Use VS Code Agents, GitHub Copilot, and the MSSQL extension to build AI apps | Data Exposed

From the creator

The MSSQL extension for VS Code keeps unlocking new ways to take you from idea to a working SQL application faster. In this episode, we walk through the new agentic experience in VS Code Insiders by going from a written plan to a fully working semantic search app powered by SQL Server 2025's native vectors. We start with the TalkScout app, a CFP finder for upcoming developer conferences. The data layer is already in place, scaffolded from a one-page plan written with GitHub Copilot in plan mode: a typed schema, 91 conferences seeded with embeddings generated locally via Ollama, and a ranked vector query against SQL Server. The search UI itself is missing, and that is what we build next. From there, we open the new Agents window in VS Code Insiders and use a spec-driven workflow to generate the proposal, design, and tasks for the missing UI, then write the code, approving each diff in the Changes panel as the agent works. Once the app is alive, we use GitHub Copilot Chat in plain English to fix a closed-CFP filter directly in our only SQL file, and inspect the VECTOR(768) column live in the MSSQL Schema Designer. Learn how the new agentic experience in VS Code and spec-driven development come together to ship a real semantic search app in this episode of Data Exposed. ✅ Chapters: 0:00 Intro 1:01 Scenario overview, plan mode 4:08 New Agents experience in VS Code 5:08 App is running, how did it use openspec 7:57 openspec propose 11:00 openspec apply 13:00 investigate and

Published 2026-05-21 · Watch on YouTube ↗
Availability and playback restrictions are controlled by the original channel.