Add a busy timeout for the LocalAI watchdog

LOCALAI_WATCHDOG_BUSY had no matching timeout setting, so the role
could only ever use LocalAI's built-in 5 minutes. On the throttled
Tesla T4 a legitimate request can take more than 6 minutes, so default
to 10 minutes, which also matches the ingress proxy-read-timeout.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

defaults/main.yml

@@ -8,6 +8,7 @@ localai_api_keys:
 localai_watchdog_idle: "false"
 localai_watchdog_idle_timeout: "15m"
 localai_watchdog_busy: "true"
+localai_watchdog_busy_timeout: "10m"
 localai_image_tag: latest-gpu-nvidia-cuda-13
 localai_debug: "false"
 localai_p2p: "false"

tasks/main.yml

@@ -34,6 +34,8 @@
             value: "{{ localai_watchdog_idle_timeout }}"
           - name: LOCALAI_WATCHDOG_BUSY
             value: "{{ localai_watchdog_busy }}"
+          - name: LOCALAI_WATCHDOG_BUSY_TIMEOUT
+            value: "{{ localai_watchdog_busy_timeout }}"
           - name: DEBUG
             value: "{{ localai_debug }}"
           - name: LOCALAI_P2P