[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-langwatch--langwatch":3,"tool-langwatch--langwatch":61},[4,18,26,36,44,53],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":17},4358,"openclaw","openclaw\u002Fopenclaw","OpenClaw 是一款专为个人打造的本地化 AI 助手，旨在让你在自己的设备上拥有完全可控的智能伙伴。它打破了传统 AI 助手局限于特定网页或应用的束缚，能够直接接入你日常使用的各类通讯渠道，包括微信、WhatsApp、Telegram、Discord、iMessage 等数十种平台。无论你在哪个聊天软件中发送消息，OpenClaw 都能即时响应，甚至支持在 macOS、iOS 和 Android 设备上进行语音交互，并提供实时的画布渲染功能供你操控。\n\n这款工具主要解决了用户对数据隐私、响应速度以及“始终在线”体验的需求。通过将 AI 部署在本地，用户无需依赖云端服务即可享受快速、私密的智能辅助，真正实现了“你的数据，你做主”。其独特的技术亮点在于强大的网关架构，将控制平面与核心助手分离，确保跨平台通信的流畅性与扩展性。\n\nOpenClaw 非常适合希望构建个性化工作流的技术爱好者、开发者，以及注重隐私保护且不愿被单一生态绑定的普通用户。只要具备基础的终端操作能力（支持 macOS、Linux 及 Windows WSL2），即可通过简单的命令行引导完成部署。如果你渴望拥有一个懂你",349277,3,"2026-04-06T06:32:30",[13,14,15,16],"Agent","开发框架","图像","数据工具","ready",{"id":19,"name":20,"github_repo":21,"description_zh":22,"stars":23,"difficulty_score":10,"last_commit_at":24,"category_tags":25,"status":17},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,"2026-04-05T11:01:52",[14,15,13],{"id":27,"name":28,"github_repo":29,"description_zh":30,"stars":31,"difficulty_score":32,"last_commit_at":33,"category_tags":34,"status":17},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",143909,2,"2026-04-07T11:33:18",[14,13,35],"语言模型",{"id":37,"name":38,"github_repo":39,"description_zh":40,"stars":41,"difficulty_score":32,"last_commit_at":42,"category_tags":43,"status":17},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107888,"2026-04-06T11:32:50",[14,15,13],{"id":45,"name":46,"github_repo":47,"description_zh":48,"stars":49,"difficulty_score":32,"last_commit_at":50,"category_tags":51,"status":17},4721,"markitdown","microsoft\u002Fmarkitdown","MarkItDown 是一款由微软 AutoGen 团队打造的轻量级 Python 工具，专为将各类文件高效转换为 Markdown 格式而设计。它支持 PDF、Word、Excel、PPT、图片（含 OCR）、音频（含语音转录）、HTML 乃至 YouTube 链接等多种格式的解析，能够精准提取文档中的标题、列表、表格和链接等关键结构信息。\n\n在人工智能应用日益普及的今天，大语言模型（LLM）虽擅长处理文本，却难以直接读取复杂的二进制办公文档。MarkItDown 恰好解决了这一痛点，它将非结构化或半结构化的文件转化为模型“原生理解”且 Token 效率极高的 Markdown 格式，成为连接本地文件与 AI 分析 pipeline 的理想桥梁。此外，它还提供了 MCP（模型上下文协议）服务器，可无缝集成到 Claude Desktop 等 LLM 应用中。\n\n这款工具特别适合开发者、数据科学家及 AI 研究人员使用，尤其是那些需要构建文档检索增强生成（RAG）系统、进行批量文本分析或希望让 AI 助手直接“阅读”本地文件的用户。虽然生成的内容也具备一定可读性，但其核心优势在于为机器",93400,"2026-04-06T19:52:38",[52,14],"插件",{"id":54,"name":55,"github_repo":56,"description_zh":57,"stars":58,"difficulty_score":10,"last_commit_at":59,"category_tags":60,"status":17},4487,"LLMs-from-scratch","rasbt\u002FLLMs-from-scratch","LLMs-from-scratch 是一个基于 PyTorch 的开源教育项目，旨在引导用户从零开始一步步构建一个类似 ChatGPT 的大型语言模型（LLM）。它不仅是同名技术著作的官方代码库，更提供了一套完整的实践方案，涵盖模型开发、预训练及微调的全过程。\n\n该项目主要解决了大模型领域“黑盒化”的学习痛点。许多开发者虽能调用现成模型，却难以深入理解其内部架构与训练机制。通过亲手编写每一行核心代码，用户能够透彻掌握 Transformer 架构、注意力机制等关键原理，从而真正理解大模型是如何“思考”的。此外，项目还包含了加载大型预训练权重进行微调的代码，帮助用户将理论知识延伸至实际应用。\n\nLLMs-from-scratch 特别适合希望深入底层原理的 AI 开发者、研究人员以及计算机专业的学生。对于不满足于仅使用 API，而是渴望探究模型构建细节的技术人员而言，这是极佳的学习资源。其独特的技术亮点在于“循序渐进”的教学设计：将复杂的系统工程拆解为清晰的步骤，配合详细的图表与示例，让构建一个虽小但功能完备的大模型变得触手可及。无论你是想夯实理论基础，还是为未来研发更大规模的模型做准备",90106,"2026-04-06T11:19:32",[35,15,13,14],{"id":62,"github_repo":63,"name":64,"description_en":65,"description_zh":66,"ai_summary_zh":66,"readme_en":67,"readme_zh":68,"quickstart_zh":69,"use_case_zh":70,"hero_image_url":71,"owner_login":64,"owner_name":72,"owner_avatar_url":73,"owner_bio":74,"owner_company":75,"owner_location":75,"owner_email":75,"owner_twitter":75,"owner_website":75,"owner_url":76,"languages":77,"stars":114,"forks":115,"last_commit_at":116,"license":117,"difficulty_score":118,"env_os":119,"env_gpu":120,"env_ram":120,"env_deps":121,"category_tags":129,"github_topics":131,"view_count":32,"oss_zip_url":75,"oss_zip_packed_at":75,"status":17,"created_at":145,"updated_at":146,"faqs":147,"releases":148},5083,"langwatch\u002Flangwatch","langwatch","The platform for LLM evaluations and AI agent testing","LangWatch 是一个专为大语言模型（LLM）评估与 AI 智能体测试打造的开源平台。它旨在帮助团队在应用发布前及生产环境中，对 AI 智能体进行端到端的测试、模拟、评估与监控，解决开发者无需自行构建复杂工具即可实现回归测试和系统可观测性的难题。\n\n无论是正在开发复杂 AI 应用的工程师，还是负责模型效果优化的研究人员，LangWatch 都能提供全方位的支持。其核心亮点在于支持“全栈”智能体模拟，能够运行包含工具调用、状态管理及用户模拟的真实场景，精准定位智能体决策失误的根本原因。平台独创性地将追踪、数据集管理、自动化评估、提示词优化整合为闭环工作流，消除了不同工具间的割裂感。\n\n此外，LangWatch 基于 OpenTelemetry 标准构建，天然兼容各类框架与大模型提供商，避免了厂商锁定风险。它还特别注重团队协作效率，支持通过标注功能让领域专家标记边缘案例，并结合 GitHub 集成将提示词版本与运行追踪直接关联，确保迭代过程透明可控。通过本地部署或云端服务，LangWatch 让用户在掌握系统控制权的同时，系统性提升 AI 应用的可靠性、性能并优化成本。","\u003Cimg width=\"1212\" height=\"395\" alt=\"012d1688-24ae-4759-ae70-5f8f81a13c0e\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flangwatch_langwatch_readme_93f72f04a2b9.png\" \u002F>\n\n\n\u003Ch3 align=\"center\">\n    \u003Ca href=\"https:\u002F\u002Flangwatch.ai\">Website\u003C\u002Fa> · \u003Ca href=\"https:\u002F\u002Fdocs.langwatch.ai\">Docs\u003C\u002Fa> · \u003Ca href=\"https:\u002F\u002Fdiscord.gg\u002FkT4PhDS2gH\">Discord\u003C\u002Fa> · \u003Ca href=\"https:\u002F\u002Fdocs.langwatch.ai\u002Fself-hosting\u002Foverview\">Self-hosting\u003C\u002Fa>\n\u003C\u002Fh3>\n\n\u003Cp align=\"center\">\n\u003Ca href=\"https:\u002F\u002Fdiscord.gg\u002FkT4PhDS2gH\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fdiscord\u002F1227886780536324106?logo=discord&labelColor=%20%235462eb&logoColor=%20%23f5f5f5&color=%20%235462eb\" alt=\"chat on Discord\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fpypi.python.org\u002Fpypi\u002Flangwatch\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Flangwatch?logo=python&logoColor=white&label=pypi%20langwatch&color=blue\" alt=\"langwatch Python package on PyPi\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Flangwatch\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fnpm\u002Fdm\u002Flangwatch?logo=npm&logoColor=white&label=npm%20langwatch&color=blue\" alt=\"langwatch npm package\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Ftwitter.com\u002Fintent\u002Ffollow?screen_name=langwatchai\" target=\"_blank\">\n   \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Ftwitter\u002Ffollow\u002Flangwatchai?logo=X&color=%20%23f5f5f5\"\n      alt=\"follow on X\">\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003Cvideo src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fff49882d-4e9d-4b7c-819b-be690fba9387\" autoplay loop muted playsinline width=\"100%\" style=\"display: block; aspect-ratio: 16 \u002F 9;\">\u003C\u002Fvideo>\n\n## Why LangWatch?\n\nThe platform for LLM evaluations and AI agent testing.\nWe help teams test, simulate, evaluate, and monitor LLM-powered agents end-to-end — before release and in production.\nBuilt for teams that need regression testing, simulations, and production observability without building custom tooling.\n\n- [**End-to-end agent simulations**](https:\u002F\u002Flangwatch.ai\u002Fscenario\u002F)\n  Run realistic scenarios against your **full stack** (tools, state, user simulator, judge) and pinpoint where your agents break, and why? down to each decision.\n\n- **Eval + observability + prompts in one loop**\n  [Trace](https:\u002F\u002Fdocs.langwatch.ai\u002Fintegration\u002Foverview) → [dataset](https:\u002F\u002Fdocs.langwatch.ai\u002Fdatasets\u002Foverview) → [evaluate](https:\u002F\u002Fdocs.langwatch.ai\u002Fllm-evaluation\u002Foffline-evaluation) → [optimize prompts\u002Fmodels](https:\u002F\u002Fdocs.langwatch.ai\u002Foptimization-studio\u002Foverview) → re-test. No glue code, no tool sprawl.\n\n- [**Open standards, no lock-in**](https:\u002F\u002Fdocs.langwatch.ai\u002Fintegration\u002Fopentelemetry\u002Fguide)\n  OpenTelemetry\u002FOTLP-native. Framework- and LLM-provider agnostic by design.\n\n- [**Collaboration that doesn't slow shipping**](https:\u002F\u002Fdocs.langwatch.ai\u002Ffeatures\u002Fannotations)\n  Review runs, annotate failures, and ship fixes faster. Let domain experts label edge cases with [annotations & queues](https:\u002F\u002Fdocs.langwatch.ai\u002Ffeatures\u002Fannotations), keep prompts in Git with the [GitHub integration](https:\u002F\u002Fdocs.langwatch.ai\u002Fprompt-management\u002Ffeatures\u002Fessential\u002Fgithub-integration), and [link prompt versions to traces](https:\u002F\u002Fdocs.langwatch.ai\u002Fprompt-management\u002Ffeatures\u002Fadvanced\u002Flink-to-traces).\n\nLangWatch gives you full visibility into agent behavior and the tools to systematically improve reliability, performance, and cost, while keeping you in control of your AI system\n\n## Getting Started\n\n### Cloud ☁️\n\nThe easiest way to get started with LangWatch.\n\n[Create a free account](https:\u002F\u002Fapp.langwatch.ai) → create a project → get started\u002F copy your API key.\n\n### Local setup 💻\n\nGet up and running on your own machine using docker compose:\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch.git\ncd langwatch\ncp langwatch\u002F.env.example langwatch\u002F.env\ndocker compose up -d --wait --build\n```\nOnce running, LangWatch will be available at `http:\u002F\u002Flocalhost:5560`, where you can create your first project and API key.\n\n### Deployment options ⚓️\n\nRun LangWatch on your own infrastructure:\n\n- [Docker Compose](https:\u002F\u002Fdocs.langwatch.ai\u002Fself-hosting\u002Fopen-source#docker-compose) - Run LangWatch on your own machine.\n- [Kubernetes (Helm)](https:\u002F\u002Fdocs.langwatch.ai\u002Fself-hosting\u002Fopen-source#helm-chart-for-langwatch) - Run LangWatch on a Kubernetes cluster using Helm.\n- [OnPrem](https:\u002F\u002Fdocs.langwatch.ai\u002Fself-hosting\u002Fonprem) - Cloud-specific setups for AWS, Google Cloud, and Azure.\n\n\u003Cdetails>\n\u003Csummary>Hybrid (OnPrem data) 🔀\u003C\u002Fsummary>\n\nFor companies that have strict data residency and control requirements, without needing to go fully on-prem.\n\nRead more about it on our [docs](https:\u002F\u002Fdocs.langwatch.ai\u002Fself-hosting\u002Fhybrid).\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Local Development 👩‍💻\u003C\u002Fsummary>\n\nYou can also run LangWatch locally without docker to develop and help contribute to the project.\n\nStart just the databases using docker and leave it running:\n\n```bash\ndocker compose up redis postgres opensearch\n```\n\nThen, on another terminal, install the dependencies and start LangWatch:\n\n```bash\nmake install\nmake start\n```\n\n\u003C\u002Fdetails>\n\n## 🚀 Quick Start\n\nShip safer agents in minutes. [Create a free account](https:\u002F\u002Fapp.langwatch.ai), then dive into these guides:\n\n- **[Run your first agent simulation](https:\u002F\u002Flangwatch.ai\u002Fscenario\u002Fintroduction\u002Fgetting-started)** - Test agents against realistic scenarios before production\n- **[Set up evaluations](https:\u002F\u002Fdocs.langwatch.ai\u002Fllm-evaluation\u002Foffline-evaluation)** - Measure quality, performance, and reliability\n- **[Send your first traces](https:\u002F\u002Fdocs.langwatch.ai\u002Fintegration\u002Foverview)** - Integrate LangWatch with your stack\n- **[Get started with LangWatch MCP](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fmcp)** - Use LangWatch in Claude Desktop and other MCP clients\n\n## 🗺️ Integrations\n\n\nLangWatch builds and maintains several integrations listed below. Our tracing platform is built on top of [OpenTelemetry](https:\u002F\u002Fopentelemetry.io\u002F), so we support any OpenTelemetry-compatible library out of the box.\n\n**Frameworks:**  \n[LangChain](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Flangchain) ·\n[LangGraph](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Flanggraph) ·\n[Vercel AI SDK](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Ftypescript\u002Fintegrations\u002Fvercel-ai) ·\n[Mastra](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Ftypescript\u002Fintegrations\u002Fmastra) ·\n[CrewAI](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Fcrewai) ·\n[Google ADK](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Fgoogle-ai)\n\n**Model Providers:**  \n[OpenAI](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Fopenai) ·\n[Anthropic](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Fanthropic) ·\n[Azure](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Fazure) ·\n[Google Cloud](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Fgoogle-cloud) ·\n[AWS](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Faws) ·\n[Groq](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Fgroq) ·\n[Ollama](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Follama)\n\n### Platforms\n\n[LangFlow](https:\u002F\u002Fdocs.langwatch.ai\u002Fintegration\u002Flangflow) · [Flowise](https:\u002F\u002Fdocs.langwatch.ai\u002Fintegration\u002Fflowise) · [n8n](https:\u002F\u002Fdocs.langwatch.ai\u002Fintegration\u002Fn8n)\n\n*and many more…*\n\nAre you using a platform that could benefit from a direct LangWatch integration? We'd love to hear from you, please [**fill out this very quick form.**](https:\u002F\u002Fwww.notion.so\u002F1e35e165d48280468247fcbdc3349077?pvs=21)\n\n## 💬 Support\n\nHave questions or need help? We're here to support you in multiple ways:\n\n- **Documentation:** Our comprehensive [documentation](https:\u002F\u002Fdocs.langwatch.ai) covers everything from getting started to advanced features.\n- **Discord Community:** Join our [Discord server](https:\u002F\u002Fdiscord.gg\u002FkT4PhDS2gH) for real-time help from our team and community.\n- **X (Twitter):** Follow us on [X](https:\u002F\u002Fx.com\u002FLangWatchAI) for updates and announcements.\n- **GitHub Issues:** Report bugs or request features through our [GitHub repository](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch).\n- **Enterprise Support:** Enterprise customers receive priority support with dedicated response times. Our [pricing page](https:\u002F\u002Flangwatch.ai\u002Fpricing) contains more information.\n\n## 🤝 Collaborating\n\nContributions are what make the open-source community such an amazing place to learn, inspire, and create. Any contributions you make are **greatly appreciated**.\n\nPlease read our [Contribution Guidelines](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fblob\u002Fmain\u002FCONTRIBUTING.md) for details on our code of conduct, and the process for submitting pull requests.\n\n## ✍️ License\n\nPlease read our [LICENSE.md](\u002FLICENSE.md) file.\n\n## 👮‍♀️ Security + Compliance\n\nAs a platform that has access to data that is highly likely to be sensitive, we take security incredibly seriously and treat it as a core part of our culture.\n\n| Legal Framework | Current Status                                                                 |\n| --------------- | ------------------------------------------------------------------------------ |\n| GDPR            | Compliant. DPA available upon request.                                         |\n| ISO 27001       | Certified. Certification report available upon request on our Enterprise plan. |\n\nPlease refer to our Security page for more information. Contact us at [security@langwatch.ai](mailto:security@langwatch.ai) if you have any further questions.\n\n### Vulnerability Disclosure\n\nIf you need to do a responsible disclosure of a security vulnerability, you may do so by email to [security@langwatch.ai](mailto:security@langwatch.ai), or if you prefer you can reach out to one of our team privately on [Discord](https:\u002F\u002Fdiscord.com\u002Finvite\u002FkT4PhDS2gH).\n","\u003Cimg width=\"1212\" height=\"395\" alt=\"012d1688-24ae-4759-ae70-5f8f81a13c0e\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flangwatch_langwatch_readme_93f72f04a2b9.png\" \u002F>\n\n\n\u003Ch3 align=\"center\">\n    \u003Ca href=\"https:\u002F\u002Flangwatch.ai\">官网\u003C\u002Fa> · \u003Ca href=\"https:\u002F\u002Fdocs.langwatch.ai\">文档\u003C\u002Fa> · \u003Ca href=\"https:\u002F\u002Fdiscord.gg\u002FkT4PhDS2gH\">Discord\u003C\u002Fa> · \u003Ca href=\"https:\u002F\u002Fdocs.langwatch.ai\u002Fself-hosting\u002Foverview\">自托管\u003C\u002Fa>\n\u003C\u002Fh3>\n\n\u003Cp align=\"center\">\n\u003Ca href=\"https:\u002F\u002Fdiscord.gg\u002FkT4PhDS2gH\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fdiscord\u002F1227886780536324106?logo=discord&labelColor=%20%235462eb&logoColor=%20%23f5f5f5&color=%20%235462eb\" alt=\"在Discord上聊天\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fpypi.python.org\u002Fpypi\u002Flangwatch\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Flangwatch?logo=python&logoColor=white&label=pypi%20langwatch&color=blue\" alt=\"PyPI上的langwatch Python包\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fwww.npmjs.com\u002Fpackage\u002Flangwatch\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fnpm\u002Fdm\u002Flangwatch?logo=npm&logoColor=white&label=npm%20langwatch&color=blue\" alt=\"langwatch npm包\">\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Ftwitter.com\u002Fintent\u002Ffollow?screen_name=langwatchai\" target=\"_blank\">\n   \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Ftwitter\u002Ffollow\u002Flangwatchai?logo=X&color=%20%23f5f5f5\"\n      alt=\"在X上关注\">\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003Cvideo src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fff49882d-4e9d-4b7c-819b-be690fba9387\" autoplay loop muted playsinline width=\"100%\" style=\"display: block; aspect-ratio: 16 \u002F 9;\">\u003C\u002Fvideo>\n\n## 为什么选择LangWatch？\n\nLLM评估与AI智能体测试的平台。\n我们帮助团队在发布前和生产环境中，对基于LLM的智能体进行端到端的测试、模拟、评估和监控。专为需要回归测试、仿真和生产环境可观测性，但又不想构建自定义工具的团队打造。\n\n- [**端到端智能体仿真**](https:\u002F\u002Flangwatch.ai\u002Fscenario\u002F)\n  针对你的**完整堆栈**（工具、状态、用户模拟器、评判者）运行真实场景，精准定位智能体在何处出错以及原因——精确到每一个决策步骤。\n\n- **评估 + 可观测性 + 提示词一体化闭环**\n  [追踪](https:\u002F\u002Fdocs.langwatch.ai\u002Fintegration\u002Foverview) → [数据集](https:\u002F\u002Fdocs.langwatch.ai\u002Fdatasets\u002Foverview) → [评估](https:\u002F\u002Fdocs.langwatch.ai\u002Fllm-evaluation\u002Foffline-evaluation) → [优化提示词\u002F模型](https:\u002F\u002Fdocs.langwatch.ai\u002Foptimization-studio\u002Foverview) → 再次测试。无需胶水代码，避免工具泛滥。\n\n- [**开放标准，无厂商锁定**](https:\u002F\u002Fdocs.langwatch.ai\u002Fintegration\u002Fopentelemetry\u002Fguide)\n  原生支持OpenTelemetry\u002FOTLP。设计上不依赖特定框架或LLM提供商。\n\n- [**协作加速交付，不拖慢进度**](https:\u002F\u002Fdocs.langwatch.ai\u002Ffeatures\u002Fannotations)\n  审核运行结果，标注失败点，更快修复问题。让领域专家通过[注释与队列](https:\u002F\u002Fdocs.langwatch.ai\u002Ffeatures\u002Fannotations)标记边缘场景，借助[GitHub集成](https:\u002F\u002Fdocs.langwatch.ai\u002Fprompt-management\u002Ffeatures\u002Fessential\u002Fgithub-integration)将提示词保存在Git中，并将[提示词版本与追踪记录关联](https:\u002F\u002Fdocs.langwatch.ai\u002Fprompt-management\u002Ffeatures\u002Fadvanced\u002Flink-to-traces)。\n\nLangWatch为你提供对智能体行为的全面可见性，以及系统化提升可靠性、性能和成本效益的工具，同时让你始终掌控自己的AI系统。\n\n## 开始使用\n\n### 云端 ☁️\n\n开始使用LangWatch最简单的方式。\n\n[创建免费账户](https:\u002F\u002Fapp.langwatch.ai) → 创建项目 → 开始使用\u002F复制API密钥。\n\n### 本地部署 💻\n\n使用Docker Compose在你自己的机器上快速启动：\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch.git\ncd langwatch\ncp langwatch\u002F.env.example langwatch\u002F.env\ndocker compose up -d --wait --build\n```\n运行成功后，LangWatch将在`http:\u002F\u002Flocalhost:5560`可用，你可以在此创建第一个项目和API密钥。\n\n### 部署选项 ⚓️\n\n在你自己的基础设施上运行LangWatch：\n\n- [Docker Compose](https:\u002F\u002Fdocs.langwatch.ai\u002Fself-hosting\u002Fopen-source#docker-compose) - 在本地机器上运行LangWatch。\n- [Kubernetes (Helm)](https:\u002F\u002Fdocs.langwatch.ai\u002Fself-hosting\u002Fopen-source#helm-chart-for-langwatch) - 使用Helm在Kubernetes集群上运行LangWatch。\n- [OnPrem](https:\u002F\u002Fdocs.langwatch.ai\u002Fself-hosting\u002Fonprem) - 针对AWS、Google Cloud和Azure等云服务的专用部署方案。\n\n\u003Cdetails>\n\u003Csummary>混合模式（OnPrem数据） 🔀\u003C\u002Fsummary>\n\n适用于对数据驻留和控制有严格要求，但又不需要完全私有化部署的企业。\n\n更多详情请参阅我们的[文档](https:\u002F\u002Fdocs.langwatch.ai\u002Fself-hosting\u002Fhybrid)。\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>本地开发 👩‍💻\u003C\u002Fsummary>\n\n你也可以不使用Docker，在本地运行LangWatch进行开发并参与项目贡献。\n\n只需使用Docker启动数据库并保持运行：\n\n```bash\ndocker compose up redis postgres opensearch\n```\n\n然后在另一个终端安装依赖并启动LangWatch：\n\n```bash\nmake install\nmake start\n```\n\n\u003C\u002Fdetails>\n\n## 🚀 快速入门\n\n几分钟内即可部署更安全的智能体。[创建免费账户](https:\u002F\u002Fapp.langwatch.ai)，然后参考以下指南：\n\n- **[运行首次智能体仿真](https:\u002F\u002Flangwatch.ai\u002Fscenario\u002Fintroduction\u002Fgetting-started)** - 在生产前用真实场景测试智能体\n- **[设置评估](https:\u002F\u002Fdocs.langwatch.ai\u002Fllm-evaluation\u002Foffline-evaluation)** - 衡量质量、性能和可靠性\n- **[发送首次追踪](https:\u002F\u002Fdocs.langwatch.ai\u002Fintegration\u002Foverview)** - 将LangWatch集成到你的技术栈中\n- **[开始使用LangWatch MCP](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fmcp)** - 在Claude Desktop及其他MCP客户端中使用LangWatch\n\n## 🗺️ 集成\n\n\nLangWatch构建并维护了以下多种集成。我们的追踪平台基于[OpenTelemetry](https:\u002F\u002Fopentelemetry.io\u002F)，因此原生支持任何兼容OpenTelemetry的库。\n\n**框架：**\n[LangChain](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Flangchain) ·\n[LangGraph](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Flanggraph) ·\n[Vercel AI SDK](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Ftypescript\u002Fintegrations\u002Fvercel-ai) ·\n[Mastra](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Ftypescript\u002Fintegrations\u002Fmastra) ·\n[CrewAI](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Fcrewai) ·\n[Google ADK](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Fgoogle-ai)\n\n**模型提供商：**\n[OpenAI](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Fopenai) ·\n[Anthropic](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Fanthropic) ·\n[Azure](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Fazure) ·\n[Google Cloud](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Fgoogle-cloud) ·\n[AWS](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Faws) ·\n[Groq](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Fgroq) ·\n[Ollama](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fpython\u002Fintegrations\u002Follama)\n\n### 平台\n\n[LangFlow](https:\u002F\u002Fdocs.langwatch.ai\u002Fintegration\u002Flangflow) · [Flowise](https:\u002F\u002Fdocs.langwatch.ai\u002Fintegration\u002Fflowise) · [n8n](https:\u002F\u002Fdocs.langwatch.ai\u002Fintegration\u002Fn8n)\n\n*还有更多…*\n\n您是否正在使用某个平台，并认为直接集成 LangWatch 会带来益处？我们非常期待您的反馈，请[**填写这份简短的表格。**](https:\u002F\u002Fwww.notion.so\u002F1e35e165d48280468247fcbdc3349077?pvs=21)\n\n## 💬 支持\n\n您有疑问或需要帮助吗？我们提供多种方式来支持您：\n\n- **文档：** 我们的全面[文档](https:\u002F\u002Fdocs.langwatch.ai)涵盖了从入门到高级功能的方方面面。\n- **Discord 社区：** 加入我们的[Discord 服务器](https:\u002F\u002Fdiscord.gg\u002FkT4PhDS2gH)，即可获得来自团队和社区的实时帮助。\n- **X (Twitter)：** 关注我们在[X](https:\u002F\u002Fx.com\u002FLangWatchAI)上的账号，获取最新动态和公告。\n- **GitHub Issues：** 您可以通过我们的[Github 仓库](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch)提交 bug 报告或功能请求。\n- **企业级支持：** 企业客户可享受优先支持及专属响应时间。更多信息请参阅我们的[定价页面](https:\u002F\u002Flangwatch.ai\u002Fpricing)。\n\n## 🤝 合作\n\n正是贡献让开源社区成为一个充满学习、启发与创造的绝佳场所。您所做的任何贡献都将受到**高度赞赏**。\n\n请阅读我们的[贡献指南](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fblob\u002Fmain\u002FCONTRIBUTING.md)，了解我们的行为准则以及提交 Pull Request 的流程。\n\n## ✍️ 许可证\n\n请阅读我们的[LICENSE.md](\u002FLICENSE.md)文件。\n\n## 👮‍♀️ 安全与合规\n\n作为一家能够访问高度敏感数据的平台，我们对安全问题极为重视，并将其视为企业文化的核心组成部分。\n\n| 法律框架       | 当前状态                                                                 |\n| --------------- | -------------------------------------------------------------------------- |\n| GDPR            | 符合要求。可根据需求提供 DPA。                                             |\n| ISO 27001       | 已通过认证。企业版用户可根据需求索取认证报告。                             |\n\n更多信息请参阅我们的安全页面。如有任何进一步的问题，请联系 [security@langwatch.ai](mailto:security@langwatch.ai)。\n\n### 漏洞披露\n\n如果您需要负责任地披露安全漏洞，可通过电子邮件发送至 [security@langwatch.ai](mailto:security@langwatch.ai)，或者您也可以选择在[Discord](https:\u002F\u002Fdiscord.com\u002Finvite\u002FkT4PhDS2gH)上私下联系我们的团队成员。","# LangWatch 快速上手指南\n\nLangWatch 是一个专为 LLM（大语言模型）评估和 AI Agent 测试设计的开源平台。它帮助团队在发布前和生产环境中对 AI Agent 进行端到端的测试、模拟、评估和监控，无需构建自定义工具即可实现回归测试和生产可观测性。\n\n## 环境准备\n\n在开始之前，请确保您的开发环境满足以下要求：\n\n*   **操作系统**：Linux, macOS 或 Windows (需安装 WSL2)\n*   **核心依赖**：\n    *   [Docker](https:\u002F\u002Fwww.docker.com\u002F) 及 Docker Compose (推荐 v2.0+)\n    *   Git\n*   **网络环境**：由于项目源位于 GitHub，国内用户建议配置代理或使用镜像加速以确保 `git clone` 和 `docker pull` 顺利进行。\n*   **可选依赖**（用于本地开发贡献）：\n    *   Python 3.10+\n    *   Node.js 18+\n    *   Make\n\n## 安装步骤\n\n您可以选择使用云端服务快速体验，或在本地通过 Docker 私有化部署。\n\n### 方式一：云端快速开始（推荐新手）\n\n无需安装任何环境，直接注册即可使用：\n\n1.  访问 [LangWatch Cloud](https:\u002F\u002Fapp.langwatch.ai) 创建免费账户。\n2.  创建一个新项目 (Project)。\n3.  复制您的 **API Key** 以便后续集成。\n\n### 方式二：本地私有化部署 (Docker Compose)\n\n适合需要在内网运行或数据隐私要求较高的场景。\n\n1.  **克隆项目仓库**：\n    ```bash\n    git clone https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch.git\n    cd langwatch\n    ```\n\n2.  **配置环境变量**：\n    复制示例配置文件为正式配置文件：\n    ```bash\n    cp langwatch\u002F.env.example langwatch\u002F.env\n    ```\n    *(注：如有需要，可编辑 `.env` 文件修改端口或密钥配置)*\n\n3.  **启动服务**：\n    使用 Docker Compose 构建并启动所有服务：\n    ```bash\n    docker compose up -d --wait --build\n    ```\n\n4.  **访问平台**：\n    服务启动后，在浏览器打开 `http:\u002F\u002Flocalhost:5560`。\n    首次登录时，请在界面中创建第一个项目和 API Key。\n\n> **提示**：如果您希望参与源码开发而非仅运行服务，请参考原文 \"Local Development\" 部分，仅启动数据库容器 (`redis`, `postgres`, `opensearch`)，然后在本地终端运行 `make install` 和 `make start`。\n\n## 基本使用\n\n安装完成后，您可以通过以下三个核心步骤将 LangWatch 集成到您的 AI 工作流中：\n\n### 1. 发送追踪数据 (Tracing)\n\nLangWatch 基于 **OpenTelemetry** 标准，支持主流框架（LangChain, LangGraph, Vercel AI SDK, CrewAI 等）和模型提供商（OpenAI, Anthropic, Azure 等）。\n\n以 Python + LangChain 为例，安装 SDK 并初始化：\n\n```bash\npip install langwatch\n```\n\n在代码中配置追踪：\n\n```python\nimport langwatch\nfrom langchain_openai import ChatOpenAI\nfrom langchain_core.prompts import ChatPromptTemplate\n\n# 初始化 LangWatch\nlangwatch.init(\n    api_key=\"YOUR_API_KEY\", \n    project_id=\"YOUR_PROJECT_ID\"\n)\n\n# 您的常规 LLM 调用代码\nllm = ChatOpenAI(model=\"gpt-4o\")\nprompt = ChatPromptTemplate.from_messages([(\"human\", \"Hello, world!\")])\nchain = prompt | llm\n\n# 执行调用，LangWatch 会自动捕获 Trace\nresponse = chain.invoke({})\nprint(response)\n```\n\n*更多框架集成指南请参阅官方文档中的 [Integrations](https:\u002F\u002Fdocs.langwatch.ai\u002Fintegration\u002Foverview) 部分。*\n\n### 2. 运行 Agent 模拟测试 (Simulations)\n\n在将 Agent 发布到生产环境前，使用 LangWatch 的场景模拟功能进行测试：\n\n1.  登录 Web 界面，进入 **Scenarios** 模块。\n2.  定义测试场景（包括用户模拟器、评判器 Judge、工具集等）。\n3.  运行模拟，系统将自动分析 Agent 在复杂交互中的决策路径， pinpoint 失败原因。\n\n### 3. 设置评估与优化 (Evaluation & Optimization)\n\n建立“追踪 -> 数据集 -> 评估 -> 优化”的闭环：\n\n1.  **创建数据集**：从已收集的 Trace 中筛选典型用例生成数据集。\n2.  **配置评估指标**：设置自动化评估标准（如准确性、延迟、成本、幻觉检测等）。\n3.  **迭代优化**：根据评估结果调整 Prompt 或模型参数，并在 Git 中管理 Prompt 版本（支持 GitHub 集成），链接 Prompt 版本与具体 Trace 进行对比分析。\n\n---\n\n**下一步**：\n*   查看 [完整文档](https:\u002F\u002Fdocs.langwatch.ai) 了解高级功能。\n*   加入 [Discord 社区](https:\u002F\u002Fdiscord.gg\u002FkT4PhDS2gH) 获取实时支持。\n*   尝试 [LangWatch MCP](https:\u002F\u002Flangwatch.ai\u002Fdocs\u002Fintegration\u002Fmcp) 在 Claude Desktop 中直接使用。","某电商团队正在开发一个能处理退货、换货及投诉的智能客服 Agent，需要在上线前确保其决策逻辑准确且成本可控。\n\n### 没有 langwatch 时\n- **测试覆盖不全**：只能手动构造少量对话测试，难以模拟用户复杂的连环追问或边缘场景，导致上线后频繁出现“死循环”或错误调用工具。\n- **故障定位困难**：当 Agent 回答错误时，开发者无法直观看到是提示词（Prompt）问题、模型选择失误还是工具参数传递错误，排查如同“盲人摸象”。\n- **优化缺乏依据**：调整 Prompt 后无法系统性地对比新旧版本效果，往往凭感觉修改，导致性能提升不明显甚至引发回归倒退。\n- **协作效率低下**：产品经理发现坏案例后，只能通过截图反馈给开发，缺乏统一的标注和队列机制，修复进度难以追踪。\n\n### 使用 langwatch 后\n- **全链路仿真测试**：利用端到端场景模拟功能，自动运行数百个包含复杂状态和多轮交互的测试用例，精准定位 Agent 在特定决策节点的崩溃原因。\n- **深度可观测性**：通过完整的 Trace 追踪，一键查看从用户输入、Prompt 渲染、模型输出到工具调用的全过程，迅速锁定故障根源。\n- **闭环迭代优化**：建立“追踪 - 数据集 - 评估 - 优化”的自动化闭环，量化对比不同 Prompt 版本的准确率与成本，确保每次修改都有数据支撑。\n- **高效协同评审**：产品专家可直接在平台上对失败案例进行标注并加入修复队列，开发人员在 GitHub 中关联 Prompt 版本与具体轨迹，大幅缩短修复周期。\n\nlangwatch 将原本黑盒且分散的 AI 开发流程转变为透明、可量化且高效协同的工程化闭环，显著提升了智能代理的可靠性与交付速度。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flangwatch_langwatch_93f72f04.png","LangWatch","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Flangwatch_187b6b32.png","",null,"https:\u002F\u002Fgithub.com\u002Flangwatch",[78,82,86,90,94,98,102,105,109,111],{"name":79,"color":80,"percentage":81},"TypeScript","#3178c6",72.7,{"name":83,"color":84,"percentage":85},"Jupyter Notebook","#DA5B0B",15.2,{"name":87,"color":88,"percentage":89},"Python","#3572A5",5.5,{"name":91,"color":92,"percentage":93},"Gherkin","#5B2063",4.8,{"name":95,"color":96,"percentage":97},"Go","#00ADD8",0.9,{"name":99,"color":100,"percentage":101},"JavaScript","#f1e05a",0.3,{"name":103,"color":104,"percentage":101},"Shell","#89e051",{"name":106,"color":107,"percentage":108},"MDX","#fcb32c",0.1,{"name":110,"color":96,"percentage":108},"Go Template",{"name":112,"color":113,"percentage":108},"Jinja","#a52a22",3185,307,"2026-04-07T11:02:16","NOASSERTION",4,"Linux, macOS, Windows","未说明",{"notes":122,"python":120,"dependencies":123},"该工具主要基于 Docker Compose 部署，核心依赖为容器化环境而非特定的 Python 包或 GPU。本地开发模式下需手动安装依赖（通过 make install）。支持多种 LLM 框架（如 LangChain, CrewAI）和模型提供商的集成，底层追踪基于 OpenTelemetry 标准。",[124,125,126,127,128],"Docker","Docker Compose","Redis","PostgreSQL","OpenSearch",[15,13,16,35,130,14],"其他",[132,133,134,135,136,137,138,139,140,141,142,143,144],"ai","analytics","datasets","evaluation","gpt","llm","observability","openai","prompt-engineering","dspy","llmops","low-code","llm-ops","2026-03-27T02:49:30.150509","2026-04-08T01:01:40.906700",[],[149,154,159,164,169,174,179,184,189,194,199,204,209,214,219,223,228,232,237,241],{"id":150,"version":151,"summary_zh":152,"released_at":153},137023,"clickhouse-serverless-0.1.0-next.1","ClickHouse，具备基于CPU\u002F内存的自动调优、分层存储，并可通过内置的Keeper实现可选的复制功能。","2026-04-07T08:36:51",{"id":155,"version":156,"summary_zh":157,"released_at":158},137024,"typescript-sdk@v0.20.0","## [0.20.0](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcompare\u002Ftypescript-sdk@v0.19.0...typescript-sdk@v0.20.0)（2026-03-29）\n\n\n### 功能特性\n\n* **技能：** 基于功能映射和场景测试的代理技能上线流程（[#2377](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2377)）（[6f6abdb](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F6f6abdb67b97fcb7c2958dbc193dcde37d4c82a8)）\n* **追踪：** 显式指定应用来源，以防止评估竞争条件（[#2346](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2346)）（[348874e](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F348874e48ce79537596d28887c6ed9c719200b29)）\n\n\n### Bug 修复\n\n* **SDK+技能：** 扩大 peer dependencies，修复场景配置，并强化技能测试（[#2485](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2485)）（[c77237c](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fc77237c8f5690f4eafcc96ed3334e3c6d61a7249)）\n* **SDK：** 在 fetch-policy 分支中抛出 PromptsError 而不是 Error（[#2691](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2691)）（[9388ab6](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F9388ab63a77f32f8888b204f462f8ec3cc88abcc)），关闭了 [#976](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F976)\n* **TypeScript SDK：** 处理 Zod 优先的 evaluations.ts 文件以及 filter types 的导入（[#2352](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2352)）（[516d4e8](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F516d4e839388ae954b42ba5f742ef6b96c0beacf)）","2026-03-29T17:05:44",{"id":160,"version":161,"summary_zh":162,"released_at":163},137025,"skills@v0.2.0","## [0.2.0](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcompare\u002Fskills@v0.1.0...skills@v0.2.0) (2026-03-29)\n\n\n### 功能特性\n\n* 在 langwatch\u002Fskills 仓库中添加技能发布流水线 ([#2415](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2415)) ([9b769fe](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F9b769fe8b6e11a18434e967564697a26c48f8aa0))\n* 通过 Git 钩子和 CI 自动重新生成文档提示词及 llms.txt 文件 ([#2686](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2686)) ([9f78ffa](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F9f78ffae2257c8a233b318957f06aa264f9ccac1))\n* **技能：** 基于功能图和场景测试的代理技能引导式接入 ([#2377](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2377)) ([6f6abdb](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F6f6abdb67b97fcb7c2958dbc193dcde37d4c82a8))\n\n\n### 错误修复\n\n* 在技能发布同步中包含 version.txt 文件 ([#2417](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2417)) ([1cd6606](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F1cd66066a90984530059eef502d29359c14bf0ea))\n* **SDK+技能：** 扩大对等依赖范围，修复场景配置，强化技能测试 ([#2485](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2485)) ([c77237c](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fc77237c8f5690f4eafcc96ed3334e3c6d61a7249))","2026-03-29T17:05:19",{"id":165,"version":166,"summary_zh":167,"released_at":168},137026,"python-sdk@v0.17.0","## [0.17.0](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcompare\u002Fpython-sdk@v0.16.0...python-sdk@v0.17.0) (2026-03-18)\n\n\n### 功能特性\n\n* **追踪：** 显式指定应用来源，以防止评估竞争条件（[#2346](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2346)）（[348874e](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F348874e48ce79537596d28887c6ed9c719200b29)）\n\n\n### 错误修复\n\n* **Python SDK：** 将 langchain-core 设为可选依赖项（[#2464](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2464)）（[cde8406](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fcde8406990f5e19af98048a43a0d7d9782553a48)）\n* 在遇到 Cerebras 的 RateLimitError 时跳过 litellm_bot 测试（[#2355](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2355)）（[76bf87f](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F76bf87f1db51b0fab681c593cd5515c66d0908f1)）","2026-03-18T17:29:35",{"id":170,"version":171,"summary_zh":172,"released_at":173},137027,"mcp-server@v0.6.1","## [0.6.1](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcompare\u002Fmcp-server@v0.6.0...mcp-server@v0.6.1) (2026-03-13)\n\n\n### Bug修复\n\n* 修复文档中场景相对路径的解析问题 ([315abfd](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F315abfd4c9d21641f0c1845fcf29b6337673ce6d))","2026-03-13T15:27:34",{"id":175,"version":176,"summary_zh":177,"released_at":178},137028,"mcp-server@v0.6.0","## [0.6.0](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcompare\u002Fmcp-server@v0.5.0...mcp-server@v0.6.0) (2026-03-13)\n\n\n### 功能特性\n\n* 为评估者和模型提供商添加平台 MCP 工具 ([#1762](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1762)) ([6339001](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F6339001ffae3c83f7c3ea0cdadf480f4ec47f3cc))\n* 在平台 MCP 工具前加上 `platform_` 前缀，并添加消除歧义的说明 ([#1760](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1760)) ([91851ee](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F91851ee28aff1b3b75d92ce46aec6ae5072ff625))\n\n\n### 错误修复\n\n* **依赖:** 为 Dependabot 高严重性告警固定传递性 npm 依赖项 ([#2220](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2220)) ([a179da1](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fa179da1d55859b8c9a476a6443dbfaf6bf632af3))\n* **MCP:** 将提示工具模式与后端 API 对齐 ([#2326](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2326)) ([65279f3](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F65279f38687bae5c7dc7a13db33dcb2f7d135397))\n* **MCP:** 解决 get_trace 404 以及 search_traces 元数据过滤器的 bug ([#2214](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2214)) ([a0362e6](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fa0362e60aa54e7c6e5dac9f241b206f4f4cd0b5f))\n* 为 litellm 路由规范化 defaultModel，使其带有 provider 前缀 ([cc7798e](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fcc7798ecb229654f371912e7863ca20364ceb7fa))\n\n\n### 其他\n\n* **依赖:** 批量安全地升级 JS 依赖（一级 a） ([#2169](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2169)) ([8b5b891](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F8b5b8915941f19ca07019b330e276f01ae3bface))\n* **依赖:** 将 mcp-server 的 zod 升级至 v4 ([#1977](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1977)) ([eb89188](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Feb891883eb47db01bfee4d184b1c9b19e3cc60d7))\n* **依赖:** 批次升级 Node.js 依赖项第 2 批次 ([#1947](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1947)) ([98d9f5c](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F98d9f5cb53dc13051c500359def11b35762fd5e5))\n* **依赖:** 批量升级安全的 npm 依赖项（共 19 个 Dependabot PR） ([#1931](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1931)) ([4d7607e](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F4d7607e5d467749e2e47fc3b0b2b58c212bc8721))\n* **依赖:** 在一个目录中批量升级 npm_and_yarn 组的 5 个依赖项 ([#1948](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1948)) ([9227fb4](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F9227fb4b96184a82c87ab95f00fb985491df7224))","2026-03-13T13:17:38",{"id":180,"version":181,"summary_zh":182,"released_at":183},137029,"python-sdk@v0.16.0","## [0.16.0](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcompare\u002Fpython-sdk@v0.15.0...python-sdk@v0.16.0) (2026-03-13)\n\n\n### 功能特性\n\n* **追踪：** 显式指定应用来源，以防止评估竞争条件 ([#2325](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2325)) ([b11b499](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fb11b49990bdb1ccf1bf302b9a54b442d3b6f3be3))\n\n\n### 错误修复\n\n* 自定义 SDK 评估 - 灰色徽章、卡在“处理中”状态、前置条件崩溃 ([#2234](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2234)) ([77cba68](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F77cba6844de4e58a14bac4348254ef93213d4079))\n* **MCP：** 将提示工具模式与后端 API 对齐 ([#2326](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2326)) ([65279f3](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F65279f38687bae5c7dc7a13db33dcb2f7d135397))\n* **安全：** 升级 Dependabot 检测到的严重漏洞（unstructured + nltk）([#2275](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2275)) ([cc1f004](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fcc1f004d2ec710c9eaaa762e336516d447338090))\n* **安全：** 升级 python-sdk 中的中低危漏洞 ([#2299](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2299)) ([da288b2](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fda288b2559e01e0af6046c19eec39eb3d4678d77))","2026-03-13T13:03:13",{"id":185,"version":186,"summary_zh":187,"released_at":188},137030,"typescript-sdk@v0.19.0","## [0.19.0](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcompare\u002Ftypescript-sdk@v0.18.0...typescript-sdk@v0.19.0) (2026-03-13)\n\n\n### 功能\n\n* **追踪：** 显式指定应用来源，以防止评估竞争条件（[#2325](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2325)）（[b11b499](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fb11b49990bdb1ccf1bf302b9a54b442d3b6f3be3)）","2026-03-13T13:02:25",{"id":190,"version":191,"summary_zh":192,"released_at":193},137031,"python-sdk@v0.15.0","## [0.15.0](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcompare\u002Fpython-sdk@v0.14.0...python-sdk@v0.15.0) (2026-03-11)\n\n\n### 功能\n\n* **提示词：** 当存在提示词句柄时，从追踪中打开现有提示词（[#2223](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2223)）（[428f8ff](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F428f8ff22b8cb5a4be5942d5b26bf2d62f44d7be)）","2026-03-11T09:26:27",{"id":195,"version":196,"summary_zh":197,"released_at":198},137032,"typescript-sdk@v0.18.0","## [0.18.0](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcompare\u002Ftypescript-sdk@v0.17.0...typescript-sdk@v0.18.0)（2026-03-11）\n\n\n### 功能特性\n\n* **提示词：** 当存在提示词句柄时，从追踪中打开现有提示词（[#2223](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2223)）（[428f8ff](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F428f8ff22b8cb5a4be5942d5b26bf2d62f44d7be)）\n\n\n### 错误修复\n\n* **依赖项：** 为 Dependabot 高严重性告警固定传递性 npm 依赖项（[#2220](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2220)）（[a179da1](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fa179da1d55859b8c9a476a6443dbfaf6bf632af3)）","2026-03-11T09:25:35",{"id":200,"version":201,"summary_zh":202,"released_at":203},137033,"typescript-sdk@v0.17.0","## [0.17.0](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcompare\u002Ftypescript-sdk@v0.16.1...typescript-sdk@v0.17.0) (2026-03-08)\n\n\n### Features\n\n* **examples:** add metadata and labels examples for TypeScript and Python SDKs ([#1585](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1585)) ([7d09ab8](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F7d09ab805146542921e8b1f1258d5e6e59462bfe))\n* **traces:** add langwatch.origin attribute for trace origin classification ([#2066](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2066)) ([1d94865](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F1d9486548e994d8d2f933a7f530196df8d255e43))\n* updated mastra example in typescript sdk to use 1.0 api ([#1872](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1872)) ([f738633](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Ff738633951c6f2d0db3372f9be913e9f0c7d5d51))\n\n\n### Bug Fixes\n\n* get SDK e2e CI tests passing ([#1648](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1648)) ([2a4583e](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F2a4583e4fb7ccb4674e64f0154f62fd990a2d8aa))\n* typescript sdk uses correct attributes now sdk metadata ([#1651](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1651)) ([c499e26](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fc499e26924a87cffe6abb37be8bdc739094a4568))\n* **typescript-sdk:** improve auto-shutdown signals and default to batch processor ([#1851](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1851)) ([fcf6bb7](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Ffcf6bb751eaa895c874d3ed96b455966fd8d462a))\n* **typescript-sdk:** move @opentelemetry\u002Fapi to peerDependencies ([#2072](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2072)) ([c93fc98](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fc93fc98ba06d1a90ee615e367e327f88554ced68))\n\n\n### Miscellaneous\n\n* **deps:** bump Node.js dependencies - batch 3 ([#1957](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1957)) ([e500557](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fe500557ac7514ffc52ce26738d74e292e5428d2d)), closes [#1518](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1518)\n* **deps:** bump safe npm dependencies (19 dependabot PRs) ([#1931](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1931)) ([4d7607e](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F4d7607e5d467749e2e47fc3b0b2b58c212bc8721))\n* **deps:** bump the npm_and_yarn group across 1 directory with 9 updates ([#1965](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1965)) ([3674bcc](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F3674bccb17d52d6436dc75518aeaefe75942a318))\n\n\n### Code Refactoring\n\n* derive response_format from outputs, eliminate duality ([#1647](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1647)) ([856e4e7](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F856e4e7350b26f25a44da919e4e7e7ffb5a0a0bd))\n* derive response_format from outputs, eliminate stored duality ([856e4e7](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F856e4e7350b26f25a44da919e4e7e7ffb5a0a0bd))","2026-03-08T12:21:50",{"id":205,"version":206,"summary_zh":207,"released_at":208},137034,"python-sdk@v0.14.0","## [0.14.0](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcompare\u002Fpython-sdk@v0.13.0...python-sdk@v0.14.0) (2026-03-08)\n\n\n### Features\n\n* **examples:** add metadata and labels examples for TypeScript and Python SDKs ([#1585](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1585)) ([7d09ab8](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F7d09ab805146542921e8b1f1258d5e6e59462bfe))\n* **traces:** add langwatch.origin attribute for trace origin classification ([#2066](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F2066)) ([1d94865](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F1d9486548e994d8d2f933a7f530196df8d255e43))\n\n\n### Miscellaneous\n\n* **deps:** bump Python (uv) dependencies across langwatch_nlp, python-sdk, langevals ([#1940](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1940)) ([d103f89](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fd103f89fc1936dc8dae73f2b90885a28b97e2775))\n* **deps:** bump Python uv lockfile packages (v2) ([#1954](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1954)) ([23b3a5c](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F23b3a5ce9f16d6a722690cbdd86cb095f01dba08))","2026-03-08T12:20:42",{"id":210,"version":211,"summary_zh":212,"released_at":213},137035,"langwatch-helm-2.6.0","Helm chart for LangWatch application","2026-03-02T07:00:15",{"id":215,"version":216,"summary_zh":217,"released_at":218},137036,"langwatch@v2.6.0","## [2.6.0](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcompare\u002Flangwatch@v2.5.0...langwatch@v2.6.0) (2026-03-01)\n\n\n### Features\n\n* add new event sourcing observability tooling, called skynet ([#1865](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1865)) ([b759012](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fb75901283d238dbf32df4f80e3eda4f95c561232))\n* **billing:** add subscription override support for all plan limit fields ([3ac5ba7](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F3ac5ba768a790761f32850a113ddfcac0970374d))\n* **billing:** add subscription override support for all plan limit fields ([#1862](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1862)) ([3ac5ba7](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F3ac5ba768a790761f32850a113ddfcac0970374d))\n* encrypt model provider API keys at rest ([#1764](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1764)) ([7f194b2](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F7f194b27579c0c26ffe351c778ab89d47d1cb545))\n* evaluation fixes  ([#1839](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1839)) ([a31f460](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fa31f460490b03bf66857ab21c88eaa135cb2ad06))\n* **licensing:** add usageUnit to license schema and generation ([#1860](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1860)) ([2fc56ec](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F2fc56ecad0512f92e87179f1b5a79cbc74d9e9b6))\n* otel trace context propagation for HTTP scenario targets ([#1840](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1840)) ([5b0507d](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F5b0507dddf88c1d600416e4c5866549a6f5dfcca))\n* **subscription:** implement new billing model ([#1742](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1742)) ([bda81e8](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fbda81e836568327a210d8cf5bcb6e2469c59e6bb))\n* **suites:** add time filter for suite runs ([#1827](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1827)) ([8b8cc84](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F8b8cc84cef6375a359efe35bacabe4232cc94fce))\n\n\n### Bug Fixes\n\n* actually use, not just enable, async inserts in clickhouse inserts ([#1876](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1876)) ([745a1a8](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F745a1a83c71683c855acbbc5616bb5d3750c8b02))\n* **billing:** make FREE plan limits pricing-model-aware for SEAT_EVENT orgs ([c7b2f74](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fc7b2f74d81dc77e5c555e12580c522e26bb8f815))\n* **billing:** make plan limits as pricing-model-aware for orgs ([#1850](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1850)) ([c7b2f74](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fc7b2f74d81dc77e5c555e12580c522e26bb8f815))\n* change default org pricing model from TIERED to SEAT_EVENT ([209c21f](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F209c21fbaba0c5abdc6bcd3b879716e3358bdcf9))\n* change default org pricing model to SEAT_EVENT ([#1832](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1832)) ([209c21f](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F209c21fbaba0c5abdc6bcd3b879716e3358bdcf9))\n* **contact:** update contact link from mail to HubSpot link ([#1835](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1835)) ([c919520](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fc9195206f3d84bfa99f1a2dae0e3db2239083170))\n* **contact:** update contact linkt from mail to hubspot link ([c919520](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fc9195206f3d84bfa99f1a2dae0e3db2239083170))\n* **event-sourcing:** increase global concurrency ([#1877](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1877)) ([5f76fb3](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F5f76fb37c9d80d94ba877bcf8b26737b93b8ad67))\n* **event-sourcing:** race condition in redis lua group queues handling ([#1873](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1873)) ([82a6e26](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F82a6e26b3a160e82b4f784cc17c2d1088be787c6))\n* missing started at in simulation projection ([#1874](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1874)) ([da8e7c0](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fda8e7c0a56452b97251143de54a3682c3f9dc40c))\n* **subscription:** correct legacy paid plan detection logic ([de73e28](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fde73e28de1f856514a0b894a695b70d752d99d22))\n* **subscription:** correct legacy paid plan detection logic ([#1842](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1842)) ([de73e28](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fde73e28de1f856514a0b894a695b70d752d99d22))\n* **suites:** use query param routing with slugs for direct suite access ([#1826](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1826)) ([2613a77](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F2613a77b30b7e63b87ebc28d406bc5efa168bb5a))\n\n\n### Miscellaneous\n\n* add length limit to PII detection, matching the limit of the library itself ([#1864](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1864)) ([2f14579](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F2f14579d13fe554d641a4e7d1e6269ca5e05669f))\n* **billing:** update stripe catalog with new gr","2026-03-02T06:42:38",{"id":220,"version":221,"summary_zh":212,"released_at":222},137037,"langwatch-helm-2.5.0","2026-02-25T10:31:10",{"id":224,"version":225,"summary_zh":226,"released_at":227},137038,"langwatch@v2.5.0","## [2.5.0](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcompare\u002Flangwatch@v2.4.2...langwatch@v2.5.0) (2026-02-25)\n\n\n### Features\n\n* add \u002Fwatch-ci and \u002Flaunch skills ([#1784](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1784)) ([ba1e76a](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fba1e76a37e1d7b5d999804925c90df5724f177c7))\n* add group-by selector for suite run history ([#1743](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1743)) ([ba7fee5](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fba7fee5f1753708e8e144fe3eb8884bfef6896f9))\n* add platform MCP tools for evaluators and model providers ([#1762](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1762)) ([6339001](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F6339001ffae3c83f7c3ea0cdadf480f4ec47f3cc))\n* **billing:** add billing foundation for seat+event pricing model ([#1800](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1800)) ([48f2e54](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F48f2e54c8931fa641218f0769362b1068cc94f4f))\n* **billing:** add usage reporting worker and billing meter dispatch ([#1733](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1733)) ([d8e5fa7](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fd8e5fa7ea35c4cea490cf354eb3972025b8ef41e))\n* collapsible suite sidebar ([#1817](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1817)) ([6a28431](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F6a284315c504e9fe48c22664fb099c6230309a16))\n* convert scenario BatchActionBar to floating bottom bar ([#1766](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1766)) ([5063299](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F50632992a0ed10664ed20b77dad8370b720c394a))\n* convert scenario BatchActionBar to floating bottom bar ([#1787](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1787)) ([5063299](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F50632992a0ed10664ed20b77dad8370b720c394a))\n* default to \"All Runs\" view when opening Suites page ([c2dd1dd](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fc2dd1dd1cc57b2f9b6640594884bc5c2b454a72f)), closes [#1771](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1771)\n* default to All Runs view when opening Suites page ([#1813](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1813)) ([c2dd1dd](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fc2dd1dd1cc57b2f9b6640594884bc5c2b454a72f))\n* defer scenario persistence until explicit Save ([#1783](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1783)) ([b21337d](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fb21337da242cc05760d629e13113f013b86e6216)), closes [#1769](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1769)\n* display scenario names in All Runs row headers ([#1816](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1816)) ([618dfda](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F618dfda09dc093650033e38f3541332d2ff42842))\n* remove label generation from AI scenario creation ([#1770](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1770)) ([112cee9](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F112cee9833062520fc230825bafda9384b9d6bd1))\n* remove label generation from AI scenario creation ([#1788](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1788)) ([112cee9](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F112cee9833062520fc230825bafda9384b9d6bd1))\n* rework event sourcing, move product to dual write, improve scenarios and platform performance ([#1704](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1704)) ([b0048b8](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fb0048b8dfb39f0343a87966dbebf951d978fe52d))\n* **skills:** check for existing feature files before creating new ones ([#1791](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1791)) ([e41a7ac](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fe41a7acaf852beb713d9e5dc50c366c229d70834))\n* skip hostname and TLS validation in scenario runner for on-prem ([#1818](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1818)) ([93b2a7a](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F93b2a7aa34ea1002baebb12eb2f16bd61a24a21e))\n* skip hostname\u002FTLS validation in scenario runner for on-prem ([#1821](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1821)) ([93b2a7a](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F93b2a7aa34ea1002baebb12eb2f16bd61a24a21e))\n* standardize search fields with shared SearchInput component ([#1672](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1672)) ([72e82e2](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F72e82e2657782f1f5fa2485a8e07369d52ab1ce5))\n* standardize search fields with shared SearchInput component ([#1815](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1815)) ([72e82e2](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F72e82e2657782f1f5fa2485a8e07369d52ab1ce5))\n* suite sidebar shows pass count, recency, and three-dot menu ([#1776](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1776)) ([a3d3d96](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002Fa3d3d9605f702353c1bafe388509ed9b25e99c83))\n* **suites:** archive instead of hard delete ([#1808](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1808)) ([8314ad9](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F8314ad9c1eb095ac49a4c0d740a","2026-02-25T10:12:27",{"id":229,"version":230,"summary_zh":212,"released_at":231},137039,"langwatch-helm-2.4.2","2026-02-20T19:40:15",{"id":233,"version":234,"summary_zh":235,"released_at":236},137040,"langwatch@v2.4.2","## [2.4.2](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcompare\u002Flangwatch@v2.4.1...langwatch@v2.4.2) (2026-02-20)\n\n\n### Bug Fixes\n\n* clean up release_ui_simulations_menu_enabled feature flag ([6ae6175](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F6ae6175aabd99f9c962fc36bbe3dbdbecf852be3))","2026-02-20T19:25:28",{"id":238,"version":239,"summary_zh":212,"released_at":240},137041,"langwatch-helm-2.4.1","2026-02-20T19:13:38",{"id":242,"version":243,"summary_zh":244,"released_at":245},137042,"langwatch@v2.4.1","## [2.4.1](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcompare\u002Flangwatch@v2.4.0...langwatch@v2.4.1) (2026-02-20)\n\n\n### Bug Fixes\n\n* copy SDK package files for generate-sdk-versions.sh in Docker build ([455c0ca](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F455c0caa0e982c29f3cf30cec6b2c3b5ae58eda7))\n* copy SDK package files into Docker image for build ([#1753](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fissues\u002F1753)) ([455c0ca](https:\u002F\u002Fgithub.com\u002Flangwatch\u002Flangwatch\u002Fcommit\u002F455c0caa0e982c29f3cf30cec6b2c3b5ae58eda7))","2026-02-20T18:16:28"]