audit: clean Docker architecture, consolidate test fixtures, add containerized test runner (#94)

docker/scripts/init-ollama.sh

@@ -2,11 +2,20 @@
 # ── Ollama Initialization Script ──────────────────────────────────────────────
 #
 # Starts Ollama and pulls models on first run.
+# Requires: curl (ships with the ollama image).
+# jq is installed at runtime if missing so we can parse /api/tags reliably
+# instead of fragile grep-based JSON extraction.
 
 set -e
 
 echo "🚀 Ollama startup — checking for models..."
 
+# ── Ensure jq is available (ollama image is Debian-based) ────────────────────
+if ! command -v jq &>/dev/null; then
+  echo "📦 Installing jq for reliable JSON parsing..."
+  apt-get update -qq && apt-get install -y -qq jq >/dev/null 2>&1 || true
+fi
+
 # Start Ollama in background
 ollama serve &
 OLLAMA_PID=$!
@@ -18,15 +27,26 @@ for i in {1..60}; do
     echo "✓ Ollama is ready"
     break
   fi
+  if [ "$i" -eq 60 ]; then
+    echo "❌ Ollama failed to start after 60 s"
+    exit 1
+  fi
   echo "  Attempt $i/60..."
   sleep 1
 done
 
-# Check if models are already present
+# Check if models are already present (jq with grep fallback)
 echo "📋 Checking available models..."
-MODELS=$(curl -s http://localhost:11434/api/tags | grep -o '"name":"[^"]*"' | wc -l)
+TAGS_JSON=$(curl -s http://localhost:11434/api/tags)
 
-if [ "$MODELS" -eq 0 ]; then
+if command -v jq &>/dev/null; then
+  MODELS=$(echo "$TAGS_JSON" | jq '.models | length')
+else
+  # Fallback: count "name" keys (less reliable but functional)
+  MODELS=$(echo "$TAGS_JSON" | grep -o '"name"' | wc -l)
+fi
+
+if [ "${MODELS:-0}" -eq 0 ]; then
   echo "📥 No models found. Pulling llama3.2..."
   ollama pull llama3.2 || echo "⚠️  Failed to pull llama3.2 (may already be pulling)"
 else