Poyo's Potion Dashboard

Inspiration & Problem

Poyo's Potion Dashboard tackles a critical logistics challenge at Poyo's Potion Factory preventing overflow disasters while detecting potion theft. With 12 magical cauldrons continuously filling, each requiring timely collection by witch couriers with maximum limited 100L carrying capacity, the factory faces a complex optimization problem. Traditional manual scheduling led to overflows, wasted potion, and worse—discrepancies between actual drain events and transport tickets suggested potential fraud. We built Poyo's Potion Dashboard to transform this chaotic operation into a data-driven, optimized system that ensures no cauldron overflows while minimizing the witch workforce needed.

What we built

Poyo's Potion Dashboard is a full-stack monitoring and optimization platform featuring a real-time dashboard with interactive map visualization, historical playback via time-slider controls, and robust discrepancy detection. The system integrates uses the provided API calls to fetch live cauldron levels, transport tickets, and network topology. It also applies sophisticated algorithms including Dijkstra's shortest-path routing and custom drain-event detection to analyze operations. Our advanced route optimizer is able to calculate the minimum number of witch couriers needed for sustainable 24/7 operations while respecting network constraints and courier capacity limits. The reconciliation feature matches detected drain events with transport tickets, flagging over- and under-reporting of cauldron drainage.

Problems we faced

Our biggest challenge was designing the route optimization algorithm from scratch. We had to balance many competing constraints like courier capacity, travel times, and overflow prevention to minimize the number of witches needed to sustain the cauldrons. After multiple iterations, we settled on a greedy approach with Dijkstra's shortest-path algorithm, but getting it to produce sustainable schedules took extensive debugging and verification. In addition, building the interactive SVG map in a user-friendly manner was also difficult, as it required custom cauldron icons, elliptical layouts, and hover interactions that work seamlessly across different screen sizes. Calculating accurate drain rates from noisy historical data was another challenge we faced because we had to account for concurrent filling during drain events and filter out sensor noise to extract meaningful patterns.