Deployment

LeanCore runs as a containerized application stack deployed via Docker Compose.

Production Environment

Component	URL
Web Application	app.leancoreai.com
API Server	server.leancoreai.com
Monitoring	ops.leancoreai.com

Architecture

The production deployment includes:

• Backend API -- Java Spring Boot application
• Frontend -- Vue 3 SPA served via Cloudflare Pages
• Database -- MySQL 8 with multi-tenant schemas
• Vector Store -- Apache Cassandra for semantic search
• MCP Connectors -- 20+ Node.js MCP servers
• Messaging -- Evolution API for WhatsApp
• Observability -- Langfuse for AI tracing
• Monitoring -- Prometheus + Grafana + Tempo

Health Checks

Service	Health Endpoint
Backend API	GET /actuator/health
MCP Servers	GET /health
Monitoring	Grafana dashboards

Deployment Process

1. Build -- compile backend JAR + build frontend assets
2. Docker Build -- create container images
3. Deploy -- update running containers
4. Verify -- health checks on all services
5. Post-Deploy -- reseed skills, generate routing cards if needed

Post-Deploy Steps

After deploying code changes:

1. Reseed Default Skills -- ensures new skills are available to all specialists
2. Generate Routing Cards -- creates tool discovery metadata for any new tools
3. Audit Routing Cards -- verifies all tools have proper routing metadata

Monitoring

Three pre-built Grafana dashboards provide operational visibility:

System Overview

• CPU and memory usage
• Disk utilization
• Network I/O
• Container health and restarts

Application Performance

• JVM heap and garbage collection
• HTTP request rate and latency (p50, p95, p99)
• Error rates by status code
• Database connection pool status

MCP Health Grid

• Server UP/DOWN status (green/red grid)
• Health response times
• Container restarts
• Memory usage per connector