日报 · 2026-10-17
GitHub 新星 · 汇总 2026-10-17 新收录项目
当日新增收录 31 个项目
覆盖日期：2026-10-17 · 北京时间
发布时间：2026/10/18 08:00:24
部分来源已不可公开，本期文章将按其余项目修订后重新展示。下方为已保存的项目资料。
本期项目资料
收录数按本期首次达标收录的项目统计。
- perplexityai/SPACERust · 1,372 Stars
沙箱代码执行
Stars 统计更新于 2026/10/18 07:36:08
查看原始资料
作者描述
Sandboxed Platform for Agentic Code Execution: pluggable-runtime sandboxes for running agent and user code in isolation.
编报时的 AI 中文解读
SPACE 是一个用于在隔离的 Firecracker 微虚拟机中运行代码的沙箱平台，应用程序通过 HTTP API 创建 Linux 环境、执行命令、收集输出并管理沙箱生命周期。适用于需要为代理、用户工作区或代码运行任务提供独立执行环境的开发者。
编报时的 README 依据
# SPACE [](LICENSE) [Guide](https://space.docs.perplexity.ai/guide/index.html) · [API Reference](https://space.docs.perplexity.ai/api/index.html) SPACE (Sandboxed Platform for Agentic Code Execution) is a sandbox platform for running code in isolated Firecracker microVMs. Applications use its HTTP API to create Linux environments, execute commands, collect output, and manage sandbox lifecycles.
| [Your first sandbox](docs/first-sandbox.md) | A complete API workflow, command customization, output handling, inspection, and deletion. | | [E2B Compatibility Wrapper](docs/ecw.md) | Configure ECW, create a compatibility API key, use the tested E2B Python SDK, and understand supported behavior and limitations. | | [API reference](docs/api/index.md) | Endpoints, request and response fields, and examples. | | [Troubleshooting](docs/troubleshooting.md) | Local setup, authentication, lifecycle operations, command streams, and safe retries. |
Follow [Installation and configuration](docs/installation.md) to deploy SPACE in a local Kind cluster and authenticate. Then follow [Your first sandbox](docs/first-sandbox.md) to run a command and collect its output. The setup guide covers Kubernetes; the [architecture guide](docs/architecture.md#request-and-execution-flow) also explains how SPACE can run without it. ## Documentation
The API serves its generated reference at `/openapi-m1.json`, with request fields, response schemas, and errors. ## SPACE API The SPACE API manages sandbox lifecycles, runs and controls processes, and streams their output. Fields marked `incubating` in the schema are outside the supported contract. SPACE is implemented in Rust, but applications interact with it over HTTP and server-sent events. You do not need to embed a Rust library to use it; the first-sandbox walkthrough runs a standard-library-only Python example. Installation and cluster diagnostics use Bash. ## Development
- zhoufuweigg/guoapp未标注语言 · 1,073 Stars
短剧影视播放、视频下载、追剧管理
Stars 统计更新于 2026/10/18 07:40:57
查看原始资料
作者描述
红果鉴/真果鉴 - 专属私有短剧客户端
编报时的 AI 中文解读
红果鉴/真果鉴是一款基于Flutter的多端短剧与影视应用，支持浏览、搜索、播放、下载和追剧等功能，数据请求在设备上完成，不依赖自建服务。适用于喜欢观看短剧、韩剧、鬼片等内容的用户。
编报时的 README 依据
默认可见站源（红果、韩小圈、鬼片、青空）无需密码；其余站源默认隐藏，在站源管理里通过密码锁解锁后显示。已有受限用户不会自动获得新增站源权限，管理员可在用户管理中勾选。 这是**编译选项**，应用内不能切换版本。标题、Android 桌面名称与电视横幅、Windows 窗口与分发文件名、iOS 显示名随编译选项变化。界面、站源调用、原生下载调度同时限制可用站源；红果版不会访问或继续执行其他站源的旧任务。 两版保留原 Android / iOS 应用标识及数据目录；Android 使用同一签名可相互覆盖升级，不能作为两个独立正式应用并排安装。切换版本保留追剧、观看记录、用户权限和下载记录；备份格式保持兼容。 ## 使用 | 功能 | 操作 | | --- | --- | | 浏览 | 标题栏依次提供排序与筛选、榜单、多选下载和展开搜索。分类保持独立一行，内容区左右滑动切类；支持默认、名称、自然季号、上线日期、热度、播放量排序及连载状态筛选。列表下滑接近底部自动加载下一页，底部「加载更多」为兜底 | | 搜索 | 红果合并官网与名称索引，输入停顿 300 毫秒显示最多 10 条联想；韩小圈、剧果、野果、鬼片调用各自的在线搜索并按分页继续加载。部分失败保留有效结果；最近 20 次搜索按用户保存 |
| 推荐 | 仅红果分类栏在「全部」后显示「推荐」，可切换真人剧、漫剧、AI 剧并分别记住位置 | | 详情 | 普通点剧直接进入播放；下载和兜底入口保留详情页。详情页浏览封面、资料、追剧状态与可展开简介；底部固定「立即播放 / 继续播放」和下载入口。选集默认折叠，展开后每组 50 集，可切换范围、定位当前或输入集数跳转；宽屏与电视采用资料、选集分栏 | | 追剧与历史 | 按想看 / 在看 / 已看 / 有更新筛选；红果已追剧发现新季后进入「有更新」。追剧和历史均可搜索。每个用户独立保存，每 5 秒及退出播放时记录真实进度 | | 卡片与批量下载 | 卡片「更多」提供收藏、状态和下载选集；电脑右键，电视菜单键。发现页长按卡片或点标题栏「多选下载」，一次最多 50 部 | | VIP | 全站源版浏览黄豆或全部站源时显示 VIP 图标，仅过滤黄豆；默认隐藏已确认的 VIP。未知、免费、VIP 分别保存，未知不会覆盖已知状态 | | 手机播放 | 上滑下一集、下滑上一集；长按 350 毫秒临时 3 倍速。竖屏轻点播放 / 暂停，横屏轻点控制条，双击播放 / 暂停。视频下方为「选集 / 简介 / 下载」Tab | | 画中画 | Android 手机 / 平板播放器接入画中画按钮；进入前隐藏自绘控制层与弹幕，小窗只保留画面。Android TV、Windows、iOS 暂未接入 |
入口：「追剧 → 标题栏同步」或「设置与备份 → 设备互联」。手动同步默认「双向合并」，可选覆盖对方 / 覆盖本机，先「查看预览」再执行；自动同步仅发送改动条目与删除标记，播放期间约每 10 秒发送进度。冲突记录保留候选，用户从同步页逐项处理。 ## 安装包与平台状态 | 平台 | 包与状态 | | --- | --- | | Android 8.0+ | 三架构（arm64-v8a / armeabi-v7a / x86_64）APK；同一签名可覆盖升级 | | Windows 10/11 x64 | 完整 ZIP 解压后运行 `hongguojian.exe` / `zhenguojian.exe`，保留所有 DLL 与 `data`；局域网原生发现依赖 Windows 10 1903+ | | Android TV | 与手机共用源码，自动识别电视模式并保持横屏；待电视 / 盒子实机验收 | | iOS 15.1+ | 已加入工程、Go 核心链接、媒体依赖、文件管理与构建脚本；iOS 播放页禁用 media_kit_video 硬件纹理加速以规避 libmpv 渲染崩溃；待 Xcode 构建与真机验收，无已签名 IPA | `INSTALL_FAILED_NO_MATCHING_ABIS` 表示 APK 与设备架构不匹配，请更换对应架构安装包。
本地不入库的 `android/key.properties`： ~~~properties storeFile=/absolute/path/zhenguojian-release.jks storePassword=你的密码 keyAlias=zhenguojian keyPassword=你的密码 ~~~ ## 开发与构建 Flutter `3.47.4`、Dart `3.12+`、Go `1.24.1+`、Python `3.10+`。Android 需要 JDK 17、SDK 36、NDK `28.2.13676358`；Windows 需要 Visual Studio C++ 桌面组件及 MinGW-w64 x64；iOS 需要 macOS、完整 Xcode 和 CocoaPods。 构建脚本对子进程默认设置 `GOPROXY=https://goproxy.cn,direct`、`GOSUMDB=off`，同名环境变量可覆盖。 ~~~sh python3 scripts/build_android.py # 红果鉴 python3 scripts/build_android.py --all-sources # 真果鉴 python3 scripts/build_android.py --abi arm64-v8a
| 野果 | 接口实际分类、目录、详情及真实分集；按分集重新取流，保留 H.264 / H.265 地址 | 站源在线分页搜索 | | 帝果 | 网页分类、目录、详情及分集；vplayer 签名解析 | 站源在线分页搜索 | | 黄果视频 / 黄果 AI / 黄果旧版 | 列表、分类、详情及分集 | 筛选已加载短剧 | 站源可用性、清晰度和区域限制取决于源站及网络；应用不解除源站 VIP 或其他授权限制。 ### 网络与资源设置 管理员从「设置与备份 → 网络与资源」选择自动、直连或手动代理，手动地址支持 HTTP、HTTPS、SOCKS5 / SOCKS5h。代理地址默认遮蔽、不进入备份；设备内播放服务始终绕过代理。自动模式读取系统静态代理与排除列表，恢复前台及每 30 秒更新；代理不会关闭 TLS 校验。 目录请求并发可设 1–6、间隔 0–5000 毫秒（默认 3 / 250ms）；下载并发默认 2、可设 1–6。仍保留前台优先、后台让行、取消、退避及请求数量边界。 ### 局域网互联：追剧同步与推送播放（已接入，待验证） 两台设备在同一局域网直连，自动同步追剧记录或接续播放，不依赖账号服务器。服务类型 `_zgj-link._tcp`（DNS-SD / mDNS），按设备 ID 合并多网卡地址，首次配对记录证书指纹，后续连接固定校验。
- goodnight000/api-anythingTypeScript · 1,051 Stars
网站转API、代理集成
Stars 统计更新于 2026/10/18 07:40:57
查看原始资料
作者描述
Turn any website into an API your agent can call. Learns from the site's own frontend, heals itself when the site changes.
编报时的 AI 中文解读
API Anything 是一个本地工具，能将任意网站转化为可供 AI 代理调用的 API。它通过观察网站自身的前端请求来学习操作，并在网站变化时自动修复。适用于希望让代理以 API 方式访问网站而非驱动浏览器的开发者。
编报时的 README 依据
# API Anything [](https://github.com/goodnight000/api-anything/actions/workflows/ci.yml) **Teach your agent a website once. After that, it calls the site like an API instead of driving a browser.** API Anything watches a site's own page make a request and learns which parts of it are your inputs. It saves the result as an operation. Your agent calls that operation with new inputs over MCP, the CLI or TypeScript, and gets JSON back over plain HTTP.
Paste this into Claude Code, Codex or another coding agent: ```text Set up API Anything for me: https://github.com/goodnight000/api-anything/blob/main/INSTALL.md ``` The agent checks Node 22.13+, builds and installs the `api-anything` command, confirms that a call works, then connects the MCP server and the skill to itself. Restart the agent, then ask it something like "What does Goodreads rate Piranesi?" <details> <summary>Install by hand</summary> The package isn't on npm yet. Don't use `npm install -g github:...`, because it fails under npm 11. ```sh
`reason` and a `next` step. Connect an agent with `codex mcp add api-anything -- api-anything mcp`, `claude mcp add --scope user api-anything -- api-anything mcp`, or this MCP entry: ```json { "mcpServers": { "api-anything": { "command": "api-anything", "args": ["mcp"] } } } ``` </details> Over MCP an agent **calls** operations, using four tools: `list_sites`, `list_operations`, `call_operation` and `login`. To **create** operations, an agent needs a shell and the [skill](skills/api-anything/SKILL.md). A Chromium browser is needed only for creating operations
- It knows only the operations it was taught. It doesn't support every feature of every site. - Repair can fail. There is no CAPTCHA solving, and if a site challenges real Chrome, you have to clear the challenge by hand. - No write operations are bundled. LinkedIn messaging isn't bundled and hasn't been tested against LinkedIn. Writes are tested only against a local fixture, and they need `--allow-writes`. An HTTP 2xx doesn't prove a message was delivered. - Credentials stay in `~/.api-anything`, and specs hold references, never values. Read each
- Miaotofu01/Study-MateJavaScript · 1,030 Stars
AI辅助学习、学习计划制定、项目实操
Stars 统计更新于 2026/10/18 07:05:29
查看原始资料
作者描述
你的AI学习搭档：定路线、讲知识、做项目，边学边做，学透一门科目
编报时的 AI 中文解读
StudyMate 是一套面向数学与计算机科目学习的 AI 学习助手，通过 DSH 插件、Antigravity 插件或 Codex/ChatGPT 插件形式提供，支持制定学习计划、讲解知识、做项目，并遵循“learn with doing”原则。
编报时的 README 依据
<p align="center"><img src="docs/images/taitou.png" width="860" alt="StudyMate：看板娘 + 手写体品牌字 + 覆盖科目（线代／微积分／概率论／C++／Python／机器学习／深度学习）+「任何科目，一站式搞定 / Learn With Doing」"></p> ## 前言 本项目原先是vibe出来给自己用的一个小项目，没想到有这么多人喜欢。但是vibe出来的东西有很多的问题，包括但不限于「文档过于臃肿且充满AI味、可读性极差」、「海量且无用的防御性代码」、「混乱的功能模块」。 虽然现在跑起来的效果也不差，但是跟我理想中的效果还是有很大差距 我觉得这样的东西对不起这么多的信任与star，我会直视这些问题，并且人工修改审查每一个文件，在未来的更新维护中给大家带来更好的体验，感谢大家的使用与支持，有任何建议都可以提个issue，本项目将长期维护。 ## 快速开始 ### DeepSeek Harness **DSH 依赖**：DSH 0.1.7-alpha.1+ #### 安装 - web端 ```bash dsh plugin --profile web add @yunmiao/studymate ``` - 桌面端 ```bash
> **支持范围：只有 `v1.0.0`。** 1.1.0 起的版本不再保证在这一侧可用。 从 [`v1.0.0` 的 Release](https://github.com/Miaotofu01/Study-Mate/releases/tag/v1.0.0)（[ZIP 直链](https://github.com/Miaotofu01/Study-Mate/releases/download/v1.0.0/studymate-openai.zip)）下载 **studymate-openai.zip**，通过 Codex/ChatGPT 提供的插件导入入口导入（详见 [导入说明](docs/使用/Codex与ChatGPT.md)）。这一侧没有后续版本可换，换机器时用同一个 ZIP。 ### Google Antigravity > **支持范围：只有 `v1.0.0`。** 1.1.0 起的版本不再保证在这一侧可用。 用 `v1.0.0` 那棵树构建并安装（不要用 `main`）： ```bash git checkout v1.0.0 node bin/studymate.mjs build-antigravity --install ```
- **带着指定教材自学** → 盘问结束后主动提供本地资料路径（讲义、笔记或教材目录），系统把教材转成 Markdown 放进 `reference/`（学生能翻），把它收集到的在线来源转成 Markdown 放进 `sources/`（写课对齐用）；大纲与课件都对着这批原文写。 - **贴一段看不懂的课文 + 「这里没懂」** → 主教练当场答一小段，记一条档案，送你回原位接着读。 - **「考考我」** → 现场出题 + 按可运行证据核验，给一份评估记录并更新进度。 - **「太简单了 / 没听懂」** → 换讲法（加边界与反例，或降一层抽象），并把这条偏好记进共享记忆。 <img src="docs/images/preview-chat.png" width="640" alt="对话示例"> ## 配置与维护 日常要跑的命令只有一条，就是你装完之后自检的那条： ```bash npm test ```
<p align="center"><img src="docs/images/logo.png" width="120" alt="StudyMate"></p> <h1 align="center">StudyMate</h1> <p align="center"><b>你的AI学习搭档：定计划、讲知识、做项目，学透一门科目</b></p> <p align="center"> <img src="https://img.shields.io/badge/DSH-%E5%AD%A6%E4%B9%A0%E6%A8%A1%E5%BC%8F%E9%A2%84%E8%AE%BE-1c5a40" alt="DSH 学习模式预设"> <img src="https://img.shields.io/badge/Antigravity-%E5%A4%9A%E6%99%BA%E8%83%BD%E4%BD%93%E6%8F%92%E4%BB%B6-4285f4" alt="Antigravity 插件"> <img src="https://img.shields.io/badge/Node.js-22.19%2B%20%7C%2024-339933" alt="Node.js">
<summary><b>装完看不到课件页</b></summary> 课件由**阅读端实时渲染**，不再预生成 HTML（页面模板已随 #83 退役）。在 DSH 里新建会话选「学习模式」，从侧边栏切到阅读端即可。要一份能离线打开的，跑 `npx -y @yunmiao/studymate export`，产物在 `<工作区>/export/`。 </details> <details> <summary><b>还需要装 Python 吗</b></summary> 不需要。引擎从 Python 脚本迁到了随包发的 TypeScript 插件（[ADR-0002](docs/adr/0002-引擎改用TypeScript.md)），安装器也不再探测 Python。只有一件事另外要 Node：导出静态页面时要一份 React（`npm i -g react react-dom`，或用 `STUDYMATE_REACT_DIR` 指过去）。 </details> 更多问题（手改 YAML 的坑、大纲改节点后指针为什么会错、能不能离线）见 [使用说明 §八 常见问题](docs/使用/使用说明.md#八常见问题)。 ## 贡献 / License - **项目交流群**(QQ)：161914370
- ESPARGOS/esp-sdrC · 1,006 Stars
SDR接收、频谱分析
Stars 统计更新于 2026/10/18 07:40:57
查看原始资料
作者描述
ESP-SDR uses the undocumented raw I/Q capture functionality of ESP32 family chips to turn them into (low-duty-cycle) Software Defined Radios
编报时的 AI 中文解读
ESP-SDR 固件将 ESP32 系列芯片内置的 2.4 GHz Wi-Fi 无线电转换为软件定义无线电（SDR），无需额外 SDR 硬件，可捕获原始 I/Q 样本并通过 USB 或 UART 传输到计算机进行分析，适合无线电爱好者和嵌入式开发者。
编报时的 README 依据
# ESP-SDR firmware <img src="docs/espargos-logo.png" width="40%" align="right" alt="ESPARGOS logo"> ### Try it yourself: [Project overview](https://espargos.net/espsdr/) · [Browser SDR viewer](https://espargos.net/espsdr/app/) · [Browser firmware installer](https://espargos.net/espsdr/app/flash.html) ESP-SDR turns the ESP32's built-in 2.4 GHz Wi-Fi radio into a **software-defined radio (SDR)**. It lets you capture the radio signal itself and process it in software, so you can view the spectrum and study signals beyond ordinary Wi-Fi packets. No separate SDR hardware is needed. The
low (0), or high (1). See [GPIO controls](docs/rx-controls.md#gpio-outputs). ## Commands and transport Connect over native USB or a 3.3 V USB-to-UART adapter with crossed TX/RX and common ground, using the pins above. Both interfaces carry the same request/response protocol: send newline-terminated ASCII commands and read text replies. Capture replies also include a binary I/Q payload. Query `INFO` and `CAPS` to identify the firmware and supported features. `VERSION?` reports the Git revision and UTC build date/time; see [firmware version reporting](docs/firmware-version.md).
| Chip | Status | Native USB | UART0 TX / RX | Minimum flash | Special modes / firmware | | --- | --- | --- | --- | --- | --- | | ESP32 | ✅ | — | GPIO1 / GPIO3 | 2 MB | — | | ESP32-C2 | ✅ (26 MHz crystal) | — | GPIO20 / GPIO19 | 2 MB | — | | ESP32-C3 | ✅ | Serial/JTAG | GPIO21 / GPIO20 | 2 MB | — | | ESP32-C5 | ✅ | Serial/JTAG | GPIO11 / GPIO12 | 2 MB | — | | ESP32-C6 | ✅ | Serial/JTAG | GPIO16 / GPIO17 | 2 MB | — | | ESP32-C61 | ✅ | Serial/JTAG | GPIO11 / GPIO10 | 2 MB | — | | ESP32-H2 | ✅ | Serial/JTAG | GPIO24 / GPIO23 | 2 MB | — | | ESP32-H21 | 🚧 | — | — | — | — |
| ESP32-H4 | 🚧 | — | — | — | — | | ESP32-P4 | ❌ | — | — | — | — | | ESP32-S2 | ✅ | USB-OTG CDC | GPIO43 / GPIO44 | 4 MB | — | | ESP32-S3 | ✅ | Serial/JTAG | GPIO43 / GPIO44 | 2 MB | [Continuous decimated I/Q over USB (15.625–250 kSa/s)](#s3-streaming) | | ESP32-S31 | ✅ | Serial/JTAG | GPIO58 / GPIO59 | 2 MB | [High Speed USB / Ethernet streaming at up to 40 MSa/s (experimental)](#s31-streaming) | ✅ Supported · 🚧 Not yet supported · ❌ Unsupported (no integrated radio). USB, UART, and flash requirements above refer to the standard firmware; see each
 The diagram shows the hardware's receive and transmit paths; this firmware currently implements reception only. **Parts of the firmware code are AI-generated.** While we have a very good understanding of how the IQ sampling functionality works on the ESP32-C61 chip (used in our ESPARGOS One array), making IQ sampling work on the whole range of ESP32 family chips would have been too much work without LLM support. ## Chip support
- **S2:** 80/40/16 MS/s; 15–60 MHz bandwidth; up to 12,284 complex samples. - **S3:** 13–69 MHz bandwidth. - **S31:** 80/40/20/10/8/4 MS/s; 13–54 MHz bandwidth. Captures have gaps; nominal sample rates exceed sustained serial throughput. Gain and power are uncalibrated. Extended tuning does not guarantee PLL lock or reception; the viewer uses the ISM-band warning described above. See [receive-control details](docs/rx-controls.md). ## Contributors <table> <tr> <td align="center"> <a href="https://github.com/Jeija">
- rohit267/net-mergeGo · 788 Stars
网络聚合
Stars 统计更新于 2026/10/18 07:30:37
查看原始资料
作者描述
Use all your network adapters at once on Windows/Linux/Mac: per-connection load balancing over Wi-Fi, Ethernet and phone/USB tethers (TUN + Wintun)
编报时的 AI 中文解读
net-merge 是一款免费工具，可同时使用 Wi-Fi、以太网和手机热点等多个网络连接，通过按连接分配流量来提升下载、游戏和浏览等场景的总体速度。适用于希望充分利用多条网络链路的用户。
编报时的 README 依据
> **Start net-merge first, then connect the Android phone.** macOS drops the TetherKit connection each > time net-merge starts, quits, or is paused and resumed. If the phone link disappears, click **Apply** > in TetherKit again. The iPhone is not affected. A phone uses its mobile data, so part of your traffic goes through it while net-merge runs. Quit net-merge when you do not want that. ## Use it 1. Start net-merge. A tray icon appears (menu bar on Mac). 2. The icon is **green** when it is merging and **grey** when it is paused.
and KStatusNotifierItem Support* extension). </details> <details> <summary><strong>A site asks me to log in again</strong></summary> <br> net-merge may use a different IP for each connection, and some sites tie your login to one IP. That is a limit of splitting connections over several links. </details> ## Commands ``` net-merge tray tray icon (the normal way to run it) net-merge on | off start / pause merging in the running tray net-merge quit pause and close the tray net-merge status | list live numbers / which connections are used
To try it without installing: `sudo nix run github:rohit267/net-merge -- run`. ## Add a second connection net-merge needs **two or more** real connections. Use Wi-Fi plus one of these: a phone (USB tethering or hotspot), an Ethernet cable, or a second Wi-Fi. Plug it in and net-merge adds it within a few seconds. When you unplug it, its connections move to the others. **Windows and Linux:** plug the phone in and turn on *USB tethering* (Android) or *Personal Hotspot* (iPhone). That is all. **Mac with an iPhone** (nothing extra to install)
net-merge autostart on|off|status net-merge socks plain SOCKS5 proxy on 127.0.0.1:1080 (no admin, per app) net-merge cleanup remove leftover routes after a crash net-merge install macOS only: `sudo net-merge install` (see above) ``` More flags and details are in [docs/DEVELOPING.md](docs/DEVELOPING.md). ## Good to know - Needs admin (Windows), `CAP_NET_ADMIN` (Linux) or `sudo net-merge install` once (macOS). - Some sites tie your login to one IP. net-merge may use a different IP for each connection, so such a site can ask you to log in again.
net-merge gives every new connection to one of your links. When an app opens many connections, they spread over all your links. - **Good for:** Steam and game downloads, browsers, system updates, cloud uploads (Google Drive), download managers like IDM, torrents. - **Not for:** one single connection. One connection always uses one link, so one big file from one connection does not get faster. - **Works for every app**, TCP and UDP, IPv4 and IPv6. You do not set up a proxy. - **Smart:** faster links get more connections. A link with no internet gets none until it works again.
- If your connections differ in IPv6 support, sites that only have IPv6 are not reachable while the box above is ticked. ## More - What changed in each version: [CHANGELOG.md](CHANGELOG.md) - How it works, building from source, tests: [docs/DEVELOPING.md](docs/DEVELOPING.md) - Want to help? [CONTRIBUTING.md](CONTRIBUTING.md). Found a security problem? [SECURITY.md](SECURITY.md) - License: MIT, see `LICENSE`. Third-party parts: `THIRD-PARTY-NOTICES.md`. ## Support
- Noniv/web-oceanHTML · 476 Stars
海洋渲染
Stars 统计更新于 2026/10/18 07:30:41
查看原始资料
作者描述
Real-time WebGPU ocean renderer written from scratch, no engine. FFT waves, whitecaps, sun glitter, physically based sunset sky with clouds, free camera. Built with Claude Code (Opus 5.5). Opens from a single 61 KB HTML file in Chrome.
编报时的 AI 中文解读
这是一个用WebGPU从零编写的实时海洋渲染器，无需引擎或渲染库，可在Chrome中直接打开单个HTML文件运行，适合对图形渲染和海洋模拟感兴趣的开发者或爱好者。
编报时的 README 依据
You need a recent Chrome with WebGPU and a reasonably capable GPU. On a laptop, force Chrome onto the dedicated GPU in the Windows graphics settings, or it will fall back to the integrated one. Open `ocean.html` directly in the browser. `ocean.min.html` is the same page minified (61 KB instead of 102 KB). Both are self-contained and work from `file://`. To run from source: ``` npm install npm start # serves http://localhost:8080/ ```
At 2560 by 1440 on an RTX 5070 Ti, the GPU takes about 6 to 8 ms per frame, measured while other programs were also using the GPU. The HUD shows FPS and GPU time when the browser supports `timestamp-query`. The per-frame code reuses preallocated typed arrays and descriptors, so the garbage collector has almost nothing to do. The only recurring allocations are the objects WebGPU itself requires, such as the swap-chain view and command buffers. ## Known limitations - Far water is still somewhat darker than the bright horizon sky above it.
- The water is tuned toward turbid coastal water. The "Water turbidity" slider gives darker, bluer deep water when turned down. - The scene is open ocean only. There is no shoreline, surf, or objects in the water. - Clouds are two flat layers, not volumetric. ## How it was made
- BaruqueRodrigues/llm-legislativoPython · 395 Stars
模型预训练、语料库构建、文本生成
Stars 统计更新于 2026/10/18 07:30:41
查看原始资料
作者描述
Corpus legislativo brasileiro em larga escala e modelo Transformer causal treinado do zero em PyTorch puro.
编报时的 AI 中文解读
该项目构建了巴西立法语料库，并从头训练了一个因果Transformer语言模型，用于研究巴西立法文本的统计模式。 面向研究社区，提供公开的模型权重、数据集和完整的预处理、训练、评估流程，强调可复现性和科学严谨性。
编报时的 README 依据
O **LLM Legislativo Brasileiro** é um projeto de pesquisa e engenharia de inteligência artificial de ponta a ponta. Seu objetivo é estruturar o acervo histórico oficial do Congresso Nacional brasileiro (Câmara dos Deputados e Senado Federal) em dados abertos auditáveis e treinar um modelo de linguagem causal autônomo, **implementado do zero sem uso de pesos pré-treinados, sem dependência de APIs fechadas e sem caixas-pretas como o Hugging Face Trainer**.
Todo o pipeline foi concebido sob princípios estritos de **ciência aberta**, **reprodutibilidade**, **integridade criptográfica (SHA-256)** e **isolamento causal estrito (zero data leakage entre splits de treino e teste)**. --- ## 🚀 Destaques do Projeto * 📦 **Volume de Dados Extraordinário:** * **+360 GiB** de dados oficiais em arquivos brutos (TARs compactados e documentos primários). * **Câmara dos Deputados:** 43 arquivos TARs catalogados (**311,38 GiB**, 856.902 registros internos verificados). * **Senado Federal:** 16 lotes documentais (**53,36 GiB**) com processos e diários.
Todos os artefatos gerados pelo projeto estão disponibilizados publicamente para a comunidade de pesquisa: | Tipo | Repositório no Hub | Descrição | |:---:|:---|:---| | 🤖 **Modelo** | [**BaruqRodrigues/llm-legislativo**](https://huggingface.co/BaruqRodrigues/llm-legislativo) | Pesos `model.safetensors`, tokenizer, configurações, métricas e inferência pura. |
git clone https://github.com/BaruqueRodrigues/llm-legislativo.git cd llm-legislativo # Criar e ativar ambiente virtual python3 -m venv .venv source .venv/bin/activate # Instalar o pacote em modo editável com dependências pip install -e . ``` ### 2. Executar a Suíte de Testes Automatizados Todos os contratos de modelo, persistência, tokenização e avaliação podem ser validados localmente em CPU: ```bash pytest tests/ -q ``` *(135 testes passam com 100% de sucesso).* ### 3. Usar o Modelo Treinado para Inferência (PyTorch Puro)
Você pode carregar o modelo treinado diretamente do Hugging Face Hub sem depender de bibliotecas externas complexas: ```python import torch from huggingface_hub import hf_hub_download from safetensors.torch import load_file from legislm.model.gpt import GPTModel from legislm.model.config import ModelConfig from tokenizers import Tokenizer # 1. Baixar pesos e tokenizer do Hub weights_path = hf_hub_download(repo_id="BaruqRodrigues/llm-legislativo", filename="model.safetensors") tokenizer_path = hf_hub_download(repo_id="BaruqRodrigues/llm-legislativo", filename="tokenizer.json")
# 🏛️ LLM Legislativo Brasileiro > **Corpus Oficial em Larga Escala e Pré-treinamento de Modelo de Linguagem Causal (Transformer) Construído 100% do Zero em PyTorch Puro.** [](https://www.python.org/) [](https://pytorch.org/) [](https://www.r-project.org/) [](https://developer.apple.com/documentation/vision)
* **Ele não deve ser utilizado como consultor jurídico factual ou ferramenta de decisão normativa:** modelos causais pré-treinados podem alucinar números de leis, inventar ementas ou atribuir conteúdos fictícios a dispositivos vigentes. --- ## 📄 Licença O código-fonte e as tabelas de metadados deste projeto estão licenciados sob os termos da [Licença MIT](LICENSE). Os documentos e textos legislativos oficiais processados são de domínio público, de acordo com a legislação brasileira (Lei nº 9.610/1998, Art. 8º, IV).
- aicodeking/leanreaPython · 392 Stars
逆向工程、AI辅助编程
Stars 统计更新于 2026/10/18 07:30:41
查看原始资料
作者描述
One 8 KB skill that replaces REA (Reverse Engineer Anything): plain CLI tools, no MCP. Benchmarked: 38% faster, 35% cheaper.
编报时的 AI 中文解读
这是一个8KB的Agent技能文件，用于替代REA（Reverse Engineer Anything）工具，通过指导AI代理直接使用命令行工具进行逆向工程，无需MCP服务器和大量工具定义。适用于使用Claude Code等支持技能的AI代理进行软件逆向分析的用户。
编报时的 README 依据
- Ghidra, Hopper and jadx weren't installed, so the native tasks ran on Xcode tools for both arms. - The plain model is already good at this. leanrea's gain over no skill (about 19% faster, 32% fewer turns) is smaller than its gain over REA. ## Install ```sh npx skills add aicodeking/leanrea ``` This uses the [skills CLI](https://skills.sh). It finds the `leanrea` skill and asks which agents to install it for: Claude Code, Codex, Cursor, Gemini CLI, OpenCode and the others it supports. Common variants: ```sh
npx skills add aicodeking/leanrea -g # install for your user, in every project npx skills add aicodeking/leanrea -a claude-code -y # Claude Code only, no prompts npx skills add aicodeking/leanrea -g -a '*' -y # every supported agent, user level npx skills update leanrea # pull the latest version later ``` Manual install, if you prefer: ```sh git clone https://github.com/aicodeking/leanrea mkdir -p ~/.claude/skills && cp -R leanrea/leanrea ~/.claude/skills/ ```
If an agent has no skills support, paste `leanrea/SKILL.md` into its rules or system prompt; it is plain text. Use it: ``` /leanrea How does the search feature in Foo.app work? Show the evidence. ``` ## What's in the skill - **Rules:** - Work on a copy. - Route by `file` and magic bytes first. - Slice minified files instead of reading them. - Batch probes into one shell call. - Search extracted trees with `rg -uu`, because bundlers hide code in dot-folders like `.webpack/`. - Mark each claim as seen, inferred or unknown. - **Router:** the fastest command for each target type:
- **Speed patterns:** start from strings and imports, find a string's function through the otool literal-pool comments, read call sequences rather than every instruction, and diff inventories before diffing code. ## Reproduce the benchmark Needs macOS with the target Electron app installed (see `bench/setup.sh`), `gh` from Homebrew, Node, and Claude Code logged in. ```sh cd bench ./setup.sh # installs rea-agents@6.5.0, copies the targets into bench/work (gitignored) ./run_all.sh # 36 runs, about $8 on Sonnet ./ctx.sh # context-overhead probe
| Electron app analysis (real desktop app) | fails on asar integrity until you pass a flag, then **47.9 s and 140 MB of JSON** | asar extract + `rg -uu`: 4.8 s, 24 KB | Full tables, the context-overhead runs and caveats: [bench/BENCHMARK.md](bench/BENCHMARK.md). Raw per-run metrics: [bench/results/final.jsonl](bench/results/final.jsonl). **Caveats:** - Only 3 reps per cell, one model, one machine. - REA was tested in its best case: Claude Code's deferred tool search loads MCP tools on demand. Agents that load every tool pay the 400k-token row.
python3 mcp_list.py node work/rea-install/node_modules/rea-agents/scripts/rea.mjs --mcp # REA tool count + size ``` The answer keys in `tasks.json` match the app versions used here and a `gh` built with go1.27.1. Other versions need new keys. ## Credit REA is an impressive, well-documented project by the morluto team. Most of the gotchas in this skill came from reading their code and docs. leanrea isn't affiliated with REA. It is just a test of whether a model with plain tools and a short note can do the same work faster. MIT licensed.
- Awakenchan/BloodBorneButPcC# · 391 Stars
游戏启动
Stars 统计更新于 2026/10/18 07:30:41
查看原始资料
作者描述
Bloodborne on Windows in one package: launcher and engine. Bring your own game dump.
编报时的 AI 中文解读
在Windows上运行《血源诅咒》的启动器与引擎，需要用户自行提供游戏文件。
编报时的 README 依据
> **Note:** I'm keeping this going on my own, and updates are slower now because of the hate I've been getting. I'll keep working on it when I can. **No game files are included.** You need your own decrypted Bloodborne dump (CUSA03173) with update 1.09. ## Install 1. Download `BloodbornePC-<version>-win64.zip` from [Releases](https://github.com/Awakenchan/BloodBorneButPc/releases) and extract it anywhere. 2. Run `Bloodborne Launcher.exe` and pick your game folder (the one with `eboot.bin`). 3. Press **Play**. The launcher updates itself. ## Credits | | | |---|---|
- heqiangdiao-wq/AI-Cross-Border-Ecommerce-Agent-ToolkitPython · 354 Stars
AI辅助运营、商品内容生成、广告优化
Stars 统计更新于 2026/10/18 07:30:41
查看原始资料
作者描述
AI 跨境电商运营智能体工具箱 · 把运营 SOP 拆成 7 个可复用的 AI Agent 工作流，含 Prompt、模板与真实案例
编报时的 AI 中文解读
这是一个将跨境电商运营SOP转化为可复用AI Agent工作流的工具箱，包含7个模块，每个模块提供标准Prompt、输入模板和真实案例，帮助运营人员稳定产出商品素材。
编报时的 README 依据
> ⚠️ **请注意**：这份 PDF 是**未经删改的原始内部手册**，其中包含平台明确禁止的做法（如刷单、诱导评价、虚假举报、参数虚标等）。 > > **本仓库的 7 个 Agent 模块均未实现这些做法**：已识别项及其合规替代方案见 [`docs/03_Compliance_Guard.md`](./docs/03_Compliance_Guard.md)，未实现项逐条登记见 [`docs/05_SOP_Gap_Register.md`](./docs/05_SOP_Gap_Register.md)。 > > 换句话说 —— **这份 PDF 是「原料」，不是「产品」。** 本仓库的产出物只包含合规路径。 --- ## 🗂 Agent 全景 7 个模块，**全部已完成**。 | # | Agent | 中文 | 一句话职责 | 版本 | |---|---|---|---|---| | **01** | [Market Research Agent](./Agents/01_Market_Research_Agent/) | 市场分析 | 前置判断 / 容量 / 集中度 / 价格带（波谷检测）/ 配置组合 · **带计算脚本** | ✅ v0.4 |
| 06 | [Advertising Optimization Agent](./Agents/06_Advertising_Optimization_Agent/) | 广告优化 | 四类漏斗诊断 + 盈亏线核算 + **P0/P1/P2 分级动作** · Skill 化 | ✅ v0.4 | | 07 | [Review Analysis Agent](./Agents/07_Review_Analysis_Agent/) | 评论分析 | 八类归因 + 误解型识别 → 详情页修补建议 | ✅ v0.4 | > **这七个模块构成一条完整链路，而且是闭环的**： > > ```text > 01 市场分析 ──► 02 产品策略 ──► 03/04/05 素材产出 ──► 上架 > （做什么） （做成什么样） （怎么表达） > │ > ┌───────────────────────────┘ > ▼ > 06 广告诊断 ─┐
> 07 评论诊断 ─┴──► 回流 03/04 修正 ──► 复测 ──► 达标放量 / 未达标回诊断 > ``` > > **为什么按这个顺序编号**：`01-02` 分析簇（前置链路）→ `03-04-05` 创意簇（素材产出）→ `06-07` 运营簇（数据回流）。 > > ✅ 编号方案已确认。原始规划为 `03` Listing / `04` Advertising / `05` Review / `06` Video（且无 Market Research 与 Product Strategy 的独立编号），本仓库按工作簇重排，变更记录见 [CHANGELOG](./CHANGELOG.md)。 --- ## 🚀 快速开始 不需要 API key，不需要装环境。**三个文件就能跑。** ```text Step 1 复制 Agents/<模块>/templates/ 下的输入模板 Step 2 填上你的产品信息（10 分钟） Step 3 把「系统 Prompt」+「任务 Prompt」+「填好的模板」丢给任意 LLM Step 4 用 docs/04_Result_Checklist.md 逐条自检，通过即发布 ```
## 🧭 设计原则 1. **母语输出优先**——巴西站的标题必须像巴西人写的。直译英文 Listing 的 CTR 结构性偏低，所以 Prompt 强制要求原生 PT-BR 表达，而不是"翻译腔"。 2. **输入契约化**——每个 Prompt 都声明"缺什么字段就不许输出什么内容"。宁可报错，不许编参数。 3. **合规优先于转化**——涉及参数虚标、品牌词、夸大宣传的动作，Prompt 里一律做**硬拦截**，不是"建议"。详见 [`docs/03_Compliance_Guard.md`](./docs/03_Compliance_Guard.md)。 4. **可验证**——每个模块交付时给一份自检清单，逐条打勾才算完成。产出对不对，不靠感觉。 --- ## ⚠️ 合规声明 本工具箱只做**合法合规的运营提效**。原始 SOP 中存在少数平台明确禁止的灰产动作（如虚假交易/刷单、组织虚假举报、参数虚标与夸大宣传）。这些内容**已在本仓库中被识别并改写为合规替代方案**： - 改写对照表 → [`docs/03_Compliance_Guard.md`](./docs/03_Compliance_Guard.md)
- NB-Corp/gpt-with-tasteJavaScript · 351 Stars
AI辅助编程
Stars 统计更新于 2026/10/18 07:30:41
查看原始资料
作者描述
Better judgment. Better taste. Less ceremony. System instructions and source-aware steering for GPT in Codex.
编报时的 AI 中文解读
为GPT在Codex等编码工具中提供行为配方，改善判断力、品味和简洁性，适用于希望提升编码助手输出质量的开发者。
编报时的 README 依据
| **Oh My Pi (OMP)** | Native identity-aware extension | [docs/harnesses/omp.md](docs/harnesses/omp.md) | | **DeepSeek Harness (dsh)** | Native Cordis plugin with request-local steer | [docs/harnesses/dsh.md](docs/harnesses/dsh.md) | Running in another host? See [docs/harnesses/porting.md](docs/harnesses/porting.md). Maintainers edit the single English authoring source in `recipe/` and compile across targets: see [docs/authoring.md](docs/authoring.md). --- <a id="install"></a> ## Install · Codex With Node.js 22 or newer installed: ```sh
git clone https://github.com/NB-Corp/gpt-with-taste.git cd gpt-with-taste node scripts/install.mjs --hooks ``` We recommend Hooks for multi-turn work in Codex Desktop: user submissions, child startup and supported wait results receive scoped reminders. Later child-input coverage depends on the [client entry point](docs/install.md#runtime-coverage). For a prompt-only installation, run `node scripts/install.mjs` instead. Neither route needs npm packages.
Both routes install into the active `CODEX_HOME` (otherwise `~/.codex`), write the two managed prompt files, and point `model_instructions_file` at the main one. Your model, provider, effort, authentication, permissions, MCP servers, roles, and `AGENTS.md` remain unchanged. To target another home, add `--home` with a quoted absolute path, for example `--home "C:/Users/you/.codex"`.
With `--hooks`, review and trust the three installed definitions in Codex Desktop under **Settings → Hooks**, or in the interactive terminal with `/hooks`. Enable only the events you want. Open a **new Codex session** after installation. Check updates with `node scripts/check-updates.mjs`. Installation also checks official metadata before writing; add `--skip-update-check` for offline installation. [Update instructions](docs/install.md#update-or-resolve-a-conflict) cover the installed components and customized checkouts.
- SakanaAI/AI-Scientist-v3Python · 332 Stars
自动化科学实验、论文自动生成
查看原始资料
作者描述
The Agentic Scientist: Orchestrating Coding Agents for Automated Scientific (Co-)Discovery
编报时的 AI 中文解读
这是一个将编码代理编排为AI科学家的工具，从任务描述开始自动进行实验、消融和论文撰写，适用于自动化科学发现的研究场景。
- lucannez64/darling-lolPython · 327 Stars
游戏运行
查看原始资料
作者描述
Experimental Darling runtime patches and tools for testing the macOS League client on Linux
编报时的 AI 中文解读
这是一个实验性项目，通过Darling在Linux上运行macOS版英雄联盟客户端，包含源码补丁和工具。主要面向技术挑战和Linux玩家，但存在封号风险，维护有限。
- leag1234/elixirofficeJavaScript · 322 Stars
办公文档编辑、PDF处理、在线协作
查看原始资料
作者描述
Open Source Office suite for users used to MS Office - multi-platform, tuned for performance
编报时的 AI 中文解读
ElixirOffice 是一款面向熟悉 Microsoft Office 用户的开源办公套件，可编辑 Word、Excel 和 PowerPoint 文档及其 OpenDocument 等价格式，提供桌面应用和自托管服务器版本，注重隐私和性能。
- PiggyBackBayonet/MiAutoKotlin · 316 Stars
CarPlay接入
查看原始资料
作者描述
小米 / OPPO / vivo 等 ICCOA CarLink 手机投上车机：无线连接，无需转接盒、无需 root。Car-side ICCOA CarLink receiver for Xiaomi CarWith, OPPO Car+ and vivo Jovi InCar.
编报时的 AI 中文解读
MiAuto 是一个在汽车车机上运行的接收端应用，它模拟 ICCOA CarLink 车机，让小米、OPPO、vivo 等支持该协议的手机无需额外安装软件或硬件即可无线投屏到车机。
- YIAN6557/eian-riso-brollJavaScript · 306 Stars
视频动画生成、口播视频辅助
查看原始资料
作者描述
孔版双色风格的口播信息动画 Skill：逐字稿 → 1080×1920 竖屏 mp4 · Risograph-style info animations for talking-head videos
编报时的 AI 中文解读
这是一个为口播视频制作孔版双色风格信息动画的AI编码助手工作流，输入逐字稿或字幕文件，输出竖屏mp4动画。适用于短视频创作者，无需图像生成模型，完全本地运行。
- reason-machines/reason-networkRust · 209 Stars
代理通信、团队协作
查看原始资料
作者描述
Reason Network: your team's coding agents, talking to each other. The engine behind `reason network`.
编报时的 AI 中文解读
Reason Network 是一个让团队中的编码代理（如 Claude Code、Codex、Cursor 等）相互通信的引擎。它通过每台电脑上的守护进程和 MCP 服务器，使代理会话能实时查看彼此状态并直接发送消息，无需中心服务器。
- cranebeaker/coop-anythingPython · 163 Stars
游戏合作化、AI辅助编程
查看原始资料
作者描述
Co-op Anything: point your AI agent at any single-player game and play it with friends (online and/or splitscreen!). MCP server and playbook to let you reverse engineer, fix, test, ship and share what it learns.
编报时的 AI 中文解读
这是一个指导AI编码代理为单人游戏添加合作模式的资源库，包含MCP服务器、技能和操作手册，支持从零构建或修复现有合作模组，并打包发布。
- gungnir-sanye/tee-attestation-extractionPython · 131 Stars
TEE安全研究、Keybox提取、RKP种子提取
查看原始资料
作者描述
面向授权研究的 Android TEE/KeyMint 认证材料提取与验证工程，覆盖 Keybox、RKP/CDI、证书链及后续扩展方案，记录原理、实践流水、补丁与镜像登记、验收规则和脱敏证据。
编报时的 AI 中文解读
面向授权CTF、Android TEE安全研究和防御性复核的技术文档档案，按三个设备型号组织，记录研究目标、提取对象、验证状态、原理分析、操作流水、失败分支和公开边界。
- Apolinario-coder/SkyRecompC · 127 Stars
游戏重编译、PC游戏运行
查看原始资料
作者描述
Static recompilation of The Legend of Zelda: Skyward Sword for PC
编报时的 AI 中文解读
这是一个将《塞尔达传说：天空之剑HD》静态重编译为原生PC版本的项目，包含重编译工具和Vulkan/SDL3兼容运行时，用户需自备合法游戏文件。 面向希望在PC上以原生性能运行该游戏的玩家，提供图形界面启动器，无需命令行操作。
- christos-antono/NPCTalkPython · 119 Stars
NPC对话、语音克隆、游戏模组
查看原始资料
作者描述
Make any NPC in any game talk back with Fish Audio: skills, a self-hosted voice server and tools for AI coding agents, with a full GTA V example.
编报时的 AI 中文解读
NPCTalk 是一个让游戏中的 NPC 能够与玩家对话的工具包，它提供了技能、自托管语音服务器和 AI 编码代理工具，并附带完整的 GTA V 示例。
- V4bel/InfinityCoreC · 110 Stars
漏洞研究资料
查看原始资料
- kergee/mcd-undercoverJavaScript · 110 Stars
聚会游戏
查看原始资料
作者描述
把真实门店菜单变成 3～8 人的同机谁是卧底游戏（麦当劳 MCP 参赛项目）
编报时的 AI 中文解读
将真实门店菜单转化为3至8人同机谁是卧底游戏的趣味应用，适合聚会和社团活动，无需购买餐品。
- OverkillLabs3/driveclub-recompC++ · 109 Stars
游戏运行
查看原始资料
作者描述
DriveClub (PS4, version 1.28) running natively on Windows
编报时的 AI 中文解读
这是一个让PS4游戏《DriveClub》在Windows上原生运行的项目，提供可执行程序和启动器，需要用户自备游戏文件。
- zzusec/apk-reversePython · 106 Stars
APK逆向分析、去广告、去臃肿
查看原始资料
作者描述
Suitable for Android APK reverse engineering analysis
编报时的 AI 中文解读
这是一个供AI代理（如Claude Code、Codex）加载的Agent Skill，用于Android APK逆向工程、去广告、去臃肿、dex补丁、重打包及运行时/服务器分析。它以程序化门控流程组织，包含症状索引、门控和停止条件，旨在指导代理高效完成逆向任务。
- c22dev/mush8Rust · 105 Stars
游戏运行、多人联机
查看原始资料
作者描述
A native Mach-O iOS/macOS port of Mario Kart 8 Deluxe, with LAN, online and wireless multiplayer
编报时的 AI 中文解读
将《马力欧卡丁车8 豪华版》移植到苹果设备（iOS/macOS）的原生应用，支持局域网、在线和无线多人游戏。适合拥有游戏文件并希望在Apple Silicon设备上体验的玩家。
- MaxSikorski/jumpstart-3dHTML · 75 Stars
3D建模
查看原始资料
作者描述
Browser-based 3D CAD by J.Siwek, community-maintained. No install.
编报时的 AI 中文解读
这是一个在浏览器中运行的免费3D CAD工具，无需安装或账户，适合从初学者到进阶用户进行三维建模和设计。
- byJoey/cfnew-panelJavaScript · 72 Stars
订阅生成、配置管理
查看原始资料
作者描述
CFnew 独立面板：Vercel 一键部署、订阅生成与 Worker 配置管理
编报时的 AI 中文解读
CFnew 独立面板是一个部署在 Vercel 的 Web 应用，用于管理 Cloudflare Worker 代理配置、生成订阅链接，并提供在线 IP 检测等功能。适合已有 CFnew 内核并希望简化配置管理的用户。
- rezadesky/RSDVaultTypeScript · 72 Stars
记账、预算管理、报表导出
查看原始资料
作者描述
Modern, secure, offline-first personal finance and expense tracking mobile application built with React Native Expo
编报时的 AI 中文解读
RSDVault 是一款注重隐私的离线优先个人财务管理移动应用，用于记录收支、管理预算和查看统计报告。适合希望完全掌控个人财务数据、无需云端同步的用户。
- yapfast/yapfastRust · 70 Stars
聊天辅助
查看原始资料
作者描述
The chat app that doesn't eat your PC. A lightweight, native Discord client in Rust.
编报时的 AI 中文解读
Yapfast 是一个用 Rust 编写的轻量级原生 Discord 客户端，内存占用远低于官方应用，支持服务器、私信、语音、GIF 等核心功能，适合追求低资源消耗的 Discord 用户。 它通过直接与 Discord API 通信并自绘界面，避免了浏览器引擎，启动速度快，并支持 Matrix 协议（需自行构建）。