[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-lean-dojo--LeanCopilot":3,"tool-lean-dojo--LeanCopilot":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":76,"owner_avatar_url":77,"owner_bio":78,"owner_company":79,"owner_location":79,"owner_email":79,"owner_twitter":79,"owner_website":80,"owner_url":81,"languages":82,"stars":102,"forks":103,"last_commit_at":104,"license":105,"difficulty_score":106,"env_os":107,"env_gpu":108,"env_ram":109,"env_deps":110,"category_tags":121,"github_topics":122,"view_count":10,"oss_zip_url":79,"oss_zip_packed_at":79,"status":16,"created_at":130,"updated_at":131,"faqs":132,"releases":160},3542,"lean-dojo\u002FLeanCopilot","LeanCopilot","LLMs as Copilots for Theorem Proving in Lean","LeanCopilot 是一款专为 Lean 定理证明器设计的智能辅助工具，旨在将大型语言模型（LLM）深度集成到形式化验证的工作流中。它主要解决了传统定理证明过程中依赖人工手动推导、效率低下且门槛极高的问题，通过自动化建议证明策略（tactics）、筛选关键前提以及自动搜索完整证明路径，显著提升了数学定理和软件正确性验证的效率。\n\n该工具特别适合形式化方法研究人员、数学家以及从事高可靠性系统开发的工程师使用。无论是希望利用现有模型加速研究，还是想要部署自定义模型的开发者，都能从中受益。LeanCopilot 的独特亮点在于其高度的灵活性与原生集成能力：用户既可以直接调用来自 LeanDojo 的预训练模型，也能轻松接入自己在本地（支持 CPU 或 GPU）或云端运行的私有模型。此外，它跨平台支持 Linux、macOS 及 Windows，并提供了清晰的 API 接口，让大模型真正成为了定理证明过程中的得力“副驾驶”，在保持严谨逻辑的同时大幅降低了人机交互的认知负担。","Lean Copilot: LLMs as Copilots for Theorem Proving in Lean\n==========================================================\n\n🚩**News**: [Our paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2404.12534) is accepted to the International Conference on Neuro-symbolic Systems (NeuS), 2025. See you in Philadelphia!\n\nLean Copilot allows large language models (LLMs) to be used natively in Lean for proof automation, e.g., suggesting tactics\u002Fpremises and searching for proofs. You can use our built-in models from [LeanDojo](https:\u002F\u002Fleandojo.org\u002F) or bring your own models that run either locally (w\u002F or w\u002Fo GPUs) or on the cloud.\n\n\u003Chttps:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fassets\u002F114432581\u002Fee0f56f8-849e-4099-9284-d8092cbd22a3>\n\n## Table of Contents\n\n1. [Requirements](#requirements)  \n1. [Using Lean Copilot in Your Project](#using-lean-copilot-in-your-project)\n   1. [Adding Lean Copilot as a Dependency](#adding-lean-copilot-as-a-dependency)\n   1. [Getting Started with Lean Copilot](#getting-started-with-lean-copilot)\n      1. [Tactic Suggestion](#tactic-suggestion)\n      1. [Proof Search](#proof-search)\n      1. [Premise Selection](#premise-selection)\n1. [Advanced Usage](#advanced-usage)\n   1. [Tactic APIs](#tactic-apis)\n   1. [Model APIs](#model-apis)\n   1. [Bring Your Own Model](#bring-your-own-model)\n1. [Caveats](#caveats)\n1. [Getting in Touch](#getting-in-touch)\n1. [Acknowledgements](#acknowledgements)\n1. [Citation](#citation)\n\n## Requirements\n\n* Supported platforms: Linux, macOS, Windows and [Windows WSL](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fwindows\u002Fwsl\u002Finstall).\n* [Git LFS](https:\u002F\u002Fgit-lfs.com\u002F).\n* Optional (recommended if you have a [CUDA-enabled GPU](https:\u002F\u002Fdeveloper.nvidia.com\u002Fcuda-gpus)): CUDA and [cuDNN](https:\u002F\u002Fdeveloper.nvidia.com\u002Fcudnn).\n* Required for building Lean Copilot itself (rather than a downstream package): CMake >= 3.7 and a C++17 compatible compiler.\n\n## Using Lean Copilot in Your Project\n\n:warning: Your project must use a Lean version of at least `lean4:v4.3.0-rc2`.\n\n### Adding Lean Copilot as a Dependency\n\n1. Add the package configuration option `moreLinkArgs := #[\"-L.\u002F.lake\u002Fpackages\u002FLeanCopilot\u002F.lake\u002Fbuild\u002Flib\", \"-lctranslate2\"]` to lakefile.lean. For example,\n\n```lean\npackage «my-package» {\n  moreLinkArgs := #[\n    \"-L.\u002F.lake\u002Fpackages\u002FLeanCopilot\u002F.lake\u002Fbuild\u002Flib\",\n    \"-lctranslate2\"\n  ]\n}\n```\n\nAlternatively, if your project uses lakefile.toml, it should include:\n\n```toml\nmoreLinkArgs = [\"-L.\u002F.lake\u002Fpackages\u002FLeanCopilot\u002F.lake\u002Fbuild\u002Flib\", \"-lctranslate2\"]\n```\n\n2. Add the following line to lakefile.lean, including the quotation marks:\n\n```lean\nrequire LeanCopilot from git \"https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot.git\" @ \"LEAN_COPILOT_VERSION\"\n```\n\nFor stable Lean versions (e.g., `v4.28.0`), set `LEAN_COPILOT_VERSION` to be that version. For the latest unstable Lean versions (e.g., `v4.29.0-rc1`), set `LEAN_COPILOT_VERSION` to `main`. In either case, make sure the version is compatible with other dependencies such as mathlib. If your project uses lakefile.toml instead of lakefile.lean, it should include:\n\n```toml\n[[require]]\nname = \"LeanCopilot\"\ngit = \"https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot.git\"\nrev = \"LEAN_COPILOT_VERSION\"\n```\n\n3. If you are using native Windows, add `\u003Cpath_to_your_project>\u002F.lake\u002Fpackages\u002FLeanCopilot\u002F.lake\u002Fbuild\u002Flib` to your `Path` variable in Advanced System Settings > Environment Variables... > System variables. \n\n4. Run `lake update LeanCopilot`.\n\n5. Run `lake exe LeanCopilot\u002Fdownload` to download the built-in models from Hugging Face to `~\u002F.cache\u002Flean_copilot\u002F`. *Alternatively*, you can download the models from Hugging Face manually from\n\n* [ct2-leandojo-lean4-tacgen-byt5-small](https:\u002F\u002Fhuggingface.co\u002Fkaiyuy\u002Fct2-leandojo-lean4-tacgen-byt5-small)\n* [ct2-leandojo-lean4-retriever-byt5-small](https:\u002F\u002Fhuggingface.co\u002Fkaiyuy\u002Fct2-leandojo-lean4-retriever-byt5-small)\n* [premise-embeddings-leandojo-lean4-retriever-byt5-small](https:\u002F\u002Fhuggingface.co\u002Fkaiyuy\u002Fpremise-embeddings-leandojo-lean4-retriever-byt5-small)\n* [ct2-byt5-small](https:\u002F\u002Fhuggingface.co\u002Fkaiyuy\u002Fct2-byt5-small)\n\n6. Run `lake build`.\n\n[Here](https:\u002F\u002Fgithub.com\u002Fyangky11\u002Flean4-example\u002Fblob\u002FLeanCopilot-demo) is an example of a Lean package depending on Lean Copilot. If you have problems building the project, our [Dockerfile](.\u002FDockerfile), [build.sh](scripts\u002Fbuild.sh) or [build_example.sh](scripts\u002Fbuild_example.sh) may be helpful.\n\n### Getting Started with Lean Copilot\n\n#### Tactic Suggestion\n\nAfter `import LeanCopilot`, you can use the tactic `suggest_tactics` to generate tactic suggestions. You can click on any of the suggested tactics to use it in the proof.\n\n\u003Cimg width=\"977\" alt=\"suggest_tactics\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flean-dojo_LeanCopilot_readme_56e1b86bc57b.png\">\n\nYou can provide a prefix (e.g., `simp`) to constrain the generated tactics:\n\n\u003Cimg width=\"915\" alt=\"suggest_tactics_simp\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flean-dojo_LeanCopilot_readme_527b265ac454.png\">\n\n#### Proof Search\n\nThe tactic `search_proof` combines LLM-generated tactics with [aesop](https:\u002F\u002Fgithub.com\u002Fleanprover-community\u002Faesop) to search for multi-tactic proofs. When a proof is found, you can click on it to insert it into the editor.\n\n\u003Cimg width=\"824\" alt=\"search_proof\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flean-dojo_LeanCopilot_readme_beb412f6132f.png\">\n\n#### Premise Selection\n\nThe `select_premises` tactic retrieves a list of potentially useful premises. Currently, it uses the retriever in [LeanDojo](https:\u002F\u002Fleandojo.org\u002F) to select premises from a fixed snapshot of Lean and [mathlib4](https:\u002F\u002Fgithub.com\u002Fleanprover-community\u002Fmathlib4\u002Ftree\u002F3ce43c18f614b76e161f911b75a3e1ef641620ff).\n\n![select_premises](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flean-dojo_LeanCopilot_readme_7e31cd29c656.png)\n\n#### Running LLMs\n\nYou can also run the inference of any LLMs in Lean, which can be used to build customized proof automation or other LLM-based applications (not limited to theorem proving). It's possible to run arbitrary models either locally or remotely (see [Bring Your Own Model](#bring-your-own-model)).\n\n\u003Cimg width=\"1123\" alt=\"run_llms\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flean-dojo_LeanCopilot_readme_29c8812aa433.png\">\n\n## Advanced Usage\n\n**This section is only for advanced users who would like to change the default behavior of `suggest_tactics`, `search_proof`, or `select_premises`, e.g., to use different models or hyperparameters.**\n\n### Tactic APIs\n\n* Examples in [TacticSuggestion.lean](LeanCopilotTests\u002FTacticSuggestion.lean) showcase how to configure `suggest_tactics`, e.g., to use different models or generate different numbers of tactics.\n* Examples in [ProofSearch.lean](LeanCopilotTests\u002FProofSearch.lean) showcase how to configure `search_proof` using options provided by [aesop](https:\u002F\u002Fgithub.com\u002Fleanprover-community\u002Faesop).\n* Examples in [PremiseSelection.lean](LeanCopilotTests\u002FPremiseSelection.lean) showcase how to set the number of retrieved premises for `select_premises`.\n\n### Model APIs\n\n**Examples in [ModelAPIs.lean](LeanCopilotTests\u002FModelAPIs.lean) showcase how to run the inference of different models and configure their parameters (temperature, beam size, etc.).**\n\nLean Copilot supports two kinds of models: generators and encoders. Generators must implement the `TextToText` interface:\n\n```lean\nclass TextToText (τ : Type) where\n  generate (model : τ) (input : String) (targetPrefix : String) : IO $ Array (String × Float)\n```\n\n* `input` is the input string\n* `targetPrefix` is used to constrain the generator's output. `\"\"` means no constraint.\n* `generate` should return an array of `String × Float`. Each `String` is an output from the model, and `Float` is the corresponding score.\n\nWe provide three types of Generators:\n\n* [`NativeGenerator`](LeanCopilot\u002FModels\u002FNative.lean) runs locally powered by [CTranslate2](https:\u002F\u002Fgithub.com\u002FOpenNMT\u002FCTranslate2) and is linked to Lean using Foreign Function Interface (FFI).\n* [`ExternalGenerator`](LeanCopilot\u002FModels\u002FExternal.lean) is hosted either locally or remotely. See [Bring Your Own Model](#bring-your-own-model) for details.\n* [`GenericGenerator`](LeanCopilot\u002FModels\u002FGeneric.lean) can be anything that implements the `generate` function in the `TextToText` typeclass.\n\nEncoders must implement `TextToVec`:\n\n```lean\nclass TextToVec (τ : Type) where\n  encode : τ → String → IO FloatArray\n```\n\n* `input` is the input string\n* `encode` should return a vector embedding produced by the model.\n\nSimilar to generators, we have `NativeEncoder`, `ExternalEncoder`, and `GenericEncoder`.\n\n### Bring Your Own Model\n\nIn principle, it is possible to run any model using Lean Copilot through `ExternalGenerator` or `ExternalEncoder` (examples in [ModelAPIs.lean](LeanCopilotTests\u002FModelAPIs.lean)). To use a model, you need to wrap it properly to expose the APIs in [external_model_api.yaml](.\u002Fexternal_model_api.yaml). As an example, we provide a [Python API server](.\u002Fpython) and use it to run a few models.\n\n## Caveats\n\n* `select_premises` always retrieves the original form of a premise. For example, `Nat.add_left_comm` is a result of the theorem below. In this case, `select_premises` retrieves `Nat.mul_left_comm` instead of `Nat.add_left_comm`.\n\n```lean\n@[to_additive]\ntheorem mul_left_comm : ∀ a b c : G, a * (b * c) = b * (a * c)\n```\n\n* In some cases, `search_proof` produces an erroneous proof with error messages like `fail to show termination for ...`. A temporary workaround is changing the theorem's name before applying `search_proof`. You can change it back after `search_proof` completes.\n\n## Getting in Touch\n\n* For general questions and discussions, please use [GitHub Discussions](https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fdiscussions).  \n* To report a potential bug, please open an issue. In the issue, please include your OS information, the exact steps to reproduce the error on **the latest stable version of Lean Copilot**, and complete logs preferrably in debug mode. **Important: If your issue cannot be reproduced easily, it will be unlikely to receive help.**\n* Feature requests and contributions are warmly welcome. Please feel free to start a [discussion](https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fdiscussions) or open a [pull request](https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpulls).\n\n## Acknowledgements\n\n* We thank Scott Morrison for suggestions on simplifying Lean Copilot's installation and Mac Malone for helping implement it. Both Scott and Mac work for the [Lean FRO](https:\u002F\u002Flean-fro.org\u002F).\n* We thank Jannis Limperg for supporting our LLM-generated tactics in Aesop (\u003Chttps:\u002F\u002Fgithub.com\u002Fleanprover-community\u002Faesop\u002Fpull\u002F70>).\n\n## Citation\n\nIf you find our work useful, please consider citing [our paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2404.12534):\n\n```BibTeX\n@article{song2024lean,\n  title={Lean copilot: Large language models as copilots for theorem proving in lean},\n  author={Song, Peiyang and Yang, Kaiyu and Anandkumar, Anima},\n  journal={arXiv preprint arXiv:2404.12534},\n  year={2024}\n}\n```\n","Lean Copilot：在 Lean 中作为定理证明助手的大型语言模型\n==========================================================\n\n🚩**新闻**：[我们的论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2404.12534)已被 2025 年神经符号系统国际会议（NeuS）接收。费城见！\n\nLean Copilot 允许大型语言模型（LLMs）原生地集成到 Lean 中，用于证明自动化，例如建议策略\u002F前提以及搜索证明。您可以使用来自 [LeanDojo](https:\u002F\u002Fleandojo.org\u002F) 的内置模型，也可以引入您自己的模型，这些模型可以在本地运行（无论是否有 GPU）或在云端运行。\n\n\u003Chttps:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fassets\u002F114432581\u002Fee0f56f8-849e-4099-9284-d8092cbd22a3>\n\n## 目录\n\n1. [要求](#requirements)  \n1. [在您的项目中使用 Lean Copilot](#using-lean-copilot-in-your-project)\n   1. [将 Lean Copilot 添加为依赖项](#adding-lean-copilot-as-a-dependency)\n   1. [开始使用 Lean Copilot](#getting-started-with-lean-copilot)\n      1. [策略建议](#tactic-suggestion)\n      1. [证明搜索](#proof-search)\n      1. [前提选择](#premise-selection)\n1. [高级用法](#advanced-usage)\n   1. [策略 API](#tactic-apis)\n   1. [模型 API](#model-apis)\n   1. [引入您自己的模型](#bring-your-own-model)\n1. [注意事项](#caveats)\n1. [联系我们](#getting-in-touch)\n1. [致谢](#acknowledgements)\n1. [引用](#citation)\n\n## 要求\n\n* 支持的平台：Linux、macOS、Windows 和 [Windows WSL](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fwindows\u002Fwsl\u002Finstall)。\n* [Git LFS](https:\u002F\u002Fgit-lfs.com\u002F)。\n* 可选（如果您拥有支持 CUDA 的 GPU，建议安装）：CUDA 和 [cuDNN](https:\u002F\u002Fdeveloper.nvidia.com\u002Fcudnn)。\n* 构建 Lean Copilot 本身（而非下游包）所需的条件：CMake ≥ 3.7 以及兼容 C++17 的编译器。\n\n## 在您的项目中使用 Lean Copilot\n\n:warning: 您的项目必须使用至少 `lean4:v4.3.0-rc2` 版本的 Lean。\n\n### 将 Lean Copilot 添加为依赖项\n\n1. 在 lakefile.lean 中添加软件包配置选项 `moreLinkArgs := #[\"-L.\u002F.lake\u002Fpackages\u002FLeanCopilot\u002F.lake\u002Fbuild\u002Flib\", \"-lctranslate2\"]`。例如：\n\n```lean\npackage «my-package» {\n  moreLinkArgs := #[\n    \"-L.\u002F.lake\u002Fpackages\u002FLeanCopilot\u002F.lake\u002Fbuild\u002Flib\",\n    \"-lctranslate2\"\n  ]\n}\n```\n\n或者，如果您的项目使用 lakefile.toml，则应包含以下内容：\n\n```toml\nmoreLinkArgs = [\"-L.\u002F.lake\u002Fpackages\u002FLeanCopilot\u002F.lake\u002Fbuild\u002Flib\", \"-lctranslate2\"]\n```\n\n2. 在 lakefile.lean 中添加以下一行，包括引号：\n\n```lean\nrequire LeanCopilot from git \"https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot.git\" @ \"LEAN_COPILOT_VERSION\"\n```\n\n对于稳定的 Lean 版本（例如 `v4.28.0`），将 `LEAN_COPILOT_VERSION` 设置为该版本。对于最新的不稳定版本（例如 `v4.29.0-rc1`），将 `LEAN_COPILOT_VERSION` 设置为 `main`。无论哪种情况，都请确保版本与其他依赖项（如 mathlib）兼容。如果您的项目使用 lakefile.toml 而不是 lakefile.lean，则应包含：\n\n```toml\n[[require]]\nname = \"LeanCopilot\"\ngit = \"https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot.git\"\nrev = \"LEAN_COPILOT_VERSION\"\n```\n\n3. 如果您使用的是原生 Windows，请将 `\u003Cpath_to_your_project>\u002F.lake\u002Fpackages\u002FLeanCopilot\u002F.lake\u002Fbuild\u002Flib` 添加到“高级系统设置”→“环境变量”→“系统变量”中的 `Path` 变量中。\n\n4. 运行 `lake update LeanCopilot`。\n\n5. 运行 `lake exe LeanCopilot\u002Fdownload`，以从 Hugging Face 下载内置模型到 `~\u002F.cache\u002Flean_copilot\u002F`。*或者*，您也可以手动从以下链接下载模型：\n\n* [ct2-leandojo-lean4-tacgen-byt5-small](https:\u002F\u002Fhuggingface.co\u002Fkaiyuy\u002Fct2-leandojo-lean4-tacgen-byt5-small)\n* [ct2-leandojo-lean4-retriever-byt5-small](https:\u002F\u002Fhuggingface.co\u002Fkaiyuy\u002Fct2-leandojo-lean4-retriever-byt5-small)\n* [premise-embeddings-leandojo-lean4-retriever-byt5-small](https:\u002F\u002Fhuggingface.co\u002Fkaiyuy\u002Fpremise-embeddings-leandojo-lean4-retriever-byt5-small)\n* [ct2-byt5-small](https:\u002F\u002Fhuggingface.co\u002Fkaiyuy\u002Fct2-byt5-small)\n\n6. 运行 `lake build`。\n\n[这里](https:\u002F\u002Fgithub.com\u002Fyangky11\u002Flean4-example\u002Fblob\u002FLeanCopilot-demo)是一个依赖 Lean Copilot 的 Lean 包示例。如果您在构建项目时遇到问题，我们的 [Dockerfile](.\u002FDockerfile)、[build.sh](scripts\u002Fbuild.sh) 或 [build_example.sh](scripts\u002Fbuild_example.sh) 可能会有所帮助。\n\n### 开始使用 Lean Copilot\n\n#### 策略建议\n\n在 `import LeanCopilot` 之后，您可以使用策略 `suggest_tactics` 来生成策略建议。您可以点击任何一条建议的策略将其应用于证明中。\n\n\u003Cimg width=\"977\" alt=\"suggest_tactics\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flean-dojo_LeanCopilot_readme_56e1b86bc57b.png\">\n\n您还可以提供一个前缀（例如 `simp`），以限制生成的策略范围：\n\n\u003Cimg width=\"915\" alt=\"suggest_tactics_simp\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flean-dojo_LeanCopilot_readme_527b265ac454.png\">\n\n#### 证明搜索\n\n策略 `search_proof` 将 LLM 生成的策略与 [aesop](https:\u002F\u002Fgithub.com\u002Fleanprover-community\u002Faesop) 结合起来，搜索多步策略证明。当找到证明时，您可以点击它将其插入编辑器中。\n\n\u003Cimg width=\"824\" alt=\"search_proof\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flean-dojo_LeanCopilot_readme_beb412f6132f.png\">\n\n#### 前提选择\n\n策略 `select_premises` 会检索一份可能有用的前提列表。目前，它使用 [LeanDojo](https:\u002F\u002Fleandojo.org\u002F) 中的检索器，从 Lean 和 [mathlib4](https:\u002F\u002Fgithub.com\u002Fleanprover-community\u002Fmathlib4\u002Ftree\u002F3ce43c18f614b76e161f911b75a3e1ef641620ff) 的固定快照中选择前提。\n\n![select_premises](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flean-dojo_LeanCopilot_readme_7e31cd29c656.png)\n\n#### 运行 LLMs\n\n您还可以在 Lean 中运行任何 LLM 的推理，这可用于构建自定义的证明自动化或其他基于 LLM 的应用（不仅限于定理证明）。您可以选择在本地或远程运行任意模型（详见 [引入您自己的模型](#bring-your-own-model)）。\n\n\u003Cimg width=\"1123\" alt=\"run_llms\" src=\"https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flean-dojo_LeanCopilot_readme_29c8812aa433.png\">\n\n## 高级用法\n\n**本节仅适用于希望更改 `suggest_tactics`、`search_proof` 或 `select_premises` 默认行为的高级用户，例如使用不同的模型或超参数。**\n\n### 策略 API\n\n* [TacticSuggestion.lean](LeanCopilotTests\u002FTacticSuggestion.lean) 中的示例展示了如何配置 `suggest_tactics`，例如使用不同的模型或生成不同数量的策略。\n* [ProofSearch.lean](LeanCopilotTests\u002FProofSearch.lean) 中的示例展示了如何使用 [aesop](https:\u002F\u002Fgithub.com\u002Fleanprover-community\u002Faesop) 提供的选项来配置 `search_proof`。\n* [PremiseSelection.lean](LeanCopilotTests\u002FPremiseSelection.lean) 中的示例展示了如何设置 `select_premises` 的前提条件检索数量。\n\n### 模型 API\n\n**[ModelAPIs.lean](LeanCopilotTests\u002FModelAPIs.lean) 中的示例展示了如何运行不同模型的推理，并配置其参数（温度、束宽等）。**\n\nLean Copilot 支持两种类型的模型：生成器和编码器。生成器必须实现 `TextToText` 接口：\n\n```lean\nclass TextToText (τ : Type) where\n  generate (model : τ) (input : String) (targetPrefix : String) : IO $ Array (String × Float)\n```\n\n* `input` 是输入字符串。\n* `targetPrefix` 用于约束生成器的输出。`\"\"` 表示无约束。\n* `generate` 应返回一个 `String × Float` 数组。每个 `String` 是模型的输出，`Float` 是对应的得分。\n\n我们提供了三种类型的生成器：\n\n* [`NativeGenerator`](LeanCopilot\u002FModels\u002FNative.lean) 在本地运行，由 [CTranslate2](https:\u002F\u002Fgithub.com\u002FOpenNMT\u002FCTranslate2) 提供支持，并通过外部函数接口 (FFI) 与 Lean 链接。\n* [`ExternalGenerator`](LeanCopilot\u002FModels\u002FExternal.lean) 可以托管在本地或远程。详情请参阅“自带模型”部分。\n* [`GenericGenerator`](LeanCopilot\u002FModels\u002FGeneric.lean) 可以是任何实现了 `TextToText` 类型类中 `generate` 函数的对象。\n\n编码器必须实现 `TextToVec`：\n\n```lean\nclass TextToVec (τ : Type) where\n  encode : τ → String → IO FloatArray\n```\n\n* `input` 是输入字符串。\n* `encode` 应返回由模型生成的向量嵌入。\n\n与生成器类似，我们也提供 `NativeEncoder`、`ExternalEncoder` 和 `GenericEncoder`。\n\n### 自带模型\n\n原则上，可以通过 `ExternalGenerator` 或 `ExternalEncoder` 使用任何模型运行 Lean Copilot（示例见 [ModelAPIs.lean](LeanCopilotTests\u002FModelAPIs.lean)）。要使用某个模型，需要对其进行适当的封装，以暴露 [external_model_api.yaml](.\u002Fexternal_model_api.yaml) 中定义的 API。作为示例，我们提供了一个 [Python API 服务器](.\u002Fpython)，并用它来运行几个模型。\n\n## 注意事项\n\n* `select_premises` 始终会检索前提的原始形式。例如，`Nat.add_left_comm` 是下面定理的结果。在这种情况下，`select_premises` 会检索 `Nat.mul_left_comm` 而不是 `Nat.add_left_comm`。\n\n```lean\n@[to_additive]\ntheorem mul_left_comm : ∀ a b c : G, a * (b * c) = b * (a * c)\n```\n\n* 在某些情况下，`search_proof` 会生成包含错误信息（如 `fail to show termination for ...`）的错误证明。临时解决方法是在应用 `search_proof` 之前更改定理名称。待 `search_proof` 完成后，可以再将其改回原名。\n\n## 联系我们\n\n* 如有任何一般性问题或讨论，请使用 [GitHub Discussions](https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fdiscussions)。\n* 如需报告潜在的 bug，请提交 issue。请在 issue 中提供您的操作系统信息、在 **Lean Copilot 最新稳定版本** 上复现该错误的详细步骤，以及完整的日志（最好为调试模式）。**重要提示：如果您的问题无法轻松复现，我们将很难为您提供帮助。**\n* 我们热烈欢迎功能请求和贡献。请随时发起 [讨论](https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fdiscussions) 或提交 [拉取请求](https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpulls)。\n\n## 致谢\n\n* 我们感谢 Scott Morrison 对简化 Lean Copilot 安装流程提出的建议，以及 Mac Malone 在实施过程中的帮助。Scott 和 Mac 均就职于 [Lean FRO](https:\u002F\u002Flean-fro.org\u002F)。\n* 我们感谢 Jannis Limperg 在 Aesop 中支持我们基于 LLM 的策略（\u003Chttps:\u002F\u002Fgithub.com\u002Fleanprover-community\u002Faesop\u002Fpull\u002F70>）。\n\n## 引用\n\n如果您认为我们的工作有所帮助，请考虑引用我们的论文：\n\n```BibTeX\n@article{song2024lean,\n  title={Lean copilot: Large language models as copilots for theorem proving in lean},\n  author={Song, Peiyang and Yang, Kaiyu and Anandkumar, Anima},\n  journal={arXiv preprint arXiv:2404.12534},\n  year={2024}\n}\n```","# LeanCopilot 快速上手指南\n\nLeanCopilot 是一个允许在 Lean 证明助手中原生使用大语言模型（LLM）的工具，支持战术建议、自动证明搜索和前提选择等功能。\n\n## 环境准备\n\n### 系统要求\n*   **操作系统**：Linux, macOS, Windows 或 [Windows WSL](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fwindows\u002Fwsl\u002Finstall)。\n*   **Lean 版本**：项目必须使用 `lean4:v4.3.0-rc2` 或更高版本。\n\n### 前置依赖\n*   **Git LFS**：必须安装 [Git LFS](https:\u002F\u002Fgit-lfs.com\u002F) 以拉取模型文件。\n*   **编译工具**（仅当需要从头构建 LeanCopilot 时）：CMake >= 3.7 和支持 C++17 的编译器。\n*   **GPU 加速**（可选但推荐）：如果拥有 [CUDA -enabled GPU](https:\u002F\u002Fdeveloper.nvidia.com\u002Fcuda-gpus)，建议安装 CUDA 和 [cuDNN](https:\u002F\u002Fdeveloper.nvidia.com\u002Fcudnn) 以提升推理速度。\n\n## 安装步骤\n\n### 1. 配置项目依赖\n在你的 Lean 项目根目录下，编辑 `lakefile.lean`（或 `lakefile.toml`）。\n\n**对于 `lakefile.lean`：**\n添加链接参数并引入 LeanCopilot 依赖：\n\n```lean\npackage «my-package» {\n  moreLinkArgs := #[\n    \"-L.\u002F.lake\u002Fpackages\u002FLeanCopilot\u002F.lake\u002Fbuild\u002Flib\",\n    \"-lctranslate2\"\n  ]\n}\n\nrequire LeanCopilot from git \"https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot.git\" @ \"LEAN_COPILOT_VERSION\"\n```\n\n> **注意**：将 `LEAN_COPILOT_VERSION` 替换为与你当前 Lean 版本兼容的标签。\n> *   稳定版 Lean (如 `v4.28.0`)：使用对应的版本号。\n> *   最新开发版 Lean (如 `v4.29.0-rc1`)：使用 `main`。\n\n**对于 `lakefile.toml`：**\n\n```toml\nmoreLinkArgs = [\"-L.\u002F.lake\u002Fpackages\u002FLeanCopilot\u002F.lake\u002Fbuild\u002Flib\", \"-lctranslate2\"]\n\n[[require]]\nname = \"LeanCopilot\"\ngit = \"https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot.git\"\nrev = \"LEAN_COPILOT_VERSION\"\n```\n\n### 2. Windows 用户额外步骤\n如果你使用的是原生 Windows（非 WSL），需要将以下路径添加到系统的 `Path` 环境变量中：\n`\u003C你的项目路径>\u002F.lake\u002Fpackages\u002FLeanCopilot\u002F.lake\u002Fbuild\u002Flib`\n\n### 3. 更新与构建\n在项目根目录终端执行以下命令：\n\n```bash\n# 更新依赖\nlake update LeanCopilot\n\n# 下载内置模型 (默认下载至 ~\u002F.cache\u002Flean_copilot\u002F)\n# 如果下载速度慢，可手动从 HuggingFace 下载后放入对应目录\nlake exe LeanCopilot\u002Fdownload\n\n# 构建项目\nlake build\n```\n\n> **提示**：若 `lake exe LeanCopilot\u002Fdownload` 连接 HuggingFace 超时，可手动访问以下链接下载模型文件并放置于 `~\u002F.cache\u002Flean_copilot\u002F`：\n> *   [ct2-leandojo-lean4-tacgen-byt5-small](https:\u002F\u002Fhuggingface.co\u002Fkaiyuy\u002Fct2-leandojo-lean4-tacgen-byt5-small)\n> *   [ct2-leandojo-lean4-retriever-byt5-small](https:\u002F\u002Fhuggingface.co\u002Fkaiyuy\u002Fct2-leandojo-lean4-retriever-byt5-small)\n> *   [premise-embeddings-leandojo-lean4-retriever-byt5-small](https:\u002F\u002Fhuggingface.co\u002Fkaiyuy\u002Fpremise-embeddings-leandojo-lean4-retriever-byt5-small)\n\n## 基本使用\n\n在您的 `.lean` 文件顶部导入库：\n\n```lean\nimport LeanCopilot\n```\n\n### 1. 战术建议 (Tactic Suggestion)\n使用 `suggest_tactics` 让 LLM 生成下一步可能的战术。点击编辑器中生成的建议即可直接应用。\n\n```lean\nexample : ∀ n : Nat, n + 0 = n := by\n  intro n\n  suggest_tactics -- 生成战术建议\n  -- 点击建议中的 `simp` 或其他战术继续证明\n```\n\n你也可以指定前缀来限制生成的战术类型：\n```lean\n  suggest_tactics \"simp\" -- 仅生成以 simp 开头的战术\n```\n\n### 2. 自动证明搜索 (Proof Search)\n使用 `search_proof` 结合 LLM 生成的战术和 `aesop` 进行多步证明搜索。找到证明后可一键插入。\n\n```lean\nexample : ∀ n m : Nat, n + m = m + n := by\n  search_proof -- 自动搜索完整证明路径\n```\n\n### 3. 前提选择 (Premise Selection)\n使用 `select_premises` 检索当前上下文中可能有用到的引理或前提。\n\n```lean\nexample : some_theorem := by\n  select_premises -- 列出相关前提供选择\n  -- 选择后自动引入到上下文\n  sorry\n```\n\n### 4. 运行自定义模型\nLeanCopilot 支持在 Lean 内部直接运行任意 LLM 推理（本地或远程），可用于构建自定义自动化流程。\n\n```lean\n-- 示例：运行模型推理\n#eval LeanCopilot.Models.NativeGenerator.run \"input text\"\n```","一位形式化验证工程师正在使用 Lean 4 将复杂的代数几何定理转化为机器可检查的代码，但在证明过程中陷入了繁琐的战术（tactic）选择困境。\n\n### 没有 LeanCopilot 时\n- **盲目试错效率低**：面对数百种可能的证明战术，工程师只能凭经验手动尝试，经常花费数小时在死胡同里反复调试。\n- **前提检索困难**：在庞大的 mathlib 库中，难以快速定位当前证明步骤所需的关键引理或前置条件，容易遗漏重要线索。\n- **思维中断频繁**：由于缺乏即时反馈，开发者需要不断切换上下文去查阅文档或搜索社区案例，导致逻辑推导过程频繁被打断。\n- **本地模型缺失**：若想利用 AI 辅助，往往依赖云端 API，存在数据隐私顾虑且受网络延迟影响，无法在离线环境下流畅工作。\n\n### 使用 LeanCopilot 后\n- **智能战术推荐**：LeanCopilot 直接在编辑器中预测并推荐最可能的下一步战术，将原本数小时的试错缩短为几分钟的确认过程。\n- **精准前提定位**：内置的检索模型能自动分析当前目标，从海量库中精准筛选出相关引理，显著降低了查找门槛。\n- **自动化证明搜索**：对于常规子目标，LeanCopilot 可自动执行搜索策略生成完整证明片段，让工程师专注于核心逻辑架构而非琐碎细节。\n- **灵活本地部署**：支持加载本地模型或利用 GPU 加速，既保障了代码隐私安全，又实现了零延迟的实时交互体验。\n\nLeanCopilot 通过将大语言模型深度集成到 Lean 工作流中，把形式化证明从“手工编织”升级为\"AI 协同驾驶”，极大提升了数学定理机器验证的生产力。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Flean-dojo_LeanCopilot_7e31cd29.png","lean-dojo","LeanDojo","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Flean-dojo_35b749b4.jpg","Machine Learning for Theorem Proving in Lean",null,"https:\u002F\u002Fleandojo.org\u002F","https:\u002F\u002Fgithub.com\u002Flean-dojo",[83,87,90,94,98],{"name":84,"color":85,"percentage":86},"C++","#f34b7d",91.8,{"name":88,"color":79,"percentage":89},"Lean",5.1,{"name":91,"color":92,"percentage":93},"Python","#3572A5",2.9,{"name":95,"color":96,"percentage":97},"Shell","#89e051",0.1,{"name":99,"color":100,"percentage":101},"Dockerfile","#384d54",0,1258,123,"2026-04-04T19:10:23","MIT",4,"Linux, macOS, Windows, Windows WSL","非必需。若使用本地加速推理，推荐配备支持 CUDA 的 NVIDIA GPU 及 cuDNN（具体型号和显存未说明）。","未说明",{"notes":111,"python":112,"dependencies":113},"1. 该项目是 Lean4 的插件，需通过 Lake 包管理器安装。2. 若在原生 Windows 环境下运行，需手动将构建生成的 lib 目录添加到系统环境变量 Path 中。3. 首次使用需运行命令下载内置模型（或从 Hugging Face 手动下载），模型基于 CTranslate2 格式。4. 若需自行编译 Lean Copilot 本身（而非仅作为依赖使用），必须安装 CMake >= 3.7 和 C++17 编译器。5. 支持通过外部 API 接入自定义模型（示例代码使用 Python 编写服务端）。","未说明（项目核心为 Lean4\u002FC++，仅自带的外部模型示例服务使用 Python）",[114,115,116,117,118,119,120],"Lean4 (>= v4.3.0-rc2)","Git LFS","CMake (>= 3.7)","C++17 兼容编译器","CTranslate2","CUDA (可选，用于 GPU 加速)","cuDNN (可选，用于 GPU 加速)",[26,51,13],[123,124,125,126,127,128,129],"lean","llm-inference","machine-learning","theorem-proving","lean4","formal-mathematics","llm","2026-03-27T02:49:30.150509","2026-04-06T08:08:34.434639",[133,138,143,148,152,156],{"id":134,"question_zh":135,"answer_zh":136,"source_url":137},16228,"为什么运行 `lake update` 后会出现 Lean 版本不匹配或构建错误？","这是因为 `lakefile.lean` 中的依赖项检出标志如果设置为最新提交，当新的 Lean 版本发布时，运行 `lake update` 会将依赖项更新到最新版本，导致与 Lean Copilot 所需的 Lean 版本不匹配。解决方案是使用 Lean Copilot 的最新稳定版本（如 v1.5.0），该版本已修复此问题，会始终检出与发布版本对应的特定提交，确保 Lean 版本一致。请避免随意运行 `lake update`，除非明确需要更新依赖。","https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fissues\u002F75",{"id":139,"question_zh":140,"answer_zh":141,"source_url":142},16229,"在 GitHub Codespace 中构建 LeanInfer 时遇到 `fatal error: 'atomic' file not found` 错误怎么办？","该错误是因为 Codespace 环境中的 Clang++ 编译器无法找到 `atomic` 头文件，通常由 `-stdlib=libc++` 参数引起。解决方法是从 LeanInfer 的 `lakefile` 中移除所有 `-stdlib=libc++` 编译选项。此外，简化安装流程可能无需手动执行 `lake update`，只需运行 `git lfs install && git clone \u003C模型仓库>` 下载模型，其余步骤可由 VSCode Lean 4 扩展自动处理。","https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fissues\u002F3",{"id":144,"question_zh":145,"answer_zh":146,"source_url":147},16230,"如何在开发 std 库时使用 Lean Copilot，避免循环依赖问题？","由于 Lean Copilot 依赖于 std 库，而 std 库不能直接依赖 Copilot，因此无法直接在 std 仓库中使用。推荐的变通方案是创建一个新的仓库（例如 `MyStdLib`），让它同时依赖 `std` 和 `LeanCopilot`。在这个新仓库中开发增量式的证明或功能，验证通过后再将代码移植回标准的 `std` 库。这种方法虽然仅限于新增内容，但能有效利用 Copilot 辅助开发。","https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fissues\u002F54",{"id":149,"question_zh":150,"answer_zh":151,"source_url":142},16231,"安装 Lean Copilot 时是否必须运行 `lake update`？","不一定。在某些情况下（如使用简化安装流程或在 GitHub Codespace 中），`lake update` 并非必需，甚至可能因意外更新其他依赖（如 Mathlib4）而导致问题。如果项目配置正确，通常只需克隆模型仓库并让 VSCode Lean 4 扩展自动处理构建即可。如果遇到构建挂起或依赖冲突，尝试跳过 `lake update` 步骤。",{"id":153,"question_zh":154,"answer_zh":155,"source_url":142},16232,"Lean Copilot 是否支持在 WSL (Windows Subsystem for Linux) 上运行？","虽然官方未对 WSL 进行正式测试，但根据在 GitHub Codespace 中的成功运行经验，Lean Copilot 很可能也可以在 WSL 上正常工作。用户可以尝试按照 Linux 的安装步骤进行操作，如果遇到类似 Codespace 中的编译问题（如缺少头文件），可参考移除 `-stdlib=libc++` 等解决方案。",{"id":157,"question_zh":158,"answer_zh":159,"source_url":137},16233,"如何在不修改标准库源码的情况下体验 Lean Copilot 的新功能或外部模型？","维护者正在开发支持插件化加载外部模型的功能（见 PR #97）。在该功能合并之前，用户可以通过创建独立的项目来测试不同模型，而不是直接修改核心库。关注后续的稳定版本发布，新版本将支持原生在 Lean 中插入和比较多种外部模型，无需复杂的本地构建配置。",[161,166,171,176,181,186,191,196,201,206,211,216,221,226,231,236,241,246,251,256],{"id":162,"version":163,"summary_zh":164,"released_at":165},97214,"v4.28.0","## 变更内容\n* 由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F190 中升级至最新稳定版本 Lean 4.28.0\n\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv4.27.0...v4.28.0","2026-02-16T23:28:40",{"id":167,"version":168,"summary_zh":169,"released_at":170},97215,"v4.27.0","## 变更内容\n* 升级至最新稳定版本 Lean 4.27.0，并由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F189 中修复因依赖项升级引入的 bug\n\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv4.26.0...v4.27.0","2026-02-11T08:52:57",{"id":172,"version":173,"summary_zh":174,"released_at":175},97216,"v4.26.0","## 变更内容\n* 由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F186 中升级至最新的稳定版 Lean 版本\n\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv4.25.0...v4.26.0","2025-12-18T08:48:08",{"id":177,"version":178,"summary_zh":179,"released_at":180},97217,"v4.25.0","## 变更内容\n* 由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F184 中升级至最新稳定版本 Lean v4.25.0\n\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv4.24.0...v4.25.0","2025-11-17T22:59:57",{"id":182,"version":183,"summary_zh":184,"released_at":185},97218,"v4.24.0","## 变更内容\n* 由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F183 中升级至最新的稳定版 Lean 版本（v4.24.0）\n\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv4.23.0...v4.24.0","2025-10-14T20:59:57",{"id":187,"version":188,"summary_zh":189,"released_at":190},97219,"v4.23.0","## 变更内容\n* 由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F174 中提交的最新稳定版本\n* 由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F179 中提交的最新稳定版本\n* 由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F182 中升级至最新稳定版本 Lean 4.23.0\n\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv4.22.0...v4.23.0","2025-09-15T02:33:58",{"id":192,"version":193,"summary_zh":194,"released_at":195},97220,"v4.22.0","## 变更内容\n* 由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F181 中升级至最新稳定版本 Lean v4.23.0-rc1（约等于 v4.22.0）\n\n注意：该标签实际使用的是 Lean v4.23.0-rc1，它与 Lean v4.22.0 几乎同时发布，因此我们的依赖直接升级到了 v4.23.0-rc1，跳过了 v4.22.0 版本。\n\n**完整变更日志**：https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv4.21.0...v4.22.0","2025-08-14T08:52:07",{"id":197,"version":198,"summary_zh":199,"released_at":200},97221,"v4.21.0","## 变更内容\n* 由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F175 中升级至最新不稳定版本\n* 由 @durant42040 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F176 中引入 Kimina 推理引擎 API\n* 由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F178 中升级至 Lean v4.21.0\n\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv4.20.0...v4.21.0","2025-07-01T03:48:53",{"id":202,"version":203,"summary_zh":204,"released_at":205},97222,"v4.20.0","## 变更内容\n* 由 @durant42040 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F170 中更新了 Windows 平台的 README 文件\n* 由 @durant42040 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F171 中使用下载的 C++ 头文件编译 ct2.o\n* 由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F173 中升级到最新的稳定版 Lean\n\n**完整变更日志**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv4.19.0...v4.20.0","2025-06-06T05:32:18",{"id":207,"version":208,"summary_zh":209,"released_at":210},97223,"v4.19.0","## 变更内容\n* 最新稳定版本 v4.17.0，由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F150 中发布\n* 将 main 分支合并到 stable 分支，由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F158 中完成\n* 小幅调整：修改 CI 测试触发的分支，由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F160 中完成\n* 升级至最新版本，由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F164 中完成\n* 小修复，由 @Peiyang-Song 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F165 中完成\n* 为 v4.19.0 更新编译后的 ct2 文件，由 @durant42040 在 https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F167 中完成\n\n\n**完整变更日志**：https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv4.18.0...v4.19.0","2025-05-02T21:51:52",{"id":212,"version":213,"summary_zh":214,"released_at":215},97224,"v4.18.0","## What's Changed\r\n* Bump to latest by @Peiyang-Song in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F151\r\n* Windows support by @durant42040 in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F152\r\n\r\n## New Contributors\r\n* @durant42040 made their first contribution in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F152\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv4.17.0...v4.18.0","2025-04-09T09:48:09",{"id":217,"version":218,"summary_zh":219,"released_at":220},97225,"v4.17.0","## What's Changed\r\n* Bump to latest: v4.17.0-rc1 by @Peiyang-Song in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F147\r\n* Bump to latest stable version (Lean v4.17.0) by @Peiyang-Song in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F149\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv4.16.0...v4.17.0","2025-03-03T22:47:21",{"id":222,"version":223,"summary_zh":224,"released_at":225},97226,"v4.16.0","## What's Changed\r\n* Bump to latest (v4.16.0-rc1) by @Peiyang-Song in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F142\r\n* merge main into stable by @Peiyang-Song in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F145\r\n* Bump to Lean v4.16.0 stable by @Peiyang-Song in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F146\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv4.15.0...v4.16.0","2025-02-03T11:02:37",{"id":227,"version":228,"summary_zh":229,"released_at":230},97227,"v4.15.0","## What's Changed\r\n* Update README.md by @ldct in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F139\r\n* Bump to v4.15.0 by @Peiyang-Song in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F141\r\n\r\n## New Contributors\r\n* @ldct made their first contribution in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F139\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv4.14.0...v4.15.0","2025-01-04T08:29:14",{"id":232,"version":233,"summary_zh":234,"released_at":235},97228,"v4.14.0","## What's Changed\r\n* bump to v4.14.0 by @yangky11 in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F138\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv1.6.0...v4.14.0","2024-12-25T18:45:41",{"id":237,"version":238,"summary_zh":239,"released_at":240},97229,"v1.6.0","## What's Changed\r\n* Bump to Lean v4.11.0 with deps by @Peiyang-Song in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F123\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv1.5.3...v1.6.0","2024-09-02T09:55:55",{"id":242,"version":243,"summary_zh":244,"released_at":245},97230,"v1.5.3","## What's Changed\r\n* Bump to Lean v4.11.0-rc3 with dependencies by @Peiyang-Song in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F121\r\n* Support external models by @Peiyang-Song in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F122\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv1.5.2...v1.5.3","2024-09-01T08:13:37",{"id":247,"version":248,"summary_zh":249,"released_at":250},97231,"v1.5.2","## What's Changed\r\n* Add model manual download instructions by @Peiyang-Song in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F118\r\n* Fix CI by @Peiyang-Song in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F119\r\n* Bump to Lean v4.11.0-rc2 with dependencies by @Peiyang-Song in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F120\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv1.5.1...v1.5.2","2024-08-29T23:50:20",{"id":252,"version":253,"summary_zh":254,"released_at":255},97232,"v1.5.1","## What's Changed\r\n* Fix string UTF-8 misformatting PANIC by @Peiyang-Song in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F114\r\n* Bump to Lean v4.11.0-rc1 by @Peiyang-Song in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F116\r\n\r\n## Note\r\nThis version has a known issue that the Lean server occasionally fails on certain hardware architectures. This issue stems from the new version of Lean4 itself, and disappears in the next pre-release of Lean4 (Lean v4.11.0-rc2), which corresponds to Lean Copilot v1.5.2. Thus the issue should only appear in this release.\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv1.5.0...v1.5.1","2024-08-14T23:57:01",{"id":257,"version":258,"summary_zh":259,"released_at":260},97233,"v1.5.0","## What's Changed\r\n* Bump to Lean v4.10.0 by @Peiyang-Song in https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fpull\u002F112\r\n\r\n\r\n**Full Changelog**: https:\u002F\u002Fgithub.com\u002Flean-dojo\u002FLeanCopilot\u002Fcompare\u002Fv1.4.2...v1.5.0","2024-08-14T08:47:56"]