尚未偵測到更新。GitHub 發布版本後會顯示在這裡。
個人 GitHub 帳號下公開儲存庫的擁有者可以登入並驗證所有權,自行管理介紹,無需管理員核准。
登入並產生驗證檔案,在 30 分鐘內將其提交至此儲存庫的預設分支。其他分支或分叉儲存庫中的檔案無效。然後在下方檢查檔案,即可無需管理員核准直接管理專案。請保留檔案以維持管理權限。
如需登入,將前往登入頁面。登入後請回到此處,再次點擊所選的驗證按鈕。
在專案資料夾中執行 /nomorevibe verify,即可驗證所有權並自行管理介紹。
一分鐘完成驗證Resume parser micro-agent: PDF in, scored candidate profile out (Next.js, Groq).
翻譯準備中
取自儲存庫 README
正在載入完整 README。
檢視完整 README ↗A single-page Next.js app for the Prolanz BOS platform. A recruiter drops in an applicant's resume PDF and gets a structured candidate profile: contact details, core skills, work history, education, a 0–100 fit score and a short recruiter verdict.
pdfjs-dist (loaded on first use).
Ligatures and run spacing are rebuilt from glyph positions, so "fi ve" reads "five".app/api/parse) via groq-sdk, in JSON mode.
The key never reaches the browser.normalizeProfile (lib/schema.ts): missing fields become empty values, numbers
are clamped, names the model reworded are snapped back to the resume's own words,
and roles are ordered most recent first. The UI cannot receive a malformed profile.npm install
cp .env.example .env.local # then paste a free key from https://console.groq.com/keys
npm run dev # http://localhost:3000
| Variable | Required | |
|---|---|---|
GROQ_API_KEY | yes | Server-side only. |
GROQ_MODEL | no | The model to try first. |
The PRD names llama-3.3-70b-versatile. Groq has retired it for some keys, so the
route tries it first and falls back to openai/gpt-oss-120b, with low reasoning effort.
In testing, that model parsed a real two-page resume in about 1.8 s. It remembers which
model answered, so a retired one costs a single extra round trip.
Import the repo in Vercel, add GROQ_API_KEY under Project → Settings →
Environment Variables, and deploy. Nothing else is needed. Groq's free tier means
no API charges, though it is rate-limited. When the limit is hit, the app says so
and asks the user to wait a few seconds.
app/
page.tsx the single page
api/parse/route.ts Groq call, model fallback, error mapping
components/
Workspace.tsx state: upload → read → parse → profile; the candidate list
UploadZone.tsx drag-and-drop, file picker, sample button
CandidateProfile.tsx the dashboard
Shell.tsx sidebar / phone drawer
ui.tsx panels, stat cards, pills, buttons, score ring
lib/
pdf.ts PDF → text in the browser
prompt.ts system prompt and models
schema.ts CandidateProfile type and normalizer
sample.ts the sample resume
The interface follows a vendor-portal style: roomy white panels on a soft page,
hairlines instead of shadows, and pill actions in one red accent. Colours are CSS
variables in app/globals.css; change --accent-hue to re-hue everything.