initial: import DOLPHIN baseline 2026-04-21 from dolphinng5_predict working tree

Includes core prod + GREEN/BLUE subsystems:
- prod/ (BLUE harness, configs, scripts, docs)
- nautilus_dolphin/ (GREEN Nautilus-native impl + dvae/ preserved)
- adaptive_exit/ (AEM engine + models/bucket_assignments.pkl)
- Observability/ (EsoF advisor, TUI, dashboards)
- external_factors/ (EsoF producer)
- mc_forewarning_qlabs_fork/ (MC regime/envelope)

Excludes runtime caches, logs, backups, and reproducible artifacts per .gitignore.

prod/services/ARCHITECTURE_CHOICE.md

@@ -0,0 +1,119 @@
+# Service Architecture Options
+
+## Option 1: Single Supervisor (Recommended for You)
+**One systemd service → Manages multiple internal components**
+
+```
+dolphin-supervisor.service
+├── ExF Component (thread)
+├── OB Component (thread)
+├── Watchdog Component (thread)
+└── MC Component (thread)
+```
+
+**Pros:**
+- One systemd unit to manage
+- Components share memory efficiently
+- Centralized health monitoring
+- Built-in restart per component
+- Lower system overhead
+
+**Cons:**
+- Single process (if it crashes, all components stop)
+- Less isolation between components
+
+**Use when:** Components are tightly coupled, share data
+
+**Commands:**
+```bash
+systemctl --user start dolphin-supervisor
+journalctl --user -u dolphin-supervisor -f
+```
+
+---
+
+## Option 2: Multiple Separate Services
+**Each component = separate systemd service**
+
+```
+dolphin-exf.service
+├── ExF Component
+
+dolphin-ob.service
+├── OB Component
+
+dolphin-watchdog.service
+├── Watchdog Component
+```
+
+**Pros:**
+- Full isolation between components
+- Independent restart/failure domains
+- Can set different resource limits per service
+- Systemd handles everything
+
+**Cons:**
+- More systemd units to manage
+- Higher memory overhead (separate processes)
+- IPC needed for shared data
+
+**Use when:** Components are independent, need strong isolation
+
+**Commands:**
+```bash
+./service_manager.py start
+./service_manager.py status
+```
+
+---
+
+## Option 3: Hybrid (Single Supervisor + Critical Services Separate)
+
+```
+dolphin-supervisor.service
+├── ExF Component
+├── OB Component
+└── MC Component (scheduled)
+
+dolphin-watchdog.service (separate - critical!)
+└── Watchdog Component
+```
+
+**Use when:** One component is critical/safety-related
+
+---
+
+## Recommendation
+
+For your Dolphin system, **Option 1 (Single Supervisor)** is likely best because:
+
+1. **Tight coupling**: ExF, OB, Watchdog all need Hazelcast
+2. **Data sharing**: Components share state via memory
+3. **Simplicity**: One command to start/stop everything
+4. **Resource efficiency**: Lower overhead than separate processes
+
+The supervisor handles:
+- Auto-restart of failed components
+- Health monitoring
+- Structured logging
+- Graceful shutdown
+
+---
+
+## Quick Start: Single Supervisor
+
+```bash
+# 1. Enable and start
+cd /mnt/dolphinng5_predict/prod/services
+systemctl --user enable dolphin-supervisor
+systemctl --user start dolphin-supervisor
+
+# 2. Check status
+systemctl --user status dolphin-supervisor
+
+# 3. View logs
+journalctl --user -u dolphin-supervisor -f
+
+# 4. Stop
+systemctl --user stop dolphin-supervisor
+```