Input Sanitization → Semantic Analysis → Behavioral Analysis → Meta-Observer (O₂)

┌──────────────────────────────────────────────┐
│  LAYER 1: Input Sanitization                 │
│  Deobfuscation (homoglyphs, leetspeak, etc.) │
├──────────────────────────────────────────────┤
│  LAYER 2: Semantic Analysis                  │
│  4-Classifier Ensemble: ONNX_BERT (~50ms)    │
│  + RuleBased + Heuristic + LLM_SEMANTIC      │
│  Dual classifier OR-logic (BERT+RuleBased)   │
├──────────────────────────────────────────────┤
│  LAYER 3: Behavioral Analysis                │
│  16 AoC Defense Modules (11 original + 5     │
│  extended), Pattern matching, Trajectory     │
│  anomaly detection, ThoughtVirus defense     │
├──────────────────────────────────────────────┤
│  LAYER 4: Meta-Observer (O₂)                 │
│  Transfer entropy, BFT consensus,            │
│  semantic entanglement, causal graphs,       │
│  Kill-Switch, SVETILO value verification     │
├──────────────────────────────────────────────┤
│  C4 Core Engine (Z₃³)                        │
│  pipeline_orchestrator.py, event_bus.py      │
│  c4_meta_monitor.py — self-awareness deque  │
├──────────────────────────────────────────────┤
│  Defenses: Anti-Deadlock, Anti-Emergence,    │
│  Anti-Hijack, Circuit Breaker, O₂ Kill-Switch│
├──────────────────────────────────────────────┤
│  Red Team Lab: AOC scenarios, experiment     │
│  runner, LLM client, adapters                │
├──────────────────────────────────────────────┤
│  Routing: Smart Router, Quarantine,          │
│  Antifragile Scoring (capped growth)         │
└──────────────────────────────────────────────┘

c4-meta-system/
├── v4_1/
│   ├── core/
│   │   ├── pipeline.py                  # Main entry points (re-exports)
│   │   ├── __main__.py                  # HTTP server entrypoint for Docker
│   │   ├── pipeline_stages.py           # Individual processing stages
│   │   ├── pipeline_orchestrator.py      # Main orchestration (thread-safe)
│   │   ├── result_factory.py             # Standardized C4v4Result factory
│   │   ├── event_bus.py                  # Organic event bus (atexit cleanup)
│   │   └── c4_meta_monitor.py            # Z³ self-awareness (deque bounded)
│   ├── security/
│   │   ├── o2_engine.py                  # O₂ defense (kill-switch self-DoS fixed)
│   │   ├── explainable_o2.py             # O₂ explainability (sampling inverted)
│   │   ├── o2_shared.py                  # Window structures (@mention comms)
│   │   ├── semantic_detector.py          # Concept graphs (normalized entanglement)
│   │   ├── secure_debug_endpoints.py     # Debug endpoints (UTC + rate limits)
│   │   ├── hardening.py                  # Model signing / admin token verification
│   │   ├── behavioral_profiler.py        # Drift detection (thread-safe singleton)
│   │   ├── distributed_o2.py             # SQLite/Redis backend (BEGIN IMMEDIATE)
│   │   ├── swarm_orchestrator.py         # Anti-virus swarm
│   │   └── ...
│   ├── defenses/
│   │   ├── anti_deadlock.py              # Resource deadlock prevention
│   │   ├── anti_emergence.py             # State convergence (async release fixed)
│   │   └── ...
│   ├── redteam/
│   │   ├── orchestrator.py              # Main orchestrator (target_callback parsing)
│   │   ├── scenario_manager.py           # AOC scenarios management
│   │   ├── adapters/                     # LLM backend adapters
│   │   ├── experiment_executor.py        # Async execution (FPR logic fixed)
│   │   ├── experiment_services.py       # Service locator
│   │   ├── experiment_runner.py         # Web UI + REST API
│   │   ├── llm_client.py                 # Async-safe LLM client (empty choices guarded)
│   │   └── ...
│   ├── classifiers/                      # 4-classifier ensemble
│   ├── config/                           # Configuration management
│   ├── access/                           # Access control
│   ├── explainability/                   # C4 explainability
│   ├── learning/                         # Learning loop
│   ├── plugins/                          # Plugin system
│   ├── quarantine/                       # Quarantine management
│   ├── router/                           # Smart routing
│   ├── scoring/                          # Antifragile scoring
│   └── tests/                            # 240 tests (19 test files)
├── formal/                               # TLA+ specifications
├── models/                               # ONNX model + tokenizer
├── archive/Dockerfile.prepared          # Multi-stage production build (archived)
├── Dockerfile.distroless                # Distroless-ready builder pattern
├── archive/docker-compose.yml.prepared  # Full stack (Ollama + UI + Monitoring) (archived)
├── .dockerignore                        # Security-hardened exclusion list
├── infra/k8s/                           # Kubernetes manifests (hardened)
└── README.md                            # This file

# Full stack (C4-META + Ollama + UI)
docker compose --profile experiment up -d

# Build image
docker build -t c4-meta-system -f archive/Dockerfile.prepared .

# Run with local Ollama
OLLAMA_BASE_URL=http://host.docker.internal:11434 docker run -p 8080:8080 c4-meta-system

# Red Team runner
docker build -t c4-redteam -f redteam/Dockerfile.redteam.prepared .
docker run -p 8081:8081 c4-redteam

pip install -r requirements.txt
# Optional Docker deps
pip install -r requirements-docker.txt

python -m v4_1.core
# or explicitly
python -m v4_1.core.__main__

python -m v4_1.redteam.experiment_runner --web --port 8080