Enter Hub →

The Phoenix Architecture

P31 2.0 represents an evolutionary leap in neurodivergent support technology. Three pillars, one unified system. Edge-native. Open source. Actually secure.

🔮
PHOS Oracle
Predictive Intelligence
v2.0 Live
🕸️
PHOS Weave
Collaborative Mesh
v2.0 Live
👤
PHOS Avatar
Embodied Presence
v2.1 In Progress

Live Agent Fleet

10 Active 12 Total
PHOS Oracle
Predictive Intelligence
Spoon forecasting and crisis prediction
v2.0.0
Health →
PHOS Weave
Family Mesh Coordinator
Spoon lending and collective crisis response
v2.0.0
Health →
K4 Cage
Family Mesh
Secure family tetrahedron coordination
Health →
K4 Personal
Personal Agent
Individual mesh pillar agent
Health →
K4 Hubs
Context Router
Life-context hub management
Health →
Orchestrator
System Coordinator
Cross-worker coordination and health monitoring
Health →
Ecosystem Bridge
External Integration
Third-party service integration
Health →
Agent Hub
Agent Registry
Central agent coordination point
Health →
Buffer API
Communication Filter
Message voltage scoring and rewriting
Health →
Geodesic Room
Collaborative Space
Shared 3D workspace for mesh members
Health →
PHOS Avatar
Embodied AI
Voice-first embodied assistant (v2.1 in progress)
v2.1-beta
Node Zero
Hardware Bridge
ESP32-based sensor hub (firmware v1.2)
v1.2
🔮
PHOS Oracle

Predictive Intelligence

Continuous biometric + behavioral pattern analysis enables pre-crash intervention. 30-minute, 2-hour, and 12-hour predictions using LSTM-inspired models.

https://phos-oracle.trimtab-signal.workers.dev →

Key Features

  • Spoon depletion forecasting
  • Crisis probability calculation
  • Calcium correlation detection (hypoparathyroidism)
  • Circadian rhythm analysis
  • HRV biofeedback integration

Components

Pattern Analyzer Prediction Engine Crisis Predictor
🕸️
PHOS Weave

Collaborative Mesh

Family K4 vertices share anonymized spoon states with EigenTrust-based task delegation. Spoon borrowing protocol enables collective crisis response.

https://phos-weave.trimtab-signal.workers.dev →

Key Features

  • Family mesh synchronization
  • Trust-based spoon lending
  • Automated crisis response teams
  • EigenTrust reputation system
  • Edge-native secure coordination

Components

Mesh Sync Spoon Lending Collective Response
👤
PHOS Avatar

Embodied Presence

Persistent 3D avatar in Geodesic Room with voice-first interface. Procedural animation reflects spoon state and emotional condition.

https://p31ca.org/centaur →

Key Features

  • 3D avatar with aura visualization
  • Voice interface with Whisper.cpp WASM
  • Procedural emotional animation
  • Real-time mesh presence
  • SOULSAFE safety constraints

Components

Avatar Core Voice Interface Aura Renderer

Technology Stack

Frontend
Astro, React, Three.js, Tailwind CSS
Edge
Cloudflare Workers, Durable Objects, KV, R2
AI/ML
ONNX Runtime, Ollama fleet, Whisper.cpp WASM
Database
D1 (SQLite), Durable Objects
Build
pnpm workspaces, Turbo
Hardware
ESP32, LoRa, Node Zero (v1.2)

Active Worker Fleet

PHOS Oracle
Spoon forecasting
Health Check →
PHOS Weave
Family mesh
Health Check →
K4 Cage
Family coordination
Health Check →
K4 Personal
Individual agent
Health Check →
Orchestrator
System coordination
Health Check →

Data Flow Architecture

Biometric Stream → Pattern Analyzer → Prediction Engine → Family Mesh
                                     ↓
                              Crisis Predictor → Collective Response
                                     ↓
                              Avatar Aura → Voice Interface → User
      

Zero-knowledge architecture: Sensitive biometric data never leaves your device. Only anonymized pattern metadata flows to the mesh.

Security & Privacy

🔐 Zero-Knowledge
Sensitive biometric data never leaves your device
🛡️ PQC-Ready
Post-quantum cryptography for verifiable credentials
🏠 Edge-Native
No data center required - runs on Cloudflare edge
📜 Open Source
Full code transparency - auditable by anyone
Security Program →

Explore the Code

P31 is fully open source under AGPL v3. Explore, fork, and contribute. 22 research publications available on Zenodo.

View on GitHub →