Office Energy Monitor

Office Energy Monitor tracks the energy usage of 15 devices across 3 rooms in real time. A live dashboard is driven entirely by Server-Sent Events — no polling — and a Discord bot (!status, !room, !usage) answers with LLM-phrased summaries backed by 3-key Groq API failover. Both interfaces share one backend, so they can never contradict each other. A centralized alerts engine posts deduplicated after-hours and prolonged-usage alerts to Discord, and energy accounting is persisted to PostgreSQL with explicit downtime-gap handling so totals survive restarts. The whole system runs under Docker Compose.

Device energy usage in an office is invisible until the bill arrives; polling dashboards waste resources and lag behind reality, naive alerting spams the same warning repeatedly, and in-memory accounting silently loses data every time the server restarts.

One backend as the single source of truth, pushing state to an SSE dashboard and a Discord bot simultaneously; a centralized alerts engine that deduplicates before posting; and restart-safe accounting persisted to PostgreSQL with explicit handling of downtime gaps, so recorded totals stay honest across crashes and deploys.

I wanted to design a real-time system properly — one source of truth serving two very different interfaces, and accounting that stays correct across restarts instead of just looking correct while the process is up.

• React
• Node.js
• Express
• PostgreSQL
• Prisma
• SSE
• discord.js
• Groq LLM API
• Docker

Tried this project or read the case study? I'd love to hear your thoughts.