[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-bentoml--BentoML":3,"tool-bentoml--BentoML":64},[4,17,27,35,43,56],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},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。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},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",140436,2,"2026-04-05T23:32:43",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"status":16},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。",107662,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":23,"last_commit_at":41,"category_tags":42,"status":16},3704,"NextChat","ChatGPTNextWeb\u002FNextChat","NextChat 是一款轻量且极速的 AI 助手，旨在为用户提供流畅、跨平台的大模型交互体验。它完美解决了用户在多设备间切换时难以保持对话连续性，以及面对众多 AI 模型不知如何统一管理的痛点。无论是日常办公、学习辅助还是创意激发，NextChat 都能让用户随时随地通过网页、iOS、Android、Windows、MacOS 或 Linux 端无缝接入智能服务。\n\n这款工具非常适合普通用户、学生、职场人士以及需要私有化部署的企业团队使用。对于开发者而言，它也提供了便捷的自托管方案，支持一键部署到 Vercel 或 Zeabur 等平台。\n\nNextChat 的核心亮点在于其广泛的模型兼容性，原生支持 Claude、DeepSeek、GPT-4 及 Gemini Pro 等主流大模型，让用户在一个界面即可自由切换不同 AI 能力。此外，它还率先支持 MCP（Model Context Protocol）协议，增强了上下文处理能力。针对企业用户，NextChat 提供专业版解决方案，具备品牌定制、细粒度权限控制、内部知识库整合及安全审计等功能，满足公司对数据隐私和个性化管理的高标准要求。",87618,"2026-04-05T07:20:52",[13,26],{"id":44,"name":45,"github_repo":46,"description_zh":47,"stars":48,"difficulty_score":23,"last_commit_at":49,"category_tags":50,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,51,52,53,15,54,26,13,55],"数据工具","视频","插件","其他","音频",{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"status":16},3128,"ragflow","infiniflow\u002Fragflow","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。",77062,"2026-04-04T04:44:48",[15,14,13,26,54],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":69,"readme_en":70,"readme_zh":71,"quickstart_zh":72,"use_case_zh":73,"hero_image_url":74,"owner_login":75,"owner_name":67,"owner_avatar_url":76,"owner_bio":77,"owner_company":78,"owner_location":78,"owner_email":78,"owner_twitter":79,"owner_website":80,"owner_url":81,"languages":82,"stars":117,"forks":118,"last_commit_at":119,"license":120,"difficulty_score":23,"env_os":121,"env_gpu":122,"env_ram":121,"env_deps":123,"category_tags":129,"github_topics":130,"view_count":23,"oss_zip_url":78,"oss_zip_packed_at":78,"status":16,"created_at":146,"updated_at":147,"faqs":148,"releases":177},2252,"bentoml\u002FBentoML","BentoML","The easiest way to serve AI apps and models - Build Model Inference APIs, Job queues, LLM apps, Multi-model pipelines, and more!","BentoML 是一个专为人工智能应用打造的统一模型服务框架，旨在让开发者以最简单的方式将各类 AI 模型转化为生产级的在线服务。无论是构建模型推理 API、任务队列、大语言模型（LLM）应用，还是复杂的多模型流水线，它都能轻松胜任。\n\n在开发过程中，团队常面临环境依赖冲突、部署流程繁琐以及硬件资源利用率低等痛点。BentoML 通过标准化的配置自动管理依赖并生成 Docker 镜像，彻底解决了“依赖地狱”问题，确保从本地开发到云端部署的一致性。同时，它内置了动态批处理、模型并行及多阶段流水线编排等高级特性，能最大化 CPU 和 GPU 的计算效率，显著提升推理性能。\n\n这款工具主要面向 AI 工程师、数据科学家及后端开发者。无论您使用的是 PyTorch、TensorFlow 等主流框架，还是自定义的模型脚本，只需几行代码配合标准的 Python 类型提示，即可快速定义出高性能的服务接口。BentoML 支持在本地便捷调试，也能无缝对接生产环境，是连接模型训练与真实业务场景的高效桥梁，帮助团队更专注于核心算法逻辑而非基础设施的琐碎细节。","\u003Cpicture>\n    \u003Csource media=\"(prefers-color-scheme: dark)\" srcset=\"https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fassets\u002F489344\u002Fd3e6c95d-d224-49a5-9cff-0789f094e127\">\n    \u003Csource media=\"(prefers-color-scheme: light)\" srcset=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbentoml_BentoML_readme_374f9dc76a9f.png\">\n    \u003Cimg alt=\"BentoML: Unified Model Serving Framework\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbentoml_BentoML_readme_374f9dc76a9f.png\" width=\"370\" style=\"max-width: 100%;\">\n\u003C\u002Fpicture>\n\n## Unified Model Serving Framework\n\n🍱 Build model inference APIs and multi-model serving systems with any open-source or custom AI models. 👉 [Join our forum](https:\u002F\u002Fforum.modular.com\u002Fc\u002Fbento\u002F31)!\n\n[![License: Apache-2.0](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-Apache%202-green.svg)](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML?tab=Apache-2.0-1-ov-file)\n[![Releases](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fv\u002Frelease\u002Fbentoml\u002Fbentoml.svg)](https:\u002F\u002Fgithub.com\u002Fbentoml\u002Fbentoml\u002Freleases)\n[![CI](https:\u002F\u002Fgithub.com\u002Fbentoml\u002Fbentoml\u002Factions\u002Fworkflows\u002Fci.yml\u002Fbadge.svg?branch=main)](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Factions\u002Fworkflows\u002Fci.yml?query=branch%3Amain)\n[![Twitter](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbentoml_BentoML_readme_1df618eb85f5.png)](https:\u002F\u002Ftwitter.com\u002Fbentomlai)\n\n## What is BentoML?\n\nBentoML is a Python library for building online serving systems optimized for AI apps and model inference.\n\n- **🍱 Easily build APIs for Any AI\u002FML Model.** Turn any model inference script into a REST API server with just a few lines of code and standard Python type hints.\n- **🐳 Docker Containers made simple.** No more dependency hell! Manage your environments, dependencies and model versions with a simple config file. BentoML automatically generates Docker images, ensures reproducibility, and simplifies how you deploy to different environments.\n- **🧭 Maximize CPU\u002FGPU utilization.** Build high performance inference APIs leveraging built-in serving optimization features like dynamic batching, model parallelism, multi-stage pipeline and multi-model inference-graph orchestration.\n- **👩‍💻 Fully customizable.** Easily implement your own APIs or task queues, with custom business logic, model inference and multi-model composition. Supports any ML framework, modality, and inference runtime.\n- **🚀 Ready for Production.** Develop, run and debug locally. Seamlessly deploy to production with Docker containers or [BentoCloud](https:\u002F\u002Fwww.bentoml.com\u002F).\n\n## Getting started\n\nInstall BentoML:\n\n```\n# Requires Python≥3.9\npip install -U bentoml\n```\n\nDefine APIs in a `service.py` file.\n\n```python\nimport bentoml\n\n@bentoml.service(\n    image=bentoml.images.Image(python_version=\"3.11\").python_packages(\"torch\", \"transformers\"),\n)\nclass Summarization:\n    def __init__(self) -> None:\n        import torch\n        from transformers import pipeline\n\n        device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n        self.pipeline = pipeline('summarization', device=device)\n\n    @bentoml.api(batchable=True)\n    def summarize(self, texts: list[str]) -> list[str]:\n        results = self.pipeline(texts)\n        return [item['summary_text'] for item in results]\n```\n\n### 💻 Run locally\n\nInstall PyTorch and Transformers packages to your Python virtual environment.\n\n```bash\npip install torch transformers  # additional dependencies for local run\n```\n\nRun the service code locally (serving at http:\u002F\u002Flocalhost:3000 by default):\n\n```bash\nbentoml serve\n```\n\nYou should expect to see the following output.\n\n```\n[INFO] [cli] Starting production HTTP BentoServer from \"service:Summarization\" listening on http:\u002F\u002Flocalhost:3000 (Press CTRL+C to quit)\n[INFO] [entry_service:Summarization:1] Service Summarization initialized\n```\n\nNow you can run inference from your browser at http:\u002F\u002Flocalhost:3000 or with a Python script:\n\n```python\nimport bentoml\n\nwith bentoml.SyncHTTPClient('http:\u002F\u002Flocalhost:3000') as client:\n    summarized_text: str = client.summarize([bentoml.__doc__])[0]\n    print(f\"Result: {summarized_text}\")\n```\n\n### 🐳 Deploy using Docker\n\nRun `bentoml build` to package necessary code, models, dependency configs into a Bento - the standardized deployable artifact in BentoML:\n\n```bash\nbentoml build\n```\n\nEnsure [Docker](https:\u002F\u002Fdocs.docker.com\u002F) is running. Generate a Docker container image for deployment:\n\n```bash\nbentoml containerize summarization:latest\n```\n\nRun the generated image:\n\n```bash\ndocker run --rm -p 3000:3000 summarization:latest\n```\n\n### ☁️ Deploy on BentoCloud\n\n[BentoCloud](https:\u002F\u002Fwww.bentoml.com) provides compute infrastructure for rapid and reliable GenAI adoption. It helps speed up your BentoML development process leveraging cloud compute resources, and simplify how you deploy, scale and operate BentoML in production.\n\n[Sign up for BentoCloud](https:\u002F\u002Fcloud.bentoml.com\u002Fsignup) for personal access; for enterprise use cases, [contact our team](https:\u002F\u002Fwww.bentoml.com\u002Fcontact).\n\n```bash\n# After signup, run the following command to create an API token:\nbentoml cloud login\n\n# Deploy from current directory:\nbentoml deploy\n```\n\n![bentocloud-ui](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbentoml_BentoML_readme_062912152aa7.png)\n\nFor detailed explanations, read the [Hello World example](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fget-started\u002Fhello-world.html).\n\n## Examples\n\n- LLMs: [Llama 3.2](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoVLLM\u002Ftree\u002Fmain\u002Fllama3.2-11b-vision-instruct), [Mistral](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoVLLM\u002Ftree\u002Fmain\u002Fministral-8b-instruct-2410), [DeepSeek Distil](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoVLLM\u002Ftree\u002Fmain\u002Fdeepseek-r1-distill-llama3.1-8b-tool-calling), and more.\n- Image Generation: [Stable Diffusion 3 Medium](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoDiffusion\u002Ftree\u002Fmain\u002Fsd3-medium), [Stable Video Diffusion](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoDiffusion\u002Ftree\u002Fmain\u002Fsvd), [Stable Diffusion XL Turbo](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoDiffusion\u002Ftree\u002Fmain\u002Fsdxl-turbo), [ControlNet](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoDiffusion\u002Ftree\u002Fmain\u002Fcontrolnet), and [LCM LoRAs](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoDiffusion\u002Ftree\u002Fmain\u002Flcm).\n- Embeddings: [SentenceTransformers](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoSentenceTransformers) and [ColPali](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoColPali)\n- Audio: [ChatTTS](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoChatTTS), [XTTS](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoXTTS), [WhisperX](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoWhisperX), [Bark](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoBark)\n- Computer Vision: [YOLO](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoYolo) and [ResNet](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoResnet)\n- Advanced examples: [Function calling](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoFunctionCalling), [LangGraph](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoLangGraph), [CrewAI](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoCrewAI)\n\nCheck out the [full list](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fexamples\u002Foverview.html) for more sample code and usage.\n\n## Advanced topics\n\n- [Model composition](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fget-started\u002Fmodel-composition.html)\n- [Workers and model parallelization](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fbuild-with-bentoml\u002Fparallelize-requests.html)\n- [Adaptive batching](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fget-started\u002Fadaptive-batching.html)\n- [GPU inference](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fbuild-with-bentoml\u002Fgpu-inference.html)\n- [Distributed serving systems](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fbuild-with-bentoml\u002Fdistributed-services.html)\n- [Concurrency and autoscaling](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fscale-with-bentocloud\u002Fscaling\u002Fautoscaling.html)\n- [Model loading and Model Store](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fbuild-with-bentoml\u002Fmodel-loading-and-management.html)\n- [Observability](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fbuild-with-bentoml\u002Fobservability\u002Findex.html)\n- [BentoCloud deployment](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fget-started\u002Fcloud-deployment.html)\n\nSee [Documentation](https:\u002F\u002Fdocs.bentoml.com) for more tutorials and guides.\n\n## Community\n\nGet involved and join our [Community Forum 💬](https:\u002F\u002Fforum.modular.com\u002Fc\u002Fbento\u002F31), where thousands of AI\u002FML engineers help each other, contribute to the project, and talk about building AI products.\n\nTo report a bug or suggest a feature request, use\n[GitHub Issues](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fissues\u002Fnew\u002Fchoose).\n\n### Contributing\n\nThere are many ways to contribute to the project:\n\n- Report bugs and \"Thumbs up\" on [issues](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fissues) that are relevant to you.\n- Investigate [issues](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fissues) and review other developers' [pull requests](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpulls).\n- Contribute code or [documentation](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Findex.html) to the project by submitting a GitHub pull request.\n- Check out the [Contributing Guide](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fblob\u002Fmain\u002FCONTRIBUTING.md) and [Development Guide](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fblob\u002Fmain\u002FDEVELOPMENT.md) to learn more.\n- Share your feedback and discuss roadmap plans in our [forum](https:\u002F\u002Fforum.modular.com\u002Fc\u002Fbento\u002F31).\n\nThanks to all of our amazing contributors!\n\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fgraphs\u002Fcontributors\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbentoml_BentoML_readme_a3d4e920276f.png\" \u002F>\n\u003C\u002Fa>\n\n### Usage tracking and feedback\n\nThe BentoML framework collects anonymous usage data that helps our community improve the product. Only BentoML's internal API calls are being reported. This excludes any sensitive information, such as user code, model data, model names, or stack traces. Here's the [code](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fblob\u002Fmain\u002Fsrc\u002Fbentoml\u002F_internal\u002Futils\u002Fanalytics\u002Fusage_stats.py) used for usage tracking. You can opt-out of usage tracking by the `--do-not-track` CLI option:\n\n```bash\nbentoml [command] --do-not-track\n```\n\nOr by setting the environment variable:\n\n```bash\nexport BENTOML_DO_NOT_TRACK=True\n```\n\n### License\n\n[Apache License 2.0](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fblob\u002Fmain\u002FLICENSE)\n","\u003Cpicture>\n    \u003Csource media=\"(prefers-color-scheme: dark)\" srcset=\"https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fassets\u002F489344\u002Fd3e6c95d-d224-49a5-9cff-0789f094e127\">\n    \u003Csource media=\"(prefers-color-scheme: light)\" srcset=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbentoml_BentoML_readme_374f9dc76a9f.png\">\n    \u003Cimg alt=\"BentoML：统一的模型推理框架\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbentoml_BentoML_readme_374f9dc76a9f.png\" width=\"370\" style=\"max-width: 100%;\">\n\u003C\u002Fpicture>\n\n## 统一的模型推理框架\n\n🍱 使用任何开源或自定义的 AI 模型构建模型推理 API 和多模型服务系统。👉 [加入我们的论坛](https:\u002F\u002Fforum.modular.com\u002Fc\u002Fbento\u002F31)！\n\n[![许可证：Apache-2.0](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-Apache%202-green.svg)](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML?tab=Apache-2.0-1-ov-file)\n[![发布版本](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fv\u002Frelease\u002Fbentoml\u002Fbentoml.svg)](https:\u002F\u002Fgithub.com\u002Fbentoml\u002Fbentoml\u002Freleases)\n[![CI](https:\u002F\u002Fgithub.com\u002Fbentoml\u002Fbentoml\u002Factions\u002Fworkflows\u002Fci.yml\u002Fbadge.svg?branch=main)](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Factions\u002Fworkflows\u002Fci.yml?query=branch%3Amain)\n[![Twitter](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbentoml_BentoML_readme_1df618eb85f5.png)](https:\u002F\u002Ftwitter.com\u002Fbentomlai)\n\n## 什么是 BentoML？\n\nBentoML 是一个 Python 库，用于构建针对 AI 应用和模型推理优化的在线服务系统。\n\n- **🍱 轻松为任何 AI\u002FML 模型构建 API。** 只需几行代码和标准的 Python 类型提示，即可将任何模型推理脚本转换为 REST API 服务器。\n- **🐳 简化 Docker 容器管理。** 再也不用担心依赖地狱！通过一个简单的配置文件管理你的环境、依赖项和模型版本。BentoML 自动生成 Docker 镜像，确保可重复性，并简化你在不同环境中的部署流程。\n- **🧭 最大化 CPU\u002FGPU 利用率。** 利用内置的服务优化功能（如动态批处理、模型并行、多阶段流水线和多模型推理图编排）构建高性能的推理 API。\n- **👩‍💻 完全可定制。** 可以轻松实现自己的 API 或任务队列，支持自定义业务逻辑、模型推理和多模型组合。支持任何 ML 框架、模态和推理运行时。\n- **🚀 生产就绪。** 在本地开发、运行和调试。无缝部署到生产环境，使用 Docker 容器或 [BentoCloud](https:\u002F\u002Fwww.bentoml.com\u002F)。\n\n## 快速开始\n\n安装 BentoML：\n\n```\n# 需要 Python≥3.9\npip install -U bentoml\n```\n\n在 `service.py` 文件中定义 API。\n\n```python\nimport bentoml\n\n@bentoml.service(\n    image=bentoml.images.Image(python_version=\"3.11\").python_packages(\"torch\", \"transformers\"),\n)\nclass Summarization:\n    def __init__(self) -> None:\n        import torch\n        from transformers import pipeline\n\n        device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n        self.pipeline = pipeline('summarization', device=device)\n\n    @bentoml.api(batchable=True)\n    def summarize(self, texts: list[str]) -> list[str]:\n        results = self.pipeline(texts)\n        return [item['summary_text'] for item in results]\n```\n\n### 💻 本地运行\n\n将 PyTorch 和 Transformers 包安装到你的 Python 虚拟环境中。\n\n```bash\npip install torch transformers  # 本地运行所需的额外依赖\n```\n\n在本地运行服务代码（默认监听 http:\u002F\u002Flocalhost:3000）：\n\n```bash\nbentoml serve\n```\n\n你应该会看到以下输出。\n\n```\n[INFO] [cli] 从 \"service:Summarization\" 启动生产 HTTP BentoServer，监听 http:\u002F\u002Flocalhost:3000（按 CTRL+C 退出）\n[INFO] [entry_service:Summarization:1] 服务 Summarization 已初始化\n```\n\n现在你可以通过浏览器访问 http:\u002F\u002Flocalhost:3000 进行推理，或者使用 Python 脚本：\n\n```python\nimport bentoml\n\nwith bentoml.SyncHTTPClient('http:\u002F\u002Flocalhost:3000') as client:\n    summarized_text: str = client.summarize([bentoml.__doc__])[0]\n    print(f\"结果：{summarized_text}\")\n```\n\n### 🐳 使用 Docker 部署\n\n运行 `bentoml build` 将必要的代码、模型和依赖配置打包成一个 Bento——BentoML 中的标准可部署工件：\n\n```bash\nbentoml build\n```\n\n确保 [Docker](https:\u002F\u002Fdocs.docker.com\u002F) 正在运行。生成用于部署的 Docker 容器镜像：\n\n```bash\nbentoml containerize summarization:latest\n```\n\n运行生成的镜像：\n\n```bash\ndocker run --rm -p 3000:3000 summarization:latest\n```\n\n### ☁️ 在 BentoCloud 上部署\n\n[BentoCloud](https:\u002F\u002Fwww.bentoml.com) 提供计算基础设施，助力快速可靠地采用 GenAI 技术。它利用云上计算资源加速你的 BentoML 开发流程，并简化你在生产环境中部署、扩展和运维 BentoML 的方式。\n\n[注册 BentoCloud](https:\u002F\u002Fcloud.bentoml.com\u002Fsignup) 获取个人访问权限；对于企业级应用，请 [联系我们的团队](https:\u002F\u002Fwww.bentoml.com\u002Fcontact)。\n\n```bash\n# 注册后，运行以下命令创建 API 令牌：\nbentoml cloud login\n\n# 从当前目录部署：\nbentoml deploy\n```\n\n![bentocloud-ui](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbentoml_BentoML_readme_062912152aa7.png)\n\n有关详细说明，请阅读 [Hello World 示例](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fget-started\u002Fhello-world.html)。\n\n## 示例\n\n- LLM：[Llama 3.2](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoVLLM\u002Ftree\u002Fmain\u002Fllama3.2-11b-vision-instruct)、[Mistral](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoVLLM\u002Ftree\u002Fmain\u002Fministral-8b-instruct-2410)、[DeepSeek Distil](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoVLLM\u002Ftree\u002Fmain\u002Fdeepseek-r1-distill-llama3.1-8b-tool-calling) 等。\n- 图像生成：[Stable Diffusion 3 Medium](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoDiffusion\u002Ftree\u002Fmain\u002Fsd3-medium)、[Stable Video Diffusion](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoDiffusion\u002Ftree\u002Fmain\u002Fsvd)、[Stable Diffusion XL Turbo](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoDiffusion\u002Ftree\u002Fmain\u002Fsdxl-turbo)、[ControlNet](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoDiffusion\u002Ftree\u002Fmain\u002Fcontrolnet) 和 [LCM LoRAs](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoDiffusion\u002Ftree\u002Fmain\u002Flcm) 等。\n- 嵌入：[SentenceTransformers](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoSentenceTransformers) 和 [ColPali](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoColPali)。\n- 音频：[ChatTTS](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoChatTTS)、[XTTS](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoXTTS)、[WhisperX](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoWhisperX) 和 [Bark](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoBark)。\n- 计算机视觉：[YOLO](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoYolo) 和 [ResNet](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoResnet)。\n- 高级示例：[函数调用](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoFunctionCalling)、[LangGraph](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoLangGraph) 和 [CrewAI](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoCrewAI)。\n\n查看 [完整列表](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fexamples\u002Foverview.html)，获取更多示例代码和使用方法。\n\n## 高级主题\n\n- [模型组合](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fget-started\u002Fmodel-composition.html)\n- [工作进程与模型并行化](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fbuild-with-bentoml\u002Fparallelize-requests.html)\n- [自适应批处理](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fget-started\u002Fadaptive-batching.html)\n- [GPU 推理](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fbuild-with-bentoml\u002Fgpu-inference.html)\n- [分布式服务系统](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fbuild-with-bentoml\u002Fdistributed-services.html)\n- [并发与自动扩缩容](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fscale-with-bentocloud\u002Fscaling\u002Fautoscaling.html)\n- [模型加载与 Model Store](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fbuild-with-bentoml\u002Fmodel-loading-and-management.html)\n- [可观测性](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fbuild-with-bentoml\u002Fobservability\u002Findex.html)\n- [BentoCloud 部署](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Fget-started\u002Fcloud-deployment.html)\n\n更多教程和指南请参阅 [文档](https:\u002F\u002Fdocs.bentoml.com)。\n\n## 社区\n\n欢迎参与并加入我们的 [社区论坛 💬](https:\u002F\u002Fforum.modular.com\u002Fc\u002Fbento\u002F31)，这里汇聚了数千名 AI\u002FML 工程师，大家互相帮助、共同为项目贡献力量，并探讨如何构建 AI 产品。\n\n如需报告 bug 或提出功能请求，请使用 [GitHub Issues](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fissues\u002Fnew\u002Fchoose)。\n\n### 贡献方式\n\n您可以通过多种方式为项目做出贡献：\n\n- 报告与您相关的 bug，并为 [issues](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fissues) 点赞。\n- 查看 [issues](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fissues) 并评审其他开发者的 [pull requests](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpulls)。\n- 通过提交 GitHub pull request，为项目贡献代码或 [文档](https:\u002F\u002Fdocs.bentoml.com\u002Fen\u002Flatest\u002Findex.html)。\n- 请查阅 [贡献指南](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fblob\u002Fmain\u002FCONTRIBUTING.md) 和 [开发指南](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fblob\u002Fmain\u002FDEVELOPMENT.md)，了解更多详情。\n- 在我们的 [论坛](https:\u002F\u002Fforum.modular.com\u002Fc\u002Fbento\u002F31) 分享您的反馈，并讨论项目路线图计划。\n\n感谢所有优秀的贡献者！\n\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fgraphs\u002Fcontributors\">\n  \u003Cimg src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbentoml_BentoML_readme_a3d4e920276f.png\" \u002F>\n\u003C\u002Fa>\n\n### 使用情况跟踪与反馈\n\nBentoML 框架会收集匿名的使用数据，以帮助社区改进产品。仅记录 BentoML 的内部 API 调用，不包含任何敏感信息，例如用户代码、模型数据、模型名称或堆栈跟踪。用于使用情况跟踪的 [代码](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fblob\u002Fmain\u002Fsrc\u002Fbentoml\u002F_internal\u002Futils\u002Fanalytics\u002Fusage_stats.py) 如下。您可以通过 `--do-not-track` CLI 选项选择退出使用情况跟踪：\n\n```bash\nbentoml [命令] --do-not-track\n```\n\n或者设置环境变量：\n\n```bash\nexport BENTOML_DO_NOT_TRACK=True\n```\n\n### 许可证\n\n[Apache License 2.0](https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fblob\u002Fmain\u002FLICENSE)","# BentoML 快速上手指南\n\nBentoML 是一个用于构建和优化 AI 应用模型推理服务的统一框架。它能帮助你将任意机器学习模型快速转换为高性能的 REST API，并轻松打包为 Docker 容器或部署到云端。\n\n## 环境准备\n\n在开始之前，请确保你的开发环境满足以下要求：\n\n*   **操作系统**：Linux, macOS, 或 Windows (推荐 WSL2)\n*   **Python 版本**：Python ≥ 3.9\n*   **前置依赖**：\n    *   `pip` (Python 包管理工具)\n    *   (可选) **Docker**：如果你计划将服务容器化部署，需预先安装并运行 Docker Desktop 或 Docker Engine。\n\n> **💡 国内开发者提示**：为避免下载依赖包时速度过慢，建议配置国内镜像源（如清华源或阿里源）。\n> ```bash\n> pip install -U bentoml -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n> ```\n\n## 安装步骤\n\n使用 pip 安装最新版本的 BentoML：\n\n```bash\npip install -U bentoml\n```\n\n## 基本使用\n\n以下是构建一个基于 Hugging Face Transformers 的文本摘要服务的极简流程。\n\n### 1. 定义服务\n\n创建名为 `service.py` 的文件，编写以下代码。BentoML 允许你使用标准的 Python 类和方法来定义 API，并通过装饰器自动处理批处理和依赖管理。\n\n```python\nimport bentoml\n\n@bentoml.service(\n    image=bentoml.images.Image(python_version=\"3.11\").python_packages(\"torch\", \"transformers\"),\n)\nclass Summarization:\n    def __init__(self) -> None:\n        import torch\n        from transformers import pipeline\n\n        device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n        self.pipeline = pipeline('summarization', device=device)\n\n    @bentoml.api(batchable=True)\n    def summarize(self, texts: list[str]) -> list[str]:\n        results = self.pipeline(texts)\n        return [item['summary_text'] for item in results]\n```\n\n### 2. 本地运行\n\n首先安装运行所需的额外依赖（PyTorch 和 Transformers）：\n\n```bash\npip install torch transformers\n```\n\n启动本地服务（默认监听 `http:\u002F\u002Flocalhost:3000`）：\n\n```bash\nbentoml serve\n```\n\n当看到以下日志时，表示服务已就绪：\n```\n[INFO] [cli] Starting production HTTP BentoServer from \"service:Summarization\" listening on http:\u002F\u002Flocalhost:3000 (Press CTRL+C to quit)\n[INFO] [entry_service:Summarization:1] Service Summarization initialized\n```\n\n### 3. 调用测试\n\n你可以直接在浏览器访问 `http:\u002F\u002Flocalhost:3000` 查看交互式文档，或使用 Python 脚本进行调用：\n\n```python\nimport bentoml\n\nwith bentoml.SyncHTTPClient('http:\u002F\u002Flocalhost:3000') as client:\n    summarized_text: str = client.summarize([bentoml.__doc__])[0]\n    print(f\"Result: {summarized_text}\")\n```\n\n### 4. 打包与部署 (可选)\n\nBentoML 可将上述服务一键打包为标准化的构建产物（Bento），并生成 Docker 镜像：\n\n```bash\n# 1. 构建 Bento 包\nbentoml build\n\n# 2. 生成 Docker 镜像\nbentoml containerize summarization:latest\n\n# 3. 运行 Docker 容器\ndocker run --rm -p 3000:3000 summarization:latest\n```","某电商初创团队急需将最新的开源大语言模型集成到客服系统中，以实现对用户咨询的自动摘要和智能回复。\n\n### 没有 BentoML 时\n- **环境依赖地狱**：数据科学家在本地能跑通的模型，部署到服务器时因 PyTorch、CUDA 版本不一致频繁报错，排查耗时数天。\n- **高性能并发难实现**：面对突发流量，手动编写代码实现动态批处理（Dynamic Batching）极其复杂，导致 GPU 利用率低且响应延迟高。\n- **多模型协作混乱**：需要串联“意图识别”和“文本生成”两个模型时，不得不维护多个独立的 Flask\u002FFastAPI 服务，链路调用逻辑繁琐且难以调试。\n- **交付标准不统一**：每次上线都需要人工编写 Dockerfile，缺乏标准化的打包流程，导致开发环境与生产环境行为不一致。\n\n### 使用 BentoML 后\n- **一键容器化部署**：仅需几行代码定义依赖，BentoML 自动生成包含特定 Python 版本和模型权重的 Docker 镜像，彻底消除环境差异。\n- **内置性能优化**：通过 `@bentoml.api(batchable=True)` 装饰器轻松开启动态批处理，无需修改业务逻辑即可最大化 GPU 吞吐量，显著降低延迟。\n- **原生支持多模型编排**：在一个服务类中直接实例化多个模型并编排推理流水线，像调用本地函数一样简洁地构建复杂的 AI 应用。\n- **标准化生产流程**：提供从本地调试到云端部署的一致体验，开发者只需关注模型逻辑，剩余的工程化难题由框架自动解决。\n\nBentoML 让团队从繁琐的基础设施运维中解放出来，将模型从实验代码转化为高可用生产服务的周期从数周缩短至数小时。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fbentoml_BentoML_06291215.png","bentoml","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fbentoml_c36be549.jpg","Build fast and reliable model serving systems",null,"bentomlai","https:\u002F\u002Fbentoml.com","https:\u002F\u002Fgithub.com\u002Fbentoml",[83,87,91,95,99,103,107,110,114],{"name":84,"color":85,"percentage":86},"Python","#3572A5",96.6,{"name":88,"color":89,"percentage":90},"Shell","#89e051",1.8,{"name":92,"color":93,"percentage":94},"Jinja","#a52a22",0.6,{"name":96,"color":97,"percentage":98},"Starlark","#76d275",0.4,{"name":100,"color":101,"percentage":102},"Dockerfile","#384d54",0.3,{"name":104,"color":105,"percentage":106},"HTML","#e34c26",0.1,{"name":108,"color":109,"percentage":106},"Makefile","#427819",{"name":111,"color":112,"percentage":113},"CSS","#663399",0,{"name":115,"color":116,"percentage":113},"JavaScript","#f1e05a",8555,945,"2026-04-05T10:10:58","Apache-2.0","未说明","非必需（支持 CPU 运行），若使用 GPU 需 NVIDIA 显卡并安装 PyTorch CUDA 版本，具体型号和显存取决于所选模型",{"notes":124,"python":125,"dependencies":126},"BentoML 是一个用于构建 AI 模型推理 API 的 Python 库。它支持动态批处理、模型并行和多模型编排以优化性能。本地运行时需手动安装特定模型的依赖（如 torch, transformers）。生产环境部署可自动生成 Docker 镜像以确保环境一致性。支持通过 BentoCloud 进行云端部署和管理。","3.9+",[75,127,128],"torch","transformers",[13,26,54],[131,132,133,134,135,136,137,138,139,140,141,142,143,144,145],"model-serving","mlops","llmops","generative-ai","llm-inference","model-inference-service","inference-platform","deep-learning","llm-serving","machine-learning","python","multimodal","ml-engineering","llm","ai-inference","2026-03-27T02:49:30.150509","2026-04-06T11:30:50.533801",[149,154,159,164,169,173],{"id":150,"question_zh":151,"answer_zh":152,"source_url":153},10356,"使用 BentoML >= 1.2 版本时遇到内存泄漏问题怎么办？","该问题已在后续版本中修复。如果您遇到内存持续增加的情况，请尝试升级到最新版本或使用主分支（main branch）的代码进行验证。维护者确认相关 PR 已合并，将在下一个发布版本中可用。临时解决方案是避免使用有问题的旧版本，或按照社区反馈检查是否因未正确打包导致（例如直接运行 `bentoml serve` 而非容器化部署时表现不同）。","https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fissues\u002F4760",{"id":155,"question_zh":156,"answer_zh":157,"source_url":158},10357,"在生产模式（--production）下运行 Bento 服务时报 AnyIO 错误如何解决？","在某些环境下，直接使用 `bentoml serve --production` 可能会触发 AnyIO 相关错误。一个有效的解决方法是将服务容器化后通过 Docker 运行。用户反馈表明，虽然直接在主机上运行生产模式会失败，但构建为 Docker 镜像并作为容器运行时工作正常。建议优先使用容器化部署以确保生产环境的稳定性。","https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fissues\u002F2271",{"id":160,"question_zh":161,"answer_zh":162,"source_url":163},10358,"如何在受限网络环境中处理 Conda 依赖的自定义频道（channels）问题？","在无法访问默认 Conda 频道（如 anaconda.com）的环境中，可以通过设置 `@env(conda_override_channels=True)` 来解决。此配置会在生成的 `environment.yml` 文件中自动添加 `nodefaults` 条目，从而阻止 Conda 使用默认的 \"defaults\" 频道，仅使用您指定的频道。这确保了在离线或受限网络环境下的构建能够成功。","https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fissues\u002F1429",{"id":165,"question_zh":166,"answer_zh":167,"source_url":168},10359,"为什么添加 --production 标志后模型推理速度变慢了？","`--production` 标志旨在优化高并发场景，但在某些旧版本或特定配置下可能导致单次请求延迟增加。如果 CPU 或 GPU 利用率未饱和，瓶颈可能在于 API 或 Runner 代码本身。建议检查是否使用了异步 API 函数，并确保 Runner 配置合理。该问题在较新版本中已得到大部分解决，如遇此情况建议升级到最新版或新建 Issue 提供详细基准测试数据。","https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fissues\u002F2267",{"id":170,"question_zh":171,"answer_zh":172,"source_url":153},10360,"BentoML 服务在 Kubernetes 中部署时内存持续增长的原因是什么？","如果在 Kubernetes 中部署 BentoML 服务时发现内存随负载测试（如 Locust）持续上升，这通常是由于特定版本（如 1.2）的内存泄漏 bug 导致的。解决方案是升级到修复了该问题的最新版本。此外，确保使用正确的容器化命令 `bentoml build -f bentofile.yaml --containerize` 构建镜像，并在 Deployment 配置中正确设置存活探针（livenessProbe）和就绪探针（readinessProbe）。",{"id":174,"question_zh":175,"answer_zh":176,"source_url":158},10361,"如何在没有 Docker 的情况下解决生产模式下的运行错误？","虽然容器化是解决生产模式运行错误（如 AnyIO 错误）的推荐方法，但如果暂时无法使用 Docker，可以尝试检查环境依赖版本兼容性（如 Python 3.8, AnyIO 3.5.0 等）。不过，根据社区反馈，最稳妥的方式仍是尽快迁移到 Docker 环境，因为许多生产模式的底层优化是围绕容器环境设计的。如果必须在不使用 Docker 的情况下运行，请确保所有依赖项与生产模式完全兼容，并考虑回退到开发模式直到环境问题解决。",[178,183,188,193,198,203,208,213,218,223,228,233,238,243,248,253,258,263,268,273],{"id":179,"version":180,"summary_zh":181,"released_at":182},107589,"v1.4.38","## What's Changed\r\n* feat: native support for src-layout projects by @VedantMadane in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5555\r\n* fix: switch to SandboxedEnvironment and remove unused Jinja2 extensions by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5591\r\n* fix: use correct NVIDIA CUDA base images for debian distro by @mattmorganpdx in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5590\r\n\r\n## New Contributors\r\n* @mattmorganpdx made their first contribution in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5590\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.37...v1.4.38","2026-04-02T06:12:17",{"id":184,"version":185,"summary_zh":186,"released_at":187},107590,"v1.4.37","## What's Changed\r\n* docs: Update slack to forum link by @Sherlock113 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5567\r\n* fix: correctly load `bentoml.models.HuggingFaceModel` in container by @nickthegroot in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5582\r\n* chore(deps): bump docker\u002Fsetup-qemu-action from 3 to 4 by @dependabot[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5564\r\n* chore(deps): bump marocchino\u002Fsticky-pull-request-comment from 2 to 3 by @dependabot[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5571\r\n* chore(deps): bump docker\u002Fsetup-buildx-action from 3 to 4 by @dependabot[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5565\r\n* fix: resolve SQLite 'database is locked' errors under high concurrency by @VedantMadane in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5558\r\n* fix: shell quote system package names in Dockerfile templates and image commands by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5583\r\n* fix: reorder histogram samples in multiprocess prometheus output by @saivedant169 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5570\r\n\r\n## New Contributors\r\n* @nickthegroot made their first contribution in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5582\r\n* @saivedant169 made their first contribution in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5570\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.36...v1.4.37","2026-03-25T01:20:40",{"id":189,"version":190,"summary_zh":191,"released_at":192},107591,"v1.4.36","## What's Changed\r\n* fix: correct typo 'seperators' to 'separators' by @thecaptain789 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5546\r\n* Reapply \"feat: new image API: build_include\" (#5531) by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5539\r\n* fix: resolve AnyIO NoEventLoopError when calling sync API from async API by @paipeline in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5550\r\n* fix: Set SQLite busy_timeout and WAL mode to prevent 'database is locked' under concurrency by @VedantMadane in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5551\r\n* Fix memory leak in readiness checks with remote dependencies by @paipeline in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5553\r\n* fix: validate symlink targets in safe_extract_tarfile by @q1uf3ng in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5548\r\n* Revert \"fix: resolve AnyIO NoEventLoopError when calling sync API from async API\" by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5554\r\n* chore: update AWS BYOC doc to v10 by @sauyon in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5559\r\n* chore(deps): bump actions\u002Fdownload-artifact from 7 to 8 by @dependabot[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5561\r\n* chore(deps): bump actions\u002Fupload-artifact from 6 to 7 by @dependabot[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5560\r\n* ci: pre-commit autoupdate [skip ci] by @pre-commit-ci[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5562\r\n\r\n## New Contributors\r\n* @thecaptain789 made their first contribution in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5546\r\n* @paipeline made their first contribution in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5550\r\n* @VedantMadane made their first contribution in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5551\r\n* @q1uf3ng made their first contribution in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5548\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.35...v1.4.36","2026-03-06T04:25:28",{"id":194,"version":195,"summary_zh":196,"released_at":197},107592,"v1.4.35","## What's Changed\r\n* fix(client): recreate connector on session refresh to prevent closed session errors by @Wirg in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5541\r\n* ci: pre-commit autoupdate [skip ci] by @pre-commit-ci[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5542\r\n* fix: Support importing multiple Bentos with shared models by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5543\r\n\r\n## New Contributors\r\n* @Wirg made their first contribution in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5541\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.34...v1.4.35","2026-02-03T05:06:31",{"id":199,"version":200,"summary_zh":201,"released_at":202},107593,"v1.4.34","- Fixed a security issue when resolving file paths input by user.\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.33...v1.4.34","2026-01-26T03:19:56",{"id":204,"version":205,"summary_zh":206,"released_at":207},107594,"v1.4.33","## What's Changed\r\n* fix: update installation logic for Python packages in Dockerfile template by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5534\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.32...v1.4.33","2026-01-12T10:26:47",{"id":209,"version":210,"summary_zh":211,"released_at":212},107595,"v1.4.32","## What's Changed\r\n* fix: ensure at least one CPU worker is assigned when CPU resources are zero by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5528\r\n* fix: update installation path logic for Python packages in Dockerfile template by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5529\r\n* Revert \"feat: new image API: build_include\" by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5531\r\n* ci: pre-commit autoupdate [skip ci] by @pre-commit-ci[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5530\r\n* fix: sanitize filename with path separators in FileSchema.decode by @Anri-Lombard in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5533\r\n\r\n## New Contributors\r\n* @Anri-Lombard made their first contribution in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5533\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.31...v1.4.32","2026-01-09T02:22:49",{"id":214,"version":215,"summary_zh":216,"released_at":217},107596,"v1.4.31","## What's Changed\r\n* chore: update pre-commit config and fix lint errors by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5511\r\n* ci: pre-commit autoupdate [skip ci] by @pre-commit-ci[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5512\r\n* doc(gateway): add gateway documentation by @ssheng in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5513\r\n* docs: Add gateway screenshots by @Sherlock113 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5515\r\n* feat: add build time secrets by @jianshen92 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5514\r\n* feat: API token sdk by @jianshen92 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5516\r\n* docs: API token SDK documentation by @jianshen92 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5517\r\n* feat: new image API: build_include by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5518\r\n* feature: Add native prefix routing (root_path) support for BentoML by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5521\r\n* fix: support max age and max requets for runner connections by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5522\r\n* fix: improve the error message for importing bento by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5523\r\n* chore(deps): bump actions\u002Fdownload-artifact from 6 to 7 by @dependabot[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5520\r\n* chore(deps): bump actions\u002Fupload-artifact from 5 to 6 by @dependabot[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5519\r\n* feat: support stage in secret CLI by @jianshen92 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5526\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.30...v1.4.31","2025-12-23T06:51:00",{"id":219,"version":220,"summary_zh":221,"released_at":222},107597,"v1.4.30","## What's Changed\r\n* chore(deps): bump actions\u002Fcheckout from 5 to 6 by @dependabot[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5505\r\n* fix: aiohttp impl of RemoteProxy for dependency call by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5502\r\n* docs: update AWS CloudFormation template URL to v9 by @sauyon in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5506\r\n* fix: improve process termination handling in create_proxy_app by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5508\r\n* fix: cli deploy to inject deployment env by @jianshen92 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5507\r\n* fix: set default worker count based on available CPU resources by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5509\r\n* fix: ensure at least one worker is set based on CPU resources by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5510\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.29...v1.4.30","2025-11-27T03:15:11",{"id":224,"version":225,"summary_zh":226,"released_at":227},107598,"v1.4.29","## What's Changed\r\n* docs: Add byoc link by @Sherlock113 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5498\r\n* ci: pre-commit autoupdate [skip ci] by @pre-commit-ci[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5500\r\n* feat: associate bento service instance with the service class by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5501\r\n* fix: update apt-get install command to use Dpkg options by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5503\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.28...v1.4.29","2025-11-17T12:02:22",{"id":229,"version":230,"summary_zh":231,"released_at":232},107599,"v1.4.28","## What's Changed\r\n* fix: update cuda base image and base image docs by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5492\r\n* fix: use dir() instead of vars() to walk through the MRO by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5495\r\n* chore(deps): bump actions\u002Fdownload-artifact from 5 to 6 by @dependabot[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5494\r\n* chore(deps): bump actions\u002Fupload-artifact from 4 to 5 by @dependabot[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5493\r\n* fix: feature: build error message with missing BentoModel by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5496\r\n* feat: multi-worker support for custom command service by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5497\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.27...v1.4.28","2025-10-29T06:09:36",{"id":234,"version":235,"summary_zh":236,"released_at":237},107600,"v1.4.27","## What's Changed\r\n* chore(deps): bump github\u002Fcodeql-action from 3 to 4 by @dependabot[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5486\r\n* chore(deps): bump astral-sh\u002Fsetup-uv from 6 to 7 by @dependabot[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5487\r\n* docs: Update vllm inference doc by @Sherlock113 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5488\r\n* feat: support python 3.14 in build by @jianshen92 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5489\r\n* fix: exclude metrics for endpoints by @bojiang in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5491\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.26...v1.4.27","2025-10-20T06:57:33",{"id":239,"version":240,"summary_zh":241,"released_at":242},107601,"v1.4.26","## What's Changed\r\n* fix: avoid circular reference in response background handling by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5481\r\n* fix: error handling in proxy request processing by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5482\r\n* ci: pre-commit autoupdate [skip ci] by @pre-commit-ci[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5485\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.25...v1.4.26","2025-10-10T07:28:39",{"id":244,"version":245,"summary_zh":246,"released_at":247},107602,"v1.4.25","## What's Changed\r\n* fix: Adding secret as an argument to create deployment by @sean-hickey-wf in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5478\r\n* docs: Add equivalent code note by @Sherlock113 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5476\r\n* skip configured livez and readyz endpoints in traffic and access middlewares by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5473\r\n* fix: Ensure sdk respects _cloud_client when performing deployment operations by @sean-hickey-wf in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5474\r\n* feat: add bentoml deployment start command by @jianshen92 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5475\r\n* fix: handle compatibility with click 8.3+ by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5480\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.24...v1.4.25","2025-09-24T08:51:56",{"id":249,"version":250,"summary_zh":251,"released_at":252},107603,"v1.4.24","## What's Changed\r\n* doc: update gpu options for on-demand p instances by @ssheng in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5456\r\n* docs: Add labels by @Sherlock113 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5457\r\n* chore: add gpu literals by @ssheng in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5459\r\n* fix: use configurable readyz endpoint in HTTPClient classes by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5458\r\n* fix: use abspath instead of realpath in safe_extract_tarfile function by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5461\r\n* chore(deps): bump actions\u002Fsetup-python from 5 to 6 by @dependabot[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5462\r\n* fix: replace questionary with rich-toolkit by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5460\r\n* fix: adding build args option to apply cli command by @sean-hickey-wf in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5468\r\n* feat: Adding label as cli command for cloud deployment by @sean-hickey-wf in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5467\r\n* docs: Add split staging and prod doc by @Sherlock113 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5463\r\n* feat: Allow bento description to be passed to service decorator by @sean-hickey-wf in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5470\r\n* feat: implement reverse proxy for custom command services by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5465\r\n* docs: Add traffic config ref link by @Sherlock113 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5471\r\n* feat: enforce requirement for at least one dependency attribute in Dependency class by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5472\r\n\r\n## New Contributors\r\n* @sean-hickey-wf made their first contribution in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5468\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.23...v1.4.24","2025-09-17T09:30:34",{"id":254,"version":255,"summary_zh":256,"released_at":257},107604,"v1.4.23","## What's Changed\r\n* feat: add support for extra_ports in service configuration by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5441\r\n* docs: Fix typo in DataFrame annotation for classify method by @jianshen92 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5442\r\n* fix: call lifespans on mounted starlette apps by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5444\r\n* refactor: remove unused endpoint labels from service configuration and add endpoints field to BentoManifestSchema by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5448\r\n* feat: add a worker script for process runner by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5447\r\n* docs: Add Prophet example by @Sherlock113 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5446\r\n* feat: add endpoints field to Bento class and update related methods by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5449\r\n* ci: pre-commit autoupdate [skip ci] by @pre-commit-ci[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5450\r\n* feat: get_hosts via runner-lb API by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5445\r\n* fix: revert get_hosts to only returning hostnames by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5451\r\n* fix: remove unnecessary command from create_dependency_watcher function by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5452\r\n* refactor: remove extra_ports and endpoints from Bento class and related schemas by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5453\r\n* doc: update AWS byoc docs to use role by @sauyon in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5454\r\n* fix: update service host return values to include port for UDS connections by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5455\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.22...v1.4.23","2025-09-05T00:51:38",{"id":259,"version":260,"summary_zh":261,"released_at":262},107605,"v1.4.22","## What's Changed\r\n* docs: Update diagram image by @Sherlock113 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5431\r\n* chore(deps): bump actions\u002Fcheckout from 4 to 5 by @dependabot[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5432\r\n* fix: Correct argument merging logic in set_arguments function by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5433\r\n* feat: Add new GPU type 'amd-mi300x' to GpuLiteralType by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5435\r\n* feat: allow to get command later for service by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5437\r\n* fix: Add error handling for failed file downloads in JSONSerde and MultipartSerde by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5436\r\n* docs: Add examples by @Sherlock113 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5439\r\n* fix: respect the metrics namespace passed from config by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5440\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.21...v1.4.22","2025-08-27T01:20:34",{"id":264,"version":265,"summary_zh":266,"released_at":267},107606,"v1.4.21","## What's Changed\r\n* fix: Move wheels directory if src\u002Fwheels exists by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5430\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.20...v1.4.21","2025-08-14T01:07:36",{"id":269,"version":270,"summary_zh":271,"released_at":272},107607,"v1.4.20","## What's Changed\r\n* docs: Add rollback section by @Sherlock113 in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5423\r\n* feat: Replace usages of fs library with fsspec and pathlib by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5424\r\n* ci: pre-commit autoupdate [skip ci] by @pre-commit-ci[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5425\r\n* fix: Add env var to disable local bentoml URL in requirements by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5427\r\n* chore(deps): bump actions\u002Fdownload-artifact from 4 to 5 by @dependabot[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5428\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.19...v1.4.20","2025-08-13T08:02:48",{"id":274,"version":275,"summary_zh":276,"released_at":277},107608,"v1.4.19","## What's Changed\r\n* fix: run post setup commands for codespace by @frostming in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5414\r\n* fix: Ignore python.packages when requirements_txt is provided by @qimcis in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5421\r\n* chore(deps): bump the pip group across 2 directories with 1 update by @dependabot[bot] in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5413\r\n\r\n## New Contributors\r\n* @qimcis made their first contribution in https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fpull\u002F5421\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fbentoml\u002FBentoML\u002Fcompare\u002Fv1.4.18...v1.4.19","2025-07-29T03:33:55"]