An executive command center for an integrated energy major: one live view across Trading & Supply, Refining, and Retail — and a running answer to the question that frames the whole demo, "If the Strait of Hormuz closes, what is our margin at risk, and what do we do about it?" Built as a Databricks App for an Executive Briefing Center session.
Note: This is a public demo build. All data is synthetic and generated in-app — Helios Energy is a fictional company. Run it locally with
apx dev start, or deploy it to your own Databricks workspace as a Databricks App.
Helios is the operations cockpit for Helios Energy, a synthetic integrated oil major modeled on a Databricks lakehouse. The audience is a VP of Trading & Supply and their leadership team. Every surface is built to take an executive from a headline number to a decision in a few clicks:
- One integrated picture of the physical and commercial business — maritime crude voyages and chokepoint exposure, refinery throughput and margin, and retail fuel demand.
- A resilience spine: a Strait of Hormuz closure puts a $2.17B quarterly margin at risk (60-day downside), and the app carries that single figure consistently from the chat headline, to the scenario model, to the escalation brief, to Sol's live answers.
- A conversational analyst, Sol, backed by a Databricks Multi-Agent Supervisor that routes across Genie spaces and Unity Catalog functions over the same governed data the cockpit shows.
The landing view: a six-tile KPI pulse, a live globe of crude voyage arcs with the Hormuz flash point, and Sol docked as the integrated-operations assistant. Ask in plain language; get an executive answer with its source and a next action.
Crude flows, chokepoints, and closure exposure on an operational map, with filters by chokepoint, region, grade, corridor, and period. Toggling Close Strait of Hormuz reroutes exposed cargoes around the Cape and lights up the at-risk corridors.
| Default | Hormuz closed |
|---|---|
 |
 |
Five levers (Brent shock, demand growth, freight, transit days, refinery outage) drive a live margin-at-risk number and a decomposition across Crude Trade, Refining Margin, and Retail. The Strait of Hormuz closure preset lands on the calibrated $2.17B headline with ranked mitigations.
Turns the analysis into an escalation: a brief that states who raised it and who owns it, an audit trail across the Supply & Ops and Trading desks, and a ranked set of quantified mitigation actions.
A live diagram of how the system is wired — frontend, the FastAPI BFF, the pluggable data provider, Lakebase, the Delta lakehouse, and the Genie-backed supervisor. (Open the Architecture tab in the running app to see it.)
A presenter-driven Demo mode spotlights each surface in turn, and Sol answers live from the Multi-Agent Supervisor over the governed lakehouse data.
| Guided walkthrough | Live Sol answer |
|---|---|
 |
 |
- Stack: APX Databricks App — FastAPI + Pydantic backend, React 19 + TanStack Router (file routes) + Tailwind v4 + shadcn/ui frontend.
- Data path (this build): the app runs fully self-contained on a synthetic in-app dataset
(
HELIOS_DATA_SOURCE=mock) — global crude-voyage flows and chokepoint exposure, refinery/retail KPIs, the region scorecard, forecast series, and scenario presets are all generated inbackend/providers/mock.py. No external data source is required to run or demo it. - Wiring live data (optional): the provider layer is pluggable. To back the surfaces with real
data in your own workspace, add a provider under
backend/providers/, branch onHELIOS_DATA_SOURCEinfactory.py, and declare the workspace resources (SQL warehouse, Genie spaces, Lakebase, a Multi-Agent Supervisor serving endpoint) indatabricks.yml. - Sol: the conversational analyst is designed to route through a Databricks Multi-Agent
Supervisor (Genie spaces + Unity Catalog functions). In this public build Sol answers from the
synthetic dataset; the
margin_at_riskheadline ($2.17B for a 60-day Hormuz closure) is the single figure carried consistently across the Scenario Lab, the escalation brief, and Sol.
| Action | Command (from repo root unless noted) |
|---|---|
| Dev server | apx dev start / apx dev restart / apx dev status — http://localhost:9000 |
| Backend tests | uv run pytest -q |
| Frontend tests | cd src/helios_v2/ui && bun test |
| Typecheck | cd src/helios_v2/ui && bun run tsc --noEmit |
| Build | apx build |
| Deploy (Databricks Apps) | databricks bundle deploy -t dev then databricks apps deploy helios-energy --source-code-path <your workspace .build path> |
App source lives under src/helios_v2/ (backend backend/, frontend ui/). See CLAUDE.md for
the data-provider chain and how to wire live data.
Synthetic energy-operations demo · built on Databricks with apx