All case studies
Delivered

Wine Review Intelligence Platform

Collects expert critic reviews from four paywalled sources, matches each to the industry wine ID and runs from a chat command.

Industry
Wine Trade
Region
Europe
Type
Data platform + AI agent
Status
Delivered

Overview

A wine intelligence business needed a large, clean database of professional critic reviews. We built a scraping and enrichment platform that pulls reviews from four leading critic sources, matches every review to the LWIN11 wine identifier, checks quality automatically and is controlled through a dashboard and a plain-language AI agent.

The Challenge

The best wine reviews sit behind paywalls, logins and bot protection across several sites with different score scales. Collecting them by hand for tens of thousands of wines was impossible.

The Solution

Four-source scraping engine

For Jancis Robinson, Robert Parker, James Suckling and Decanter, each as its own plug-in module.

Wine matching and extraction

Score, critic, review text, drinking window, vintage and source URL, matched to LWIN11.

Automatic quality control

Catches paywalled rows, duplicates and stub entries.

Dashboard and AI control agent

Monitor jobs, download files and type commands like "start Decanter from row 3000".

How It Works

1

Wine list (CSV / batch)

2

Source modules

3

Authenticated scraping

4

Match to LWIN11

5

Quality control

6

Dashboard + export

Architecture

Backend

Flask API with blueprints and dynamic source dispatch.

Database

SQLite in WAL mode with resume snapshots for long batches.

Source modules

One module per critic site, each handling its own login and scale.

De-duplication

Note fingerprinting so repeated reviews are stored once.

Live updates

Socket.IO pushes job progress to the dashboard in real time.

Dashboard

React and Vite app with job control, wine list and file export.

Engineering Highlights

  • Batches of 5,000 wines with resume support, so a long run never restarts from zero.
  • Handles four different score scales (/20 and /100) in one consistent schema.
  • Codebase audit and redesign: fixed CORS, removed hardcoded secrets and committed cookies, added a modern dashboard.
  • Docker deployment on a VPS with a custom domain.

Outcome

  • 35,000+ wines processed across four critic sources.
  • 12,000+ verified reviews delivered in clean, structured files.
  • Delivered in paid milestones.

Tech Stack

PythonFlaskSQLiteSocket.IOReactVitePlaywrightDockerUbuntu VPS

More Case Studies

Free Strategy Session

Ready to Automate Your Business?

Book a free 30-minute strategy Meet. We'll analyze your processes, identify automation opportunities, and show you exactly how Nyvvo can transform your operations.