DataSentry

DataSentry is a privacy-conscious dataset quality workspace. It profiles CSV and XLSX files, calculates an explainable quality score, surfaces high-priority risks, and exports a reusable JSON report without persisting uploaded data.

MVP workflow

1. Upload a UTF-8/CP949 CSV or XLSX workbook up to 10 MB and 100,000 rows.
2. Review completeness, uniqueness, validity, and consistency scores.
3. Inspect missing values, duplicates, mixed formats, outliers, and likely PII.
4. Explore column profiles and a sanitized eight-row preview.
5. Export the complete report as JSON.
6. Reopen recent reports and review source metadata from browser-only history.
7. Switch the complete workspace between English and Korean; the browser remembers the selected language.

Architecture

Browser (Next.js) -> multipart CSV -> FastAPI
                                      |-> Pandas profiling
                                      |-> DuckDB duplicate checks
                                      `-> JSON quality report

Uploads are held in memory only for the duration of the request. The MVP does not use a database, object storage, cookies, or third-party AI APIs. Recent report history excludes row previews and sample values and remains only in the current browser's local storage.

Run locally

Backend:

cd backend
python3.12 -m venv .venv
source .venv/bin/activate
pip install -r requirements-dev.txt
uvicorn app.main:app --reload

Frontend:

cd frontend
npm install
npm run dev

Open http://localhost:3000. The API runs at http://localhost:8000.

Verification

cd backend && .venv/bin/python -m pytest -q
cd frontend && npm run lint && npm run build

Deployment

• Deploy frontend/ to Vercel and set NEXT_PUBLIC_API_URL to the Render URL.
• Deploy the repository using render.yaml and set CORS_ORIGINS to the Vercel origin.