diff --git a/llm_server/llm/info.py b/llm_server/llm/info.py
index 0718dfd..b26093d 100644
--- a/llm_server/llm/info.py
+++ b/llm_server/llm/info.py
@@ -9,14 +9,14 @@ def get_running_model():
 
     if opts.mode == 'oobabooga':
         try:
-            backend_response = requests.get(f'{opts.backend_url}/api/v1/model', timeout=3, verify=opts.verify_ssl)
+            backend_response = requests.get(f'{opts.backend_url}/api/v1/model', timeout=10, verify=opts.verify_ssl)
             r_json = backend_response.json()
             return r_json['result'], None
         except Exception as e:
             return False, e
     elif opts.mode == 'vllm':
         try:
-            backend_response = requests.get(f'{opts.backend_url}/model', timeout=3, verify=opts.verify_ssl)
+            backend_response = requests.get(f'{opts.backend_url}/model', timeout=10, verify=opts.verify_ssl)
             r_json = backend_response.json()
             return r_json['model'], None
         except Exception as e:
diff --git a/llm_server/llm/oobabooga/generate.py b/llm_server/llm/oobabooga/generate.py
index ffa83e5..c4736af 100644
--- a/llm_server/llm/oobabooga/generate.py
+++ b/llm_server/llm/oobabooga/generate.py
@@ -9,7 +9,7 @@ from llm_server import opts
 
 def generate(json_data: dict):
     try:
-        r = requests.post(f'{opts.backend_url}/api/v1/generate', json=json_data, verify=opts.verify_ssl)
+        r = requests.post(f'{opts.backend_url}/api/v1/generate', json=json_data, verify=opts.verify_ssl, timeout=120)
     except Exception as e:
         return False, None, f'{e.__class__.__name__}: {e}'
     if r.status_code != 200:
diff --git a/llm_server/llm/vllm/generate.py b/llm_server/llm/vllm/generate.py
index b740679..0a689d8 100644
--- a/llm_server/llm/vllm/generate.py
+++ b/llm_server/llm/vllm/generate.py
@@ -79,7 +79,7 @@ def transform_prompt_to_text(prompt: list):
 
 def handle_blocking_request(json_data: dict):
     try:
-        r = requests.post(f'{opts.backend_url}/generate', json=prepare_json(json_data), verify=opts.verify_ssl)
+        r = requests.post(f'{opts.backend_url}/generate', json=prepare_json(json_data), verify=opts.verify_ssl, timeout=120)
     except Exception as e:
         return False, None, f'{e.__class__.__name__}: {e}'
 
diff --git a/llm_server/netdata.py b/llm_server/netdata.py
index 44c68ea..d3abf32 100644
--- a/llm_server/netdata.py
+++ b/llm_server/netdata.py
@@ -44,7 +44,7 @@ def get_gpu_wh(gpu_id: int):
         "format": "json",
         "options": "absolute|jsonwrap"
     }
-    response = requests.get(f'{opts.netdata_root}/api/v1/data', params=params)
+    response = requests.get(f'{opts.netdata_root}/api/v1/data', params=params, timeout=10)
     data = json.loads(response.text)
     total_power_usage_watts = sum(point[1] for point in data['result']['data'])
     # total_power_usage_watt_hours = round(total_power_usage_watts / 3600, 1)
diff --git a/other/vllm/vllm.service b/other/vllm/vllm.service
new file mode 100644
index 0000000..d8da27c
--- /dev/null
+++ b/other/vllm/vllm.service
@@ -0,0 +1,15 @@
+[Unit]
+Description=VLLM Backend
+Wants=basic.target
+After=basic.target network.target
+
+[Service]
+User=USERNAME
+Group=USERNAME
+# Can add --disable-log-requests when I know the backend won't crash
+ExecStart=/storage/vllm/venv/bin/python /storage/vllm/api_server.py --model /storage/oobabooga/one-click-installers/text-generation-webui/models/TheBloke_MythoMax-L2-13B-GPTQ/ --host 0.0.0.0 --port 7000 --max-num-batched-tokens 24576
+Restart=always
+RestartSec=2
+
+[Install]
+WantedBy=multi-user.target
diff --git a/other/vllm/vllm_api_server.py b/other/vllm/vllm_api_server.py
index 98a1f7d..f5b5f45 100644
--- a/other/vllm/vllm_api_server.py
+++ b/other/vllm/vllm_api_server.py
@@ -12,8 +12,6 @@ from vllm.engine.async_llm_engine import AsyncLLMEngine
 from vllm.sampling_params import SamplingParams
 from vllm.utils import random_uuid
 
-# python api_server.py --model /storage/oobabooga/one-click-installers/text-generation-webui/models/TheBloke_MythoMax-L2-13B-GPTQ/ --host 0.0.0.0 --port 7000 --max-num-batched-tokens 24576
-
 TIMEOUT_KEEP_ALIVE = 5  # seconds.
 TIMEOUT_TO_PREVENT_DEADLOCK = 1  # seconds.
 app = FastAPI()