feat: per-node host_vars for Artemis, Mark44, Mark5, Bones; dynamic local.yml

host_vars/artemis.ai.home.yml

@@ -0,0 +1,31 @@
+---
+# Artemis (AI Foreman) — Control node, no NVIDIA GPU
+node_type: foreman
+has_gpu: false
+
+# Artemis-specific packages (monitoring and control)
+extra_packages:
+  - nvtop                    # GPU monitoring (uses AMD iGPU info if available)
+  - nethogs                  # Per-process network monitoring
+  - iotop                    # Per-process I/O monitoring
+  - lm-sensors               # Temperature/fan monitoring
+  - stress-ng                # Load testing
+  - cockpit                  # Web-based system management
+
+# Services to manage (not auto-started, just ensure packages installed)
+managed_services:
+  - name: hermes-gateway
+    enabled: true
+  - name: hermes-dashboard
+    enabled: true
+
+# Ollama models for Artemis (CPU inference, small models only)
+ollama_models:
+  - gemma3:4b                # Small enough for CPU
+  - phi4-mini:latest         # Tiny, fast
+
+# Hermes configuration
+hermes_config:
+  provider: openrouter
+  model: openai/gpt-4o-mini
+  context_length: 128000

host_vars/bones.yml

@@ -0,0 +1,27 @@
+---
+# Bones (Mark XLI) — Headless CPU-only node
+node_type: headless
+has_gpu: false
+
+# Headless essentials
+extra_packages:
+  - cpufrequtils              # CPU frequency management
+  - lm-sensors               # Temperature monitoring
+  - smartmontools             # Disk health monitoring
+  - hdparm                    # Disk performance tuning
+  - netdata                   # lightweight monitoring (optional)
+
+# Services managed on Bones
+managed_services:
+  - name: jarvis              # Paperclip + Ollama + PostgreSQL stack
+    enabled: true
+  - name: ollama              # CPU inference only
+    enabled: true
+
+# Ollama config (CPU mode, very small models)
+ollama_models:
+  - gemma3:1b                # Ultra-tiny for CPU
+
+# Node-specific vars
+bones_storage: "256GB SSD"
+jvm_heap: "512m"

host_vars/hulkbuster.ai.home.yml

@@ -0,0 +1,35 @@
+---
+# Mark44 (Hulkbuster) — Heavy GPU compute node
+node_type: gpu_heavy
+has_gpu: true
+gpu_type: nvidia
+gpu_model: "RTX 4070"
+vram_mb: 12282
+
+# GPU-specific packages
+extra_packages:
+  - nvidia-driver             # NVIDIA driver (if not already installed via proprietary)
+  - cuda-toolkit              # CUDA toolkit for ML workloads
+  - nvtop                     # GPU monitoring (better than nvidia-smi for live view)
+  - nethogs                   # Network per-process monitoring
+  - iotop                     # I/O per-process monitoring
+
+# Ollama models — largest VRAM headroom, can run big models
+ollama_models:
+  - gemma4:e4b               # Already pulled — keep it
+  - qwen2.5-coder:14b        # Primary coding model
+  - qwen2.5:14b             # General purpose large model
+  - mistral:7b               # Fallback fast model
+
+# Services
+managed_services:
+  - name: ollama
+    enabled: true
+  - name: open-webui
+    enabled: true
+  - name: hermes-gateway
+    enabled: true
+
+# Ollama port override (standard)
+ollama_port: 11434
+open_webui_port: 8080

host_vars/mark44.yml

@@ -0,0 +1,35 @@
+---
+# Mark44 (Hulkbuster) — Heavy GPU compute node
+node_type: gpu_heavy
+has_gpu: true
+gpu_type: nvidia
+gpu_model: "RTX 4070"
+vram_mb: 12282
+
+# GPU-specific packages
+extra_packages:
+  - nvidia-driver             # NVIDIA driver (if not already installed via proprietary)
+  - cuda-toolkit              # CUDA toolkit for ML workloads
+  - nvtop                     # GPU monitoring (better than nvidia-smi for live view)
+  - nethogs                   # Network per-process monitoring
+  - iotop                     # I/O per-process monitoring
+
+# Ollama models — largest VRAM headroom, can run big models
+ollama_models:
+  - gemma4:e4b               # Already pulled — keep it
+  - qwen2.5-coder:14b        # Primary coding model
+  - qwen2.5:14b             # General purpose large model
+  - mistral:7b               # Fallback fast model
+
+# Services
+managed_services:
+  - name: ollama
+    enabled: true
+  - name: open-webui
+    enabled: true
+  - name: hermes-gateway
+    enabled: true
+
+# Ollama port override (standard)
+ollama_port: 11434
+open_webui_port: 8080

host_vars/mark5.ai.home.yml

@@ -0,0 +1,30 @@
+---
+# Mark5 (Suitcase) — Mobile/light GPU node
+node_type: gpu_light
+has_gpu: true
+gpu_type: nvidia
+gpu_model: "RTX 4060 Laptop"
+vram_mb: 8188
+
+# Laptop-specific packages
+extra_packages:
+  - nvtop                     # GPU monitoring
+  - powertop                  # Power management analysis
+  - tlp                       # Laptop power management
+  - htop                      # Already baseline, ensure present
+
+# Ollama models — limited VRAM, smaller models only
+ollama_models:
+  - qwen2.5-coder:7b         # Small coding model
+  - gemma3:4b                # Tiny, fast
+  - llama3.1:8b              # Balanced
+
+# Services
+managed_services:
+  - name: ollama
+    enabled: true
+  - name: hermes-gateway
+    enabled: true
+
+# Ollama port
+ollama_port: 11434