diff --git a/docs/.vitepress/config.mts b/docs/.vitepress/config.mts
index 70a280d..ef5a772 100644
--- a/docs/.vitepress/config.mts
+++ b/docs/.vitepress/config.mts
@@ -1,7 +1,8 @@
+import { text } from 'stream/consumers'
 import { defineConfig } from 'vitepress'
 import { tabsMarkdownPlugin } from 'vitepress-plugin-tabs'
 
-// https://vitepress.dev/reference/site-config
+// https://vitepress.dev/references/site-config
 export default defineConfig({
   markdown: {
     config(md) {
@@ -13,7 +14,7 @@ export default defineConfig({
   lastUpdated: true,
   ignoreDeadLinks: [
     // Site Config | VitePress
-    // https://vitepress.dev/reference/site-config#ignoredeadlinks
+    // https://vitepress.dev/references/site-config#ignoredeadlinks
     //
     // ignore all localhost links
     /^https?:\/\/localhost/,
@@ -82,7 +83,8 @@ export default defineConfig({
           {
             text: 'Reference',
             items: [
-              { text: 'CRD definition', link: '/pages/en/references/crd' },
+              { text: 'CLI Reference', link: '/pages/en/references/cli/' },
+              { text: 'CRD Reference', link: '/pages/en/references/crd/' },
               { text: 'Architectural Design', link: '/pages/en/references/architectural-design' },
             ]
           },
@@ -110,7 +112,27 @@ export default defineConfig({
           {
             text: 'Reference',
             items: [
-              { text: 'CRD definition', link: '/pages/en/references/crd' },
+              {
+                text: 'CLI Reference',
+                items: [
+                  { text: 'Commands list', link: '/pages/en/references/cli/' },
+                  {
+                    text: 'Commands',
+                    items: [
+                      { text: 'kollama deploy', link: '/pages/en/references/cli/commands/deploy' },
+                      { text: 'kollama undeploy', link: '/pages/en/references/cli/commands/undeploy' },
+                      { text: 'kollama expose', link: '/pages/en/references/cli/commands/undeploy' }
+                    ]
+                  }
+                ]
+              },
+              {
+                text: 'CRD Reference',
+                items: [
+                  { text: 'CRD list', link: '/pages/en/references/crd/' },
+                  { text: 'Model', link: '/pages/en/references/crd/model' }
+                ]
+              },
               { text: 'Architectural Design', link: '/pages/en/references/architectural-design' },
             ]
           },
@@ -147,7 +169,13 @@ export default defineConfig({
           {
             text: '参考',
             items: [
-              { text: 'CRD 定义', link: '/pages/zh-CN/references/crd' },
+              {
+                text: 'CLI 参考',
+                items: [
+                  { text: '命令列表', link: '/pages/zh-CN/references/cli/' },
+                ]
+              },
+              { text: 'CRD 参考', link: '/pages/zh-CN/references/crd/' },
               { text: '架构设计', link: '/pages/zh-CN/references/architectural-design' },
             ]
           },
@@ -175,7 +203,27 @@ export default defineConfig({
           {
             text: '参考',
             items: [
-              { text: 'CRD 定义', link: '/pages/zh-CN/references/crd' },
+              {
+                text: 'CLI 参考',
+                items: [
+                  { text: '命令列表', link: '/pages/zh-CN/references/cli/' },
+                  {
+                    text: '子命令',
+                    items: [
+                      { text: 'kollama deploy', link: '/pages/zh-CN/references/cli/commands/deploy' },
+                      { text: 'kollama undeploy', link: '/pages/zh-CN/references/cli/commands/undeploy' },
+                      { text: 'kollama expose', link: '/pages/zh-CN/references/cli/commands/undeploy' }
+                    ]
+                  }
+                ]
+              },
+              {
+                text: 'CRD 参考',
+                items: [
+                  { text: 'CRD 列表', link: '/pages/zh-CN/references/crd/' },
+                  { text: 'Model 模型资源', link: '/pages/zh-CN/references/crd/model' }
+                ]
+              },
               { text: '架构设计', link: '/pages/zh-CN/references/architectural-design' },
             ]
           },
diff --git a/docs/.vitepress/theme/components/GettingStartedBlocksEn.vue b/docs/.vitepress/theme/components/GettingStartedBlocksEn.vue
index 3566cbc..16004d3 100644
--- a/docs/.vitepress/theme/components/GettingStartedBlocksEn.vue
+++ b/docs/.vitepress/theme/components/GettingStartedBlocksEn.vue
@@ -1,7 +1,7 @@
 <template>
   <TitleBlockContainerGroup>
     <TitleBlockContainer
-      href="/pages/en/guide/getting-started/cli"
+      href="/pages/en/guide/getting-started/cli/"
       bg="zinc-50 dark:zinc-800"
       border="transparent solid 2 hover:zinc-200 dark:hover:zinc-800"
       class="[&_.title-block-container]:bg-zinc-200 [&_.title-block-container]:dark:bg-zinc-600 [&_.title-block-container]:text-zinc-900 [&_.title-block-container]:dark:text-zinc-100"
@@ -32,7 +32,7 @@
       </div>
     </TitleBlockContainer>
     <TitleBlockContainer
-      href="/pages/en/guide/getting-started/crd"
+      href="/pages/en/guide/getting-started/crd/"
       bg="zinc-50 dark:zinc-800"
       border="transparent solid 2 hover:zinc-200 dark:hover:zinc-800"
       class="[&_.title-block-container]:bg-zinc-200 [&_.title-block-container]:dark:bg-zinc-600 [&_.title-block-container]:text-zinc-900 [&_.title-block-container]:dark:text-zinc-100"
diff --git a/docs/.vitepress/theme/components/GettingStartedBlocksZhCn.vue b/docs/.vitepress/theme/components/GettingStartedBlocksZhCn.vue
index e70d38b..b794f3b 100644
--- a/docs/.vitepress/theme/components/GettingStartedBlocksZhCn.vue
+++ b/docs/.vitepress/theme/components/GettingStartedBlocksZhCn.vue
@@ -1,7 +1,7 @@
 <template>
   <TitleBlockContainerGroup>
     <TitleBlockContainer
-      href="/pages/zh-CN/guide/getting-started/cli"
+      href="/pages/zh-CN/guide/getting-started/cli/"
       bg="zinc-50 dark:zinc-800"
       border="transparent solid 2 hover:zinc-200 dark:hover:zinc-800"
       class="[&_.title-block-container]:bg-zinc-200 [&_.title-block-container]:dark:bg-zinc-600 [&_.title-block-container]:text-zinc-900 [&_.title-block-container]:dark:text-zinc-100"
@@ -32,7 +32,7 @@
       </div>
     </TitleBlockContainer>
     <TitleBlockContainer
-      href="/pages/zh-CN/guide/getting-started/crd"
+      href="/pages/zh-CN/guide/getting-started/crd/"
       bg="zinc-50 dark:zinc-800"
       border="transparent solid 2 hover:zinc-200 dark:hover:zinc-800"
       class="[&_.title-block-container]:bg-zinc-200 [&_.title-block-container]:dark:bg-zinc-600 [&_.title-block-container]:text-zinc-900 [&_.title-block-container]:dark:text-zinc-100"
diff --git a/docs/pages/en/guide/getting-started/cli.md b/docs/pages/en/guide/getting-started/cli.md
index 73d5051..397540a 100644
--- a/docs/pages/en/guide/getting-started/cli.md
+++ b/docs/pages/en/guide/getting-started/cli.md
@@ -10,12 +10,16 @@ We have a CLI called `kollama` here to simplify the deployment process. It is a
 go install github.com/nekomeowww/ollama-operator/cmd/kollama@latest
 ```
 
+> To learn about the supported commands, please refer to [`kollama`](/pages/en/references/cli/).
+
 2. Deploy a model:
 
 ```shell
 kollama deploy phi --expose --node-port 30001
 ```
 
+> For more information about the `deploy` command, please refer to [`kollama deploy`](/pages/en/references/cli/commands/deploy).
+
 That's it.
 
 3. Interact with the model:
diff --git a/docs/pages/en/guide/getting-started/crd.md b/docs/pages/en/guide/getting-started/crd.md
index bda641f..cdc716b 100644
--- a/docs/pages/en/guide/getting-started/crd.md
+++ b/docs/pages/en/guide/getting-started/crd.md
@@ -2,7 +2,7 @@
 
 ## Deploy the model
 
-1. Create one `Model` CRD to rule them all.
+1. Create one [`Model` CRD](/pages/en/references/crd/model) to rule them all.
 
 ::: tip What is CRD?
 
@@ -40,7 +40,7 @@ spec:
 
 :::
 
-Copy the following command to create a phi model CRD:
+Copy the following command to create a phi [`Model` CRD](/pages/en/references/crd/model):
 
 ```shell
 cat <<EOF >> ollama-model-phi.yaml
@@ -70,13 +70,13 @@ spec: # [!code ++]
 
 :::
 
-2. Apply the `Model` CRD to your Kubernetes cluster:
+2. Apply the [`Model` CRD](/pages/en/references/crd/model) to your Kubernetes cluster:
 
 ```shell
 kubectl apply -f ollama-model-phi.yaml
 ```
 
-3. Wait for the model to be ready:
+1. Wait for the [`Model`](/pages/en/references/crd/model) to be ready:
 
 ```shell
 kubectl wait --for=jsonpath='{.status.readyReplicas}'=1 deployment/ollama-model-phi
diff --git a/docs/pages/en/references/cli/commands/deploy.md b/docs/pages/en/references/cli/commands/deploy.md
new file mode 100644
index 0000000..c9c2218
--- /dev/null
+++ b/docs/pages/en/references/cli/commands/deploy.md
@@ -0,0 +1,163 @@
+# `kollama deploy`
+
+`kollama deploy` command is used to deploy a [`Model`](/pages/en/references/crd/model) to the Kubernetes cluster. It's basic a wrapper and utility CLI to interact with the Ollama Operator by manipulating CRD resources.
+
+[[toc]]
+
+## Use cases
+
+### Deploy model that lives on [registry.ollama.ai](https://registry.ollama.ai)
+
+```shell
+kollama deploy phi
+```
+
+### Deploy to a specific namespace
+
+```shell
+kollama deploy phi --namespace=production
+```
+
+### Deploy [`Model`](/pages/en/references/crd/model) that lives on a custom registry
+
+```shell
+kollama deploy phi --image=registry.example.com/library/phi:latest
+```
+
+### Deploy [`Model`](/pages/en/references/crd/model) with exposed NodePort service for external access
+
+```shell
+kollama deploy phi --expose
+```
+
+### Deploy [`Model`](/pages/en/references/crd/model) with exposed LoadBalancer service for external access
+
+```shell
+kollama deploy phi --expose --service-type=LoadBalancer
+```
+
+## Flags
+
+### `--namespace`
+
+If present, the namespace scope for this CLI request.
+
+### `--image`
+
+Default: `registry.ollama.ai/library/<model name>:latest`
+
+```shell
+kollama deploy phi --image=registry.ollama.ai/library/phi:latest
+```
+
+Model image to deploy.
+
+- If not specified, the [`Model`](/pages/en/references/crd/model) name will be used as the image name (will be pulled from `registry.ollama.ai/library/<model name>` by default if no registry is specified). For example, if the [`Model`](/pages/en/references/crd/model) name is `phi`, the image name will be `registry.ollama.ai/library/phi:latest`.
+- If not specified, the tag will be latest.
+
+### `--storage-class`
+
+```shell
+kollama deploy phi --storage-class=standard
+```
+
+[`StorageClass`](https://kubernetes.io/docs/concepts/storage/storage-classes/#storageclass-objects) to use for the [`Model`](/pages/en/references/crd/model)'s associated [`PersistentVolumeClaim`](https://kubernetes.io/docs/concepts/storage/persistent-volumes/#persistentvolumeclaims).
+
+If not specified, the [default `StorageClass`](https://kubernetes.io/docs/concepts/storage/storage-classes/#default-storageclass) will be used.
+
+### `--pv-access-mode`
+
+```shell
+kollama deploy phi --pv-access-mode=ReadWriteMany
+```
+
+[Access mode](https://kubernetes.io/docs/concepts/storage/persistent-volumes/#access-modes) for Ollama Operator created image store (to cache pulled images)'s [`StatefulSet`](https://kubernetes.io/docs/concepts/workloads/controllers/statefulset/) resource associated [`PersistentVolume`](https://kubernetes.io/docs/concepts/storage/persistent-volumes/#introduction).
+
+If not specified, the access mode will be [`ReadWriteOnce`](https://kubernetes.io/docs/concepts/storage/persistent-volumes/#access-modes).
+
+If you are deploying [`Model`](/pages/en/references/crd/model)s into default deployed [kind](https://kind.sigs.k8s.io/) and [k3s](https://k3s.io/) clusters, you should keep it as [`ReadWriteOnce`](https://kubernetes.io/docs/concepts/storage/persistent-volumes/#access-modes). If you are deploying [`Model`](/pages/en/references/crd/model)s into a custom cluster, you can set it to [`ReadWriteMany`](https://kubernetes.io/docs/concepts/storage/persistent-volumes/#access-modes) if [`StorageClass`](https://kubernetes.io/docs/concepts/storage/storage-classes/#storageclass-objects) supports it.
+
+### `--expose`
+
+Default: `false`
+
+```shell
+kollama deploy phi --expose
+```
+
+Whether to expose the [`Model`](/pages/en/references/crd/model) through a Service for external access and makes it easy to interact with the [`Model`](/pages/en/references/crd/model).
+
+::: info Actually, when creating a Model resource, a `ClusterIP` type service will be created
+
+At the case where users didn't supply either `--expose` flag, Ollama Operator will create a associated service for the [`Model`](/pages/en/references/crd/model) with the type of `ClusterIP` with the same name as the corresponding Deployment by default, and the service will be used for internal communication between the [`Model`](/pages/en/references/crd/model) and other services in the cluster.
+
+:::
+
+By default, [`--expose`](#expose) will create a [`NodePort`](https://kubernetes.io/docs/concepts/services-networking/service/#type-nodeport) service.
+
+Use `--expose=LoadBalancer` to create a [`LoadBalancer`](https://kubernetes.io/docs/concepts/services-networking/service/#loadbalancer) service.
+
+### `--service-type`
+
+```shell
+kollama deploy phi --expose --service-type=NodePort
+```
+
+Default: `NodePort`
+
+Type of the [`Service`](https://kubernetes.io/docs/concepts/services-networking/service/) to expose the [`Model`](/pages/en/references/crd/model). **Only valid when [`--expose`](#expose) is specified.**
+
+If not specified, the service will be exposed as [`NodePort`](https://kubernetes.io/docs/concepts/services-networking/service/#type-nodeport).
+
+::: tip To understand how many Services are associated to [`Model`](/pages/en/references/crd/model)...
+
+```shell
+kubectl get svc --selector model.ollama.ayaka.io
+```
+
+:::
+
+Use [`LoadBalancer`](https://kubernetes.io/docs/concepts/services-networking/service/#loadbalancer) to expose the service as [`LoadBalancer`](https://kubernetes.io/docs/concepts/services-networking/service/#loadbalancer).
+
+### `--service-name`
+
+```shell
+kollama deploy phi --expose --service-name=phi-svc-nodeport
+```
+
+Default: `ollama-model-<model name>-<service type>`
+
+Name of the [`Service`](https://kubernetes.io/docs/concepts/services-networking/service/) to expose the [`Model`](/pages/en/references/crd/model).
+
+If not specified, the [`Model`](/pages/en/references/crd/model) name will be used as the service name with `-nodeport` as the suffix for [`NodePort`](https://kubernetes.io/docs/concepts/services-networking/service/#type-nodeport).
+
+### `--node-port`
+
+```shell
+kollama deploy phi --expose --service-type=NodePort --node-port=30000
+```
+
+Default: Random port
+
+::: tip To understand what NodePort is used for the [`Model`](/pages/en/references/crd/model)...
+
+```shell
+kubectl get svc --selector model.ollama.ayaka.io/name=<model name> -o json | jq ".spec.ports[0].nodePort"
+```
+
+:::
+
+::: warning You can't simply specify a port number!
+
+There are several restrictions:
+
+1. By default, `30000-32767` is the `NodePort` port range in the Kubernetes cluster. If you want to use ports outside this range, you need to configure the `--service-node-port-range` parameter for the cluster.
+2. You can't use the port number already occupied by other services.
+
+For more information about choosing your own port number, please refer to [Chapter of Kubernetes Official Document about `nodePort`](https://kubernetes.io/docs/concepts/services-networking/service/#nodeport-custom-port).
+
+:::
+
+[`nodePort`](https://kubernetes.io/docs/concepts/services-networking/service/#nodeport-custom-port) to expose the [`Model`](/pages/en/references/crd/model).
+
+If not specified, a random port will be assigned. Only valid when [`--expose`](#expose) is specified, and [`--service-type`](#service-type) is set to NodePort.
diff --git a/docs/pages/en/references/cli/commands/expose.md b/docs/pages/en/references/cli/commands/expose.md
new file mode 100644
index 0000000..b0952fe
--- /dev/null
+++ b/docs/pages/en/references/cli/commands/expose.md
@@ -0,0 +1,118 @@
+# `kollama expose`
+
+`kolamma expose` is a command to expose a [`Model`](/pages/en/references/crd/model) as a service.
+
+[[toc]]
+
+## Use cases
+
+### Expose a [`Model`](/pages/en/references/crd/model) service
+
+```shell
+kollama expose phi
+```
+
+### Expose a [`Model`](/pages/en/references/crd/model) service in a specific namespace
+
+```shell
+kollama expose phi --namespace=production
+```
+
+### Expose the service as a [`LoadBalancer`](https://kubernetes.io/docs/concepts/services-networking/service/#loadbalancer) type
+
+```shell
+kollama expose phi --service-type=LoadBalancer
+```
+
+## Flags
+
+### `--namespace`
+
+If present, the namespace scope for this CLI request.
+
+### `--service-type`
+
+```shell
+kollama deploy phi --expose --service-type=NodePort
+```
+
+Default: `NodePort`
+
+Type of the [`Service`](https://kubernetes.io/docs/concepts/services-networking/service/) to expose the model. **Only valid when [`--expose`](#expose) is specified.**
+
+If not specified, the service will be exposed as [`NodePort`](https://kubernetes.io/docs/concepts/services-networking/service/#type-nodeport).
+
+::: tip To understand how many Services are associated to models...
+
+```shell
+kubectl get svc --selector model.ollama.ayaka.io
+```
+
+:::
+
+Use [`LoadBalancer`](https://kubernetes.io/docs/concepts/services-networking/service/#loadbalancer) to expose the service as [`LoadBalancer`](https://kubernetes.io/docs/concepts/services-networking/service/#loadbalancer).
+
+### `--service-type`
+
+```shell
+kollama deploy phi --expose --service-type=NodePort
+```
+
+Default: `NodePort`
+
+Type of the [`Service`](https://kubernetes.io/docs/concepts/services-networking/service/) to expose the [`Model`](/pages/en/references/crd/model). **Only valid when [`--expose`](#expose) is specified.**
+
+If not specified, the service will be exposed as [`NodePort`](https://kubernetes.io/docs/concepts/services-networking/service/#type-nodeport).
+
+::: tip To understand how many Services are associated to [`Model`](/pages/en/references/crd/model)...
+
+```shell
+kubectl get svc --selector model.ollama.ayaka.io
+```
+
+:::
+
+Use [`LoadBalancer`](https://kubernetes.io/docs/concepts/services-networking/service/#loadbalancer) to expose the service as [`LoadBalancer`](https://kubernetes.io/docs/concepts/services-networking/service/#loadbalancer).
+
+### `--service-name`
+
+```shell
+kollama deploy phi --expose --service-name=phi-svc-nodeport
+```
+
+Default: `ollama-model-<model name>-<service type>`
+
+Name of the [`Service`](https://kubernetes.io/docs/concepts/services-networking/service/) to expose the [`Model`](/pages/en/references/crd/model).
+
+If not specified, the [`Model`](/pages/en/references/crd/model) name will be used as the service name with `-nodeport` as the suffix for [`NodePort`](https://kubernetes.io/docs/concepts/services-networking/service/#type-nodeport).
+
+### `--node-port`
+
+```shell
+kollama deploy phi --expose --service-type=NodePort --node-port=30000
+```
+
+Default: Random port
+
+::: tip To understand what NodePort is used for the [`Model`](/pages/en/references/crd/model)...
+
+```shell
+kubectl get svc --selector model.ollama.ayaka.io/name=<model name> -o json | jq ".spec.ports[0].nodePort"
+```
+
+:::
+
+::: warning You can't simply specify a port number!
+
+There are several restrictions:
+
+1. By default, `30000-32767` is the `NodePort` port range in the Kubernetes cluster. If you want to use ports outside this range, you need to configure the `--service-node-port-range` parameter for the cluster.
+2. You can't use the port number already occupied by other services.
+
+For more information about choosing your own port number, please refer to [Chapter of Kubernetes Official Document about `nodePort`](https://kubernetes.io/docs/concepts/services-networking/service/#nodeport-custom-port).
+
+:::
+
+[`nodePort`](https://kubernetes.io/docs/concepts/services-networking/service/#nodeport-custom-port) to expose the [`Model`](/pages/en/references/crd/model).
+
+If not specified, a random port will be assigned. Only valid when [`--expose`](#expose) is specified, and [`--service-type`](#service-type) is set to NodePort.
diff --git a/docs/pages/en/references/cli/commands/undeploy.md b/docs/pages/en/references/cli/commands/undeploy.md
new file mode 100644
index 0000000..0dbecfd
--- /dev/null
+++ b/docs/pages/en/references/cli/commands/undeploy.md
@@ -0,0 +1,39 @@
+# `kollama undeploy`
+
+::: info No need to worry about deleting past [`Model`](/pages/en/references/crd/model) deployments and having to re-download the model image!
+
+Ollama Operator deploys a separate [`StatefulSet`](https://kubernetes.io/zh-cn/docs/concepts/workloads/controllers/statefulset/) resource for storing downloaded Ollama model images and its corresponding storage resource in addition to the normal [`Deployment`](https://kubernetes.io/zh-cn/docs/concepts/workloads/controllers/deployment/) type resource for [`Model`](/pages/en/references/crd/model)s.
+
+Therefore, even if the deployment of a [`Model`](/pages/en/references/crd/model) is deleted, it will not affect the model images that have already been downloaded. They are stored in a separate resource called `ollama-models-store` until manually deleted.
+
+You can check the status of `ollama-models-store` with the following command:
+
+```shell
+kubectl describe statefulset ollama-models-store
+```
+
+:::
+
+`kollama undeploy` is used to delete the deployment of a [`Model`](/pages/en/references/crd/model).
+
+[[toc]]
+
+## Use cases
+
+### Delete the deployment of a [`Model`](/pages/en/references/crd/model)
+
+```shell
+kollama undeploy phi
+```
+
+### Delete the deployment of a [`Model`](/pages/en/references/crd/model) in a specific namespace
+
+```shell
+kollama undeploy phi --namespace=production
+```
+
+## Flags
+
+### `--namespace`
+
+If present, the namespace scope for this CLI request.
diff --git a/docs/pages/en/references/cli/index.md b/docs/pages/en/references/cli/index.md
new file mode 100644
index 0000000..73b6e05
--- /dev/null
+++ b/docs/pages/en/references/cli/index.md
@@ -0,0 +1,5 @@
+# `kollama` CLI Reference
+
+- [`kollama deploy`](/pages/en/references/cli/commands/deploy)
+- [`kollama undeploy`](/pages/en/references/cli/commands/undeploy)
+- [`kollama expose`](/pages/en/references/cli/commands/expose)
diff --git a/docs/pages/en/references/crd.md b/docs/pages/en/references/crd.md
deleted file mode 100644
index 9b582c9..0000000
--- a/docs/pages/en/references/crd.md
+++ /dev/null
@@ -1,20 +0,0 @@
-### CRD reference
-
-```yaml
-apiVersion: ollama.ayaka.io/v1
-kind: Model
-metadata:
-  name: phi
-spec:
-  # Scale the model to 2 replicas
-  replicas: 2
-  # Use the model image `phi`
-  image: phi
-  imagePullPolicy: IfNotPresent
-  storageClassName: local-path
-  # If you have your own PersistentVolumeClaim created
-  persistentVolumeClaim: your-pvc
-  # If you need to specify the access mode for the PersistentVolume
-  persistentVolume:
-    accessMode: ReadWriteOnce
-```
diff --git a/docs/pages/en/references/crd/index.md b/docs/pages/en/references/crd/index.md
new file mode 100644
index 0000000..1b0b6af
--- /dev/null
+++ b/docs/pages/en/references/crd/index.md
@@ -0,0 +1,3 @@
+# CRD Reference
+
+- [`Model`](/pages/en/references/crd/model)
diff --git a/docs/pages/en/references/crd/model.md b/docs/pages/en/references/crd/model.md
new file mode 100644
index 0000000..014e5ea
--- /dev/null
+++ b/docs/pages/en/references/crd/model.md
@@ -0,0 +1,28 @@
+# `Model`
+
+`Model` is a custom resource definition (CRD) that represents a Ollama server instance in the cluster.
+
+When created, the creation of `Model` triggers the creation of [`Deployment`](https://kubernetes.io/docs/concepts/workloads/controllers/deployment/), [`Service`](https://kubernetes.io/docs/concepts/services-networking/service/).
+
+> If it is created for the first time, an additional StatefulSet and PersistentVolumeClaim will be created to store and persist the downloaded Ollama image.
+
+## Full CRD Reference
+
+```yaml
+apiVersion: ollama.ayaka.io/v1
+kind: Model
+metadata:
+  name: phi
+spec:
+  # Scale the model to 2 replicas
+  replicas: 2
+  # Use the model image `phi`
+  image: phi
+  imagePullPolicy: IfNotPresent
+  storageClassName: local-path
+  # If you have your own PersistentVolumeClaim created
+  persistentVolumeClaim: your-pvc
+  # If you need to specify the access mode for the PersistentVolume
+  persistentVolume:
+    accessMode: ReadWriteOnce
+```
diff --git a/docs/pages/zh-CN/guide/getting-started/cli.md b/docs/pages/zh-CN/guide/getting-started/cli.md
index 7151a28..29f8714 100644
--- a/docs/pages/zh-CN/guide/getting-started/cli.md
+++ b/docs/pages/zh-CN/guide/getting-started/cli.md
@@ -10,12 +10,16 @@
 go install github.com/nekomeowww/ollama-operator/cmd/kollama@latest
 ```
 
+> 要了解 `kollama` CLI 支持的命令，请参阅 [`kollama`](/pages/zh-CN/references/cli/)。
+
 2. 部署 Ollama 模型 CRD 到 Kubernetes 集群：
 
 ```shell
 kollama deploy phi --expose --node-port 30001
 ```
 
+> 有关 `deploy` 命令的更多信息，请参阅 [`kollama deploy`](/pages/zh-CN/references/cli/commands/deploy)。
+
 3. 开始与模型进行交互吧：
 
 ```shell
diff --git a/docs/pages/zh-CN/guide/getting-started/crd.md b/docs/pages/zh-CN/guide/getting-started/crd.md
index ee212d6..ea94712 100644
--- a/docs/pages/zh-CN/guide/getting-started/crd.md
+++ b/docs/pages/zh-CN/guide/getting-started/crd.md
@@ -2,7 +2,7 @@
 
 ## 部署模型
 
-1. 创建一个 `Model` 类型的 CRD 资源
+1. 创建一个 [`Model`](/pages/zh-CN/references/crd/model) 类型的 CRD 资源
 
 ::: tip 什么是 CRD？
 
@@ -40,7 +40,7 @@ spec:
 
 :::
 
-复制以下命令以创建一个名为 phi 的模型 CRD：
+复制以下命令以创建一个名为 phi 的 [`Model` CRD](/pages/zh-CN/references/crd/model)：
 
 ```shell
 cat <<EOF >> ollama-model-phi.yaml
@@ -70,13 +70,13 @@ spec: # [!code ++]
 
 :::
 
-2. 将 `Model` CRD 应用到 Kubernetes 集群：
+1. 将 [`Model` CRD](/pages/zh-CN/references/crd/model) 应用到 Kubernetes 集群：
 
 ```shell
 kubectl apply -f ollama-model-phi.yaml
 ```
 
-3. 等待模型就绪：
+3. 等待 [`Model` CRD](/pages/zh-CN/references/crd/model) 就绪：
 
 ```shell
 kubectl wait --for=jsonpath='{.status.readyReplicas}'=1 deployment/ollama-model-phi
diff --git a/docs/pages/zh-CN/references/cli/commands/deploy.md b/docs/pages/zh-CN/references/cli/commands/deploy.md
new file mode 100644
index 0000000..313c008
--- /dev/null
+++ b/docs/pages/zh-CN/references/cli/commands/deploy.md
@@ -0,0 +1,179 @@
+# `kollama deploy`
+
+`kollama deploy` 命令用于将 [`Model`](/pages/zh-CN/references/crd/model) 部署到 Kubernetes 集群。它是通过操作 CRD 资源与 Ollama Operator 交互的基本封装和工具类型 CLI。
+
+[[toc]]
+
+## 用例
+
+### 部署存储于 [registry.ollama.ai](https://registry.ollama.ai) 镜像仓库上的模型
+
+```shell
+kollama deploy phi
+```
+
+### 部署到特定命名空间（namespace）
+
+```shell
+kollama deploy phi --namespace=production
+```
+
+### 部署存储于自定义镜像仓库上的模型
+
+```shell
+kollama deploy phi --image=registry.example.com/library/phi:latest
+```
+
+### 部署带有暴露 NodePort 服务的 [`Model`](/pages/zh-CN/references/crd/model) 以供外部访问
+
+```shell
+kollama deploy phi --expose
+```
+
+::: tip 了解 [`Model`](/pages/zh-CN/references/crd/model) 使用的 NodePort 端口号...
+
+```shell
+kubectl get svc --selector model.ollama.ayaka.io/name=<model name> -o json | jq ".spec.ports[0].nodePort"
+```
+
+:::
+
+::: tip 配置期望分配的 NodePort 端口号...
+
+```shell
+kollama deploy phi --expose --node-port=30000
+```
+
+:::
+
+### 部署带有暴露 LoadBalancer 服务的 [`Model`](/pages/zh-CN/references/crd/model) 以供外部访问
+
+```shell
+kollama deploy phi --expose --service-type=LoadBalancer
+```
+
+## 选项
+
+### `--namespace`
+
+如果配置了该参数，将会在指定的命名空间中部署 [`Model`](/pages/zh-CN/references/crd/model)。
+
+### `--image`
+
+默认：`registry.ollama.ai/library/<model name>:latest`
+
+```shell
+kollama deploy phi --image=registry.ollama.ai/library/phi:latest
+```
+
+要部署的模型镜像。
+
+- 如果未指定，将使用 [`Model`](/pages/zh-CN/references/crd/model) 名称作为镜像名称（如果未指定镜像仓库（Registry），这个时候会默认从 `registry.ollama.ai/library/<model name>` 拉取）。例如，如果 [`Model`](/pages/zh-CN/references/crd/model) 名称是 `phi`，最终获取的镜像名称将是 `registry.ollama.ai/library/phi:latest`。
+- 如果没有指定，标签将会使用 `latest` 的。
+
+### `--storage-class`
+
+```shell
+kollama deploy phi --storage-class=standard
+```
+
+要给 [`Model`](/pages/zh-CN/references/crd/model) 部署服务关联的 [`PersistentVolumeClaim`](https://kubernetes.io/zh-cn/docs/concepts/storage/persistent-volumes/#persistentvolumeclaims) 使用的 [`StorageClass`](https://kubernetes.io/zh-cn/docs/concepts/storage/storage-classes/#storageclass-objects)。
+
+如果没有指定，则会使用 [默认的 `StorageClass`](https://kubernetes.io/zh-cn/docs/concepts/storage/storage-classes/#default-storageclass)。
+
+### `--pv-access-mode`
+
+```shell
+kollama deploy phi --pv-access-mode=ReadWriteMany
+```
+
+用于给 Ollama Operator 所创建的 image store 的 [`StatefulSet`](https://kubernetes.io/zh-cn/docs/concepts/workloads/controllers/statefulset/) 资源所关联的 [`PersistentVolume`](https://kubernetes.io/zh-cn/docs/concepts/storage/persistent-volumes/#introduction) 所使用的 [访问模式（Access mode）](https://kubernetes.io/zh-cn/docs/concepts/storage/persistent-volumes/#access-modes)。
+
+如果未指定，则默认使用 [`ReadWriteOnce`](https://kubernetes.io/zh-cn/docs/concepts/storage/persistent-volumes/#access-modes) 作为[访问模式（Access mode）](https://kubernetes.io/zh-cn/docs/concepts/storage/persistent-volumes/#access-modes)的值。
+
+如果你会将 [`Model`](/pages/zh-CN/references/crd/model) 部署到 Ollama Operator 默认支持的 [kind](https://kind.sigs.k8s.io/) 和 [k3s](https://k3s.io/) 集群，你应该保持其当前的默认值 [`ReadWriteOnce`](https://kubernetes.io/zh-cn/docs/concepts/storage/persistent-volumes/#access-modes)。有且仅有当你部署到一个自托管的集群的时候，且 [`StorageClass`](https://kubernetes.io/zh-cn/docs/concepts/storage/storage-classes/#storageclass-objects) 支持时，就可以指定访问模式为 [`ReadWriteMany`](https://kubernetes.io/zh-cn/docs/concepts/storage/persistent-volumes/#access-modes)。
+
+### `--expose`
+
+默认：`false`
+
+```shell
+kollama deploy phi --expose
+```
+
+是否通过服务公开 [`Model`](/pages/zh-CN/references/crd/model) 所暴露的接口供外部访问，使其方便与 [`Model`](/pages/zh-CN/references/crd/model) 交互。
+
+::: info 其实创建 Model 资源时，也会创建一个 `ClusterIP` 类型的服务
+
+在没有指定 `--expose` 情况下，在为 [`Model`](/pages/zh-CN/references/crd/model) 创建资源时，Ollama Operator 也默认将为 [`Model`](/pages/zh-CN/references/crd/model) 创建一个用于集群中其他服务内部直接请求到 [`Model`](/pages/zh-CN/references/crd/model) 的关联服务，方便其他服务直接与之进行集成，其类型为 `ClusterIP`，名称与 [`Model`](/pages/zh-CN/references/crd/model) 所关联的 Deployment 的名称（也就是 `ollama-model-<model name>`）相同。
+
+:::
+
+默认情况下，[`--expose`](#expose) 参数包含在内时，将会创建一个类型为 [`NodePort`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#type-nodeport) 的 [`Service`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/)。
+
+你可以使用 [`--service-type`](#service-type) 参数加上 `LoadBalancer` 值（也就是 `--service-type=LoadBalancer`）来创建一个 [`LoadBalancer`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#loadbalancer) 类型的 [`Service`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/)。
+
+### `--service-type`
+
+```shell
+kollama deploy phi --expose --service-type=NodePort
+```
+
+默认：`NodePort`
+
+暴露所部署的 [`Model`](/pages/zh-CN/references/crd/model) 服务时所连带创建的 [`Service`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/) 类型。**有且仅有当 [`--expose`](#expose) 被指定时，该参数才会生效。**
+
+如果没有指定该参数，那么创建时，将会默认使用 [`NodePort`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#type-nodeport) 类型的服务。
+
+::: tip 了解有多少服务与 [`Model`](/pages/zh-CN/references/crd/model) 相关联...
+
+```shell
+kubectl get svc --selector model.ollama.ayaka.io
+```
+
+:::
+
+可以指定为 [`LoadBalancer`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#loadbalancer) 来暴露一个 [`LoadBalancer`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#loadbalancer) 类型的服务。
+
+### `--service-name`
+
+```shell
+kollama deploy phi --expose --service-name=phi-svc-nodeport
+```
+
+默认：`ollama-model-<model name>-<service type>`
+
+暴露 [`Model`](/pages/zh-CN/references/crd/model) 所使用的 [`Service`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/) 的名称。
+
+如果未指定，则将使用 [`Model`](/pages/zh-CN/references/crd/model) 名称作为服务名称，并将 `-nodeport` 作为 [`NodePort`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#type-nodeport) 类型的服务的后缀。
+
+### `--node-port`
+
+```shell
+kollama deploy phi --expose --service-type=NodePort --node-port=30000
+```
+
+默认：随机协商的端口
+
+::: tip 了解 [`Model`](/pages/zh-CN/references/crd/model) 使用的 `NodePort` 端口号...
+
+```shell
+kubectl get svc --selector model.ollama.ayaka.io/name=<model name> -o json | jq ".spec.ports[0].nodePort"
+```
+
+:::
+
+::: warning 并不可以随便指定端口号哦！
+
+有这么几个限制是存在的：
+
+1. 默认情况下 `30000-32767` 是 Kubernetes 集群中的 `NodePort` 端口范围。如果你想要使用这个范围之外的端口，你需要在集群中配置 `--service-node-port-range` 参数。
+2. 你不能使用已经被其他服务占用的端口号。
+
+有关自己选择端口号的更多信息，请参考 [Kubernetes 官方文档有关 `nodePort` 的章节](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#nodeport-custom-port)。
+
+:::
+
+暴露 [`Model`](/pages/zh-CN/references/crd/model) 为 [`NodePort`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#type-nodeport) 类型的服务时所使用的指定的 [`nodePort`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#nodeport-custom-port) 端口号。
+
+如果没有指定，将分配一个随机端口。该参数有且仅有在 [`--expose`](#expose) 指定，或者 [`--service-type`](#service-type) 设置为 `NodePort` 时才有效。
diff --git a/docs/pages/zh-CN/references/cli/commands/expose.md b/docs/pages/zh-CN/references/cli/commands/expose.md
new file mode 100644
index 0000000..eaa171d
--- /dev/null
+++ b/docs/pages/zh-CN/references/cli/commands/expose.md
@@ -0,0 +1,96 @@
+# `kollama expose`
+
+`kollama expose` 用于暴露 [`Model`](/pages/zh-CN/references/crd/model) 服务。
+
+[[toc]]
+
+## 用例
+
+### 暴露 [`Model`](/pages/zh-CN/references/crd/model) 服务
+
+```shell
+kollama expose phi
+```
+
+### 暴露指定命名空间中的 [`Model`](/pages/zh-CN/references/crd/model) 服务
+
+```shell
+kollama expose phi --namespace=production
+```
+
+### 暴露为 [`LoadBalancer`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#loadbalancer) 类型的服务
+
+```shell
+kollama expose phi --service-type=LoadBalancer
+```
+
+## 选项
+
+### `--namespace`
+
+如果配置了该参数，将会在指定的命名空间中操作 [`Model`](/pages/zh-CN/references/crd/model)。
+
+### `--service-type`
+
+```shell
+kollama deploy phi --expose --service-type=NodePort
+```
+
+默认：`NodePort`
+
+暴露所部署的 [`Model`](/pages/zh-CN/references/crd/model) 服务时所连带创建的 [`Service`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/) 类型。**有且仅有当 [`--expose`](#expose) 被指定时，该参数才会生效。**
+
+如果没有指定该参数，那么创建时，将会默认使用 [`NodePort`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#type-nodeport) 类型的服务。
+
+::: tip 了解有多少服务与 [`Model`](/pages/zh-CN/references/crd/model) 相关联...
+
+```shell
+kubectl get svc --selector model.ollama.ayaka.io
+```
+
+:::
+
+可以指定为 [`LoadBalancer`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#loadbalancer) 来暴露一个 [`LoadBalancer`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#loadbalancer) 类型的服务。
+
+### `--service-name`
+
+```shell
+kollama deploy phi --expose --service-name=phi-svc-nodeport
+```
+
+默认：`ollama-model-<model name>-<service type>`
+
+暴露 [`Model`](/pages/zh-CN/references/crd/model) 所使用的 [`Service`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/) 的名称。
+
+如果未指定，则将使用 [`Model`](/pages/zh-CN/references/crd/model) 名称作为服务名称，并将 `-nodeport` 作为 [`NodePort`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#type-nodeport) 类型的服务的后缀。
+
+### `--node-port`
+
+```shell
+kollama deploy phi --expose --service-type=NodePort --node-port=30000
+```
+
+默认：随机协商的端口
+
+::: tip 了解 [`Model`](/pages/zh-CN/references/crd/model) 使用的 `NodePort` 端口号...
+
+```shell
+kubectl get svc --selector model.ollama.ayaka.io/name=<model name> -o json | jq ".spec.ports[0].nodePort"
+```
+
+:::
+
+::: warning 并不可以随便指定端口号哦！
+
+有这么几个限制是存在的：
+
+1. 默认情况下 `30000-32767` 是 Kubernetes 集群中的 `NodePort` 端口范围。如果你想要使用这个范围之外的端口，你需要在集群中配置 `--service-node-port-range` 参数。
+2. 你不能使用已经被其他服务占用的端口号。
+
+有关自己选择端口号的更多信息，请参考 [Kubernetes 官方文档有关 `nodePort` 的章节](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#nodeport-custom-port)。
+
+:::
+
+暴露 [`Model`](/pages/zh-CN/references/crd/model) 为 [`NodePort`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#type-nodeport) 类型的服务时所使用的指定的 [`nodePort`](https://kubernetes.io/zh-cn/docs/concepts/services-networking/service/#nodeport-custom-port) 端口号。
+
+如果没有指定，将分配一个随机端口。该参数有且仅有在 [`--expose`](#expose) 指定，或者 [`--service-type`](#service-type) 设置为 `NodePort` 时才有效。
diff --git a/docs/pages/zh-CN/references/cli/commands/undeploy.md b/docs/pages/zh-CN/references/cli/commands/undeploy.md
new file mode 100644
index 0000000..4628fa2
--- /dev/null
+++ b/docs/pages/zh-CN/references/cli/commands/undeploy.md
@@ -0,0 +1,39 @@
+# `kollama undeploy`
+
+::: info 无需担心删除过去的 [`Model`](/pages/zh-CN/references/crd/model) 部署后会需要重新下载模型镜像哦！
+
+Ollama Operator 在为 [`Model`](/pages/zh-CN/references/crd/model) 部署的普通的 [`Deployment`](https://kubernetes.io/zh-cn/docs/concepts/workloads/controllers/deployment/) 类型资源外，还会部署单独的用于存储下载过的模型的 [`StatefulSet`](https://kubernetes.io/zh-cn/docs/concepts/workloads/controllers/statefulset/) 资源和其对应的存储资源。
+
+因此，即使删除了模型的部署，也不会影响到已经下载的模型文件。他们都会存放在名为 `ollama-models-store` 的单独的资源中，直到手动删除。
+
+你可以通过下面的命令来查看 `ollama-models-store` 的状态
+
+```shell
+kubectl describe statefulset ollama-models-store
+```
+
+:::
+
+`kollama undeploy` 用于删除 [`Model`](/pages/zh-CN/references/crd/model) 的部署。
+
+[[toc]]
+
+## 用例
+
+### 删除 [`Model`](/pages/zh-CN/references/crd/model) 的部署
+
+```shell
+kollama undeploy phi
+```
+
+### 删除指定命名空间中的 [`Model`](/pages/zh-CN/references/crd/model) 的部署
+
+```shell
+kollama undeploy phi --namespace=production
+```
+
+## 选项
+
+### `--namespace`
+
+如果配置了该参数，将会在指定的命名空间中删除模型的部署。
diff --git a/docs/pages/zh-CN/references/cli/index.md b/docs/pages/zh-CN/references/cli/index.md
new file mode 100644
index 0000000..bc28465
--- /dev/null
+++ b/docs/pages/zh-CN/references/cli/index.md
@@ -0,0 +1,5 @@
+# `kollama` CLI 参考
+
+- [`kollama deploy`](/pages/zh-CN/references/cli/commands/deploy)
+- [`kollama undeploy`](/pages/zh-CN/references/cli/commands/undeploy)
+- [`kollama expose`](/pages/zh-CN/references/cli/commands/expose)
diff --git a/docs/pages/zh-CN/references/crd.md b/docs/pages/zh-CN/references/crd.md
deleted file mode 100644
index 7d68c62..0000000
--- a/docs/pages/zh-CN/references/crd.md
+++ /dev/null
@@ -1,20 +0,0 @@
-### 完整 CRD 参考
-
-```yaml
-apiVersion: ollama.ayaka.io/v1
-kind: Model
-metadata:
-  name: phi
-spec:
-  # Scale the model to 2 replicas
-  replicas: 2
-  # Use the model image `phi`
-  image: phi
-  imagePullPolicy: IfNotPresent
-  storageClassName: local-path
-  # If you have your own PersistentVolumeClaim created
-  persistentVolumeClaim: your-pvc
-  # If you need to specify the access mode for the PersistentVolume
-  persistentVolume:
-    accessMode: ReadWriteOnce
-```
diff --git a/docs/pages/zh-CN/references/crd/index.md b/docs/pages/zh-CN/references/crd/index.md
new file mode 100644
index 0000000..8d6d6c9
--- /dev/null
+++ b/docs/pages/zh-CN/references/crd/index.md
@@ -0,0 +1,3 @@
+# CRD 参考
+
+- [`Model`](/pages/zh-CN/references/crd/model)
diff --git a/docs/pages/zh-CN/references/crd/model.md b/docs/pages/zh-CN/references/crd/model.md
new file mode 100644
index 0000000..237063c
--- /dev/null
+++ b/docs/pages/zh-CN/references/crd/model.md
@@ -0,0 +1,28 @@
+# `Model` 模型资源
+
+`Model` 是一个自定义资源定义（CRD），代表集群中的一个 Ollama 推理服务实例。
+
+创建时，`Model` 的创建会触发对 [`Deployment`](https://kubernetes.io/zh-cn/docs/concepts/workloads/controllers/deployment/)，[`Service`](https://kubernetes.io/docs/concepts/services-networking/service/) 的创建。
+
+> 如果是首次创建，还会额外创建一个用于存储和持久化已下载 Ollama 镜像的 StatefulSet 和其对应的 PersistentVolumeClaim。
+
+## 完整 CRD 参考
+
+```yaml
+apiVersion: ollama.ayaka.io/v1
+kind: Model
+metadata:
+  name: phi
+spec:
+  # Scale the model to 2 replicas
+  replicas: 2
+  # Use the model image `phi`
+  image: phi
+  imagePullPolicy: IfNotPresent
+  storageClassName: local-path
+  # If you have your own PersistentVolumeClaim created
+  persistentVolumeClaim: your-pvc
+  # If you need to specify the access mode for the PersistentVolume
+  persistentVolume:
+    accessMode: ReadWriteOnce
+```
diff --git a/internal/cli/kollama/cmd_deploy.go b/internal/cli/kollama/cmd_deploy.go
index 0e7c3ad..65b4fa5 100644
--- a/internal/cli/kollama/cmd_deploy.go
+++ b/internal/cli/kollama/cmd_deploy.go
@@ -144,37 +144,37 @@ func NewCmdDeploy(streams genericiooptions.IOStreams) *cobra.Command {
 	)
 
 	cmd.Flags().StringVarP(&o.storageClass, "storage-class", "", "", ""+
-		"Storage class to use for the model's persistent volume claim. If not specified, "+
-		"the default storage class will be used.",
+		"StorageClass to use for the model's associated PersistentVolumeClaim. If not specified, "+
+		"the default StorageClass will be used.",
 	)
 
 	cmd.Flags().StringVarP(&o.pvAccessMode, "pv-access-mode", "", "", ""+
-		"Access mode for the model's persistent volume for image store StatefulSet. If not "+
-		"specified, the access mode will be ReadWriteOnce. If you are deploying models "+
-		"into default deployed kind and k3s clusters, you should keep it as ReadWriteOnce. "+
-		"If you are deploying models into a custom cluster, you can set it to ReadWriteMany "+
-		"if storage class supports it.",
+		"Access mode for Ollama Operator created image store (to cache pulled images)'s StatefulSet "+
+		"resource associated PersistentVolume. If not specified, the access mode will be ReadWriteOnce. "+
+		"If you are deploying models into default deployed kind and k3s clusters, you should keep "+
+		"it as ReadWriteOnce. If you are deploying models into a custom cluster, you can set it to "+
+		"ReadWriteMany if StorageClass supports it.",
 	)
 
 	cmd.Flags().BoolVar(&o.expose, "expose", false, ""+
 		"Whether to expose the model through a service for external access and makes it "+
 		"easy to interact with the model. By default, --expose will create a NodePort "+
-		"service. Use --expose=LoadBalancer to create a LoadBalancer service",
+		"service. Use --service-type=LoadBalancer to create a LoadBalancer service",
 	)
 
 	cmd.Flags().StringVar(&o.serviceType, "service-type", "", ""+
-		"Type of the service to expose the model. If not specified, the service will be "+
+		"Type of the Service to expose the model. If not specified, the service will be "+
 		"exposed as NodePort. Use LoadBalancer to expose the service as LoadBalancer.",
 	)
 
 	cmd.Flags().StringVar(&o.serviceName, "service-name", "", ""+
-		"Name of the service to expose the model. If not specified, the model name will "+
+		"Name of the Service to expose the model. If not specified, the model name will "+
 		"be used as the service name with -nodeport as the suffix for NodePort.",
 	)
 
 	cmd.Flags().Int32Var(&o.nodePort, "node-port", 0, ""+
 		"NodePort to expose the model. If not specified, a random port will be assigned."+
-		"Only valid when --expose is set to true, and --service-type is set to NodePort.",
+		"Only valid when --expose is specified, and --service-type is set to NodePort.",
 	)
 
 	o.configFlags.AddFlags(cmd.Flags())
@@ -236,7 +236,7 @@ func (o *CmdDeployOptions) runE(cmd *cobra.Command, args []string) error {
 		o.kubeClient,
 		namespace,
 		modelName,
-		corev1.ServiceType("NodePort"),
+		lo.Ternary(o.serviceType == "", corev1.ServiceTypeNodePort, corev1.ServiceType(o.serviceType)),
 		o.serviceName,
 		o.nodePort,
 	)
diff --git a/internal/cli/kollama/cmd_expose.go b/internal/cli/kollama/cmd_expose.go
index c76e0fb..c80f81e 100644
--- a/internal/cli/kollama/cmd_expose.go
+++ b/internal/cli/kollama/cmd_expose.go
@@ -90,16 +90,16 @@ func NewCmdExpose(streams genericiooptions.IOStreams) *cobra.Command {
 	}
 
 	cmd.Flags().StringVar(&o.serviceType, "service-type", "", ""+
-		"Type of the service to expose the model. If not specified, the service will be "+
+		"Type of the Service to expose the model. If not specified, the service will be "+
 		"exposed as NodePort. Use LoadBalancer to expose the service as LoadBalancer.",
 	)
 	cmd.Flags().StringVar(&o.serviceName, "service-name", "", ""+
-		"Name of the service to expose the model. If not specified, the model name will "+
+		"Name of the Service to expose the model. If not specified, the model name will "+
 		"be used as the service name with -nodeport as the suffix for NodePort.",
 	)
 	cmd.Flags().Int32Var(&o.nodePort, "node-port", 0, ""+
 		"NodePort to expose the model. If not specified, a random port will be assigned."+
-		"Only valid when --expose is set to true, and --service-type is set to NodePort.",
+		"Only valid when --expose is specified, and --service-type is set to NodePort.",
 	)
 
 	o.configFlags.AddFlags(cmd.Flags())
@@ -149,7 +149,7 @@ func (o *CmdExposeOptions) runE(cmd *cobra.Command, args []string) error {
 		o.kubeClient,
 		namespace,
 		modelName,
-		corev1.ServiceType("NodePort"),
+		lo.Ternary(o.serviceType == "", corev1.ServiceTypeNodePort, corev1.ServiceType(o.serviceType)),
 		o.serviceName,
 		o.nodePort,
 	)
diff --git a/pkg/model/image_store.go b/pkg/model/image_store.go
index 8ee0d41..0cbc869 100644
--- a/pkg/model/image_store.go
+++ b/pkg/model/image_store.go
@@ -18,9 +18,8 @@ import (
 )
 
 const (
-	imageStorePVCName       = "ollama-models-store-pvc"
-	imageStoreStatefulSet   = "ollama-models-store"
-	imagePullingJobImageKey = "ollama.ayaka.io/image"
+	imageStorePVCName     = "ollama-models-store-pvc"
+	imageStoreStatefulSet = "ollama-models-store"
 )
 
 func getImageStorePVC(ctx context.Context, client client.Client, namespace string) (*corev1.PersistentVolumeClaim, error) {
@@ -66,8 +65,8 @@ func EnsureImageStorePVCCreated(
 
 	pvc = &corev1.PersistentVolumeClaim{
 		ObjectMeta: metav1.ObjectMeta{
-			Labels:      make(map[string]string),
-			Annotations: make(map[string]string),
+			Labels:      ModelLabels(imageStoreStatefulSet, imageStoreStatefulSet, true),
+			Annotations: ModelAnnotations(imageStoreStatefulSet, true),
 			Name:        imageStorePVCName,
 			Namespace:   namespace,
 		},
@@ -143,10 +142,8 @@ func EnsureImageStoreStatefulSetCreated(
 
 	statefulSet = &appsv1.StatefulSet{
 		ObjectMeta: metav1.ObjectMeta{
-			Labels: map[string]string{
-				"app": imageStoreStatefulSet,
-			},
-			Annotations: make(map[string]string),
+			Labels:      ModelLabels(imageStoreStatefulSet, imageStoreStatefulSet, true),
+			Annotations: ModelAnnotations(imageStoreStatefulSet, true),
 			Name:        imageStoreStatefulSet,
 			Namespace:   namespace,
 		},
@@ -159,10 +156,8 @@ func EnsureImageStoreStatefulSetCreated(
 			},
 			Template: corev1.PodTemplateSpec{
 				ObjectMeta: metav1.ObjectMeta{
-					Labels: map[string]string{
-						"app": imageStoreStatefulSet,
-					},
-					Annotations: make(map[string]string),
+					Labels:      ModelLabels(imageStoreStatefulSet, imageStoreStatefulSet, true),
+					Annotations: ModelAnnotations(imageStoreStatefulSet, true),
 				},
 				Spec: corev1.PodSpec{
 					Containers: []corev1.Container{
@@ -257,8 +252,8 @@ func EnsureImageStoreServiceCreated(
 
 	service = &corev1.Service{
 		ObjectMeta: metav1.ObjectMeta{
-			Labels:      map[string]string{},
-			Annotations: map[string]string{},
+			Labels:      ModelLabels(imageStoreStatefulSet, imageStoreStatefulSet, true),
+			Annotations: ModelAnnotations(imageStoreStatefulSet, true),
 			Name:        imageStoreStatefulSet,
 			Namespace:   namespace,
 			OwnerReferences: []metav1.OwnerReference{{
diff --git a/pkg/model/model.go b/pkg/model/model.go
index b898d34..7897745 100644
--- a/pkg/model/model.go
+++ b/pkg/model/model.go
@@ -21,6 +21,21 @@ func ModelAppName(name string) string {
 	return fmt.Sprintf("ollama-model-%s", name)
 }
 
+func ModelLabels(appName, name string, imageStore bool) map[string]string {
+	return map[string]string{
+		"app":                        name,
+		"model.ollama.ayaka.io":      name,
+		"model.ollama.ayaka.io/type": lo.Ternary(imageStore, "image-store", "model"),
+	}
+}
+
+func ModelAnnotations(name string, imageStore bool) map[string]string {
+	return map[string]string{
+		"model.ollama.ayaka.io/name": name,
+		"model.ollama.ayaka.io/type": lo.Ternary(imageStore, "image-store", "model"),
+	}
+}
+
 func getDeployment(ctx context.Context, c client.Client, namespace string, name string) (*appsv1.Deployment, error) {
 	var deployment appsv1.Deployment
 
@@ -56,8 +71,8 @@ func EnsureDeploymentCreated(
 
 	deployment = &appsv1.Deployment{
 		ObjectMeta: metav1.ObjectMeta{
-			Labels:      make(map[string]string),
-			Annotations: make(map[string]string),
+			Labels:      ModelLabels(ModelAppName(name), name, false),
+			Annotations: ModelAnnotations(ModelAppName(name), false),
 			Name:        ModelAppName(name),
 			Namespace:   namespace,
 			OwnerReferences: []metav1.OwnerReference{{
@@ -77,9 +92,8 @@ func EnsureDeploymentCreated(
 			},
 			Template: corev1.PodTemplateSpec{
 				ObjectMeta: metav1.ObjectMeta{
-					Labels: map[string]string{
-						"app": ModelAppName(name),
-					},
+					Labels:      ModelLabels(ModelAppName(name), name, false),
+					Annotations: ModelAnnotations(ModelAppName(name), false),
 				},
 				Spec: corev1.PodSpec{
 					InitContainers: []corev1.Container{
@@ -203,8 +217,8 @@ func getService(ctx context.Context, c client.Client, namespace string, name str
 func NewServiceForModel(namespace, name string, deployment *appsv1.Deployment, serviceType corev1.ServiceType) *corev1.Service {
 	return &corev1.Service{
 		ObjectMeta: metav1.ObjectMeta{
-			Labels:      make(map[string]string),
-			Annotations: make(map[string]string),
+			Labels:      ModelLabels(ModelAppName(name), name, false),
+			Annotations: ModelAnnotations(ModelAppName(name), false),
 			Name:        ModelAppName(name),
 			Namespace:   namespace,
 			OwnerReferences: []metav1.OwnerReference{{