Windows-GPU-UNLOCK
Windows GPU UNLOCK: RTX3090 UEFI port, Korean no-reset Windows boot guide, verified source and benchmark evidence
翻译准备中
个人 GitHub 账号下公开仓库的所有者可以登录并验证所有权,自行管理介绍,无需管理员批准。
登录并生成验证文件,在 30 分钟内将其提交到此仓库的默认分支。其他分支或分叉仓库中的文件无效。然后在下方检查文件,即可无需管理员批准直接管理项目。请保留文件以维持管理权限。
如需登录,将前往登录页面。登录后请回到此处,再次点击所选的验证按钮。
在项目文件夹中运行 /nomorevibe verify,即可验证所有权并自行管理介绍。
一分钟完成验证本地 macOS 软件,用于 AI 账户、特定提供商配额和模型路由预览。
AI 翻译
A local macOS app for AI accounts, provider-specific quotas, and model route previews
取自仓库 README
正在加载完整 README。 以下为已保存的摘录。
A local macOS menu bar app for AI accounts, subscriptions and quota observations. Each provider keeps its own balance, quota windows, model scope and reset semantics.
Ant 0.1.0 configures and observes resources and provides read-only model route previews. The supply engine has native modules for Z.ai, DeepSeek, Kimi and Alibaba Bailian, alongside additional observational adapters. Capabilities differ by provider.查看完整 README ↗A local macOS menu bar app for AI accounts, subscriptions and quota observations. Each provider keeps its own balance, quota windows, model scope and reset semantics.
Download Ant for macOS · Releases · Report a problem
Ant 0.1.0 configures and observes resources and provides read-only model route previews. The supply engine has native modules for Z.ai, DeepSeek, Kimi and Alibaba Bailian, alongside additional observational adapters. Capabilities differ by provider.
Real request consumption, automatic failover and usage attribution are still planned. A route preview does not send a billable inference request. The app is free to download; provider subscriptions and API charges are separate.
Download the universal DMG from v0.1.0, open it, drag Ant into Applications and launch it. Requires macOS 13 or later; includes Apple Silicon and Intel builds. A ZIP and SHA256SUMS are also available. The application is Developer ID signed, notarized by Apple and includes a stapled ticket. The published app passed macOS Gatekeeper assessment.
Gemini CLI token refresh reads the OAuth client from the installed CLI on PATH; Ant does not embed Google OAuth client credentials. If the CLI is unavailable, its fresh access token can still be read, while expired-token refresh reports an actionable error. A complete GEMINI_OAUTH_CLIENT_ID / GEMINI_OAUTH_CLIENT_SECRET environment pair can override discovery.
The app can discover supported local sign-ins and make read-only usage/balance requests. Use its account management to connect supported credentials. Configuration and database files live in ~/.config/ant; a cached panel reading lives under ~/Library/Caches/local.ant.app. No account database is included in the download. The website has no credential form or analytics.
Quit Ant and remove the application to uninstall. Your local data remains until you deliberately remove it. Never attach credentials, cookies, tokens or unredacted account exports to an issue.
Install Go matching server/go.mod and Apple's Xcode command line tools, then run:
sh scripts/build-app.sh
open bin/Ant.app
The local build is ad-hoc signed. For a universal build:
ANT_UNIVERSAL=1 ANT_VERSION=0.1.0 ANT_BUILD=1 sh scripts/build-app.sh
For Developer ID distribution, also set ANT_SIGN_IDENTITY to your installed certificate identity. Apple notarization is a separate release step; a signed local build is not automatically notarized.
Run the core checks:
cd server
go test -race -count=1 -p 1 ./...
go vet ./...
For a fixture-only browser preview, use an isolated new database:
cd server
go run ./cmd/ant-server -demo -db /tmp/ant-demo.db -addr 127.0.0.1:17890
-demo shows synthetic accounts and does not read this machine's provider sign-ins.
The Go core separates provider-native observations from normalized account/resource records, quota readings, offer eligibility and route ordering. SQLite supplies serialized writes, append-only observations and persistent preferences. The macOS panel and local web UI read the service's projections.
Design index · Current status · Router boundaries · Route preferences
The long-term goal is extensible subscription and resource management across AI providers and other service categories. This release exposes the current configuration and observation work so it can be tried and reviewed.