docs: Update multimodal-translator README with DOKS setup instructions

multimodal-translator/kubernetes/README.md

@@ -4,5 +4,127 @@
 
 - kubectl
 - NVIDIA GPU Operator
-- SMB StorageClass
+- DigitalOcean account
 
+## Setup Instructions
+
+### 1. Create NFS Server Droplet
+
+1. Create a new Droplet in DigitalOcean using doctl:
+
+   ```bash
+   doctl compute droplet create nfs-server \
+     --image ubuntu-22-04-x64 \
+     --size s-2vcpu-2gb \
+     --region <your-region>
+   ```
+
+   Make sure to:
+   - Use Ubuntu 22.04 LTS image
+   - Select at least Basic plan (2GB RAM)
+   - Choose same datacenter region as your K8s cluster
+
+2. SSH into the Droplet and set up NFS server:
+
+   ```bash
+   # Install NFS server
+   sudo apt update
+   sudo apt install nfs-kernel-server -y
+
+   # Create export directory
+   sudo mkdir -p /mnt/nfs_share
+   sudo chown nobody:nogroup /mnt/nfs_share
+   sudo chmod 777 /mnt/nfs_share
+
+   # Configure exports
+   echo "/mnt/nfs_share *(rw,sync,no_subtree_check,no_root_squash)" | sudo tee -a /etc/exports
+   sudo exportfs -a
+   sudo systemctl restart nfs-kernel-server
+   ```
+
+3. Note down the Droplet's IP address - you'll need it for the next step.
+
+### 2. Set up NFS StorageClass in Kubernetes
+
+1. Install the NFS client provisioner:
+
+   ```bash
+   helm repo add nfs-subdir-external-provisioner https://kubernetes-sigs.github.io/nfs-subdir-external-provisioner/
+   helm repo update
+   ```
+
+2. Create a values.yaml file:
+
+   ```yaml
+   nfs:
+     server: <YOUR-NFS-SERVER-IP>
+     path: /mnt/nfs_share
+   storageClass:
+     name: nfs
+     defaultClass: false
+   ```
+
+3. Install the helm chart:
+
+   ```bash
+   helm install nfs-subdir-external-provisioner nfs-subdir-external-provisioner/nfs-subdir-external-provisioner -f values.yaml
+   ```
+
+## Manifest Documentation
+
+The `manifest.yaml` file defines the following resources:
+
+### Namespace
+
+- Creates `nim-apps` namespace for all resources
+
+### Persistent Volume Claims
+
+Creates four PVCs for model caches and NVS storage:
+
+- `parakeet-cache-pvc`: 100GB for Parakeet model cache
+- `parakeet-nvs-pvc`: 100GB for Parakeet NVS storage
+- `megatron-cache-pvc`: 100GB for Megatron model cache
+- `megatron-nvs-pvc`: 100GB for Megatron NVS storage
+
+### Parakeet CTC Deployment
+
+Deploys the Parakeet ASR (Automatic Speech Recognition) model:
+
+- Image: `nvcr.io/nim/nvidia/parakeet-ctc-1.1b-asr:1.0.0`
+- GPU requirements: Uses NVIDIA H100 GPUs
+- Ports:
+  - HTTP API: 9000
+  - gRPC API: 50051
+- Volume mounts for cache and NVS storage
+- Environment configuration for NGC API and model profile
+
+### Megatron Deployment
+
+Deploys the Megatron Neural Machine Translation model:
+
+- Image: `nvcr.io/nim/nvidia/megatron-1b-nmt:1.0.0`
+- GPU requirements: Uses NVIDIA GPUs
+- Ports:
+  - HTTP API: 9000 (exposed as 9001)
+  - gRPC API: 50051 (exposed as 50052)
+- Volume mounts for cache and NVS storage
+- Environment configuration for NGC API and model profile
+
+### Services
+
+Creates LoadBalancer services for both deployments:
+
+- `parakeet-service`: Exposes Parakeet ASR endpoints
+  - HTTP: 9000
+  - gRPC: 50051
+- `megatron-service`: Exposes Megatron NMT endpoints
+  - HTTP: 9001 → 9000
+  - gRPC: 50052 → 50051
+
+## Important Notes
+
+1. Replace `your-api-key` in the manifest with your actual NGC API key
+2. Ensure your Kubernetes cluster has NVIDIA GPUs available
+3. The manifest uses DigitalOcean-specific annotations for LoadBalancer configuration
+4. Both services are exposed via LoadBalancer - ensure proper security measures are in place

multimodal-translator/kubernetes/manifest.yaml

@@ -14,7 +14,7 @@ spec:
   resources:
     requests:
       storage: 100Gi
-  storageClassName: smb
+  storageClassName: nfs
 ---
 apiVersion: v1
 kind: PersistentVolumeClaim
@@ -27,7 +27,7 @@ spec:
   resources:
     requests:
       storage: 100Gi
-  storageClassName: smb
+  storageClassName: nfs
 ---
 apiVersion: v1
 kind: PersistentVolumeClaim
@@ -40,7 +40,7 @@ spec:
   resources:
     requests:
       storage: 100Gi
-  storageClassName: smb
+  storageClassName: nfs
 ---
 apiVersion: v1
 kind: PersistentVolumeClaim
@@ -53,7 +53,7 @@ spec:
   resources:
     requests:
       storage: 100Gi
-  storageClassName: smb
+  storageClassName: nfs
 ---
 apiVersion: apps/v1
 kind: Deployment