokens": [ 50715, 5562, 13824, 5233, 235, 5233, 255, 42766, 36700, 8082, 9722, 12088, 2930, 231, 11, 42766, 36700, 8082, 7792, 106, 34490, 1543, 50865 ], "temperature": 0, "avg_logprob": -0.13110137472347339, "compression_ratio": 1.2396166134185302, "no_speech_prob": 1.6149904330520215e-11, "source": "whisper" }, { "id": 41, "seek": 35698, "start": 366.98, "end": 371.98, "text": "看起来是不是更加的接近于Codex和Cloud Code了呢?", "tokens": [ 50865, 4200, 9147, 6912, 23034, 19002, 9990, 1546, 14468, 17463, 37732, 34, 1429, 87, 12565, 32787, 15549, 2289, 6240, 30, 51115 ], "temperature": 0, "avg_logprob": -0.13110137472347339, "compression_ratio": 1.2396166134185302, "no_speech_prob": 1.6149904330520215e-11, "source": "whisper" }, { "id": 42, "seek": 35698, "start": 371.98, "end": 376.98, "text": "现在状态是好看的。但是文本输出还是千篇一律全是文字。", "tokens": [ 51115, 25040, 35276, 114, 3757, 223, 1541, 2131, 4200, 1546, 1543, 11189, 17174, 8802, 9830, 241, 7781, 45726, 20787, 20878, 229, 2257, 2172, 233, 11319, 1541, 17174, 22381, 1543, 51365 ], "temperature": 0, "avg_logprob": -0.13110137472347339, "compression_ratio": 1.2396166134185302, "no_speech_prob": 1.6149904330520215e-11, "source": "whisper" }, { "id": 43, "seek": 35698, "start": 376.98, "end": 380.98, "text": "对于Markdown这样的东西,没有办法输出相应的表格,也不够美化。", "tokens": [ 51365, 8713, 37732, 15168, 5093, 21209, 1546, 38409, 16220, 11, 17944, 39453, 11148, 9830, 241, 7781, 15106, 44297, 1546, 17571, 30921, 11, 6404, 1960, 1787, 253, 9175, 23756, 1543, 51565 ], "temperature": 0, "avg_logprob": -0.13110137472347339, "compression_ratio": 1.2396166134185302, "no_speech_prob": 1.6149904330520215e-11, "source": "whisper" }, { "id": 44, "seek": 35698, "start": 380.98, "end": 383.98, "text": "我们需要装一个副文本的显示插件。", "tokens": [ 51565, 15003, 35748, 8083, 227, 20182, 5935, 107, 17174, 8802, 1546, 1431, 122, 25696, 11673, 240, 20485, 1543, 51715 ], "temperature": 0, "avg_logprob": -0.13110137472347339, "compression_ratio": 1.2396166134185302, "no_speech_prob": 1.6149904330520215e-11, "source": "whisper" }, { "id": 45, "seek": 38398, "start": 383.98, "end": 388.98, "text": "这样我们就可以像Mermud这样的流程图,进行可视化了。", "tokens": [ 50365, 21209, 15003, 36781, 12760, 44, 966, 532, 21209, 1546, 27854, 29649, 3919, 122, 11, 36700, 8082, 4429, 40656, 23756, 2289, 1543, 50615 ], "temperature": 0, "avg_logprob": -0.11965421325001636, "compression_ratio": 1.4619883040935673, "no_speech_prob": 1.4197129583060075e-11, "source": "whisper" }, { "id": 46, "seek": 38398, "start": 388.98, "end": 393.98, "text": "同时表格文字它还给你加一些底色,高亮,排版美化等等。", "tokens": [ 50615, 13089, 15729, 17571, 30921, 17174, 22381, 11284, 14852, 23197, 2166, 9990, 38515, 22816, 17673, 11, 12979, 22991, 11, 44647, 42096, 9175, 23756, 36000, 1543, 50865 ], "temperature": 0, "avg_logprob": -0.11965421325001636, "compression_ratio": 1.4619883040935673, "no_speech_prob": 1.4197129583060075e-11, "source": "whisper" }, { "id": 47, "seek": 38398, "start": 393.98, "end": 396.98, "text": "更重要的是它还可以输出图表一样的内容。", "tokens": [ 50865, 19002, 24928, 24620, 11284, 14852, 6723, 9830, 241, 7781, 3919, 122, 17571, 2257, 14496, 1546, 34742, 25750, 1543, 51015 ], "temperature": 0, "avg_logprob": -0.11965421325001636, "compression_ratio": 1.4619883040935673, "no_speech_prob": 1.4197129583060075e-11, "source": "whisper" }, { "id": 48, "seek": 38398, "start": 396.98, "end": 400.98, "text": "我们直接让它分析一个内容,它就可以直接用图表的形式给你呈现。", "tokens": [ 51015, 15003, 43297, 33650, 11284, 6627, 7360, 238, 20182, 34742, 25750, 11, 11284, 36781, 43297, 9254, 3919, 122, 17571, 1546, 30900, 27584, 23197, 2166, 3606, 230, 20204, 1543, 51215 ], "temperature": 0, "avg_logprob": -0.11965421325001636, "compression_ratio": 1.4619883040935673, "no_speech_prob": 1.4197129583060075e-11, "source": "whisper" }, { "id": 49, "seek": 38398, "start": 400.98, "end": 402.98, "text": "这种方式我太喜欢了。", "tokens": [ 51215, 5562, 39810, 9249, 27584, 1654, 9455, 41606, 2289, 1543, 51315 ], "temperature": 0, "avg_logprob": -0.11965421325001636, "compression_ratio": 1.4619883040935673, "no_speech_prob": 1.4197129583060075e-11, "source": "whisper" }, { "id": 50, "seek": 38398, "start": 402.98, "end": 405.98, "text": "最后一个我们把终端给它集成进来。", "tokens": [ 51315, 8661, 13547, 20182, 15003, 16075, 10115, 230, 11957, 107, 23197, 11284, 26020, 11336, 36700, 6912, 1543, 51465 ], "temperature": 0, "avg_logprob": -0.11965421325001636, "compression_ratio": 1.4619883040935673, "no_speech_prob": 1.4197129583060075e-11, "source": "whisper" }, { "id": 51, "seek": 38398, "start": 405.98, "end": 408.98, "text": "刚刚特别难自带的终端有一些问题。", "tokens": [ 51465, 49160, 49160, 17682, 18453, 46531, 9722, 4845, 99, 1546, 10115, 230, 11957, 107, 32241, 13824, 34069, 1543, 51615 ], "temperature": 0, "avg_logprob": -0.11965421325001636, "compression_ratio": 1.4619883040935673, "no_speech_prob": 1.4197129583060075e-11, "source": "whisper" }, { "id": 52, "seek": 38398, "start": 408.98, "end": 412.98, "text": "我们独自安装一个终端,让这个终端也是一个命令就可以安装好。", "tokens": [ 51615, 15003, 18637, 105, 9722, 16206, 8083, 227, 20182, 10115, 230, 11957, 107, 11, 33650, 15368, 10115, 230, 11957, 107, 22021, 20182, 25236, 49061, 36781, 16206, 8083, 227, 2131, 1543, 51815 ], "temperature": 0, "avg_logprob": -0.11965421325001636, "compression_ratio": 1.4619883040935673, "no_speech_prob": 1.4197129583060075e-11, "source": "whisper" }, { "id": 53, "seek": 41298, "start": 412.98, "end": 416.98, "text": "而且安装好以后它就会出现在我们页面下半。", "tokens": [ 50365, 22942, 16206, 8083, 227, 2131, 3588, 13547, 11284, 3111, 12949, 7781, 25040, 15003, 10178, 113, 8833, 4438, 30018, 1543, 50565 ], "temperature": 0, "avg_logprob": -0.11234694719314575, "compression_ratio": 1.244131455399061, "no_speech_prob": 1.5080235671316444e-11, "source": "whisper" }, { "id": 54, "seek": 41298, "start": 416.98, "end": 418.98, "text": "默认状态下它是收起来的。", "tokens": [ 50565, 6173, 246, 7422, 97, 35276, 114, 3757, 223, 4438, 11284, 1541, 18681, 9147, 6912, 1546, 1543, 50665 ], "temperature": 0, "avg_logprob": -0.11234694719314575, "compression_ratio": 1.244131455399061, "no_speech_prob": 1.5080235671316444e-11, "source": "whisper" }, { "id": 55, "seek": 41298, "start": 418.98, "end": 421.98, "text": "点开我们就可以看到,非常方便。", "tokens": [ 50665, 12579, 18937, 15003, 36781, 18032, 11, 14392, 9249, 27364, 1543, 50815 ], "temperature": 0, "avg_logprob": -0.11234694719314575, "compression_ratio": 1.244131455399061, "no_speech_prob": 1.5080235671316444e-11, "source": "whisper" }, { "id": 56, "seek": 41298, "start": 421.98, "end": 425.98, "text": "在上面你可以写命令,跑脚本,都是没有问题的。", "tokens": [ 50815, 3581, 49750, 42766, 5676, 247, 25236, 49061, 11, 32585, 27067, 248, 8802, 11, 22796, 17944, 34069, 1546, 1543, 51015 ], "temperature": 0, "avg_logprob": -0.11234694719314575, "compression_ratio": 1.244131455399061, "no_speech_prob": 1.5080235671316444e-11, "source": "whisper" }, { "id": 57, "seek": 41298, "start": 425.98, "end": 431.98, "text": "上面的这些插件安装完以后,你会发现一件事情。", "tokens": [ 51015, 49750, 1546, 5562, 13824, 11673, 240, 20485, 16206, 8083, 227, 14128, 3588, 13547, 11, 2166, 12949, 28926, 20204, 2257, 49459, 1543, 51315 ], "temperature": 0, "avg_logprob": -0.11234694719314575, "compression_ratio": 1.244131455399061, "no_speech_prob": 1.5080235671316444e-11, "source": "whisper" }, { "id": 58, "seek": 43198, "start": 431.98, "end": 434.98, "text": "DeepseekHarness它不是一个软件。", "tokens": [ 50365, 11089, 595, 405, 916, 39, 16937, 11284, 7296, 20182, 17819, 107, 20485, 1543, 50515 ], "temperature": 0, "avg_logprob": -0.17536012331644693, "compression_ratio": 1.1938775510204083, "no_speech_prob": 1.4288795840977642e-11, "source": "whisper" }, { "id": 59, "seek": 43198, "start": 434.98, "end": 438.98, "text": "它是一个像落高的底座,插件就是它的积木。", "tokens": [ 50515, 11284, 1541, 20182, 12760, 30848, 12979, 1546, 22816, 6346, 100, 11, 11673, 240, 20485, 5620, 45224, 8204, 107, 41437, 1543, 50715 ], "temperature": 0, "avg_logprob": -0.17536012331644693, "compression_ratio": 1.1938775510204083, "no_speech_prob": 1.4288795840977642e-11, "source": "whisper" }, { "id": 60, "seek": 43198, "start": 438.98, "end": 444.98, "text": "现在AI已经从谁的大脑更聪明,变成了谁的身体更好用。", "tokens": [ 50715, 25040, 48698, 49161, 35630, 33556, 1546, 3582, 27067, 239, 19002, 8171, 103, 11100, 11, 2129, 246, 11336, 2289, 33556, 1546, 19847, 29485, 19002, 2131, 9254, 1543, 51015 ], "temperature": 0, "avg_logprob": -0.17536012331644693, "compression_ratio": 1.1938775510204083, "no_speech_prob": 1.4288795840977642e-11, "source": "whisper" }, { "id": 61, "seek": 43198, "start": 444.98, "end": 448.98, "text": "模型决定了Agent多聪明,插件决定了Agent也多能干。", "tokens": [ 51015, 41908, 39823, 5676, 111, 12088, 2289, 32, 6930, 6392, 8171, 103, 11100, 11, 11673, 240, 20485, 5676, 111, 12088, 2289, 32, 6930, 6404, 6392, 8225, 26111, 1543, 51215 ], "temperature": 0, "avg_logprob": -0.17536012331644693, "compression_ratio": 1.1938775510204083, "no_speech_prob": 1.4288795840977642e-11, "source": "whisper" }, { "id": 62, "seek": 44898, "start": 448.98, "end": 452.98, "text": "当然还有很多好玩的插件,也可以自己去专一行。", "tokens": [ 50365, 40486, 35091, 20778, 2131, 19912, 1546, 11673, 240, 20485, 11, 47935, 17645, 6734, 940, 241, 2257, 8082, 1543, 50565 ], "temperature": 0, "avg_logprob": -0.1473502000172933, "compression_ratio": 1.0178571428571428, "no_speech_prob": 1.71493184236704e-11, "source": "whisper" }, { "id": 63, "seek": 44898, "start": 452.98, "end": 458.98, "text": "如果你喜欢,下期我教你如何写一个自己的DeepseekHarness插件。", "tokens": [ 50565, 45669, 41606, 11, 4438, 16786, 1654, 21936, 2166, 43526, 5676, 247, 20182, 17645, 1546, 11089, 595, 405, 916, 39, 16937, 11673, 240, 20485, 1543, 50865 ], "temperature": 0, "avg_logprob": -0.1473502000172933, "compression_ratio": 1.0178571428571428, "no_speech_prob": 1.71493184236704e-11, "source": "whisper" }, { "id": 64, "seek": 44898, "start": 458.98, "end": 460.98, "text": "点个关注,别错过了。", "tokens": [ 50865, 12579, 7549, 28053, 26432, 11, 18453, 29900, 16866, 2289, 1543, 50965 ], "temperature": 0, "avg_logprob": -0.1473502000172933, "compression_ratio": 1.0178571428571428, "no_speech_prob": 1.71493184236704e-11, "source": "whisper" } ], "transcript_selection": { "schema": "videonote.transcript_selection.v1", "status": "SELECTED", "requested_language": "zh", "duration": 461.450188, "reference_path": null, "reference_kind": null, "reference_language": null, "selected_source": "whisper_segmented", "decision_reason": "source_unavailable_or_untrusted_whisper_passed", "whisper_rejected": false, "source_profile": { "segment_count": 0, "duration_coverage": 0.0, "usable_full": false, "usable_partial": false, "hallucination_region_count": 0, "hallucination_regions": [], "language": { "requested": "zh", "matched": false } }, "whisper_profile": { "segment_count": 65, "multilingual": false, "first_start": 0.0, "final_end": 460.98, "duration_coverage": 0.998981, "maximum_gap_seconds": 1.58, "inverted_count": 0, "language": { "requested": "zh", "matched": true, "latin": 329, "han": 2027, "kana": 0, "hangul": 0 }, "hallucination_region_count": 0, "hallucination_regions": [], "usable_full": true, "usable_partial": true }, "comparison": { "window_seconds": 30.0, "compared_window_count": 0, "median_score": 0.0, "divergent_window_count": 0, "divergent_ratio": 0.0, "divergent_windows": [] }, "timestamp_alignment": { "applied": false, "estimated_offset_seconds": 0.0, "matched_segment_count": 0, "match_scores": [] }, "caption_timing_repair": null, "filled_source_profile": null, "whisper_duration_guard": { "media_duration": 461.45, "rejected_segment_count": 14, "rejected_segments": [ { "index": 65, "start": 460.98, "end": 462.98, "reason": "material_end_timestamp_overrun" }, { "index": 66, "start": 462.98, "end": 464.98, "reason": "starts_at_or_after_media_end" }, { "index": 67, "start": 464.98, "end": 466.98, "reason": "starts_at_or_after_media_end" }, { "index": 68, "start": 466.98, "end": 468.98, "reason": "starts_at_or_after_media_end" }, { "index": 69, "start": 468.98, "end": 470.98, "reason": "starts_at_or_after_media_end" }, { "index": 70, "start": 470.98, "end": 472.98, "reason": "starts_at_or_after_media_end" }, { "index": 71, "start": 472.98, "end": 474.98, "reason": "starts_at_or_after_media_end" }, { "index": 72, "start": 474.98, "end": 476.98, "reason": "starts_at_or_after_media_end" }, { "index": 73, "start": 476.98, "end": 478.98, "reason": "starts_at_or_after_media_end" }, { "index": 74, "start": 478.98, "end": 480.98, "reason": "starts_at_or_after_media_end" }, { "index": 75, "start": 480.98, "end": 482.98, "reason": "starts_at_or_after_media_end" }, { "index": 76, "start": 482.98, "end": 484.98, "reason": "starts_at_or_after_media_end" }, { "index": 77, "start": 484.98, "end": 486.98, "reason": "starts_at_or_after_media_end" }, { "index": 78, "start": 486.98, "end": 488.98, "reason": "starts_at_or_after_media_end" } ] } }, "presegmentation_correction": { "applied": true, "correction_count": 37 } }