diff --git a/.github/workflows/test-playbooks.yml b/.github/workflows/test-playbooks.yml
index 1d7b8905..a15ac2a0 100644
--- a/.github/workflows/test-playbooks.yml
+++ b/.github/workflows/test-playbooks.yml
@@ -139,7 +139,7 @@ jobs:
       - name: Setup Python
         uses: actions/setup-python@v5
         with:
-          python-version: ${{ matrix.playbook == 'open-webui-chat' && '3.12' || '3.13' }}
+          python-version: ${{ (matrix.playbook == 'open-webui-chat' || matrix.playbook == 'vllm-inference') && '3.12' || '3.13' }}
 
       - name: Install test dependencies
         run: |
diff --git a/playbooks/supplemental/vllm-inference/README.md b/playbooks/supplemental/vllm-inference/README.md
index 29b36438..57e85743 100644
--- a/playbooks/supplemental/vllm-inference/README.md
+++ b/playbooks/supplemental/vllm-inference/README.md
@@ -38,15 +38,56 @@ For this playbook, we'll use the prebuilt AMD ROCm wheel from AMD's package inde
 
 ### Install vLLM
 
+<!-- @test:id=verify-scripts timeout=30 hidden=True -->
+```python
+import os
+import sys
+import ast
+
+# Check that required script files exist
+scripts = ['chat_with_model.py', 'curl_script.sh']
+missing = [s for s in scripts if not os.path.exists(s)]
+
+if missing:
+    print(f"FAIL: Missing files: {missing}")
+    sys.exit(1)
+print("PASS: All required script files exist")
+
+# Verify Python scripts have valid syntax
+for script in ['chat_with_model.py']:
+    with open(script, 'r') as f:
+        ast.parse(f.read())
+    print(f"PASS: {script} has valid syntax")
+```
+<!-- @test:end -->
+
+<!-- @test:id=vllm-assets-linux timeout=30 hidden=True -->
+```bash
+set -euo pipefail
+
+test -f curl_script.sh
+test -f chat_with_model.py
+
+python -m py_compile chat_with_model.py
+bash curl_script.sh --help >/dev/null
+
+echo "PASS: Required vLLM asset scripts are present and valid"
+```
+<!-- @test:end -->
+
 Create a Python 3.12 virtual environment and activate it:
 
+<!-- @test:id=create-venv timeout=30 hidden=True -->
 ```bash
-python -m venv .venv
+python3 -m venv .venv
 source .venv/bin/activate
 ```
+<!-- @test:end -->
+<!-- @setup:id=activate-venv command="source .venv/bin/activate" -->
 
 Install PyTorch 2.9.1 built for ROCm 7.12.0, along with the required ROCm Python packages, in the virtual environment:
 
+<!-- @test:id=pytorch-install timeout=600 hidden=True setup=activate-venv -->
 ```bash
 python -m pip install \
   --index-url https://repo.amd.com/rocm/whl/gfx1151/ \
@@ -54,14 +95,17 @@ python -m pip install \
   "torchaudio==2.9.0+rocm7.12.0" \
   "torchvision==0.24.0+rocm7.12.0"
 ```
+<!-- @test:end -->
 
 Install vLLM from the prebuilt ROCm wheel:
 
+<!-- @test:id=vLLM-install timeout=600 hidden=True setup=activate-venv -->
 ```bash
 python -m pip install \
   --extra-index-url https://rocm.frameworks.amd.com/whl/gfx1151/ \
   "vllm==0.16.1.dev10+g11515110f.d20260323.rocm712"
 ```
+<!-- @test:end -->
 
 Set the environment variables required by the ROCm pip packages before starting vLLM:
 
@@ -78,6 +122,24 @@ echo "=== PyTorch ===" && python -c "import torch; print('PyTorch:', torch.__ver
 echo "=== flash-attn ===" && python -c "import flash_attn; print('flash-attn:', flash_attn.__version__)"
 ```
 
+<!-- @test:id=python-env-check-linux timeout=30 hidden=True setup=activate-venv -->
+```bash
+set -euo pipefail
+python3 --version
+which python3
+```
+<!-- @test:end --> 
+
+<!-- @test:id=set-env-var-and-check-install timeout=600 hidden=True setup=activate-venv -->
+```bash
+export PYTHONPATH=.venv/lib/python3.12/site-packages/_rocm_sdk_core/share/amd_smi
+export FLASH_ATTENTION_TRITON_AMD_ENABLE=TRUE
+echo "=== vLLM ===" && python -c "import vllm; print('vLLM version:', vllm.__version__)"
+echo "=== PyTorch ===" && python -c "import torch; print('PyTorch:', torch.__version__); print('HIP available:', torch.cuda.is_available()); print('HIP built:', torch.backends.hip.is_built() if hasattr(torch.backends, 'hip') else 'N/A')"
+echo "=== flash-attn ===" && python -c "import flash_attn; print('flash-attn:', flash_attn.__version__)"
+```
+<!-- @test:end -->
+
 ## Quick Start
 
 ### 1. Start the vLLM Server
@@ -108,7 +170,7 @@ vllm serve Qwen/Qwen3-1.7B \
 You can test the server using the curl script:
 
 ```bash
-./assets/curl_script.sh
+./curl_script.sh
 ```
 
 Or use the curl command directly:
@@ -174,8 +236,177 @@ for chunk in response:
 Run the script:
 
 ```bash
-python assets/chat_with_model.py
+python chat_with_model.py
+```
+
+<!-- @test:id=vllm-server-smoke-linux timeout=1800 hidden=True -->
+```bash
+set -euo pipefail
+
+source .venv/bin/activate
+
+python -m pip install openai
+
+export PYTHONPATH="$PWD/.venv/lib/python3.12/site-packages/_rocm_sdk_core/share/amd_smi"
+export FLASH_ATTENTION_TRITON_AMD_ENABLE=TRUE
+
+server_pid=""
+
+cleanup() {
+  if [ -n "${server_pid:-}" ] && kill -0 "$server_pid" 2>/dev/null; then
+    kill "$server_pid" 2>/dev/null || true
+    sleep 2
+    kill -9 "$server_pid" 2>/dev/null || true
+  fi
+}
+trap cleanup EXIT
+
+vllm serve Qwen/Qwen3-1.7B \
+  --host 127.0.0.1 \
+  --port 8000 \
+  --max-model-len 2048 \
+  --gpu-memory-utilization 0.7 \
+  >/tmp/vllm-test.log 2>&1 &
+
+server_pid=$!
+
+ready=false
+for i in $(seq 1 300); do
+  code="$(curl -s -o /dev/null -w "%{http_code}" --max-time 2 http://127.0.0.1:8000/health || true)"
+  if [ "$code" = "200" ]; then
+    ready=true
+    break
+  fi
+  sleep 2
+done
+
+if [ "$ready" != "true" ]; then
+  echo "FAIL: vLLM server did not become ready on http://127.0.0.1:8000/health"
+  echo "Last 200 lines of vLLM log:"
+  tail -n 200 /tmp/vllm-test.log || true
+  exit 1
+fi
+
+echo "PASS: vLLM server is responding on /health"
+
+models_json="$(curl -s --max-time 10 http://127.0.0.1:8000/v1/models || true)"
+
+if [ -z "$models_json" ]; then
+  echo "FAIL: Empty response from /v1/models"
+  exit 1
+fi
+
+export MODELS_JSON="$models_json"
+python - <<'PY'
+import json
+import os
+import sys
+
+data = json.loads(os.environ["MODELS_JSON"])
+model_ids = [item.get("id") for item in data.get("data", [])]
+
+print("Available models:", model_ids)
+
+if "Qwen/Qwen3-1.7B" not in model_ids:
+    print("FAIL: Qwen/Qwen3-1.7B was not listed by /v1/models")
+    sys.exit(1)
+
+print("PASS: Qwen/Qwen3-1.7B is listed by /v1/models")
+PY
+
+direct_response="$(curl -s -X POST --max-time 300 http://127.0.0.1:8000/v1/chat/completions \
+  -H "Content-Type: application/json" \
+  -d '{
+    "model": "Qwen/Qwen3-1.7B",
+    "messages": [
+      {
+        "role": "user",
+        "content": "Reply with exactly: OK"
+      }
+    ],
+    "temperature": 0,
+    "max_tokens": 32
+  }')"
+
+export DIRECT_RESPONSE="$direct_response"
+python - <<'PY'
+import json
+import os
+import sys
+
+data = json.loads(os.environ["DIRECT_RESPONSE"])
+content = data["choices"][0]["message"]["content"]
+
+print("Direct curl response:", content)
+
+if "OK" not in content:
+    print(f"FAIL: Expected direct curl response to contain OK, got: {content!r}")
+    sys.exit(1)
+
+print("PASS: Direct curl chat completion worked")
+PY
+
+script_response="$(bash curl_script.sh \
+  --model Qwen/Qwen3-1.7B \
+  --prompt "Reply with exactly: OK" \
+  --temperature 0 \
+  --max-tokens 32)"
+
+export SCRIPT_RESPONSE="$script_response"
+python - <<'PY'
+import json
+import os
+import sys
+
+raw = os.environ["SCRIPT_RESPONSE"]
+
+try:
+    data = json.loads(raw)
+except json.JSONDecodeError:
+    print("FAIL: curl_script.sh did not return valid JSON")
+    print(raw)
+    sys.exit(1)
+
+content = data["choices"][0]["message"]["content"]
+
+print("curl_script.sh response:", content)
+
+if "OK" not in content:
+    print(f"FAIL: Expected curl_script.sh response to contain OK, got: {content!r}")
+    sys.exit(1)
+
+print("PASS: curl_script.sh chat completion worked")
+PY
+
+python - <<'PY'
+from openai import OpenAI
+import sys
+
+client = OpenAI(
+    base_url="http://localhost:8000/v1",
+    api_key="EMPTY",
+)
+
+response = client.chat.completions.create(
+    model="Qwen/Qwen3-1.7B",
+    messages=[
+        {"role": "user", "content": "Reply with exactly: OK"},
+    ],
+    temperature=0,
+    max_tokens=32,
+)
+
+content = response.choices[0].message.content or ""
+print("OpenAI Python API response:", content)
+
+if "OK" not in content:
+    print(f"FAIL: Expected OpenAI Python API response to contain OK, got: {content!r}")
+    sys.exit(1)
+
+print("PASS: OpenAI Python API chat completion worked")
+PY
 ```
+<!-- @test:end -->
 
 ## Troubleshooting
 
diff --git a/playbooks/supplemental/vllm-inference/playbook.json b/playbooks/supplemental/vllm-inference/playbook.json
index ae562754..66d42973 100644
--- a/playbooks/supplemental/vllm-inference/playbook.json
+++ b/playbooks/supplemental/vllm-inference/playbook.json
@@ -8,8 +8,7 @@
       "linux"
     ],
     "halo_box": [
-      "linux",
-      "windows"
+      "linux"
     ],
     "stx": [
       "linux"
@@ -18,6 +17,19 @@
       "linux"
     ]
   },
+  "tested_platforms": {
+    "halo": [
+      "linux"
+    ]
+  },
+  "required_platforms": {
+    "halo": [
+      "linux"
+    ],
+    "halo_box": [
+      "linux"
+    ]
+  },
   "difficulty": "beginner",
   "isNew": false,
   "isFeatured": true,