REI - Riopele Energy Intelligence

Riopele Energy Intelligence (REI)

Problem

Riopele, a Portuguese vertical textile company, spends about €7.5 million a year on energy, roughly 7.6% of its turnover. Dozens of meters measure that energy, but the data is spread across factory equipment, databases and production systems, and the cost arrives once a month on a supplier invoice. Questions such as "where is our energy going right now?" or "which section used more this week, and why?" take spreadsheets, time and guesswork.

Solution. REI is a web platform that brings every meter into one place, collects readings automatically, explains consumption and allocates its cost.

Features

• Meter acquisition: Modbus TCP, validated read-only SQL queries, and virtual meters built from safe formulas.
• Scheduling: each meter has its own cron schedule that triggers the read and computes consumption per interval.
• Time series: hourly, daily and monthly views with sum, average, minimum and maximum.
• Dashboards: fully customisable per user, with six widget types, templates and drag-and-drop, all on one single source of truth.
• Electricity invoices: upload a PDF or image, and Grok reads the period, kWh and amount. The user confirms, and the cost is split to the cent across meters by measured kWh.
• Deviation alerts: automatic emails when consumption deviates from normal.
• Security: JWT authentication and role-based access and read-only users.

AI

REI has a Grok layer built on the official xAI SDK. It answers questions about consumption, generates insights compared with the previous period, and chats about the dashboard. It also builds cron schedules, formulas and SQL from plain language. It works only on the real meter catalogue and real data. Every output is validated before it is applied, requests are rate-limited, and answers come in the user's language.

Additionally REI watches consumption on its own, so nobody has to dig through every data source looking for problems. It compares each meter's readings with that meter's usual behaviour, and when consumption deviates or peaks, it emails the right people automatically. The email says which meter it is, when it happened and how far consumption is from normal. Problems that used to surface weeks later on the invoice now reach the team while they can still act.

Architecture. Data flows from the meters to a per-meter scheduler, into a MySQL time series, through a Node.js/Express REST API in TypeScript, and on to a React web app with the Grok layer. SQL runs on a separate read-only connection with a strict whitelist, and formulas use our own parser and never execute code. REI is packaged with Docker, with continuous integration and a staging environment on Google Cloud.

Tracks covered

• Data handling: multi-source acquisition into one source of truth.
• Internal processing: invoice reading and cent-exact cost allocation.
• Automation: scheduled readings and deviation emails.
• Other business needs: cost allocation per meter, roles and read-only access.
• Tools and features: custom dashboards and Grok assistants.

Impact

REI gives Riopele continuous visibility on a €7.5 million cost. Invoices are allocated automatically instead of in spreadsheets, alerts replace manual analysis, and every user has their own view of the same numbers. The is room for several other features to be implemented, which measures the great potential the solution has.

Useful links:

https://github.com/riopele/spaceXAI_BACKEND/

https://github.com/riopele/spaceXAI_FRONTEND/