From 1bcb3ec4ce062bc989655116f2c5f4723b43ec81 Mon Sep 17 00:00:00 2001 From: bjwswang Date: Mon, 18 Mar 2024 06:24:17 +0000 Subject: [PATCH] fix: set/deploy a default embedder for kubeagi Signed-off-by: bjwswang --- .../arcadia_v1alpha1_worker_baichuan2-7b.yaml | 2 +- ...dia_v1alpha1_worker_bge-large-zh-v1.5.yaml | 2 +- .../arcadia_v1alpha1_worker_qwen-7b-chat.yaml | 2 +- deploy/charts/arcadia/Chart.yaml | 2 +- deploy/charts/arcadia/README.md | 1 - deploy/charts/arcadia/templates/config.yaml | 11 ++++++++-- .../arcadia/templates/post-embedder.yaml | 19 +++++++++++++++++ deploy/charts/arcadia/values.yaml | 21 ++++++++++++------- 8 files changed, 46 insertions(+), 14 deletions(-) create mode 100644 deploy/charts/arcadia/templates/post-embedder.yaml diff --git a/config/samples/arcadia_v1alpha1_worker_baichuan2-7b.yaml b/config/samples/arcadia_v1alpha1_worker_baichuan2-7b.yaml index 806b55639..f60224138 100644 --- a/config/samples/arcadia_v1alpha1_worker_baichuan2-7b.yaml +++ b/config/samples/arcadia_v1alpha1_worker_baichuan2-7b.yaml @@ -25,7 +25,7 @@ spec: image: kubeagi/minio-mc:RELEASE.2023-01-28T20-29-38Z imagePullPolicy: IfNotPresent runner: - image: kubeagi/arcadia-fastchat-worker:v0.2.0 + image: kubeagi/arcadia-fastchat-worker:v0.2.36 imagePullPolicy: IfNotPresent resources: limits: diff --git a/config/samples/arcadia_v1alpha1_worker_bge-large-zh-v1.5.yaml b/config/samples/arcadia_v1alpha1_worker_bge-large-zh-v1.5.yaml index f3ed77e3d..f1ddd8de3 100644 --- a/config/samples/arcadia_v1alpha1_worker_bge-large-zh-v1.5.yaml +++ b/config/samples/arcadia_v1alpha1_worker_bge-large-zh-v1.5.yaml @@ -12,7 +12,7 @@ spec: image: kubeagi/minio-mc:RELEASE.2023-01-28T20-29-38Z imagePullPolicy: IfNotPresent runner: - image: kubeagi/arcadia-fastchat-worker:v0.2.0 + image: kubeagi/arcadia-fastchat-worker:v0.2.36 imagePullPolicy: IfNotPresent model: kind: "Models" diff --git a/config/samples/arcadia_v1alpha1_worker_qwen-7b-chat.yaml b/config/samples/arcadia_v1alpha1_worker_qwen-7b-chat.yaml index 61dbfba44..f3c99ed12 100644 --- a/config/samples/arcadia_v1alpha1_worker_qwen-7b-chat.yaml +++ b/config/samples/arcadia_v1alpha1_worker_qwen-7b-chat.yaml @@ -15,7 +15,7 @@ spec: image: kubeagi/minio-mc:RELEASE.2023-01-28T20-29-38Z imagePullPolicy: IfNotPresent runner: - image: kubeagi/arcadia-fastchat-worker:v0.2.0 + image: kubeagi/arcadia-fastchat-worker:v0.2.36 imagePullPolicy: IfNotPresent resources: limits: diff --git a/deploy/charts/arcadia/Chart.yaml b/deploy/charts/arcadia/Chart.yaml index 8ed823f39..110168873 100644 --- a/deploy/charts/arcadia/Chart.yaml +++ b/deploy/charts/arcadia/Chart.yaml @@ -2,7 +2,7 @@ apiVersion: v2 name: arcadia description: A Helm chart(Also a KubeBB Component) for KubeAGI Arcadia type: application -version: 0.3.19 +version: 0.3.20 appVersion: "0.2.0" keywords: diff --git a/deploy/charts/arcadia/README.md b/deploy/charts/arcadia/README.md index af4ee551b..efdc09492 100644 --- a/deploy/charts/arcadia/README.md +++ b/deploy/charts/arcadia/README.md @@ -17,7 +17,6 @@ global settings of arcadia chart. | Parameter | Description | Default | | ------------------------ | ------------------------------------------------------------ | ----------- | -| `oss.bucket` | Name of the bucket where data is stored | `"arcadia"` | | `defaultVectorStoreType` | Defines the default vector database type, currently `chroma` and `pgvector` are available | `pgvector` | ### controller diff --git a/deploy/charts/arcadia/templates/config.yaml b/deploy/charts/arcadia/templates/config.yaml index 0ec121b5a..1e7f2f859 100644 --- a/deploy/charts/arcadia/templates/config.yaml +++ b/deploy/charts/arcadia/templates/config.yaml @@ -38,11 +38,18 @@ data: name: '{{ .Release.Name }}-pgvector-vectorstore' {{- end }} namespace: '{{ .Release.Namespace }}' -{{- if .Values.rerank.enabled }} +{{- if .Values.config.embedder.enabled }} + embedder: + apiGroup: arcadia.kubeagi.k8s.com.cn/v1alpha1 + kind: Embedder + name: {{ .Release.Name }}-embedder + namespace: {{ .Release.Namespace }} +{{- end }} +{{- if .Values.config.rerank.enabled }} rerank: apiGroup: arcadia.kubeagi.k8s.com.cn/v1alpha1 kind: Model - name: bge-reranker-large + name: {{ .Values.config.rerank.model }} namespace: {{ .Release.Namespace }} {{- end }} diff --git a/deploy/charts/arcadia/templates/post-embedder.yaml b/deploy/charts/arcadia/templates/post-embedder.yaml new file mode 100644 index 000000000..ef7416d4f --- /dev/null +++ b/deploy/charts/arcadia/templates/post-embedder.yaml @@ -0,0 +1,19 @@ +{{- if .Values.config.embedder.enabled }} +apiVersion: arcadia.kubeagi.k8s.com.cn/v1alpha1 +kind: Worker +metadata: + name: {{ .Release.Name }}-embedder + namespace: {{ .Release.Namespace }} + annotations: + "helm.sh/hook": post-install + "helm.sh/hook-weight": "2" +spec: + displayName: 系统默认的Embedding模型服务 + description: "这是系统默认使用的Embedding模型服务" + type: "fastchat" + replicas: 1 + model: + kind: "Models" + name: {{ .Values.config.embedder.model }} + namespace: {{ .Release.Namespace }} +{{- end }} \ No newline at end of file diff --git a/deploy/charts/arcadia/values.yaml b/deploy/charts/arcadia/values.yaml index 5de0a17c0..4fbd69f7d 100644 --- a/deploy/charts/arcadia/values.yaml +++ b/deploy/charts/arcadia/values.yaml @@ -1,6 +1,4 @@ global: - oss: - bucket: &default-oss-bucket "arcadia" ## @param global.defaultVectorStoreType Defines the default vector database type, currently `chroma` and `pgvector` are available ## When the option is `chroma`, it needs `chromadb.enabled` to be `true` as well to work. ## When the option is `pgvector`, it needs `postgresql.enabled` to be `true` as well to work. @@ -20,6 +18,18 @@ global: - postgresql..nip.io ip: +# @section config is used to update configurations for this platform +config: + # embedder model by default + embedder: + enabled: false + model: "bge-large-zh-v1.5" + # reranking model by default + rerank: + enabled: true + model: "bge-reranker-large" + + # @section controller is used as the core controller for arcadia # @param image Image to be used # @param imagePullPolcy ImagePullPolicy @@ -115,7 +125,7 @@ minio: - minio-console..nip.io buckets: # Name of the bucket - - name: *default-oss-bucket + - name: "arcadia" # Policy to be set on the # bucket [none|download|upload|public|custom] # if set to custom, customPolicy must be set. @@ -188,7 +198,4 @@ ray: - name: 3090-2-GPUs headAddress: raycluster-kuberay-head-svc.kuberay-system.svc:6379 pythonVersion: 3.9.18 - dashboardHost: raycluster-kuberay-head-svc.kuberay-system.svc:8265 - -rerank: - enabled: true + dashboardHost: raycluster-kuberay-head-svc.kuberay-system.svc:8265 \ No newline at end of file