Oswaldo Gomez Kagent + M3 MCP server

Description

# AI-Powered Medical Database Queries with Kubernetes: M3 + Kagent Demo 🏥 Watch how to deploy and use M3 (MIMIC-IV MCP Server) on Kubernetes with AI agent integration! In this demo, I show you a complete GitOps workflow for deploying a medical database query system using: - **M3**: MCP server with pre-loaded MIMIC-IV demo database (100 patients, 668K+ clinical events) - **Kagent**: AI agent that queries medical data using natural language - **ArgoCD**: GitOps deployment automation for Kubernetes - **Kubernetes**: Container orchestration with intra-cluster service mesh ## 🎯 What You'll See ✅ Building containerized M3 MCP server with MIMIC-IV demo data ✅ Deploying to Kubernetes cluster via ArgoCD GitOps workflow ✅ AI agent (kagent) connecting to M3 over HTTP/SSE transport ✅ Natural language queries transforming into SQL against real clinical data ✅ Complex medical queries: patient timelines, lab trends, medication records ## 🔧 Tech Stack - **M3 MCP Server**: FastMCP-based server exposing MIMIC-IV database - **MIMIC-IV Demo**: Real de-identified ICU/hospital data (100 patients) - **Kubernetes**: Container orchestration platform - **ArgoCD**: Declarative GitOps CD for Kubernetes - **Helm**: Kubernetes package manager for M3 deployment - **Kagent**: AI agent with MCP client capabilities - **Docker/Podman**: Container runtime (supports both!) ## 💡 Use Cases Demonstrated 🔹 Patient medical record retrieval 🔹 ICU stay analysis with abnormal lab values 🔹 Medication timeline creation 🔹 Multi-table JOINs across clinical datasets 🔹 Real-time AI-powered clinical data exploration ## 📚 Resources - M3 Repository: https://github.com/rafiattrach/m3 - Fork with K8s enhancements: https://github.com/papagala/m3 - MIMIC-IV Dataset: https://physionet.org/content/mimiciv-demo/ - FastMCP Framework: https://github.com/jlowin/fastmcp ## 🚀 Try It Yourself ```bash # Clone the repository git clone https://github.com/papagala/m3 # Build Docker image (will prompt for your registry) cd m3 make all # Deploy with your Helm charts helm install m3 ./path-to-helm-charts --namespace kagent # Connect your MCP client to: # http://m3.kagent.svc.cluster.local:3000/sse