chore: update rating doc

kubebb · Jan 10, 2024 · adc4cad · adc4cad
1 parent cb33b53
commit adc4cad
Showing 1 changed file with 128 additions and 21 deletions.
diff --git a/docs/develop-guid/core/rating/enablerating.md b/docs/develop-guid/core/rating/enablerating.md
@@ -21,42 +21,81 @@ sidebar_position: 1
 
 参考[安装内核](../../../quick-start/quick-install.md)完成内核(未启用Rating)安装。
 
-## 2. 创建命名空间 `kubebb-addons`
+## 2. 创建需要的命名空间 `kubebb-addons`, `tekton`
 
 :::info
+tekton 会安装在 `tekton` 的命名空间下。
 后续安装的组件都会添加到`kubebb-addons`命名空间中，作为kubebb的扩展
 :::
 
 ```shell
-kubectl create namespace kubebb-addons
+kubectl create ns kubebb-addons tekton
 ```
 
 ## 3. 安装Tekton流水线
 
-官方组件仓库提供了[Tekton组件安装包](https://github.com/kubebb/components/tree/main/charts/tekton-operator)和[Tekton安装示例](https://github.com/kubebb/components/tree/main/examples/tekton-operator)。可以使用内核快速完成tekton的安装。
+```yaml
+# install-tekton.yaml
+apiVersion: core.kubebb.k8s.com.cn/v1alpha1
+kind: ComponentPlan
+metadata:
+  name: tekton
+  namespace: tekton
+spec:
+  approved: true
+  name: tekton # similar to helm release name
+  version: 0.64.0
+  component:
+    name: kubebb.tekton-operator
+    namespace: kubebb-system
+```
+
+在 `kubebb core` 内核安装完成后，会自动创建一个系统的 `Repository`, 直接利用这个 `Repository` 来创建 tekton。
 
 > 执行以下命令前需要确保组件kubebb.tekton-operator已经同步完成
 
 命令如下:
 
 ```shell
-# 默认安装到default命名空间
-kubectl apply -f https://raw.githubusercontent.com/kubebb/components/main/examples/tekton-operator/componentplan.yaml
+kubectl apply -f install-tekton.yaml
 ```
 
 查看安装状态:
 
 ```shell
-kubectl get pods --watch
+kubectl -ntekton get pods --watch
 ```
 
 如果安装完成,输出如下:
 
 ```shell
-❯ kubectl get pods
-NAME                                                 READY   STATUS    RESTARTS   AGE
-my-tekton-tekton-operator-68bdffc888-8dtfx           2/2     Running   0          25m
-my-tekton-tekton-operator-webhook-78bdfcbc77-6k6cx   1/1     Running   0          25m
+root@macbookpro:~/workspace/test-arcadia-helm/rating# kubectl get po -ntekton 
+NAME                                              READY   STATUS    RESTARTS   AGE
+tekton-tekton-operator-7754767b9f-hjzjq           2/2     Running   0          2m35s
+tekton-tekton-operator-webhook-66c8448c8b-qmzz7   1/1     Running   0          2m35s
+```
+
+安装完成后，检查 tekton 安装的 CRD.
+
+```shell
+root@macbookpro:~/workspace/test-arcadia-helm/rating# kubectl get crd|grep tekton
+clustertasks.tekton.dev                    2024-01-10T06:41:24Z
+customruns.tekton.dev                      2024-01-10T06:41:25Z
+pipelineresources.tekton.dev               2024-01-10T06:41:25Z
+pipelineruns.tekton.dev                    2024-01-10T06:41:25Z
+pipelines.tekton.dev                       2024-01-10T06:41:25Z
+resolutionrequests.resolution.tekton.dev   2024-01-10T06:41:25Z
+runs.tekton.dev                            2024-01-10T06:41:25Z
+taskruns.tekton.dev                        2024-01-10T06:41:25Z
+tasks.tekton.dev                           2024-01-10T06:41:25Z
+tektonchains.operator.tekton.dev           2024-01-10T06:41:19Z
+tektonconfigs.operator.tekton.dev          2024-01-10T06:41:19Z
+tektondashboards.operator.tekton.dev       2024-01-10T06:41:19Z
+tektonhubs.operator.tekton.dev             2024-01-10T06:41:19Z
+tektoninstallersets.operator.tekton.dev    2024-01-10T06:41:19Z
+tektonpipelines.operator.tekton.dev        2024-01-10T06:41:19Z
+tektonresults.operator.tekton.dev          2024-01-10T06:41:19Z
+tektontriggers.operator.tekton.dev         2024-01-10T06:41:19Z
 ```
 
 如果长时间安装未完成,可查看对应的Componentplan资源的状态。
@@ -89,24 +128,40 @@ arcadia.llms         32s
 
 2. 安装Arcadia AI组件
 
-> 此处建议采用组件订阅自动安装的模式，默认安装到default命名空间
+```yaml
+# arcadia-sub.yaml
+apiVersion: core.kubebb.k8s.com.cn/v1alpha1
+kind: Subscription
+metadata:
+  name: arcadia
+  namespace: kubebb-system
+spec:
+  component:
+    name: arcadia.arcadia
+    namespace: kubebb-system
+  componentPlanInstallMethod: auto
+  name: arcadia
+```
 
 ```shell
-kubectl apply -f https://raw.githubusercontent.com/kubebb/components/main/examples/arcadia/subscription.yaml
+kubectl apply -f arcadia-sub.yaml
 ```
 
 3. 查看安装状态
 
 ```shell
-❯ kubectl get pods --watch
-NAME                                                 READY   STATUS    RESTARTS   AGE
-my-tekton-tekton-operator-68bdffc888-8dtfx           2/2     Running   0          48m
-my-tekton-tekton-operator-webhook-78bdfcbc77-6k6cx   1/1     Running   0          48m
-arcadia-5cb86f8787-jvd7j                             0/1     Pending   0          0s
-arcadia-5cb86f8787-jvd7j                             0/1     Pending   0          0s
-arcadia-5cb86f8787-jvd7j                             0/1     ContainerCreating   0          0s
-arcadia-5cb86f8787-jvd7j                             0/1     Running             0          20s
-arcadia-5cb86f8787-jvd7j                             1/1     Running             0          30s
+kubectl get po -nkubebb-system 
+
+NAME                                   READY   STATUS              RESTARTS   AGE
+arcadia-apiserver-559974c74b-r8jrp     0/1     ContainerCreating   0          5m5s
+arcadia-chromadb-0                     0/1     Pending             0          5m4s
+arcadia-controller-85cbddc8fd-bbh4q    1/1     Running             0          5m5s
+arcadia-dataprocess-57b99c84bb-sgr9f   0/1     ContainerCreating   0          5m5s
+arcadia-fastchat-6b4b454d69-mqjsl      0/2     ContainerCreating   0          5m5s
+arcadia-minio-58d6b4885c-vdgzx         0/1     Pending             0          5m5s
+arcadia-portal-64c5b7f986-5dc8x        1/1     Running             0          5m5s
+arcadia-postgresql-0                   0/1     Pending             0          5m4s
+kubebb-core-679594d66f-r769g           1/1     Running             0          69m
 ```
 
 ## 5. 更新内核
@@ -137,3 +192,55 @@ kubebb-core-6d78d7d8fd-vxbc6   0/1     Terminating   0          2m3s
 1.6935407235060694e+09 INFO Starting EventSource {"controller": "rating", "controllerGroup": "core.kubebb.k8s.com.cn", "controllerKind": "Rating", "source": "kind source: *v1alpha1.Rating"}
 1.6935407235063274e+09 INFO Starting EventSource {"controller": "rating", "controllerGroup": "core.kubebb.k8s.com.cn", "controllerKind": "Rating", "source": "kind source: *v1beta1.PipelineRun"}
 ```
+
+## 6 创建系统 LLM
+
+因为要进行 AI 评测，为了保证评测结果的公平，需要全局配置一个 `LLM`, 这个资源的 yaml 格式如下
+
+```yaml
+apiVersion: arcadia.kubeagi.k8s.com.cn/v1alpha1
+kind: LLM
+metadata:
+  name: zhipuai # LLM 的名字
+  namespace: kubebb-system # LLM的namespace
+spec:
+  provider:
+    endpoint:
+      authSecret: # authSecret 是当我们选择了一个模型服务后，例如 openai，zhipuai 等，会给用户一个 token 进行 api调用
+        kind: secret
+        name: zhipuai
+      url: https://open.bigmodel.cn/api/paas/v3/model-api
+  type: zhipuai
+```
+
+这里需要关注的就是 `spec.privider.endpoint`, 其中 `authSecret` 是配置一个 secret 的名字，这个secret中保存了访问 AI 服务的 `Token`, `Token` 在 secret的存储字段是 `apiKey`
+
+```yaml
+apiVersion: v1
+data:
+  apiKey: YmFzZTY0Cg==
+kind: Secret
+metadata:
+  name: zhipuai
+  namespace: kubebb-system
+type: Opaque
+```
+
+`url` 字段，就是模型服务提供的调用接口的地址。
+
+目前平台支持 `openai` 和 `zhipuai`，所以 `spec.type` 字段就是这两个值可选。
+因为系统中可能存在多个 `LLM`, 所以我们需要一个配置指明 AI 要使用的 `LLM`。所以我们需要在创建一个 configmap 名字叫 `system-llm`，名字是固定的。
+包含两个字段：
+
+- name 使用的LLM的名字
+- namespace 使用的LLM所在的namespace
+
+```yaml
+kind: ConfigMap
+metadata:
+  name: system-llm
+  namespace: kubebb-system
+data:
+  name: "zhipuai"
+  namespace: "kubebb-system"
+```