ArkonisConfig

API: arkonis.dev/v1alpha1 Kind: ArkonisConfig Short name: aocfg

Analogous to a Kubernetes ConfigMap. Stores reusable prompt fragments and model settings that can be referenced from multiple ArkonisDeployment resources. Keeps configuration DRY — define a persona or output format once and share it across many agents.

Example

apiVersion: arkonis.dev/v1alpha1
kind: ArkonisConfig
metadata:
  name: analyst-base
  namespace: default
spec:
  temperature: "0.3"
  outputFormat: structured-json
  memoryBackend: in-context
  promptFragments:
    persona: "You are an expert analyst."
    outputRules: "Always cite sources. Flag uncertainty explicitly."

Spec fields

Field	Type	Required	Description
temperature	string	no	Sampling temperature passed to the Anthropic API. String-encoded float (e.g. "0.7").
outputFormat	string	no	Requested output format. structured-json instructs the agent to return JSON. markdown and plain are also accepted.
memoryBackend	string	no	Memory strategy for the agent. in-context — conversation history kept in the context window. vector-store and redis planned for v1beta1.
promptFragments	map[string]string	no	Named prompt fragments merged into the effective system prompt.

promptFragments

A key-value map of prompt fragment names to text values. When an ArkonisDeployment references this config via spec.configRef, the operator appends each fragment to the agent’s system prompt in definition order.

Reserved keys with defined merge behavior:

Key	Merge position
persona	Prepended to the system prompt.
outputRules	Appended after the system prompt.

Other keys are appended in alphabetical order.

Referencing ArkonisConfig from ArkonisDeployment

apiVersion: arkonis.dev/v1alpha1
kind: ArkonisDeployment
metadata:
  name: analyst-agent
spec:
  model: claude-sonnet-4-20250514
  systemPrompt: |
    Analyze the provided dataset and identify key trends.
  configRef: analyst-base   # references the ArkonisConfig above

The operator resolves the configRef during reconciliation and builds the effective system prompt by merging promptFragments. The temperature and outputFormat values are injected as environment variables into agent pods alongside the standard operator-managed variables.

ArkonisConfig and the referencing ArkonisDeployment must be in the same namespace.