StadiumOps Copilot (FIFA World Cup 2026)

GenAI-enabled copilot that enhances stadium operations and the tournament experience for fans, organizers, volunteers, and venue staff during FIFA World Cup 2026-style match days.

Problem statement → feature mapping

Challenge pillar	How this app addresses it
Navigation	Step-free/fastest route engine over a stadium graph (copilot/routing.py), turned into plain-language directions by the LLM
Crowd management	Simulated real-time gate density/queue/incident telemetry + risk scoring, surfaced in the Ops Dashboard
Accessibility	Step-free routing, accessible-queue timing, Sensory Relief Room + Guest Services routing, plus a WCAG-oriented UI (high-contrast/large-text modes, focus outlines)
Transportation	Transit / Park & Ride / Rideshare-aware routing and last-mile guidance
Sustainability	Live arrival-mode split (transit/park & ride/rideshare) and estimated CO2 avoided, shown in the Ops Dashboard; refill/waste-sorting guidance for fans
Multilingual assistance	Fan responses in English/Spanish/French/Arabic; Ops responses in English/Spanish/French
Operational intelligence	Per-gate risk scoring, incident counts, and prioritized heuristic actions feeding the GenAI layer
Real-time decision support	GenAI-generated 15-minute action plans and shift briefings grounded in the live telemetry snapshot
Audience coverage	Fans (Fan Assistant) and Organizers / Volunteers / Venue Staff (Ops Dashboard, with a "Viewing as" role selector)

What it delivers

• Fan Assistant
  • Multilingual help (English/Spanish/French/Arabic)
  • Accessibility-aware navigation (step-free routing, accessible queues, sensory support)
  • Crowd-aware reroutes using real-time (simulated) gate congestion/queue telemetry
  • Transportation + sustainability nudges (transit/park&ride, refill + waste sorting guidance)
• Ops Dashboard
  • Real-time decision support: gate risk ranking, queues, incidents (simulated telemetry)
  • Sustainability snapshot: arrival-mode split and estimated CO2 avoided
  • Role-aware GenAI action plans: 15-minute congestion reduction plan + push/PA comms snippet, tailored for Venue Staff / Volunteer / Organizer
  • Shift briefing generation: 60-second briefing from telemetry + playbook context

How GenAI is used

• A live LLM generates natural-language guidance and operations recommendations.
• A lightweight RAG layer retrieves relevant playbook/venue policy excerpts from data/knowledge/*.md and injects them into the prompt.
• A small routing engine computes step-free vs fastest routes over a sample stadium graph (data/stadium_map.json) and feeds that route to the LLM for user-friendly directions.
• Resilient by design: if the LLM provider times out, rate-limits, or is unreachable, the app returns a clear fallback message instead of crashing (see tests/test_llm_resilience.py).

Run locally

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
streamlit run app.py

Run tests

source .venv/bin/activate
python -m unittest discover -s tests -p "test_*.py" -v

Enable GenAI (choose one)

Option A: Ollama (recommended for local demo, no API key)

ollama serve
ollama pull llama3.1
export LLM_PROVIDER=ollama
export OLLAMA_MODEL=llama3.1

Option B: OpenAI-compatible

export LLM_PROVIDER=openai
export OPENAI_API_KEY="..."
export OPENAI_MODEL="gpt-4o-mini"

If no live LLM is reachable, the app runs in mock mode (UI still works, but responses won’t be genuinely generative).

Deploy to Streamlit Community Cloud

1. Go to share.streamlit.io and sign in with GitHub.
2. New app → repository sachinML/stadiumops-copilot, branch main, main file app.py.
3. In Advanced settings → Secrets, paste (optional, enables live GenAI on the cloud since Ollama isn't reachable there):

LLM_PROVIDER = "openai"
OPENAI_API_KEY = "sk-..."
OPENAI_MODEL = "gpt-4o-mini"

4. Click Deploy. You'll get a public URL to use as your "live preview" link.

Accessibility & inclusive design

The app implements accessibility on two levels:

1. Domain-level (stadium accessibility for fans)

• Step-free routing engine (copilot/routing.py) that avoids stairs when a fan indicates a mobility need
• Accessible-queue timing surfaced alongside standard queue times in the Ops Dashboard
• Guest Services & Accessibility Desk, Sensory Relief Room routing and guidance
• Multilingual assistance (English/Spanish/French/Arabic)

2. UI-level (WCAG-oriented interface design)

• Sidebar "High-contrast mode" toggle (black/white/yellow palette, targets AAA-level contrast)
• Sidebar "Large text mode" toggle (WCAG 1.4.4 Resize Text)
• Always-visible, high-contrast keyboard focus outlines app-wide (WCAG 2.4.7 Focus Visible)
• No color-only indicators: status badges (e.g. "Live"/"Mock") always pair color with explicit text
• All interactive controls (buttons, selects, inputs) have explicit visible text labels — no icon-only controls
• Configured via .streamlit/config.toml with a WCAG AA-oriented default theme

Suggested demo prompts

• Fan: “How do I get from Central Station to Section 120? I need a step-free route.”
• Fan: “Where is the sensory room and how crowded is the nearest gate right now?”
• Ops: “What should we do in the next 15 minutes to reduce congestion safely?”
• Ops: “Generate a short push notification to redirect fans away from the most congested gate.”