"Imagine 350mm of torrential rain paralyzing Hyderabad in under 12 hours. Streets turn into rivers. Power grids fail. Emergency hotlines receive 10,000+ unstructured panic calls. Inter-agency dispatch between Police, Fire, Hospitals, and NDRF stalls into critical bottlenecks.
RescueNet-AI autonomously ingests raw distress feeds, calculates survivor decay curves, matches real geo-tagged emergency assets, and generates a life-critical dispatch plan in under 45 seconds."
┌──────────────────────────────────────────────────────────────────────────────────────────────────┐
│ RAW DISTRESS FEEDS (SMS / Phone Calls / WhatsApp / 112 Transcripts) │
└─────────────────────────────────┬────────────────────────────────────────────────────────────────┘
▼
┌──────────────────────────────────────────────────────────────────────────────────────────────────┐
│ 🤖 10-AGENT AUTONOMOUS DISASTER ORCHESTRATION PIPELINE (CrewAI + Pydantic v2) │
│ │
│ [01 Incident] ──▶ [02 Intel] ──▶ [03 Survivor] ──▶ [04 Triage] ──▶ [05 Priority] │
│ │ │
│ [10 Master Plan] ◀── [09 Alert] ◀── [08 Risk] ◀── [07 Hospital] ◀─────┴──▶ [06 Resource] │
└─────────────────────────────────┬────────────────────────────────────────────────────────────────┘
▼
┌──────────────────────────────────────────────────────────────────────────────────────────────────┐
│ TELANGANA EMERGENCY INFRASTRUCTURE GROUNDING │
│ • 100+ Hospitals & ICU Beds • 108 Ambulance Units • NDRF Battalions • Fire Depots │
└─────────────────────────────────┬────────────────────────────────────────────────────────────────┘
▼
┌──────────────────────────────────────────────────────────────────────────────────────────────────┐
│ AUTONOMOUS MULTI-CHANNEL DISPATCH (<45 Seconds) │
│ 📲 Automated Twilio SMS • 📞 Synthetic IVR Voice Calls • 🗺️ Live Geospatial Command Map │
└──────────────────────────────────────────────────────────────────────────────────────────────────┘
flowchart TD
classDef input fill:#1e293b,stroke:#64748b,stroke-width:2px,color:#f8fafc;
classDef agent fill:#0f172a,stroke:#ef4444,stroke-width:2px,color:#f8fafc;
classDef data fill:#1e1e38,stroke:#3b82f6,stroke-width:2px,color:#f8fafc;
classDef output fill:#064e3b,stroke:#10b981,stroke-width:2px,color:#f8fafc;
IN1[Distress Calls / SMS / Audio]:::input --> API[FastAPI Ingestion Gateway]:::input
API --> ORCH[CrewAI Orchestrator Engine]:::agent
subgraph AgentPipeline ["🤖 10-Agent Swarm Execution Sequence"]
direction TB
A1["01. Incident Understanding<br/><i>Extracts coordinates, victims, hazard</i>"]:::agent
A2["02. Disaster Intelligence<br/><i>Calculates flood velocity & infrastructure risks</i>"]:::agent
A3["03. Survivor Probability<br/><i>Evaluates time-decay exposure curves</i>"]:::agent
A4["04. Medical Triage<br/><i>Assigns RED / YELLOW / GREEN triage tags</i>"]:::agent
A5["05. Priority Scoring<br/><i>Calculates composite life-threat index</i>"]:::agent
A6["06. Resource Allocation<br/><i>Matches nearest boat & ambulance depots</i>"]:::agent
A7["07. Hospital Coordination<br/><i>Reserves ICU beds & blood bank units</i>"]:::agent
A8["08. Risk Prediction<br/><i>Identifies submerged roads & route hazards</i>"]:::agent
A9["09. Alert Communication<br/><i>Synthesizes localized emergency broadcasts</i>"]:::agent
A10["10. Command Orchestrator<br/><i>Synthesizes immutable master dispatch plan</i>"]:::agent
A1 --> A2 --> A3 --> A4 --> A5 --> A6 --> A7 --> A8 --> A9 --> A10
end
ORCH --> AgentPipeline
subgraph TelanganaGIS ["🗺️ Real Regional Emergency Assets"]
HOSP[(100+ Hospitals & Trauma Units)]:::data
AMB[(108 Ambulance Hubs)]:::data
NDRF[(NDRF Flood Rescue Units)]:::data
FIRE[(Regional Fire Stations)]:::data
end
TelanganaGIS -.-> A6
TelanganaGIS -.-> A7
subgraph DispatchOut ["🚀 Autonomous Action Outputs"]
SMS[Twilio SMS Alerts to First Responders]:::output
VOICE[Synthetic Voice Phone Calls]:::output
DASH[Live Incident Command Geospatial Map]:::output
end
A10 ==> SMS
A10 ==> VOICE
A10 ==> DASH
Each agent is governed by deterministic Pydantic v2 schemas, eliminating hallucination in life-critical decision pathways:
| # | Agent | Operational Responsibility | Schema Contract / Output |
|---|---|---|---|
| 01 | Incident Understanding | Extracts structured geo-coordinates, disaster taxonomy, and victim counts from chaotic distress messages. | hazard_type, latitude, longitude, casualty_estimate |
| 02 | Disaster Intelligence | Models rainfall rates, flood propagation speed, and structural integrity of surrounding bridges and power grids. | flood_velocity_index, structural_hazard_score |
| 03 | Survivor Probability | Applies mathematical survival decay functions based on water level, exposure duration, hypothermia, and age. | survival_probability_score (0-100), critical_rescue_window_minutes |
| 04 | Medical Triage | Applies standard START mass-casualty triage protocols across reported casualties. | triage_category (RED | YELLOW | GREEN | BLACK) |
| 05 | Priority Scoring | Computes a composite multi-factor rescue priority weighting. | composite_priority_index (1-100) |
| 06 | Resource Allocation | Calculates geodesic distances to dispatch nearest available flood rescue boats, ambulances, and NDRF teams. | assigned_units, depot_id, estimated_eta_mins |
| 07 | Hospital Coordination | Queries live hospital registries across Telangana to reserve ICU beds, burn units, and matching blood units. | target_hospital_id, icu_bed_reserved, blood_bank_match |
| 08 | Risk Prediction | Evaluates route access constraints (inundated underpasses, fallen electric poles) and recommends safe detours. | impassable_roads, recommended_safe_corridor |
| 09 | Alert Communication | Generates dynamic emergency SMS messages and synthesized speech scripts in Telugu, Hindi, and English. | sms_payload, voice_ivr_script, target_recipients |
| 10 | Command Orchestrator | Synthesizes all individual agent outputs into an immutable, verifiable, and auditable Master Dispatch Plan. | master_rescue_plan_id, execution_log, audit_trail |
$ python trigger_demo.py --incident hyd-flood-001
[10:42:01] INGESTION: Distress Signal #hyd-flood-001 (Begumpet, Hyderabad | Water Level: 5.2ft)
[10:42:05] [Agent 01] Incident Understanding :: Parsed 14 victims trapped on single-story roof. Lat: 17.444, Lon: 78.468.
[10:42:11] [Agent 02] Disaster Intelligence :: Inundation rate +0.4 ft/hr. Drainage canal overflow detected.
[10:42:16] [Agent 03] Survivor Probability :: Survival Probability: 78% | Critical window: 42 mins.
[10:42:20] [Agent 04] Medical Triage :: 3 RED (Elderly hypothermia + Infant), 8 YELLOW, 3 GREEN.
[10:42:24] [Agent 05] Priority Scoring :: Priority Index: 94/100 [CRITICAL EMERGENCY].
[10:42:29] [Agent 06] Resource Allocation :: Dispatched NDRF Unit #03 (Sanathnagar) + 2 Advanced Life Support Ambulances.
[10:42:33] [Agent 07] Hospital Coordination :: Reserved 3 ICU Beds at Gandhi Hospital (ETA: 11 mins).
[10:42:37] [Agent 08] Risk Prediction :: Begumpet flyover underpass SUBMERGED. Routing via Prakash Nagar.
[10:42:41] [Agent 09] Alert Communication :: Twilio SMS dispatched to NDRF Team Lead & 108 Base.
[10:42:44] [Agent 10] Command Orchestrator :: ✅ Master Dispatch Plan #RP-9921 committed in 43.2 seconds.RescueNet-AI is grounded in vetted, regional geospatial datasets:
- 🏥 Hospitals & Trauma Units: Curated registry of 100+ public and private hospitals across Telangana with ICU bed quotas, burn units, and ventilator status.
- 🩸 Blood Banks: Real-time regional blood bank registry with blood component inventory tracking.
- 🚒 Fire Stations: Coordinates and equipment rosters of regional fire control stations equipped with high-capacity dewatering pumps.
- 🚑 108 Emergency Response Ambulances: Geographic coordinates of primary ambulance staging points.
- 🛡️ NDRF & SDRF Units: Battalion locations of specialized disaster and flood rescue squads.
Source datasets are normalized in database/emergency_assets_master.csv and seeded directly via database/emergency_assets.sql.
Languages & Core: Python 3.11+, TypeScript, SQL, Bash
Agent Orchestration: CrewAI, LangChain, Amazon Bedrock (Llama 3 / Claude 3)
Backend REST API: FastAPI, Pydantic v2, SQLAlchemy, Uvicorn
Database: PostgreSQL 15 (Production) / SQLite (Local Test Mode)
Telephony & Alerts: Twilio API (Programmable SMS, Voice IVR, WhatsApp)
Geospatial Frontend: Next.js 14, Tailwind CSS, Leaflet.js
Containerization & CI: Docker, Docker Compose, GitHub Actions, AWS EC2, AWS Amplify
git clone https://github.com/Vikram30069/RescueNet-AI.git
cd RescueNet-AIdocker-compose up --build- Interactive OpenAPI Swagger Docs:
http://localhost:8000/docs - Service Health Status:
http://localhost:8000/health
cd backend
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
pip install -r requirements.txt
cp ../.env.example .env
uvicorn app.main:app --reload --port 8000(Default configuration runs in offline mock mode; no paid API keys required for testing).
RescueNet-AI includes a comprehensive automated test suite verifying agent execution, health, and triage schema contracts:
cd backend
pytest tests/ -vtests/test_health.py::test_health_check PASSED [ 14%]
tests/test_health.py::test_list_incidents_returns_seed_data PASSED [ 28%]
tests/test_health.py::test_create_incident_success PASSED [ 42%]
tests/test_health.py::test_agent_execute_with_seed_incident PASSED [ 57%]
tests/test_health.py::test_list_hospitals_filter_by_city PASSED [ 71%]
tests/test_health.py::test_list_resources_returns_seed_data PASSED [ 85%]
tests/test_health.py::test_agent_log_populated_after_execute PASSED [100%]
============================== 15 passed in 2.14s ==============================
This project is licensed under the MIT License — see the LICENSE file for details.
RescueNet-AI • Architected by Vikram Banerjee • LinkedIn