BurgReport: Fine Wine Pricing Intelligence

Fine-wine pricing is scattered across sources and slow to evaluate. Buyers often have to cross-reference multiple merchant listings, inconsistent critic scores, and scattered vintage guidance just to get to a usable pricing view. That slows decision-making for collectors, sommeliers, and retailers, and increases the odds of weak pricing decisions. BurgReport was built to compress that scattered research into a single benchmark view that can be used in seconds.

BurgReport acts as a specialized aggregation and pricing-reference layer for fine wine. It provides:

• Live price snapshots with average, minimum, and maximum retail pricing.
• Merchant-count visibility.
• A seen-price checker to compare an observed price against market benchmarks.
• Normalized critic consensus across major sources.
• Vintage and drinking-window context to make the pricing signal more useful.

This project was less about model development and more about live data acquisition, normalization, and fast search-to-insight UX. Key technical decisions:

• Python FastAPI backend on Railway with three routers (search, wines, vintages).
• Uses the OpenAI Responses API web-search tool (with a Tavily fallback) to fetch live pricing from public merchant listings — unvalidated estimates parsed from public web pages, not a licensed feed or Wine-Searcher integration.
• Supabase Postgres for a 24-hour TTL price cache to keep response times sub-3 seconds.
• Airtable REST API for curated Grand Cru content (descriptions, producers, pairings) with a 1-hour in-memory cache.
• Nightly Railway cron refreshes 34 Grand Crus across 5 vintages so the cache stays warm.
• Designed for near-instant utility with no login wall or onboarding friction.