fix(qmd): configure LiteLLM embedding gateway and inject auth token

roles/vhosts/qmd/defaults/main.yml

@@ -22,8 +22,9 @@ qmd_service_uid: ""
 qmd_mcp_host: 127.0.0.1
 qmd_mcp_port: 8181
 qmd_mcp_url: "http://localhost:{{ qmd_mcp_port }}/mcp"
-qmd_embed_api_base_url: https://integrate.api.nvidia.com/v1
-qmd_embed_model: nvidia/llama-nemotron-embed-1b-v2
+qmd_embed_api_base_url: "http://127.0.0.1:4000/v1"
+qmd_embed_model: "text-embedding-3-small"
+qmd_embed_api_key: "{{ lookup('ansible.builtin.env', 'AI_WORKSPACE_AUTH_TOKEN') | default('', true) }}"
 qmd_collections:
   openclaw-workspace:
     path: "{{ qmd_home }}/.openclaw/workspace"

roles/vhosts/qmd/templates/qmd.env.j2

@@ -1,3 +1,5 @@
 QMD_EMBED_API_BASE_URL={{ qmd_embed_api_base_url }}
 QMD_EMBED_MODEL={{ qmd_embed_model }}
-
+{% if qmd_embed_api_key is defined and qmd_embed_api_key != '' %}
+QMD_EMBED_API_KEY={{ qmd_embed_api_key }}
+{% endif %}