Skip to main content
🚀 Phoenix Rooivalk: Revolutionary SAE Level 4 Autonomous Counter-UAS Defense Platform
Phoenix Rooivalk
Documentation
For Investors
Executive Summary
Investor Summary
Pitch Deck
Market Analysis
ROI Analysis
For Engineers
Technical Architecture
API Documentation
Blockchain Architecture
System Architecture
Glossary
For Operations
Deployment Guide
Operations Manual
Training Materials
Compliance Framework
Calendar
News
Support
On this page
Loading...
Table of Contents
1. Why the System is Structured This Way
Design Philosophy: Hardware-Agnostic Edge Processing
Why Factory Pattern with Hot-Swappable Components?
Hardware Profile Detection
Multi-Backend Inference Strategy
Layered Architecture Rationale
2. Where State Actually Lives
Layer 1: Edge State (Python Detector)
Frame-Level State
Detection State
Tracking State
Targeting State
Layer 2: Backend State (Rust API)
Layer 3: Cloud State (Azure Cosmos DB)
Layer 4: Blockchain State (Immutable)
State Summary Matrix
3. How Failure Propagates
Edge Pipeline Failure Modes
Failure Mode Table
Alert Throttling
Targeting System Safety Interlocks
Backend Failure Propagation
API Startup Failures
Keeper Job Processing
Cloud Function Failure Propagation
Health Check Pattern
Cache Failures Are Non-Fatal
Failure Propagation Summary
4. What Changes Under Load or Latency
Frame Buffering Behavior
Inference Latency Scaling
Tracking Drift Under Fast Motion
Webhook Latency Impact
Cosmos DB Throughput Behavior
Blockchain Confirmation Latency
Load Impact Summary
5. Which Constraints Are Real vs Imagined
Real Constraints
Hardware Memory (Edge)
Camera Frame Rate (Hardware-Limited)
Inference Quantization Trade-off
Tracking Frame Rate Dependency
Fire Net Hardware Safety
Cosmos DB Partition Key
RPC Rate Limits
Imagined Constraints
"Must Use Coral TPU for Speed"
"Tracking Requires Kalman Filter"
"Evidence Must Go to Blockchain Immediately"
"Must Support Video Streaming"
"Cannot Run Without a Model File"
Constraint Classification Matrix
Configuration Reference
Environment Variables
Hard-Coded Constraints
Conclusion