[Docs]: Local inference on K8s not documented

[e-minguez]

Documentation Location

https://docs.nvidia.com/aiq-blueprint/latest/customization/swapping-models.html#using-downloadable-nims-self-hosted

Issue

To do local inference, the configuration cannot be customized easily (the config file needs to be created manually and store it as a configmap)

https://docs.nvidia.com/aiq-blueprint/latest/deployment/kubernetes.html#configuration

Suggested Fix

Add a section on the docs on how to create a custom configmap with custom tweaks on the docs.

This probably will require changing the helm chart as well to expose a configmap reference instead a file reference at aiq.apps.backend.env.CONFIG_FILE

Pre-Submission Checks

• I searched existing issues and pull requests.
• This report does not include secrets, credentials, or private data.

[e-minguez]

I guess this is similar to #243

[mbelur]

I confirmed that mounting a custom ConfigMap as a volume works.

I also support the proposed fix in this issue: updating the Helm chart to expose a ConfigMap reference instead of a file reference via aiq.apps.backend.env.CONFIG_FILE. This would provide a cleaner and more Kubernetes-native mechanism for customizing the configuration.

For reference, these are the steps I used to validate the ConfigMap approach.

1. Create a custom ConfigMap

I created a modified version of configs/config_web_frag.yml (config_web_frag_2_1_0_updated.yaml) that updates the NIM base URLs, then created a ConfigMap:

kubectl create configmap aiq-web-frag-config \
  -n ns-aiq \
  --from-file=config_web_frag.yml=config_web_frag_2_1_0_updated.yaml

2. Override the AIQ Helm values

aiq:
  apps:
    backend:
      env:
        CONFIG_FILE: configs/config_web_frag.yml
        RAG_SERVER_URL: http://rag-server.nvidia-blueprint-rag.svc.cluster.local:8081/v1
        RAG_INGEST_URL: http://ingestor-server.nvidia-blueprint-rag.svc.cluster.local:8082/v1
        EMBED_BASE_URL: http://nemotron-vlm-embedding-ms.nvidia-blueprint-rag.svc.cluster.local:8000/v1
        EMBED_MODEL: nvidia/llama-nemotron-embed-vl-1b-v2

      volumes:
        - name: postgres-init
          configMap:
            name: aiq-postgres-init

        # Override the default config with our custom ConfigMap
        - name: aiq-web-frag-config
          configMap:
            name: aiq-web-frag-config

    volumeMounts:
      # Replace only the bundled config file while leaving the rest of
      # /app/configs unchanged.
      - name: aiq-web-frag-config
        mountPath: /app/configs/config_web_frag.yml
        subPath: config_web_frag.yml

With this configuration, the backend successfully uses the custom config_web_frag.yml from the ConfigMap.