Stop Burning RUs: Live AI App Code Review with the Cosmos DB Agent Kit | Azure Cosmos DB Conf 2026

Microsoft Developer · Intermediate ·💻 AI-Assisted Coding ·1h ago
"Did I build it right?" — the question every developer asks before shipping an AI app to production. At scale, getting it wrong gets expensive fast: the wrong partition key, the wrong index, the wrong data model — that's RUs burning in production. Andrew Liu, Principal PM Manager for Azure Cosmos DB, shows how the Cosmos DB Agent Kit answers that question without waiting a week for a senior architect. In this demo from Azure Cosmos DB Conf 2026, Andrew walks through a real multi-agent travel-planner app — a 5-day LA family trip assistant that remembers where you left off using short-term, long-term, and vector memory stored in Azure Cosmos DB. Then he runs a live, in-editor code review of the entire data layer. You'll see: • The Azure Cosmos DB VS Code extension — query, browse, and inspect your account inline, no portal context-switch • Multi-agent memory patterns on Azure Cosmos DB: declarative facts, procedural preferences, episodic trip data, and vectors • Installing the Cosmos DB Agent Kit with one CLI command — project or global scope • How agent skills package decades of Azure Cosmos DB expertise (data modeling, partitioning, indexing, RU economics) as specialized context for your AI assistant • A live review that reads your actual code AND your Bicep IaC — correlating partition keys, index policies, and document shapes against your real data access patterns • Concrete findings: missing composite indexes on ORDER BY queries, a wrong partition key on the memories container, and a prioritized fix list ranked by RU savings • Why catching this BEFORE production is the difference between a $50/month app and a runaway bill at scale Two takeaways: 1. New to Azure Cosmos DB? You don't have to learn partitioning and data modeling alone — the Agent Kit gives you a senior architect on demand, in your editor. 2. Existing Azure Cosmos DB developer? Point it at your past projects and find performance and cost wins. 👤 Connect with Andrew Liu 📝 Andrew Liu is Principal
Watch on YouTube ↗ (saves to browser)
Sign in to unlock AI tutor explanation · ⚡30

Related AI Lessons

Beyond the Spectrogram: How AI is Finally Solving the “Cocktail Party Problem”
AI is solving the 'Cocktail Party Problem' by isolating individual voices in noisy environments, revolutionizing audio processing
Medium · Deep Learning
Async Python for AI: Building High-Concurrency AI Applications
Learn to build high-concurrency AI applications using Async Python to improve performance and efficiency
Dev.to · ZNY
How to Lower Transcription Latency in Voice AI Systems: Practical Tips
Lower transcription latency in voice AI systems to 80-150ms using streaming STT and partial transcripts
Dev.to AI
Lore as Code: How I Used SDD to'Compile' a 30-Chapter Novel
Learn how to apply software engineering principles to transmedia storytelling using AI as a compiler for complex narratives
Dev.to AI
Up next
Update and audit a finance model in Excel with ChatGPT
OpenAI
Watch →