https://i68.ltd/notes/Recent content on i68,爱六八,链接你我他-i68.ltd-随笔Hugozh-Hanshttps://i68.ltd/notes/posts/20250408-llm-code-read/Sat, 26 Apr 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250408-llm-code-read/git仓库代码阅读神器 GitDiagram-GitHub仓库秒变交互式架构图 让代码阅读体验彻底颠覆的神器 能将任何GitHub代码库秒变交互式https://i68.ltd/notes/posts/20250407-llm-person-kb/Sat, 26 Apr 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250407-llm-person-kb/第二大脑-个人知识库 Quivr-第二大脑一样的个人助手，利用AI技术增强个人生产力 将 GenAI 集成到您的应用程序中的个性化 RAG,专注于您的产品而非https://i68.ltd/notes/posts/20250406-llm-rag2/Sun, 20 Apr 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250406-llm-rag2/RAG框架精选2 清华Hyper-RAG 使用超图驱动的RAG对抗LLM幻觉 论文链接:https://arxiv.org/pdf/2504.08https://i68.ltd/notes/posts/20250405-visual-agent/Tue, 15 Apr 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250405-visual-agent/可视化的智能体 pyspur-可视化的智能体游乐场 智能体工作流的可视化游乐场，10倍加速智能体迭代 官网:https://www.pyspur.https://i68.ltd/notes/posts/20250404-llm-framework3/Sun, 13 Apr 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250404-llm-framework3/llm开发框架新秀 google开源ADK-Agent Development Kit 开源的、代码优先的 Python 工具包，用于构建、评估和部署具有灵活性和控制力的复杂智能体 项目仓https://i68.ltd/notes/posts/20250403-llm-docs-parse/Sun, 13 Apr 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250403-llm-docs-parse/llm与文档解析处理神器 免费全能的文档解析神器-MinerU 一站式开源高质量数据提取工具，将PDF转换成Markdown和JSON格式。 精准https://i68.ltd/notes/posts/20250402-multi-agent/Sun, 06 Apr 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250402-multi-agent/多智能体优秀开发框架 motia-面向软件工程师的智能体框架 项目仓库:https://github.com/MotiaDev/motia 1.3k 官方https://i68.ltd/notes/posts/20250401-deep-research3/Sun, 06 Apr 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250401-deep-research3/深度研究deep-research优秀开源项目 DeepResearcher-基于LLM的深度研究代理端到端训练的综合框架 论文链接:httpshttps://i68.ltd/notes/posts/20250312-llm-rag2/Sun, 30 Mar 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250312-llm-rag2/llm检索增强创新 HippoRAG 2 海马体记忆索引 HippoRAG 2是LLMs的一个强大的记忆框架，它增强了他们识别和利用新知识中的联系的能力，这反映了人类长期记忆的一https://i68.ltd/notes/posts/20250312-llm-framework2/Sun, 30 Mar 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250312-llm-framework2/llm主流开发框架 AutoGen-微软推出多智能体应用程序开发框架 AutoGen 是用于创建多智能体 AI 应用程序的开发框架，这些应用程序可以自主行动或与人类https://i68.ltd/notes/posts/20250313-rust-wasm-fw/Sun, 30 Mar 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250313-rust-wasm-fw/rust服务应用开发框架 spin 开源开发工具，用于构建和运行由 WebAssembly 提供支持的无服务器应用程序 Spin 是一个框架，用于使用 WebAssembly 构建，部署和运行快速，安全和可组https://i68.ltd/notes/posts/20250308-deepseek-news/Sun, 23 Mar 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250308-deepseek-news/DeepSeek满血推理集锦 美团Int8满血DeepSeek R1部署 美团基于SGLang提供INT8无损满血版DeepSeek R1部署方案 对https://i68.ltd/notes/posts/20250311-llm-framework/Sun, 23 Mar 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250311-llm-framework/llm优秀开发框架 Eino Eino 是字节跳动内部大模型应用的首选全代码开发框架，已有包括豆包、抖音、扣子等多条业务线、数百个服务接入使用 Eino 是覆盖 devops 全流程https://i68.ltd/notes/posts/20250309-llm-tuning/Sun, 23 Mar 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250309-llm-tuning/大语言模型训练优化 fp8训练COAT 减少40%显存占用，训练速度提高1.4倍，精度接近无损 论文链接:https://arxiv.org/pdhttps://i68.ltd/notes/posts/20250307-search-r1/Fri, 14 Mar 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250307-search-r1/Search-R1、ReSearch、R1-Searcher 和 Search-o1 Search-R1 DeepSeek团队开发的SEARCH-R1模型通过强化学习，让AI学会了https://i68.ltd/notes/posts/20250305-deep-research2/Fri, 14 Mar 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250305-deep-research2/深度研究deep-research优秀开源项目 港大开源AI科研神器AI-Researcher 项目仓库:https://github.com/https://i68.ltd/notes/posts/20250310-llm-db/Mon, 10 Mar 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250310-llm-db/llm数据存储基础设施 infinity 专为LLM应用程序构建的AI原生数据库，可提供对密集向量、稀疏向量、张量（多向量）和全文的快速混合搜索 项目仓库:hthttps://i68.ltd/notes/posts/250306-opensource-agi-agent/Fri, 07 Mar 2025 00:00:00 +0000https://i68.ltd/notes/posts/250306-opensource-agi-agent/Manus开源平替-开源通用智能体 OWL-比Manus还强的全能开源Agent OWL: Optimized Workforce Learning for General Multi-Agent Assistance in Real-World Task Automation，现实世界中执行自动https://i68.ltd/notes/posts/20250303-llm-small+large/Sun, 02 Mar 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250303-llm-small+large/SLM+LLM-大小搭配，干活不累 minions-本地小模型与云端大模型结合 把设备端的LLMs（如ollama）与云端前沿的LLMs（像ophttps://i68.ltd/notes/posts/20250304-llm-fine-tuning-reason/Sun, 02 Mar 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250304-llm-fine-tuning-reason/模型推理增强微调 知识融合FuseAI 能融合多个模型，降低训练成本，提升推理性能 论文链接:https://arxiv.org/abs/2408https://i68.ltd/notes/posts/20250302-llm-reason-news/Sat, 01 Mar 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250302-llm-reason-news/推理增强框架 TPO（Test-Time Prompt Optimization）推理即优化 论文链接:https://arxiv.org/abs/2501.https://i68.ltd/notes/posts/20250301-deep-research/Sat, 01 Mar 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250301-deep-research/香港大学deep-research开源项目 Deep Research — 搜索的终结，研究的开始 OpenAI的Deep Research，既具备人类专家般深度与细腻度，https://i68.ltd/notes/posts/20250103-i68qa/Tue, 14 Jan 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250103-i68qa/i68问答 网站主页:https://qa.i68.ltd 实时网页搜索 答案引用参考 上下文内追问 界面简洁明了 响应轻便快速 提供更准确、更新鲜的信息https://i68.ltd/notes/posts/20250104-slm/Tue, 14 Jan 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250104-slm/slm-小马拉大车,小车扛大炮 简要 slm,Small Language Models,小语言模型 模型虽小，能力很强 rStar-Math 7B小模型数学能力直逼o1 论文地址:https://arxhttps://i68.ltd/notes/posts/20250102-llm-story/Sat, 04 Jan 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250102-llm-story/故事可视化AI StoryWeaver故事可视化 通过知识增强的角色定制技术，实现高质量的故事可视化 论文链接:https://arxiv.orghttps://i68.ltd/notes/posts/20250101-vide-translation/Wed, 01 Jan 2025 00:00:00 +0000https://i68.ltd/notes/posts/20250101-vide-translation/视频智能翻译 EasyVideoTrans英文视频转换成中文视频 EasyVideoTrans简要 最快的英文视频转中文方案 由B站多位程序员Up主https://i68.ltd/notes/posts/20241206-llm-enh/Wed, 25 Dec 2024 00:00:00 +0000https://i68.ltd/notes/posts/20241206-llm-enh/LLM思维增强 连续思维链Coconut Coconut简要 Coconut:Chain of Continuous Thought，连续思维链 论文链接:https://arxiv.org/pdfhttps://i68.ltd/notes/posts/20241207-mcp-list/Wed, 25 Dec 2024 00:00:00 +0000https://i68.ltd/notes/posts/20241207-mcp-list/MCP模型上下文协议集锦 MCP模型上下文协议简介 MCP模型上下文协议（Model-Context-Protocol，MCP），能够实现大型语https://i68.ltd/notes/posts/241205-doc2md/Mon, 23 Dec 2024 00:00:00 +0000https://i68.ltd/notes/posts/241205-doc2md/转化为MarkDown markitdown markitdown简介 将文件和办公文档，如pdf,doc等转换为Markdown的Python工具 项目仓库:htthttps://i68.ltd/notes/posts/241204-llm-fin/Fri, 20 Dec 2024 00:00:00 +0000https://i68.ltd/notes/posts/241204-llm-fin/LLM与金融 RD-Agent RD-Agent简介 项目仓库:https://github.com/microsoft/rd-agent 项目网址:https:https://i68.ltd/notes/posts/241203-llms-uni/Fri, 13 Dec 2024 00:00:00 +0000https://i68.ltd/notes/posts/241203-llms-uni/LLM统一接口 one-api OpenAI 接口管理 one-api 简介 OpenAI 接口管理 & 分发系统 项目仓库 支持 Azure、Anthropic Claude、Google PaLM 2 & Gemini、https://i68.ltd/notes/posts/241202-agent-integration/Thu, 12 Dec 2024 00:00:00 +0000https://i68.ltd/notes/posts/241202-agent-integration/智能体Agent集成整合 智能体Agent趋势 比尔盖茨：Agent 将是 AI 最大的赛道 Agent应用开发技术-全景图 2024年AI智能体(Agenhttps://i68.ltd/notes/posts/241201-cloud-ai-ide/Sat, 07 Dec 2024 00:00:00 +0000https://i68.ltd/notes/posts/241201-cloud-ai-ide/云端智能IDE 谷歌IDX 谷歌IDX简介 官网:https://idx.google.com 谷歌推出新一代 AI 代码编辑器，不仅支持自然语言交互，还https://i68.ltd/notes/posts/241112-tools-myip/Tue, 26 Nov 2024 00:00:00 +0000https://i68.ltd/notes/posts/241112-tools-myip/超好用的Ip工具箱-MyIp MyIp简要介绍 一个开源免费的 IP工具箱。 轻松检查本地的 IP 信息，包含IP 地理位置，DNS泄漏，WebRTC 连接，https://i68.ltd/notes/posts/241111-critics/Mon, 25 Nov 2024 00:00:00 +0000https://i68.ltd/notes/posts/241111-critics/CritiCS创意故事生成框架 CritiCS简要 CritiCS:Collective Critics for Creative Story Generation，协作式创意故事生成 论文链接:https://arxiv.orhttps://i68.ltd/notes/posts/241110-generative-ai-for-beginners/Sat, 23 Nov 2024 00:00:00 +0000https://i68.ltd/notes/posts/241110-generative-ai-for-beginners/面向初学者的生成式人工智能课程 内容摘要 最懂开发者的微软开源的大模型入门课，面向初学者的生成式人工智能课程！通过实战项目方式，渐进式引入大模型https://i68.ltd/notes/posts/241109-llm-mcts/Thu, 21 Nov 2024 00:00:00 +0000https://i68.ltd/notes/posts/241109-llm-mcts/MCTS蒙特卡罗树搜索 相关介绍 蒙特卡罗树搜索（Monte Carlo Tree Search MCTS）是一种广泛应用于博弈和复杂决策过程的决策算法，它通过构建搜索树并模拟https://i68.ltd/notes/posts/241108-llm-medical/Wed, 20 Nov 2024 00:00:00 +0000https://i68.ltd/notes/posts/241108-llm-medical/llm与医疗 MMedAgent医疗领域的多模态医疗AI智能体 MMedAgent简介 论文链接:https://arxiv.org/html/2https://i68.ltd/notes/posts/241107-llm-rag/Tue, 19 Nov 2024 00:00:00 +0000https://i68.ltd/notes/posts/241107-llm-rag/RAG框架精选 EasyRAG：重新定义网络操作的高效轻量级生成框架 EasyRAG简介 高效的检索增强生成框架，用于自动化网络操作 论文链接:hthttps://i68.ltd/notes/posts/241106-llm-deploy/Fri, 15 Nov 2024 00:00:00 +0000https://i68.ltd/notes/posts/241106-llm-deploy/LLM部署最佳实践 Ollama简单易用 官网:https://ollama.com 仓库:https://github.com/ollama/ohttps://i68.ltd/notes/posts/241105-philosophy-software-design/Thu, 14 Nov 2024 00:00:00 +0000https://i68.ltd/notes/posts/241105-philosophy-software-design/软件设计的哲学 A Philosophy of Software Design 内容简介 斯坦福教授、Tcl 语言发明者 John Ousterhout 的著作《A Philosophy of Software Design》，自出版以来，好评如潮。按照 IT 图书出版的惯例，https://i68.ltd/notes/posts/241104-self-discover/Tue, 12 Nov 2024 00:00:00 +0000https://i68.ltd/notes/posts/241104-self-discover/Google’s Self-Discover 简要介绍 Self-Discover: Large Language Models Self-Compose Reasoning Structures 大语言模型自组织推理结构 论文链接:https://arxiv.org/pdf/2402.03620 参考实现:htthttps://i68.ltd/notes/posts/241103-simple-mind-map/Mon, 11 Nov 2024 00:00:00 +0000https://i68.ltd/notes/posts/241103-simple-mind-map/simple-mind-map简单而强大的Web思维导图 简介 简单而强大的Web思维导图 项目仓库:https://github.com/wanhttps://i68.ltd/notes/posts/241102-prompt-auto/Mon, 11 Nov 2024 00:00:00 +0000https://i68.ltd/notes/posts/241102-prompt-auto/提示工程自动优化框架 DSPy: Programming—not prompting—Language Models DSPy简介 项目主站:https://dspy.ahttps://i68.ltd/notes/posts/241101-magentic-one/Sat, 09 Nov 2024 00:00:00 +0000https://i68.ltd/notes/posts/241101-magentic-one/微软开源多智能体系统Magentic-One 简介 Magentic-One 是微软开源的通用多智能体系统，能够像人类专家一样解决复杂的网络和文件任务，性能比肩 SOTAhttps://i68.ltd/notes/posts/241001-qwen-2-5-code-interpreter/Tue, 22 Oct 2024 00:00:00 +0000https://i68.ltd/notes/posts/241001-qwen-2-5-code-interpreter/Qwen 2.5 Code Interpreter 简介 项目仓库:https://github.com/cfahlgren1/qwen-2.5-code-interpreter Qwen 2.5 Coder 1.5Bhttps://i68.ltd/notes/posts/240916-qwen2.5/Thu, 19 Sep 2024 00:00:00 +0000https://i68.ltd/notes/posts/240916-qwen2.5/qwen2.5 简介 项目仓库:https://github.com/QwenLM/Qwen2.5 Qwen2.5-Coder:https://github.com/QwenLM/Qwen2.5-Coder 相关文章:https://qwenlm.github.ihttps://i68.ltd/notes/posts/240915-openai-o1/Wed, 18 Sep 2024 00:00:00 +0000https://i68.ltd/notes/posts/240915-openai-o1/OpenAI o1开源实现 Hugging Face开源Llama版o1 简要 Hugging Face官方发文，开源了扩展测试时计算的方法 项目仓库:https://github.comhttps://i68.ltd/notes/posts/240914/Wed, 18 Sep 2024 00:00:00 +0000https://i68.ltd/notes/posts/240914/移动云智算平台 简介 模型开发（IDE） 智算-910B型 华为昇腾 910B 智算-A800型 NVIDIA A800 智算-BIV150型 天数智芯 智算-BR104P型 壁仞科技,https://i68.ltd/notes/posts/240913-canvas-confetti/Fri, 13 Sep 2024 00:00:00 +0000https://i68.ltd/notes/posts/240913-canvas-confetti/canvas-confetti 简介 Canvas Confetti 是一个基于 JavaScript 的特效动画库，可以在网页界面上轻松地实现五彩纸屑飞舞的庆祝场景特效 项目站点:https://www.kirilv.cohttps://i68.ltd/notes/posts/240912/Tue, 03 Sep 2024 00:00:00 +0000https://i68.ltd/notes/posts/240912/轻量级的 AI 证件照制作工具：HivisionIDPhotos 项目地址:https://github.com/Zeyi-Lin/Hivisionhttps://i68.ltd/notes/posts/240910/Tue, 03 Sep 2024 00:00:00 +0000https://i68.ltd/notes/posts/240910/一条 SQL 生成网页的工具：SQLpage 项目地址:https://github.com/lovasoa/SQLpage 用 Rust 编写的基于 SQL 的 Web 应用构https://i68.ltd/notes/posts/240911/Tue, 03 Sep 2024 00:00:00 +0000https://i68.ltd/notes/posts/240911/在浏览器里玩《暗黑破坏神》：diabloweb 项目地址:<github.com/d07RiV/diabloweb> 将经典的动作角https://i68.ltd/notes/posts/240810/Thu, 29 Aug 2024 00:00:00 +0000https://i68.ltd/notes/posts/240810/开源自动驾驶平台:Donkeycar Donkeycar简介 Donkeycar是一个专为爱好者和学生设计的自动驾驶小车平台，它不仅仅是一个Py