[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-zkonduit--ezkl":3,"tool-zkonduit--ezkl":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",138956,2,"2026-04-05T11:33:21",[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":76,"owner_avatar_url":77,"owner_bio":78,"owner_company":79,"owner_location":79,"owner_email":79,"owner_twitter":79,"owner_website":79,"owner_url":80,"languages":81,"stars":98,"forks":99,"last_commit_at":100,"license":79,"difficulty_score":10,"env_os":101,"env_gpu":102,"env_ram":101,"env_deps":103,"category_tags":116,"github_topics":117,"view_count":10,"oss_zip_url":79,"oss_zip_packed_at":79,"status":16,"created_at":122,"updated_at":123,"faqs":124,"releases":152},794,"zkonduit\u002Fezkl","ezkl","ezkl is an engine for doing inference for deep learning models and other computational graphs in a zk-snark (ZKML). Use it from Python, Javascript, or the command line. ","ezkl 是一个专注于零知识机器学习（ZKML）推理的开源引擎。它让开发者能够轻松地将深度学习模型或其他计算图转化为 zk-snark 电路，从而在保护隐私的前提下验证计算结果。\n\n传统机器学习模型往往面临数据隐私和结果可信度的挑战。ezkl 通过生成零知识证明，让用户可以证明“我在私有数据上运行了公开模型”或“我正确执行了特定算法”，而无需暴露输入数据或模型权重。这种机制特别适用于区块链智能合约、隐私计算等场景，因为生成的证明可以在资源受限的环境（如浏览器或以太坊虚拟机）中快速验证。\n\nezkl 支持主流框架如 PyTorch 和 TensorFlow，只需导出为 ONNX 格式即可接入。其底层采用 Halo2 证明系统，兼顾了效率与安全性。无论是希望通过 Python 或 JavaScript 库进行集成，还是使用命令行工具快速测试，ezkl 都能提供灵活的开发体验。对于关注隐私保护、希望将 AI 能力安全部署到去中心化网络的研究人员和开发者来说，ezkl 是一个值得尝试的选择。","\u003Ch1 align=\"center\">\n\t\u003Cbr>\n\t :thought_balloon:\n\t\u003Cbr>\n\t\u003Cbr>\nEZKL\n\t\u003Cbr>\n\t\u003Cbr>\n\t\u003Cbr>\n\u003C\u002Fh1>\n\n> Easy Zero-Knowledge Inference\n\n[![Test](https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fworkflows\u002FRust\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Factions?query=workflow%3ARust)\n\n`ezkl` is a library and command-line tool for doing inference for deep learning models and other computational graphs in a zk-snark (ZKML). It enables the following workflow:\n\n1. Define a computational graph, for instance a neural network (but really any arbitrary set of operations), as you would normally in pytorch or tensorflow.\n2. Export the final graph of operations as an [.onnx](https:\u002F\u002Fonnx.ai\u002F) file and some sample inputs to a `.json` file.\n3. Point `ezkl` to the `.onnx` and `.json` files to generate a ZK-SNARK circuit with which you can prove statements such as:\n\n> \"I ran this publicly available neural network on some private data and it produced this output\"\n\n[![Notebook](https:\u002F\u002Fcolab.research.google.com\u002Fassets\u002Fcolab-badge.svg)](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fzkonduit\u002Fezkl\u002Fblob\u002Fmain\u002Fexamples\u002Fnotebooks\u002Fsimple_demo_public_network_output.ipynb) \n\n> \"I ran my private neural network on some public data and it produced this output\"\n\n[![Notebook](https:\u002F\u002Fcolab.research.google.com\u002Fassets\u002Fcolab-badge.svg)](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fzkonduit\u002Fezkl\u002Fblob\u002Fmain\u002Fexamples\u002Fnotebooks\u002Fsimple_demo_public_input_output.ipynb) \n\n> \"I correctly ran this publicly available neural network on some public data and it produced this output\"\n\n[![Notebook](https:\u002F\u002Fcolab.research.google.com\u002Fassets\u002Fcolab-badge.svg)](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fzkonduit\u002Fezkl\u002Fblob\u002Fmain\u002Fexamples\u002Fnotebooks\u002Fsimple_demo_all_public.ipynb) \n\nIn the backend we use the collaboratively-developed [Halo2](https:\u002F\u002Fgithub.com\u002Fprivacy-scaling-explorations\u002Fhalo2) as a proof system.\n\nThe generated proofs can then be verified with much less computational resources, including on-chain (with the Ethereum Virtual Machine), in a browser, or on a device. \n\n- If you have any questions, we'd love for you to open up a discussion topic in [Discussions](https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fdiscussions). Alternatively, you can join the ✨[EZKL Community Telegram Group](https:\u002F\u002Ft.me\u002F+QRzaRvTPIthlYWMx)💫.\n\n- For more technical writeups and details check out our [blog](https:\u002F\u002Fblog.ezkl.xyz\u002F).\n\n- To see what you can build with ezkl, check out [cryptoidol.tech](https:\u002F\u002Fcryptoidol.tech\u002F) where ezkl is used to create an AI that judges your singing ... forever.\n\n----------------------\n\n### Getting Started ⚙️\n\nThe easiest way to get started is to try out a notebook. \n\n#### Python\nInstall the python bindings by calling.\n\n```bash\npip install ezkl\n```\nOr for the GPU:\n\n```bash\npip install ezkl-gpu\n```\n\nGoogle Colab Example to learn how you can train a neural net and deploy an inference verifier onchain for use in other smart contracts. [![Notebook](https:\u002F\u002Fcolab.research.google.com\u002Fassets\u002Fcolab-badge.svg)](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fzkonduit\u002Fezkl\u002Fblob\u002Fmain\u002Fexamples\u002Fnotebooks\u002Fezkl_demo.ipynb) \n\n\nMore notebook tutorials can be found within `examples\u002Fnotebooks`.\n\n#### CLI\nInstall the CLI\n``` shell\ncurl https:\u002F\u002Fraw.githubusercontent.com\u002Fzkonduit\u002Fezkl\u002Fmain\u002Finstall_ezkl_cli.sh | bash\n```\n\nhttps:\u002F\u002Fuser-images.githubusercontent.com\u002F45801863\u002F236771676-5bbbbfd1-ba6f-418a-902e-20738ce0e9f0.mp4\n\nFor more details visit the [docs](https:\u002F\u002Fdocs.ezkl.xyz). The CLI is faster than Python, as it has less overhead. For even more speed and convenience, check out the [remote proving service](https:\u002F\u002Fei40vx5x6j0.typeform.com\u002Fto\u002FsFv1oxvb), which feels like the CLI but is backed by a tuned cluster.\n\nBuild the auto-generated rust documentation and open the docs in your browser locally. `cargo doc --open`\n\n### Building the Project 🔨\n\n#### Rust CLI\n\nYou can install the library from source\n\n```bash\ncargo install --locked --path .\n```\n\n`ezkl` now auto-manages solc installation for you.\n\n\n\n\n#### Building Python Bindings\nPython bindings exists and can be built using `maturin`. You will need `rust` and `cargo` to be installed.\n\n```bash\npython -m venv .env\nsource .env\u002Fbin\u002Factivate\npip install -r requirements.txt\nmaturin develop --release --features python-bindings\n# dependencies specific to tutorials\npip install torch pandas numpy seaborn jupyter onnx kaggle py-solc-x web3 librosa tensorflow keras tf2onnx\n```\n\n### GPU Acceleration\n\nIf you have access to NVIDIA GPUs, you can enable acceleration by building with the feature `icicle` and setting the following environment variable:\n\n```sh\nexport ENABLE_ICICLE_GPU=true\n```\n\nGPU acceleration is provided by [Icicle](https:\u002F\u002Fgithub.com\u002Fingonyama-zk\u002Ficicle)\n\nTo go back to running with CPU, the previous environment variable must be **unset** instead of being switch to a value of false:\n\n```sh\nunset ENABLE_ICICLE_GPU\n```\n\n**NOTE:** Even with the above environment variable set, icicle is disabled for circuits where k \u003C= 8. To change the value of `k` where icicle is enabled, you can set the environment variable `ICICLE_SMALL_K`.\n\n### Contributing 🌎\n\nIf you're interested in contributing and are unsure where to start, reach out to one of the maintainers:\n\n* dante (alexander-camuto)\n* jason (jasonmorton)\n\nMore broadly:\n\n- See currently open issues for ideas on how to contribute.\n\n- For PRs we use the [conventional commits](https:\u002F\u002Fwww.conventionalcommits.org\u002Fen\u002Fv1.0.0\u002F) naming convention.\n\n- To report bugs or request new features [create a new issue within Issues](https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fissues) to inform the greater community.\n\n\nAny contribution intentionally submitted for inclusion in the work by you shall be licensed to Zkonduit Inc. under the terms and conditions specified in the [CLA](https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fblob\u002Fmain\u002Fcla.md), which you agree to by intentionally submitting a contribution. In particular, you have the right to submit the contribution and we can distribute it, among other terms and conditions. \n\n\n### Audits & Security\n\n[v21.0.0](https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Freleases\u002Ftag\u002Fv21.0.0) has been audited by Trail of Bits, the report can be found [here](https:\u002F\u002Fgithub.com\u002Ftrailofbits\u002Fpublications\u002Fblob\u002Fmaster\u002Freviews\u002F2025-03-zkonduit-ezkl-securityreview.pdf).\n\n> NOTE: Because operations are quantized when they are converted from an onnx file to a zk-circuit, outputs in python and ezkl may differ slightly.\n\n\nCheck out `docs\u002Fadvanced_security` for more advanced information on potential threat vectors that are specific to zero-knowledge inference, quantization, and to machine learning models generally.\n\n\n### No Warranty\n\nTHE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\nFITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.\n\nCopyright (c) 2026 Zkonduit Inc. \n\n","\u003Ch1 align=\"center\">\n\t\u003Cbr>\n\t :thought_balloon:\n\t\u003Cbr>\n\t\u003Cbr>\nEZKL\n\t\u003Cbr>\n\t\u003Cbr>\n\t\u003Cbr>\n\u003C\u002Fh1>\n\n> Easy Zero-Knowledge Inference\n\n[![Test](https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fworkflows\u002FRust\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Factions?query=workflow%3ARust)\n\n`ezkl` 是一个库和命令行工具，用于在 zk-snark（零知识简洁非交互式论证，ZK-SNARK）环境中进行深度学习模型和其他计算图的推理（ZKML 零知识机器学习）。它支持以下工作流程：\n\n1. 定义计算图，例如神经网络（但实际上可以是任意一组操作），就像你在 pytorch 或 tensorflow 中通常所做的那样。\n2. 将最终的操作图导出为 [.onnx](https:\u002F\u002Fonnx.ai\u002F) 文件和一些样本输入到 `.json` 文件中。\n3. 将 `ezkl` 指向 `.onnx` 和 `.json` 文件以生成一个 ZK-SNARK 电路，你可以用它来证明诸如以下的陈述：\n\n> \"I ran this publicly available neural network on some private data and it produced this output\"\n\n> “我在一些私有数据上运行了这个公开可用的神经网络，并产生了这个输出”\n\n[![Notebook](https:\u002F\u002Fcolab.research.google.com\u002Fassets\u002Fcolab-badge.svg)](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fzkonduit\u002Fezkl\u002Fblob\u002Fmain\u002Fexamples\u002Fnotebooks\u002Fsimple_demo_public_network_output.ipynb) \n\n> \"I ran my private neural network on some public data and it produced this output\"\n\n> “我在一些公共数据上运行了我的私有神经网络，并产生了这个输出”\n\n[![Notebook](https:\u002F\u002Fcolab.research.google.com\u002Fassets\u002Fcolab-badge.svg)](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fzkonduit\u002Fezkl\u002Fblob\u002Fmain\u002Fexamples\u002Fnotebooks\u002Fsimple_demo_public_input_output.ipynb) \n\n> \"I correctly ran this publicly available neural network on some public data and it produced this output\"\n\n> “我正确地在一些公共数据上运行了这个公开可用的神经网络，并产生了这个输出”\n\n[![Notebook](https:\u002F\u002Fcolab.research.google.com\u002Fassets\u002Fcolab-badge.svg)](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fzkonduit\u002Fezkl\u002Fblob\u002Fmain\u002Fexamples\u002Fnotebooks\u002Fsimple_demo_all_public.ipynb) \n\n在后台，我们使用协作开发的 [Halo2](https:\u002F\u002Fgithub.com\u002Fprivacy-scaling-explorations\u002Fhalo2) 作为证明系统。\n\n生成的证明随后可以用更少的计算资源进行验证，包括链上（使用以太坊虚拟机 EVM）、在浏览器中，或在设备上。 \n\n- 如果您有任何问题，我们非常欢迎您前往 [Discussions](https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fdiscussions) 开启讨论话题。或者，您可以加入 ✨[EZKL 社区 Telegram 群组](https:\u002F\u002Ft.me\u002F+QRzaRvTPIthlYWMx)💫。\n\n- 欲了解更多技术文章和详情，请查看我们的 [博客](https:\u002F\u002Fblog.ezkl.xyz\u002F)。\n\n- 想了解用 ezkl 能构建什么，请查看 [cryptoidol.tech](https:\u002F\u002Fcryptoidol.tech\u002F)，ezkl 在那里被用来创建一个永远评判你唱歌水平的 AI。\n\n----------------------\n\n### Getting Started ⚙️\n\n最简单的方法是尝试一下笔记本。 \n\n#### Python\n通过调用以下命令安装 Python 绑定。\n\n```bash\npip install ezkl\n```\n或者针对 GPU：\n\n```bash\npip install ezkl-gpu\n```\n\nGoogle Colab 示例，学习如何训练神经网络并在链上部署推理验证器以供其他智能合约使用。[![Notebook](https:\u002F\u002Fcolab.research.google.com\u002Fassets\u002Fcolab-badge.svg)](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fzkonduit\u002Fezkl\u002Fblob\u002Fmain\u002Fexamples\u002Fnotebooks\u002Fezkl_demo.ipynb) \n\n\n更多笔记本教程可以在 `examples\u002Fnotebooks` 中找到。\n\n#### CLI\n安装 CLI\n``` shell\ncurl https:\u002F\u002Fraw.githubusercontent.com\u002Fzkonduit\u002Fezkl\u002Fmain\u002Finstall_ezkl_cli.sh | bash\n```\n\nhttps:\u002F\u002Fuser-images.githubusercontent.com\u002F45801863\u002F236771676-5bbbbfd1-ba6f-418a-902e-20738ce0e9f0.mp4\n\n更多详情请访问 [文档](https:\u002F\u002Fdocs.ezkl.xyz)。CLI 比 Python 更快，因为它的开销更少。为了获得更快的速度和便利性，请查看 [远程证明服务](https:\u002F\u002Fei40vx5x6j0.typeform.com\u002Fto\u002FsFv1oxvb)，它感觉像 CLI 但由调优后的集群提供支持。\n\n构建自动生成的 Rust 文档并在本地浏览器中打开文档。`cargo doc --open`\n\n### Building the Project 🔨\n\n#### Rust CLI\n\n您可以从源码安装库\n\n```bash\ncargo install --locked --path .\n```\n\n`ezkl` 现在会自动为您管理 solc 的安装。\n\n\n\n\n#### Building Python Bindings\nPython 绑定已存在，可以使用 `maturin` 构建。您需要安装 `rust` 和 `cargo`。\n\n```bash\npython -m venv .env\nsource .env\u002Fbin\u002Factivate\npip install -r requirements.txt\nmaturin develop --release --features python-bindings\n# dependencies specific to tutorials\npip install torch pandas numpy seaborn jupyter onnx kaggle py-solc-x web3 librosa tensorflow keras tf2onnx\n```\n\n### GPU Acceleration\n\n如果您可以访问 NVIDIA GPU，您可以通过使用 `icicle` 特性构建并设置以下环境变量来启用加速：\n\n```sh\nexport ENABLE_ICICLE_GPU=true\n```\n\nGPU 加速由 [Icicle](https:\u002F\u002Fgithub.com\u002Fingonyama-zk\u002Ficicle) 提供。\n\n若要回退到使用 CPU 运行，必须 **取消设置** 之前的环境变量，而不是将其切换为 false 值：\n\n```sh\nunset ENABLE_ICICLE_GPU\n```\n\n**NOTE:** 即使设置了上述环境变量，对于 k \u003C= 8 的电路，icicle 也会被禁用。要更改启用 icicle 的 `k` 值，您可以设置环境变量 `ICICLE_SMALL_K`。\n\n### Contributing 🌎\n\n如果您有兴趣贡献且不确定从哪里开始，请联系以下维护者之一：\n\n* dante (alexander-camuto)\n* jason (jasonmorton)\n\n更广泛地说：\n\n- 查看当前开放的问题以获取如何贡献的想法。\n\n- 对于 PR，我们使用 [约定式提交](https:\u002F\u002Fwww.conventionalcommits.org\u002Fen\u002Fv1.0.0\u002F) 命名约定。\n\n- 要报告错误或请求新功能，请在 [Issues](https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fissues) 内创建新问题以告知更广泛的社区。\n\n\n您有意提交纳入本作品的任何贡献，将根据 [CLA](https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fblob\u002Fmain\u002Fcla.md) 中规定的条款和条件授权给 Zkonduit Inc.，您通过有意提交贡献即表示同意。特别是，您有权提交贡献，我们可以分发它，以及其他条款和条件。 \n\n\n### Audits & Security\n\n[v21.0.0](https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Freleases\u002Ftag\u002Fv21.0.0) 已由 Trail of Bits 审计，报告可在 [此处](https:\u002F\u002Fgithub.com\u002Ftrailofbits\u002Fpublications\u002Fblob\u002Fmaster\u002Freviews\u002F2025-03-zkonduit-ezkl-securityreview.pdf) 找到。\n\n> NOTE: 由于操作在从 onnx 文件转换为 zk-circuit 时被量化，python 和 ezkl 的输出可能会有细微差别。\n\n\n查看 `docs\u002Fadvanced_security` 以获取更多关于特定于零知识推理、量化以及一般机器学习模型的潜在威胁向量的高级信息。\n\n### 无担保\n\n本软件按“原样”提供，不提供任何形式的明示或暗示担保，包括但不限于适销性、特定用途适用性及非侵权性担保。在任何情况下，作者或版权持有人均不对任何索赔、损害或其他责任负责，无论是因合同、侵权还是其他原因，或因使用或处理本软件而产生的。\n\n版权所有 (c) 2026 Zkonduit Inc.","# ezkl 快速上手指南\n\n## 环境准备\n\n- **操作系统**：Linux \u002F macOS \u002F Windows\n- **基础依赖**：\n  - [Rust](https:\u002F\u002Fwww.rust-lang.org\u002F) 及 `cargo`（构建 CLI 或源码所需）\n  - [Python](https:\u002F\u002Fwww.python.org\u002F) 3.x（使用 Python 绑定所需）\n- **硬件加速（可选）**：\n  - NVIDIA GPU（需安装对应驱动）\n  - 相关环境变量配置见“基本使用”部分\n\n## 安装步骤\n\n### 方式一：Python 库（推荐新手）\n直接通过 pip 安装 Python 绑定，适合快速集成到现有 Python 项目中。\n\n```bash\n# CPU 版本\npip install ezkl\n\n# GPU 版本（需 NVIDIA 显卡）\npip install ezkl-gpu\n```\n\n### 方式二：命令行工具（CLI）\n适合需要更高性能的场景，执行速度通常快于 Python 绑定。\n\n```bash\ncurl https:\u002F\u002Fraw.githubusercontent.com\u002Fzkonduit\u002Fezkl\u002Fmain\u002Finstall_ezkl_cli.sh | bash\n```\n\n### 方式三：从源码构建\n如需自定义功能或参与开发，可从源码编译。\n\n```bash\ncargo install --locked --path .\n```\n\n## 基本使用\n\nezkl 的核心工作流是将深度学习模型转换为零知识证明电路。\n\n1.  **定义模型**：使用 PyTorch 或 TensorFlow 定义计算图（神经网络）。\n2.  **导出文件**：将模型导出为 `.onnx` 文件，并将样本输入保存为 `.json` 文件。\n3.  **生成证明**：使用 ezkl 处理上述文件，生成 ZK-SNARK 电路以验证推理结果。\n\n**运行示例：**\n官方提供了详细的 Jupyter Notebook 教程，展示了如何训练神经网络并部署链上验证器。\n\n- [Google Colab 演示笔记](https:\u002F\u002Fcolab.research.google.com\u002Fgithub\u002Fzkonduit\u002Fezkl\u002Fblob\u002Fmain\u002Fexamples\u002Fnotebooks\u002Fsimple_demo_public_network_output.ipynb)\n- [更多示例](https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Ftree\u002Fmain\u002Fexamples\u002Fnotebooks)\n\n**GPU 加速配置（可选）：**\n若使用 NVIDIA GPU 并希望启用 Icicle 加速，请在运行前设置环境变量：\n\n```sh\nexport ENABLE_ICICLE_GPU=true\n```\n\n> 注意：当 k \u003C= 8 时，Icicle 默认禁用。如需调整阈值，可设置 `ICICLE_SMALL_K` 环境变量。\n\n---\n*详细文档请访问：[docs.ezkl.xyz](https:\u002F\u002Fdocs.ezkl.xyz)*","某去中心化借贷平台计划接入 AI 信用评估模型，但面临用户隐私保护与计算结果可信度的双重挑战。\n\n### 没有 ezkl 时\n- 用户必须将敏感银行流水上传至中心化服务器，存在数据泄露和滥用的高风险。\n- 平台无法向链上合约证明 AI 推理过程未被篡改，结果可信度完全依赖第三方背书。\n- 链下计算结果难以直接写入智能合约，需依赖高风险的中心化预言机节点中转。\n- 每次发生争议时需人工调取日志审计，流程繁琐且无法实现即时验证。\n\n### 使用 ezkl 后\n- 用户只需提交零知识证明，无需透露任何原始财务数据即可完成信用评分验证。\n- 智能合约能直接验证证明，确保 AI 模型运行逻辑正确且输出结果真实可靠。\n- 证明验证计算开销极低，支持在以太坊等公链上低成本实时确认，节省大量 Gas。\n- 通过 ezkl 导出 ONNX 文件自动生成电路，开发者无需精通密码学即可快速集成现有模型。\n\nezkl 通过零知识证明技术，在不泄露隐私的前提下实现了 AI 推理结果的链上可验证性。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fzkonduit_ezkl_dbf04c3d.png","zkonduit","Zkonduit","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fzkonduit_13244586.png","Making ezkl",null,"https:\u002F\u002Fgithub.com\u002Fzkonduit",[82,86,90,94],{"name":83,"color":84,"percentage":85},"Rust","#dea584",96.7,{"name":87,"color":88,"percentage":89},"Python","#3572A5",2.4,{"name":91,"color":92,"percentage":93},"Shell","#89e051",0.9,{"name":95,"color":96,"percentage":97},"Metal","#8f14e9",0,1190,200,"2026-04-03T03:29:43","未说明","可选，需 NVIDIA GPU 配合 Icicle 库，具体型号、显存及 CUDA 版本未说明",{"notes":104,"python":101,"dependencies":105},"构建 Python 绑定需安装 Rust 和 Cargo；CLI 模式性能优于 Python 绑定；支持通过环境变量 ENABLE_ICICLE_GPU 启用 GPU 加速；量化转换可能导致输出与原生模型略有差异；solc 由工具自动管理",[106,107,108,109,110,111,112,113,114,115],"torch","onnx","numpy","pandas","tensorflow","keras","web3","py-solc-x","librosa","jupyter",[13,15,14],[118,119,120,121],"ai","cryptography","zero-knowledge","zkml","2026-03-27T02:49:30.150509","2026-04-06T07:14:02.379936",[125,130,134,139,143,148],{"id":126,"question_zh":127,"answer_zh":128,"source_url":129},3415,"运行 `create-evm-verifier` 时报错 `Result::unwrap() on an 'Err' value: OddLength'` 是什么原因？","该错误通常是因为远程机器上未正确安装 `solc` (Solidity 编译器)。维护者在评论中指出，如果远程机器没有 `solc`，会导致操作系统找不到相关命令从而引发此类错误。","https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fissues\u002F128",{"id":131,"question_zh":132,"answer_zh":133,"source_url":129},3416,"生成 EVM 验证器合约前必须安装哪些依赖工具？","必须安装 `solc`。维护者确认了这是一个必需的依赖，并已提交 PR #135 来明确这一点。用户需要在本地和远程机器上都安装并更新 `solc` 才能成功运行 EVM 验证流程。",{"id":135,"question_zh":136,"answer_zh":137,"source_url":138},3417,"`ezkl setup` 进程运行一段时间后突然被系统杀死（Killed），该如何解决？","这通常是由于内存不足（OOM）导致的。用户反馈显示即使模型文件不大（约 17 MB），Setup 阶段仍可能因内存耗尽被杀。解决方法是增加机器的可用 RAM 内存，或者优化模型配置以减少内存占用。","https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fissues\u002F987",{"id":140,"question_zh":141,"answer_zh":142,"source_url":138},3418,"硬件资源有限时，如何调整 `ezkl` 配置以完成 Setup 过程？","可以通过修改 `settings.json` 中的 `logrows` 参数来降低资源需求。例如，`logrows` 设置为 25 可能需要超过 256GB 的内存。建议尝试降低 `logrows` 值，或者改用平均绝对误差（Mean Absolute Error）代替余弦相似度，并对数据进行更多下采样处理。",{"id":144,"question_zh":145,"answer_zh":146,"source_url":147},3419,"Python 中使用 `Gen_Settings` 函数时，为何会在特定 ONNX 节点（如 Mul, Squeeze）上分析失败？","这通常是因为模型内部存在不一致的形状（inconsistent shapes）或操作定义问题。维护者指出需要解析并修复这些节点以确保图结构的一致性。如果是特定模型（如 IIN），可能需要参考相关的 Tract 库 issue 来解决形状兼容性问题。","https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fissues\u002F411",{"id":149,"question_zh":150,"answer_zh":151,"source_url":147},3420,"模型规格中混合使用变量 Batch Size 和静态 Batch Size 会导致什么问题？","这会导致 `Gen_Settings` 或其他设置生成步骤失败。维护者指出，模型不应混合变量 `batch_size` 与指定的静态 `batch_size`（如 20）。这种规格不统一会被视为模型定义不当，导致无法生成有效的设置文件或证明。",[153,158,162,166,170,174,178,182,186,190,194,198,202,206,210,214,219,224,229,234],{"id":154,"version":155,"summary_zh":156,"released_at":157},103030,"v16.2.7","## What's Changed\r\n* fix: add version string and sed by @JSeam2 in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F893\r\n* fix: `get_slice` should not use intermediate `Vec` by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F894\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fcompare\u002Fv16.2.5...v16.2.7","2024-12-28T04:27:27",{"id":159,"version":160,"summary_zh":79,"released_at":161},103011,"v23.0.5","2026-02-20T19:34:09",{"id":163,"version":164,"summary_zh":79,"released_at":165},103012,"v23.0.3","2025-10-26T18:35:13",{"id":167,"version":168,"summary_zh":79,"released_at":169},103013,"v23.0.2","2025-10-26T03:23:49",{"id":171,"version":172,"summary_zh":79,"released_at":173},103014,"v22.3.0","2025-10-08T11:39:54",{"id":175,"version":176,"summary_zh":79,"released_at":177},103015,"v22.2.4","2025-10-07T18:37:55",{"id":179,"version":180,"summary_zh":79,"released_at":181},103016,"v22.2.1","2025-07-30T18:55:45",{"id":183,"version":184,"summary_zh":79,"released_at":185},103017,"v22.2.0","2025-07-27T20:06:28",{"id":187,"version":188,"summary_zh":79,"released_at":189},103018,"v22.0.1","2025-04-23T08:22:11",{"id":191,"version":192,"summary_zh":79,"released_at":193},103019,"v21.0.4","2025-04-06T18:11:18",{"id":195,"version":196,"summary_zh":79,"released_at":197},103020,"v21.0.3","2025-03-25T19:35:43",{"id":199,"version":200,"summary_zh":79,"released_at":201},103021,"v21.0.0","2025-03-19T10:52:58",{"id":203,"version":204,"summary_zh":79,"released_at":205},103022,"v20.2.7","2025-03-17T14:49:03",{"id":207,"version":208,"summary_zh":79,"released_at":209},103023,"v20.2.6","2025-03-14T16:16:08",{"id":211,"version":212,"summary_zh":79,"released_at":213},103024,"v20.2.4","2025-03-08T18:31:30",{"id":215,"version":216,"summary_zh":217,"released_at":218},103025,"v20.2.0","## What's Changed\r\n* ci: change to sha hashes by @JSeam2 in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F922\r\n* chore: add model input\u002Foutput types to settings by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F933\r\n* fix: broken links in polycommit.rs and poseidon.rs by @cypherpepe in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F932\r\n* docs: cat-dog example by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F929\r\n* chore: fix typos in comments and docs by @rebustron in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F934\r\n* refactor: enforce max decomp base\u002Flegs in args by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F936\r\n* feat: generalize conv mem layout and ND by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F935\r\n* refactor: rm tolerance parameter by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F937\r\n* feat: pass data directly in cli by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F939\r\n\r\n## New Contributors\r\n* @cypherpepe made their first contribution in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F932\r\n* @rebustron made their first contribution in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F934\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fcompare\u002Fv19.0.7...v20.2.0","2025-02-20T18:40:17",{"id":220,"version":221,"summary_zh":222,"released_at":223},103026,"v19.0.7","## What's Changed\r\n* fix: version interpolation in npm publishing by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F917\r\n* fix: make flushing err more informative by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F919\r\n* docs: improve quality + code quality fixes by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F920\r\n* fix: range-check graph input and outputs by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F921\r\n* fix: force prover det on argmax\u002Fmin for collisions by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F923\r\n* fix: use onnx convention when integer dividing by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F925\r\n* docs: advanced security notices by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F926\r\n* fix: use variable len domain for poseidon by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F927\r\n* refactor: serial lookup commits for metal by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F928\r\n\r\nBREAKING CHANGES: pk and vk are not backwards compatible due to argument changes, nor is settings file\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fcompare\u002Fv18.1.5...v19.0.7","2025-02-06T04:05:07",{"id":225,"version":226,"summary_zh":227,"released_at":228},103027,"v18.1.5","## What's Changed\r\n* fix: node parsing should not panic by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F912\r\n* fix: syn-sel should be range-checked when overflow by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F913\r\n* fix: strict cvx opt bounds to stop prover non-det by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F914\r\n* fix: patch pypi whl version labels by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F916\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fcompare\u002Fv18.1.1...v18.1.5","2025-01-28T01:26:15",{"id":230,"version":231,"summary_zh":232,"released_at":233},103028,"v18.1.1","## What's Changed\r\n* refactor: pregen mv-lookup blinds by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F900\r\n* feat: add gen-random-data helpers func by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F901\r\n* refactor: range check sanity toggled by CHECKMODE by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F902\r\n* fix: eager exec of `ok_or` error prints by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F903\r\n* fix: swift pm ci issue with updating testing files by @ElusAegis in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F905\r\n* fix: apply `zizmor` suggestions to CI by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F906\r\n* fix: update test files in CI pipeline by @ElusAegis in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F908\r\n* fix!: shuffle argument should include an incrementing index  by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F904\r\n* feat: metal acceleration for MSM solving by @ElusAegis in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F909\r\n* fix: rm macos metal bindings from python by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F911\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fcompare\u002Fv17.0.0...v18.1.1","2025-01-21T05:37:54",{"id":235,"version":236,"summary_zh":237,"released_at":238},103029,"v17.0.0","## What's Changed\r\n* refactor!: simplified decompose op by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F892\r\n* fix: clearer duplication functions by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F895\r\n* refactor: batched poly reads by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F897\r\n* chore: version mismatch warnings for artifacts by @alexander-camuto in https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fpull\u002F899\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Fzkonduit\u002Fezkl\u002Fcompare\u002Fv16.2.7...v17.0.0","2025-01-06T16:02:52"]