openclaw: use smaller local triage model

services/openclaw/configmap.yaml

@@ -13,10 +13,10 @@ data:
         "defaults": {
           "workspace": "/home/node/.openclaw/workspace",
           "model": {
-            "primary": "ollama-cluster/qwen2.5:7b-instruct-q4_0"
+            "primary": "ollama-cluster/qwen2.5:1.5b-instruct-q4_0"
           },
           "models": {
-            "ollama-cluster/qwen2.5:7b-instruct-q4_0": {}
+            "ollama-cluster/qwen2.5:1.5b-instruct-q4_0": {}
           }
         },
         "list": [
@@ -62,8 +62,8 @@ data:
             "apiKey": "ollama",
             "models": [
               {
-                "id": "qwen2.5:7b-instruct-q4_0",
+                "id": "qwen2.5:1.5b-instruct-q4_0",
-                "name": "qwen2.5:7b-instruct-q4_0 (Titan local)",
+                "name": "qwen2.5:1.5b-instruct-q4_0 (Titan local)",
                 "contextWindow": 32768,
                 "maxTokens": 4096,
                 "input": ["text"],
@@ -108,4 +108,3 @@ data:
     Do not run mutating commands such as `kubectl apply`, `delete`, `scale`,
     `patch`, `cordon`, `uncordon`, `drain`, or `rollout restart`. Do not read
     Kubernetes Secret values. Draft repo changes or operator steps instead.
-

services/openclaw/ollama-deployment.yaml

@@ -19,7 +19,7 @@ spec:
       labels:
         app: openclaw-ollama
       annotations:
-        ai.bstein.dev/model: qwen2.5:7b-instruct-q4_0
+        ai.bstein.dev/model: qwen2.5:1.5b-instruct-q4_0
         ai.bstein.dev/gpu: Jetson pool (titan-20/21)
     spec:
       runtimeClassName: nvidia
@@ -46,7 +46,7 @@ spec:
             - name: OLLAMA_MODELS
               value: /root/.ollama
             - name: OLLAMA_MODEL
-              value: qwen2.5:7b-instruct-q4_0
+              value: qwen2.5:1.5b-instruct-q4_0
             - name: NVIDIA_VISIBLE_DEVICES
               value: all
             - name: NVIDIA_DRIVER_CAPABILITIES