Engineering Architecture

The internal-system names on this site are labels for discrete software components, not marketing concepts. Each component below is implemented as a Netlify Function with specific inputs, outputs, and data sources. This page exists so county officials, first responders, and technical reviewers can see the engineering underneath the names.

Read this first. YANA-IGY is a supplemental alerting service built on the Netlify platform: static pages, serverless functions, Netlify Blobs for persistence, and Netlify Identity for admin access. Every named subsystem on this site maps to a source file at netlify/functions/<name>.js. Components consume public government feeds (USGS, NOAA, NWS, NASA FIRMS, NDBC) and user-opt-in device telemetry, pass results through integrity and compliance gates, and emit alerts via Twilio SMS with a Bluetooth/Wi-Fi mesh fallback. Nothing below is a black box — the function files are in this repository.

Detection & Intake Layer

USA Alert Shield

Nationwide public-feed poller.

Continuously polls federal and state hazard APIs and normalizes the responses into a common event schema.

Inputs: USGS earthquake feed, NOAA/NWS CAP alerts, NASA FIRMS satellite fire detections, NDBC buoy observations, USGS water-level gauges.
Outputs: Event records tagged with hazard type, location (lat/lon, FIPS), severity, source, and retrieval timestamp.
Cadence: Short polling intervals per feed (seconds to minutes) with per-source rate-limit respect.
Failure mode: Per-source degradation — one feed going down never blocks the others.

netlify/functions/usa-alert-shield.js

DAFT — Depth-Aware Flood Technology

Terrain-aware flood and fire spread modeling.

Combines live NOAA wind data, USGS elevation grids, and river-gauge levels to produce vector fields that indicate likely spread direction and arrival windows.

Inputs: NOAA wind vectors, USGS DEM elevation samples, USGS stream-gauge stage height, NASA FIRMS hotspots.
Outputs: Spread-vector field, slope-aware arrival estimate, inundation depth estimate (flood) or fire-front advance rate.
Uses: Feeds Threat Compass with bearing/ETA and feeds Future Sense with forecast arrivals.

netlify/functions/daft.js

DolphinGemma

Coastal sentinel scoring from live NDBC + USGS feeds.

Pulls the last ~24 h of NDBC buoy observations and cross-references live USGS submarine/coastal seismic events (≤ 800 km) to compute a coastal-threat score anchored only in real measurements. Optional caller-supplied acoustic/behavioral observations are carried through but never fabricated.

Inputs: NDBC realtime2 text feed (wave height, wave period, barometric pressure, wind), USGS FDSN event API (M ≥ 4.0, 1-day lookback), optional caller observations.
Outputs: Sentinel score with per-axis breakdown (pressure drop, wave period shift, wave height rise, NDBC tsunami indicator, nearby seismic magnitude) and a verified/clear verdict.
Uses: Feeds Master System as one of several evidence streams — it never fires alerts on its own.

netlify/functions/dolphin-gemma.js

NATURA Engine

On-device + crowdsourced fusion with live corroboration.

Accepts device accelerometer/barometer readings and citizen hazard reports, persists both to Netlify Blobs, and fuses them with live USGS seismic (≤ 300 km), NWS alerts, and Open-Meteo point weather so a raw reading is only amplified when independent public feeds corroborate it.

Actions: citizen-report stores a reporter's hazard observation; analyze (default) fuses device physics + citizen reports + live feeds into a single confidence score.
Inputs: Smartphone accelerometer (MyShake-style P-wave capture), barometric pressure, citizen reports within 75 km / 3 h, USGS + NWS + Open-Meteo live feeds.
Outputs: Per-location confidence with per-signal components (accelScore, baroScore, seismicScore, nwsScore, citizenScore, weatherAnomalyScore).
Grounding: Each signal class on the Proof-of-Concept page is backed by a DOI link — no novel sensor claims are made.

netlify/functions/natura-engine.js

Reasoning & Fusion Layer

Threat Compass

Directional geometry for alerts.

Takes a hazard event and a subscriber location, returns the geometric facts the subscriber needs — bearing, distance, spread cone, and estimated time to arrival.

Inputs: Hazard location and vector from DAFT / USA Alert Shield; subscriber county centroid or mesh-reported location.
Outputs: Cardinal direction, bearing in degrees, distance to zone edge, spread-cone width, ETA to the subscriber zone.
Constraint: Outputs facts only. It does not emit guidance, routes, or instructions — that is handled by MSAD downstream.

netlify/functions/threat-compass.js

Future Sense (FTS)

Short-horizon forecasting grounded in live public feeds.

Composes four axes (Spatial, Temporal, Causal, Convergence) directly from live USGS earthquakes, NWS alerts, NASA FIRMS fires, NDBC buoys, NOAA SPC storm reports, and Open-Meteo hourly wind/humidity/pressure. No fabricated coefficients — every axis is anchored to a real measurement.

Inputs: USGS FDSN events, NWS active alerts, NASA FIRMS active fires, NDBC buoys, NOAA SPC storm reports, Open-Meteo point forecast.
Outputs: Forecast path, speed, direction, confidence interval, and arrival ETA per county, each with the underlying evidence trace.
Consumer: TESSERACT DIMENSIONS uses FTS outputs as one of its input dimensions.

netlify/functions/future-sense.js

TESSERACT DIMENSIONS

Multi-axis composition of detection, forecast, and integrity.

Composes the Spatial, Temporal, Causal, and Convergence axes into a single reviewable object for a given alert candidate. Not a black box — each axis is the output of a concrete upstream component listed on this page.

Inputs: Threat Compass geometry, Future Sense forecast, HGIP custody chain, MSAD compliance state.
Outputs: Composite alert candidate with a per-axis trace, consumed by Master System.

netlify/functions/tesseract-dimensions.js

Ripple Principle

Concentric-ring propagation model.

Implements the prioritization rule that an alert radiates outward in concentric rings from the source, with closer rings delivered first.

Inputs: Alert origin, ring radii, subscriber locations per county.
Outputs: Ordered delivery queue so that the most-at-risk residents receive SMS/mesh messages ahead of those in outer rings.

netlify/functions/ripple-principle.js

Integrity, Compliance & Failure-Handling Layer

EID — Encryption Integrity Directive

Payload encryption and custody.

AES-256-GCM encryption plus HMAC authentication on every alert payload, with a hop-by-hop custody log from origin to SMS.

Inputs: Alert candidate from TESSERACT, per-hop identifiers.
Outputs: Encrypted payload + signed custody chain. Tampered payloads are rejected at the next gate.

netlify/functions/eid-chain.js

HGIP — Holographic Integrity Protocol

Source authenticity, freshness, and anti-false-alert checks.

Rejects any alert whose source is not on the upstream allow-list, whose SHA-256 content hash fails, whose timestamp is stale (> 10 min) or dated in the future, whose confidence is below 0.65, or whose EID envelope / signed-event custody fails verification. Adds itself as a custody hop to every signed event it sees.

Inputs: Alert candidate with alertId, source, timestamp, payload, optional EID envelope and signed event.
Outputs: verified boolean, content hash, per-check breakdown, and an audit-log reason on rejection.

netlify/functions/hgip.js

TFID — True/False Integrity Directive

Binary pre-release gate.

Hard pass/fail check just before an alert is released: required subsystems alive, confidence above floor, payload not stale, custody chain intact, HGIP & MSAD both green.

Inputs: Subsystem heartbeat table, confidence score, freshness timestamp, upstream gate results.
Outputs: Boolean allow/deny with the specific failed check named. If denied, nothing is sent.

netlify/functions/tfid.js

MSAD — Mission Standard Alert Directive

Regulatory and message-content compliance.

Enforces the "alert-to-move only" rule. Blocks guidance, route instructions, or advisory language; rewrites non-compliant messages before release.

Inputs: Candidate SMS body, template ID, hazard metadata.
Outputs: Compliant message body (or rejection with reason). Protects YANA-IGY from providing emergency guidance it is not authorized to give.

netlify/functions/msad.js

MIARP — Model Integrity Audit & Replacement Protocol

Continuous model health monitoring with persistent state.

Tracks accuracy, drift, staleness, bias, and latency of detection/forecasting models. Quarantines any model below thresholds and routes traffic to the next-best fallback. Model registry, audit log, replacement queue, and quarantine set are persisted to Netlify Blobs so the audit trail survives function cold-starts.

Inputs: Rolling accuracy / false-positive / false-negative counts, inference latency, last-update timestamp.
Outputs: Per-model status (healthy / watch / quarantined) and a routing hint for Aim.

netlify/functions/miarp.js

Liquid Technology

Adaptive breach-patching with durable patch history.

When a gate (HGIP / TFID / MSAD) or data source fails, Liquid picks an adaptive strategy — confidence boost from corroborating evidence, fallback routing, redundant verification, model swap, data interpolation, or encryption bridge. Active patches and patch history are persisted to Netlify Blobs so repairs survive cold-starts.

Inputs: Failing gate/source identifier, available alternatives, policy limits.
Outputs: Patched path that either satisfies the failed check or escalates to Aim for a system-level response.

netlify/functions/liquid-technology.js

Electric Shield

Propagation-channel integrity from live space-weather feeds.

Reads live NOAA SWPC planetary Kp-index and GOES primary X-ray flux (6-hour) to derive an interference level and engage protective propagation when needed. No caller-supplied EM level — it refuses to guess, returning 502 if both public feeds are unavailable.

Inputs: alertId, sourcePhone, hazardType. Live feeds: SWPC planetary_k_index_1m, GOES xrays-6-hour.
Outputs: shieldActive boolean, shieldStrength 0–100, Kp + X-ray class snapshot, audit record in the shield-events blob store.
Threshold: Full shield engages when derived interference ≥ 40; below that, standard propagation is used.

netlify/functions/electric-shield.js

Aim — Autonomous Integrity Monitor

System-level supervisor with durable state.

Watches the union of EID, TFID, MIARP, Liquid Technology, MSAD, and Rotating Alert System. Its job is to notice silent failure — a component that's "up" but producing bad output — and to isolate it before it touches a live alert. Subsystem registry, breach log, Enforcer directives, EID conformance snapshot, and continuous-monitoring history are persisted to Netlify Blobs so the guardian retains memory across serverless invocations.

Inputs: Heartbeats, gate decisions, per-model MIARP status.
Outputs: System-health summary, quarantine actions, and escalation events into the audit log.

netlify/functions/aim.js

Delivery Layer

Master System

Orchestrator.

Routes an alert candidate through the EID → HGIP → TFID → MSAD chain and, on success, hands it off to the Rotating Alert System for delivery.

Inputs: TESSERACT composite, gate references.
Outputs: Either a released alert with a full audit trail, or a rejection with the exact failed gate named.

netlify/functions/master-system.js

Rotating Alert System

Two-tier channel failover.

Primary channel is Twilio SMS. Fallback is the Ripple Mesh Relay (Bluetooth Low Energy + Wi-Fi Direct on opted-in phones) for regions where cellular is down. Channel state is persisted to Netlify Blobs so retries survive function cold-starts.

Inputs: Compliant alert message, subscriber roster for the target county, current channel health.
Outputs: Per-recipient delivery attempts with status; fallback to mesh on Twilio or carrier failure.

netlify/functions/rotating-alert-system.js

Ripple Mesh Relay

Offline peer-to-peer alert propagation.

Opted-in phones rebroadcast encrypted alert payloads over Bluetooth Low Energy and Wi-Fi Direct when cellular is unavailable, then re-upload when any node regains signal. Payload only — no personal data transits the mesh.

Inputs: EID-encrypted alert payload; opt-in status and battery policy on each participating device.
Outputs: Hop-by-hop delivery log; server-side re-upload of collected alerts via mesh-relay-upload.js.

netlify/functions/mesh-relay.js · netlify/functions/mesh-relay-upload.js

Vibe Text (Dual-Sense)

Accessibility delivery mode for blind and low-vision subscribers.

Delivers each alert as vibrating Braille patterns, spoken audio, or both simultaneously. Subscribers pick the mode and the speech rate at opt-in.

Runs in: The browser (PWA), using the Vibration API and the Web Speech API.
Modes: Vibration-only (silent Braille), speech-only, or both (default).

vibe-braille.js (client-side)

End-to-end pipeline for a single alert

Detection — USA Alert Shield pulls a hazard event; DAFT, DolphinGemma, and NATURA contribute supporting evidence per hazard type.
Reasoning — Threat Compass computes geometry for affected counties; Future Sense forecasts path and ETA; TESSERACT composes these into a reviewable alert candidate.
Integrity — EID encrypts the payload; HGIP verifies the source; MIARP confirms the models used are healthy; MSAD rewrites or rejects non-compliant wording.
Gate — TFID makes a binary pass/fail decision. Aim watches the whole chain for silent failure. Liquid patches recoverable failures without bypassing the actual check.
Delivery — Master System hands the compliant, sealed alert to the Rotating Alert System. Twilio SMS is primary; Ripple Mesh Relay is the fallback. Ripple Principle orders delivery by proximity to the source. Vibe Text handles accessibility modes on the receiving device.
Audit — Every gate decision and delivery attempt is written to Netlify Blobs with a timestamp and reason, so a county can audit any alert after the fact.

Public data sources, named

USGS Earthquake Hazards Program — seismic events (earthquake.usgs.gov).
NOAA / National Weather Service — CAP severe-weather watches, warnings, advisories (api.weather.gov).
NASA FIRMS — near-real-time satellite fire detections.
NOAA NDBC — moored-buoy observations for coastal threat scoring.
USGS Water Resources — stream gauges and flood stages (waterservices.usgs.gov).
Open-Meteo — supplemental weather fields where NWS coverage is thin.
Peer-reviewed precursor research — DOI-cited methodology maintained internally on the Command Center.

What this page is not. This is not a trade-secret disclosure: specific thresholds, confidence weights, and fusion rules stay inside the function source and internal documentation. What is disclosed here is the shape of each component — what it reads, what it decides, what it hands to the next stage, and the file it lives in — so that a county IT director or procurement officer can review the architecture at the level a reasonable purchasing decision requires.