How to build AI-powered VS Code extensions with local models
summary
Case study of a VS Code extension that embeds an OpenMetadata catalog and Airflow lineage, queries Elasticsearch and summarizes results with Gemini, and adds a real‑time offline dictation feature using Whisper.js and transformers.js that downloads ~200 MB model weights once. Claims: reduce context switching (~40%) and cut typing time up to 4×; internal rollout shows adoption growth within five weeks. Code is open source and runs locally with no external data egress.
AI-written summary · about
chapters (7)
- 0:00AI IDES and workflow
- 1:31The role of VS Code
- 3:25Data developer challenges
- 6:12Building the catalog tool
- 7:52Implementing local AI
- 10:43Live extension demo
- 12:21Real world impact
related talks
37mEnhancing VS Code extensions with GitHub Copilot
11mAll about AI Toolkit for Visual Studio Code
45mDevelopment of VSCode extensions
58mWriting VS Code extensions for fun and profit
26mAI-powered Productivity using Open-Source LLMs
12mAI-generated recap of Devoxx Belgium 2024 opening keynote : Genesis
29mUpgrade Your Toolbox with IDE Extensions
06h 22mUsing AI to Enhance Developer Productivity