Helm chart values

The OX AI-Service integrates 3rd party AIs in App Suite UI, currently ChatGPT from OpenAI and LocalAI compatible models.

Please check out the full documentation here: OX AI-Service documentation

Chart Configuration

Parameter	Description	Default
image.repository	The image to be used for the deployment	appsuite-core-internal/ai-service
image.pullPolicy	The imagePullPolicy for the deployment	IfNotPresent
image.tag	The image tag, defaults to app version	""
defaultRegistry	Registry the image is pulled from	registry.open-xchange.com
imagePullSecrets	Image pull secrets, e.g. `[{name: ai-service-pullsecret}]`	[]
nameOverride	Override the chart name	""
fullnameOverride	Override the full name of the release resources	""
replicaCount	Number of pods; ignored when autoscaling is enabled	1
containerPort	Container port of the HTTP API	8080
resources	Container resource requests and limits	requests: cpu 500m, memory 384Mi; limits: memory 512Mi
autoscaling.enabled	Create a HorizontalPodAutoscaler	false
autoscaling.minReplicas	Minimum number of pods	1
autoscaling.maxReplicas	Maximum number of pods	100
autoscaling.targetMemoryUtilizationPercentage	Target memory utilization	80
autoscaling.targetCPUUtilizationPercentage	Target CPU utilization; not set by default	""
nodeSelector	Node selector for the pod	{}
tolerations	Tolerations for the pod	[]
affinity	Affinity rules for the pod	{}
podAnnotations	Annotations added to the pod	`logging.open-xchange.com/format: "appsuite-json"`
probe.liveness	Liveness probe on `/live` (port 9090); `enabled` plus any probe field	enabled, periodSeconds 10, failureThreshold 15
probe.readiness	Readiness probe on `/ready` (port 9090); `enabled` plus any probe field	enabled, initialDelaySeconds 1, periodSeconds 5, failureThreshold 2, timeoutSeconds 5
service.type	Kubernetes service type	ClusterIP
service.port	Kubernetes service port	80
bindAddr	IP address to listen for connections	"::"
origins	Allowed origins for CORS	*
logLevel	specify log level for service	"info"
exposeApiDocs	Expose API documentation via Swagger UI at /api-docs	false
ingress.enabled	Generate ingress resource	false
ingress.className	Ingress class name	""
ingress.annotations	Map of key-value pairs that will be added as annotations to the ingress resource	{}
ingress.hosts	Hosts the ingress serves (`- host: <name>`)	chart-example.local
ingress.tls	TLS configuration of the ingress	[]
istio.enabled	Create an Istio VirtualService	false
istio.hostname	Host of the VirtualService	""
istio.gateway	Gateway the VirtualService binds to	"istio-system/default-gateway"
istio.oauth	Protect `/admin` with an oauth2-proxy registered as Istio ext_authz provider, route `/oauth2` to it and label the pod for sidecar injection	false
istio.oauthProvider	Name of the ext_authz provider in the mesh config	"oauth2-proxy"
istio.oauthHostname	Service of the oauth2-proxy (port 4180)	"oauth2-proxy.oauth2-proxy.svc.cluster.local"
oidc.issuer	OIDC issuer URL, e.g. https://example.com/appsuite/ui	""
openaiSecret.enabled	Enable the secret for openai	true
openaiAPIKey	OpenAI API Key	""
openaiModerationModel	OpenAI moderation model	"omni-moderation-latest"
localaiSecret.enabled	Enable the secret for localai. May be optional, depending on your used service and model.	false
localaiAPIKey	localAI API Key. May be optional, depending on your used service and model.	""
mysqlSecret.enabled	Create the kubernetes secret for mysql (enable if you want to use the DB or provide own secret)	true
overrides.mysqlSecret	If you provide your own secret for mysql put the secret name here	""
overrides.openaiSecret	If you provide your own secret for OpenAI put the secret name here (key `openaiAPIKey`)	""
overrides.awsSecret	If you provide your own secret for AWS put the secret name here (keys `awsAccessKey`, `awsSecretKey`)	""
overrides.localaiSecret	If you provide your own secret for LocalAI put the secret name here (key `localaiAPIKey`)	""
overrides.mcpSessionEncryptionSecret	If you provide your own secret for the MCP session encryption key put the secret name here (key `mcpSessionEncryptionKey`)	""
mcpSessionEncryptionSecret.enabled	Create the kubernetes secret for the MCP session encryption key (from mcpSessionEncryptionKey)	false
mcpSessionEncryptionKey	MCP session encryption key (AES-256-GCM seed); generate with `openssl rand -hex 32`	""
mcp.allowedApiHosts	Comma-separated allow-list of Middleware hostnames for MCP auth (`*.example.com` wildcards supported); required when MCP is enabled	""
mcp.rateLimitAuthPerMin	Rate limit for `POST /api/mcp/auth` per user per minute	"5"
mcp.rateLimitToolsPerMin	Rate limit for MCP tool calls per user per minute	"60"
mcp.maxConcurrentCalls	MCP requests one user may have in flight at the same time (per pod)	"4"
mcp.allowedOrigins	Comma-separated origins allowed to call `POST /api/mcp/` from a browser; empty sends no CORS headers	""
mcp.publicUrl	MCP endpoint exactly as users enter it in a client, e.g. `https://appsuite.example.com/api/ai/mcp` (env `MCP_PUBLIC_URL`). OAuth for claude.ai and Claude Code is on only when this and `mcp.oauthAppSuiteUrl` are set; the discovery paths sit below this URL, and for claude.ai Anthropic's outbound range (see documentation/mcp.md) must reach the service	""
mcp.oauthAppSuiteUrl	App Suite that shows the OAuth consent dialog, e.g. `https://appsuite.example.com` (env `MCP_OAUTH_APP_SUITE_URL`); set together with `mcp.publicUrl`	""
mcp.maxGrantsPerUser	Connections (OAuth clients plus manual keys) one user may hold (env `MCP_MAX_GRANTS_PER_USER`); rotating the MCP session encryption key invalidates registered OAuth client ids	"10"
mcp.oauthAllowedRedirects	Security setting, recommended in production: comma-separated hostnames (`https`, exact match) and private-use schemes with their colon that OAuth clients may redirect to, e.g. `claude.ai,cursor:` (env `MCP_OAUTH_ALLOWED_REDIRECTS`); loopback is always allowed; empty allows every redirect, see documentation/mcp.md	""
mcp.externalInvitations	Covers event invitations, mail recipients, vacation notice text, and text sent to external attendees or organizers (event and task text changes, the display name of an invited attendee, comments on replies, declines and cancellations), and booked resources whose address is outside the user's mail domains and their subdomains. false refuses addresses that are not listed as users in the global address book, and addresses the lookup cannot check	false
mcp.scopeTiers	Object mapping the key tiers `read`, `write` and `destructive` to Middleware module scopes (env `MCP_SCOPE_TIERS`); a tier left out keeps the mapping derived from the tools; invalid scopes stop the service at startup	{}
database.enabled	Use Database. This is mandatory for AI chat, usage tracking, and rate limiting and therefore deprecated. With App Suite 8.34, the default changed to true.	true
database.host	SQL server hostname	RELEASE-NAME-ai-service-sql
database.port	SQL server port	3306
database.name	Database name	RELEASE-NAME-ai-service
database.connections	Number of concurrent connections to the DB server	"10"
database.user	DB User with access rights to sqlDB	""
database.password	Password of the database user	""
migration.jobId	Rotation token for the migrate Job name suffix (e.g. $CI_JOB_ID); falls back to release revision	""
cron.cleanupDb	Database cleanup interval (Cron notation)	"0 4 * * *"
cron.sidecarInjection.disabled	Disable Istio Sidecar Injection for CronJobs	true
openaiBaseUrl	Url of the OpenAI service (internal or localAI use only)	""
localaiBaseUrl	Url of the localAI service. Falls back to to openaiBaseUrl if configured.	""
aiEgressProxy.url	Outbound proxy for the AI provider APIs (OpenAI, LocalAI, Azure, AWS Bedrock)	""
aiEgressProxy.noProxy	Hosts that stay direct; NO_PROXY syntax (host, .domain, .domain, host:port, IP, ``), evaluated the same way for every provider including Bedrock	""
aiEgressProxy.secret.enabled	Render aiEgressProxy.url into a secret; required when the url carries credentials	false
overrides.aiEgressProxySecret	Your own secret for the proxy url (key `aiEgressProxyUrl`); needs secret.enabled=false	""
extraCaCerts.certificates	Inline PEM bundle of additional CA certificates; the chart creates the Secret	""
extraCaCerts.existingSecret	Name of an existing Secret holding the PEM bundle; takes precedence over `certificates`	""
extraCaCerts.key	Key inside the Secret that holds the bundle	"ca.crt"
localaiModerationDisabled	Legacy switch; LocalAI models are never moderated, regardless of this value	true
accounting.enabled	Enable the accounting feature	false
chatAssistantContext	First (system) message that is sent to the ai chat to setup the mood and general context.	"You are a helpful AI assistant."
enforceDefaultModel	Always use the configured default model regardless of the clients request	false
disableAPI	Disable all API endpoints. Only UI source files will be delivered	false
moderationDisabled	Disable moderation for all models (testing only)	false
awsSecret.enabled	Enable the aws Secret: static keys (renders `AWS_ACCESS_KEY_ID`/`AWS_SECRET_ACCESS_KEY`) or a Bedrock API key (`awsBearerToken`); leave false for IRSA (`serviceAccount.annotations`) or EKS Pod Identity (association on `serviceAccount.name`)	false
awsRegion	Region of the AWS user with Bedrock & Claude enabled; rendered whenever set, independent of awsSecret.enabled	""
awsAccessKey	AWS IAM access key with Bedrock & Claude enabled (rendered as `AWS_ACCESS_KEY_ID`)	""
awsSecretKey	AWS IAM secret key with Bedrock & Claude enabled (rendered as `AWS_SECRET_ACCESS_KEY`)	""
awsBearerToken	Bedrock API key, rendered as `AWS_BEARER_TOKEN_BEDROCK` from the aws Secret (third path next to static keys and IAM roles; needs awsSecret.enabled)	""
awsClaudeUrl	URL of the AWS Bedrock/Claude service (internal use only)	""
bedrock.guardrail.id	Bedrock Guardrail identifier for output moderation; unset falls back to the classifier	""
bedrock.guardrail.version	Bedrock Guardrail version	""
bedrock.moderationModelId	Model or inference profile ID for the classifier moderation fallback	""
bedrock.moderationDisabled	Disable Bedrock output moderation independently of moderationDisabled	false
bedrock.moderationFailClosed	Retract the answer in the UI (`{reset}`) when the Bedrock moderation call fails instead of delivering it unchecked; errors count in `ai_moderation_errors_count` either way	false
serviceAccount.create	Create a dedicated ServiceAccount for the pods; set false and name an existing one in `serviceAccount.name` when an operator already created the SA (e.g. for IRSA), otherwise the upgrade fails because the SA is not managed by Helm	true
serviceAccount.name	Name of the ServiceAccount; defaults to the chart fullname	""
serviceAccount.annotations	Annotations on the ServiceAccount, e.g. `eks.amazonaws.com/role-arn` for IRSA	{}
serviceAccount.automountServiceAccountToken	Set on the ServiceAccount resource itself	false
automountServiceAccountToken	Pod-level token automount; this service never calls the Kubernetes API	false
basicAuth.enabled	Enable the secret for basic authentication (Admin routes)	false
basicAuth.user	Username for basic authentication (Admin routes)	""
basicAuth.password	Password for basic authentication (Admin routes)	""
overrides.basicAuthSecret	If you provide your own secret for basic auth put the secret name here	""
bearerAuth.enabled	Enable the secret for bearer authentication (Admin routes)	false
bearerAuth.keys	Comma separated keys to use with bearer auth (Admin Routes)	""
overrides.bearerAuthSecret	If you provide your own secret for bearer auth put the secret name here	""
qdrant.enabled	Enable Qdrant vector database for semantic search	false
qdrant.url	URL of the Qdrant instance	""
qdrant.secret.enabled	Create the kubernetes secret for the Qdrant API key (from qdrant.apiKey)	false
qdrant.apiKey	Qdrant API Key (used when qdrant.secret.enabled is true)	""
qdrant.timeoutMs	Timeout in milliseconds for Qdrant requests	""
qdrant.collectionName	Collection name used for mail embeddings	"appsuite_mail_embeddings"
qdrant.distance	Distance metric for vector similarity	"Cosine"
qdrant.mSize	HNSW `M` parameter for the main vector index	"16"
qdrant.efConstruct	HNSW `ef_construct` parameter for index build	"100"
qdrant.payloadMSize	HNSW `M` parameter for the payload index	"16"
qdrant.embeddingsOnDisk	Store vectors on disk in Qdrant	"false"
qdrant.hnswIndexOnDisk	Store the HNSW index on disk in Qdrant	"false"
overrides.qdrantSecret	If you provide your own secret for Qdrant put the secret name here (key `qdrantAPIKey`)	""
pgvector.enabled	Enable PostgreSQL + pgvector as vector store (alternative to Qdrant, enable only one)	false
pgvector.url	PostgreSQL connection string without the password	""
pgvector.secret.enabled	Create the kubernetes secret for the pgvector password (from pgvector.password)	false
pgvector.password	Password for the pgvector database user	""
overrides.pgvectorSecret	If you provide your own secret for pgvector put the secret name here (key `pgvectorPassword`)	""
openaiEmbeddingModel	Embedding model used for Smart Search when using OpenAI/localAI compatible embeddings	""
embeddingSize	Embedding vector size written to the vector store	384
mailFetchConcurrency	Number of concurrent mail fetch operations during embedding/backfill	5
mailEmbedConcurrency	Number of concurrent embedding operations during embedding/backfill	2
embeddingQueueTtlMs	TTL for embedding jobs in milliseconds	7200000
embedRetryMax	Maximum number of retries for failed embedding jobs	4
oauth2.enabled	Enable OAuth2/OIDC authentication for the admin routes	false
oauth2.domain	Domain of the openid connect capable OAuth2 server	""
plans	List of plans with limits for users and brands	See full documentation for details
modelsConfig	List of models that the service supports	See full documentation for details
config	Application configuration that is also relevent for the UI parts	See full documentation for details
monitoring.grafana.enabled	Enable Grafana integration for dashboards and init job	false
monitoring.grafana.grafanaOperator	Enable Grafana Operator integration (adds label and skips Node.js job)	false
monitoring.grafana.url	URL of the Grafana instance	""
monitoring.grafana.secretName	Name of the Kubernetes secret for Grafana credentials	""
monitoring.grafana.username	Username for Grafana API access (if required)	""
monitoring.grafana.password	Password for Grafana API access (if required)	""
monitoring.grafana.apiToken	API token for Grafana (preferred for automation)	""
monitoring.grafana.folder	Target folder in Grafana for dashboards	""
monitoring.grafana.ttlInitJob	TTL in seconds for the Grafana init job after completion	600
monitoring.prometheus.url	URL of the Prometheus instance used for metrics collection	""
monitoring.prometheus.dataSource	Override the Prometheus datasource name in Grafana	""
monitoring.prometheus.kube-prometheus	Enable ServiceMonitor creation for kube-prometheus environments	false
monitoring.prometheus.serviceMonitor.namespace	Target namespace for ServiceMonitor	""
monitoring.prometheus.serviceMonitor.interval	Scrape interval for ServiceMonitor	30s
monitoring.prometheus.serviceMonitor.scrapeTimeout	Scrape timeout for ServiceMonitor	""
monitoring.prometheus.serviceMonitor.honorLabels	Honor labels for ServiceMonitor	false
monitoring.prometheus.serviceMonitor.additionalLabels	Additional labels for ServiceMonitor	{}
monitoring.prometheus.serviceMonitor.annotations	Annotations for ServiceMonitor	{}
monitoring.prometheus.serviceMonitor.relabelings	Relabeling configuration for ServiceMonitor	maps `__meta_kubernetes_pod_label_app_kubernetes_io_name` to `app`
monitoring.prometheus.serviceMonitor.metricRelabelings	Metric relabeling configuration for ServiceMonitor	[]

The cleanup CronJob and the migration Job run outside the Deployment, so their pods do not carry the Deployment's pod labels (app.kubernetes.io/name plus instance is the Service selector). Kubernetes only adds job-name and controller-uid, which change on every run. To select the job pods in a NetworkPolicy, use the stable labels the chart sets:

podSelector:
  matchLabels:
    app.kubernetes.io/instance: as8          # your Helm release name
    app.kubernetes.io/component: cleanupdb   # or migratedb

AWS Bedrock credentials

claude-aws models pick up credentials from the AWS SDK default chain, so the chart supports four ways to deploy them:

Static keys: set awsSecret.enabled=true (or overrides.awsSecret for a pre-existing Secret) plus awsAccessKey/awsSecretKey, and awsRegion as a plain value. The pod gets AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY from the Secret.
Bedrock API key: set awsSecret.enabled=true and awsBearerToken; the pod gets AWS_BEARER_TOKEN_BEDROCK from the Secret and no static keys, even when awsAccessKey and awsSecretKey are still set (the token takes precedence, so stale keys cannot interfere). Use a long-term key bound to an IAM user: short-term keys expire after about 12 hours and are valid in one region only, after which every Bedrock call fails with an authentication error. With a pre-existing overrides.awsSecret only the static keys are read from it and awsBearerToken is ignored.
IRSA: leave awsSecret.enabled=false and set serviceAccount.annotations["eks.amazonaws.com/role-arn"] to the IAM role ARN. No static keys are rendered; the SDK picks up AWS_ROLE_ARN/AWS_WEB_IDENTITY_TOKEN_FILE that the EKS webhook injects into the ServiceAccount's pods.
EKS Pod Identity: leave awsSecret.enabled=false, no annotation; create the Pod Identity association for the chart's ServiceAccount (serviceAccount.name, default the chart fullname). The SDK reads the credentials from the container credential endpoint.

In both role-based modes awsRegion still needs to be set (or AWS_REGION supplied by the cluster's environment), since it is rendered independently of the secret.

Security Context

Security contexts follow the ox-common pattern with defaults for Wolfi-based images (nonroot UID 65532):

Parameter	Description	Default
podSecurityContext	Explicit pod security context (overrides defaults)	{}
securityContext	Explicit container security context (overrides defaults)	{}
defaultPodSecurityContext.fsGroup	Default pod fsGroup (Wolfi nonroot)	65532
defaultSecurityContext.runAsNonRoot	Default run as non-root	true
defaultSecurityContext.runAsUser	Default user ID (Wolfi nonroot)	65532
defaultSecurityContext.runAsGroup	Default group ID (Wolfi nonroot)	65532
defaultSecurityContext.allowPrivilegeEscalation	Default prevent privilege escalation	false
defaultSecurityContext.readOnlyRootFilesystem	Default read-only root filesystem	true
defaultSecurityContext.capabilities.drop	Default dropped capabilities	[ALL]
global.useDefaultSecurityContext	Enable default security contexts	false

To enable Wolfi security defaults, set global.useDefaultSecurityContext=true.

Helm chart values

Chart Configuration

Job pod labels

AWS Bedrock credentials

Security Context