Automated EDA, interactive charts, a lightweight forecast, and an AI narrative report from your CSV/Excel. No setup. Instant results.

• Live: (https://query-lens-test.vercel.app/)
• API Docs (in‑app): https://(https://query-lens-test.vercel.app/)/apidocs
• © 2025 Himesh. All rights reserved under MIT license.

Upload a file → get answers. QueryLens runs automated exploratory data analysis (EDA), builds interactive charts, explains key insights in plain English, and projects a short forecast when a time series is detected. Open an AI narrative report in a new tab and print to PDF, or grab a server‑rendered PDF in one click.

• Upload CSV/Excel (robust encoding and date parsing)
• EDA summary: missingness, types, duplicates, quick stats
• Interactive charts: histograms, bar/pie, scatter, correlation heatmap, time series
• AI insights: strongest relationships, category dominance, skew, trends
• Quick forecast: trend + simple seasonality with MAE/RMSE/MAPE
• AI Report: clean HTML narrative, printable to PDF
• Backend PDF: ReportLab‑generated, printer‑friendly
• Mobile‑friendly UI, light/dark theme
• Minimal REST API you can integrate anywhere

• Frontend: Vite + React + Tailwind (deployed on Vercel)
• Backend: FastAPI + NumPy/Pandas + ReportLab (host anywhere: HF Spaces/Render/Railway)
• Optional AI narrative: OpenAI/compatible model via /ai-report
• Privacy‑first: analyses cached in memory (short‑lived), not persisted by default
• Production: frontend proxies /api/* to backend to keep your origin private

frontend/
  src/
    components/   # NavBar, Logo, SummaryCards, ChartGrid, etc.
    pages/        # Landing, Upload, Dashboard, Report, Features, ApiDocs
    services/     # api.js (API_BASE -> "/api" in prod)
  index.html      # SEO, OG tags, JSON-LD
vercel.json       # /api proxy + SPA fallback
backend/
  main.py         # FastAPI with /analyze, /analysis/{id}, /ai-report, /download-report

Prereqs: Node 18+, Python 3.10+, pip

cd backend
python -m venv .venv && source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -U fastapi uvicorn[standard] numpy pandas reportlab openai
# Optional for AI narrative:
# export OPENAI_API_KEY=your_key
uvicorn main:app --host 0.0.0.0 --port 8000 --reload

cd frontend
cp .env.example .env
# For local dev, point to your local backend:
# VITE_API_URL=http://localhost:8000
npm install
npm run dev
# open http://localhost:5173

1. Ensure the frontend defaults to a relative API in production:

// src/services/api.js
export const API_BASE = import.meta.env.VITE_API_URL || "/api";

2. Add vercel.json (at repo root):

{
  "version": 2,
  "rewrites": [
    { "source": "/api/(.*)", "destination": "https://YOUR_BACKEND_URL/$1" },
    { "source": "/(.*)", "destination": "/" }
  ]
}

• /api/* is proxied to your backend (origin stays private).
• SPA fallback ensures deep links like /report/:id render.

3. Vercel project settings:

• Framework: Vite
• Build command: npm run build
• Output directory: dist
• Env var: VITE_API_URL=/api
• Deploy and attach your custom domain (optional).

Base (prod): https://YOUR_VERCEL_URL/api (proxied)

• POST /analyze (multipart/form‑data; field file)

  • Returns:

    {
      "analysis_id": "uuid",
      "summary": { "rows": 123, "columns": 7, "missing_total": 12, "duplicate_rows": 0, "dtypes": {} },
      "columns": [{ "name": "col", "dtype": "float64", "missing": 0, "unique": 12, "sample_values": [] }],
      "charts": { "histograms": [], "barCounts": [], "pie": null, "heatmap": null, "scatter": null, "timeseries": null },
      "forecast": { /* optional */ },
      "insights": ["..."],
      "detected": { "dateCol": "OrderDate", "target": "Sales", "agg": "sum", "freq": "D" },
      "forecast_reason": "",
      "preview_rows": [{ /* first 50 rows */ }]
    }

• GET /analysis/{analysis_id}

  • Returns the cached analysis (~24h in‑memory)

• POST /ai-report

  • Body: { "analysis_id": "<uuid>" } or { "analysis": { ...full analysis... } }
  • Returns: { "html": "<semantic HTML narrative>" }

• GET /download-report?analysis_id=<uuid>

  • Returns a printer‑friendly PDF (ReportLab)

cURL snippets:

curl -X POST -F "[email protected]" https://YOUR_VERCEL_URL/api/analyze
curl https://YOUR_VERCEL_URL/api/analysis/1234-uuid
curl -X POST https://YOUR_VERCEL_URL/api/ai-report \
  -H "Content-Type: application/json" -d '{"analysis_id":"1234-uuid"}'
curl -L -o report.pdf "https://YOUR_VERCEL_URL/api/download-report?analysis_id=1234-uuid"

Frontend:

• VITE_API_URL

  • Local dev: http://localhost:8000
  • Vercel prod: /api

Backend:

• OPENAI_API_KEY (optional) — enables AI narrative
• OPENAI_BASE_URL (optional) — custom endpoint/Azure/proxy
• LLM_MODEL (optional) — default gpt-4o-mini

• frontend/index.html includes title/description, canonical, OG/Twitter tags, JSON‑LD (Organization, WebSite, SoftwareApplication, FAQ).
• Add public/robots.txt and public/sitemap.xml for indexing.
• Publish helpful pages targeting intent keywords: “AI EDA tool”, “Upload CSV for data analysis”, “time series forecast online”.

• Analyses are cached in memory for convenience; not persisted by default.
• For persistence or multi‑replica, plug in Redis/DB.
• Frontend calls /api on the same origin; the backend origin stays hidden behind Vercel rewrites.

• 404 for /analysis/{id}: ID was created on a different server or cache expired. Re‑analyze or pass the full analysis to /ai-report.
• CORS errors: ensure requests go to /api (proxied), not the raw backend hostname.
• PDF download issues: use fetch → blob download in the UI to avoid exposing the origin.

• Advanced forecasting adapters (Prophet/ARIMA/LSTM)
• Saved projects & auth
• Column profiling (drift, anomalies)
• Exports (Excel/Parquet)

PRs welcome! Please open an issue for proposals, keep PRs focused, and add screenshots for UI changes.

This project is released under the MIT License.

Copyright © 2025 Himesh

See LICENSE for the full text.