Hugging Face's logo

Spaces:
ALAMDIENG
/
waste-prediction-api
Running

App Files Community
3
waste-prediction-api / docs /UI_UX_AUDIT.md
ALAMDIENG's picture
ALAMDIENG
feat(ui): complete Smart City Decision Intelligence HUD overhaul on 12-col grid with Chart.js curve and provenance audit
41b21d8
preview code
|
Raw
History Blame Contribute Delete
11 kB

AETERNA AI v4.1 — UI/UX Audit & Refactor Report

Branch: refactor/smartcity-forecast-ui
Date: 2026-08-30
Role: Senior Product Designer, UI/UX Engineer & Responsible AI Reviewer
Scope: Complete UI/UX redesign and structural refactor of the Waste Forecast Intelligence page into a professional Smart City Decision Support & Operational Command Center.

1. Executive Summary & Design Vision

AETERNA AI is an environmental decision-intelligence platform and research prototype exploring spatial-temporal waste forecasting across the 44 sub-districts (kecamatan) of DKI Jakarta.

The previous iteration suffered from a vertically stacked output architecture that dumped raw technical data, creating excessive vertical scrolling, unbalanced visual hierarchy, and confusing layout scaling on laptops and mobile devices.

This refactor transforms the Forecast page into an executive Smart City Decision Intelligence Dashboard (inspired by Palantir / municipal command center interfaces). It prioritizes Decision Clarity over Raw Data Density, ensuring that stakeholders can immediately answer:

  1. What will happen? (Tonnage volume, trend curve, daily averages)
  2. Where will it happen? (High-contrast spatial routing, sub-district baseline comparisons)
  3. How serious is it? (Operational risk thresholds: Safe / Warning / Critical)
  4. What operational response is suggested? (15T compactor fleet sizing, crew allocation, collection hours)
  5. How reliable is the data? (Unambiguous provenance labels: Live Observed, Reference, Estimated, Forecast, Simulation, Mode B Pending).

2. Audit: UX Problems Identified & Solved

# Problem in Previous UI Solution in Smart City Command Center Refactor
1 Excessive Vertical Scrolling Unified into a structured 12-column responsive layout with 6 clear hierarchical sections.
2 Confusing Grid Cell Placement Fixed CSS Grid distributing mismatched elements by establishing strict column containers (.col-8, .col-4, .col-6, etc.).
3 Buried Decision Outputs Placed 4 high-contrast KPI cards (Forecast Volume, Risk Level, Suggested Fleet, Forecast Readiness) immediately below the control panel.
4 Overly Bulky Config Panel Streamlined into a single-row control grid with quick horizon pills ([1D] [3D] [7D] [14D] [30D]) and collapsible scenario settings.
5 Action Button Equality Made RUN FORECAST the dominant primary CTA (with glow & play icon), demoting Export CSV to a clean secondary action.
6 Map Too Small / Distorted Expanded map to 8 columns (~440px desktop height), integrated Esri World Dark Canvas (zero watermark), and added a floating route status overlay.
7 Crude Daily Timeline Cards Replaced compressed text boxes with an interactive Chart.js Area/Line Chart showing curve trends, peak day, minimum day, and daily averages.
8 Plain-Text Waste Composition Converted into a segmented multi-bar visual and 2-column detailed progress bars with exact tonnages and percentages.
9 Raw Debug-Style Logistics Block Redesigned into a 6-card operational grid with clear vehicle capacity (15T), crew requirements, throughput duration, and an expandable formula drawer.
10 Misleading Statistical Confidence Renamed false "Confidence/Reliability" score to Forecast Readiness (Model Indicator) to honestly reflect model & data health rather than statistical probability.
11 Excessive Monospace & Neon Glow Established modern typography hierarchy (Outfit / Space Grotesk for UI and headings, JetBrains Mono strictly for metadata/tags) and restrained borders.
12 Mobile / Laptop Clipping Implemented fluid responsive breakpoints: 12-col desktop (>=1200px), 2-col tablet, and 1-col mobile with 0 horizontal overflow.

3. Information Architecture & Section Hierarchy 

┌────────────────────────────────────────────────────────────────────────────────────────┐
│  PAGE HEADER: Waste Forecast Intelligence (Context Pill: Menteng · 7D · Stacking v1.0) │
├────────────────────────────────────────────────────────────────────────────────────────┤
│  1. FORECAST CONTROL PANEL                                                             │
│     [Location Select] [Model Select] [Horizon Pills: 1D | 3D | 7D | 14D | 30D]        │
│     [▸ Advanced Scenario Settings (Rainfall / Headcount Auto vs Manual Toggles)]       │
│     [ ▶ RUN FORECAST ] (Dominant CTA)    [ ⬇ Export CSV ] (Secondary)                 │
├────────────────────────────────────────────────────────────────────────────────────────┤
│  2. DECISION SUMMARY — 4 COMPACT KPI CARDS                                             │
│     [ FORECAST VOLUME ]      [ RISK LEVEL ]       [ SUGGESTED FLEET ]   [ READINESS ] │
│     539.48 T [FORECAST]     SAFE [DERIVED]       40 Trucks [SIMULATION] 85.4% [MODEL] │
├────────────────────────────────────────────────────────────────────────────────────────┤
│  3. GEOSPATIAL & CONTEXT INTELLIGENCE (8 cols + 4 cols)                                │
│     Left (8 cols): Large Interactive Map (~440px) with Bantargebang Route & Overlay    │
│     Right (4 cols): Live Weather (Open-Meteo) & Local Event / Crowd Activity Card      │
├────────────────────────────────────────────────────────────────────────────────────────┤
│  4. FORECAST TIMELINE & TREND INTELLIGENCE (12 cols)                                   │
│     Chart.js Area Curve with Tooltips | Summary: Total, Daily Avg, Peak Day, Min Day   │
├────────────────────────────────────────────────────────────────────────────────────────┤
│  5. OPERATIONAL SCENARIO & WASTE COMPOSITION (6 cols + 6 cols)                         │
│     Left (6 cols): 6 Operational Metric Cards (Fleet, 15T Cap, Crew, Duration, Loads) │
│     Right (6 cols): Segmented Multi-Bar & Detailed Composition Breakdown (Organic...) │
├────────────────────────────────────────────────────────────────────────────────────────┤
│  6. HOURLY DISPATCH RISK & DATA PROVENANCE (7 cols + 5 cols)                           │
│     Left (7 cols): 24-Hour Diurnal Pressure Blocks with Tooltips & Legend              │
│     Right (5 cols): Data Transparency Table (Live, Ref, Est, Forecast) & Notice Box    │
└────────────────────────────────────────────────────────────────────────────────────────┘

4. Scientific Data Integrity & Provenance Guarantee

The UI refactor strictly adheres to Responsible AI and Scientific Transparency guidelines:

  1. Explicit Data Badges Maintained:
    • [LIVE OBSERVED]: Real-time weather data fetched from Open-Meteo API.
    • [REFERENCE]: Demographic headcount baseline from BPS DKI Jakarta.
    • [ESTIMATED]: Baseline historical normal volumes.
    • [FORECAST]: Model outputs from StackingRegressor or Chronos-T5.
    • [SIMULATION]: Deterministic logistics calculations based on prototype operational assumptions.
    • [DERIVED PROPORTIONS]: Fixed solid waste characterization percentages.
    • [MODE B PENDING]: DLH DKI Jakarta field observation data pending data-sharing agreement.
  2. No Fabricated Accuracy Claims:
    • Model readiness is framed as a Model & Data Health Indicator, not an operational efficiency percentage or real-world accuracy guarantee.
    • Logistics outputs carry the prominent notice: ⚠️ SIMULATION — NOT OFFICIAL DLH OPERATIONAL INSTRUCTION.
  3. No Retraining or Parameter Alteration:
    • The ML inference pipeline, FastAPI endpoints (/api/v1/predict, /api/v1/kecamatan, /api/v1/predict/csv), and mathematical formulas in services/logistics_engine.py remain 100% untouched.

5. Files Modified in this Refactor

  • frontend/index.html:
    • Integrated Chart.js (chart.umd.min.js).
    • Restructured page-predictor into the 12-column Smart City Command Center layout.
    • Implemented Horizon quick pills, collapsible scenario drawer, 4 KPI cards, Chart canvas, and refined provenance table.
  • frontend/style.css:
    • Implemented design tokens (--bg-primary, --bg-surface, --text-primary, --accent-primary, --status-live, etc.).
    • Added 12-column responsive grid classes and media queries for desktop, tablet, and mobile.
    • Cleaned all legacy duplicate rules and CSS invert filters.
  • frontend/app.js:
    • Implemented renderForecastChart() using Chart.js with gradient fill and summary bar calculations.
    • Added event listeners for quick horizon pills, scenario auto/manual toggles, and context badge updates.
    • Enhanced map routing to display real-time distance and travel time in the floating map overlay.
    • Preserved all existing API contracts and state handling.
  • docs/UI_UX_AUDIT.md:
    • Comprehensive documentation of design decisions, architecture, and verification.

6. Verification & Test Results

  • Automated Test Suite:
    python -m pytest tests/ -v
    Result: 14 passed in 26.41s (100% PASS)
    • Schema integrity, provenance fields, logistics formulas, and news validation all passed.
  • Multi-Device Responsiveness:
    • Desktop (1440px / 1920px): Full 12-column HUD layout with high visual density.
    • Laptop (1024px / 1280px): Compact, zero-clipping layout with smooth scaling.
    • Mobile (375px - 428px): Single-column linear layout with touch-friendly controls and zero horizontal overflow.