Using Postgres to locate the best coffee near you!
summary
Uses PostgreSQL 16.4 with the Yelp open dataset (≈6.9M reviews, ~150k businesses), adds a TSVector full‑text column, then creates embeddings stored with pgvector/Azure OpenAI embeddings and a DISKANN vector index to run hybrid semantic + geospatial queries (PostGIS points) that return 20 ranked cafes near Delaware Art Museum matching “pet friendly” and “outdoor seating.” Results feed a retrieval‑augmented GPT‑4.0 LLM for grounded recommendations; a Python app and web demo are built and the approach works end‑to‑end. Yelp open dataset GPT-4.0
AI-written summary · about
chapters (9)
- 0:00Introduction
- 3:06Coffee search strategy
- 5:31Cafe dataset overview
- 8:15Basic pattern matching
- 11:40Full-text search
- 16:00Vector search fundamentals
- 22:18Semantic and hybrid search
- 28:49Retrieval Augmented Generation
- 37:45Summary and conclusion
related talks
42mUsing Postgres to locate the best coffee near you!
44mPostgres and OpenAI - a match made in Azure
45mUsing Postgres + OpenAI to power your AI Recommendation...
45mUsing Postgres + OpenAI to power your AI Recommendation...
05mCheeses, vectors and PostgreSQL
25mSemantic search with Django, PostgreSQL, & pgvector
12mThe mechanics of integrating AI and vector search into Azure Database for PostgreSQL
46mPostgreSQL as a Graph Database: Who Grabbed a Beer Together?