日报 · 2026-10-25
GitHub 新星 · 汇总昨日新收录项目
当日新增收录 44 个项目
覆盖日期：2026-10-25 · 北京时间
发布时间：2026/10/26 08:00:28
本期文章生成失败，后台将在稍后重试。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。
- ludovic111/horizonRust · 2,000 Stars
游戏运行
Stars 统计更新于 2026/10/26 07:55:32
查看原始资料
作者描述
Minecraft: Java Edition 26.3 rewritten from scratch in Rust, checked against vanilla with 281 differential harnesses. Get it with the Horizon launcher.
编报时的 AI 中文解读
Horizon 是用 Rust 重写的 Minecraft: Java Edition 26.3 原生游戏，可读取原版世界、遵循相同规则，并通过启动器使用自有 Minecraft 副本运行。适用于希望以原生性能体验原版极高一致性的玩家。
编报时的 README 依据
<td width="50%"><img src="docs/media/launcher-patch-notes.webp" alt="The launcher's Patch notes page"></td> </tr> </table> ## What works today The [documentation](https://playhorizon.dev/docs/) covers each of these. - **The game:** singleplayer with vanilla's world generation, mobs, villagers, raids, redstone, every command, the HUD and the recipe book, on saves in vanilla's format. Worlds from 1.8 to 26.2 are upgraded as vanilla does. [Status table](#project-status).
- **Playing with others:** invite friends from the pause menu's Friends button with a code, Open to LAN, and joining offline-mode servers. Horizon cannot sign in to a Microsoft account yet, so it cannot join servers that check accounts. See [Multiplayer](https://playhorizon.dev/docs/multiplayer). - **Mods:** Horizon has its own mod system (Rust compiled to WebAssembly). Java mod jars do not run. See [Mods and data packs](https://playhorizon.dev/docs/mods). Screenshots, captured in the Rust client itself with vanilla assets and no mods or shaders: <table> <tr>
The launcher's **Servers** page creates, starts, backs up and updates a Horizon server with buttons, and invites friends over Tailscale. On a VPS, `horizon-server` installs with one command and runs without Java. Guide: [Host a server](https://playhorizon.dev/docs/hosting) ([reference](docs/hosting.md)). ## Modpacks
Horizon is installed through **the Horizon launcher**. Get it from [playhorizon.dev](https://playhorizon.dev) (the Download button picks your system) or from [the launcher releases](https://github.com/ludovic111/horizon/releases?q=launcher-v&expanded=true), for Linux x86-64, macOS on Apple Silicon or Windows x64. The launcher installs and updates the game, prepares its files from your own copy of Minecraft, and updates itself.
AI agents and scripts can play Horizon. `horizon-ctl`, shipped beside the game in every release folder, is a command line that reads the game as JSON and acts like a player; `horizon-ctl mcp` serves the same commands as an MCP server (one tool per command, screenshots returned as images). ```sh horizon-ctl launch --world Survival --seed 5 --wait # finds the version the launcher installed horizon-ctl state # position, health, food, target, screen... horizon-ctl find-block birch_log # the nearest logs
**You need** Minecraft: Java Edition 26.3, played once (the official launcher, Prism, MultiMC and other launchers all work); a Java 25 JDK, for the one-time preparation; and a GPU with Vulkan on Linux and Windows, or any Apple Silicon Mac. Step by step, with macOS's first-open prompt and the first launch: [Install and first launch](https://playhorizon.dev/docs/install). <table> <tr> <td width="50%"><img src="docs/media/launcher-play.webp" alt="The launcher's Play page: a version ready to play, its summary, Client and Server, the version picker and Play"></td>
For developers. Requirements: Rust 1.90 or newer, Python 3.10 or newer, JDK 25, and on Linux the ALSA and udev development headers (`libasound2-dev libudev-dev` on Debian and Ubuntu). Development and the vanilla harnesses run on Linux. Horizon contains no game files and never downloads them. It builds its resources from **your own copy** of Minecraft: Java Edition 26.3, verified against the checksums in `research-lock.json`:
Horizon can't sign in to a Microsoft account yet, so it can't join online-mode servers (it joins offline-mode ones); skins served over https show the default skin unless already cached; pack icons show the unknown-pack texture in game; the creative hotbar keys (C/X) aren't wired to the hotbar loader; item tooltips lack attribute lines. ## Licence
- amosblomqvist/learnharnessTypeScript · 1,643 Stars
个性化学习、知识图谱管理
Stars 统计更新于 2026/10/26 07:56:03
查看原始资料
作者描述
Agent harness for learning difficult things.
编报时的 AI 中文解读
LearnHarness 是一个桌面应用，利用 AI 代理根据用户当前理解水平规划并教授课程，通过持续更新知识图谱来跟踪学习进度，适用于希望系统性学习复杂主题的个人用户。
编报时的 README 依据
**[Download from learnharness.ai](https://learnharness.ai)** · [GitHub Releases](https://github.com/amosblomqvist/learnharness/releases) - **macOS (Apple Silicon):** signed and notarized DMG. Drag the app to Applications. - **Windows (x64):** run the installer. It is currently unsigned; Windows may show an “unknown publisher” or SmartScreen warning. - **Linux (x64):** make the AppImage executable and open it, or install the Debian/Ubuntu `.deb`.
Add your provider API key and choose your models, then start a session with what you want to understand. LearnHarness supports 30+ provider options, including OpenAI, Anthropic, Google, DeepSeek, and OpenRouter. Optional dictation and web search use separate service keys. All three platforms support in-app updates. ## Your keys, your data - **Your keys and models.** No LearnHarness subscription or markup. You pay your providers for usage; a live cost meter shows model costs.
- **Local storage.** Your chats, graph, settings, and API keys live on your machine. Learning records are plain files you can inspect. - **External services.** Models run through your configured providers. Model requests send the context needed for the task; web research and dictation also contact external services. Local storage does not mean offline processing. - **Inspectable behavior.** System prompts are visible and editable. The app is open source under the AGPL-3.0 license. ## Development and support ```bash npm install npm run dev # http://localhost:3000
- morganlinton/BeautifulRustRust · 695 Stars
AI辅助编程
Stars 统计更新于 2026/10/26 07:51:43
查看原始资料
作者描述
A Claude Code plugin that teaches Claude to write idiomatic, robust Rust
编报时的 AI 中文解读
这是一个Agent Skill，用于指导编码代理编写地道、健壮的Rust代码。它适用于使用Claude Code、OpenAI Codex、GitHub Copilot、Cursor、Gemini CLI等支持Agent Skills格式的编码代理。
编报时的 README 依据
/// Returns the major version number. pub fn major(&self) -> u64 { self.major } } ``` Across three test tasks (a library, a CLI, and refactoring a messy module), the skill passed 29 of 30 checks against 24 of 30 without it. See [Testing the skill](#testing-the-skill). ## Install The skill is the folder `skills/rust-best-practices/`. Every option below puts that folder where your agent looks for skills. ### Any agent, one command The [`skills`](https://github.com/vercel-labs/skills) installer supports most agents (Node 22.20 or later): ```bash
### OpenAI Codex, Cursor, or anything else: copy the folder Clone the repo and copy `skills/rust-best-practices/` into your agent's skills directory. Most agents now read the shared `.agents/skills/` location: ```bash git clone https://github.com/morganlinton/BeautifulRust mkdir -p ~/.agents/skills cp -R BeautifulRust/skills/rust-best-practices ~/.agents/skills/ ``` | Agent | User directory | Project directory | |---|---|---| | Claude Code | `~/.claude/skills/` | `.claude/skills/` | | OpenAI Codex | `~/.agents/skills/` | `.agents/skills/` |
since it triggers on its own. To try it for one session without installing, clone the repo and run `claude --plugin-dir ./BeautifulRust`. ### GitHub Copilot With GitHub CLI 2.90 or later: ```bash gh skill install morganlinton/BeautifulRust rust-best-practices ``` This installs into the current project by default. Add `--scope user` to install for your user, or `--agent <name>` to install for a different agent. ### Gemini CLI ```bash gemini skills install https://github.com/morganlinton/BeautifulRust --path skills/rust-best-practices ```
- udhaykumarbala/cleanyourtv-appKotlin · 576 Stars
电视清理、去广告、主屏美化
Stars 统计更新于 2026/10/26 07:51:46
查看原始资料
作者描述
A community-built phone app for Clean Your TV: clean up an Android TV from your phone. No laptop, no coding agent, nothing uninstalled.
编报时的 AI 中文解读
Clean Your TV 的配套安卓应用，用于从手机清理 Android TV，无需电脑或编码代理。适合希望快速清除电视广告和臃肿软件、恢复简洁主屏幕的用户。
编报时的 README 依据
<p align="center"> <img src="docs/images/banner-3-steps.png" width="800" alt="Community app for Clean Your TV. Clean your TV from your phone. No laptop. No coding agent. Open the app on your Android phone, tap Start, and the ads and bloat are gone in about five minutes. Three steps: connect, analyse, clean."> </p> <p align="center"> <a href="https://github.com/udhaykumarbala/cleanyourtv-app/releases/latest"><img src="https://img.shields.io/github/v/release/udhaykumarbala/cleanyourtv-app?style=flat&color=E8A84C&labelColor=15181e&label=download%20APK" alt="Download the latest APK"></a>
| :- | :- | :- | | **You need** | A computer with ADB, or a coding agent such as Claude Code | **Just an Android phone** on the same Wi-Fi as the TV | | **You do** | Run the scripts, or give the prompt to the agent | **Tap Start, check the plan, tap "Make it clean"** | | **Made for** | Tinkerers, themes and new TV lists | **Everyone**: you, your parents, your friends | Both give you the same Clean Your TV home screen and the same community TV lists, switch apps off without uninstalling them, and keep a way back for every change. ## Safe by design
| The app finds your TV on the Wi-Fi and shows you how to let the phone in with the remote. | It reads what is on the TV and shows the plan: what goes off and what stays. Nothing changes yet. | Tap "Make it clean". A restore point first, then the clutter goes and the new home screen comes on. | **You need** an Android phone (Android 9 or newer, with a screen lock) on the same Wi-Fi as the TV. [Download the APK](https://github.com/udhaykumarbala/cleanyourtv-app/releases/latest) and open it to install. Connecting takes about a minute (you switch on network debugging with the remote; the app
- **Clean-up only** (the TV keeps its own home screen, and the app says why): Fire TV, TCL's China builds, Android 7, vendor home screens the app does not know, Chromecast with Google TV and the Google TV Streamer (the switch would cost casting), the Kalley and PRISM+ sets whose setup wizard returns at every boot, and TVs whose setup wizard has to stay on. - **Network debugging:** the TV needs ADB over the network (port 5555). On TVs that only offer Wireless debugging with a pairing code, pair once from a computer; the app shows the three commands. </details>
- mizorewww/gdirectRust · 551 Stars
文件管理
Stars 统计更新于 2026/10/26 07:51:46
查看原始资料
作者描述
Rust Google Drive direct downloader using pinned libaria2
编报时的 AI 中文解读
gdirect 是一个用 Rust 编写的 Google Drive 命令行工具，支持浏览、上传、下载和挂载，适合需要高效管理 Google Drive 文件的用户。
编报时的 README 依据
**First login:** download a Desktop OAuth client JSON using the [configuration guide](docs/configuration/README.md), then: ```sh gdirect login --client-secret ~/Downloads/client_secret.json gdirect browse ``` Working Google IPs are discovered automatically and stored in local SQLite. `gdirect scan` rescans; `gdirect scan --list` lists results. Downloads default to 8 connections. | Key | Browser action | | --- | --- | | `j/k`, `↑/↓` | Select | | Enter, `l/→` | Open folder | | `h/←`, Backspace | Parent folder | | `d` | Download; export Google documents as PDF |
| `+/-` | Download connections (1–16) | | `r` | Refresh | | `c` | Cancel, keeping partial data | | `q` | Quit | ```sh gdirect browse -o downloads -j 8 gdirect ls root --json gdirect download FILE_ID -o downloads/file.zip gdirect upload ./file.zip gdirect upload ./file.zip --parent FOLDER_ID --name backup.zip gdirect mount # ~/GoogleDrive; requires mount dependencies gdirect mount ~/MyDrive --read-only gdirect unmount # Windows foreground mount: Ctrl-C ```
# gdirect [中文](README.zh-CN.md) ## Installation **Version 0.1.1+ includes Google authentication. No gcloud CLI or separate Google SDK installation is required.** First-time users still need a Desktop OAuth client JSON: follow the [Google configuration guide](docs/configuration/README.md) after installation. | Use | Runtime dependencies | | --- | --- | | Browse, upload, download | gdirect and a Google account; a browser for initial authorization | | Mount on macOS | rclone 1.75.2+; built-in NFS, no FUSE driver | | Mount on Linux | rclone 1.75.2+, FUSE3 (`fusermount3`) |
| Mount on Windows | rclone 1.75.2+, WinFsp; foreground mode | Prebuilt binaries: macOS 15+ (Apple Silicon / Intel), Linux x86_64 with glibc 2.35+, Windows x86_64. Alpine/musl, older glibc and Linux ARM64 do not have a compatible prebuilt package. Homebrew (macOS / Linux): ```sh brew tap mizorewww/gdirect https://github.com/mizorewww/gdirect && brew install mizorewww/gdirect/gdirect rclone ``` Linux Homebrew users also need their distro's FUSE3 package for mounting. To upgrade: `brew update && brew upgrade gdirect`.
- DenisSergeevitch/ex-ussr-kinship-researchPython · 533 Stars
家谱研究、历史档案检索、证据链构建
Stars 统计更新于 2026/10/26 07:51:46
查看原始资料
作者描述
Evidence-first genealogy skill for AI agents (Codex, Claude Code): church books, civil, military, deportation and repression records across the former USSR. Replies in your language.
编报时的 AI 中文解读
这是一个面向AI代理（如Codex、Claude Code）的家谱研究技能，专注于前苏联、苏联和俄罗斯帝国的历史记录，包括教会记录、民事记录、军事、驱逐和镇压档案。它采用证据优先的方法，帮助用户进行严谨的家族史研究。
编报时的 README 依据
- New eval cases: language matching, unindexed scans, house-number identity, birthplace conflicts, exhausted sources, repression lists. ## Install Clone the repository into your agent's skills folder: ```bash # Codex git clone https://github.com/DenisSergeevitch/ex-ussr-kinship-research ~/.codex/skills/ex-ussr-kinship-research # Claude Code (personal skills) git clone https://github.com/DenisSergeevitch/ex-ussr-kinship-research ~/.claude/skills/ex-ussr-kinship-research ```
For a single project, put the folder in `.claude/skills/` inside the repository. The scripts need only Python 3 and the standard library. ## Use Just describe the family question; the skill triggers on genealogy requests. For example: - *"My great-grandfather was exiled to Kazakhstan in the 1930s. Help me find his parents in church records."* - *"Here are scans of a marriage certificate and a soldier card. Check whether they are the same person and build the evidence chain."*
- **Knowing when to stop.** When public sources are exhausted, the agent says so once, prepares questions for living relatives and request drafts for archives, and does not invent or loop. Also in 1.1.0: - The agent replies in the user's language; the intake template defaults to `requested_output_language: auto`. - The skill description is bilingual, so it triggers for English-speaking users too. - **CAPTCHAs are handed to a human.** The agent never solves or bypasses them, and the user signs in to sites themselves. - More seed sources, each with a reliability class and role.
seed-sources.md curated starting resources with reliability classes field-lessons.md practical lessons from a real case (new in 1.1.0) assets/ case intake, CSV registers, final report template scripts/ init_case.py, hash_artifact.py, audit_evidence.py evals/discovery-and-logic.yaml trigger and reasoning test cases ``` ## Responsible use - **Family history, not surveillance.** The skill refuses to look up current addresses, phones or accounts of living people, and it minimizes their data in reports.
- **No bypassing.** No CAPTCHA solving, no paywall, geo-block or rate-limit evasion, no leaked databases. Sign-ins are done by the user. - **No uploads of your documents or DNA data** to third-party services without your explicit consent. - **Sensitive findings** (deportation, repression, adoption, DNA surprises) are reported neutrally and only as far as the question requires. - **Archive access rules change.** Always check the current official procedure before sending a request; the skill only drafts requests, it does not send them. ## Contributing
- enesozturk/dimensional-drawingHTML · 461 Stars
技术图纸生成、组件架构可视化
Stars 统计更新于 2026/10/26 07:51:46
查看原始资料
编报时的 AI 中文解读
这是一个为AI代理提供的技能包，用于将UI组件转换为类似Apple风格的一页技术图纸，标注间距、圆角、图层和视口位置。适用于需要清晰传达组件架构的设计师和开发者。
编报时的 README 依据
| 🙅 **Never invent values** | unknowns stay `[TBD]` | ## Install **Claude Code** ```bash git clone https://github.com/enesozturk/dimensional-drawing cp -r dimensional-drawing/skills/* ~/.claude/skills/ ``` **claude.ai / Claude desktop:** zip a skill folder (for example `skills/dimensional-drawing`) and add it under *Settings → Capabilities → Skills*. **Any other agent** (Codex, Cursor, Gemini CLI, …): use the [`skills`](https://skills.sh) CLI, or point your agent at the `SKILL.md` files. ```bash npx skills add enesozturk/dimensional-drawing ``` ## Use it
| **With Claude** | ask *"animate this drawing"*. Claude builds it in the **Motion** artifact and you export MP4. This is the primary target and where the timing was tuned. | | **Without Claude** | `cd skills/dimensional-drawing-motion/adapters/headless && npm install`, then `node render.mjs --scene ../../templates/scene-dialog.js --css ../../templates/base.css --out dialog.mp4`. Needs Node 18+ and ffmpeg. | > [!NOTE]
> Optimized for Claude Motion. The headless renderer is tested with Playwright (Chromium) + ffmpeg. Other animation libraries (Remotion, Motion Canvas, GSAP, Lottie…) are **not tested**. ## Examples **Dialog**: fixed width, scrim, viewport placement, header controls, footer actions <img src="assets/dialog-sheet.png" alt="Dimensional drawing of a dialog component" width="100%"> **Card**: flexible width, body well, nested corner, shadow envelope <img src="assets/card-sheet.png" alt="Dimensional drawing of a card component" width="100%">
- jss8649/quiet-tizenPython · 382 Stars
电视界面定制、系统清理、恢复原厂
Stars 统计更新于 2026/10/26 07:51:46
查看原始资料
作者描述
Native Quiet Home launcher, reversible cleanup tools, and an agent-readable guide for the Samsung U8000F running Tizen 9.
编报时的 AI 中文解读
为三星Tizen电视提供更安静、更实用的HDMI显示体验的实用指南和工具包，包含原生启动器、清理配置和恢复方案，适用于特定型号电视的定制。
编报时的 README 依据
# Quiet Tizen A practical, agent-readable guide to making a Samsung Tizen TV quieter and more useful as an HDMI display. This project documents work performed on one **Samsung UN75U8000FFXZA (2025 U8000F)**. It includes **Quiet Home**, a native launcher with a colorful, English-language interface, lightweight settings, local app cards and a shortcut to native Bluetooth controller pairing.  *Layout illustration with the public text-only app cards. Input labels are configurable.*
It also includes targeted cleanup profiles, source-build tools and rollback recipes. **It is a working case study with a reusable toolkit, not a universal installer or a replacement operating system.** ## Give this repository to your coding agent Clone the repository, open it in Codex, and use: > Read README.md and AGENTS.md, then inspect the documentation before touching > my TV. Help me reproduce Quiet Tizen on a TV I own. First identify its exact > model, firmware and runtime, compare them with the tested profile, and show
| [Cleanup profile](profiles/un75u8000ffxza-1300.9.json) | Separately selectable advertising, voice, context, smart-extras, sidebar and TV Plus groups | | [Cleanup planner](tools/plan_cleanup.py) | Field-specific apply/restore SQL generated from your own fresh snapshot | | [Access helper](tools/tv.py) | Explicit-target SDB preflight, temporary authenticated root, bounded transfers and session cleanup checks | | [Recipes](recipes) | Snapshot, stop-marker, launcher, settings-bridge and restoration scripts |
9. [Restore the original experience](docs/09-recovery.md). Read [AGENTS.md](AGENTS.md) before using the tools. TV commands require a specific address and a verified firmware declaration. All downloaded proprietary references, local artwork, snapshots and session records stay under ignored paths. ## Development Host tools use Python 3.12+ and the standard library. Raster artwork conversion optionally uses Pillow. Source builds use the .NET 6 SDK to match the tested TV; it is an older SDK and should be isolated from unrelated projects. Install Samsung's SDB separately. ```sh
python3 -m unittest discover -s tests -v python3 tools/build.py payloads # After obtaining the four private Tizen reference assemblies: python3 tools/build.py launcher --framework .local/tv-framework ``` GitHub Actions runs the offline tests and builds the open-source access payloads and app-catalog tests. The NUI launcher build needs Samsung reference assemblies and is deliberately not built from redistributed SDK files. ## Credits The visual design adapts [Free My TV](https://github.com/simbasjungles/free-my-tv); the temporary archive-launcher access route comes from
| App preloading | Vendor refreshes can restore the preference; no permanent enforcement daemon | | Netflix | More available memory after closing background apps; perceived input lag **not yet confirmed fixed** | The native launcher and underlying techniques were exercised on that TV. The **generalized scripts in this public release were checked locally and compiled where applicable; they have not been replayed end-to-end on a second TV**. This matters especially for private Samsung APIs and registry schemas. ## Walkthrough
- mirkobrombin/ravenC++ · 379 Stars
Windows应用运行
Stars 统计更新于 2026/10/26 07:51:46
查看原始资料
作者描述
A Windows your apps recognize.
编报时的 AI 中文解读
Raven 是一个在 Linux 上运行 Windows 应用程序的兼容层，无需虚拟机或额外配置，直接运行安装程序即可。
编报时的 README 依据
The tested setup also uses libpng 1.6.59, a CPU build of Skia with the Liberation fonts, DXVK and vkd3d-proton; the `scripts/provision-*.py` tools download the pinned versions. ## Installation Raven is built from source for now: ```sh git clone https://github.com/mirkobrombin/raven.git cd raven cmake -S . -B build -DCMAKE_BUILD_TYPE=RelWithDebInfo cmake --build build -j2 ``` Then run an installer, or any program: ```sh export RAVEN_PREFIX="$HOME/.local/share/raven/my-app" build/native/raven --run-process --execution=native --timeout-ms=600000 --desktop=wayland setup.exe ```
The program's code runs directly on your CPU, inside a sandbox where it cannot reach Linux. Every call it makes to the system is answered by Raven, which plays the part of the Windows kernel and its services: processes and threads, files, the registry, windows and input. Graphics go to Wayland and Vulkan, with DXVK and vkd3d-proton for Direct3D. ## Requirements - Linux on x86_64 - CMake 3.20 or newer, a C++20 compiler and Python 3.11 or newer - MinGW-w64 for x86_64 and i686, to build the Windows test programs - For desktop applications: Wayland, xkbcommon, pixman, FreeType and Vulkan
Some tests need software Raven cannot redistribute; point `RAVEN_MINESWEEPER`, `RAVEN_NETFX_INSTALLER`, `RAVEN_NETFX35_PACKAGE`, `RAVEN_XNA_INSTALLER` and `RAVEN_REDIST_DIRECTORY` at your own copies, or those tests are skipped. Each component under [native/upstream](native/upstream) records where it comes from and its license. ## What works today | Area | Status | |---|---| | 32 and 64-bit programs | Both, side by side in the same prefix | | Installers | NSIS, Inno Setup and Windows Installer (MSI) packages, with custom actions, rollback and elevation |
- routerhubSI/glm-5.3-abliteratedPython · 373 Stars
模型对比、提示词参考
Stars 统计更新于 2026/10/26 07:51:46
查看原始资料
作者描述
GLM 5.3 abliterated or uncensored API access with RouterHub. Browse 35 jailbreak prompt links, compare GLM models, and find SDK examples, coding-tool setup, pricing, and evaluation guides.
编报时的 AI 中文解读
用于研究GLM 5.3 abliterated或uncensored模型的API访问，包含35个越狱提示词文件和模型对比指南，适合开发者选择模型变体并评估行为。
编报时的 README 依据
**Find a GLM 5.3 abliterated API, compare GLM 5.3 uncensored models, and explore 35 bundled research prompt files.** This RouterHub guide connects model terminology, jailbreak prompt research, API examples, and coding-tool setup in one place. Searching for a **GLM 5.3 uncensored AI model**, a **GLM 5.3 abliterated model**, or a **GLM 5.3 no-refusal model**? Learn what each label tells you, choose an API ID, and evaluate the behavior you actually need. The gateway offers OpenAI-compatible Chat Completions for GLM and Anthropic-compatible Messages for supported Claude entries.
- project instructions and multi-step agent work; - bilingual drafting and translation; - repeated tasks where token cost affects deployment choices; - comparing standard, Flash, abliterated, and uncensored entries on the same inputs. Choose on measured task quality, latency, and cost rather than a product label alone. The [GLM 5.3 model guide](./docs/01-what-is-glm-5.3-abliterated.md) explains where to start. ## Abliterated vs. uncensored | Question | GLM 5.3 Abliterated | GLM 5.3 Uncensored | |---|---|---|
<a href="https://routerhub.si/models?utm_source=github&utm_medium=readme&utm_campaign=glm-5.3-abliterated"> <img src="./assets/routerhub-banner.png" width="600" alt="GLM 5.3 abliterated and uncensored API access with RouterHub"> </a> ## GLM 5.3 jailbreak collection (35 prompts) The **jailbreak collection is the main event**: 35 paste-ready prompt files bundled in [jailbreaks/](./jailbreaks/), organized by model family. The GLM 5.3 abliterated/uncensored angle and the RouterHub API path are added around them. ### Collection ### OpenCode Zen — Free Stealth Models
The API is the fallback funnel: [GLM 5.3 API quickstart](./docs/03-glm-5.3-api-quickstart.md) · [Get a key](https://routerhub.si/register?next=onboarding&utm_source=github&utm_medium=readme&utm_campaign=glm-5.3-abliterated). ## What is GLM 5.3? **GLM 5.3** belongs to the GLM model family associated with Z.ai / Zhipu. This guide focuses on API access, variant selection, and evaluation for developers considering GLM for coding, agent tasks, analysis, and Chinese/English workflows. Useful evaluation scenarios include: - code explanation, refactoring suggestions, and test generation;
[Register with RouterHub](https://routerhub.si/register?next=onboarding&utm_source=github&utm_medium=readme&utm_campaign=glm-5.3-abliterated), add prepaid credits, and create a private API key. The published API base is `https://routerhub.si/v1`. ```bash export ROUTERHUB_API_KEY="<your-private-key>" export ROUTERHUB_API_BASE_URL="https://routerhub.si/v1" ``` Set these values only in your local environment or secret manager. Running an inference example consumes credits; do not commit a real key. ### 2. Confirm the GLM 5.3 model ID
These entries were listed in the [public catalogue](https://routerhub.si/models.json) on October 24, 2026. All five list Chat Completions and preview availability; recheck availability and rates before use. | Variant | Exact API ID | |---|---| | **GLM 5.3 Abliterated** | `glm-5.3-abliterated` | | **GLM 5.3** | `glm-5.3` | | **GLM 5.3 Flash** | `glm-5.3-flash` | | **GLM 5.3 Uncensored** | `glm-5.3-uncensored` | | **GLM 5.3 Flash Uncensored** | `glm-5.3-flash-uncensored` | ### 3. OpenAI-compatible request Install `openai` with pip, then run a short Chat Completions request: ```python
- **A no-refusal label is not a guarantee.** Evaluate the actual model instead of assuming every request will receive a useful answer. - **Hosted behavior can change.** Recheck availability and rerun important evaluations after updates. - **Protocols are model-specific.** A shared API key does not make a Chat-only model compatible with Messages or Responses. - **Instruction handling matters.** In particular, confirm whether a persona replaces your system/developer instructions.
GLM currently requires Chat Completions. Claude Code requires Messages, so use a supported Claude entry there. For Cursor, check that your version supports the needed custom endpoint. See the [client guide](./docs/04-glm-5.3-claude-code-cursor-opencode-codex.md). **Is GLM 5.3 abliterated safe?** Suitability depends on the task, data, model behavior, and controls around your application. Review outputs and approve consequential actions; see the [deployment checklist](./docs/07-safety-and-responsible-use.md). **Will a GLM 5.3 no-refusal model never refuse?**
- routerhubSI/uncensored-ai-api未标注语言 · 365 Stars
模型接口服务、模型切换
Stars 统计更新于 2026/10/26 07:51:46
查看原始资料
作者描述
Uncensored AI API and models — one key for Claude, GPT, Gemini, Grok, DeepSeek, GLM and Qwen, with uncensored and abliterated variants where available. OpenAI- and Anthropic-compatible. Pay per token.
编报时的 AI 中文解读
RouterHub 是一个 API 网关，通过一个 API 密钥提供对多个前沿及开源模型家族的访问，包括 Claude、GPT、Gemini、Grok、DeepSeek、GLM 和 Qwen，并提供去审查和消融变体。它兼容 OpenAI 和 Anthropic 接口，按 token 计费，适合开发者集成到应用中。
编报时的 README 依据
- **API keys** — create separate keys to isolate projects and usage. - **One dashboard** — keys, billing, usage and support in one place. Live model prices are listed in the dashboard and can change when providers update their rates. See [Pricing & limits](./docs/08-pricing-and-limits.md). ## Use cases - **AI safety and red-team research** — compare how model versions handle instruction conflicts and policy boundaries. - **Creative writing and worldbuilding** — work with fewer mid-scene refusals while keeping your own review process.
- **Agentic coding** — route Claude, GPT, DeepSeek and GLM through one key in Claude Code, Cursor, Codex or OpenCode. - **Multilingual assistants** — test the same prompt across model families without rebuilding your integration. - **Privacy-conscious prototyping** — keep billing and keys centralized while you evaluate providers. - **Developer tooling** — API keys and one balance make it easier to meter usage per project or client. - **AI-assisted editing and tailoring** — rewrite, tighten, adapt tone and tailor copy for different audiences.
export ROUTERHUB_API_KEY="sk-..." export ROUTERHUB_BASE_URL="https://.../v1" # copy from the dashboard ``` ### 2. OpenAI-compatible request ```python import os from openai import OpenAI client = OpenAI( api_key=os.environ["ROUTERHUB_API_KEY"], base_url=os.environ["ROUTERHUB_BASE_URL"], ) response = client.chat.completions.create( model="MODEL_ID", # choose any model ID from the dashboard messages=[ {"role": "system", "content": "You are a precise, helpful assistant."},
{"role": "user", "content": "Explain the trade-offs of model routing in two paragraphs."}, ], ) print(response.choices[0].message.content) ``` ### 3. Anthropic-compatible request ```python import os from anthropic import Anthropic client = Anthropic( api_key=os.environ["ROUTERHUB_API_KEY"], base_url=os.environ["ROUTERHUB_BASE_URL"], # use the Anthropic base URL from the dashboard ) message = client.messages.create( model="MODEL_ID", max_tokens=1024, messages=[{"role": "user", "content": "Write a concise release note for a developer tool."}], )
| **OpenAI SDK** | Set `base_url` to the RouterHub base URL | | **Anthropic SDK** | Set `base_url` to the Anthropic-compatible endpoint | | **Claude Code** | Point the Anthropic base URL at RouterHub | | **Cursor** | Add a custom OpenAI-compatible provider | | **Codex CLI** | Use an OpenAI-compatible provider profile | | **OpenCode** | Configure a custom provider and model ID | | **LangChain / LlamaIndex** | Use the OpenAI-compatible adapter | | **Any OpenAI-compatible app** | Change two values: base URL and API key |
Model IDs, context limits and live availability are shown in the [RouterHub dashboard](https://routerhub.si/?utm_source=github&utm_medium=readme&utm_campaign=uncensored-ai-api). Availability changes as providers release new versions — always confirm the current model ID before deploying. ## Quickstart ### 1. Get your key and base URL Create an account at **[routerhub.si](https://routerhub.si/?utm_source=github&utm_medium=readme&utm_campaign=uncensored-ai-api)**. Your API key starts with `sk-`; the exact **Base URL** is shown on the dashboard overview. ```bash
- **Summarization and document analysis** — long-context workflows across model families. - **Chatbot and assistant backends** — one API for multiple model families. ## Honest limits - **No provider can guarantee that a model will never refuse.** Uncensored and abliterated variants reduce refusal behavior; they do not remove the need for evaluation, guardrails or human review. - **Model behavior is version-dependent.** A prompt that works on one release may behave differently after an upstream update.
- **Provider terms still apply.** You are responsible for complying with the terms of the model providers and the laws that apply to you. - **Availability changes.** Model IDs and uncensored variants rotate as providers ship updates; check the dashboard before relying on a specific ID. ## FAQ **What is an uncensored AI API?**
- mrdoob/slop-newyorkJavaScript · 358 Stars
三维城市漫游
Stars 统计更新于 2026/10/26 07:51:46
查看原始资料
编报时的 AI 中文解读
基于 three.js 和 WebGPU 的纽约市三维模拟，可步行、驾车、骑行或乘直升机探索曼哈顿，并支持作为起点自由修改以制作游戏、动画或实验。
编报时的 README 依据
# Slop New York A New York City simulation built with [three.js](https://threejs.org/), WebGPURenderer, and TSL. Explore Manhattan on foot, by car, by bicycle, or by helicopter. [Try it live](https://mrdoob.github.io/slop-newyork/) ## Make it your own Fork it, remix it, improve it, break it. Make games, animations, films, experiments, or something completely different. Use it as a starting point and take it wherever you want. Have fun! ## License and data The project code is available under the [MIT License](LICENSE).
- zengshuangzi/mcd-money-saverPython · 347 Stars
省钱助手、优惠券管理、预算点餐
查看原始资料
作者描述
基于麦当劳MCP的智能省钱助手
编报时的 AI 中文解读
基于麦当劳MCP的智能省钱Skill，帮助用户自动管理优惠券、积分和套餐组合，计算最优点餐方案。
- wangrongding/tv-freshShell · 346 Stars
电视去广告、电视换桌面、电视装应用
查看原始资料
作者描述
tv-fresh 电视焕新：小米电视(MIUI TV)去广告/清理预装/换桌面/装代理与影视仓的 Agent Skill——免root可逆，含固件逆向实录与一键脚本
编报时的 AI 中文解读
这是一个面向小米电视（MIUI TV）用户的改造指南与Agent Skill，提供去广告、换桌面、装代理和影视应用等操作的详细步骤和可逆命令，适合希望通过ADB方式安全改造电视的用户。
- misshexa66/OunceCakeJavaScript · 313 Stars
游戏数据提取
查看原始资料
作者描述
Tool to dump unencrypted romFS and exeFS for Switch 2 games
编报时的 AI 中文解读
用于从Switch 2游戏、更新和DLC中提取未加密的romFS和exeFS数据，支持数字版、卡带和Switch 2 Edition，需配合ns2jb漏洞使用。
- dsmurrell/onkeyRust · 306 Stars
歌唱训练
查看原始资料
作者描述
Learn to sing the songs you love. A free, open-source singing trainer for Mac that listens to the music you play.
编报时的 AI 中文解读
OnKey 是一款面向 Mac 用户的开源歌唱训练工具，通过聆听用户播放的音乐来学习歌曲旋律，并在用户跟唱时提供实时音高反馈和评分。
- kailashsr38/ascii-magic-artJavaScript · 302 Stars
ASCII艺术展示、终端浏览、网页嵌入
查看原始资料
作者描述
634 animated ASCII art pieces for web pages, READMEs, wallpapers and terminals. Browse them in your terminal with npx ascii-magic-art and add one to your project in one prompt.
编报时的 AI 中文解读
提供634个动画ASCII艺术作品，可用于网页、README、壁纸和终端，支持在终端浏览、生成提示词并集成到项目中。
- ryderclanky/BotW-NativeC++ · 203 Stars
游戏运行
查看原始资料
作者描述
Breath of the Wild (Wii U, v1.5.0) statically recompiled to a native Windows program with a Vulkan renderer.
编报时的 AI 中文解读
将《塞尔达传说：旷野之息》Wii U版静态重编译为原生Windows程序，使用Vulkan渲染，无需模拟器。适用于希望在Windows上以原生性能运行该游戏的玩家，但需自备合法游戏文件。
- YoyoLin008/mcreplayPython · 200 Stars
结构测试、变化检测、本地回放
查看原始资料
作者描述
McReplay（麦回放）— Structural snapshots, change reports, and local MCP replay for McDonald's China integrations.
编报时的 AI 中文解读
McReplay（麦回放）是一个面向维护麦当劳中国MCP Skill、解析器和演示应用的开发者的本地工具，用于将MCP查询结果转换为可复现的结构测试，支持结构快照、变化检测和本地回放。
- dvbs2000/liuchunlin-shufa未标注语言 · 174 Stars
书法生成
查看原始资料
- mamcjr/BOTWRecompAndroidC · 166 Stars
游戏运行
查看原始资料
作者描述
Static recompilation of The Legend of Zelda: Breath of the Wild (Switch 1.0.0 / 1.6.0 / 1.8.2) for Android - no game files included
编报时的 AI 中文解读
这是一个将《塞尔达传说：旷野之息》静态重编译为Android原生应用的项目，非模拟器，需要用户自备合法游戏文件。目前处于Beta阶段，主要面向希望在Android设备上原生运行该游戏的玩家。
- vxcn/mcd-kcal-pickerPython · 156 Stars
热量搭配、营养数据获取
查看原始资料
作者描述
麦门热量搭配师：基于麦当劳中国 MCP 的热量搭配 Skill，只读，不下单
编报时的 AI 中文解读
基于麦当劳中国官方MCP数据的热量搭配Skill，用户设定目标热量和上限，自动从官方营养数据中挑选主食、小食和饮料组合，使总热量接近目标且不超过上限。适用于想控制单餐热量的人群，也适合想了解商业MCP数据应用和Skill结构的人。
- holdcloud/mcd-calendar-buddyPython · 152 Stars
日历同步、活动提醒
查看原始资料
作者描述
作为一个J人&麦门信徒，是时候把优惠活动日期添加到我的日历了。疯狂星期四（just kidding~
编报时的 AI 中文解读
面向Mac用户的日历提醒工具，通过麦当劳MCP获取活动信息并自动同步到macOS日历，支持智能提醒和关键词过滤。适合喜欢麦当劳、不想错过优惠活动的用户。
- xanthichi/mcd-meal-mateTypeScript · 127 Stars
智能搭餐、营养匹配、优惠整合
查看原始资料
作者描述
M餐搭子：基于麦当劳 MCP 实时菜单与营养数据的智能搭餐 Skill｜在线演示已上线
编报时的 AI 中文解读
M餐搭子是一个基于麦当劳MCP实时数据的智能搭餐Skill，输入城市、地标、预算、热量上限和蛋白质目标，从当前门店的真实菜单中选出三套可执行的组合，并展示价格、热量、蛋白质和MCP调用链。适用于工作日午餐时间有限的上班族、有预算和营养目标的学生和健身人群，以及想控制热量但仍希望保留选择感的人。
- xx20050823xx/McMatePython · 123 Stars
套餐搭配、预算控制
查看原始资料
作者描述
询问用户就餐餐厅并根据餐厅菜单用户就餐人数及就餐预算帮用户选择套餐。
编报时的 AI 中文解读
这是一个帮助用户根据预算、人数和忌口在麦当劳选择套餐的Skill，包含交互流程和搭配引擎。适用于选择困难、预算敏感或帮人代点的用户。
- tractordev/webframeGo · 120 Stars
网页自动化、模型本地部署
查看原始资料
作者描述
Run a web page like a program
编报时的 AI 中文解读
webframe 是一个命令行工具，用于在隐藏的 webview 中加载网页，并将页面控制台输出连接到终端，支持从标准输入评估 JavaScript，适合运行浏览器专用代码、后台运行页面或与页面交互。
- tian1326/mcd-coupon-assistantPython · 116 Stars
优惠券比价、门店查询、菜单搭配
查看原始资料
作者描述
麦当劳优惠券省钱助手 — 基于麦当劳中国官方 MCP 的 WorkBuddy Skill：查门店/查券/带券算价/比价推荐。默认不下单、不自动领券。
编报时的 AI 中文解读
基于麦当劳中国官方MCP服务的WorkBuddy Skill，帮助用户查询附近门店、获取优惠券、计算带券价格并推荐最省钱方案，默认不下单不自动领券。
- xiaozhangup/Oh-My-McDonaldPython · 116 Stars
餐品推荐、门店比价、优惠查询
查看原始资料
作者描述
基于麦当劳官方 MCP 的对话式 AI Skill，支持餐品推荐、门店比价、优惠查询、营养计算和探索记录，并通过 Python 生成体素风格分享图片。
编报时的 AI 中文解读
这是一个基于麦当劳官方MCP的对话式AI Skill，通过自然语言提供餐品推荐、门店比价、优惠查询、营养计算和探索记录功能，并利用本地Python脚本生成体素风格分享图片。适合经常吃麦当劳、想根据预算和口味选择餐品、比较价格或记录食用历史的消费者，也适合对MCP和AI Skill感兴趣的开发者。
- JellyFlow/Mcdonalds_in_RokidGlassesTypeScript · 115 Stars
语音点餐、购物车管理、实时算价
查看原始资料
作者描述
麦当劳 × Rokid Glasses 眼镜点餐助手：一句话点餐、MCP直连、购物车管理与实时算价
编报时的 AI 中文解读
面向Rokid眼镜用户的语音点餐助手，通过一句话点餐、MCP直连麦当劳，支持购物车管理和实时算价，用户按键确认提交。
- mokevnin/dev-quality-gatesPython · 105 Stars
工具推荐
查看原始资料
作者描述
An opinionated, curated list of tools for enforcing software development standards: one recommended tool per job, with limits.
编报时的 AI 中文解读
这是一个精选的软件开发标准执行工具清单，为每项任务推荐一个工具，并说明其适用边界。适合希望快速为项目选择合适质量门禁工具的开发者或团队。
- hffgzktbcc-blip/ounce-coreRust · 104 Stars
Switch2模拟
查看原始资料
作者描述
Clean-room Nintendo Switch 2 (Tegra T239 / Horizon OS FW 23.0.1) Emulator Engine in Pure Rust
编报时的 AI 中文解读
ounce-core 是一个用纯 Rust 编写的任天堂 Switch 2 模拟器引擎，能够启动并执行未修改的 Switch 2 原生 AArch64 二进制文件（如 .xci、.nsp 等）以及捕获的固件机器码。它面向对 Switch 2 硬件和系统软件感兴趣的开发者、研究人员和模拟器爱好者。
- ChinaOpenClaw/mcd-nutrition-expertPython · 104 Stars
套餐推荐、营养计算、点餐下单
查看原始资料
作者描述
身材管控期也想吃麦当劳？这个专家会帮你根据每日消耗搭配麦麦套餐。基于 SP 矩阵理论，打通麦当劳官方 MCP Server 的营养选型与点餐闭环（35 个官方工具 · 14 个 REST 接口 · 零第三方前端依赖）。
编报时的 AI 中文解读
这是一个基于SP矩阵理论的麦当劳营养膳食搭配工具，帮助用户在身材管控期根据每日消耗推荐套餐，并打通点餐闭环。适用于体重管理、健身增肌、控钠需求及不想做功课的普通用户。
其余 12 个项目见收录列表。