opea-project · lianhao · Jan 17, 2025 · Jan 10, 2025
@@ -1,31 +1,38 @@
 # retriever-usvc
 
-Helm chart for deploying Retriever microservice.
+Helm chart for deploying OPEA retriever-usvc microservice.
 
-retriever-usvc depends on redis and tei, you should set these endpoints before start.
+## Installing the chart
 
-## (Option1): Installing the chart separately
+`retriever-usvc` will use TEI for embedding service, and support different vector DB backends.
 
-First, you need to install the tei and redis-vector-db chart, refer to the [tei](../tei/README.md) and [redis-vector-db](../redis-vector-db/README.md) for more information.
+- TEI: please refer to the [tei](../tei) for more information.
 
-After you've deployed the tei and redis-vector-db chart successfully, run `kubectl get svc` to get the service endpoint and URL respectively, i.e. `http://tei`, `redis://redis-vector-db:6379`.
+- Redis vector DB: please refer to [redis-vector-db](../redis-vector-db/) for more information.
 
-To install retriever-usvc chart, run the following:
+- Milvus DB: please refer to [milvus-helm](https://github.com/zilliztech/milvus-helm/tree/milvus-4.2.12) for more information.
 
-```console
-cd GenAIInfra/helm-charts/common/retriever-usvc
-export REDIS_URL="redis://redis-vector-db:6379"
-export TEI_EMBEDDING_ENDPOINT="http://tei"
-helm dependency update
-helm install retriever-usvc . --set REDIS_URL=${REDIS_URL} --set TEI_EMBEDDING_ENDPOINT=${TEI_EMBEDDING_ENDPOINT}
-```
+First, you need to install the `tei` helm chart and one of the vector DB service, i.e. `redis-vector-db` chart.
 
-## (Option2): Installing the chart with dependencies automatically
+After you've deployed dependency charts successfully, please run `kubectl get svc` to get the service endpoint URL respectively, i.e. `http://tei:80`, `redis://redis-vector-db:6379`.
+
+To install `retriever-usvc` chart, run the following:
 
 ```console
 cd GenAIInfra/helm-charts/common/retriever-usvc
 helm dependency update
-helm install retriever-usvc . --set tei.enabled=true --set redis-vector-db.enabled=true
+export HFTOKEN="insert-your-huggingface-token-here"
+export TEI_EMBEDDING_ENDPOINT="http://tei"
+
+# Install retriever-usvc with Redis DB backend
+export RETRIEVER_BACKEND="REDIS"
+export DB_HOST="redis-vector-db"
+helm install retriever-usvc . --set TEI_EMBEDDING_ENDPOINT=${TEI_EMBEDDING_ENDPOINT} --set global.HUGGINGFACEHUB_API_TOKEN=${HF_TOKEN} --set RETRIEVER_BACKEND=${RETRIEVER_BACKEND} --set REDIS_HOST=${DB_HOST}
+
+# Install retriever-usvc with Milvus DB backend
+# export RETRIEVER_BACKEND="MILVUS"
+# export DB_HOST="milvus"
+# helm install retriever-usvc . --set TEI_EMBEDDING_ENDPOINT=${TEI_EMBEDDING_ENDPOINT} --set global.HUGGINGFACEHUB_API_TOKEN=${HF_TOKEN} --set RETRIEVER_BACKEND=${RETRIEVER_BACKEND} --set MILVUS_HOST=${DB_HOST}
 ```
 
 ## Verify
@@ -46,14 +53,16 @@ curl http://localhost:7000/v1/retrieval  \
 
 ## Values
 
-| Key                    | Type   | Default                | Description |
-| ---------------------- | ------ | ---------------------- | ----------- |
-| image.repository       | string | `"opea/retriever-tgi"` |             |
-| service.port           | string | `"7000"`               |             |
-| REDIS_URL              | string | `""`                   |             |
-| TEI_EMBEDDING_ENDPOINT | string | `""`                   |             |
-| global.monitoring      | bool   | `false`                |             |
+| Key                             | Type   | Default   | Description                                                                                             |
+| ------------------------------- | ------ | --------- | ------------------------------------------------------------------------------------------------------- |
+| global.HUGGINGFACEHUB_API_TOKEN | string | `""`      | Your own Hugging Face API token                                                                         |
+| service.port                    | string | `"7000"`  |                                                                                                         |
+| RETRIEVER_BACKEND               | string | `"REDIS"` | vector DB backend to use, one of "REDIS", "MILVUS"                                                      |
+| REDIS_HOST                      | string | `""`      | Redis service URL host, only valid for Redis, please see `values.yaml` for other Redis configuration    |
+| MILVUS_HOST                     | string | `""`      | Milvus service URL host, only valid for Milvus, please see `values.yaml` for other Milvus configuration |
+| TEI_EMBEDDING_ENDPOINT          | string | `""`      |                                                                                                         |
+| global.monitoring               | bool   | `false`   |                                                                                                         |
 
 ## Milvus support
 
-Refer to the milvus-values.yaml for milvus configurations.
+Refer to the milvus-values.yaml for Milvus configurations.
@@ -1,33 +1,34 @@
 # Copyright (C) 2024 Intel Corporation
 # SPDX-License-Identifier: Apache-2.0
 
-# Default values for retriever-usvc.
-# This is a YAML-formatted file.
-# Declare variables to be passed into your templates.
-
 milvus:
   enabled: true
+  # Milvus config for standalone mode with no PVC which has minimum requirements for the K8s cluster.
+  # Check https://github.com/zilliztech/milvus-helm/tree/milvus-4.2.12/charts/milvus for more production level configuration.
   cluster:
     enabled: false
   etcd:
     replicaCount: 1
+    persistence:
+      enabled: false
   pulsar:
     enabled: false
   minio:
     mode: standalone
+    persistence:
+      enabled: false
+  standalone:
+    persistence:
+      enabled: false
 redis-vector-db:
-    enabled: false
+  enabled: false
 tei:
   enabled: true
 
-image:
-  repository: opea/retriever-milvus
-port: 7000
+RETRIEVER_BACKEND: "MILVUS"
 # text embedding inference service URL, e.g. http://<service-name>:<port>
-#TEI_EMBEDDING_ENDPOINT: "http://dataprep-tei:80"
+#TEI_EMBEDDING_ENDPOINT: "http://retriever-tei:80"
 # milvus DB configurations
-#MILVUS_HOST: "dataprep-milvus"
-MILVUS_PORT: "19530"
+#MILVUS_HOST: "retriever-milvus"
+MILVUS_PORT: 19530
 COLLECTION_NAME: "rag_milvus"
-MOSEC_EMBEDDING_ENDPOINT: ""
-MOSEC_EMBEDDING_MODEL: ""
@@ -8,41 +8,47 @@ metadata:
   labels:
     {{- include "retriever-usvc.labels" . | nindent 4 }}
 data:
-  {{- if .Values.MOSEC_EMBEDDING_ENDPOINT }}
-  MOSEC_EMBEDDING_ENDPOINT: {{ .Values.MOSEC_EMBEDDING_ENDPOINT | quote}}
-  MOSEC_EMBEDDING_MODEL: {{ .Values.MOSEC_EMBEDDING_MODEL | quote}}
-  {{- else if .Values.TEI_EMBEDDING_ENDPOINT }}
-  TEI_EMBEDDING_ENDPOINT: {{ .Values.TEI_EMBEDDING_ENDPOINT | quote }}
+  {{- if .Values.TEI_EMBEDDING_ENDPOINT }}
+  TEI_EMBEDDING_ENDPOINT: {{ tpl .Values.TEI_EMBEDDING_ENDPOINT . | quote }}
   {{- else if not .Values.LOCAL_EMBEDDING_MODEL }}
   TEI_EMBEDDING_ENDPOINT: "http://{{ .Release.Name }}-tei"
   {{- end }}
-  {{- if .Values.LOCAL_EMBEDDING_MODEL }}
-  EMBED_MODEL: {{ .Values.LOCAL_EMBEDDING_MODEL | quote }}
-  LOCAL_EMBEDDING_MODEL: {{ .Values.LOCAL_EMBEDDING_MODEL | quote }}
-  {{- end }}
-  {{- if .Values.REDIS_URL }}
-  REDIS_URL: {{ .Values.REDIS_URL | quote}}
+  {{- if eq "REDIS" .Values.RETRIEVER_BACKEND }}
+  RETRIEVER_COMPONENT_NAME: "OPEA_RETRIEVER_REDIS"
+  {{- if .Values.REDIS_HOST }}
+  REDIS_HOST: {{ tpl .Values.REDIS_HOST . | quote}}
   {{- else }}
-  REDIS_URL: "redis://{{ .Release.Name }}-redis-vector-db:6379"
+  REDIS_HOST: "{{ .Release.Name }}-redis-vector-db"
   {{- end }}
+  REDIS_PORT: {{ .Values.REDIS_PORT | quote }}
+  REDIS_SSL: {{ .Values.REDIS_SSL | quote }}
+  REDIS_PASSWORD: {{ .Values.REDIS_PASSWORD | quote }}
+  REDIS_USERNAME: {{ .Values.REDIS_USERNAME | quote }}
   INDEX_NAME: {{ .Values.INDEX_NAME | quote }}
+  EMBED_MODEL: {{ .Values.LOCAL_EMBEDDING_MODEL | quote }}
+  {{- else if eq "MILVUS" .Values.RETRIEVER_BACKEND }}
+  RETRIEVER_COMPONENT_NAME: "OPEA_RETRIEVER_MILVUS"
   {{- if .Values.MILVUS_HOST }}
-  MILVUS_HOST: {{ .Values.MILVUS_HOST | quote }}
+  MILVUS_HOST: {{ tpl .Values.MILVUS_HOST . | quote }}
   {{- else }}
   MILVUS_HOST: "{{ .Release.Name }}-milvus"
   {{- end }}
-  MILVUS: {{ .Values.MILVUS_HOST | quote }}
   MILVUS_PORT: {{ .Values.MILVUS_PORT | quote }}
-  {{- if .Values.COLLECTION_NAME }}
   COLLECTION_NAME: {{ .Values.COLLECTION_NAME | quote }}
+  LOCAL_EMBEDDING_MODEL: {{ .Values.LOCAL_EMBEDDING_MODEL | quote }}
+  {{- else }}
+  {{- cat "Invalid RETRIEVER_BACKEND:" .Values.RETRIEVER_BACKEND | fail }}
+  {{- end }}
+  {{- if .Values.global.HF_ENDPOINT }}
+  HF_ENDPOINT: {{ .Values.global.HF_ENDPOINT | quote}}
   {{- end }}
   EASYOCR_MODULE_PATH: "/tmp/.EasyOCR"
   http_proxy: {{ .Values.global.http_proxy | quote }}
   https_proxy: {{ .Values.global.https_proxy | quote }}
-  {{- if and (not .Values.REDIS_URL) (and (not .Values.TEI_EMBEDDING_ENDPOINT) (or .Values.global.http_proxy .Values.global.https_proxy)) }}
-  no_proxy: "{{ .Release.Name }}-tei,{{ .Release.Name }}-redis-vector-db,{{ .Values.global.no_proxy }}"
+  {{- if and (and (not .Values.MILVUS_HOST ) (not .Values.REDIS_HOST)) (and (not .Values.TEI_EMBEDDING_ENDPOINT) (or .Values.global.http_proxy .Values.global.https_proxy)) }}
+  no_proxy: "{{ .Release.Name }}-tei,{{ .Release.Name }}-redis-vector-db,{{ .Release.Name }}-milvus,{{ .Values.global.no_proxy }}"
   {{- else }}
-  no_proxy: {{ .Values.global.no_proxy | quote }}
+  no_proxy: {{ tpl .Values.global.no_proxy . | quote }}
   {{- end }}
   HF_HOME: "/tmp/.cache/huggingface"
   HUGGINGFACEHUB_API_TOKEN: {{ .Values.global.HUGGINGFACEHUB_API_TOKEN | quote}}

@@ -28,8 +28,36 @@ spec:
       serviceAccountName: {{ include "retriever-usvc.serviceAccountName" . }}
       securityContext:
         {{- toYaml .Values.podSecurityContext | nindent 8 }}
+      initContainers:
+        - name: wait-for-db
+          envFrom:
+            - configMapRef:
+                name: {{ include "retriever-usvc.fullname" . }}-config
+            {{- if .Values.global.extraEnvConfig }}
+            - configMapRef:
+                name: {{ .Values.global.extraEnvConfig }}
+                optional: true
+            {{- end }}
+          image: busybox:1.36
+          command: ["sh", "-c"]
+          args:
+            - |
+              {{- if eq "REDIS" .Values.RETRIEVER_BACKEND }}
+              TESTHOST=$(REDIS_HOST);
+              TESTPORT=$(REDIS_PORT);
+              {{- else if eq "MILVUS" .Values.RETRIEVER_BACKEND }}
+              TESTHOST=$(MILVUS_HOST);
+              TESTPORT=$(MILVUS_PORT);
+              {{- end }}
+              retry_count={{ .Values.retryCount | default 60 }};
+              j=1;
+              while ! nc -z ${TESTHOST} ${TESTPORT}; do
+                [[ $j -ge ${retry_count} ]] && echo "ERROR: ${TESTHOST}:${TESTPORT} is NOT reachable in $j seconds!" && exit 1;
+                j=$((j+1)); sleep 1;
+              done;
+              echo "${TESTHOST}:${TESTPORT} is reachable within $j seconds.";
       containers:
-        - name: {{ .Release.Name }}
+        - name: {{ .Chart.Name }}
           envFrom:
             - configMapRef:
                 name: {{ include "retriever-usvc.fullname" . }}-config

@@ -5,27 +5,36 @@
 # This is a YAML-formatted file.
 # Declare variables to be passed into your templates.
 
-tei:
-  enabled: false
-milvus:
-  enabled: false
-redis-vector-db:
-  enabled: false
-
 replicaCount: 1
 
-# Set it as a non-null string, such as true, if you want to enable logging facility,
-# otherwise, keep it as "" to disable it.
+# Set it as a non-null string, such as true, if you want to enable logging.
 LOGFLAG: ""
 
+# retriever-usvc needs to use different vector-DB backend service, e.g. Redis, Milvus
+# Default is to use Redis as vector-DB backend
+RETRIEVER_BACKEND: "REDIS"
+REDIS_HOST: ""
+REDIS_PORT: 6379
+REDIS_SSL: false
+REDIS_PASSWORD: ""
+REDIS_USERNAME: ""
+INDEX_NAME: "rag_redis"
+
+# Uncomment and set the following settings to use Milvus as vector-DB backend
+# RETRIEVER_BACKEND: "MILVUS"
+MILVUS_HOST: ""
+MILVUS_PORT: 19530
+COLLECTION_NAME: "rag_milvus"
+
+# retriever-usvc can do embedding locally or talk to remote TEI service to do embedding.
+# TEI_EMBEDDING_ENDPOINT takes precedence over LOCAL_EMBEDDING_MODEL.
+# text embedding inference service URL, e.g. http://<service-name>:<port>
 TEI_EMBEDDING_ENDPOINT: ""
+# local embedding model
 LOCAL_EMBEDDING_MODEL: ""
 
-REDIS_URL: ""
-INDEX_NAME: "rag-redis"
-
 image:
-  repository: opea/retriever-redis
+  repository: opea/retriever
   # Uncomment the following line to set desired image pull policy if needed, as one of Always, IfNotPresent, Never.
   # pullPolicy: ""
   # Overrides the image tag whose default is the chart appVersion.
@@ -52,7 +61,8 @@ podSecurityContext: {}
   # fsGroup: 2000
 
 securityContext:
-  readOnlyRootFilesystem: true
+  # NOTE: many dependent python modules(e.g. numpy, mpl, haystack, etc.) need to write to local disk
+  readOnlyRootFilesystem: false
   allowPrivilegeEscalation: false
   runAsNonRoot: true
   runAsUser: 1000
@@ -69,24 +79,13 @@ service:
   port: 7000
 
 resources:
-  # We usually recommend not to specify default resources and to leave this as a conscious
-  # choice for the user. This also increases chances charts run on environments with little
-  # resources, such as Minikube. If you do want to specify resources, uncomment the following
-  # lines, adjust them as necessary, and remove the curly braces after 'resources:'.
   # limits:
   #   cpu: 100m
   #   memory: 128Mi
   requests:
     cpu: 100m
     memory: 128Mi
 
-livenessProbe:
-  httpGet:
-    path: v1/health_check
-    port: retriever-usvc
-  initialDelaySeconds: 5
-  periodSeconds: 5
-  failureThreshold: 24
 readinessProbe:
   httpGet:
     path: v1/health_check
@@ -107,13 +106,6 @@ tolerations: []
 
 affinity: {}
 
-# milvus DB configurations
-MILVUS_HOST: ""
-MILVUS_PORT: ""
-COLLECTION_NAME: ""
-MOSEC_EMBEDDING_ENDPOINT: ""
-MOSEC_EMBEDDING_MODEL: ""
-
 global:
   http_proxy: ""
   https_proxy: ""
@@ -129,3 +121,11 @@ global:
 
   # Prometheus Helm install release name for serviceMonitor
   prometheusRelease: prometheus-stack
+
+# For CI tests only
+tei:
+  enabled: false
+milvus:
+  enabled: false
+redis-vector-db:
+  enabled: false