The Burrito Bot - AI-Powered Search in SQL Server 2025 | MVP Unplugged

Justin Garrett sits down with Microsoft MVP Andrew Pruski to explain and demo “The Burrito Bot,” a practical semantic search app built around SQL Server 2025’s vector search capabilities.

Overview

What the episode is about

The conversation focuses on using vector search in SQL Server 2025 to move beyond keyword matching and instead search by meaning (semantic similarity). Andrew demonstrates a sample application (“Burrito Bot”) that recommends restaurants using concepts such as “cozy atmosphere” or “romantic date spot.”

Core concepts covered

• Vector search vs. keyword search

  • Keyword search matches literal terms.
  • Vector search enables semantic (meaning-based) search by comparing embeddings.

• Embeddings and how they capture meaning

  • The episode explains how embeddings represent text (and concepts) as high-dimensional vectors.
  • Similarity between items can be computed by comparing vectors.

• Similarity and distance measures

  • The discussion calls out cosine similarity and vector distance as the basis for ranking results.

• Scaling semantic search in SQL Server

  • Best practices are discussed for scaling with vector indexes.
  • The episode highlights approximate nearest neighbor (ANN) search as a technique to improve performance at scale.

Building smarter experiences with RAG

The episode connects vector search to building AI applications using Retrieval-Augmented Generation (RAG), including using Azure AI Foundry and Azure OpenAI Service to create richer, “smarter search” experiences.

Model selection trade-offs

The conversation also touches on choosing the right model for the job, including the practical trade-off between cost vs. performance in AI solutions.

Resources

• Free Microsoft Foundry trial: https://aka.ms/devrelft
• Visualization app: https://projector.tensorflow.org
• Andrew’s repo: https://github.com/dbafromthecold/aipoweredsearch
• SQL Server vector search docs: https://learn.microsoft.com/sql/sql-server/ai/vectors?view=sql-server-ver17#vector-search