JWIS Winning System (Jakarta Waste Intelligence System)

AI command center prototype for the AI Open Innovation Challenge 2026 DLH waste case (Case 1 & Case 2).

What This Replaces

This version replaces the legacy Streamlit code with a high-performance React/Vite command center, a Python FastAPI backend, and a lightweight Node.js WhatsApp Gateway (Baileys). The interface is designed with a premium, professional SaaSAble layout, following the impeccable visual hierarchy and taste-skill typography standards.

Project Architecture

• Frontend: React + Vite, MapLibre GL, and CSS variables for a clean SaaSAble dashboard (Inter & Plus Jakarta Sans).
• Backend: FastAPI, OR-Tools CP-SAT (Integrated Planning optimizer), Prophet + XGBoost (Waste Forecast models), and an OpenAI Assistant route configured to stream via 9Router.
• WhatsApp Gateway: A standalone Express + @whiskeysockets/baileys gateway running on port 2785 for direct WhatsApp alert dispatching (no Puppeteer/headless browser overhead).

Complete Demo Flow

1. Sign In: Enter username dispatcher and password dispatcher-demo-pass.
2. Fleet Operations (Case 1):
   • View the full-width Live Fleet Map with real-time GPS coordinates.
   • Observe that T-047 is off-corridor (marked in yellow).
   • Click the A Simulate Jam* button. Watch T-047 dynamically calculate a new road-following route to TPA Bantargebang.
   • Look at the TPA Queue and staggered dispatch slots.
   • Click Send Alert in the Action Queue to dispatch the reroute instruction.
   • Open /field in another tab, log in as driver, and confirm the dispatch instruction.
   • Back in Fleet Operations, the confirmation is synced instantly.
3. Waste Forecast & AI Assistant (Case 2):
   • Navigate to Waste Forecast.
   • Browse/filter the 42 Kecamatan map/list. Click a Kecamatan to expand its 3-column resource optimization dashboard (predicted waste tonnage, fuel consumption, carbon emissions, crew, and fleet mix).
   • Use the Operational AI Assistant at the bottom: type a question, and it will respond via the 9Router gateway using custom domain knowledge.
4. Integrated Planning:
   • Review constraints and approve the weekly staggered queue plan.
5. Data & ML Audit:
   • Audit the Prophet/XGBoost models' accuracy metrics (WAPE, MAE), training limits, and data provenance.

Installation & Running Locally

1. WhatsApp Gateway (Baileys)

Make sure Node.js is installed. Run the gateway server:

cd "jwis/backend/wa-gateway"
npm install
node server.js

Note: A QR code will display in the terminal. Scan it with your WhatsApp app (authenticated as 6289675877496 or any driver phone).

2. Backend API

Make sure Python (3.10+) is installed. Install dependencies and run the API:

cd "jwis/backend"
pip install -r requirements.txt

# Create a .env file inside jwis/backend/ with the following:
# OPENAI_API_KEY=your-9router-api-key
# OPENAI_BASE_URL=http://100.67.31.81:20128/v1
# OPENAI_MODEL=graphify
# OPENWA_BASE_URL=http://localhost:2785/api
# OPENWA_API_KEY=your-wa-api-key
# OPENWA_SESSION_ID=default

python -m uvicorn app.main:app --port 8001

Note: The backend will warm all Prophet + XGBoost prediction caches on startup (~20-25 seconds) to ensure instant responses.

3. Frontend Web App

Run the production build preview (optimized layout):

cd "jwis/frontend"
npm install
npm run build
npm run preview -- --port 5175

Access the app at: http://localhost:5175

Testing & Verification

Run the full end-to-end Playwright tests to verify zero regressions:

cd "jwis/frontend"
npx playwright test --workers 1

(All 42 tests will pass successfully in headless mode).