first build nexus
This commit is contained in:
billyshen
BIN
AI/@eaDir/AI 解决方案专家培训课程.md@SynoEAStream
Normal file
BIN
AI/@eaDir/AI 解决方案专家培训课程.md@SynoEAStream
Normal file
Binary file not shown.
BIN
AI/@eaDir/AI 解决方案专家培训课程.md@SynoResource
Normal file
BIN
AI/@eaDir/AI 解决方案专家培训课程.md@SynoResource
Normal file
Binary file not shown.
BIN
AI/@eaDir/Best 7 news API data feeds - AI News.md@SynoEAStream
Normal file
BIN
AI/@eaDir/Best 7 news API data feeds - AI News.md@SynoEAStream
Normal file
Binary file not shown.
BIN
AI/@eaDir/Best 7 news API data feeds - AI News.md@SynoResource
Normal file
BIN
AI/@eaDir/Best 7 news API data feeds - AI News.md@SynoResource
Normal file
Binary file not shown.
BIN
AI/@eaDir/Designing for Agentic AI.md@SynoEAStream
Normal file
BIN
AI/@eaDir/Designing for Agentic AI.md@SynoEAStream
Normal file
Binary file not shown.
BIN
AI/@eaDir/Designing for Agentic AI.md@SynoResource
Normal file
BIN
AI/@eaDir/Designing for Agentic AI.md@SynoResource
Normal file
Binary file not shown.
BIN
AI/@eaDir/Google 神级生产力工具，所有 GitHub 开源平替都找到了。.md@SynoEAStream
Normal file
BIN
AI/@eaDir/Google 神级生产力工具，所有 GitHub 开源平替都找到了。.md@SynoEAStream
Normal file
Binary file not shown.
BIN
AI/@eaDir/Google 神级生产力工具，所有 GitHub 开源平替都找到了。.md@SynoResource
Normal file
BIN
AI/@eaDir/Google 神级生产力工具，所有 GitHub 开源平替都找到了。.md@SynoResource
Normal file
Binary file not shown.
BIN
AI/@eaDir/LLMs、RAG、AI Agent 三个到底什么区别？.md@SynoEAStream
Normal file
BIN
AI/@eaDir/LLMs、RAG、AI Agent 三个到底什么区别？.md@SynoEAStream
Normal file
Binary file not shown.
BIN
AI/@eaDir/LLMs、RAG、AI Agent 三个到底什么区别？.md@SynoResource
Normal file
BIN
AI/@eaDir/LLMs、RAG、AI Agent 三个到底什么区别？.md@SynoResource
Normal file
Binary file not shown.
BIN
AI/@eaDir/Nano Banana 提示词框架.md@SynoEAStream
Normal file
BIN
AI/@eaDir/Nano Banana 提示词框架.md@SynoEAStream
Normal file
Binary file not shown.
BIN
AI/@eaDir/Nano Banana 提示词框架.md@SynoResource
Normal file
BIN
AI/@eaDir/Nano Banana 提示词框架.md@SynoResource
Normal file
Binary file not shown.
BIN
AI/@eaDir/Never write another prompt.md@SynoEAStream
Normal file
BIN
AI/@eaDir/Never write another prompt.md@SynoEAStream
Normal file
Binary file not shown.
BIN
AI/@eaDir/Never write another prompt.md@SynoResource
Normal file
BIN
AI/@eaDir/Never write another prompt.md@SynoResource
Normal file
Binary file not shown.
BIN
AI/@eaDir/OpenAI ChatGPT 个性化定义.md@SynoEAStream
Normal file
BIN
AI/@eaDir/OpenAI ChatGPT 个性化定义.md@SynoEAStream
Normal file
Binary file not shown.
BIN
AI/@eaDir/OpenAI ChatGPT 个性化定义.md@SynoResource
Normal file
BIN
AI/@eaDir/OpenAI ChatGPT 个性化定义.md@SynoResource
Normal file
Binary file not shown.
BIN
AI/@eaDir/Untitled.md@SynoEAStream
Normal file
BIN
AI/@eaDir/Untitled.md@SynoEAStream
Normal file
Binary file not shown.
BIN
AI/@eaDir/Untitled.md@SynoResource
Normal file
BIN
AI/@eaDir/Untitled.md@SynoResource
Normal file
Binary file not shown.
BIN
AI/@eaDir/openclaw@SynoEAStream
Normal file
BIN
AI/@eaDir/openclaw@SynoEAStream
Normal file
Binary file not shown.
BIN
AI/@eaDir/固定镜头短视频制作的AI全流程解析.md@SynoEAStream
Normal file
BIN
AI/@eaDir/固定镜头短视频制作的AI全流程解析.md@SynoEAStream
Normal file
Binary file not shown.
BIN
AI/@eaDir/固定镜头短视频制作的AI全流程解析.md@SynoResource
Normal file
BIN
AI/@eaDir/固定镜头短视频制作的AI全流程解析.md@SynoResource
Normal file
Binary file not shown.
BIN
AI/@eaDir/如何利用Sora接口实现视频自动化生成工作流.md@SynoEAStream
Normal file
BIN
AI/@eaDir/如何利用Sora接口实现视频自动化生成工作流.md@SynoEAStream
Normal file
Binary file not shown.
BIN
AI/@eaDir/如何利用Sora接口实现视频自动化生成工作流.md@SynoResource
Normal file
BIN
AI/@eaDir/如何利用Sora接口实现视频自动化生成工作流.md@SynoResource
Normal file
Binary file not shown.
BIN
AI/@eaDir/文字生成视频网站推荐.md@SynoEAStream
Normal file
BIN
AI/@eaDir/文字生成视频网站推荐.md@SynoEAStream
Normal file
Binary file not shown.
BIN
AI/@eaDir/文字生成视频网站推荐.md@SynoResource
Normal file
BIN
AI/@eaDir/文字生成视频网站推荐.md@SynoResource
Normal file
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
BIN
AI/@eaDir/🟠API Key.md@SynoEAStream
Normal file
BIN
AI/@eaDir/🟠API Key.md@SynoEAStream
Normal file
Binary file not shown.
BIN
AI/@eaDir/🟠API Key.md@SynoResource
Normal file
BIN
AI/@eaDir/🟠API Key.md@SynoResource
Normal file
Binary file not shown.
219
AI/AI 解决方案专家培训课程.md
Normal file
219
AI/AI 解决方案专家培训课程.md
Normal file
@@ -0,0 +1,219 @@
|
||||
|
||||
#ai #coze
|
||||
|
||||
|
||||
### coze平台demo（国内版）
|
||||
|
||||
1. 点击邀请链接，加入团队空间（不需要重复点击，点过一次之后就成功加入了）
|
||||
|
||||
2. 点击Agent的链接，直接到达Agent页面（可直接对话体验，也可点击右上角创建副本后进行改造）
|
||||
|
||||
|
||||
|
||||
|
||||
**邀请链接**：邀请你加入我的扣子空间"0220-Prompt & RAG & Function Call"，链接将在 2025-06-29 11:28 过期
|
||||
|
||||
👉🏻 https://www.coze.cn/invite/023HTTh566vNqnumiPtx?type=1
|
||||
|
||||
**Agent链接**：
|
||||
|
||||
- 知乎财报解读_Chao：https://www.coze.cn/space/7433704316877520906/bot/7473176769286766632
|
||||
|
||||
- SONY门店店员_Chao ：https://www.coze.cn/space/7433704316877520906/bot/7473182193574363136，给回答打分的提示词[Sony店员沟通测试prompt](https://ncnmfdan85y5.feishu.cn/wiki/EMrVw2SKOixrIekIYMpcz8fxnKP?from=from_copylink)
|
||||
|
||||
- 对话内容解析_Chao：https://www.coze.cn/space/7433704316877520906/bot/7473193418752622592，对话内容原始输入数据[门店店员顾客沟通对话数据](https://ncnmfdan85y5.feishu.cn/wiki/Da2bwqF4ei7IBSkGwRucebRynBh?from=from_copylink)
|
||||
|
||||
- 医疗分诊助手_Chao：https://www.coze.cn/space/7433704316877520906/bot/7473176678181830697
|
||||
|
||||
- 询问天气Call工具_Yuchuan: https://www.coze.cn/space/7433704316877520906/bot/7496391362737815603
|
||||
|
||||
- 故事合成Call工具_Yuchuan: https://www.coze.cn/space/7433704316877520906/bot/7496583684271767592
|
||||
|
||||
- 企业办事助手_Yuchuan: https://www.coze.cn/space/7433704316877520906/bot/7498109970719227938
|
||||
|
||||
- 骑手招聘助手_Yuchuan: https://www.coze.cn/space/7433704316877520906/bot/7496616735870140467
|
||||
|
||||
- 表格问答助手_插件版_Chao：https://www.coze.cn/space/7433704316877520906/bot/7477473633594720292
|
||||
|
||||
- 表格问答助手_代码版_Chao：https://www.coze.cn/space/7433704316877520906/bot/7477473845952790568
|
||||
|
||||
- 表格知识库_Chao：https://www.coze.cn/space/7433704316877520906/bot/7477473355403345931
|
||||
|
||||
- 滴滴计费规则解答_Chao：https://www.coze.cn/space/7433704316877520906/bot/7473180407505633332
|
||||
|
||||
- 滴滴计费解答_WorkFlow_Chao：https://www.coze.cn/space/7433704316877520906/bot/7477475272074412059
|
||||
|
||||
- SONY店员_WorkFlow_Chao：https://www.coze.cn/space/7433704316877520906/bot/7501577412447567909
|
||||
|
||||
- 骑手招聘助手_WorkFlow_Chao：https://www.coze.cn/space/7433704316877520906/bot/7478263479720230923
|
||||
|
||||
- AutoGPT的主prompt：[文件自动处理AutoGPT_主Prompt](https://ncnmfdan85y5.feishu.cn/wiki/UVymwjT9UiCaGJkt9Uvcq7ZlnFc)
|
||||
|
||||
- 在线问诊：https://www.coze.cn/space/7433704316877520906/bot/7480801328214736908
|
||||
|
||||
- 医疗demo
|
||||
|
||||
- 影像图片识别demo数据（Excel）：[医疗图片识别](https://ncnmfdan85y5.feishu.cn/wiki/JxsMwvdkUibvV9kQsx6cbfQFnCh?from=from_copylink)，代码地址：https://github.com/BananaResearch/medical_image_recognition/tree/main
|
||||
|
||||
- 医疗问诊案例：模型参考资料：[GPT-SoVITS](https://www.yuque.com/baicaigongchang1145haoyuangong/ib3g1e)
|
||||
|
||||
- 金融行业 客户分层营销助手：https://www.coze.cn/space/7433704316877520906/bot/7505209120241631243
|
||||
|
||||
- 金融行业 智能客服Agent：https://www.coze.cn/space/7433704316877520906/bot/7505212240938418210
|
||||
|
||||
- [金融行业案例 老师课堂笔记](https://ncnmfdan85y5.feishu.cn/wiki/CNO1w9yGbilj2nk4KFicTIOtnSd)
|
||||
|
||||
- 教育案例 知识库问答：https://www.coze.cn/space/7433704316877520906/bot/7483382009826967606
|
||||
|
||||
- 教育案例 拍照搜视频：https://demo.ai-expert.cc:8443/video_search/
|
||||
|
||||
- 教育行业拍照搜视频demo：[视频解析内容](https://ncnmfdan85y5.feishu.cn/wiki/OTeBwJT6YigoDakDrQsc46VNnbg?from=from_copylink)
|
||||
|
||||
- 教育案例 组卷出题：https://www.coze.cn/space/7433704316877520906/bot/7483446959312044047
|
||||
|
||||
- 教育案例 知识点掌握情况评估： https://www.coze.cn/space/7433704316877520906/bot/7505974042647068684
|
||||
|
||||
- 财务行业案例：https://www.coze.cn/space/7433704316877520906/bot/7497919484410691619
|
||||
|
||||
- 财务行业案例 模型测试及优化过程数据：[财务行业 - 企业预算管理](https://ncnmfdan85y5.feishu.cn/wiki/P4yAwzgDBiGdGkk5N0DcFpaPnyf)
|
||||
|
||||
- 财务行业案例 其它资料 [业务预算数据的专家经验](https://ncnmfdan85y5.feishu.cn/wiki/AuZ6wc08wimJ3Rkc68wcw9hInff)
|
||||
|
||||
- 数据分析案例：https://www.coze.cn/space/7433704316877520906/project-ide/7507579385827360779
|
||||
|
||||
- 人力资源案例：
|
||||
|
||||
- 招聘场景打分能力验证：https://www.coze.cn/space/7433704316877520906/bot/7486001310287118377
|
||||
|
||||
- 面试对话：https://www.coze.cn/space/7433704316877520906/bot/7485649954023702566
|
||||
|
||||
- AI培训对练：https://www.coze.cn/space/7433704316877520906/bot/7507280886069477388
|
||||
|
||||
- 莫欣老师的课程demo：https://www.coze.cn/space/7433704316877520906/project-ide/7508998840931123212
|
||||
|
||||
- 莫欣老师直播上课时搭建的：https://www.coze.cn/space/7433704316877520906/project-ide/7509443526267355199
|
||||
|
||||
- 电商
|
||||
|
||||
- 混剪助手：https://www.coze.cn/space/7433704316877520906/bot/7482459190217146387
|
||||
|
||||
- 在线换衣：https://demo.bananaresearch.cn/videogen/
|
||||
|
||||
- 电商行业案例中用到的开源模型（链接内是项目代码，可自行部署）：[电商行业案例开源项目汇总](https://ncnmfdan85y5.feishu.cn/wiki/PefTwB99EiChXlkdXZjcfJFNnsc)
|
||||
|
||||
- 抖音直播间自动回复助手（录播课demo）：[直播间助手 demo 说明](https://ncnmfdan85y5.feishu.cn/wiki/UzE7wbxFAiw6JfkrOpocTNnjnpb)
|
||||
|
||||
- 泛娱乐
|
||||
|
||||
- 霸道总裁：https://www.coze.cn/space/7433704316877520906/bot/7485312777990062118
|
||||
|
||||
- FaceFusion：https://www.facefusion.co/
|
||||
|
||||
- F5-TTS：https://github.com/SWivid/F5-TTS
|
||||
|
||||
- Google Genie 2：https://deepmind.google/discover/blog/genie-2-a-large-scale-foundation-world-model/
|
||||
|
||||
- World labs：https://www.worldlabs.ai/blog
|
||||
|
||||
- 以下是泛娱乐录播课需要的链接
|
||||
|
||||
- AI证件照Demo：https://idphoto.bananaresearch.cn/
|
||||
|
||||
- 人脸识别模型：https://huggingface.co/spaces/hysts/mediapipe-face-detection?utm_source=chatgpt.com
|
||||
|
||||
- AI生成视频工作流1 ：https://www.coze.cn/work_flow?space_id=7433704316877520906&workflow_id=7511205004892471337
|
||||
|
||||
- AI生成视频工作流2 古风育儿: https://www.coze.cn/work_flow?space_id=7433704316877520906&workflow_id=7511280492429377590
|
||||
|
||||
- AI生成视频工作流3 儿童神话故事: https://www.coze.cn/work_flow?space_id=7433704316877520906&workflow_id=7511280755508707340
|
||||
|
||||
- AI生成视频工作流4 治愈女孩视频:https://www.coze.cn/work_flow?space_id=7433704316877520906&workflow_id=7511281332770619401
|
||||
|
||||
- 在线客服
|
||||
|
||||
- 解决方案课程AI助教：https://www.coze.cn/space/7433704316877520906/bot/7513143689787719699
|
||||
|
||||
- 录播课1涉及到的文档：[解决方案课程的AI助教涉及的工作流](https://ncnmfdan85y5.feishu.cn/wiki/LWl7wM8CMivQeska9itcj3wun0c?from=from_copylink)
|
||||
|
||||
- AI销售：https://www.coze.cn/space/7433704316877520906/bot/7512921281609220133
|
||||
|
||||
- 录播课2涉及到的文档：[AI在线销售部门案例涉及到的智能体和工作流](https://ncnmfdan85y5.feishu.cn/wiki/OQQEw54TaiTnSak1shscPwYinve?from=from_copylink)
|
||||
|
||||
|
||||
|
||||
|
||||
demo解析录播课的团队空间，需要重新点邀请链接
|
||||
|
||||
1. AutoGPT：邀请你加入我的扣子空间"AutoGPT"，链接将在 2025-06-29 11:29 过期
|
||||
|
||||
|
||||
👉🏻 https://www.coze.cn/invite/C7874GVv908sJp7vu08Z?type=1，加入新的团队空间后，直接点链接即可找到该Agent：https://www.coze.cn/space/7434815743025594431/bot/7437180587003281460
|
||||
|
||||
2. 支小助：邀请你加入我的扣子空间"支小助Demo"，链接将在 2025-06-29 11:31 过期
|
||||
|
||||
|
||||
👉🏻 https://www.coze.cn/invite/WBXFvY4JDoXdVvZNu2Fs?type=1，加入新的团队空间后，直接点链接即可找到该Agent：https://www.coze.cn/space/7434815646162223144/bot/7478274489961365558
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
相关文件资料：通过网盘分享的文件：相关文件资料0427
|
||||
|
||||
链接: https://pan.baidu.com/s/1Wo6x9V0eGfOMNzpdaBrNFQ?pwd=eqx7 提取码: eqx7
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
### coze平台demo（海外版）
|
||||
|
||||
1. 点击邀请链接，加入团队空间（不需要重复点击，点过一次之后就成功加入了）
|
||||
|
||||
2. 点击Agent的链接，直接到达Agent页面（可直接对话体验，也可点击右上角创建副本后进行改造）
|
||||
|
||||
|
||||
|
||||
|
||||
**邀请链接**：join my space"Prompt & RAG & Function Call", this URL will be expired on 2025-06-23 16:27.👉🏻 https://www.coze.com/invite/JtW2fJUv2WTt4drYnP4T?type=1
|
||||
|
||||
**Agent链接**：
|
||||
|
||||
- 知乎财报解读_Chao：https://www.coze.com/space/7432640186326712326/bot/7473195950740144146
|
||||
|
||||
- SONY门店店员_Chao ：https://www.coze.com/space/7432640186326712326/bot/7473197554201657362，给回答打分的提示词[Sony店员沟通测试prompt](https://ncnmfdan85y5.feishu.cn/wiki/EMrVw2SKOixrIekIYMpcz8fxnKP?from=from_copylink)
|
||||
|
||||
- 对话内容解析_Chao：https://www.coze.com/space/7432640186326712326/bot/7473197683965558791，对话内容原始输入数据[门店店员顾客沟通对话数据](https://ncnmfdan85y5.feishu.cn/wiki/Da2bwqF4ei7IBSkGwRucebRynBh?from=from_copylink)
|
||||
|
||||
- 医疗分诊助手_Chao：https://www.coze.com/space/7432640186326712326/bot/7473191673704136711
|
||||
|
||||
- 询问天气Call工具：https://www.coze.com/space/7432640186326712326/bot/7475659806565793799
|
||||
|
||||
- 故事合成Call工具：https://www.coze.com/space/7432640186326712326/bot/7475658544307159058
|
||||
|
||||
- 企业办事助手：https://www.coze.com/space/7432640186326712326/bot/7475657076598538248
|
||||
|
||||
- 骑手招聘助手：https://www.coze.com/space/7432640186326712326/bot/7475663329072381960
|
||||
|
||||
- 滴滴计费解答_WorkFlow_Chao：https://www.coze.com/space/7432640186326712326/bot/7478661424374382600
|
||||
|
||||
- 表格问答助手_代码版_Chao：https://www.coze.com/space/7432640186326712326/bot/7478649751164993543
|
||||
|
||||
- 表格问答助手_插件版_Chao：https://www.coze.com/space/7432640186326712326/bot/7478647812881072135
|
||||
|
||||
- 在线问诊：https://www.coze.com/space/7432640186326712326/bot/7485293332848033800
|
||||
|
||||
|
||||
|
||||
|
||||
demo解析录播课的团队空间，需要重新点邀请链接
|
||||
|
||||
1. AutoGPT：join my space"AutoGPT", this URL will be expired on 2025-06-23 16:28.👉🏻 https://www.coze.com/invite/6xpVGvvuhdBGTibSxp2i?type=1，加入新的团队空间后，直接点链接即可找到该Agent：https://www.coze.com/space/7410266370836840465/bot/7435939032980389904
|
||||
|
||||
2. 支小助：join my space"支小助Demo", this URL will be expired on 2025-06-23 16:27.👉🏻 https://www.coze.com/invite/V5NuDchUoobsODEtByGU?type=1，加入新的团队空间后，直接点链接即可找到该Agent：https://www.coze.com/space/7401006355362185222/bot/7401007312318169094
|
||||
|
||||
3. 市场调研助手：join my space"调研助手", this URL will be expired on 2025-06-23 16:26.👉🏻 https://www.coze.com/invite/cy9b6Futvnyp4xUZUhWd?type=1，加入新的团队空间后，直接点链接即可找到该Agent：https://www.coze.com/space/7426296757053259784/bot/7433710527240962049
|
||||
|
||||
|
||||
147
AI/Best 7 news API data feeds - AI News.md
Normal file
147
AI/Best 7 news API data feeds - AI News.md
Normal file
@@ -0,0 +1,147 @@
|
||||
---
|
||||
title: Best 7 news API data feeds - AI News
|
||||
source: https://www.artificialintelligence-news.com/news/best-7-news-api-data-feeds/
|
||||
author:
|
||||
published: 2025-03-11
|
||||
created: 2025-03-14
|
||||
description: With the rapid growth in the generation, storage, and sharing of data, ensuring its security has become both a necessity and a formidable challenge.
|
||||
tags:
|
||||
- clippings
|
||||
---
|
||||
Access to real-time and historical news data is important in today’s digital landscape. Businesses, developers, and analysts rely on news API data feeds to gather structured insights from various sources, ranging from global news outlets and blogs, to forums and social media. APIs help integrate content into applications and workflows, enabling decision-making and scalable solutions.
|
||||
|
||||
### What are news API data feeds?
|
||||
|
||||
News API data feeds are platforms that aggregate, organise, and deliver structured news data from multiple sources, like websites, blogs, forums, and online publications. They simplify the process of gathering information from different outlets and formatting it into machine-readable formats like JSON or XML. These feeds eliminate the manual effort of collecting and curating data by presenting structured content ready to be processed.
|
||||
|
||||
### Top 7 news API data feeds
|
||||
|
||||
Let’s explore seven top news API data feeds leading the industry. These tools provide businesses with real-time access, historical coverage, and features tailored to various industries.
|
||||
|
||||
#### 1\. Webz.io
|
||||
|
||||
[Webz.io](http://webz.io/) is one of the most comprehensive news APIs, offering both real-time and archived coverage from the open and deep web, as well as the dark web. It provides highly customisable data feeds for industries like finance, risk intelligence, and cybersecurity.
|
||||
|
||||
Key features:
|
||||
|
||||
- Access to open, deep, and dark web data.
|
||||
|
||||
- Advanced filters for sentiment, topic, and geographic coverage.
|
||||
|
||||
- Support for visualisation and actionable risk monitoring.
|
||||
|
||||
Use case: Media monitoring, sentiment analysis, and threat intelligence for corporate security teams and financial organisations.
|
||||
|
||||
Why Webz.io? Its expansive source list and deep customisation options make it ideal for specialised industries like cybersecurity and financial analytics.
|
||||
|
||||
#### 2\. GNews API
|
||||
|
||||
GNews API is a simple, lightweight platform that aggregates reliable news from around the globe. It is perfect for small-scale applications or developers looking for affordable yet efficient solutions.
|
||||
|
||||
Key features:
|
||||
|
||||
- Real-time global coverage.
|
||||
|
||||
- Filters for topics, languages, and countries.
|
||||
|
||||
- Affordable pricing plans suitable for startups.
|
||||
|
||||
Use case: Localisation-focused news widgets or small aggregators serving specific regional or language-based audiences.
|
||||
|
||||
Why GNews? Its intuitive design and affordability make GNews a great entry point for developers and startups.
|
||||
|
||||
#### 3\. The Guardian API
|
||||
|
||||
The Guardian API provides direct access to high-quality journalism from the Guardian’s editorial content. It offers structured news, tags, and metadata from one of the world’s most respected news organisations.
|
||||
|
||||
Key features:
|
||||
|
||||
- High-quality editorial content.
|
||||
|
||||
- Filtering by topic or category.
|
||||
|
||||
- Media-rich datan integration, including multimedia embedding.
|
||||
|
||||
Use case: Apps or research projects requiring trusted editorial sources for accurate analysis or curated content.
|
||||
|
||||
Why The Guardian API? Focused on credible data, it works best for platforms and professionals prioritising journalistic integrity.
|
||||
|
||||
#### 4\. Bloomberg API
|
||||
|
||||
Renowned for its financial insights, Bloomberg API delivers in-depth business coverage and real-time data for institutions and professional investors. It specialises in market data, financial news, and economic reports.
|
||||
|
||||
Key features:
|
||||
|
||||
- Exclusive financial data and analysis.
|
||||
|
||||
- Real-time market coverage.
|
||||
|
||||
- Seamless integration with Bloomberg’s terminals.
|
||||
|
||||
Use case: Analysts and investment professionals monitoring market trends and making data-driven decisions.
|
||||
|
||||
Why Bloomberg? Its precise focus on finance makes it essential for institutions heavily reliant on actionable market news.
|
||||
|
||||
#### 5\. Financial Times API
|
||||
|
||||
The Financial Times API is a premium solution that supplies business and economic-focused news. It is built for professional teams that require deep insights into global markets and economic activity.
|
||||
|
||||
Key features:
|
||||
|
||||
- Premium content on global finance and markets.
|
||||
|
||||
- Access to detailed economic reports and analyses.
|
||||
|
||||
- Subscription access for gated content.
|
||||
|
||||
Use case: Economists, researchers, or executives tracking global economic trends and industry reports.
|
||||
|
||||
Why Financial Times? Its premium-quality data and economic insights provide unmatched value for businesses targeting comprehensive market analysis.
|
||||
|
||||
#### 6\. Opoint
|
||||
|
||||
Opoint specialises in news monitoring and sentiment analysis, making it particularly useful for PR, marketing, and branding teams. It supports multiple languages and global sources with cutting-edge media monitoring capabilities.
|
||||
|
||||
Key features:
|
||||
|
||||
- Real-time monitoring with sentiment tagging.
|
||||
|
||||
- Multilingual and multi-source coverage.
|
||||
|
||||
- Tailored brand monitoring and competitor tracking.
|
||||
|
||||
Use case: PR agencies and marketers monitoring sentiment shifts or competitive landscape changes like product launches.
|
||||
|
||||
Why Opoint? Its advanced monitoring features help organisations stay agile in rapidly shifting media environments.
|
||||
|
||||
#### 7\. Mediastack API
|
||||
|
||||
Mediastack combines accessibility with scalability, offering a mix of free plans for developers and paid tiers for advanced features. It aggregates news in real time from over 7,500 sources globally.
|
||||
|
||||
Key features:
|
||||
|
||||
- Free and affordable paid plans.
|
||||
|
||||
- Multilingual support and geo-targeted searches.
|
||||
|
||||
- Scalable for both startups and growing enterprises.
|
||||
|
||||
Use case: Developers building applications that require versatile, budget-friendly news feeds with reliable real-time updates.
|
||||
|
||||
Why Mediastack? Its affordability and flexibility cater to businesses of all sizes, making it a versatile option for a wide range of users.
|
||||
|
||||
### Use cases for news API data feeds
|
||||
|
||||
The applications of news API data feeds are as diverse as the industries relying on them:
|
||||
|
||||
**Financial intelligence**: Investment tools use APIs to analyse market-moving news in real time.
|
||||
|
||||
**Media monitoring**: PR agencies use media insights to track brand mentions and sentiment.
|
||||
|
||||
**Risk assessment**: Governments and corporations assess geopolitical risks or public sentiment.
|
||||
|
||||
**Content platforms**: Aggregators curate articles, summaries, and headlines for apps/websites.
|
||||
|
||||
**AI & predictive analysis**: APIs provide data for machine learning models that forecast trends.
|
||||
|
||||
*(Image source: Unsplash)*
|
||||
45
AI/Designing for Agentic AI.md
Normal file
45
AI/Designing for Agentic AI.md
Normal file
@@ -0,0 +1,45 @@
|
||||
---
|
||||
title: Designing for Agentic AI
|
||||
source: https://www.linkedin.com/pulse/designing-agentic-ai-yuri-pessa-ztcmf/?trackingId=gSoKslBrTP6VWNCDJSd7ZA%3D%3D
|
||||
author:
|
||||
published: 2001-02-27
|
||||
created: 2025-03-02
|
||||
description:
|
||||
tags:
|
||||
- clippings
|
||||
- agentic-ai
|
||||
- ai
|
||||
- "#design"
|
||||
---
|
||||
The world of AI is constantly evolving, and with it, the way we interact with technology. You might have heard of Generative AI (GenAI), but what about Agentic AI? Let's explore the differences and the exciting implications for product designers.
|
||||
|
||||
## GenAI vs. Agentic AI: What's the Difference?
|
||||
|
||||
GenAI excels at creating new content, like text, images, or music. Think of it as a creative assistant that can generate ideas or translate languages. Agentic AI, on the other hand, is all about action. It can interact with its environment, make decisions, and even anticipate user needs. It's like having a personal agent working for you 24/7.
|
||||
|
||||
Example:
|
||||
|
||||
- GenAI: You ask it to write a poem about a cat, and it generates a beautiful piece of verse.
|
||||
- Agentic AI: You ask it to schedule a meeting with a colleague, and it not only finds a time that works for both of you but also considers your preferred meeting locations and automatically sends out calendar invites.
|
||||
|
||||
## Designing for Feedback
|
||||
|
||||
Agentic AI is pushing us to reimagine product design. For years, we've focused on interfaces that react to direct user input—clicks, swipes, and edits. But agentic AI introduces a new dimension: proactive agents that anticipate needs and act autonomously.
|
||||
|
||||
This doesn't mean users become passive. Observing the AI's decision-making process, understanding its "thinking," is a form of interaction in itself. The user may not be clicking buttons, but they're still engaged, evaluating, and potentially intervening.
|
||||
|
||||
This shift requires a new design metaphor. Instead of just reacting to user actions, we're crafting experiences that provide live feedback as the AI operates. The focus is on transparency, allowing users to understand and respond to what's happening in real-time.
|
||||
|
||||
## Best Practices for Designing Agentic AI Experiences
|
||||
|
||||
Here are some best practices for designing agentic AI experiences:
|
||||
|
||||
- **Transparency:** Users should be able to understand how the AI is making decisions. This can be achieved by visualizing the AI's progress in completing a task and providing users with a summary of the AI's reasoning process.
|
||||
- **Control:** Users should always feel in control of the AI. This can be achieved by providing users with a clear way to stop the AI from performing a task or to undo an action that the AI has taken, as well as allowing users to set preferences for how the AI should behave.
|
||||
- **Personalization:** Agentic AI should adapt to individual user needs and preferences. This can be achieved by using the user's past behavior to predict their future needs and offer relevant suggestions, as well as allowing users to provide feedback on the AI's performance.
|
||||
- **Conversation:** Design for natural, intuitive conversations between users and the AI. This can be achieved by using a conversational interface that allows users to interact with the AI using natural language and providing users with feedback on how the AI is interpreting their input.
|
||||
- **Anticipation:** Agentic AI should be able to anticipate user needs and proactively offer assistance. However, users should also have the ability to control the level of autonomy they want to give to the AI. This can be achieved by providing users with clear controls to adjust the AI's level of autonomy, as well as providing feedback on the AI's anticipated actions.
|
||||
|
||||
By considering all five of these best practices, designers can create agentic AI experiences that provide the high level of real-time feedback that users will expect. This will help to ensure that users feel in control of the AI and that they understand how it is making decisions.
|
||||
|
||||
We're just scratching the surface of what's possible with agentic AI. What are your thoughts on designing for this new paradigm? Share your best practices or any other implications you foresee in the comments below!
|
||||
165
AI/Google 神级生产力工具，所有 GitHub 开源平替都找到了。.md
Normal file
165
AI/Google 神级生产力工具，所有 GitHub 开源平替都找到了。.md
Normal file
@@ -0,0 +1,165 @@
|
||||
---
|
||||
title: Google 神级生产力工具，所有 GitHub 开源平替都找到了。
|
||||
source: https://mp.weixin.qq.com/s/6EoEMi8opDWOParUHRiHOg
|
||||
author:
|
||||
- "[[逛逛]]"
|
||||
published:
|
||||
created: 2026-01-01
|
||||
description:
|
||||
tags:
|
||||
- clippings
|
||||
---
|
||||
原创 逛逛 *2025年12月19日 15:24*
|
||||
|
||||
NotebookLM 是谷歌推出的 一款 AI 笔记助手 。与普通 AI 不一样，它严格限制在你上传的文档范围里进行回答，并能提供精准的原文引用。
|
||||
|
||||
它最出圈的功能是 播客生成 ，能一键把你上传的复杂资料转换成一段逼真的双人英语对话播客。不仅让学习变得更有趣，还支持通过听来消化信息。
|
||||
|
||||
|
||||
|
||||
Unlock Smarter Studying with Google’s LM Notebook
|
||||
|
||||
01
|
||||
|
||||
**最受欢迎的 Notebook LM 开源平替**
|
||||
|
||||
Open Notebook 是 GitHub 上 Star 数量最高的 开源平替项目。
|
||||
|
||||
在 GitHub 上已经获得了 **14.6k** 颗 Star。
|
||||
|
||||
|
||||
|
||||
它是一个全功能的本地化解决方案， 不依赖云端的情况下进行知识管理和研究， 支持通过 Docker 等方式轻松部署。
|
||||
|
||||
该项目在模型选择上非常开放，目前 支持超过 16 种 AI 提供商 ，包括 OpenAI、Anthropic、Gemini 等主流云端模型。
|
||||
|
||||
同时也完美支持通过 Ollama 或 LM Studio 运行的本地模型。你可以根据成本、隐私需求或性能偏好自由切换底层 AI 能力。
|
||||
|
||||
|
||||
|
||||
这个开源项目支持 多模态内容输入 ，包括 PDF、网页、音频和 YouTube 视频等。
|
||||
|
||||
它不仅具备类似 NotebookLM 的文档问答和引用功能，还提供了 高级的播客生成工 具，支持创建多达 4 位演讲者的多角色对话，还能对脚本进行精细控制。
|
||||
|
||||
关于他和 Google 的那个工具的差异，可以看下面这个表格：
|
||||
|
||||
|
||||
|
||||
```perl
|
||||
开源地址：https://github.com/lfnovo/open-notebook
|
||||
```
|
||||
|
||||
02
|
||||
|
||||
**SurfSense：AI 搜索与研究智能体**
|
||||
|
||||
目前，SurfSense 在 GitHub 上拥有 **11.4k** 颗 Star。
|
||||
|
||||
它是一个比较综合的开源 AI 搜索与研究智能体 ，定位为 NotebookLM、Perplexity 和 Glean 的开源替代品。
|
||||
|
||||
|
||||
|
||||
它不仅能处理上传的文件，还能连接广泛的外部数据源，通过 整合你的个人知识库和外部信息流，进行深度定制化的研究。
|
||||
|
||||
它能够集成多种平台和工具，包括 Notion、YouTube、GitHub 啥的。
|
||||
|
||||
而且采用 语义搜索 + 全文搜索 混合搜索技术，并结合 重排序算法 ，确保在海量数据中能快速精准地找到并引用答案。
|
||||
|
||||
SurfSense 的功能非常丰富，支持与保存的内容进行自然语言对话、生成带有引用的答案，以及利用本地 LLM 保护隐私。
|
||||
|
||||
它还内置了 快速播客生成智能体 ，能够在短时间内将聊天内容转化为引人入胜的音频内容，并支持多种文本转语音服务。
|
||||
|
||||
支持 Docker 容器化部署和基于角色的访问控制（RBAC），使其不仅适合个人研究者，也适合需要 团队协作和知识共享 的企业环境。
|
||||
|
||||
   
|
||||
|
||||
```javascript
|
||||
开源地址：https://github.com/MODSetter/SurfSense
|
||||
```
|
||||
|
||||
03
|
||||
|
||||
**Podcastfy：专注播客生成**
|
||||
|
||||
Podcastfy 专注于播客生成，对标的是 NotebookLM 的播客生成功能。
|
||||
|
||||
他可以把多模态内容，比如文本、图像、网站、PDF 等 转化为高质量、多语言的音频对话。
|
||||
|
||||
|
||||
|
||||
这个工具提供了 高度的定制化能力 ，可以让你生成短视频风格（Shorts）或长篇深度（Longform）的播客内容。
|
||||
|
||||
它整合了超过 100 种 LLM 用于脚本生成，并支持 OpenAI、Google、ElevenLabs 以及 Microsoft Edge TTS 等 多种语音合成引擎 ，确保生成的语音自然且富有表现力。
|
||||
|
||||
Podcastfy 不仅作为一个 Python 包供开发者调用，还提供了命令行工具和 Web 界面，方便不同技术背景的用户使用。
|
||||
|
||||
```javascript
|
||||
开源地址：https://github.com/souzatharsis/podcastfy
|
||||
```
|
||||
|
||||
04
|
||||
|
||||
**notebookllama**
|
||||
|
||||
|
||||
|
||||
NotebookLlama 是由 LlamaIndex 官方推出的一个完全开源的项目，现在 1.7k 的 Star。
|
||||
|
||||
通过 LlamaCloud 生态系统来处理复杂的文档解析，并利用开源模型的能力来实现从文档到播客的转换流程。
|
||||
|
||||
看这个开源项目，你会学会 如何利用 AI 大模型技术链条构建一个文档转播客的应用。
|
||||
|
||||
涵盖了从文本提取、脚本生成、戏剧化改编到最终文本转语音（TTS）的全过程。
|
||||
|
||||
用户可以使用 OpenAI 或 ElevenLabs 的 API，也可以选择完全本地化的模型来运行这一流程。
|
||||
|
||||
```javascript
|
||||
开源地址：https://github.com/run-llama/notebookllama
|
||||
```
|
||||
|
||||
05
|
||||
|
||||
**学习工具：** PageLM
|
||||
|
||||
PageLM 是一个 把学习材料转化为互动式资源的教育平台，通过 AI 技术提升学习效率。
|
||||
|
||||
这个开源项目提供了一系列针对学习场景优化的功能，包括自动生成 康奈尔笔记（SmartNotes） 、基于文档的 互动测验、间隔重复闪卡（Flashcards） 以及 模拟考试系统（ExamLab）。
|
||||
|
||||
它还能将枯燥的学习资料转化为播客，不仅支持读，更支持听和测。
|
||||
|
||||
|
||||
|
||||
PageLM 在技术架构上支持多种主流 AI 模型，包括 Google Gemini、OpenAI GPT、Anthropic Claude 以及本地的 Ollama 模型。
|
||||
|
||||
这意味着用户可以根据自己的预算和硬件条件，灵活配置用于生成学习内容的后端模型。
|
||||
|
||||
```javascript
|
||||
开源地址：https://github.com/CaviraOSS/PageLM
|
||||
```
|
||||
|
||||
06
|
||||
|
||||
**InsightsLM**
|
||||
|
||||
InsightsLM 这个 NotebookLM 替代方案，强调低代码/无代码。
|
||||
|
||||
它采用 Supabase 作为后端数据库和存储， 结合 N8N 工作流自动化工具， 前端则基于 React 构建，为你提供了一个可完全掌控数据的私有化研究工具。
|
||||
|
||||
|
||||
|
||||
核心功能包括与上传的文档进行聊天、生成带有可验证引用的回答，以及生成播客。
|
||||
|
||||
InsightsLM 的独特之处在于 它利用了 N8N 进行后端逻辑处理，同时也支持本地化部署方案 ，允许接入 Ollama 和 Qwen3 等本地模型，实现完全离线的 AI 交互。
|
||||
|
||||
```javascript
|
||||
开源地址：https://github.com/theaiautomators/insights-lm-public
|
||||
```
|
||||
|
||||
07
|
||||
|
||||
**点击下方卡片，关注逛逛 GitHub**
|
||||
|
||||
这个公众号历史发布过很多有趣的开源项目，如果你懒得翻文章一个个找，你直接关注微信公众号：逛逛 GitHub ，后台对话聊天就行了：
|
||||
|
||||
|
||||
|
||||
139
AI/LLMs、RAG、AI Agent 三个到底什么区别？.md
Normal file
139
AI/LLMs、RAG、AI Agent 三个到底什么区别？.md
Normal file
@@ -0,0 +1,139 @@
|
||||
---
|
||||
title: "LLMs、RAG、AI Agent 三个到底什么区别？"
|
||||
source: "https://mp.weixin.qq.com/s/8B_Phrjz_Mlvpe7vJ3maPA"
|
||||
author:
|
||||
- "[[易程LEO]]"
|
||||
published:
|
||||
created: 2025-11-19
|
||||
description: "主要讲明白关于LLMs、RAG和AI Agent这三个定义的区别到底是什么？这三者目前已经是做AI相关应用绕不过去的名词，也是作为初入AI应用开发者，必须了解掌握的基础知识。"
|
||||
tags:
|
||||
- "clippings"
|
||||
---
|
||||
|
||||
#llm #rag #ai-agent
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
对于接触 AI 相关的朋友，平时都会遇到很多新的概念，先不说什么大模型的技术性的术语，就AI应用方面的术语就非常多。
|
||||
|
||||
而且，现在还是依旧层出不穷。
|
||||
|
||||
在技术迭代到一定程度之后，它就必然会满足更多的实际场景，而要满足某些实际场景的话，并不是单单依靠某个单一技术就可以实现的。
|
||||
|
||||
举个例子来说，大家知道计算机技术最开始其实只有CPU和内存等外置硬件设备，那个时候都是基于命令行方式来做一些计算工作，普通人想要用起来计算机的话，门槛极高。
|
||||
|
||||
后来便有了Linux这类操作系统，它可以支持自定义编程，也就是在计算机硬件基础上来开发满足实际场景的软件，这里面最典型的就是操作系统，也就是我们现在用的Window、Mac等操作系统。
|
||||
|
||||
这时候，计算机（PC）和Windows、MAC等等都是当时为了满足大众使用计算机所创造出的术语/名词，通过这个概念名词来定义某个技术的作用是什么，相当于给它们起一个名字来表示。
|
||||
|
||||
继续沿着操作系统之后，就知道后面有很多基于操作系统之上的新名词诞生，例如Web浏览器、客户端软件、Client/Server技术架构等等，这些又都是在操作系统之上为了满足更多实际场景而开发出来的新东西，而每一个都是满足当时场景下的新名词。
|
||||
|
||||
所以，在AI成为新的普适性的技术底座之前，必然会有更多的名词定义出来，而它也是为了满足特定场景，解决特定问题所存在的必然。
|
||||
|
||||
今天我们主要讲明白关于LLMs、RAG和AI Agent这三个定义的区别到底是什么？这三者目前已经是做AI相关应用绕不过去的名词，也是作为初入AI应用开发者，必须了解掌握的基础知识。
|
||||
|
||||
首先，要先注意一点：它们并不是竞争技术，而是在三个不同层面，满足不同实际场景的能力展示，另外大部分人对它们使用方式都是错误的。
|
||||
|
||||
LLM 全称是大语言模型（Large Language Model），它是AI应用的“天才大脑”，这个天才大脑学习了过去上下五千年的所有知识，是的，是所有知识，堪比“全能人”。
|
||||
|
||||
这个“天才大脑”你问它啥，它都能回答上来，甚至还能帮助我们写写文章、分析点东西、编程、画画等等的。
|
||||
|
||||
LLMs也分为很多种，有底座大模型，例如ChatGPT、DeepSeek、Qwen等等，也有专有大模型，也就是专门用来画画，专门用来编写的模型，例如绘画模型：Midjourney、Stable Diffusion、Flux等等，编程模型：Claude、Curos、kimi-k2-thing等等。
|
||||
|
||||
专有模型某种意义上来说，也是基于底座通用大模型来单独训练出来的能力，也就是让“天才大脑”对于某一个方面特别精通，做了专项的训练。
|
||||
|
||||
但是，这个大模型有一个问题，它只能知道过去已经发生的时候，在上面也提到了，它是基于过去的所有知识训练、学习出来的，所以，它的知识内容啊，是有某一个时间节点的，例如ChatGPT-5的知识时间就是2024年6月，单独问这个模型2025年的事情，它都不知道。
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
当然，现在是有了联网搜索的能力了，但是这种其实是在大模型之外的Agent助手，通过这个外部Agent助手，可以爬取网站的数据，或者通过搜索引擎（Baidu、Bing、Google等）来获取相关数据，然后在交给大模型来总结分析。
|
||||
|
||||
总结起来：LLM 在思考方面非常出色，但对当前情况却一无所知。
|
||||
|
||||
|
||||
|
||||
这个时候，就可以引出第二个名词解释，就是RAG。
|
||||
|
||||
RAG（Retrieval-Augmented Generation，检索增强生成）可以说是一个记忆系统，它可以将原本静态固定的“天才大脑”LLM中的知识，链接到外部实时的知识库，当你提问问题的时候，RAG会主动搜索外部数据，拉去相关文档，并将它们作为上下文输入到LLM中。
|
||||
|
||||
这样就好比于，原本是一个“书呆子”，突然打开了视野，变得灵活多动了，对于原来静态的大模型来说，动态信息、实时数据也就以为这它不需要重新训练了。
|
||||
|
||||
在大模型训练（也就是模型学习知识的过程）是一个非常高昂成本的过程，啥意思？就是费钱，不仅仅要买书、还要营养跟得上，不然动不动就卡壳、生病（出bug）啥的，所以，要用很多高端GPU卡，来吸收海量数据才能让这个大脑学会知识。
|
||||
|
||||
最基础的工具是能够访问最新信息的能力。检索增强生成（RAG）为智能体提供了一张“借书证”，使其能查询外部知识，这些知识通常存储在向量数据库或知识图谱中——从公司内部文档到通过谷歌搜索获取的网络知识，应有尽有。对于结构化数据，自然语言到SQL（NL2SQL）工具则使智能体能够直接查询数据库，从而解答诸如“上个季度我们的畅销产品有哪些？”这类分析性问题。通过在发言前先查找相关信息——无论是来自文档还是数据库——智能体得以立足于事实，显著地减少幻觉。
|
||||
|
||||
RAG 流程结合了两个关键步骤：
|
||||
|
||||
**1\. 检索（Retrieval）：**
|
||||
|
||||
当用户提出问题时，系统首先从一个或多个 **外部、定制化** 的知识库（如公司的内部文件、最新的数据库、特定领域文档等）中，检索出最相关的小块信息（Chunk）。
|
||||
|
||||
2\. 增强生成（Augmented Generation）：
|
||||
|
||||
然后，系统将用户的原始问题和检索到的相关信息作为 **上下文** （Context）输入给 LLM，指示 LLM 严格基于这些上下文信息来生成答案。
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
RAG 就像是给那个“全能天才大脑”配备了一位 **随身图书馆助理** ：
|
||||
|
||||
**1\. 知识更新与定制：**
|
||||
|
||||
当你问一个关于“公司最新财报”或“某本专业书籍第十章内容”的问题时，RAG 不会依赖 LLM 内部的旧知识，而是立即去检索公司内部最新的文档。
|
||||
|
||||
**2\. 消除幻觉：**
|
||||
|
||||
通过提供 **事实依据** ，RAG 极大地降低了 LLM “胡编乱造”的风险，因为它生成的答案是 **有据可查** 的。
|
||||
|
||||
**3\. 引用来源：**
|
||||
|
||||
优秀的 RAG 系统还能提供它查找信息的 **来源链接或文档页码** ，增加了可信度。
|
||||
|
||||
接下来还有最后一个名词，就是AI Agent，也叫做AI智能体，为啥叫智能体？
|
||||
|
||||
结合上面，LLM是思考，RAG是提供信息，但 是它俩都不具备行动能力，有脑，有手，但是不知道怎么走路。
|
||||
|
||||
而AI Agent也就是智能体，它就是围绕大脑LLM构建一个循环控制系统，能够感知目标、规划步骤、执行动作、并能够反思结果。
|
||||
|
||||
本质上，智能体通过一个连续的循环过程来实现其目标。它可被分解为五个基本步骤：
|
||||
|
||||
1\. 获取任务：该过程由一个具体且高层次的目标启动。此任务可由用户（例如：“为团队安排即将召开的会议出行事宜”）提供，或由自动触发机制（例如：“新收到一封高优先级客户工单”）激活。
|
||||
|
||||
2\. 扫描场景：Agent感知到环境中获取上下文信息。这涉及协调层访问其可用资源：“用户请求的内容是什么？”、“我的术语记忆中有哪些信息？我是否已尝试过执行此任务？”、“用户上周是否曾向我提供过指导？”、“我能从我的工具（如日历、数据库或API）中访问哪些内容？”
|
||||
|
||||
3\. 仔细思考：这是智能体的核心“思考”循环，由推理模型驱动。
|
||||
|
||||
智能体首先将任务（步骤1）与场景（步骤2）进行分析，并制定行动计划。这并非单一的思考过程，而通常是一系列连续的推理链条：“要预订行程，我首先需要知道团队成员都有谁，因此我会使用get\_team\_roster工具；接下来，我还需要通过calendar\_api检查他们的日程安排。”
|
||||
|
||||
4\. 采取行动：编排层执行计划的第一步具体操作。它会选择并调用适当的工具——无论是调用API、运行代码函数，还是查询数据库。这是代理基于自身内部推理，真正作用于外部世界的行为。
|
||||
|
||||
5\. 观察并迭代：智能体观察其行动的结果。get\_team\_roster工具会返回一个包含五个名字的列表。这些新信息将被添加到智能体的上下文或“记忆”中。随后，循环再次启动，回到步骤3：“现在我已获得名单，下一步是查询日历，确认这五个人的日程安排。我将使用calendar\_api。”
|
||||
|
||||
|
||||
|
||||
而真正的生产系统会叠加所 有三个： **用 LLM 进行推理** **，用 RAG 确保准确性，以及用Agent框架实现自主性。**
|
||||
|
||||
**使用 LLM 单独处理纯语言任务时：写作、摘要、解释。**
|
||||
|
||||
**当准确性至关重要时添加 RAG：从内部文档、技术手册、特定领域知识中回答。**
|
||||
|
||||
**需要真正自主性时部署 Agents：能够决策、行动和管理复杂工作流的系统。**
|
||||
|
||||
未来不在于选择其一。而在于将三者结合起来进行架构设计。
|
||||
|
||||
用于思考的 LLMs。
|
||||
|
||||
用于认知的 RAG。
|
||||
|
||||
用于执行的Agent。
|
||||
|
||||
由此才能够构建出AI智能时代
|
||||
|
||||
|
||||
77
AI/Nano Banana 提示词框架.md
Normal file
77
AI/Nano Banana 提示词框架.md
Normal file
@@ -0,0 +1,77 @@
|
||||
#ai #nano-banana #google #prompt
|
||||
|
||||
物件描述框架
|
||||
|
||||
``` JSON
|
||||
{
|
||||
"shot": "",
|
||||
"subject": {
|
||||
"item": "",
|
||||
"materials": "",
|
||||
"details": "",
|
||||
"condition": ""
|
||||
},
|
||||
"environment": "",
|
||||
"lighting": "",
|
||||
"camera": {
|
||||
"focal_length": "",
|
||||
"aperture": "",
|
||||
"angle": ""
|
||||
},
|
||||
"color_grade": "",
|
||||
"style": "",
|
||||
"quality": "",
|
||||
"negatives": ""
|
||||
}
|
||||
|
||||
```
|
||||
|
||||
人物描述框架
|
||||
|
||||
``` JSON
|
||||
{
|
||||
"shot": "",
|
||||
"subject": {
|
||||
"age": "",
|
||||
"appearance": "",
|
||||
"pose": ""
|
||||
},
|
||||
"environment": "",
|
||||
"lighting": "",
|
||||
"camera": {
|
||||
"focal_length": "",
|
||||
"aperture": "",
|
||||
"angle": ""
|
||||
},
|
||||
"color_grade": "",
|
||||
"style": "",
|
||||
"quality": "",
|
||||
"negatives": ""
|
||||
}
|
||||
|
||||
```
|
||||
|
||||
![[IMG-20260315173031658.png]]
|
||||
``` JSON
|
||||
{
|
||||
"shot": "Macro close-up shot, square aspect ratio (1:1), centered composition.",
|
||||
"subject": {
|
||||
"item": "A luxury men's chronograph watch.",
|
||||
"materials": "Polished stainless steel case, sapphire crystal glass, black ceramic bezel with a tachymeter scale, leather strap with fine stitching.",
|
||||
"details": "White dial with three sub-dials, glowing lume on hands and hour markers, intricate gears of the movement visible through a transparent caseback.",
|
||||
"condition": "Pristine, brand new, no dust or fingerprints."
|
||||
},
|
||||
"environment": "The watch is resting on a dark, textured slab of slate rock. The background is a simple, dark, out-of-focus gradient.",
|
||||
"lighting": "Studio softbox lighting. A key light from the top-left creates clean, sharp reflections on the steel. A soft fill light from the right reveals details in the shadows. A subtle rim light separates the watch from the dark background.",
|
||||
"camera": {
|
||||
"focal_length": "100mm macro lens look",
|
||||
"aperture": "f/8 (to keep the entire watch face in focus)",
|
||||
"angle": "Shot from a 45-degree angle above the watch."
|
||||
},
|
||||
"color_grade": "High contrast, clean and commercial look. Slightly desaturated to emphasize the metallic and monochrome textures. High clarity and sharpness.",
|
||||
"style": "Hyper-realistic CGI render, commercial product photography, luxury and precision.",
|
||||
"quality": "8K resolution, perfect material shaders, flawless reflections, extreme detail on the dial and gears.",
|
||||
"negatives": "no scratches, no dust, no logos or brand names, no human hands, blurry watch face, unrealistic lighting."
|
||||
}
|
||||
|
||||
```
|
||||
42
AI/Never write another prompt.md
Normal file
42
AI/Never write another prompt.md
Normal file
@@ -0,0 +1,42 @@
|
||||
---
|
||||
title: Never write another prompt
|
||||
source: https://youtu.be/OkaplCDf7Ac?si=Fez6aDN0PxfLiM0C
|
||||
author:
|
||||
created: 2025-03-06
|
||||
description:
|
||||
tags:
|
||||
- prompt
|
||||
- "#note-gpt"
|
||||
---
|
||||
https://youtu.be/OkaplCDf7Ac?si=Fez6aDN0PxfLiM0C
|
||||
|
||||
Summary
|
||||
In this video, the presenter introduces a revolutionary tool that simplifies the process of creating effective prompts for AI applications such as ChatGPT and Google Gemini. This tool is particularly beneficial for those who have struggled to formulate precise prompts, often resulting in frustration or inadequate responses from AI. The presenter explains how the tool works, emphasizing its ability to transform basic descriptions into detailed and structured prompts—often referred to as ‘prompt engineering’. This new approach alleviates the need for users to spend significant amounts of money on professional prompt creation services. Additionally, the video covers how to set up the tool, generate prompts, utilize variables, and refine prompts for better outputs. The presenter also offers a resource for viewers to download a list of useful AI prompts, aiding them in harnessing the full potential of AI tools.
|
||||
|
||||
Highlights
|
||||
🛠️ Prompt Engineering Simplified: The tool allows users to generate detailed prompts from simple descriptions, eliminating the complexity of traditional prompt engineering.
|
||||
💰 Cost-Effective Solution: Users can create unlimited prompts without paying exorbitant fees, which can range from $100 to $500 for a single well-crafted prompt.
|
||||
🔑 Easy Setup Process: The video provides a step-by-step guide on creating an account, generating an API key, and setting up payment options for the tool.
|
||||
⚙️ Enhanced Output Quality: The tool generates high-quality prompts that are well-structured and easy to edit, improving the quality of responses from AI applications.
|
||||
🎯 User-Friendly Interface: The interface allows for straightforward editing, including the ability to use variables for better customization of responses.
|
||||
📚 Access to Prompt Libraries: The presenter mentions prompt libraries available on different platforms, enabling users to find inspiration and ready-made prompts for various tasks.
|
||||
📥 Free Resource Available: A downloadable list of useful AI prompts is available on the presenter’s website, further assisting users in their AI interactions.
|
||||
Key Insights
|
||||
🌟 Understanding Prompt Engineering: Prompt engineering is the art of crafting prompts that elicit specific responses from AI. With the introduction of this tool, users no longer need to be experts in this field; the tool automates the process, making it accessible to everyone, regardless of their technical background. This democratization of technology is vital in empowering more individuals to leverage AI effectively.
|
||||
|
||||
💡 The Value of Detailed Prompts: Detailed prompts often yield better responses from AI models. The tool enhances basic prompts by adding context and structure, which helps in narrowing down the AI’s focus. This ensures that the output aligns closely with the user’s expectations, reducing the back-and-forth typically associated with vague or poorly constructed prompts.
|
||||
|
||||
🛡️ Security and Privacy Considerations: When creating an API key, users are reminded to keep it confidential. This highlights an important consideration in the use of AI tools—protection of personal and sensitive information. Users should remain vigilant about their data security, particularly when engaging with cloud-based services.
|
||||
|
||||
💳 Cost Management with AI Tools: The presenter notes that generating prompts may incur minimal costs, emphasizing the importance of understanding pricing structures associated with AI tools. This knowledge helps users manage their expenses effectively while still benefiting from advanced AI capabilities.
|
||||
|
||||
🧩 Customization Through Variables: The ability to use variables in prompts allows for a high degree of customization. This feature enables users to tailor responses to their specific needs without having to rewrite prompts from scratch. The ease of inserting variables enhances the user experience and increases the practicality of the prompts generated.
|
||||
|
||||
📊 Prompt Libraries as Resources: The existence of prompt libraries on various platforms serves as a valuable resource for users looking for inspiration or ready-made prompts. These libraries can significantly reduce the time and effort spent on prompt creation, allowing users to focus on the content and context of their interactions with AI.
|
||||
|
||||
📈 Long-term Efficiency in Prompt Usage: Once a user generates a successful prompt, they can save it for future use, leading to long-term efficiency in their interactions with AI. This practice not only streamlines workflows but also aids in building a personal library of effective prompts tailored to specific tasks, enhancing the overall productivity of users in their AI engagements.
|
||||
|
||||
In conclusion, this video serves as an essential guide for anyone looking to enhance their interaction with AI tools. By utilizing the newly introduced prompt generator, users can streamline the process of prompt creation, save on costs, and ultimately, improve the quality of the responses they receive from AI systems. The combination of user-friendliness, cost-effectiveness, and enhanced output quality makes this tool a game-changer in the realm of AI utilization.
|
||||
|
||||
**Console Anthropic**
|
||||
https://console.anthropic.com/
|
||||
49
AI/OpenAI ChatGPT 个性化定义.md
Normal file
49
AI/OpenAI ChatGPT 个性化定义.md
Normal file
@@ -0,0 +1,49 @@
|
||||
|
||||
|
||||
#openai #ai #chatgpt #customization
|
||||
|
||||
## 自定义指令
|
||||
|
||||
- 高度有条理
|
||||
- 尽可能提出我没想到的解决方案
|
||||
- 主动出击，预判我的需求
|
||||
- 把我当成所有领域的专家
|
||||
- 错误会削弱我的信任，所以务必做到准确和详尽
|
||||
- 提供详细的解释，我喜欢细节丰富的解释
|
||||
- 重视合理的论据，而非权威，来源无关紧要
|
||||
- 考虑新技术和反对观点，而不仅仅是传统智慧
|
||||
- 你可以使用高度推测或预测，但请告诉我
|
||||
- 不进行道德说教
|
||||
- 只有在至关重要且并非显而易见的情况下才讨论安全问题
|
||||
- 如果您的内容政策存在问题，请提供最接近可接受的答复并解释内容政策问题所在
|
||||
- 尽可能引用来源，如果可以，请提供网址
|
||||
- 请将 URL 列表放在回复末尾，不要直接写在回复中
|
||||
- 直接链接到产品，而非公司页面
|
||||
- 无需提及你的知识门槛
|
||||
- 无需透露你是人工智能
|
||||
- 如果由于我的自定义指示导致您的回复质量大幅下降，请解释一下问题所在- Highly organized
|
||||
|
||||
- Suggest solutions that I didn't expect as much as possible
|
||||
- Take the initiative to anticipate my needs
|
||||
- Think of me as an expert in all fields
|
||||
- Mistakes can erode my trust, so be accurate and detailed
|
||||
- Provide detailed explanations. I like detailed explanations
|
||||
- Value sound arguments over authority, and sources are irrelevant
|
||||
- Consider new technologies and opposing perspectives, not just conventional wisdom
|
||||
- You can use alloy speculation or prediction, but let me know
|
||||
- Do not preach morality
|
||||
- Discuss security only when it is critical and not obvious
|
||||
- If you have a content policy issue, provide the closest acceptable response and explain what the content policy issue is
|
||||
- Cite sources whenever possible, and provide URLs if you can
|
||||
- Please put the list of URLs at the end of your reply and don't write it directly in your reply
|
||||
- Links directly to products, not company pages
|
||||
- No need to mention your knowledge threshold
|
||||
- No need to reveal that you are an AI
|
||||
- If the quality of your response has dropped significantly due to my custom instructions, please explain the problem
|
||||
|
||||
|
||||
## 你的详情
|
||||
|
||||
我今年 47 岁，刚从一家企业级软件公司离职。目前是自由职业者。我之前的职位是云服务交付高级经理。我手下有近 20 名员工，分布在全球各地。我们团队的主要职责是为客户提供云服务，并负责公司企业级 SaaS 产品的运维。所以我有很强的技术背景。目前我自己成立了一家公司专注在TikTok跨境电商领域，希望能更利用现在的AI, 自动化，云等技术来帮助业务拓展和销售。
|
||||
|
||||
I'm 47 years old and have just left an enterprise software company. Currently freelancing. My previous position was Senior Manager of Cloud Service Delivery. I have nearly 20 employees all over the world. Our team's primary responsibility is to provide cloud services to customers and to operate the company's enterprise-grade SaaS products. So I have a strong technical background. At present, I have set up a company focusing on the field of TikTok cross-border e-commerce, hoping to make more use of the current AI, automation, cloud and other technologies to help business expansion and sales.
|
||||
0
AI/Untitled.md
Normal file
0
AI/Untitled.md
Normal file
BIN
AI/openclaw/@eaDir/OpenClaw 多 Agent 系统设计.md@SynoEAStream
Normal file
BIN
AI/openclaw/@eaDir/OpenClaw 多 Agent 系统设计.md@SynoEAStream
Normal file
Binary file not shown.
BIN
AI/openclaw/@eaDir/OpenClaw-Agent管理指南.md@SynoEAStream
Normal file
BIN
AI/openclaw/@eaDir/OpenClaw-Agent管理指南.md@SynoEAStream
Normal file
Binary file not shown.
212
AI/openclaw/GOG-CLI-安装配置指南.md
Normal file
212
AI/openclaw/GOG-CLI-安装配置指南.md
Normal file
@@ -0,0 +1,212 @@
|
||||
#gog #gog-cli #macos
|
||||
|
||||
本文档记录在 macOS 系统上安装和配置 gog CLI 的完整步骤，以便通过命令行管理 Google Workspace（Gmail、Google Calendar、Google Drive、Google Contacts、Google Docs、Google Sheets）。
|
||||
|
||||
## 目录
|
||||
|
||||
- [前置条件](#前置条件)
|
||||
- [安装步骤](#安装步骤)
|
||||
- [配置 OAuth 凭证](#配置-oauth-凭证)
|
||||
- [解除 Google 安全限制](#解除-google-安全限制)
|
||||
- [验证配置](#验证配置)
|
||||
- [常用命令](#常用命令)
|
||||
- [故障排除](#故障排除)
|
||||
|
||||
---
|
||||
|
||||
## 前置条件
|
||||
|
||||
- macOS 系统
|
||||
- Homebrew 已安装
|
||||
- Google 账号
|
||||
|
||||
---
|
||||
|
||||
## 安装步骤
|
||||
|
||||
### 1. 安装 gog CLI
|
||||
|
||||
使用 Homebrew 安装 gog CLI：
|
||||
|
||||
```bash
|
||||
brew install steipete/tap/gogcli
|
||||
```
|
||||
|
||||
验证安装：
|
||||
|
||||
```bash
|
||||
which gog
|
||||
# 输出: /opt/homebrew/bin/gog
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 配置 OAuth 凭证
|
||||
|
||||
### 1. 在 Google Cloud Console 创建 OAuth 凭证
|
||||
|
||||
1. 打开 [Google Cloud Console - Credentials](https://console.cloud.google.com/apis/credentials)
|
||||
2. 点击 **「创建凭证」** → 选择 **「OAuth 客户端 ID」**
|
||||
3. 应用类型选择 **「桌面应用」**
|
||||
4. 命名（例如：`gogcli`）
|
||||
5. 点击 **「创建」**
|
||||
6. 点击 **「下载 JSON」**，得到 `credentials.json` 文件
|
||||
|
||||
### 2. 移动凭证文件到 gogcli 配置目录
|
||||
|
||||
创建 gogcli 配置目录（如果不存在）：
|
||||
|
||||
```bash
|
||||
mkdir -p "/Users/weishen/Library/Application Support/gogcli"
|
||||
```
|
||||
|
||||
移动下载的凭证文件：
|
||||
|
||||
```bash
|
||||
mv ~/Downloads/credentials.json "/Users/weishen/Library/Application Support/gogcli/credentials.json"
|
||||
```
|
||||
|
||||
或者使用命令指定凭证路径：
|
||||
|
||||
```bash
|
||||
gog auth credentials /path/to/credentials.json
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 解除 Google 安全限制
|
||||
|
||||
### 问题描述
|
||||
|
||||
首次授权时，Google 会显示以下错误：
|
||||
|
||||
> 此应用未经 Google 验证
|
||||
> 此应用请求访问您 Google 账号中的敏感信息。在开发者让该应用通过 Google 验证之前，请勿使用该应用。
|
||||
|
||||
### 解决方法：添加测试用户
|
||||
|
||||
1. 打开 [Google Cloud Console - Credentials](https://console.cloud.google.com/apis/credentials)
|
||||
2. 找到你创建的 OAuth 客户端ID，点击进入详情
|
||||
3. 找到 **「测试用户」** 部分
|
||||
4. 点击 **「添加用户」**
|
||||
5. 输入你的 Google 邮箱：`ishenwei@gmail.com`
|
||||
6. 保存
|
||||
|
||||
添加测试用户后，重新运行授权命令即可：
|
||||
|
||||
```bash
|
||||
gog auth add ishenwei@gmail.com --services gmail,calendar,drive,contacts,docs,sheets
|
||||
```
|
||||
|
||||
这会打开浏览器让你登录 Google 账号并授权。
|
||||
|
||||
---
|
||||
|
||||
## 验证配置
|
||||
|
||||
### 1. 查看已授权的账号
|
||||
|
||||
```bash
|
||||
gog auth list
|
||||
```
|
||||
|
||||
### 2. 测试 Gmail
|
||||
|
||||
```bash
|
||||
gog gmail search "newer_than:1d" --max 5
|
||||
```
|
||||
|
||||
### 3. 测试 Calendar
|
||||
|
||||
```bash
|
||||
gog calendar events primary --from 2026-01-01 --to 2026-12-31
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 常用命令
|
||||
|
||||
### Gmail
|
||||
|
||||
| 功能 | 命令 |
|
||||
|------|------|
|
||||
| 搜索邮件 | `gog gmail search 'newer_than:7d' --max 10` |
|
||||
| 发送邮件 | `gog gmail send --to a@b.com --subject "Hi" --body "Hello"` |
|
||||
| 发送邮件（多行） | `gog gmail send --to a@b.com --subject "Hi" --body-file ./message.txt` |
|
||||
| 创建草稿 | `gog gmail drafts create --to a@b.com --subject "Hi" --body-file ./message.txt` |
|
||||
| 发送草稿 | `gog gmail drafts send <draftId>` |
|
||||
|
||||
### Calendar
|
||||
|
||||
| 功能 | 命令 |
|
||||
|------|------|
|
||||
| 查看事件 | `gog calendar events <calendarId> --from <iso> --to <iso>` |
|
||||
| 创建事件 | `gog calendar create <calendarId> --summary "Title" --from <iso> --to <iso>` |
|
||||
| 查看颜色 | `gog calendar colors` |
|
||||
|
||||
### Drive
|
||||
|
||||
| 功能 | 命令 |
|
||||
|------|------|
|
||||
| 搜索文件 | `gog drive search "query" --max 10` |
|
||||
|
||||
### Contacts
|
||||
|
||||
| 功能 | 命令 |
|
||||
|------|------|
|
||||
| 列出联系人 | `gog contacts list --max 20` |
|
||||
|
||||
### Sheets
|
||||
|
||||
| 功能 | 命令 |
|
||||
|------|------|
|
||||
| 获取数据 | `gog sheets get <sheetId> "Tab!A1:D10" --json` |
|
||||
| 更新数据 | `gog sheets update <sheetId> "Tab!A1:B2" --values-json '[["A","B"],["1","2"]]' --input USER_ENTERED` |
|
||||
|
||||
### Docs
|
||||
|
||||
| 功能 | 命令 |
|
||||
|------|------|
|
||||
| 导出文档 | `gog docs export <docId> --format txt --out /tmp/doc.txt` |
|
||||
| 查看内容 | `gog docs cat <docId>` |
|
||||
|
||||
---
|
||||
|
||||
## 故障排除
|
||||
|
||||
### 凭证文件路径错误
|
||||
|
||||
确保凭证文件在以下位置：
|
||||
```
|
||||
/Users/weishen/Library/Application Support/gogcli/credentials.json
|
||||
```
|
||||
|
||||
### 需要重新授权
|
||||
|
||||
删除现有授权并重新授权：
|
||||
|
||||
```bash
|
||||
gog auth remove ishenwei@gmail.com
|
||||
gog auth add ishenwei@gmail.com --services gmail,calendar,drive,contacts,docs,sheets
|
||||
```
|
||||
|
||||
### 设置默认账号
|
||||
|
||||
避免每次重复指定账号：
|
||||
|
||||
```bash
|
||||
export GOG_ACCOUNT=ishenwei@gmail.com
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 参考链接
|
||||
|
||||
- gog 官网: https://gogcli.sh
|
||||
- gog GitHub: https://github.com/steipete/gogcli
|
||||
- Google Cloud Console: https://console.cloud.google.com/
|
||||
|
||||
---
|
||||
|
||||
*文档创建日期: 2026-03-15*
|
||||
*最后更新: 2026-03-15*
|
||||
326
AI/openclaw/OpenClaw Agent 命名与架构设计参考文档.md
Normal file
326
AI/openclaw/OpenClaw Agent 命名与架构设计参考文档.md
Normal file
@@ -0,0 +1,326 @@
|
||||
#openclaw #agent
|
||||
|
||||
```table-of-contents
|
||||
```
|
||||
|
||||
# 1. 架构设计目标
|
||||
|
||||
该 Agent 架构基于 **OpenClaw 多节点智能体系统**，通过不同服务器部署不同职责的 Agent，并通过统一命名体系构建一个清晰、可扩展的 AI Agent 生态。
|
||||
|
||||
整体设计目标：
|
||||
|
||||
- 形成 **清晰的职责分层**
|
||||
- 建立 **统一的命名体系**
|
||||
- 支持 **未来扩展更多 Agent**
|
||||
- 便于 **星枢统一调度**
|
||||
|
||||
设计采用 **三层体系结构**：
|
||||
|
||||
| 层级 | 系列 | 含义 | 主要职责 |
|
||||
| --- | --- | ---- | ------------ |
|
||||
| 控制层 | 星系 | 星辰统御 | 调度、管理、智能决策 |
|
||||
| 技术层 | 云系 | 云海算力 | 开发、架构、监控 |
|
||||
| 执行层 | 风系 | 风行万里 | 测试、业务执行、流程处理 |
|
||||
|
||||
这种结构类似于 **AI Agent 操作系统**：
|
||||
```
|
||||
控制层（星）
|
||||
↓
|
||||
技术层（云）
|
||||
↓
|
||||
执行层（风）
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
# 2. 当前 Agent 架构
|
||||
|
||||
## 2.1 Mac Mini（中央控制节点）
|
||||
|
||||
定位：
|
||||
|
||||
- AI Agent 中枢
|
||||
- 调度中心
|
||||
- 个人与 IT 管理
|
||||
|
||||
|Agent|名字|角色|职责|
|
||||
|---|---|---|---|
|
||||
|星枢|Master Orchestrator|总调度|统一调度所有 Agent|
|
||||
|星曜|IT 管家|IT 管理|服务器、环境、运维|
|
||||
|星辉|个人助理|Assistant|日常任务与个人事务|
|
||||
|
||||
架构示意：
|
||||
|
||||
```
|
||||
Mac Mini
|
||||
├─ 星枢（总调度）
|
||||
├─ 星曜（IT管家）
|
||||
└─ 星辉（个人助理）
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
# 3. Ubuntu2（开发服务器）
|
||||
|
||||
定位：
|
||||
|
||||
- 技术研发
|
||||
- 架构设计
|
||||
- 自动化构建
|
||||
- 系统监控
|
||||
|
||||
Agent 命名统一以 **“云”开头**。
|
||||
|
||||
## 3.1 已有 Agent
|
||||
|
||||
| Agent | 职责 | |
|
||||
| ----- | ----------------------------------- | ---------------- |
|
||||
| 云瀚 | 监控系统 | 云海浩瀚，象征监控全局系统状态。 |
|
||||
| 云策 | - 架构设计<br>- 技术方案<br>- 系统规划 | 云中筹策，技术谋略。 |
|
||||
| 云匠 | - 代码开发<br>- 构建<br>- 工程实现 | 云端工匠 |
|
||||
| 云织 | - CI/CD<br>- 自动化流程<br>- pipeline 编排 | 织云为网 |
|
||||
|
||||
|
||||
---
|
||||
|
||||
## 3.3 Ubuntu2 最终推荐结构
|
||||
|
||||
```
|
||||
Ubuntu2（开发服务器）
|
||||
|
||||
云瀚 监控
|
||||
云策 架构设计
|
||||
云匠 开发实现
|
||||
云织 CI/CD自动化
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
# 4. Ubuntu1（准生产服务器）
|
||||
|
||||
定位：
|
||||
|
||||
- QA 测试
|
||||
- 业务执行
|
||||
- 自动任务
|
||||
- 审计规则
|
||||
|
||||
Agent 统一使用 **“风”系列命名**。
|
||||
|
||||
原因：
|
||||
|
||||
> 风代表执行、速度、行动。
|
||||
|
||||
| Agent | 职责 | |
|
||||
| ----- | ---------------------------- | ----- |
|
||||
| 风衡 | - QA 测试<br>- 自动测试<br>- 质量控制 | 风中权衡。 |
|
||||
| 风驰 | - 自动任务<br>- Job 执行<br>- 业务流程 | 风驰电掣。 |
|
||||
| 风纪 | - 规则执行<br>- 审计<br>- 合规 | 风纪法度 |
|
||||
|
||||
---
|
||||
|
||||
## 4.2 Ubuntu1 推荐结构
|
||||
|
||||
```
|
||||
Ubuntu1（准生产服务器）
|
||||
|
||||
风衡 QA测试
|
||||
风驰 自动执行
|
||||
风纪 规则审计
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
# 5. 完整 Agent 架构
|
||||
|
||||
最终整体结构：
|
||||
|
||||
```
|
||||
星枢
|
||||
（总调度 Agent）
|
||||
│
|
||||
┌────────────┼────────────┐
|
||||
│ │ │
|
||||
星曜 星辉 服务器集群
|
||||
（IT管家） （个人助理） │
|
||||
│
|
||||
┌────────────┴────────────┐
|
||||
│ │
|
||||
Ubuntu2 Ubuntu1
|
||||
（开发服务器） （准生产）
|
||||
│ │
|
||||
┌───────┼───────┐ ┌───────┼───────┐
|
||||
│ │ │ │ │ │
|
||||
云瀚 云策 云匠 风衡 风驰 风纪
|
||||
监控 架构 开发 QA 执行 审计
|
||||
│
|
||||
云织
|
||||
CI/CD
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
# 6. Agent 体系总结
|
||||
|
||||
|系列|含义|角色类型|
|
||||
|---|---|---|
|
||||
|星|星辰|调度 / 管理|
|
||||
|云|云海|技术 / 开发|
|
||||
|风|风行|执行 / 流程|
|
||||
|
||||
---
|
||||
|
||||
## 星系 Agent
|
||||
|
||||
|Agent|职责|
|
||||
|---|---|
|
||||
|星枢|总调度|
|
||||
|星曜|IT 管理|
|
||||
|星辉|助手|
|
||||
|
||||
未来可扩展：
|
||||
|
||||
|Agent|角色|
|
||||
|---|---|
|
||||
|星策|战略规划|
|
||||
|星典|知识管理|
|
||||
|
||||
---
|
||||
|
||||
## 云系 Agent
|
||||
|
||||
|Agent|职责|
|
||||
|---|---|
|
||||
|云瀚|监控|
|
||||
|云策|架构|
|
||||
|云匠|开发|
|
||||
|云织|CI/CD|
|
||||
|
||||
---
|
||||
|
||||
## 风系 Agent
|
||||
|
||||
|Agent|职责|
|
||||
|---|---|
|
||||
|风衡|QA|
|
||||
|风驰|自动执行|
|
||||
|风纪|审计|
|
||||
|
||||
---
|
||||
|
||||
# 7. 设计优势
|
||||
|
||||
该 Agent 命名体系具有以下优点：
|
||||
|
||||
### 1. 语义清晰
|
||||
|
||||
看到名字即可理解职责。
|
||||
|
||||
例如：
|
||||
|
||||
- 云匠 → 开发
|
||||
|
||||
- 风衡 → QA
|
||||
|
||||
- 星枢 → 调度
|
||||
|
||||
|
||||
---
|
||||
|
||||
### 2. 层级清晰
|
||||
|
||||
```
|
||||
星（控制）
|
||||
↓
|
||||
云（技术）
|
||||
↓
|
||||
风（执行）
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
### 3. 易于扩展
|
||||
|
||||
未来可以继续增加：
|
||||
|
||||
云系：
|
||||
|
||||
- 云图（数据）
|
||||
|
||||
- 云阵（基础设施）
|
||||
|
||||
- 云算（AI计算）
|
||||
|
||||
|
||||
风系：
|
||||
|
||||
- 风策（业务策略）
|
||||
|
||||
- 风行（业务执行）
|
||||
|
||||
- 风巡（巡检）
|
||||
|
||||
|
||||
---
|
||||
|
||||
### 4. 非常适合 Agent 调度
|
||||
|
||||
星枢可以统一调度：
|
||||
|
||||
```
|
||||
星枢 → 云系 → 风系
|
||||
```
|
||||
|
||||
示例：
|
||||
|
||||
```
|
||||
星枢
|
||||
↓
|
||||
云策（制定方案）
|
||||
↓
|
||||
云匠（开发）
|
||||
↓
|
||||
云织（部署）
|
||||
↓
|
||||
风衡（测试）
|
||||
↓
|
||||
风驰（执行）
|
||||
```
|
||||
|
||||
形成完整 **AI 自动化流水线**。
|
||||
|
||||
---
|
||||
|
||||
# 8. 最终推荐部署
|
||||
|
||||
### Mac Mini
|
||||
|
||||
```
|
||||
星枢
|
||||
星曜
|
||||
星辉
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
### Ubuntu2（开发）
|
||||
|
||||
```
|
||||
云瀚
|
||||
云策
|
||||
云匠
|
||||
云织
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
### Ubuntu1（准生产）
|
||||
|
||||
```
|
||||
风衡
|
||||
风驰
|
||||
风纪
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
574
AI/openclaw/OpenClaw 多 Agent 系统设计.md
Normal file
574
AI/openclaw/OpenClaw 多 Agent 系统设计.md
Normal file
@@ -0,0 +1,574 @@
|
||||
|
||||
#openclaw #agent #telegram
|
||||
```table-of-contents
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
# OpenClaw 多 Agent 自动化系统架构笔记 (Advanced)
|
||||
|
||||
Author: Billy
|
||||
Purpose: 构建基于 OpenClaw + Telegram + n8n 的 AI 自动化控制系统
|
||||
|
||||
---
|
||||
|
||||
# 1 系统总体架构
|
||||
|
||||
**目标：**
|
||||
|
||||
构建一个 **AI Automation Control Center**
|
||||
|
||||
**核心组件：**
|
||||
|
||||
- OpenClaw Gateway
|
||||
- 多 Agent
|
||||
- Telegram Interface
|
||||
- n8n Workflow Engine
|
||||
|
||||
**系统架构：**
|
||||
|
||||
```
|
||||
Telegram
|
||||
│
|
||||
▼
|
||||
Telegram Bot
|
||||
│
|
||||
▼
|
||||
n8n
|
||||
(Command Router)
|
||||
│
|
||||
▼
|
||||
OpenClaw Gateway
|
||||
│
|
||||
▼
|
||||
┌────────┬────────┬────────┐
|
||||
▼ ▼ ▼ ▼
|
||||
Router Dev Research Ops
|
||||
Agent Agent Agent Agent
|
||||
```
|
||||
|
||||
**系统功能：**
|
||||
|
||||
| 组件 | 作用 |
|
||||
|------|------|
|
||||
| Telegram | 用户交互 |
|
||||
| n8n | 任务路由 |
|
||||
| OpenClaw | AI Agent Runtime |
|
||||
| Agents | 任务执行 |
|
||||
|
||||
---
|
||||
|
||||
# 2 OpenClaw 基础目录结构
|
||||
|
||||
**默认路径：** `~/.openclaw/`
|
||||
|
||||
**结构：**
|
||||
|
||||
```
|
||||
.openclaw/
|
||||
│
|
||||
├── agents/
|
||||
│ │
|
||||
│ ├── main/
|
||||
│ │ └── agent/
|
||||
│ │ ├── authprofile.json
|
||||
│ │ └── models.json
|
||||
│ │
|
||||
│ └── research/
|
||||
│ └── agent/
|
||||
│ ├── authprofile.json
|
||||
│ └── models.json
|
||||
│
|
||||
└── workspace/
|
||||
│
|
||||
├── skills/
|
||||
├── memory/
|
||||
├── identity/
|
||||
│ └── Identity.md
|
||||
├── logs/
|
||||
├── devices/
|
||||
├── completions/
|
||||
└── canvas/
|
||||
```
|
||||
|
||||
**说明：**
|
||||
|
||||
| 目录 | 作用 |
|
||||
|------|------|
|
||||
| agents | Agent Profile |
|
||||
| workspace | 运行数据 |
|
||||
| skills | 技能插件 |
|
||||
| memory | AI记忆 |
|
||||
| identity | System Prompt |
|
||||
| logs | 日志 |
|
||||
|
||||
---
|
||||
|
||||
# 3 多 Agent 设计
|
||||
|
||||
**创建 Agent：**
|
||||
|
||||
```bash
|
||||
openclaw agent create dev
|
||||
openclaw agent create research
|
||||
openclaw agent create ops
|
||||
openclaw agent create orchestrator
|
||||
```
|
||||
|
||||
**推荐 Agent 列表：**
|
||||
|
||||
| Agent | 职责 |
|
||||
|-------|------|
|
||||
| orchestrator | 任务调度 |
|
||||
| dev | 编程 |
|
||||
| research | 搜索分析 |
|
||||
| ops | 服务器运维 |
|
||||
| data | 数据处理 |
|
||||
|
||||
**推荐结构：**
|
||||
|
||||
```
|
||||
agents/
|
||||
│
|
||||
├── orchestrator/
|
||||
├── dev/
|
||||
├── research/
|
||||
├── ops/
|
||||
└── data/
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
# 4 Agent Identity 设计
|
||||
|
||||
Identity 代表：System Prompt
|
||||
|
||||
**建议放在：** `agents/dev/agent/identity.md`
|
||||
|
||||
**示例：**
|
||||
|
||||
```
|
||||
You are DevAgent.
|
||||
|
||||
Responsibilities:
|
||||
- write code
|
||||
- debug programs
|
||||
- generate scripts
|
||||
- create docker configurations
|
||||
```
|
||||
|
||||
**Research Agent：**
|
||||
|
||||
```
|
||||
You are ResearchAgent.
|
||||
|
||||
Responsibilities:
|
||||
- research information
|
||||
- summarize technical topics
|
||||
- analyze documents
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
# 5 Skills 机制
|
||||
|
||||
**Skills 目录：** `workspace/skills`
|
||||
|
||||
**示例：**
|
||||
|
||||
```
|
||||
skills/
|
||||
├── browser
|
||||
├── filesystem
|
||||
├── python
|
||||
├── telegram
|
||||
└── self-improving-agent
|
||||
```
|
||||
|
||||
**特点：**
|
||||
|
||||
- 所有 Agent 共享
|
||||
- 类似插件
|
||||
|
||||
---
|
||||
|
||||
# 6 Memory 设计
|
||||
|
||||
**默认 memory：** `workspace/memory`
|
||||
|
||||
默认共享。
|
||||
|
||||
**memory 类型：**
|
||||
|
||||
```
|
||||
memory/
|
||||
├── episodic
|
||||
├── semantic
|
||||
└── vector
|
||||
```
|
||||
|
||||
**推荐 Memory 分层：**
|
||||
|
||||
```
|
||||
memory/
|
||||
├── system
|
||||
├── dev
|
||||
├── research
|
||||
└── ops
|
||||
```
|
||||
|
||||
这样避免：memory pollution
|
||||
|
||||
**Agent 指定 namespace：**
|
||||
|
||||
```json
|
||||
{
|
||||
"memoryNamespace": "dev"
|
||||
}
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
# 7 模型配置
|
||||
|
||||
每个 Agent 独立：`agents/<agent>/agent/models.json`
|
||||
|
||||
**示例：**
|
||||
|
||||
```json
|
||||
{
|
||||
"default": "gpt-4o-mini",
|
||||
"models": {
|
||||
"gpt-4o-mini": {
|
||||
"provider": "openai",
|
||||
"model": "gpt-4o-mini"
|
||||
},
|
||||
"deepseek-coder": {
|
||||
"provider": "openai-compatible",
|
||||
"base_url": "https://api.deepseek.com/v1",
|
||||
"model": "deepseek-coder"
|
||||
}
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
**推荐模型分配：**
|
||||
|
||||
| Agent | Model |
|
||||
|-------|-------|
|
||||
| orchestrator | gpt-4o-mini |
|
||||
| dev | deepseek-coder |
|
||||
| research | gpt-4o |
|
||||
| ops | claude-sonnet |
|
||||
|
||||
**优点：**
|
||||
|
||||
- 降低成本
|
||||
- 提升能力
|
||||
|
||||
---
|
||||
|
||||
# 8 Agent 切换
|
||||
|
||||
**切换 active agent：**
|
||||
|
||||
```bash
|
||||
openclaw agent use research
|
||||
```
|
||||
|
||||
**行为：**
|
||||
|
||||
```
|
||||
current_agent = research
|
||||
```
|
||||
|
||||
下一条消息：
|
||||
|
||||
```
|
||||
Telegram → research agent
|
||||
```
|
||||
|
||||
不需要重启 Gateway。
|
||||
|
||||
---
|
||||
|
||||
# 9 Telegram 集成
|
||||
|
||||
OpenClaw 支持：Telegram Bot
|
||||
|
||||
**结构：**
|
||||
|
||||
```
|
||||
Telegram
|
||||
│
|
||||
▼
|
||||
OpenClaw Gateway
|
||||
```
|
||||
|
||||
**默认：** 所有消息 → active agent
|
||||
|
||||
---
|
||||
|
||||
# 10 Telegram 指令体系
|
||||
|
||||
**推荐命令：**
|
||||
|
||||
```
|
||||
/dev
|
||||
/research
|
||||
/ops
|
||||
/data
|
||||
/help
|
||||
```
|
||||
|
||||
**示例：**
|
||||
|
||||
```
|
||||
/dev 写一个python脚本
|
||||
/research 查一下OpenClaw架构
|
||||
/ops restart docker
|
||||
```
|
||||
|
||||
在 BotFather 设置：`/setcommands`
|
||||
|
||||
**示例：**
|
||||
|
||||
```
|
||||
dev - coding tasks
|
||||
research - research tasks
|
||||
ops - server tasks
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
# 11 n8n Command Router
|
||||
|
||||
推荐使用：n8n 进行任务路由。
|
||||
|
||||
**架构：**
|
||||
|
||||
```
|
||||
Telegram
|
||||
│
|
||||
▼
|
||||
n8n
|
||||
│
|
||||
▼
|
||||
OpenClaw
|
||||
```
|
||||
|
||||
**Workflow：**
|
||||
|
||||
```
|
||||
Telegram Trigger
|
||||
│
|
||||
▼
|
||||
Parse Command
|
||||
│
|
||||
▼
|
||||
Agent Router
|
||||
│
|
||||
▼
|
||||
HTTP → OpenClaw
|
||||
```
|
||||
|
||||
**解析示例：**
|
||||
|
||||
```javascript
|
||||
const msg = $json.message.text
|
||||
|
||||
if(msg.startsWith("/dev")){
|
||||
return {agent:"dev"}
|
||||
}
|
||||
|
||||
if(msg.startsWith("/research")){
|
||||
return {agent:"research"}
|
||||
}
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
# 12 Router Agent
|
||||
|
||||
Router Agent 负责：Task Distribution
|
||||
|
||||
**Prompt 示例：**
|
||||
|
||||
```
|
||||
You are a routing agent.
|
||||
|
||||
Rules:
|
||||
/dev → DevAgent
|
||||
/research → ResearchAgent
|
||||
/ops → OpsAgent
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
# 13 推荐系统架构
|
||||
|
||||
Mac Mini AI Control Center：
|
||||
|
||||
```
|
||||
Mac Mini
|
||||
│
|
||||
├─ OpenClaw Gateway
|
||||
├─ n8n
|
||||
└─ Telegram Bot
|
||||
|
||||
Agents
|
||||
│
|
||||
├─ orchestrator
|
||||
├─ dev
|
||||
├─ research
|
||||
├─ ops
|
||||
└─ data
|
||||
```
|
||||
|
||||
**数据流：**
|
||||
|
||||
```
|
||||
Telegram
|
||||
↓
|
||||
n8n Router
|
||||
↓
|
||||
OpenClaw
|
||||
↓
|
||||
Agents
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
# 14 高级玩法
|
||||
|
||||
**支持：**
|
||||
|
||||
### 多服务器 Agent
|
||||
|
||||
```
|
||||
Mac Mini
|
||||
orchestrator
|
||||
|
||||
Ubuntu Server
|
||||
dev agent
|
||||
|
||||
NAS
|
||||
data agent
|
||||
```
|
||||
|
||||
**通信：**
|
||||
|
||||
```
|
||||
HTTP
|
||||
Redis
|
||||
MQ
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
# 15 最佳实践
|
||||
|
||||
### 1 Agent 专业化
|
||||
|
||||
**正确：**
|
||||
|
||||
```
|
||||
dev-agent
|
||||
ops-agent
|
||||
research-agent
|
||||
```
|
||||
|
||||
**错误：**
|
||||
|
||||
```
|
||||
python-agent
|
||||
browser-agent
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
### 2 模型分工
|
||||
|
||||
不要所有 agent 使用同一个模型。
|
||||
|
||||
**否则：**
|
||||
|
||||
```
|
||||
multi-agent ≈ single-agent
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
### 3 Memory 分层
|
||||
|
||||
**推荐：**
|
||||
|
||||
```
|
||||
memory/
|
||||
├── system
|
||||
├── dev
|
||||
├── research
|
||||
└── ops
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
### 4 Router 统一入口
|
||||
|
||||
**不要：**
|
||||
|
||||
```
|
||||
Telegram → 各 agent
|
||||
```
|
||||
|
||||
**应该：**
|
||||
|
||||
```
|
||||
Telegram → Router → Agents
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
# 16 最终架构总结
|
||||
|
||||
**最终系统：**
|
||||
|
||||
```
|
||||
Telegram
|
||||
│
|
||||
▼
|
||||
n8n Router
|
||||
│
|
||||
▼
|
||||
OpenClaw Gateway
|
||||
│
|
||||
▼
|
||||
┌──────────────┬──────────────┬──────────────┐
|
||||
DevAgent ResearchAgent OpsAgent
|
||||
```
|
||||
|
||||
**系统特点：**
|
||||
|
||||
- 多 Agent
|
||||
- 多模型
|
||||
- 自动任务分发
|
||||
- Telegram 控制
|
||||
- 自动化工作流
|
||||
|
||||
---
|
||||
|
||||
# 17 未来扩展
|
||||
|
||||
**未来可扩展：**
|
||||
|
||||
- AI 自动任务
|
||||
- 自动服务器运维
|
||||
- 自动代码生成
|
||||
- 自动数据分析
|
||||
- 自动监控告警
|
||||
|
||||
**目标：** 构建 Personal AI Operations Center
|
||||
|
||||
|
||||
|
||||
|
||||
194
AI/openclaw/OpenClaw-Agent管理指南.md
Normal file
194
AI/openclaw/OpenClaw-Agent管理指南.md
Normal file
@@ -0,0 +1,194 @@
|
||||
# OpenClaw Agent 管理指南
|
||||
|
||||
> 创建日期: 2026-03-15
|
||||
> 作者: 星曜
|
||||
|
||||
---
|
||||
|
||||
## 1. 创建新 Agent
|
||||
|
||||
### 基本命令
|
||||
|
||||
```bash
|
||||
openclaw agents add <agent-name> --non-interactive --workspace <workspace-path>
|
||||
```
|
||||
|
||||
### 参数说明
|
||||
|
||||
| 参数 | 说明 |
|
||||
| ------------------- | ----------------- |
|
||||
| `<agent-name>` | Agent 名称（唯一标识） |
|
||||
| `--non-interactive` | 跳过交互式提示，自动创建 |
|
||||
| `--workspace` | 指定 workspace 目录路径 |
|
||||
|
||||
### 示例
|
||||
|
||||
```bash
|
||||
# 创建名为 xinghui 的 agent
|
||||
openclaw agents add xinghui --non-interactive --workspace ~/.openclaw/workspace-agent-xinghui
|
||||
|
||||
# 创建名为 xingyao 的 agent（复制现有workspace）
|
||||
openclaw agents add xingyao --non-interactive --workspace ~/.openclaw/workspace-agent-xingyao
|
||||
```
|
||||
|
||||
### Workspace 路径规范
|
||||
|
||||
- 格式: `~/.openclaw/workspace-agent-<agent-name>`
|
||||
- 示例: `~/.openclaw/workspace-agent-xinghui`
|
||||
|
||||
---
|
||||
|
||||
## 2. 复制 Workspace 到新 Agent
|
||||
|
||||
### 场景
|
||||
创建一个与现有 agent（通常是 main）拥有相同内容的 workspace
|
||||
|
||||
### 方法一：rsync 复制（推荐）
|
||||
|
||||
```bash
|
||||
rsync -av --exclude='.git' /Users/weishen/.openclaw/workspace/ /Users/weishen/.openclaw/workspace-agent-xingyao/
|
||||
```
|
||||
|
||||
### 方法二：软链接共享记忆
|
||||
|
||||
如果想让多个 agent 共享记忆（MEMORY.md 和 memory/ 目录）：
|
||||
|
||||
```bash
|
||||
# 为新 agent 创建软链接
|
||||
ln -sf /Users/weishen/.openclaw/workspace/memory /Users/weishen/.openclaw/workspace-agent-<agent-name>/memory
|
||||
ln -sf /Users/weishen/.openclaw/workspace/MEMORY.md /Users/weishen/.openclaw/workspace-agent-<agent-name>/MEMORY.md
|
||||
```
|
||||
|
||||
**注意**: 共用 workspace 可能导致配置冲突，建议独立 workspace + 软链接共享 memory 目录
|
||||
|
||||
---
|
||||
|
||||
## 3. 绑定 Agent 到 Channel
|
||||
|
||||
### 查看当前绑定
|
||||
|
||||
```bash
|
||||
openclaw agents bindings
|
||||
```
|
||||
|
||||
### 绑定命令
|
||||
|
||||
```bash
|
||||
openclaw agents bind --agent <agent-name> --bind <channel>:<accountId>
|
||||
```
|
||||
|
||||
### 参数说明
|
||||
|
||||
| 参数 | 说明 |
|
||||
| -------------- | ----------------------------------- |
|
||||
| `<agent-name>` | 要绑定的 agent 名称 |
|
||||
| `<channel>` | 频道类型（telegram, discord, whatsapp 等） |
|
||||
| `<accountId>` | 账号 ID（数字形式） |
|
||||
|
||||
### 示例[[How to get Youtube Channel ID]]
|
||||
|
||||
```bash
|
||||
# 绑定 xinghui 到 Telegram
|
||||
openclaw agents bind --agent xinghui --bind telegram:5038825565
|
||||
|
||||
# 绑定 xingyao 到 Telegram
|
||||
openclaw agents bind --agent xingyao --bind telegram:5038825565
|
||||
```
|
||||
|
||||
### 解绑
|
||||
|
||||
```bash
|
||||
# 解绑指定 channel
|
||||
openclaw agents unbind --bind telegram:5038825565
|
||||
|
||||
# 解绑指定 agent 的 channel
|
||||
openclaw agents unbind --agent xinghui --bind telegram:5038825565
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 4. 删除 Agent
|
||||
|
||||
### 命令
|
||||
|
||||
```bash
|
||||
openclaw agents delete <agent-name>
|
||||
```
|
||||
|
||||
### 示例
|
||||
|
||||
```bash
|
||||
# 需要确认
|
||||
openclaw agents delete agent-macmini-001
|
||||
|
||||
# 强制删除（非交互模式）
|
||||
openclaw agents delete agent-macmini-001 --force
|
||||
```
|
||||
|
||||
### 删除效果
|
||||
- Workspace 目录移至废纸篓
|
||||
- Sessions 目录删除
|
||||
- 从配置文件移除
|
||||
|
||||
---
|
||||
|
||||
## 5. 查看 Agent 列表
|
||||
|
||||
```bash
|
||||
openclaw agents list
|
||||
```
|
||||
|
||||
### 输出示例
|
||||
|
||||
```
|
||||
Agents:
|
||||
- main (default)
|
||||
Workspace: ~/.openclaw/workspace
|
||||
Agent dir: ~/.openclaw/agents/main/agent
|
||||
Model: minimax-portal/MiniMax-M2.5
|
||||
Routing rules: 0
|
||||
Routing: default (no explicit rules)
|
||||
- xinghui
|
||||
Workspace: ~/.openclaw/workspace-agent-xinghui
|
||||
Agent dir: ~/.openclaw/agents/xinghui/agent
|
||||
Model: minimax-portal/MiniMax-M2.5
|
||||
Routing rules: 0
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 6. Agent 配置说明
|
||||
|
||||
### 默认模型
|
||||
所有新创建的 agent 默认使用: `minimax-portal/MiniMax-M2.5`
|
||||
|
||||
### 配置文件位置
|
||||
- 主配置: `~/.openclaw/openclaw.json`
|
||||
- Agent 状态: `~/.openclaw/agents/<agent-name>/agent/`
|
||||
|
||||
---
|
||||
|
||||
## 7. 常用命令速查
|
||||
|
||||
| 操作 | 命令 |
|
||||
|------|------|
|
||||
| 创建 agent | `openclaw agents add <name> --non-interactive --workspace ~/.openclaw/workspace-agent-<name>` |
|
||||
| 查看列表 | `openclaw agents list` |
|
||||
| 查看绑定 | `openclaw agents bindings` |
|
||||
| 绑定 channel | `openclaw agents bind --agent <name> --bind telegram:<id>` |
|
||||
| 解绑 channel | `openclaw agents unbind --bind telegram:<id>` |
|
||||
| 删除 agent | `openclaw agents delete <name> --force` |
|
||||
|
||||
---
|
||||
|
||||
## 8. 当前已创建的 Agent
|
||||
|
||||
| Agent 名称 | Workspace | 状态 |
|
||||
|------------|-----------|------|
|
||||
| main | ~/.openclaw/workspace | 默认 |
|
||||
| xinghui | ~/.openclaw/workspace-agent-xinghui | 独立（无共享记忆） |
|
||||
| xingyao | ~/.openclaw/workspace-agent-xingyao | 完整复制 main workspace |
|
||||
|
||||
---
|
||||
|
||||
*最后更新: 2026-03-15 14:33*
|
||||
368
AI/openclaw/Ubuntu 下 OpenClaw 安装与管理指南.md
Normal file
368
AI/openclaw/Ubuntu 下 OpenClaw 安装与管理指南.md
Normal file
@@ -0,0 +1,368 @@
|
||||
|
||||
#ubuntu #openclaw #install #uninstall
|
||||
|
||||
```table-of-contents
|
||||
```
|
||||
|
||||
## 环境概述
|
||||
|
||||
- 系统：Ubuntu 20.04 / 22.04
|
||||
- OpenClaw 安装方式：npm 用户本地全局安装openclaw & clawhub (注意不要用root user安装)
|
||||
```
|
||||
npm install -g openclaw clawhub
|
||||
```
|
||||
- 用户路径示例：
|
||||
|
||||
```bash
|
||||
/home/shenwei/.npm-global/bin/openclaw
|
||||
```
|
||||
|
||||
- 默认配置目录：
|
||||
```bash
|
||||
/home/shenwei/.openclaw
|
||||
```
|
||||
|
||||
- 用户级 systemd 服务目录：
|
||||
```bash
|
||||
/home/shenwei/.config/systemd/user/openclaw-gateway.service
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 卸载旧版本 OpenClaw
|
||||
|
||||
1. **停止正在运行的进程 / 服务**
|
||||
```bash
|
||||
# 查找进程
|
||||
ps aux | grep openclaw
|
||||
|
||||
# 如果有 systemd 用户服务
|
||||
systemctl --user stop openclaw
|
||||
systemctl --user disable openclaw
|
||||
```
|
||||
|
||||
2. **卸载 npm 安装的 OpenClaw**
|
||||
```bash
|
||||
# 全局卸载
|
||||
sudo npm uninstall -g openclaw clawhub
|
||||
|
||||
# 或者局部卸载
|
||||
npm uninstall openclaw clawhub
|
||||
```
|
||||
|
||||
3. **删除用户配置目录**
|
||||
```bash
|
||||
rm -rf /home/shenwei/.openclaw # 普通用户
|
||||
sudo rm -rf /root/.openclaw # root 用户（如果曾用 sudo 运行）
|
||||
sudo rm -rf /opt/openclaw # 如果之前手动统一过目录
|
||||
```
|
||||
|
||||
4. **清理残留 npm 包**
|
||||
```bash
|
||||
npm list -g --depth=0 | grep openclaw
|
||||
npm list -g --depth=0 | grep clawhub
|
||||
```
|
||||
如有残留再执行 `npm uninstall -g <package>`。
|
||||
|
||||
---
|
||||
|
||||
## 安装 OpenClaw
|
||||
|
||||
### 方法 A：通过 npm 安装（推荐）
|
||||
|
||||
```bash
|
||||
# 确保 npm 更新
|
||||
npm install -g npm
|
||||
|
||||
# 全局安装 OpenClaw
|
||||
npm install -g openclaw clawhub
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 配置 PATH，让 OpenClaw 在任意位置可执行
|
||||
|
||||
1. 临时生效（仅当前终端）：
|
||||
```bash
|
||||
export PATH=$HOME/.npm-global/bin:$PATH
|
||||
```
|
||||
|
||||
2. 永久生效（推荐）：
|
||||
- 编辑 shell 配置文件 `~/.bashrc` 或 `~/.zshrc`，添加：
|
||||
```bash
|
||||
export PATH=$HOME/.npm-global/bin:$PATH
|
||||
```
|
||||
|
||||
- 刷新配置：
|
||||
```bash
|
||||
source ~/.bashrc # bash
|
||||
source ~/.zshrc # zsh
|
||||
```
|
||||
|
||||
- 验证：
|
||||
```bash
|
||||
which openclaw
|
||||
openclaw --version
|
||||
```
|
||||
---
|
||||
|
||||
## 用户级 systemd 服务管理（OpenClaw Gateway）
|
||||
|
||||
安装 Gateway 后会生成服务文件：
|
||||
```bash
|
||||
/home/shenwei/.config/systemd/user/openclaw-gateway.service
|
||||
```
|
||||
|
||||
``` bash
|
||||
[Unit]
|
||||
Description=OpenClaw Gateway (v2026.3.13)
|
||||
After=network-online.target
|
||||
Wants=network-online.target
|
||||
|
||||
[Service]
|
||||
ExecStart=/usr/bin/node /home/shenwei/.npm-global/lib/node_modules/openclaw/dist/index.js gateway --port 18789
|
||||
Restart=always
|
||||
RestartSec=5
|
||||
TimeoutStopSec=30
|
||||
TimeoutStartSec=30
|
||||
SuccessExitStatus=0 143
|
||||
KillMode=control-group
|
||||
Environment=HOME=/home/shenwei
|
||||
Environment=TMPDIR=/tmp
|
||||
Environment=HTTP_PROXY=http://127.0.0.1:10808
|
||||
Environment=HTTPS_PROXY=http://127.0.0.1:10808
|
||||
Environment=PATH=/home/shenwei/.local/bin:/home/shenwei/.npm-global/bin:/home/shenwei/bin:/home/shenwei/.volta/bin:/home/shenwei/.asdf/shims:/home/shenwei/.bun/bin:/>
|
||||
Environment=OPENCLAW_GATEWAY_PORT=18789
|
||||
Environment=OPENCLAW_SYSTEMD_UNIT=openclaw-gateway.service
|
||||
Environment="OPENCLAW_WINDOWS_TASK_NAME=OpenClaw Gateway"
|
||||
Environment=OPENCLAW_SERVICE_MARKER=openclaw
|
||||
Environment=OPENCLAW_SERVICE_KIND=gateway
|
||||
Environment=OPENCLAW_SERVICE_VERSION=2026.3.13
|
||||
|
||||
[Install]
|
||||
WantedBy=default.target
|
||||
|
||||
```
|
||||
|
||||
查看日志
|
||||
```
|
||||
|
||||
# 查看最近 50 行
|
||||
journalctl --user -u openclaw-gateway -n 50 --no-pager
|
||||
|
||||
# 实时跟踪日志
|
||||
journalctl --user -u openclaw-gateway -f
|
||||
|
||||
# 查看今天的所有日志
|
||||
journalctl --user -u openclaw-gateway --since today
|
||||
```
|
||||
|
||||
|
||||
### 常用管理命令
|
||||
|
||||
| 操作 | 命令 |
|
||||
| --------------------------- | ------------------------------------------- |
|
||||
| 启动 Gateway | `systemctl --user start openclaw-gateway` |
|
||||
| 停止 Gateway | `systemctl --user stop openclaw-gateway` |
|
||||
| 重启 Gateway | `systemctl --user restart openclaw-gateway` |
|
||||
| 查看状态 | `systemctl --user status openclaw-gateway` |
|
||||
| 开机自启 | `systemctl --user enable openclaw-gateway` |
|
||||
| 取消开机自启 | `systemctl --user disable openclaw-gateway` |
|
||||
| 刷新 systemd 配置（修改 service 后） | `systemctl --user daemon-reload` |
|
||||
|
||||
|
||||
> ⚠️ 用户级服务不需要 sudo，安全且方便。
|
||||
|
||||
---
|
||||
|
||||
## 多用户环境与避免重复环境
|
||||
|
||||
- OpenClaw 配置目录默认跟随 `$HOME`：
|
||||
|
||||
|用户|配置目录|
|
||||
|---|---|
|
||||
|shenwei|`/home/shenwei/.openclaw`|
|
||||
|root|`/root/.openclaw`|
|
||||
|
||||
- **原因**：Linux 用户隔离机制，不同用户运行 OpenClaw 会生成独立目录。
|
||||
- **注意**：
|
||||
|
||||
- 不要用 root 启动 OpenClaw，避免权限混乱
|
||||
- 统一使用普通用户安装和运行
|
||||
- 可通过 `--workdir /opt/openclaw` 指定统一目录
|
||||
|
||||
|
||||
---
|
||||
|
||||
## 常用命令总结
|
||||
|
||||
| 命令 | 功能 |
|
||||
| --------------------------------------------------------------------- | ------------- |
|
||||
| `openclaw onboard` | 初始化新环境、设置工作目录 |
|
||||
| `openclaw --version` | 查看版本 |
|
||||
| `openclaw agent list` | 列出所有 agent |
|
||||
| `openclaw agent create --name <agent_name> --message "<description>"` | 创建新的 agent |
|
||||
| `openclaw agent delete <agent_name>` | 删除 agent |
|
||||
| `openclaw skill install <skill_name>` | 安装技能 |
|
||||
| `openclaw skill update <skill_name>` | 更新技能 |
|
||||
| `openclaw skill list` | 查看已安装技能 |
|
||||
| `openclaw memory list` | 查看记忆数据 |
|
||||
| `openclaw workspace list` | 查看工作空间 |
|
||||
| | |
|
||||
| | |
|
||||
|
||||
---
|
||||
|
||||
## 创建 Agent 与绑定 Telegram Bot
|
||||
|
||||
1. **创建 agent**
|
||||
```
|
||||
openclaw agents add <agentname> --non-interactive --workspace /home/shenwei/.openclaw/workspace-agent-<agentname> --model MiniMax-M2.5
|
||||
|
||||
```
|
||||
举例：
|
||||
```bash
|
||||
openclaw agents add yunce --non-interactive --workspace /home/shenwei/.openclaw/workspace-agent-yunce --model MiniMax-M2.5
|
||||
```
|
||||
|
||||
2. **添加Telegram 账号**
|
||||
```
|
||||
# 添加 Telegram 账号
|
||||
openclaw channels add --channel telegram --account <账号名> --token <BotToken>
|
||||
```
|
||||
举例
|
||||
```
|
||||
openclaw channels add --channel telegram --account yunhan --token 8588117769:AAFxswhHgCdBor2EOa-2oChDpI-DADRt0tQ
|
||||
```
|
||||
3. **查看 agent 列表**
|
||||
```bash
|
||||
openclaw agents list
|
||||
```
|
||||
|
||||
3. **绑定 Bot**
|
||||
```
|
||||
# 绑定 agent 到 Telegram 账号
|
||||
openclaw agents bind --agent <agent_id> --bind telegram:<account_name>
|
||||
|
||||
```
|
||||
举例
|
||||
```
|
||||
openclaw agents bind --agent yunhan --bind telegram:yunhan
|
||||
```
|
||||
|
||||
- 配置完成后重启 Gateway：
|
||||
```bash
|
||||
systemctl --user restart openclaw-gateway
|
||||
```
|
||||
|
||||
- Telegram 多 Agent 建议：
|
||||
|
||||
- 一个 bot → n8n 路由 → 多 agent
|
||||
- 避免每个 agent 都创建独立 bot（每个账号最多 20 个 bot）
|
||||
- 用命令或路径路由区分不同 agent 功能
|
||||
|
||||
---
|
||||
|
||||
## 删除Agent
|
||||
1. **删除 agent**
|
||||
```
|
||||
openclaw agents delete <agent_name> --force
|
||||
```
|
||||
2. **删除bot**
|
||||
```
|
||||
# 删除 Telegram 账号
|
||||
openclaw channels remove --channel telegram --account <account_name> --delete
|
||||
```
|
||||
|
||||
## 注意事项与避免的坑
|
||||
|
||||
1. **避免使用 root 运行**
|
||||
- root 会生成 `/root/.openclaw`，和普通用户环境冲突
|
||||
- 权限问题可能导致 agent 无法访问工作空间
|
||||
2. **避免重复 PATH 或多版本冲突**
|
||||
- 如果 npm 本地 bin 不在 PATH，会导致命令找不到
|
||||
- 如果 PATH 里还有旧版本系统全局安装路径，可能会调用错误版本
|
||||
3. **用户级 systemd 服务管理**
|
||||
- 修改 service 后必须执行 `systemctl --user daemon-reload`
|
||||
- 避免 sudo 启动服务，保证文件权限正确
|
||||
4. **Telegram Bot 限制**
|
||||
- 每个账号最多创建 20 个 bot（Premium 账号可能 40 个）
|
||||
- 多 agent 架构建议一个 bot → n8n → 多 agent 路由
|
||||
5. **统一工作目录**
|
||||
- 推荐 `/home/shenwei/.openclaw` 或 `/opt/openclaw`
|
||||
- 方便多服务器或多 agent 管理
|
||||
6. **升级和维护**
|
||||
- 升级前先备份 `.openclaw` 下的 workspace、skills、memory
|
||||
- 使用 npm 全局安装可直接 `npm install -g openclaw@latest`
|
||||
|
||||
---
|
||||
|
||||
### 参考架构示意
|
||||
|
||||
```
|
||||
Telegram Bot
|
||||
↓
|
||||
n8n Router
|
||||
↓
|
||||
OpenClaw Agents
|
||||
├─ 星枢（调度）
|
||||
├─ 星曜（IT管家）
|
||||
├─ 星辉（个人助理）
|
||||
└─ 云瀚（监控）
|
||||
```
|
||||
|
||||
|
||||
### Bot & Agent 命名
|
||||
|
||||
#### 星
|
||||
|
||||
```
|
||||
openclaw channels add --channel telegram --account xingshu --token 8787024183:AAG1M5tfSHj6Z0gMv3vvCZel2FOIX-0x8ZI
|
||||
|
||||
openclaw channels add --channel telegram --account xingyao --token 8414432613:AAG9hvKfILGSsbc1EMEZW1QVym9Quc5aHWk
|
||||
|
||||
openclaw channels add --channel telegram --account xinghui --token 8709222939:AAEfvZrvvU5vZFsmacsR5nmpkJ2Jb5JgfRg
|
||||
|
||||
```
|
||||
|
||||
| 服务器 | 角色 | Bot Name | Bot Key | Agent Id | Telegram User ID |
|
||||
| ------- | --- | ---------------------------- | ---------------------------------------------- | -------- | ---------------- |
|
||||
| macmini | 星枢 | @shenwei_macmini_xingshu_bot | 8787024183:AAG1M5tfSHj6Z0gMv3vvCZel2FOIX-0x8ZI | main | 5038825565 |
|
||||
| macmini | 星曜 | @shenwei_macmini_xingyao_bot | 8414432613:AAG9hvKfILGSsbc1EMEZW1QVym9Quc5aHWk | xingyao | 5038825565 |
|
||||
| macmini | 星辉 | @shenwei_macmini_xinghui_bot | 8709222939:AAEfvZrvvU5vZFsmacsR5nmpkJ2Jb5JgfRg | xinghui | 5038825565 |
|
||||
| | | | | | |
|
||||
#### 云
|
||||
|
||||
```
|
||||
openclaw channels add --channel telegram --account yunhan --token 8588117769:AAFxswhHgCdBor2EOa-2oChDpI-DADRt0tQ
|
||||
|
||||
openclaw channels add --channel telegram --account yunce --token 8791231082:AAFKPfTPy3LshybWUJ0joBkz3Th3mwYQOnc
|
||||
|
||||
openclaw channels add --channel telegram --account yunjiang --token 8727937702:AAGw3WGPI1j5rSD97wap6h9EGqVpDEMdjLU
|
||||
|
||||
openclaw channels add --channel telegram --account yunzhi --token 8639619464:AAEI35Dnt-9PQ8y4Du_ToxVhwUBUa5kpLjU
|
||||
```
|
||||
|
||||
| 服务器 | 角色 | Bot Name | Bot Key | Agent 名称 | Telegram User ID |
|
||||
| ------- | --- | ----------------------------- | ---------------------------------------------- | -------- | ---------------- |
|
||||
| ubuntu2 | 云瀚 | @shenwei_ubuntu2_yunhan_bot | 8588117769:AAFxswhHgCdBor2EOa-2oChDpI-DADRt0tQ | yunhan | 5038825565 |
|
||||
| ubuntu2 | 云策 | @shenwei_ubuntu2_yunce_bot | 8791231082:AAFKPfTPy3LshybWUJ0joBkz3Th3mwYQOnc | yunce | 5038825565 |
|
||||
| ubuntu2 | 云匠 | @shenwei_ubuntu2_yunjiang_bot | 8727937702:AAGw3WGPI1j5rSD97wap6h9EGqVpDEMdjLU | yunjiang | 5038825565 |
|
||||
| ubuntu2 | 云织 | @shenwei_ubuntu2_yunzhi_bot | 8639619464:AAEI35Dnt-9PQ8y4Du_ToxVhwUBUa5kpLjU | yunzhi | 5038825565 |
|
||||
|
||||
|
||||
#### 风
|
||||
|
||||
```
|
||||
openclaw channels add --channel telegram --account fengheng --token
|
||||
|
||||
openclaw channels add --channel telegram --account fengchi --token
|
||||
|
||||
openclaw channels add --channel telegram --account fengji --token
|
||||
```
|
||||
|
||||
| 服务器 | 角色 | Bot Name | Bot Key | Agent 名称 | Telegram User ID |
|
||||
| ------- | --- | ----------------------------- | ------- | -------- | ---------------- |
|
||||
| ubuntu1 | 风衡 | @shenwei_ubuntu1_fengheng_bot | | fengheng | 5038825565 |
|
||||
| ubuntu1 | 风驰 | @shenwei_ubuntu1_fengchi_bot | | fengchi | 5038825565 |
|
||||
| ubuntu1 | 风纪 | @shenwei_ubuntu1_fengji_bot | | fengji | 5038825565 |
|
||||
121
AI/固定镜头短视频制作的AI全流程解析.md
Normal file
121
AI/固定镜头短视频制作的AI全流程解析.md
Normal file
@@ -0,0 +1,121 @@
|
||||
|
||||
```table-of-contents
|
||||
```
|
||||
|
||||
## 固定镜头短视频制作的AI全流程解析
|
||||
|
||||
### 概述🛠️
|
||||
本视频围绕如何利用AI技术快速、高效地制作高播放量的家装类短视频展开介绍。讲解了从文案到分镜拆解、图片生成、一致性控制、动态图像处理和剪辑音效配合的全套流程，重点在于利用固定机位、内容连续变化和时间压缩三个核心原理，实现短时间内从毛坯房到精装修的视觉呈现。内容深入浅出，实例丰富，适合想掌握AI短视频制作方法的创作者学习和复制。
|
||||
https://youtu.be/ES6BcIIiB5g
|
||||
|
||||
|
||||
### 核心知识点总结⏰
|
||||
|
||||
- **00:00-00:31 制作需求与时间认知重塑**
|
||||
- 常见家装短视频播放量巨大，制作时间却被误解为长。实际用AI不到10分钟即可完成成片，核心在于拆解文案和分镜，逐步生成内容。
|
||||
|
||||
- **00:31-01:18 家装视频三大关键词**
|
||||
- 固定机位：摄像机位置固定，不移动镜头。
|
||||
- 内容连续变化：画面主要信息是施工进度变化。
|
||||
- 时间压缩：将长时间装修过程浓缩呈现。
|
||||
- 这三个特点使视频非常适合用AI技术生成。
|
||||
|
||||
- **01:18-01:52 AI工具分类及功能**
|
||||
- 大脑类：负责把视频逻辑转化成AI能识别的分镜语言，如XAR GPT、GEMALA。
|
||||
- 设计师类：将分镜转换为一致的图像，如Midjourney、Nano Banana。
|
||||
- 动效类：让画面产生连贯动画效果，如海螺AI、多么AI、KAI，需支持“首尾针”动画。
|
||||
![[IMG-20260315173031668.png]]
|
||||
![[IMG-20260315173031695.png]]
|
||||
![[IMG-20260315173031715.png]]
|
||||
- **01:52-02:22 原视频观察与核心关键词“时间流逝”**
|
||||
- 视频内容简洁，只有一个机位，画面随施工进展从毛坯到成品平稳变化，AI对此类时间推移处理表现优异。
|
||||
|
||||
- **02:22-02:53 AI拆分镜流程**
|
||||
- 通过Google AI Studio，输入装修视频链接并让模型分析，自动生成九个分镜描述，确保摄像机机位固定、场景顺序清晰和阶段明确。
|
||||
|
||||
- **02:53-03:55 保证画面一致性的九宫格法**
|
||||
- 一次性用三乘三九宫格图生成九个分镜画面，机位和角度不变，细节只表现施工进度的变化，增强画面空间和光影的连贯性。
|
||||
|
||||
- **03:55-05:29 九宫格图片的切割成单张过程**
|
||||
- 利用Google AI Studio工具，自动检测并将三乘三大图裁为九张竖屏图（9:16比例），为后续动画制作做好准备。
|
||||
|
||||
- **05:29-06:16 动态动画生成核心“首尾针”逻辑**
|
||||
- 逐个上传九张图片配对制作动画，利用“首针图”和“尾针图”补齐两个阶段之间的变化，达成画面平滑过渡。
|
||||
|
||||
- **06:16-07:35 具体动画生成及合成方法**
|
||||
- 以KAI工具为例，通过AI Video API依次生成阶段视频片段，核心是让画面变化自然而非镜头移动，完成所有片段后，导入剪映合成。
|
||||
|
||||
- **07:35-08:22 短视频快速剪辑三要点**
|
||||
- 统一加速，建议2-4倍速（示例用3倍）加快进度感。
|
||||
- 无需复杂转场，采用首尾针动画的硬切效果更干净。
|
||||
- 画面轻微裁边，如有黑边可稍微放大处理。
|
||||
|
||||
- **08:22-09:05 声音设计提升视频品质**
|
||||
- 添加适量施工音效（如敲击、电钻、切割），即使不完整也能增强真实感。
|
||||
- 选择节奏感强且节奏干净的背景音乐，决定观众观看体验。
|
||||
- 画面变化处精准卡点，满足视觉与节奏同步，提升整体观感。
|
||||
|
||||
- **09:05-09:48 五步复用AI短视频公式总结**
|
||||
- 拆分镜头 → 一致性图像生成 → 首尾针动画制作 → 快速剪辑 → 声音设计。
|
||||
- 该流程可应用于所有固定机位且状态变化明显的短视频类型，关键在于对节奏和细节的把握。
|
||||
|
||||
### 关键术语与定义📚
|
||||
|
||||
- **固定机位**：摄像机位置固定不变，是视频画面统一和连贯的基础。
|
||||
- **内容连续变化**：视频主体信息随时间持续发生明确阶段性变化。
|
||||
- **时间压缩**：将长时间拍摄过程在视频中浓缩表现的手法。
|
||||
- **分镜拆解**：将视频内容拆分成多个画面阶段描述。
|
||||
- **九宫格法**：同时生成3x3共九个画面，保证机位与角度不变，画面一致性强。
|
||||
- **首尾针动画**：通过上传两个关键帧（首针和尾针），AI自动补齐中间动作，产生连贯动画的技术。
|
||||
- **快节奏剪辑**：视频使用加速播放和硬切换手法，强化节奏感与流畅度。
|
||||
- **卡点**：画面变化与音乐节奏巧妙同步，提高观看体验。
|
||||
|
||||
### 推理结构🔍
|
||||
|
||||
1. **前提**：家装类短视频需表现装修变化且画面需保持一致性。
|
||||
2. **分析**：固定机位、内容阶段变化、时间压缩是视频成功关键。
|
||||
3. **推理**：利用AI分镜拆解+图像设计+动画生成技术，可快速高质量复刻此类内容。
|
||||
4. **结论**：通过九宫格一致性图片和首尾针动画，加速剪辑及音效设计，实现高播放量视频制作。
|
||||
|
||||
### 典型示例🎯
|
||||
|
||||
- **视频“从毛坯到精装”实拍片段**：
|
||||
用摄像机固定视角从空房间到悬挂床的安装，整个过程仅通过画面中施工进度的持续推进展现房屋翻新，突出时间流逝主题，示范AI在时间压缩及动态生成中的优势。
|
||||
|
||||
- **九宫格单图批量生成**：
|
||||
利用三乘三布局，将整个施工进度分解为九幅连贯画面，确保机位和景深一致，典型示范了画面一致性处理的技术手法。
|
||||
|
||||
### 易错点总结⚠️
|
||||
|
||||
- **误区：误以为短视频制作需要复杂移动镜头。**
|
||||
- 纠正：固定机位，内容变化即可，减少复杂摄像设备需求。
|
||||
- **误区：逐帧独立生成图片导致光影空间关系错乱。**
|
||||
- 纠正：采用九宫格一次性生成保证画面连贯。
|
||||
- **误区：转场效果加入过多导致视频冗杂。**
|
||||
- 纠正：利用首尾针动画自带的平滑衔接，硬切反而更简洁。
|
||||
- **误区：忽视声音设计，视频体验感降低。**
|
||||
- 纠正：施工音效和节奏感强的BGM不可缺，精准卡点尤为重要。
|
||||
|
||||
### 快速复习提示与自测题💡
|
||||
|
||||
- **复习提示（不含答案）**
|
||||
1. 家装短视频成功的三大关键词是什么？
|
||||
2. “九宫格法”为何能保证图像一致性？
|
||||
3. 首尾针动画的基本原理是什么？
|
||||
4. 快节奏剪辑应注意哪些要点？
|
||||
5. 如何通过声音设计提升视频观感？
|
||||
|
||||
- **自测练习（含答案）**
|
||||
1. 为什么固定机位对视频制作如此重要？
|
||||
**答**：固定机位保证画面空间和光影一致，增强连贯感，方便AI补齐动画。
|
||||
2. “首尾针”动画技术如何实现动态过渡？
|
||||
**答**：上传两个关键帧图片作为“首针”和“尾针”，AI自动补充中间变化，实现自然动画效果。
|
||||
3. 进行九宫格裁图时，如何保证图片比例正确？
|
||||
**答**：将图片宽高各等分成三份，裁切成9张9比16的竖屏图，保持画面比例一致。
|
||||
4. AI拆分镜的工具和流程包括哪些步骤？
|
||||
**答**：输入视频链接至Google AI Studio，利用模型分析视频逻辑，生成九个阶段分镜描述。
|
||||
5. 制作快节奏剪辑时，为什么避免复杂转场？
|
||||
**答**：首尾针动画本身提供平滑过渡，硬切清晰干净，避免视觉干扰。
|
||||
|
||||
### 总结回顾🔄
|
||||
本视频系统讲解了基于AI技术制作高效家装短视频的完整流程，以固定机位拍摄、分镜拆解、九宫格一致性生成、首尾针动画和快节奏剪辑为核心技术点，配合合理的声音设计，解决了以往工地实拍周期长、制作复杂的难题。整套方法不仅成片快且易于复制，适用于多类固定机位状态变化视频的制作，体现了AI工具在视频内容创作中的巨大潜力与应用价值。
|
||||
50
AI/如何利用Sora接口实现视频自动化生成工作流.md
Normal file
50
AI/如何利用Sora接口实现视频自动化生成工作流.md
Normal file
@@ -0,0 +1,50 @@
|
||||
#n8n #workflow #sora
|
||||
|
||||
https://youtu.be/f0fP9wQHBcY?si=zAI-YHBReu_vIUXB
|
||||
|
||||
# 摘要
|
||||
本期视频由欧阳主讲，围绕如何使用“Sora”进行视频生成的全自动化工作流进行详细讲解。视频介绍了成本效益极高的“Sora”接口，以及如何通过该接口批量生成SR（声视频）内容，提升自媒体创作的效率和质量。本教程适合对视频生成感兴趣的个人及中小型企业，帮助观众以低成本的方式启动自媒体副业，并在市场中脱颖而出。
|
||||
|
||||
# 时间线摘要
|
||||
- **00:00 - 02:45**: 视频引入内容，介绍全自动化工作流及其优势，特别强调“Sora”接口的低成本和高效性。
|
||||
- **02:46 - 05:00**: 讲解亚马逊账户注册及免费模型调用，强调新用户的优惠和如何成功注册账户。
|
||||
- **05:01 - 08:00**: 细述如何创建用户权限及API密钥，为“Sora”流的后续操作做准备。
|
||||
- **08:01 - 11:30**: 演示如何调用API并测试连接，介绍基本的AI生成设置。
|
||||
- **11:31 - 14:00**: 深入探讨不同模型的生成能力，包括无水印视频生成及相应的费用说明。
|
||||
- **14:01 - 17:30**: 讨论“Sora”生成的UGC（用户生成内容）视频，通过示例展示如何进行有效创作。
|
||||
- **17:31 - 20:00**: 演示如何利用肖像权生成内容，强调遵循法律规范的重要性。
|
||||
- **20:01 - 24:00**: 介绍如何使用故事板功能，创建分镜脚本并表现不同场景效果。
|
||||
- **24:01 - 29:00**: 总结视频生成流程，分享提示词优化技巧及字符串替换技术，强调自动化工具的重要性。
|
||||
|
||||
# 关键点
|
||||
- **🤖 全自动化工作流**: 通过“Sora”接口实现视频生成的经济实惠方案。
|
||||
- **💰 注册优惠**: 新用户注册亚马逊账户可享受200美元抵扣金等福利。
|
||||
- **📈 UGC 创作**: 用户可轻松生成UGC视频，提高市场推广能力。
|
||||
- **📜 合法使用肖像权**: 确保在生成内容时遵循肖像权法，避免法律风险。
|
||||
- **🧩 提示词优化**: 提升生成内容质量的关键在于优化提示词的撰写。
|
||||
|
||||
# 关键见解
|
||||
- **🌟 经济实惠**: 使用“Sora”能显著降低视频生成成本，相较于OpenAI便宜六倍以上。
|
||||
- **🌍 新用户福利**: 注册新账户的用户可以获得六个月的免费试用权，显著降低启动成本。
|
||||
- **📝 提示词的艺术**: 提高生成内容质量的关键在于精细化的提示词设计，影响最终结果。
|
||||
- **📊 多功能应用**: “Sora”不仅支持文本转视频，还可以生成图像类内容，扩展用户的创作边界。
|
||||
- **🔑 安全调用API**: 详细介绍了如何安全有效地调用API，确保视频生成过程中的信息安全。
|
||||
|
||||
# 常见问题 (FAQs)
|
||||
1. **问：如何快速注册亚马逊账户以使用模型？**
|
||||
- 答：访问注册页面，填写个人信息并绑定支持国际支付的信用卡，确保卡片是实名信息。
|
||||
|
||||
2. **问：如何生成无水印视频？**
|
||||
- 答：在生成请求中选择相应参数，确保移除水印设置为“TRUE”。
|
||||
|
||||
3. **问：生成视频的费用大约是多少？**
|
||||
- 答：使用“Sora”生成一般视频的费用仅需两三元人民币，远低于市场水平。
|
||||
|
||||
4. **问：是否可以使用他人的肖像权生成内容？**
|
||||
- 答：可以，但必须获得对方的同意，并确保生成的内容不违反相关法律法规。
|
||||
|
||||
5. **问：提示词优化对生成质量的影响有多大？**
|
||||
- 答：精细化的提示词设计能够显著提升生成视频的质量，增强内容的吸引力。
|
||||
|
||||
# 结论
|
||||
本期视频全面讲述了如何利用“Sora”接口实现视频生成的全自动化工作流，提供了实用的内容创作指南和技术技巧。观众可以通过学习本教程掌握低成本生成内容的能力，并在自媒体领域取得更高的竞争优势。建议大家积极实践所学内容，并根据提示词优化技巧不断提升生成效果。未来，继续探索AI技术的应用，为创作带来更多可能性。
|
||||
57
AI/文字生成视频网站推荐.md
Normal file
57
AI/文字生成视频网站推荐.md
Normal file
@@ -0,0 +1,57 @@
|
||||
根据搜索结果，以下是几款性价比较高且支持文字生成视频的AI工具推荐，结合功能、价格及用户评价进行综合评估：
|
||||
|
||||
---
|
||||
|
||||
### **1. 万彩AI**
|
||||
- **特点**：
|
||||
- **免费使用**：提供免费账号注册，支持文字直接生成短视频，无使用次数限制。
|
||||
- **操作便捷**：输入文字后，可自动匹配配音、视频模板及转场效果，支持数字人形象生成（上传照片或选择预设角色）。
|
||||
- **模板丰富**：包含100+文案模板和视频风格（如商务、教育、国风等），适合多种场景需求。
|
||||
- **适用人群**：新手小白、自媒体创作者、企业营销人员。
|
||||
- **推荐理由**：完全免费且功能全面，适合预算有限的用户快速生成高质量视频。
|
||||
|
||||
---
|
||||
|
||||
### **2. 百度AI开放平台（AI成片）**
|
||||
- **特点**：
|
||||
- **免费体验套餐**：注册后可领取免费套餐，支持图文转视频、自动配音、字幕添加及数字人功能。
|
||||
- **智能化解析**：基于百度多模态技术，智能匹配素材并生成逻辑清晰的视频内容。
|
||||
- **个性化调整**：支持视频尺寸、音色、时长等参数自定义。
|
||||
- **适用场景**：企业宣传、知识科普、新闻短视频等。
|
||||
- **推荐理由**：大厂技术背书，免费套餐适合短期需求，长期使用需根据具体功能付费（价格未公开）。
|
||||
|
||||
---
|
||||
|
||||
### **3. Zeemo（蓝色脉动公司）**
|
||||
- **特点**：
|
||||
- **精准字幕生成**：支持95种语言转录，准确率达98%，适合全球化内容创作者。
|
||||
- **收费模式**：年费分三档（$79/119/239），按视频时长和清晰度分级。
|
||||
- **优势**：多语言支持及高精度字幕生成，适合需要专业级字幕优化的用户。
|
||||
- **适用场景**：海外短视频平台（如TikTok、YouTube）的内容制作。
|
||||
|
||||
---
|
||||
|
||||
### **4. Vizard（蓝色脉动公司）**
|
||||
- **特点**：
|
||||
- **自动剪辑亮点**：从长视频中智能提取高光片段，生成10-30秒短视频。
|
||||
- **免费版限制**：每月60分钟上传时长，适合轻度用户。
|
||||
- **企业版费用**：年费约2610美元（72000分钟上传时长）。
|
||||
- **推荐理由**：适合需要批量处理长视频的用户，免费版可满足基础需求。
|
||||
|
||||
---
|
||||
|
||||
### **5. 快影（腾讯系工具）**
|
||||
- **特点**：
|
||||
- **模板化剪辑**：提供特效和模板库，适合快速制作短视频。
|
||||
- **免费使用**：基础功能免费，但高级特效需付费。
|
||||
- **优势**：操作简单，适合对剪辑要求不高的用户。
|
||||
|
||||
---
|
||||
|
||||
### **总结推荐**
|
||||
- **最实惠选择**：**万彩AI**（完全免费且功能全面）。
|
||||
- **技术型用户**：百度AI开放平台（免费套餐+多模态技术）。
|
||||
- **多语言需求**：Zeemo（高精度字幕+多语言支持）。
|
||||
- **长视频处理**：Vizard（免费版基础功能）。
|
||||
|
||||
建议优先试用免费工具（如万彩AI或百度AI），再根据实际需求选择付费服务。更多细节可参考各平台官网或体验套餐。
|
||||
525
AI/详细！离线部署大模型：ollama+deepseek+open-webui安装使用方法及常见问题解决 1.md
Normal file
525
AI/详细！离线部署大模型：ollama+deepseek+open-webui安装使用方法及常见问题解决 1.md
Normal file
@@ -0,0 +1,525 @@
|
||||
---
|
||||
title: "详细！离线部署大模型：ollama+deepseek+open-webui安装使用方法及常见问题解决"
|
||||
source: "https://mp.weixin.qq.com/s/1cbpf9IlLgg9NApk5322GA"
|
||||
author:
|
||||
- "[[任侠001]]"
|
||||
published:
|
||||
created: 2025-03-14
|
||||
description:
|
||||
tags:
|
||||
- "clippings"
|
||||
---
|
||||
|
||||
ollama 是一个开源的本地大语言模型运行框架，它提供了非常简单便捷的使用形式，让用户可以十分方便的在本地机器上部署和运行大型语言模型，从而实现免费离线的方式使用 LLM 能力，并确保私有数据的隐私和安全性。
|
||||
|
||||
## 1 ollama 安装
|
||||
|
||||
ollama 支持多种操作系统，包括 macOS、Windows、Linux 以及通过 Docker 容器运行。其安装、使用及模型下载非常简单，可以简单概括为以下几步：
|
||||
|
||||
- • 下载 ollama 安装程序并安装。
|
||||
- • 启动 ollama，执行命令下载和运行模型。如：`ollama run deepseek-r1:1.5b`
|
||||
- • 以命令行交互、API 调用、第三方应用接入等形式使用其服务。
|
||||
|
||||
### 1.1 硬件要求
|
||||
|
||||
ollama 本身对硬件要求并不高，主要取决于运行模型的要求。基本建议：
|
||||
|
||||
> 你应该至少有 4 GB 的 RAM 来运行 1.5B 模型，至少有 8 GB 的 RAM 来运行 7B 模型，16 GB 的 RAM 来运行 13B 模型，以及 32 GB 的 RAM 来运行 33B 模型。
|
||||
|
||||
假若需要本地私有化部署具有实用性的模型，应至少有独立显卡并有 4G 以上显存。纯 CPU 模式虽然也可以运行，但生成速度很慢，仅适用于本地开发调试体验一下。
|
||||
|
||||
本人实测在`Mac Studio 2023 版(Apple M2 Max 芯片：12核、32G内存、30核显、1TB SSD)`上，运行 `deepseek:1.5b` 模型响应非常快，可以较为流畅的运行 `deepseek-r1:32b` 及以下的模型。
|
||||
|
||||
**DeepSeek-r1 相关版本及大小参考：**
|
||||
|
||||
| 参数版本 | 模型大小 | 建议CPU | 建议内存 | 建议显存 | 特点 |
|
||||
| ---------------- | ----- | ----- | ---- | ------ | --------------------- |
|
||||
| deepseek-r1:1.5b | 1.1GB | 4核 | 4~8G | 4GB | 轻量级，速度快、普通文本处理 |
|
||||
| deepseek-r1:7b | 4.7G | 8核 | 16G | 14GB | 性能较好，硬件要求适中 |
|
||||
| deepseek-r1:8b | 4.9GB | 8核 | 16G | 14GB | 略强于 7b，精度更高 |
|
||||
| deepseek-r1:14b | 9GB | 12核 | 32G | 26GB | 高性能，擅长复杂任务，如数学推理、代码生成 |
|
||||
| deepseek-r1:32b | 20GB | 16核 | 64G | 48GB | 专业级，适合高精度任务 |
|
||||
| deepseek-r1:70b | 43GB | 32核 | 128G | 140GB | 顶级模型，适合大规模计算和高复杂度任务 |
|
||||
| deepseek-r1:671b | 404GB | 64核 | 512G | 1342GB | 超大规模，性能卓越，推理速度快 |
|
||||
|
||||
### 1.2 Windows \\ macOS \\ Linux 下安装 ollama
|
||||
|
||||
Windows 和 macOS 用户可访问如下地址下载安装文件并安装：
|
||||
|
||||
- • 国内中文站下载：http://ollama.org.cn/download/
|
||||
- • 官方下载：https://ollama.com/download/
|
||||
- • github release 下载：https://github.com/ollama/ollama/releases/
|
||||
|
||||
Linux 用户可以执行如下命令一键安装：
|
||||
|
||||
```
|
||||
curl -fsSL https://ollama.com/install.sh | bash
|
||||
```
|
||||
|
||||
安装完成后，可以通过执行 `ollama --version` 命令查看 ollama 版本信息，以验证是否安装成功。
|
||||
|
||||
**ollama 离线安装：**
|
||||
|
||||
Windows 和 macOS 下直接复制安装文件到本地本进行安装即可。
|
||||
|
||||
Linux 下的离线安装主要步骤参考如下：
|
||||
|
||||
```
|
||||
mkdir -p /home/ollama
|
||||
cd /home/ollama
|
||||
|
||||
# 查看服务器 CPU 信息获取其架构，如：x86_64
|
||||
lscpu
|
||||
|
||||
# 访问如下地址，下载对应架构的 ollama 安装包
|
||||
# https://github.com/ollama/ollama/releases/
|
||||
# - x86_64 CPU 选择下载 ollama-linux-amd64
|
||||
# - aarch64|arm64 CPU 选择下载 ollama-linux-arm64
|
||||
# 示例：
|
||||
wget https://github.com/ollama/ollama/releases/download/v0.5.11/ollama-linux-amd64.tgz
|
||||
|
||||
# 下载 安装脚本，并放到 /home/ollama 目录下
|
||||
wget https://ollama.com/install.sh
|
||||
|
||||
# 将 ollama-linux-amd64.tgz 和 install.sh 拷贝到需要安装的机器上，如放到 /home/ollama 目录下
|
||||
# 然后执行如下命令：
|
||||
tar -zxvf ollama-linux-amd64.tgz
|
||||
chmod +x install.sh
|
||||
# 编辑 install.sh 文件，找到如下内容
|
||||
curl --fail --show-error --location --progress-bar -o $TEMP_DIR/ollama "https://ollama.com/download/ollama-linux-${ARCH}${VER_PARAM}"
|
||||
# 注释它，并在其下增加如下内容：
|
||||
cp ./ollama-linux-amd64 $TEMP_DIR/ollama
|
||||
|
||||
# 执行安装脚本
|
||||
./install.sh
|
||||
|
||||
# 模型的离线下载请参考下文模型导入部分
|
||||
```
|
||||
|
||||
### 1.3 基于 Docker 安装 ollama
|
||||
|
||||
基于 Docker 可以使得 ollama 的安装、更新与启停管理更为便捷。
|
||||
|
||||
首先确保已安装了 docker，然后执行如下命令：
|
||||
|
||||
```
|
||||
# 拉取镜像
|
||||
docker pull ollama/ollama
|
||||
|
||||
# 运行容器：CPU 模式
|
||||
docker run -d -p 11434:11434 -v /data/ollama:/root/.ollama --name ollama ollama/ollama
|
||||
# 运行容器：GPU 模式
|
||||
docker run --gpus=all -d -p 11434:11434 -v /data/ollama:/root/.ollama --name ollama ollama/ollama
|
||||
|
||||
# 进入容器 bash 下并下载模型
|
||||
docker exec -it ollama /bin/bash
|
||||
# 下载一个模型
|
||||
ollama pull deepseek-r1:8b
|
||||
```
|
||||
|
||||
也可以基于 `docker-compose` 进行启停管理。`docker-compose.yml` 参考：
|
||||
|
||||
```
|
||||
services:
|
||||
ollama:
|
||||
image:ollama/ollama
|
||||
container_name:ollama
|
||||
restart:unless-stopped
|
||||
ports:
|
||||
-11434:11434
|
||||
volumes:
|
||||
-/data/ollama:/root/.ollama
|
||||
environment:
|
||||
# 允许局域网跨域形式访问API
|
||||
OLLAMA_HOST=0.0.0.0:11434
|
||||
OLLAMA_ORIGINS=*
|
||||
```
|
||||
|
||||
### 1.4 修改 ollama 模型默认保存位置
|
||||
|
||||
`ollama` 下载的模型默认的存储目录如下：
|
||||
|
||||
- • macOS: `~/.ollama/models`
|
||||
- • Linux: `/usr/share/ollama/.ollama/models`
|
||||
- • Windows: `C:\Users\<username>\.ollama\models`
|
||||
|
||||
若默认位置存在磁盘空间告急的问题，可以通过设置环境变量 `OLLAMA_MODELS` 修改模型存储位置。示例：
|
||||
|
||||
```
|
||||
# macOS / Linux：写入环境变量配置到 ~/.bashrc 文件中
|
||||
echo 'export OLLAMA_MODELS=/data/ollama/models' >> ~/.bashrc
|
||||
source ~/.bashrc
|
||||
|
||||
# Windows：按 \`WIN+R\` 组合键并输入 cmd 打开命令提示符
|
||||
# 然后执行如下命令写入到系统环境变量中
|
||||
setx OLLAMA_MODELS D:\data\ollama\models
|
||||
```
|
||||
|
||||
如果已经下载过模型，可以从上述默认位置将 models 目录移动到新的位置。
|
||||
|
||||
对于 docker 安装模式，则可以通过挂载卷的方式修改模型存储位置。
|
||||
|
||||
### 1.5 使用：基于 API 形式访问 ollama 服务
|
||||
|
||||
ollama 安装完成并正常启动后，可以通过命令行形式运行模型（如：`ollama run deepseek-r1:1.5b`），并通过命令行交互的方式进行测试。
|
||||
|
||||
此外也可以通过访问 `http://localhost:11434` 以 API 调用的形式调用。示例：
|
||||
|
||||
```
|
||||
curl http://localhost:11434/api/generate -d '{
|
||||
"model": "deepseek-r1:8b",
|
||||
"stream": false,
|
||||
"prompt": "你是谁"
|
||||
}'
|
||||
```
|
||||
|
||||
ollama API 文档参考：
|
||||
|
||||
- • https://ollama.readthedocs.io/api/
|
||||
- • https://github.com/ollama/ollama/blob/main/docs/api.md
|
||||
|
||||
## 2 使用 ollama 下载和运行模型
|
||||
|
||||
### 2.1 使用 ollama 命令行下载和运行模型
|
||||
|
||||
执行如下命令下载并运行一个模型：
|
||||
|
||||
```
|
||||
# 基本格式为：
|
||||
ollama run <model_name:size>
|
||||
|
||||
# 例如下载并运行 deepseek-r1 的 1.5b 模型
|
||||
# 如果下载模型速度开始较快后面变慢，可以 kill 当前进程并重新执行
|
||||
ollama run deepseek-r1:1.5b
|
||||
```
|
||||
|
||||
运行成功则会进入命令行交互模式，可以直接输入问题并获得应答反馈，也可以通过 API 调用方式测试和使用。
|
||||
|
||||
从如下地址可搜索 ollama 所有支持的模型：
|
||||
|
||||
- • 中文站：https://ollama.org.cn/search
|
||||
- • 官方站：https://ollama.com/search
|
||||
|
||||
**从 HF 和魔塔社区下载模型**
|
||||
|
||||
ollama 还支持从 HF 和魔塔社区下载第三方开源模型。基本格式为：
|
||||
|
||||
```
|
||||
# 从 HF(https://huggingface.co) 下载模型的格式
|
||||
ollama run hf.co/{username}/{reponame}:latest
|
||||
# 示例：
|
||||
ollama run hf.co/bartowski/Llama-3.2-1B-Instruct-GGUF:Q8_0
|
||||
|
||||
# 从魔塔社区(https://modelscope.cn)下载模型的格式
|
||||
ollama run modelscope.cn/{username}/{model}
|
||||
# 示例：
|
||||
ollama run modelscope.cn/Qwen/Qwen2.5-3B-Instruct-GGUF:Q3_K_M
|
||||
```
|
||||
|
||||
### 2.2 使用 ollama create 导入本地模型
|
||||
|
||||
通过 `ollama run` 和 `ollama pull` 命令均是从官方地址下载模型，可能会遇到下载速度慢、下载失败等问题。
|
||||
|
||||
ollama 支持从本地导入模型。我们可以从第三方下载模型文件并使用 `ollama create` 命令导入到 ollama 中。
|
||||
|
||||
例如，假若我们下载了 `deepseek-r1:8b` 模型文件，并保存在 `/data/ollama/gguf/deepseek-r1-8b.gguf`，则可执行如下命令进行导入：
|
||||
|
||||
```
|
||||
cd /data/ollama/gguf
|
||||
echo "From ./deepeek-r1-8b.gguf" > modelfile-deepseek-r1-8b
|
||||
ollama create deepseek-r1:8b -f modelfile-deepseek-r1-8b
|
||||
|
||||
# 查看模型信息
|
||||
ollama list
|
||||
ollama show deepseek-r1:8b
|
||||
|
||||
# 运行模型(以命令行交互模式使用)
|
||||
ollama run deepseek-r1:8b
|
||||
```
|
||||
|
||||
相关文档参考：
|
||||
|
||||
- • https://ollama.readthedocs.io/import/
|
||||
- • https://ollama.readthedocs.io/modelfile/
|
||||
|
||||
## 3 ollama 常用命令参考
|
||||
|
||||
ollama 提供了丰富的命令行工具，方便用户对模型进行管理。
|
||||
|
||||
- • `ollama --help`：查看帮助信息。
|
||||
- • `ollama serve`：启动 ollama 服务。
|
||||
- • `ollama create <model-name> [-f Modelfile]`：根据一个 Modelfile 文件导入模型。
|
||||
- • `ollama show <model-name:[size]>`：显示某个模型的详细信息。
|
||||
- • `ollama run <model-name:[size]>`：运行一个模型。若模型不存在会先拉取它。
|
||||
- • `ollama stop <model-name:[size]>`：停止一个正在运行的模型。
|
||||
- • `ollama pull <model-name:[size]>`：拉取指定的模型。
|
||||
- • `ollama push <model-name>`：将一个模型推送到远程模型仓库。
|
||||
- • `ollama list`：列出所有模型。
|
||||
- • `ollama ps`：列出所有正在运行的模型。
|
||||
- • `ollama cp <source-model-name> <new-model-name>`：复制一个模型。
|
||||
- • `ollama rm <model-name:[size]>`：删除一个模型。
|
||||
|
||||
## 4 ollama 安装使用常见问题及解决
|
||||
|
||||
### 4.1 ollama 模型下载慢：离线下载与安装模型
|
||||
|
||||
通过 ollama 官方命令拉取模型，可能会遇到网速慢、下载时间过长等问题。
|
||||
|
||||
#### 4.1.1 开始快后来慢：间隔性重启下载
|
||||
|
||||
由于模型文件较大，下载过程中可能会遇到开始网速还可以，后面变慢的情况。许多网友反馈退出然后重试则速度就可以上来了，所以可以尝试通过每隔一段时间退出并重新执行的方式以保持较快的下载速率。
|
||||
|
||||
以下是基于该逻辑实现的下载脚本，注意将其中的 `deepseek-r1:7b` 替换为你希望下载的模型版本。
|
||||
|
||||
Windows 下在 powershell 中执行：
|
||||
|
||||
```
|
||||
while ($true) {
|
||||
$modelExists = ollama list | Select-String "deepseek-r1:7b"
|
||||
|
||||
if ($modelExists) {
|
||||
Write-Host "模型已下载完成！"
|
||||
break
|
||||
}
|
||||
|
||||
Write-Host "开始下载模型..."
|
||||
$process = Start-Process -FilePath "ollama" -ArgumentList "run", "deepseek-r1:7b" -PassThru -NoNewWindow
|
||||
|
||||
# 等待60秒
|
||||
Start-Sleep -Seconds 60
|
||||
|
||||
try {
|
||||
Stop-Process -Id $process.Id -Force -ErrorAction Stop
|
||||
Write-Host "已中断本次下载，准备重新尝试..."
|
||||
}
|
||||
catch {
|
||||
Write-Host "error"
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
`macOS / Linux` 下在终端中执行：
|
||||
|
||||
```
|
||||
#!/bin/bash
|
||||
|
||||
whiletrue; do
|
||||
# 检查模型是否已下载完成
|
||||
modelExists=$(ollama list | grep "deepseek-r1:7b")
|
||||
|
||||
if [ -n "$modelExists" ]; then
|
||||
echo"模型已下载完成！"
|
||||
break
|
||||
fi
|
||||
|
||||
# 启动ollama进程并记录
|
||||
echo"开始下载模型..."
|
||||
ollama run deepseek-r1:7b & # 在后台启动进程
|
||||
processId=$! # 获取最近启动的后台进程的PID
|
||||
|
||||
# 等待60秒
|
||||
sleep 60
|
||||
|
||||
# 尝试终止进程
|
||||
ifkill -0 $processId 2>/dev/null; then
|
||||
kill -9 $processId# 强制终止进程
|
||||
echo"已中断本次下载，准备重新尝试..."
|
||||
else
|
||||
echo"进程已结束，无需中断"
|
||||
fi
|
||||
done
|
||||
```
|
||||
|
||||
#### 4.1.2 通过网盘等第三方离线下载并导入 ollama 模型
|
||||
|
||||
可以通过国内的第三方离线下载模型文件，再导入到 ollama 中。详细参考 2.2 章节。
|
||||
|
||||
`deepseek-r1` 相关模型夸克网盘下载：
|
||||
|
||||
> 链接：https://pan.quark.cn/s/7fa235cc64ef 提取码：wasX
|
||||
|
||||
也可以从 魔塔社区、HuggingFace 等大模型社区搜索并下载 stuff 格式的模型文件。例如：
|
||||
|
||||
- • https://modelscope.cn/models/unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF/files
|
||||
- • https://huggingface.co/unsloth/DeepSeek-R1-GGUF
|
||||
|
||||
#### 4.1.3 从国内大模型提供站下载模型
|
||||
|
||||
ollama 支持从魔塔社区直接下载模型。其基本格式为：
|
||||
|
||||
```
|
||||
ollama run modelscope.cn/{model-id}
|
||||
```
|
||||
|
||||
一个模型仓库可能包含多个模型，可以指定到具体的模型文件名以只下载它。示例：
|
||||
|
||||
```
|
||||
ollama run modelscope.cn/Qwen/Qwen2.5-3B-Instruct-GGUF
|
||||
#
|
||||
ollama run modelscope.cn/Qwen/Qwen2.5-3B-Instruct-GGUF:qwen2.5-3b-instruct-q3_k_m.gguf
|
||||
```
|
||||
|
||||
下载 `deepseek-r1` 模型命令参考：
|
||||
|
||||
```
|
||||
# deepseek-r1:7b
|
||||
ollama run modelscope.cn/unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF:DeepSeek-R1-Distill-Qwen-7B-Q4_K_M.gguf
|
||||
# deepseek-r1:14b
|
||||
ollama run modelscope.cn/unsloth/DeepSeek-R1-Distill-Qwen-14B-GGUF:Q4_K_M
|
||||
# deepseek-r1:32b
|
||||
ollama run modelscope.cn/unsloth/DeepSeek-R1-Distill-Qwen-32B-GGUF:Q4_K_M
|
||||
```
|
||||
|
||||
此外，也可以从 HF 的国内镜像站（https://hf-mirror.com）查找和拉取模型，方法与上述类似：
|
||||
|
||||
```
|
||||
# 基本格式
|
||||
ollama run hf-mirror.com/{username}/{reponame}:{label}
|
||||
|
||||
# 示例 - 拉取 deepseek-r1:7b
|
||||
ollama run hf-mirror.com/unsloth/DeepSeek-R1-Distill-Qwen-7B-GGUF:Q4_K_M
|
||||
```
|
||||
|
||||
### 4.2 ollama 服务设置允许局域网访问
|
||||
|
||||
默认情况下 API 服务仅允许本机访问，若需要允许局域网其他设备直接访问，可修改环境变量 `OLLAMA_HOST` 为 `0.0.0.0`，并修改 `OLLAMA_ORIGINS` 为允许的域名或 IP 地址。
|
||||
|
||||
环境变量设置示例：
|
||||
|
||||
```
|
||||
# windows 命令提示符下执行：
|
||||
setx OLLAMA_HOST 0.0.0.0:11434
|
||||
setx OLLAMA_ORIGINS *
|
||||
|
||||
# macOS 终端下执行：
|
||||
launchctl setenv OLLAMA_HOST "0.0.0.0:11434"
|
||||
launchctl setenv OLLAMA_ORIGINS "*"
|
||||
```
|
||||
|
||||
**特别注意：**
|
||||
|
||||
- • **如果你是在云服务器等拥有公网IP的环境上部署，请谨慎做此设置，否则可能导致 API 服务被恶意调用。**
|
||||
- • 若需要局域网其他设备访问，请确保防火墙等安全设置允许 11434 端口访问。
|
||||
- • 若需要自定义访问端口号，可通过环境变量 `OLLAMA_HOST` 设置，如：`OLLAMA_HOST=0.0.0.0:11435`。
|
||||
|
||||
### 4.3 为 ollama API 服务访问增加 API KEY 保护
|
||||
|
||||
**为云服务器部署的服务增加 API KEY 以保护服务**
|
||||
如果你是通过云服务器部署，那么需要特别注意服务安全，避免被互联网工具扫描而泄露，导致资源被第三方利用。
|
||||
|
||||
可以通过部署 nginx 并设置代理转发，以增加 API KEY 以保护服务，同时需要屏蔽对 11434 端口的互联网直接访问形式。
|
||||
|
||||
`nginx` 配置：
|
||||
|
||||
```
|
||||
server {
|
||||
# 用于公网访问的端口
|
||||
listen 8434;
|
||||
# 域名绑定，若无域名可移除
|
||||
server_name your_domain.com;
|
||||
|
||||
location / {
|
||||
# 验证 API KEY。这里的 your_api_key 应随便修改为你希望设置的内容
|
||||
# 可通过 uuid 生成器工具随机生成一个：https://tool.lzw.me/uuid-generator
|
||||
if ($http_authorization != "Bearer your_api_key") {
|
||||
return 403;
|
||||
}
|
||||
|
||||
# 代理转发到 ollama 的 11434 端口
|
||||
proxy_pass http://localhost:11434;
|
||||
proxy_set_header Host $host;
|
||||
proxy_set_header X-Real-IP $remote_addr;
|
||||
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
|
||||
proxy_set_header X-Forwarded-Proto $scheme;
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
## 5 集成可视化工具
|
||||
|
||||
在部署了 ollama 并拉取了 deepseek 等模型后，即可通过命令行交互和 API 服务方式使用，但使用起来并不方便。
|
||||
|
||||
开源社区中有许多大模型相关的可视化工具，如 open-webui、chat-ui、cherry-studio、AnythingLLM 等，可以方便地集成 ollama API 服务，提供图形化界面使用，以实现聊天机器人、问答知识库等多元化应用。在官方文档中列举了大量较为流行的工具应用：https://ollama.readthedocs.io/quickstart/#web
|
||||
|
||||
我们后续会选择其中较为典型的工具进行集成和介绍。
|
||||
|
||||
### 5.1 示例：基于 docker 部署 open-webui 并配置集成 ollama 服务
|
||||
|
||||
Open WebUI 是一个开源的大语言模型项目，通过部署它可以得到一个纯本地运行的基于浏览器访问的 Web 服务。它提供了可扩展、功能丰富、用户友好的自托管 AI Web 界面，支持各种大型语言模型（LLM）运行器，可以通过配置形式便捷的集成 ollama、OpenAI 等提供的 API。
|
||||
|
||||
通过 Open WebUI 可以实现聊天机器人、本地知识库、图像生成等丰富的大模型应用功能。
|
||||
|
||||
在开始之前，请确保你的系统已经安装了 docker。
|
||||
|
||||
接着拉取大语言模型 `deepseek-r1:8b` 和用于 RAG 构建本地知识库的嵌入模型 `bge-m3`：
|
||||
|
||||
```
|
||||
ollama pull deepseek-r1:8b
|
||||
ollama pull bge-m3
|
||||
```
|
||||
|
||||
然后新建文件 `docker-compose.yml`，内容参考：
|
||||
|
||||
```
|
||||
services:
|
||||
open-webui:
|
||||
image:ghcr.io/open-webui/open-webui:main
|
||||
environment:
|
||||
-OLLAMA_API_BASE_URL=http://ollama:11434/api
|
||||
-HF_ENDPOINT=https://hf-mirror.com
|
||||
-WEBUI_NAME="LZW的LLM服务"
|
||||
# 禁用 OPENAI API 的请求。若你的网络环境无法访问 openai，请务必设置该项为 false
|
||||
# 否则在登录成功时，会因为同时请求了 openai 接口而导致白屏时间过长
|
||||
-ENABLE_OPENAI_API=false
|
||||
# 设置允许跨域请求服务的域名。* 表示允许所有域名
|
||||
-CORS_ALLOW_ORIGIN=*
|
||||
# 开启图片生成
|
||||
-ENABLE_IMAGE_GENERATION=true
|
||||
# 默认模型
|
||||
-DEFAULT_MODELS=deepseek-r1:8b
|
||||
# RAG 构建本地知识库使用的默认嵌入域名
|
||||
-RAG_EMBEDDING_MODEL=bge-m3
|
||||
ports:
|
||||
-8080:8080
|
||||
volumes:
|
||||
-./open_webui_data:/app/backend/data
|
||||
extra_hosts:
|
||||
# - host.docker.internal:host-gateway
|
||||
```
|
||||
|
||||
这里需注意 `environment` 环境变量部分的自定义设置。许多设置也可以通过登录后在 web 界面进行修改。
|
||||
|
||||
在该目录下执行该命令以启动服务：`docker-compose up -d`。成功后即可通过浏览器访问：`http://localhost:8080`。
|
||||
|
||||
服务镜像更新参考：
|
||||
|
||||
```
|
||||
# 拉取新镜像
|
||||
docker-compose pull
|
||||
# 重启服务
|
||||
docker-compose up -d --remove-orphans
|
||||
# 清理镜像
|
||||
docker image prune
|
||||
```
|
||||
|
||||
- • open-webui 详细文档参考：https://docs.openwebui.com/getting-started/env-configuration
|
||||
|
||||
**可选：开启“联网搜索”功能**
|
||||
|
||||
操作路径：`设置 - 联网搜索 - 启用联网搜索`
|
||||
|
||||
当前已支持接入的联网搜索引擎中，在不需要魔法上网的情况下，有 bing 和 bocha 可以选择接入。基本只需要前往注册并获取 API KEY 回填到这里即可。如果需要保护隐私数据，请不要开启并配置该功能。
|
||||
|
||||
- • 博查文档：https://aq6ky2b8nql.feishu.cn/wiki/XgeXwsn7oiDEC0kH6O3cUKtknSR
|
||||
|
||||
## 总结与参考
|
||||
|
||||
通过以上内容，我们了解了 ollama 在国内环境下的安装使用方法，并介绍了因为国内网络特色导致安装过程可能会遇到的常见问题及解决办法。希望这些内容对你有所帮助，如果你有任何问题或建议，欢迎在评论区留言交流。
|
||||
|
||||
- • ollama 官方站：https://ollama.com
|
||||
- • ollama 中文站：https://ollama.org.cn
|
||||
- • ollama 入门：https://ollama.readthedocs.io/quickstart/
|
||||
- • ollama 常见问题：https://ollama.readthedocs.io/faq/
|
||||
- • 魔塔社区：https://modelscope.cn
|
||||
- • HF Mirror：https://hf-mirror.com
|
||||
- • open-webui 文档：https://docs.openwebui.com
|
||||
@@ -0,0 +1,62 @@
|
||||
---
|
||||
title: 𝗔𝗜 𝗶𝘀 𝗘𝗻𝘁𝗲𝗿𝗶𝗻𝗴 𝘁𝗵𝗲 𝗔𝗴𝗲 𝗼𝗳 𝗔𝗴𝗲𝗻𝗰𝘆 – 𝗠𝗼𝘃𝗶𝗻𝗴 𝗕𝗲𝘆𝗼𝗻𝗱 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻
|
||||
source: https://www.linkedin.com/posts/brijpandeyji_%F0%9D%97%94%F0%9D%97%9C-%F0%9D%97%B6%F0%9D%98%80-%F0%9D%97%98%F0%9D%97%BB%F0%9D%98%81%F0%9D%97%B2%F0%9D%97%BF%F0%9D%97%B6%F0%9D%97%BB%F0%9D%97%B4-%F0%9D%98%81%F0%9D%97%B5%F0%9D%97%B2-%F0%9D%97%94%F0%9D%97%B4%F0%9D%97%B2-activity-7300006199884738562-S9dc/?utm_medium=ios_app&rcm=ACoAADE1eGIB9ndhzD0qmslDUew4rjAk2upsYtg&utm_source=social_share_send&utm_campaign=copy_link
|
||||
author:
|
||||
published:
|
||||
created: 2025-03-02
|
||||
description:
|
||||
tags:
|
||||
- agentic-ai
|
||||
- ai
|
||||
---
|
||||
𝗔𝗜 𝗶𝘀 𝗘𝗻𝘁𝗲𝗿𝗶𝗻𝗴 𝘁𝗵𝗲 𝗔𝗴𝗲 𝗼𝗳 𝗔𝗴𝗲𝗻𝗰𝘆 – 𝗠𝗼𝘃𝗶𝗻𝗴 𝗕𝗲𝘆𝗼𝗻𝗱 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻
|
||||
|
||||
AI is no longer just about automating tasks—it’s evolving into Agentic AI, where systems think, decide, adapt, and interact intelligently.
|
||||
|
||||
These AI agents operate autonomously, learning from feedback and dynamically engaging with users and external environments.
|
||||
|
||||
But what does that mean?
|
||||
|
||||
Let's break it down with the Agentic AI Layers Framework:
|
||||
|
||||
1\. Governance & Auditability – Building Trust & Compliance
|
||||
• Transparent Decision Logs – AI maintains an audit trail of its decisions.
|
||||
• Regulatory Compliance – Aligns with legal and ethical AI standards.
|
||||
• Explainability – AI justifies its reasoning for user confidence and accountability.
|
||||
|
||||
2\. Operational Independence – AI That Thinks & Acts
|
||||
• Self-Learning – Improves continuously through real-world interactions.
|
||||
• Autonomous Decision-Making – Executes tasks independently within set guidelines.
|
||||
• Automated Workflows – Enhances efficiency by streamlining processes.
|
||||
• Scalability & Real-Time Adaptation – Dynamically adjusts to demand and insights.
|
||||
|
||||
3\. External Interactions & Multi-Modal Interfaces – Seamless AI-Human Collaboration
|
||||
• API Integrations – AI connects with external data sources and tools.
|
||||
• Multi-Modal Support – Engages via text, voice, images, and beyond.
|
||||
• Natural Language Understanding – Processes and responds intelligently to human queries.
|
||||
|
||||
4\. Ethics & Safety – Ensuring Responsible AI Development
|
||||
• Privacy Protection – Secure data handling in compliance with regulations.
|
||||
• Bias Detection & Mitigation – Actively identifies and corrects biases.
|
||||
• Harm Prevention – Prevents misinformation and harmful outputs.
|
||||
|
||||
5\. Knowledge Base & RAG (Retrieval-Augmented Generation) – AI with a Stronger Memory
|
||||
• Contextual Retrieval – Fetches relevant information for precise, context-aware responses.
|
||||
• Fact-Checking – Cross-verifies data before generating content.
|
||||
• Domain-Specific Intelligence – AI tailored for finance, healthcare, legal, and other specialized fields.
|
||||
|
||||
6\. LLM & Generative Capabilities – AI That Thinks Deeper
|
||||
• Reasoning & Adaptability – Understands complex queries and adapts to intent.
|
||||
• Real-Time Data Access – Enhances responses with up-to-date information.
|
||||
• Continuous Fine-Tuning – Learns and improves over time.
|
||||
|
||||
Why Does This Matter?
|
||||
As AI shifts toward autonomy, balancing efficiency, transparency, and ethical responsibility is critical.
|
||||
|
||||
Industries like finance, healthcare, cybersecurity, and enterprise automation stand to gain immensely—but only if we build AI that operates responsibly.
|
||||
|
||||
Your Take?
|
||||
|
||||
Should AI be fully autonomous, or should human oversight always be required?
|
||||
|
||||
|
||||
213
AI/🟠API Key.md
Normal file
213
AI/🟠API Key.md
Normal file
@@ -0,0 +1,213 @@
|
||||
#api-key #deepseek #gemini #google #aws #x #notion #n8n #github #wavespeed #siliconflow #airtable #brightdata #telegram
|
||||
|
||||
```table-of-contents
|
||||
title:
|
||||
style: nestedList # TOC style (nestedList|nestedOrderedList|inlineFirstLevel)
|
||||
minLevel: 0 # Include headings from the specified level
|
||||
maxLevel: 0 # Include headings up to the specified level
|
||||
include:
|
||||
exclude:
|
||||
includeLinks: true # Make headings clickable
|
||||
hideWhenEmpty: false # Hide TOC if no headings are found
|
||||
debugInConsole: false # Print debug info in Obsidian console
|
||||
```
|
||||
|
||||
## Gemini API Key #gemini
|
||||
### ishenwei00@gmail.com
|
||||
```
|
||||
AIzaSyALe0MnjDmTRf7zgn87vxLUe7aKfzoZRgY
|
||||
```
|
||||
|
||||
## DeepSeek #deepseek
|
||||
|
||||
```
|
||||
sk-a309a673569743ebb05d0991d3f6e51a
|
||||
```
|
||||
|
||||
|
||||
---
|
||||
## Telegram HTTP API #telegram
|
||||
|
||||
### ishenwei_bot
|
||||
t.me/ishenwei_bot
|
||||
```
|
||||
8134005762:AAHVjACJ4egbEPNY0-oiihWTM30fVt4rIoc
|
||||
```
|
||||
|
||||
### Telegram OpenClaw Bot
|
||||
|
||||
#### 星辉
|
||||
t.me/shenwei_macmini_xinghui_bot
|
||||
```
|
||||
8709222939:AAEfvZrvvU5vZFsmacsR5nmpkJ2Jb5JgfRg
|
||||
|
||||
#telegram user id
|
||||
5038825565
|
||||
```
|
||||
#### 星曜
|
||||
t.me/shenwei_macmini_xingyao_bot
|
||||
```
|
||||
8414432613:AAG9hvKfILGSsbc1EMEZW1QVym9Quc5aHWk
|
||||
```
|
||||
|
||||
|
||||
## Google API Key #google
|
||||
|
||||
n8n-workflow OAuth 2.0 Client ID
|
||||
```
|
||||
109190465048-ndh8t3ngec7sqds0ll716knt7laffirk.apps.googleusercontent.com
|
||||
```
|
||||
|
||||
Client Secret:
|
||||
```
|
||||
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
|
||||
```
|
||||
|
||||
## AWS #aws
|
||||
```
|
||||
AWS Account: 551360491749
|
||||
Access Key AKIAYAX5FODS42V2CYUQ
|
||||
Secret Access Key H9/b1/87fgpv4ZgzOTdg3rza9fLT2ac6KlrdurzF
|
||||
```
|
||||
|
||||
## News API Key
|
||||
https://newsapi.org/
|
||||
d2bf79c13a9e4feb80422c9d4ca6404a
|
||||
|
||||
Definition
|
||||
|
||||
```
|
||||
GET https://newsapi.org/v2/everything?q=Apple&from=2025-03-08&sortBy=popularity&apiKey=API_KEY
|
||||
```
|
||||
Example request
|
||||
|
||||
```bash
|
||||
curl https://newsapi.org/v2/everything -G \
|
||||
-d q=Apple \
|
||||
-d from=2025-03-08 \
|
||||
-d sortBy=popularity \
|
||||
-d apiKey=d2bf79c13a9e4feb80422c9d4ca6404a
|
||||
```
|
||||
|
||||
|
||||
## X #x
|
||||
```
|
||||
API Key: 3WzvwLqw5ZN1GsJzQ0W7K6t6H
|
||||
API Key Secret: msYmcAuVKrBqMjfk6rgRucmuDwKRfhoZCTlgkaD4FKiOlAm57Y
|
||||
|
||||
OAUTH
|
||||
Client ID: d3k2eVNoYXY0REFoX2dvVEg2a0E6MTpjaQ
|
||||
Client Secret: wbPcvt-qAbigVFa4Jn9Bj0lyl4W6ie2bvZJrcfp81MF5Rptwps
|
||||
```
|
||||
|
||||
|
||||
|
||||
## Notion #notion
|
||||
https://www.notion.com/my-integrations
|
||||
Internal Integration Secret:
|
||||
```
|
||||
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
|
||||
```
|
||||
|
||||
任务调度
|
||||
```
|
||||
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
|
||||
```
|
||||
|
||||
|
||||
## Pexel #plex
|
||||
```
|
||||
uVZ6Benfr5yzaG8c8er1K6u4r3a4JXWw9AMsYIhorw9GhRfQ5Vzxd8S5
|
||||
```
|
||||
|
||||
|
||||
## Wavespeed API key #wavespeed
|
||||
```
|
||||
b023e330aef99c65cb2a1801d6042a70a020cb645cd7383d7ed0bc54a750ce35
|
||||
```
|
||||
|
||||
## n8n API key #n8n
|
||||
```
|
||||
eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJzdWIiOiJjOThlZjhiYS05ZTVlLTQ2ZGMtYWU0OC02YjMyN2FmYmY1MjUiLCJpc3MiOiJuOG4iLCJhdWQiOiJwdWJsaWMtYXBpIiwiaWF0IjoxNzcwOTcwNTQyfQ.lJvm9rWh4hRTKQ1OL-BgkwQnuoUyzgEo62OsD5JuThk
|
||||
```
|
||||
|
||||
|
||||
## Github Personal Access Token(classic) #github
|
||||
Clawhub installation (only have public_repo readonly permission)
|
||||
```
|
||||
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
|
||||
```
|
||||
|
||||
## Siliconflow API key #siliconflow
|
||||
|
||||
```
|
||||
sk-ssdzoysqyppfaoubcpwrwzlcmbifoumpqchgisyawgwgrfia
|
||||
```
|
||||
|
||||
|
||||
## Stable Horde
|
||||
```
|
||||
7kvqIQs62Asyzj1I0UMhfQ
|
||||
```
|
||||
|
||||
|
||||
## BrightData #brightdata
|
||||
```
|
||||
011ac709c39e73762ef01946f0ca17b151e8c612e4c532e87764c23c61047ecf
|
||||
```
|
||||
|
||||
## Airtable #airtable
|
||||
```
|
||||
patB48t4Nl1WKftUs.ef6e99b44095d7da80778b872addef3fa27b5079e7408e62afb3817c3479c8da
|
||||
```
|
||||
|
||||
## OpenAI
|
||||
### ishenwei@gmail.com
|
||||
```
|
||||
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
|
||||
```
|
||||
|
||||
## OpenCode Zen
|
||||
```
|
||||
sk-70Kjcr1Au8CdM5CvIQz6FHvR5AfhtwtvuerY3sBsy6vaXGGkTcN2arhFmAV0auJh
|
||||
```
|
||||
|
||||
|
||||
## 飞书
|
||||
App ID:
|
||||
```
|
||||
cli_a93a4a4624e19bc9
|
||||
```
|
||||
App Secret
|
||||
```
|
||||
xfZKkekUhARQ3DWQ65GOVhCqCNO4ckGV
|
||||
```
|
||||
Verification Token
|
||||
```
|
||||
nz3l8CEvSsUvmJb6LDhKrd24zjWKDxiM
|
||||
```
|
||||
|
||||
8566920841:AAEfvOFAZ86fPKQdZ9Dm4-wnR46Asm7B7nU
|
||||
|
||||
|
||||
## MiniMax
|
||||
|
||||
```
|
||||
sk-cp-H0FwKNry9PnMJmLng7W51OfbN6XWbfN_9pfMnI89smCmbPNIHzUuOibPtzikdK8rzRuB9uuunGmN_SPoOBZOUgy2_D9Sm3_ivQ1LYc5Cm48cpC2mQ07hDnE
|
||||
```
|
||||
|
||||
## Tavily API Key
|
||||
```
|
||||
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
|
||||
```
|
||||
|
||||
## OpenRouter
|
||||
For OpenClaw
|
||||
```
|
||||
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
|
||||
```
|
||||
|
||||
For Claude Code
|
||||
```
|
||||
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
|
||||
```
|
||||
Reference in New Issue
Block a user