Document and preflight all dependencies

- README: explicit install commands for Ollama (macOS/Linux/Windows) and Python
  3.12 per OS; note curl as a run.sh dependency
- setup.sh / setup.ps1: preflight checks that Ollama is installed and reachable
  (with install hints if missing); setup.sh also warns if curl is absent

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

scripts/setup.ps1

@@ -31,6 +31,20 @@ $vpy = ".\.venv\Scripts\python.exe"
 & $vpy -m pip install -q -r requirements.txt
 & $vpy -c "import litellm, PIL; print('>> deps OK (litellm, pillow)')"
 
+# preflight: Ollama (the model backend)
+if (Get-Command ollama -ErrorAction SilentlyContinue) {
+    Write-Host ">> ollama: $((ollama --version 2>$null | Select-Object -First 1))"
+    $base = if ($env:ROUTER_OLLAMA) { $env:ROUTER_OLLAMA } else { "http://127.0.0.1:11434" }
+    try {
+        Invoke-WebRequest "$base/api/version" -TimeoutSec 3 -UseBasicParsing | Out-Null
+        Write-Host ">> ollama server: reachable"
+    } catch {
+        Write-Host "!! ollama installed but not responding — start the Ollama app first."
+    }
+} else {
+    Write-Host "!! ollama NOT installed (required backend). Install:  winget install Ollama.Ollama"
+}
+
 Write-Host ""
 Write-Host "Next:"
 Write-Host "  .\scripts\pull-models.ps1    # pull a model fleet (edit the list first)"