Project RISING

Resilient public-health intelligence when climate disruption breaks normal infrastructure.

Live Dashboard · Live API · Interactive API Docs · Hackathon Submission · Source Code

RISING means Resilient Intelligent Surveillance & Integrated Next-Generation Healthcare. It combines governed ASEAN health data, resilient climate-event processing, an explainable disease-risk model, FastAPI endpoints, and an interactive Streamlit decision-support dashboard.

Quick Start

From a fresh clone with Python 3.11+:

python -m venv .venv
# Windows: .venv\Scripts\activate
# macOS/Linux: source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install -r requirements.txt
python -m pytest -q

Start the API:

python -m uvicorn main:app --reload

Open http://127.0.0.1:8000/docs, or get a real sample prediction:

curl http://127.0.0.1:8000/api/v1/disease-risk/sample

For a pilot-like local deployment, copy .env.example to .env, replace all placeholder secrets, then run docker compose up --build -d. Liveness is at /health, dependency readiness at /ready, and API-key-protected Prometheus metrics at /metrics. See the pilot deployment runbook.

Start the dashboard in a second terminal:

python -m streamlit run demo/pipeline_operations_dashboard.py

The dashboard opens at http://localhost:8501. It automatically creates the small streaming-demo outputs when they do not exist.

The Problem

Typhoons, floods, heatwaves, and rural infrastructure outages can delay public-health data. Records may arrive late, duplicate, become malformed, or disappear before decision-makers can act. Project RISING treats disruption as an expected operating condition and preserves records through validation, buffering, retry, idempotency, and dead-letter isolation.

End-to-End Solution

graph LR
    subgraph Sources [Data sources]
        A[ASEAN health CSVs]
        B[Climate and weather events]
    end

    A --> C[Batch ingestion and transformation]
    B --> D[Streaming ingestion]
    C --> E{Schema and quality validation}
    D --> E
    E -->|Invalid| F[Dead Letter Queue]
    E -->|Temporary failure| G[Local buffer and retry]
    G --> E
    E -->|Valid| H[Deduplication and checkpoints]
    H --> I[Trusted storage and PostgreSQL]
    I --> J[Feature calculation]
    J --> K[Explainable climate-health risk model]
    K --> L[FastAPI]
    K --> M[Streamlit dashboard]
    L --> N[Public-health decision support]
    M --> N

Loading

The result is both operational resilience and a visible decision-support output: a country-specific, 14-day mosquito-borne disease-risk estimate with the climate and health factors that produced it.

Disease-Risk Prediction

POST /api/v1/disease-risk/predict combines requested weather conditions with the latest available country values for malaria prevalence and infant mortality. The transparent model weights climate suitability at 70% and historical health vulnerability at 30%. It is deterministic, explainable, and intentionally labeled as a hackathon decision-support model—not a clinical or epidemiological forecast.

Example request:

curl -X POST http://127.0.0.1:8000/api/v1/disease-risk/predict \
  -H "Content-Type: application/json" \
  -d '{"country":"Philippines","disease":"dengue","temperature_c":29,"rainfall_mm":180,"humidity_pct":85}'

Example output (generated by the repository, abbreviated):

{
  "country": "Philippines",
  "disease": "Dengue",
  "forecast_window": "next 14 days",
  "risk_score": 66.9,
  "risk_probability": 0.669,
  "risk_level": "high",
  "model": {
    "name": "RISING Explainable Climate-Health Risk Model",
    "version": "1.0.0",
    "type": "deterministic statistical scoring model"
  },
  "recommendations": [
    "Escalate vector-control and case-surveillance activities.",
    "Pre-position diagnostics, treatment supplies, and response staff."
  ]
}

Run the endpoint to obtain the current exact output. The checked-in test suite verifies that the wet-weather scenario has higher risk than a dry scenario.

Dashboard

The Streamlit app provides two judge-ready views in one lightweight UI:

• Climate-health decision support: choose country, disease, temperature, rainfall, and humidity; inspect the score, level, evidence, recommendations, and API-ready JSON.
• Pipeline operations: inspect accepted events, DLQ records, checkpoints, retry recovery, record routing, and optional warehouse status.

Climate-health decision support	Pipeline operations dashboard

The deployed FastAPI includes interactive Swagger documentation at the live /docs endpoint.

Full Demo Commands

Run the batch ETL:

python -m pipelines.run_etl

Run the resilient streaming scenario:

python demo/run_streaming_demo.py

Optionally load accepted events into PostgreSQL:

docker compose up -d postgres
python demo/run_streaming_demo.py --load-postgres

Inspect warehouse rows:

docker compose exec postgres psql -U rising_user -d project_rising -c "SELECT event_id, observed_at, temperature_c, humidity_pct, rainfall_mm FROM fact_weather_observation ORDER BY observed_at DESC LIMIT 10;"

API Reference

Method	Endpoint	Purpose
GET	/health	Service health check
GET	/ready	Required data and dependency readiness
GET	/metrics	Prometheus metrics (protected when API keys are enabled)
GET	/api/v1/health-indicators	Filter processed health indicators
GET	/api/v1/climate-events	Read accepted climate events when configured
GET	/api/v1/pipeline/status	Inspect pipeline readiness
GET	/api/v1/countries/{country}/risk	Legacy comparative country score
POST	/api/v1/disease-risk/predict	Explainable climate-health prediction
GET	/api/v1/disease-risk/sample	Ready-to-show prediction sample

Example health query:

curl "http://127.0.0.1:8000/api/v1/health-indicators?country=Philippines&indicator=malaria_prevalence&limit=5"

Technology Stack

Layer	Technology
API and validation	FastAPI, Pydantic, Uvicorn
Data processing	Python, pandas, NumPy, Pandera
Decision support	Explainable deterministic statistical model
Dashboard	Streamlit, Plotly
Streaming resilience	JSONL event stream, retry queue, DLQ, checkpoints
Warehouse	PostgreSQL, SQLAlchemy, Docker Compose
Quality	pytest, GitHub Actions

Project Structure

Project-Rising/
├── api/                 FastAPI routes and services
│   ├── routes/          Health, climate, pipeline, and risk endpoints
│   └── services/        Data access and disease-risk model
├── data/                Raw and processed ASEAN datasets
├── demo/                Stream resilience demo and Streamlit dashboard
├── docs/                Architecture, governance, methodology, demo assets
├── pipelines/           Batch extract, transform, validate, and load stages
├── schemas/             Health and weather data contracts
├── sql/                 Warehouse dimensions, facts, and metadata DDL
├── streaming/           Producer, consumer, retry, DLQ, and deduplication
├── tests/               API, model, pipeline, and resilience tests
├── warehouse/           PostgreSQL connectivity and weather loader
├── main.py              FastAPI application entry point
└── docker-compose.yml   Local PostgreSQL service

Minimum Viable Product (MVP): Capability Features

Requirement	Implementation evidence	Status
ADHCRI / climate-resilient healthcare	Resilient hybrid pipeline, retries, DLQ, checkpoints	Implemented
Health and climate data integration	Processed ASEAN indicators plus climate-event scenarios	Implemented
AI / intelligent decision support	Explainable disease-risk endpoint and interactive controls	Implemented
Working backend	FastAPI app and Swagger/OpenAPI docs	Implemented
Data engineering / ETL	Batch pipeline, schema validation, metadata, warehouse loading	Implemented
Interactive presentation	Streamlit decision-support and operations dashboard	Implemented
End-to-end story	Source-to-decision diagram, API, dashboard, demo script	Implemented
Reliability and cybersecurity	Validation, isolation, idempotency, security architecture	Implemented/design documented
Testing and reproducibility	pytest suite and GitHub Actions	Implemented
UN SDG alignment	SDG 3, 9, 13, and 17	Aligned
Public deployment	API and Streamlit dashboard	Implemented

Documentation

• System architecture
• End-to-end data flow
• AI and decision-support architecture
• Security architecture
• Climate-resilience architecture
• Original resilience demo guide

Responsible Use and MVP Proficiency

• The model uses aggregated public-health data and user-supplied weather conditions.
• It provides an explainable preparedness signal, not patient-level advice.
• It has not been trained or clinically validated against outbreak labels.
• Simulated climate events prove resilience behavior; a live weather provider remains a production integration.
• Human public-health teams remain responsible for decisions and local verification.

License

Released under the MIT License.