Architecture¶

This section orients you to how ElevateNow actually runs. It links to the canonical technical artifacts — demoboards, specs, and governance documents — rather than duplicating them here.

Boundary rule in action

Architecture diagrams, component specs, and data schemas live in their source repos and published demoboards. This page links to them. If you find a diagram copied into a wiki page, flag it for removal.

The stack in one view¶

CONSUMER LAYER
  Workbench (React 19 · Cloudflare Pages)
    Underwriting Hub · Claims Hub · PSU Hub · Toolkit · Projects
      ↑
API LAYER
  Workbench Backend (Express · EC2)
    /api/submissions · /api/claims · /api/tools · /api/psu-submissions
      ↑
ORCHESTRATION LAYER
  Netflix Conductor (EC2 · proxied via Cloudflare Workers)
    GKR_Submission_Intake v5 · WC FNOL Recipe · Auto FNOL Recipe
      ↑
TOOL RUNTIME
  Flask Tool Server (EC2:7000)
    Tool_Chest/Tools/ — Python tools registered in Agentic Studio
      ↑
KNOWLEDGE LAYER
  CurationStudio (EC2:8001 · gkr-curationstudio.elevatenow.tech)
    elevatenow_gkr (MongoDB Atlas) — chunks, edges, reference data
      ↑
DATA LAYER
  MongoDB Atlas — 3 clusters
    elevatenow_gkr · submission_intake_gkr · wc_fnol_platform

End-to-end change cascade¶

A change at any layer affects everything downstream. When you touch a chunk, a tool, a recipe, or a UI component, trace this chain:

CurationStudio (chunk edit)
  → elevatenow_gkr.knowledge_chunks
    → Python tools read chunks at runtime (Flask:7000)
      → Conductor recipes orchestrate tools
        → AGENT_RESPONSES (submission_intake_gkr)
          → submission_unified view (MongoDB aggregation)
            → Workbench Backend API (Express:3001)
              → Workbench Frontend (React:5173)

Published demoboards¶

Title	Audience	Link
enCODE Semantic Layer — Executive Pre-read	External / exec	Open
enCODE Engineering Architecture	Internal team	Open
Binding Authority InfoSec Architecture	Internal / infosec	Open
Agent Fleet	Internal team	Open
AI Engineering Fleet Guide	Internal team	Open

Full asset index → Asset Library

Key infrastructure¶

Service	Endpoint	Purpose
MongoDB Atlas	`artifi.2vi2m.mongodb.net`	3 databases: gkr, submissions, claims
Conductor Proxy	`conductor-proxy.vs-ca9.workers.dev`	Workflow orchestration
Auth0	`dev-y34ckcr0xp2b4064.us.auth0.com`	Authentication
CurationStudio	`gkr-curationstudio.elevatenow.tech`	Knowledge management
Flask Tool Server	EC2:7000	Python tool runtime
Workbench (prod)	Cloudflare Pages	Frontend

Agent Fleet¶

The platform runs 21 specialized AI agents. Each agent owns a domain and has a defined input/output contract. See the Agent Fleet demoboard for the full roster.

When you need an agent for internal work, routing rules are in Ways of Working.