🌐 OMIREACH: Simulated Disaster-Response Workforce

OMIREACH is an autonomous disaster-response system designed as a workforce of specialized agents, not a chatbot. It bridges the Analysis-to-Action (A2A) Gap by observing global incidents, reasoning across competing humanitarian needs, and generating simulated robotic logistics and operational handoff artifacts.

"OMIREACH does not stop at analysis. It moves beyond 'summarizing' a disaster to 'solving' it through a high-fidelity agentic ecosystem that creates structured, field-ready incident exports."

---

🏗️ Workforce Architecture

OMIREACH utilizes the Google Agent Development Kit (ADK) to define distinct, decoupled worker lanes.

• Sentinel (The Observer): Monitors live GDACS and USGS feeds to identify and normalize global incident alerts.
• Intel (The Enricher): Adds humanitarian context via ReliefWeb and environmental data using ParallelAgent workflows.
• Triage (The Strategist): Ranks zones and selects mission targets based on urgency and impact.
• Assembly (The Architect): Designs specialized kits (Medical, Food, Shelter) using LoopAgent for iterative verification.
• Logistics (The Sequencer): Calculates efficient routes and condenses kits into compact robotic pick plans.
• Robotics (The Constructor): Assigns simulated hardware and executes the physical kit build workflow.
• Delivery (The Navigator): Determines transport modes and paths by factoring in real-time weather risks.
• Action (The Operator): Generates the Operational Incident Export and dispatches partner webhooks.
• Coordinator (The Orchestrator): Manages mission state, event ordering, and cross-agent completion rules.

---

📐 The Logic of Logistics

To ensure the Logistics agent prioritizes the highest-impact missions, the system calculates the efficiency of a proposed mission $M$ using distance $d$, weather risk $\omega$, and urgency $\mu$:

$$\text{Efficiency}(M) = \frac{\mu}{d \cdot \omega}$$

---

🚦 Queue-Based Boundaries

OMIREACH models independently deployable worker lanes. The Coordinator dispatches work through named queues to ensure a clean path to microservices:

• sentinel-observer | intel-worker | triage-worker
• assembly-worker | logistics-worker | robotics-worker
• delivery-worker | action-worker

---

🛠️ Tech Stack

• Core Reasoning: Gemini API
• Orchestration: Node.js & Google Agent Development Kit (ADK)
• Operator UI: Next.js, React, TypeScript
• Geospatial: Google Maps API (Live incident visualization)
• Data Sources: GDACS (Global Disaster Alert and Coordination System), USGS, ReliefWeb

---

🚀 Getting Started

Environment Variables

Create a .env.local file with the following:

GEMINI_API_KEY=your_key_here
GOOGLE_MAPS_API_KEY=your_key_here
AEGIS_WEBHOOK_URL=optional_callback_url
AEGIS_OUTBOX_DIR=./outbox

Installation & Run

1. Install dependencies:

   npm install

2. Start the workforce:

   npm run dev

3. Validation:

   npm run lint
   npm test

Runtime APIs

• GET /api/health: Service status.
• POST /api/missions/start: Manually trigger a new autonomous mission cycle.
• GET /api/system/queues: View a live snapshot of active worker queues.

---

🚧 Challenges & Learnings

Building OMIREACH taught us that the future of relief lies in Agentic Orchestration. We overcame the "Simulated Autonomy" hurdle by building a Coordinator that handles event ordering without human intervention, ensuring a seamless hand-off between "Data Agents" and "Mechanical Agents."

Would you like me to help you draft the Action-Worker logic to format the final Incident Export packet?