fix(ai): warm fast chat model

services/ai-llm/deployment.yaml

@@ -20,7 +20,7 @@ spec:
       labels:
         app: ollama
       annotations:
-        ai.bstein.dev/model: qwen2.5:14b-instruct-q4_0
+        ai.bstein.dev/model: qwen2.5-coder:7b-instruct-q4_0,qwen2.5:14b-instruct-q4_0
         ai.bstein.dev/gpu: GPU pool (titan-20/21)
         ai.bstein.dev/restartedAt: "2026-01-26T12:00:00Z"
     spec:
@@ -53,6 +53,8 @@ spec:
               value: /root/.ollama
             - name: OLLAMA_MODEL
               value: qwen2.5:14b-instruct-q4_0
+            - name: OLLAMA_FAST_MODEL
+              value: qwen2.5-coder:7b-instruct-q4_0
           command:
             - /bin/sh
             - -c
@@ -60,7 +62,9 @@ spec:
               set -e
               ollama serve >/tmp/ollama.log 2>&1 &
               sleep 6
-              ollama pull "${OLLAMA_MODEL}"
+              for model in "${OLLAMA_FAST_MODEL}" "${OLLAMA_MODEL}"; do
+                ollama pull "${model}"
+              done
               pkill ollama || true
           volumeMounts:
             - name: models