尚未检测到更新。GitHub 发布版本后会显示在这里。
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。
一分钟完成验证正在加载完整 README。 以下为已保存的摘录。
A from-scratch rebuild of azaisai.com — an AI video/image
generation product — built as a timed take-home exercise. Full research, the
implementation plan, and every prompt/response exchanged with the coding agent while
building this are committed in this repo (see below).
Same sitemap, same generation UX, same credit economy as the original. The AI models
underneath are real, working, free-tier providers (Pollinations)
rather than licensed Sora/Veo/Runway access — that swap is disclosed in the UI, not
hidden behind the original's branding.查看完整 README ↗A from-scratch rebuild of azaisai.com — an AI video/image generation product — built as a timed take-home exercise. Full research, the implementation plan, and every prompt/response exchanged with the coding agent while building this are committed in this repo (see below).
Same sitemap, same generation UX, same credit economy as the original. The AI models underneath are real, working, free-tier providers (Pollinations) rather than licensed Sora/Veo/Runway access — that swap is disclosed in the UI, not hidden behind the original's branding.
Next.js (App Router) + TypeScript + Tailwind CSS, Postgres via Prisma, custom email +
one-time-code auth (no third-party auth service), Pollinations for images and Pixazo
for video. No Stripe, no Supabase — see docs/implementation-plan.md
for why and what that trades off.
Video generation is submit-then-poll rather than one long request: a serverless
function can't hold a connection open for the 1-2 minutes a video takes, so the route
that starts a generation returns immediately, and GET /api/generations/[id] (already
polled by the client every few seconds) advances the job by one quick status check each
time it's called. No background task ever needs to outlive a single request.
npm install
cp .env.example .env.local # fill in the API keys, SESSION_SECRET, and DATABASE_URL
npx prisma migrate deploy
npm run dev
DATABASE_URL needs a real Postgres connection string, there's no SQLite fallback
(SQLite's local file doesn't survive Vercel's serverless filesystem, so the whole app
targets Postgres from the start). The easiest source for one: deploy this to Vercel
first (see below), use the connection string it hands you for local dev too, one
database instead of two setups to keep in sync.
DATABASE_URL automatically..env.example in the project's Environment
Variables settings: POLLINATIONS_API_KEY, PIXAZO_API_KEY, SESSION_SECRET.npx prisma migrate deploy once against that same DATABASE_URL (locally,
with it set in .env.local) to create the tables, then deploy.docs/research/ — product research: sitemap, pricing, models, FAQ,
screenshots, notes on content bugs found in the live product.docs/implementation-plan.md — the plan this was built from..agent-logs/ — automatic, unedited prompt/response capture from the
coding sessions that built this (see CAPTURE-TEST.md for how it's wired).