Helm Configuration¶

Configuration reference for HolmesGPT Helm chart.

Quick Links:

Installation Tutorial - Step-by-step setup guide
values.yaml - Complete configuration reference
HTTP API Reference - Test your deployment

Basic Configuration¶

# values.yaml
# Image settings
image: holmes:0.0.0
registry: robustadev

# Logging level
logLevel: INFO

# send exceptions to sentry
enableTelemetry: true

# Resource limits
resources:
  requests:
    cpu: 100m
    memory: 1024Mi
  limits:
    memory: 1024Mi

# Enabled/disable/customize specific toolsets
toolsets:
  kubernetes/core:
    enabled: true
  kubernetes/logs:
    enabled: true
  robusta:
    enabled: true
  internet:
    enabled: true
  prometheus/metrics:
    enabled: true
  ...

Configuration Options¶

Essential Settings¶

Parameter	Description	Default
`additionalEnvVars`	Environment variables (API keys, etc.)	`[]`
`extraEnvVarsSecrets`	List of Kubernetes Secret names whose keys are auto-mounted as env vars on the Holmes pod. Enables the `envRef:VAR` sugar in `modelList`.	`[]`
`toolsets`	Enable/disable specific toolsets	(see values.yaml)
`modelList`	Configure multiple AI models for UI selection. See Using Multiple Providers	`{}`
`openshift`	Enable OpenShift compatibility mode	`false`
`image`	HolmesGPT image name	`holmes:0.0.0`
`registry`	Container registry	`robustadev`
`logLevel`	Log level (DEBUG, INFO, WARN, ERROR)	`INFO`
`enableTelemetry`	Send exception reports to sentry	`true`
`certificate`	Base64 encoded custom CA certificate for outbound HTTPS requests (e.g., LLM API via proxy)	`""`
`sentryDSN`	Sentry DSN for telemetry	(see values.yaml)

API Key Configuration¶

The most important configuration is setting up API keys for your chosen AI provider:

additionalEnvVars:
- name: OPENAI_API_KEY
  value: "your-api-key"
# Or load from secret:
# - name: OPENAI_API_KEY
#   valueFrom:
#     secretKeyRef:
#       name: holmes-secrets
#       key: openai-api-key

Simplified API Key Configuration (`extraEnvVarsSecrets` + `envRef:` sugar)¶

When you configure several models in modelList, listing every key twice — once in additionalEnvVars and once in the model config — becomes noisy. The extraEnvVarsSecrets field mounts one or more Kubernetes Secrets onto the Holmes pod with a single line each, and the envRef:VAR shorthand in modelList is rewritten at chart-render time to the runtime template {{ env.VAR }}.

Create the Secret once, with one key per API key you need:

kubectl create secret generic holmes-secrets \
  --from-literal=OPENAI_API_KEY=sk-... \
  --from-literal=ANTHROPIC_API_KEY=sk-ant-... \
  -n <namespace>

Then reference it in values.yaml:

extraEnvVarsSecrets:
  - holmes-secrets

modelList:
  gpt-4.1:
    model: openai/gpt-4.1
    api_key: envRef:OPENAI_API_KEY   # → "{{ env.OPENAI_API_KEY }}" at render time
    temperature: 0
  claude-sonnet-4:
    model: anthropic/claude-sonnet-4-5-20250929
    api_key: envRef:ANTHROPIC_API_KEY
    temperature: 1

Notes:

extraEnvVarsSecrets is a list — pass multiple Secret names to split keys across secrets (e.g. one Secret per provider). Every listed Secret is mounted via envFrom.secretRef, sharing a single env-var namespace.
The Secret keys must be valid env-var names ([A-Za-z_][A-Za-z0-9_]*) — they become environment variables verbatim via envFrom.secretRef.
envRef: sugar can be used on any string field in modelList, not just api_key (e.g. aws_access_key_id: envRef:AWS_ACCESS_KEY_ID).
Existing configs using additionalEnvVars + {{ env.OPENAI_API_KEY }} keep working unchanged — extraEnvVarsSecrets is opt-in.
extraEnvVarsSecrets can coexist with additional_env_froms; both blocks are merged into the pod's envFrom.

Toolset Configuration¶

Control which capabilities HolmesGPT has access to:

toolsets:
  kubernetes/core:
    enabled: true      # Core Kubernetes functionality
  kubernetes/logs:
    enabled: true      # Kubernetes logs access
  robusta:
    enabled: true      # Robusta platform integration
  internet:
    enabled: true      # Internet access for documentation
  prometheus/metrics:
    enabled: true      # Prometheus metrics access

Service Account Configuration¶

# Create service account (default: true)
createServiceAccount: true

# Use custom service account name
customServiceAccountName: ""

# Service account settings
serviceAccount:
  imagePullSecrets: []
  annotations: {}

# Custom RBAC rules
customClusterRoleRules: []

For detailed information about the required Kubernetes permissions, see Kubernetes Permissions.

Resource Configuration¶

resources:
  requests:
    cpu: 100m
    memory: 1024Mi
  limits:
    cpu: 100m        # Optional CPU limit
    memory: 1024Mi

Holmes also enforces a separate per-subprocess virtual memory cap on every tool command it runs (via ulimit -v), controlled by the TOOL_MEMORY_LIMIT_MB environment variable. This cap operates inside the pod's memory limit, so the two must be coordinated — keep TOOL_MEMORY_LIMIT_MB comfortably below resources.limits.memory to leave headroom for Holmes itself. If you raise the pod limit to support larger tool outputs, raise TOOL_MEMORY_LIMIT_MB in lockstep. See Tool Execution Safety for the full mechanism and tuning guidance.

Toolset Configuration¶

Enable or disable specific toolsets:

toolsets:
  kubernetes/core:
    enabled: true      # Core Kubernetes functionality
  kubernetes/logs:
    enabled: true      # Kubernetes logs access
  robusta:
    enabled: true      # Robusta platform integration
  internet:
    enabled: true      # Internet access for documentation
  prometheus/metrics:
    enabled: true      # Prometheus metrics access

Advanced Configuration¶

Scheduling¶

# Node selection
# nodeSelector:
#   kubernetes.io/os: linux

# Pod affinity/anti-affinity
affinity: {}

# Tolerations
tolerations: []

# Priority class
priorityClassName: ""

Additional Configuration¶

# Additional environment variables
additionalEnvVars: []
additional_env_vars: []  # Legacy, use additionalEnvVars instead

# Image pull secrets
imagePullSecrets: []

# Additional volumes
additionalVolumes: []

# Additional volume mounts
additionalVolumeMounts: []

# OpenShift compatibility mode
openshift: false

# Account creation
enableAccountsCreate: true

# MCP servers configuration
mcp_servers: {}

# Model list configuration for multiple AI providers (UI only)
# See: https://holmesgpt.dev/ai-providers/using-multiple-providers/
modelList: {}

Example Configurations¶

Minimal Setup¶

# values.yaml
image: holmes:0.0.0
registry: robustadev
logLevel: INFO
enableTelemetry: false

resources:
  requests:
    cpu: 100m
    memory: 512Mi
  limits:
    memory: 512Mi

toolsets:
  kubernetes/core:
    enabled: true
  kubernetes/logs:
    enabled: true
  robusta:
    enabled: false
  internet:
    enabled: false
  prometheus/metrics:
    enabled: false

Multiple AI Providers Setup¶

# values.yaml
additionalEnvVars:
  - name: OPENAI_API_KEY
    valueFrom:
      secretKeyRef:
        name: holmes-secrets
        key: openai-api-key
  - name: ANTHROPIC_API_KEY
    valueFrom:
      secretKeyRef:
        name: holmes-secrets
        key: anthropic-api-key
  - name: AWS_ACCESS_KEY_ID
    valueFrom:
      secretKeyRef:
        name: holmes-secrets
        key: aws-access-key-id
  - name: AWS_SECRET_ACCESS_KEY
    valueFrom:
      secretKeyRef:
        name: holmes-secrets
        key: aws-secret-access-key

modelList:
  gpt-4.1:
    api_key: "{{ env.OPENAI_API_KEY }}"
    model: openai/gpt-4.1
    temperature: 0
  claude-sonnet-4-5:
    api_key: "{{ env.ANTHROPIC_API_KEY }}"
    model: anthropic/claude-sonnet-4-5-20250929
    temperature: 1
    thinking:
      budget_tokens: 10000
      type: enabled
  bedrock-sonnet-4-5:
    aws_access_key_id: "{{ env.AWS_ACCESS_KEY_ID }}"
    aws_region_name: us-east-1
    aws_secret_access_key: "{{ env.AWS_SECRET_ACCESS_KEY }}"
    model: bedrock/us.anthropic.claude-sonnet-4-5-20250929-v1:0
    temperature: 1
    thinking:
      budget_tokens: 10000
      type: enabled

The same setup using extraEnvVarsSecrets + the envRef: sugar (~40% fewer lines of YAML, and adding another key never requires a new additionalEnvVars entry):

# values.yaml
# Create the secret once:
#   kubectl create secret generic holmes-secrets \
#     --from-literal=OPENAI_API_KEY=sk-... \
#     --from-literal=ANTHROPIC_API_KEY=sk-ant-... \
#     --from-literal=AWS_ACCESS_KEY_ID=... \
#     --from-literal=AWS_SECRET_ACCESS_KEY=... \
#     -n <namespace>
extraEnvVarsSecrets:
  - holmes-secrets

modelList:
  gpt-4.1:
    model: openai/gpt-4.1
    api_key: envRef:OPENAI_API_KEY
    temperature: 0
  claude-sonnet-4-5:
    model: anthropic/claude-sonnet-4-5-20250929
    api_key: envRef:ANTHROPIC_API_KEY
    temperature: 1
    thinking:
      budget_tokens: 10000
      type: enabled
  bedrock-sonnet-4-5:
    model: bedrock/us.anthropic.claude-sonnet-4-5-20250929-v1:0
    aws_access_key_id: envRef:AWS_ACCESS_KEY_ID
    aws_secret_access_key: envRef:AWS_SECRET_ACCESS_KEY
    aws_region_name: us-east-1
    temperature: 1
    thinking:
      budget_tokens: 10000
      type: enabled

OpenShift Setup¶

# values.yaml
openshift: true
createServiceAccount: true

resources:
  requests:
    cpu: 100m
    memory: 1024Mi
  limits:
    memory: 1024Mi

toolsets:
  kubernetes/core:
    enabled: true
  kubernetes/logs:
    enabled: true

Configuration Validation¶

# Validate configuration
helm template holmesgpt robusta/holmes -f values.yaml

# Dry run installation
helm install holmesgpt robusta/holmes -f values.yaml --dry-run

# Check syntax
yamllint values.yaml

Complete Reference¶

For the complete and up-to-date configuration reference, see the actual values.yaml file in the repository.

Helm Configuration¶

Basic Configuration¶

Configuration Options¶

Essential Settings¶

API Key Configuration¶

Simplified API Key Configuration (extraEnvVarsSecrets + envRef: sugar)¶

Toolset Configuration¶

Service Account Configuration¶

Resource Configuration¶

Toolset Configuration¶

Advanced Configuration¶

Scheduling¶

Additional Configuration¶

Example Configurations¶

Minimal Setup¶

Multiple AI Providers Setup¶

OpenShift Setup¶

Configuration Validation¶

Complete Reference¶

Simplified API Key Configuration (`extraEnvVarsSecrets` + `envRef:` sugar)¶