Configuration

All configuration is via Helm values. Override defaults with --set or a custom values file:

helm install gerty gerty/gerty -f my-values.yaml

Global

Parameter	Type	Default	Description
nameOverride	string	""	Override the chart name
fullnameOverride	string	""	Override the full release name
image.registry	string	ghcr.io/gerty-labs	Container image registry
image.tag	string	Chart appVersion	Image tag for all components
image.pullPolicy	string	IfNotPresent	Image pull policy

Agent

Parameter	Type	Default	Description
agent.image.repository	string	gerty-agent	Agent image name
agent.image.digest	string	""	Image digest override
agent.resources.requests.cpu	string	50m	CPU request
agent.resources.requests.memory	string	50Mi	Memory request
agent.resources.limits.cpu	string	100m	CPU limit
agent.resources.limits.memory	string	100Mi	Memory limit
agent.scrapeInterval	string	30s	Kubelet scrape interval
agent.pushInterval	string	5m	Server push interval
agent.nodeSelector	object	{}	Node selector
agent.tolerations	list	[]	Tolerations

Server

Parameter	Type	Default	Description
server.image.repository	string	gerty-server	Server image name
server.image.digest	string	""	Image digest override
server.replicas	int	1	Replica count
server.resources.requests.cpu	string	250m	CPU request
server.resources.requests.memory	string	256Mi	Memory request
server.resources.limits.cpu	string	500m	CPU limit
server.resources.limits.memory	string	512Mi	Memory limit
server.service.type	string	ClusterIP	Service type
server.service.port	int	8080	Service port
server.nodeSelector	object	{}	Node selector
server.tolerations	list	[]	Tolerations

Server Persistence

Gerty persists aggregator state so recommendations survive pod restarts. By default, it uses an embedded database backed by a PVC. For multi-AZ clusters, an external PostgreSQL database can be used instead.

Parameter	Type	Default	Description
server.persistence.enabled	bool	true	Enable state persistence
server.persistence.storageClass	string	""	PVC storage class (empty = cluster default)
server.persistence.size	string	1Gi	PVC size
server.persistence.accessModes	list	[ReadWriteOnce]	PVC access modes
server.externalDatabase.enabled	bool	false	Use external PostgreSQL instead of embedded storage
server.externalDatabase.url	string	""	PostgreSQL connection string

TIP

PVCs bind to a single availability zone. In multi-AZ clusters, either pin the server to an AZ via nodeAffinity, use a cross-AZ StorageClass, or set server.externalDatabase.enabled=true with a PostgreSQL URL.

AI Reasoning

Parameter	Type	Default	Description
slm.enabled	bool	false	Enable AI reasoning
slm.tier	string	standard	Intelligence tier (lite, standard, or premium)
slm.scaling.maxMemoryBudget	string	12Gi	Max total RAM Gerty can consume for AI reasoning
slm.scaling.maxCpuBudget	string	4	Max total CPU cores for AI reasoning
slm.scaling.minClusterHeadroom	string	20%	Never consume more than this % of free cluster resources
slm.persistence.enabled	bool	false	Enable persistent model storage
slm.persistence.size	string	5Gi	PVC size
slm.persistence.storageClass	string	""	Storage class

Intelligence Tiers

All tiers are included at every pricing level. The tier determines reasoning depth, not workload capacity.

slm.tier	Reasoning Depth	Best For
lite	Fast scanning, good recommendations	Most clusters, tight resource budgets
standard	Deeper analysis for complex workloads	Production clusters with mixed workload types
premium	Maximum reasoning quality	Large clusters with complex, heterogeneous workloads

Gerty's AI scales automatically based on demand. Standard and Premium tiers include a deeper analysis layer that scales from zero when needed, then scales back down. Gerty checks cluster headroom before scaling and degrades gracefully if resources are tight.

GitOps

Parameter	Type	Default	Description
gitops.provider	string	""	Git provider (github or gitlab). Auto-detected from repo URL if not set
gitops.token	string	""	Personal access token or deploy token
gitops.tokenSecretRef	string	""	Reference to an existing Secret containing the token
gitops.gitlab.url	string	https://gitlab.com	GitLab instance URL (for self-hosted)
gitops.prTemplate	string	""	Custom PR/MR description template (Go text/template). Uses built-in default if empty

PR Template

The default PR template includes workload name, namespace, pattern classification, confidence score, resource changes table, metrics summary, and risk assessment. All values are populated from the rules engine. If AI reasoning is enabled, an optional reasoning section is appended.

Override with a custom Go text/template. Available fields:

Field	Type	Description
.Workload	string	Workload name (e.g. deployment/api-gateway)
.Namespace	string	Kubernetes namespace
.Pattern	string	Classification (steady, burstable, batch)
.Confidence	float64	Confidence score (0.0 - 1.0)
.ObservationDays	int	Number of days of metrics data
.Changes	[]Change	List of resource changes (.Resource, .Current, .Recommended, .Delta)
.Metrics	Metrics	Summary metrics (.CPUP95, .MemP95, .Samples)
.Risk	string	Risk level (LOW, MEDIUM, HIGH)
.Reasoning	string	AI explanation (empty if AI reasoning disabled)

Integrations

Parameter	Type	Default	Description
gcpMarketplace.enabled	bool	false	Enable GCP Marketplace integration
serviceAccount.create	bool	true	Create a ServiceAccount
serviceAccount.name	string	""	ServiceAccount name override
serviceAccount.annotations	object	{}	ServiceAccount annotations
grafana.dashboards.enabled	bool	false	Deploy Grafana dashboard ConfigMap
networkPolicy.enabled	bool	true	Enable NetworkPolicy
networkPolicy.allowExternalIngress	bool	true	Allow external ingress to server
slack.enabled	bool	false	Enable Slack notifications
slack.webhookURL	string	""	Slack webhook URL
slack.channel	string	#gerty	Slack channel
slack.digestInterval	string	1h	Digest send interval
slack.minSeverity	string	optimisation	Minimum severity to notify