RelayZero: Offline Edge-AI Semantic Compression for Disaster Mesh Networks

Description and use

RelayZero compresses emergency dispatches to 25 bytes, enabling reliable transmission over low‑bandwidth mesh radios.

Problem

Centralized telecoms fail during disasters, leaving rescue teams with low‑bandwidth sub‑GHz mesh radios that choke on standard voice or text traffic.

Outcome

RelayZero enables reliable, low‑bandwidth emergency communication by compressing dispatches into a 25‑byte packet, achieving ~88% data reduction and allowing critical alerts to traverse fragile mesh radios.

Workflow

• 1. Responder speaks or types dispatch into Python UI (transmitter.py).
• 2. Local Gemma 4:E2B model (via Ollama) parses input, extracts Priority, Location, Hazard, Casualties, Requirements, and serializes into 25‑byte string.
• 3. Packet sent over local TCP socket (simulating LoRa UART).
• 4. Base station (base_station.py) receives packet on port 5000, maps coordinates on offline OpenStreetMap, draws 150‑m danger perimeter.
• 5. If packet flagged CRITICAL, frontend synthesizes 2‑second police siren using Web Audio API.
• 6. Triage officers export queue to CSV for post‑operation forensics.

Evidence

RelayZero is a 100% offline, edge‑AI semantic compression protocol.

Source

It uses Gemma 4 : E2B to strip away conversational filler, extract the vital triage facts, and compress the dispatch into a dense 25‑byte telemetry packet formatted as: PRIO:CRITICAL | LOC:LIBRARY | HAZ:FIRE | CAS:2 | REQ:MEDICS

Source

This intelligent edge compression yields an ~88% reduction in data size, allowing life‑saving SOS alerts to slip effortlessly through low‑power mesh radios.

Source