Kubernetes Helm Values Generation

Automatic generation of Kubernetes Helm values.yaml files from Jan Server configuration.

Overview

The K8s values generator maps Jan Server's unified configuration to Helm chart values, enabling:

Single source of truth: Configuration drives both local and K8s deployments
Environment-specific overrides: Development, staging, production profiles
Consistent deployment: Same config structure across all environments
Type safety: Generated from Go structs with validation

Architecture

config/defaults.yaml
 v
 Config Loader
 v
Values Generator -----> values-dev.yaml (1 replica, minimal resources)
 +------------> values-prod.yaml (3 replicas, full resources, persistence)
(Add values-staging.yaml as needed by copying values-dev.yaml)

Generated Structure

Global Values

global:
 environment: development
 imagePullPolicy: IfNotPresent
 labels:
 app.kubernetes.io/name: jan-server
 app.kubernetes.io/version: 1.0.0
 app.kubernetes.io/environment: development

Service Values

Each service gets:

Deployment: replicas, image, resources
Service: type, ports
Health Checks: liveness/readiness probes
Config: ConfigMaps for non-sensitive config
Secrets: References to K8s secrets

Example:

services:
 llm-api:
 enabled: true
 replicaCount: 2
 image:
 repository: jan-llm-api
 tag: 1.0.0
 service:
 type: ClusterIP
 port: 8080
 targetPort: 8080
 resources:
 limits:
 cpu: 1000m
 memory: 1Gi
 requests:
 cpu: 500m
 memory: 512Mi
 healthChecks:
 livenessProbe:
 httpGet:
 path: /health
 port: 8080
 initialDelaySeconds: 30
 readinessProbe:
 httpGet:
 path: /health
 port: 8080
 initialDelaySeconds: 10
 configMap:
 LOG_LEVEL: info
 LOG_FORMAT: json
 secrets:
 - database-credentials
 - keycloak-credentials

Infrastructure Values

Database and auth configuration:

infrastructure:
 database:
 postgres:
 enabled: true
 host: api-db
 port: 5432
 database: jan_llm_api
 user: jan_user
 passwordSecret: postgres-password
 maxConnections: 100
 resources:
 limits:
 cpu: 2000m
 memory: 2Gi
 persistence:
 enabled: true
 size: 10Gi
 auth:
 keycloak:
 enabled: true
 baseURL: http://keycloak:8085
 adminUser: admin
 passwordSecret: keycloak-admin-password
 resources:
 limits:
 cpu: 1000m
 memory: 1Gi

Environment Profiles

Development

Single replica per service
Minimal resources (100m CPU, 128Mi RAM requests)
imagePullPolicy: Never (use local images)
Persistence disabled
Lower health check thresholds

Staging

2 replicas per service
Moderate resources (250m CPU, 256Mi RAM requests)
imagePullPolicy: IfNotPresent
Persistence enabled (20Gi)
Production-like settings

Production

3 replicas per service
Full resources (500m CPU, 512Mi RAM requests)
imagePullPolicy: Always
Persistence enabled (50Gi)
Strict health checks
Higher failure thresholds

Usage

Programmatic Generation

package main

import (
 "context"
 "github.com/janhq/jan-server/pkg/config"
 "github.com/janhq/jan-server/pkg/config/k8s"
)

func main() {
 // Load configuration
 loader:= config.NewConfigLoader("development", "config/defaults.yaml")
 cfg, _:= loader.Load(context.Background())

 // Create generator
 generator:= k8s.NewValuesGenerator(cfg)

 // Generate base values
 generator.GenerateToFile("values.yaml")

 // Generate with environment overrides
 values, _:= generator.GenerateWithOverrides("production")
 // Use values...
}

Example Program

See pkg/config/k8s/examples/generate_values.go:

cd pkg/config/k8s/examples
go run generate_values.go

Generates:

values-dev.yaml - Development environment
values-prod.yaml - Production environment

Configuration Mapping

Config Path	Helm Values Path	Notes
`meta.version`	`global.labels["app.kubernetes.io/version"]`, `*.image.tag`	Version across all components
`meta.environment`	`global.environment`, `global.labels["app.kubernetes.io/environment"]`	Environment name
`services.llm_api.http_port`	`services.llm-api.service.port`	Service port mapping
`services.llm_api.log_level`	`services.llm-api.configMap.LOG_LEVEL`	Config as ConfigMap
`infrastructure.database.postgres.*`	`infrastructure.database.postgres.*`	Direct mapping
`infrastructure.auth.keycloak.*`	`infrastructure.auth.keycloak.*`	Direct mapping

Secret Management

Important: The generator creates references to secrets, not the secrets themselves.

Secret references in values:

services:
 llm-api:
 secrets:
 - database-credentials # -> maps to K8s Secret
 - keycloak-credentials

You must create K8s secrets separately:

kubectl create secret generic database-credentials \
 --from-literal=password=<db-password>

kubectl create secret generic keycloak-credentials \
 --from-literal=admin-password=<keycloak-password>

Secrets are managed by DevOps via Kubernetes Secrets, HashiCorp Vault, or environment variables.

Resource Sizing

Default Resources

Service	CPU Request	CPU Limit	Memory Request	Memory Limit
llm-api	500m	1000m	512Mi	1Gi
mcp-tools	250m	500m	256Mi	512Mi
media-api	250m	500m	256Mi	512Mi
response-api	250m	500m	256Mi	512Mi
postgres	1000m	2000m	1Gi	2Gi
keycloak	500m	1000m	512Mi	1Gi

Scaling Recommendations

For production workloads:

High traffic: Increase replicas (3-5+)
Heavy AI workloads: Increase llm-api resources (2-4 CPU, 2-4Gi RAM)
Large databases: Increase postgres resources and persistence size

Health Checks

Liveness Probes

Detect crashed containers
Restart unhealthy containers
Higher failure threshold (3-5)

Readiness Probes

Detect startup completion
Remove from service during issues
Lower failure threshold (3)

Generated probes:

livenessProbe:
 httpGet:
 path: /health
 port: 8080
 initialDelaySeconds: 30
 periodSeconds: 10
 timeoutSeconds: 5
 failureThreshold: 3
readinessProbe:
 httpGet:
 path: /health
 port: 8080
 initialDelaySeconds: 10
 periodSeconds: 5
 timeoutSeconds: 3
 failureThreshold: 3

Integration with Helm

Deploy with Generated Values

# Generate values
go run pkg/config/k8s/examples/generate_values.go

# Install chart with generated values
helm install jan-server k8s/jan-server \
 -f pkg/config/k8s/examples/values-prod.yaml

Override Specific Values

# Use generated base + custom overrides
helm install jan-server k8s/jan-server \
 -f values-prod.yaml \
 --set services.llm-api.replicaCount=5

CI/CD Integration

#.github/workflows/deploy.yml
- name: Generate Helm Values
 run: |
 go run pkg/config/k8s/examples/generate_values.go
 
- name: Deploy to K8s
 run: |
 helm upgrade --install jan-server k8s/jan-server \
 -f values-prod.yaml \
 --namespace jan-server

Limitations

Not a replacement for Helm templating: Generator creates values, Helm chart still needed
Basic resource sizing: May need tuning based on actual workload
Secrets not generated: K8s secrets must be created separately by DevOps team
No autoscaling: HPA configuration not generated (add manually)

Kubernetes Helm Values Generation

Kubernetes Helm Values Generation

Overview

Architecture

Generated Structure

Global Values

Service Values

Infrastructure Values

Environment Profiles

Development

Staging

Production

Usage

Programmatic Generation

Example Program

Configuration Mapping

Secret Management

Resource Sizing

Default Resources

Scaling Recommendations

Health Checks

Liveness Probes

Readiness Probes

Integration with Helm

Deploy with Generated Values

Override Specific Values

CI/CD Integration

Limitations

Future Enhancements

See Also

On this page