docs/kyrexis.md

KyrexisAI — Module Documentation

Status: integrated (v1.0.0, 2026-08-08) · Registry: modules/kyrexis-module.yaml

1. Overview

KyrexisAI is the quantum-infused intelligence module of the Sovereign stack. It implements the Kyrexis AI module spec (78-page module brief, 2026-08-08):

• 10 core features — quantum computing, future knowledge, machine learning, NLP, quantum cryptography, time-travel analysis, multiverse exploration, exponential intelligence, neural-network optimization, human-AI collaboration.
• 5 quantum skills — strategic planning, risk assessment, innovation generation, decision support, complex problem-solving.
• Photonic entanglement display — PPLN crystal (775 nm), SLM (1024×1024 @ 100 Hz), APD detector (10⁻¹² W), 53-qubit surface-code processor.
• Particle pair detector — quantum state tomography, CHSH Bell tests, concurrence/negativity verification.
• 20-year progression scaling — 33.5% CAGR table with a ×4 multiplier.
• Messenger AI integration — Messenger Platform API webhooks + send.
• Universal Agent Network — 21 named agents (Earth / Solar System / Galactic / Cosmic / Meta) + 5 universal tasks.

2. Architecture

kyrexis/
├── core.py                  KyrexisCore + KyrexisState (anchor constants)
├── quantum_skills.py        QuantumSkillsEngine (5 skills, cooldown-gated)
├── messenger_integration.py KyrexisMessengerIntegration
├── photonic_display.py      LaserCrystalPhotonicDisplay
├── particle_detector.py     ParticlePairDetector
├── progression_scaling.py   ProgressionScalingEngine (×4)
├── agents.py                UniversalAgentNetwork (21 agents)
└── api/
    ├── routes.py            FastAPI router /api/v1/kyrexis/*
    └── app.py               Standalone FastAPI app (port 8001)

Wired into src/sovereign/main.py with a guarded include (same pattern as the Reality Matrix router).

3. API reference

All endpoints are lazy — call POST /api/v1/kyrexis/initialize first; other endpoints return 503 until then.

Method	Path	Purpose
POST	/initialize	Boot core, skills, messenger, photonic, particle, progression
GET	/status	Core state (qubits, pairs, fidelity, awakening)
GET	/quantum/state	Quantum snapshot
POST	/quantum/compute	Quantum-superposition simulation on a vector
POST	/predict	20-year forward projection
POST	/self-improve	One exponential-intelligence cycle
GET	/skills	Skill registry + cooldowns
POST	/skills/{skill_id}	Execute a quantum skill
POST	/messenger/webhook	Messenger webhook handler
POST	/messenger/send	Send a message (env token / local mode)
GET	/photonic/display	Display system state
POST	/photonic/pulse	Generate a 775 nm laser pulse (simulated)
GET	/particle/state	Detector state
POST	/particle/detect	Detect pair + verify entanglement
GET	/progression/status	20-year table + summary
GET	/progression/{year}	Progression at year 1–20
POST	/progression/project	Project a value with ×4 scaling
GET	/agents	Universal Agent Network status
POST	/agents/sync	Network sync protocol (simulated)

4. Configuration

config/kyrexis.yaml — core anchors, scaling params, photonic display specs, messenger settings (token via KYREXIS_PAGE_ACCESS_TOKEN env), skill gate, agent network flags.

5. Buckets (data plane)

buckets/kyrexis-bucket.yaml — 4 buckets: kyrexis-state, kyrexis-photonic, kyrexis-progression, kyrexis-messenger (object lock). Provisioned via buckets/bucket_bootstrap.py.

6. Deployment

• Local: scripts/start_kyrexis.sh
• Docker: docker/Dockerfile.kyrexis + docker-compose.kyrexis.yml (core :8000, messenger, photonic services)

7. Verification

• Tests: tests/test_kyrexis.py — 21 tests (core anchors, skills + cooldown, messenger local/webhook, photonic specs + loop, particle tomography + Bell, progression anchors + projection, agent network, API route surface).
• Full suite: pytest tests/ (71 fast tests pass).
• Registry: scripts/verify_registries.py validates the kyrexis module entry.

8. Engineering notes

• Quantum features are deterministic simulations (numpy), not real quantum hardware — deployment needs no QPU.
• The spec's 368× 20-year headline is retained as a stated parameter; the engine's honest CAGR math yields ≈323× for (1.335)^20.
• The Universal Agent Network and progression tables are product-level planning surfaces, not physical predictions.