Build Your Own
Vertical Application
Step-by-step guide to building healthcare, legal, property, or any vertical AI application using the 18 Core Stack services. From zero to production.
Setup & Architecture
2-3 days · Foundation setup and vertical planning
📚 New to Nexus CLI?
This guide uses the Nexus CLI extensively. If you haven't installed it yet or want to learn more about all 60+ available commands, templates, and advanced features, check out the full CLI documentation.
View CLI Documentation1.1. Install Nexus-Forge
Install the Nexus-Forge CLI which provides all development tools, Docker Compose setup, and Claude Desktop integration.
npm install -g @adverant/nexus-forge
forge --version # Verify installation1.2. Create Your Vertical Project
Choose a vertical template or start blank. Templates include data models, API routes, and example queries for specific industries.
# Interactive setup
forge create my-vertical
# Or specify a template
forge create healthcare-crm --template healthcare
forge create legal-research --template legal
forge create property-mgmt --template property
# Or start completely blank
forge create custom-vertical --template blank1.3. Understand Core Services Available
All 18 core services are automatically available. Here's what each one provides for your vertical:
Data Model & Authentication
3-4 days · Define your data structures and access control
2.1. Design Your GraphRAG Schema
Define entity types, relationships, and vector embeddings for your vertical. GraphRAG provides both PostgreSQL (structured) and Neo4j (graph) storage.
// src/schema/entities.ts
export const HealthcareSchema = {
entities: {
Patient: {
properties: ['name', 'dateOfBirth', 'medicalHistory'],
embeddings: ['medicalHistory', 'symptoms'],
relationships: ['HAS_APPOINTMENT', 'HAS_DIAGNOSIS']
},
Appointment: {
properties: ['datetime', 'status', 'notes'],
embeddings: ['notes', 'diagnosis'],
relationships: ['WITH_PATIENT', 'WITH_DOCTOR']
}
}
}
// Forge automatically creates:
// - PostgreSQL tables
// - Neo4j nodes/relationships
// - Qdrant vector collections
// - VoyageAI embedding pipelines2.2. Configure Multi-Tenant Auth
Set up workspaces, roles, and permissions. Auth service handles JWT tokens, SSO integration, and workspace isolation automatically.
// src/config/auth.ts
export const authConfig = {
workspaces: {
enabled: true,
isolation: 'strict' // Each workspace = separate data
},
roles: {
admin: ['read', 'write', 'delete', 'manage_users'],
doctor: ['read', 'write', 'view_patients'],
nurse: ['read', 'write_limited'],
patient: ['read_own']
},
sso: {
providers: ['google', 'okta', 'auth0']
}
}
// Forge generates:
// - Login/signup API routes
// - JWT middleware
// - RBAC enforcement
// - Workspace switching UIAI Integration
4-5 days · Connect LLMs and build intelligent features
3.1. Configure AI Models
Choose which AI models to use for different tasks. MageAgent provides access to 320+ models from OpenRouter, Anthropic, and OpenAI with intelligent routing to save costs.
// src/config/ai.ts
export const aiConfig = {
chat: {
model: 'anthropic/claude-sonnet-4',
fallback: 'openai/gpt-4o'
},
analysis: {
model: 'anthropic/claude-opus-4',
temperature: 0.1
},
summarization: {
model: 'anthropic/claude-haiku-4', // Faster, cheaper
maxTokens: 1000
},
routing: {
enabled: true, // Intelligent cost optimization
costSavings: '45-60%'
}
}3.2. Build Multi-Agent Workflows
Create workflows where multiple AI agents collaborate. Example: medical diagnosis workflow with symptom analysis, research, and recommendation agents.
// src/workflows/diagnosis.ts
export const diagnosisWorkflow = {
name: 'Medical Diagnosis Assistant',
agents: [
{
id: 'symptom-analyzer',
model: 'claude-opus-4',
task: 'Analyze patient symptoms and medical history',
tools: ['graphrag_query', 'vector_search']
},
{
id: 'research-agent',
model: 'gpt-4o',
task: 'Research similar cases in medical literature',
tools: ['web_search', 'document_fetch']
},
{
id: 'recommendation-agent',
model: 'claude-sonnet-4',
task: 'Generate treatment recommendations',
tools: ['graphrag_query', 'guideline_lookup']
}
],
orchestration: 'parallel', // Run simultaneously
synthesis: true // Combine results into final output
}
// Frontend shows real-time progress per agentFrontend Build
4-5 days · Create the user interface
Build your React/Next.js frontend with pre-built components for chat interfaces, data tables, forms, and dashboards. TypeScript SDK provides full type safety.
- Use Forge UI component library (50+ pre-built components)
- Implement real-time chat with WebSocket streaming
- Create data tables with infinite scroll and filters
- Add file upload with VideoAgent/FileProcessAgent integration
- Build dashboards with charts and metrics
Advanced Features
3-4 days · Add specialized capabilities
Video Processing
Medical imaging, property tours, depositions
Geospatial Features
Maps, routing, location-based search
Document Intelligence
OCR, form extraction, contract analysis
Notifications
Email, SMS, push notifications, webhooks
Testing & QA
2-3 days · Ensure quality and performance
Deployment
1-2 days · Ship to production
$ forge deploy --environment production
✓ Running pre-deployment checks...
✓ Building Docker images...
✓ Pushing to registry...
✓ Updating Kubernetes...
✓ Running smoke tests...
✓ Health checks passing...
🎉 Deployed to: https://my-vertical.adverant.cloud
📊 Dashboard: https://my-vertical.adverant.cloud/admin
📈 Metrics: https://metrics.adverant.cloud/my-verticalReady to Build?
Install Nexus-Forge and start building your vertical today. All 18 core services included.