agora/contoso_hypermarket/charts/cerebral-api/templates/cerebral-api.yaml (167 lines of code) (raw):
apiVersion: apps/v1
kind: Deployment
metadata:
name: cerebral-api-deployment
spec:
replicas: 1
selector:
matchLabels:
app: cerebral-api
template:
metadata:
labels:
app: cerebral-api
spec:
initContainers:
- name: init-documents
image: mcr.microsoft.com/cbl-mariner/base/core:2.0
command:
- sh
- -c
- |
tdnf update -y
tdnf install wget -y
cd /app/docs
wget -O "Automated_Checkout_Manual.pdf" "https://download.microsoft.com/download/3ae1d7aa-a642-48cf-b848-67b4eaa81292/Automated Checkout System User Manual for Contoso Hypermarket.pdf" && \
wget -O "SmartShelf_Manual.pdf" "https://download.microsoft.com/download/3ae1d7aa-a642-48cf-b848-67b4eaa81292/SmartShelf System User Manual for Contoso Hypermarket.pdf" && \
wget -O "Refrigeration_Manual.pdf" "https://download.microsoft.com/download/3ae1d7aa-a642-48cf-b848-67b4eaa81292/Refrigeration System User Manual for Contoso Hypermarket.pdf" && \
wget -O "HVAC_Manual.pdf" "https://download.microsoft.com/download/3ae1d7aa-a642-48cf-b848-67b4eaa81292/HVAC System User Manual for Contoso Hypermarket.pdf" && \
wget -O "Scale_Manual.pdf" "https://download.microsoft.com/download/3ae1d7aa-a642-48cf-b848-67b4eaa81292/Scale System User Manual for Contoso Hypermarket.pdf"
volumeMounts:
- name: cerebral-docs
mountPath: /app/docs
- name: init-slm-model
image: mcr.microsoft.com/azurelinux/base/python:3.12
command:
- sh
- -c
- |
pip install -U "huggingface_hub[cli]"
huggingface-cli download microsoft/Phi-3-mini-4k-instruct-onnx \
--include "cpu_and_mobile/cpu-int4-rtn-block-32-acc-level-4/*" \
--local-dir /app/models
if [ ! -d "/app/models/phi3" ]; then
mkdir /app/models/phi3
fi
volumeMounts:
- name: cerebral-slm-model
mountPath: /app/models
containers:
- name: cerebral-api
image: {{ .Values.images.cerebral.repository }}:{{ .Values.images.cerebral.tag }}
imagePullPolicy: {{ .Values.images.cerebral.pullPolicy }}
ports:
- containerPort: 5003
env:
- name: AZURE_OPENAI_API_KEY
valueFrom:
secretKeyRef:
name: azure-openai-secret
key: azure-openai-key
- name: AZURE_OPENAI_ENDPOINT
valueFrom:
secretKeyRef:
name: azure-openai-secret
key: azure-openai-endpoint
- name: CHATGPT_MODEL
valueFrom:
secretKeyRef:
name: azure-openai-secret
key: azure-openai-deployment-name
- name: OPENAI_API_VERSION
valueFrom:
secretKeyRef:
name: azure-openai-secret
key: azure-openai-api-version
- name: INFLUXDB_URL
value: "http://influxdb.contoso-hypermarket.svc.cluster.local:8086"
- name: INFLUXDB_BUCKET
value: "hypermarket"
- name: INFLUXDB_TOKEN
value: "secret-token"
- name: INFLUXDB_ORG
value: "InfluxData"
- name: REDIS_URL
value: "redis://redis-service.contoso-hypermarket.svc.cluster.local:6379"
- name: SQL_DATABASENAME
value: "RetailDB"
- name: SQL_SERVER
value: "mssql-service"
- name: SQL_USERNAME
value: "SA"
- name: SQL_PASSWORD
valueFrom:
secretKeyRef:
name: azure-sqlpassword-secret
key: azure-sqlpassword-secret
- name: WHISPER_MODEL_PATH
value: ""
- name: AZURE_AI_SPEECH_KEY
valueFrom:
secretKeyRef:
name: azure-openai-secret
key: azure-openai-key
- name: AZURE_AI_SPEECH_REGION
valueFrom:
secretKeyRef:
name: azure-openai-secret
key: region
- name: USE_LOCAL_LLM
value: "false"
- name: MODEL_PATH
value: "/app/models"
- name: CHROMA_HOST
value: "chroma-db-service.contoso-hypermarket.svc.cluster.local"
- name: CHROMA_PORT
value: "8040"
- name: CHROMA_COLLECTION
value: "documents"
{{- if (lookup "v1" "Namespace" "" "gpu-operator") }}
resources:
limits:
nvidia.com/gpu: "1"
{{- end }}
volumes:
- name: cerebral-docs
persistentVolumeClaim:
claimName: cerebral-docs
- name: cerebral-slm-model
persistentVolumeClaim:
claimName: cerebral-slm-model
---
apiVersion: v1
kind: Service
metadata:
name: cerebral-api-service
spec:
type: LoadBalancer
selector:
app: cerebral-api
ports:
- protocol: TCP
port: 5003
targetPort: 5003
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: cerebral-docs
namespace: contoso-hypermarket
spec:
accessModes:
- ReadWriteOnce
resources:
requests:
storage: 1Gi
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: cerebral-slm-model
namespace: contoso-hypermarket
spec:
accessModes:
- ReadWriteOnce
resources:
requests:
storage: 10Gi