Update ISO installer tiers and backend deployment

models/pull-models.sh

@@ -1,44 +1,68 @@
 #!/usr/bin/env bash
 # Pull additional AI models into Ollama
-# Run after install: bash models/pull-models.sh --tier=entry
+# Run after install: bash models/pull-models.sh --tier=starter
 # ─────────────────────────────────────────────
-TIER=${1:-entry}
+TIER="basic"   # default tier
 
-echo "Pulling models for tier: $TIER"
+for arg in "$@"; do
+  case $arg in
+    --tier=*) TIER="${arg#*=}" ;;
+    *)        TIER="$arg" ;;     # allow positional: pull-models.sh pro
+  esac
+done
 
-entry_models=(
-  "llama3.1:8b"         # General purpose, good baseline
-  "mistral:7b"          # Fast, good for APIs
-  "llama3.1:70b"        # Larger — only if enough VRAM (3× L40S has 144GB total)
-  "nomic-embed-text"    # Embedding model for RAG
-  "codellama:13b"       # Code generation
+# Normalise legacy names
+case "$TIER" in
+  entry)    TIER="basic" ;;
+  mid)      TIER="pro" ;;
+  advanced) TIER="max" ;;
+esac
+
+# ── Model lists ───────────────────────────────
+starter_models=(
+  "phi3:mini"              # 3.8B — fits in 32 GB GDDR7 at full precision
+  "nomic-embed-text"       # Embedding model for RAG
 )
 
-mid_models=(
-  "${entry_models[@]}"
-  "llama3.1:70b"
-  "mixtral:8x7b"
-  "deepseek-coder-v2:16b"
+basic_models=(
+  "llama3.1:8b"            # General purpose, good baseline
+  "mistral:7b"             # Fast, good for APIs
+  "nomic-embed-text"       # Embedding model for RAG
+  "codellama:13b"          # Code generation
 )
 
-advanced_models=(
-  "${mid_models[@]}"
-  "llama3.1:405b"
-  "mixtral:8x22b"
+pro_models=(
+  "${basic_models[@]}"
+  "llama3.1:70b"           # Large general purpose (needs 64+ GB VRAM at 4-bit)
+  "mixtral:8x7b"           # MoE model, strong reasoning
+  "deepseek-coder-v2:16b"  # Code specialist
+)
+
+max_models=(
+  "${pro_models[@]}"
+  "llama3.1:405b"          # Flagship — needs 320+ GB VRAM or multi-node
+  "mixtral:8x22b"          # Large MoE
 )
 
 case $TIER in
-  entry)    models=("${entry_models[@]}") ;;
-  mid)      models=("${mid_models[@]}") ;;
-  advanced) models=("${advanced_models[@]}") ;;
-  *)        echo "Unknown tier: $TIER. Use entry, mid, or advanced."; exit 1 ;;
+  starter) models=("${starter_models[@]}") ;;
+  basic)   models=("${basic_models[@]}") ;;
+  pro)     models=("${pro_models[@]}") ;;
+  max)     models=("${max_models[@]}") ;;
+  *)
+    echo "Unknown tier: $TIER"
+    echo "Usage: bash pull-models.sh --tier=starter|basic|pro|max"
+    exit 1
+    ;;
 esac
 
+echo "Pulling models for tier: $TIER"
+echo ""
+
 for model in "${models[@]}"; do
-  echo ""
   echo "→ Pulling $model..."
   ollama pull "$model"
+  echo ""
 done
 
-echo ""
-echo "✓ All models pulled. List with: ollama list"
+echo "✓ Done. List installed models with: ollama list"