Full-stack outage-risk prototype for HackAI 2026.

• Frontend: React + Vite dashboard (src/ JSX files)
• Backend ingestion: Python data pipeline (src/ Python package)

• EIA grid operations data (demand / generation)
• DOE OE-417 disturbance summaries
• ORNL ODIN real-time outages (county-level)

The ingestion pipeline produces a unified CSV at data/outages_latest.csv.

1. Create and activate a virtualenv:

python -m venv .venv
source .venv/bin/activate

2. Install Python deps:

pip install -r requirements.txt

3. Add .env in project root:

EIA_API_KEY=your_eia_api_key_here
DEFAULT_REGION=USA

python -m src.pipeline.build_dataset --region "USA" --hours 24 --output data/outages_latest.csv

The backend now supports a strict timestamp-ordered replay consumer for the official feed requirement.

Configuration via environment variables:

• OFFICIAL_FEED_SOURCE: csv (default) or api
• OFFICIAL_FEED_CSV_PATH: local dataset path (default data/outages_latest.csv)
• OFFICIAL_FEED_REPLAY_URL: required when OFFICIAL_FEED_SOURCE=api
• OFFICIAL_FEED_CURSOR_PATH: persisted watermark cursor path (default backend/.replay_cursor.json)

Replay endpoints:

• GET /feed/replay/status: current cursor and pending docs
• POST /feed/replay/next with { "batch_size": N }: consume next N docs in timestamp order and run NLP -> risk -> alert
• POST /feed/replay/reset: reset cursor to replay from the start

Example:

curl http://localhost:8000/feed/replay/status
curl -X POST http://localhost:8000/feed/replay/next -H "Content-Type: application/json" -d '{"batch_size": 3}'
curl -X POST http://localhost:8000/feed/replay/reset

Install Node dependencies:

npm install

Run dev server:

npm run dev

Build production bundle:

npm run build