AgentNodeConfig

Settings that control Agent Node. Required fields are instructions, model_provider, model.

instructionsInstructions (string)required

Agent system instructions

prompts object[]

Reusable Prompt references. Rendered in list order and joined with '\n\n' as the system prefix; inline instructions is appended as a suffix. Empty = inline instructions only.

Array [

prompt_idPrompt Id (string)required

variables object

{{var}} bindings for the referenced Prompt

anyOf

• object
• null

object

]

model_providerModel Provider (string)required

LLM provider name

modelModel (string)required

LLM model name

tools object[]

Available tools

Array [

anyOf

• string
• AgentResourceRef
• AgentToolConfig
• object

string

]

actors object[]

Array [

anyOf

• string
• AgentResourceRef
• AgentActorConfig
• object

string

]

roles object[]

Array [

anyOf

• string
• AgentResourceRef
• object

string

]

outputs object

anyOf

• object
• null

object

max_iterationsMax Iterations (integer)

Maximum iteration count

Default value: 50

max_input_tokens object

Maximum input-token budget for the system prompt and conversation history before each model call. The effective limit cannot exceed the selected model's context_window; null uses the D.Hub default.

anyOf

• integer
• null

integer

Possible values: >= 1

compaction object

Optional per-agent conversation compaction settings. Omit this field to use D.Hub defaults.

anyOf

• AgentCompaction
• null

enabledEnabled (boolean)

Whether conversation compaction is enabled for the agent.

Default value: true

trigger_ratioTrigger Ratio (number)

Fraction of the input-token budget that triggers compaction.

Possible values: > 0 and < 1

Default value: 0.8

target_ratioTarget Ratio (number)

Fraction of the input-token budget targeted after compaction. It must be lower than trigger_ratio.

Possible values: > 0 and < 1

Default value: 0.5

summary_max_tokensSummary Max Tokens (integer)

Maximum number of tokens allowed in the generated compaction summary. Minimum: 128. Maximum: 8192.

Possible values: >= 128 and <= 8192

Default value: 1024

options object

anyOf

• object
• null

object

reasoning_effort object

Default reasoning effort for this agent. A value supplied with an individual request takes precedence; null uses the selected model's default.

anyOf

• string
• null

string

Possible values: [low, medium, high, xhigh]

AgentNodeConfig

{
  "instructions": "string",
  "prompts": [
    {
      "prompt_id": "string",
      "variables": {}
    }
  ],
  "model_provider": "string",
  "model": "string",
  "tools": [
    "string",
    {
      "id": "string",
      "name": "string"
    },
    {
      "name": "string",
      "description": "",
      "type": "python",
      "function": {
        "source": "string",
        "inputs": {},
        "outputs": {},
        "dependencies": [
          "string"
        ],
        "decorators": [
          {
            "decorator": "string",
            "name": "string",
            "arguments": {}
          }
        ]
      },
      "require_browser": false,
      "options": {}
    },
    {}
  ],
  "actors": [
    "string",
    {
      "id": "string",
      "name": "string"
    },
    {
      "name": "string",
      "description": "",
      "function": {
        "source": "string",
        "inputs": {},
        "outputs": {},
        "dependencies": [
          "string"
        ],
        "decorators": [
          {
            "decorator": "string",
            "name": "string",
            "arguments": {}
          }
        ]
      },
      "confirmation_policy": "confirm",
      "options": {}
    },
    {}
  ],
  "roles": [
    "string",
    {
      "id": "string",
      "name": "string"
    },
    {}
  ],
  "outputs": {},
  "max_iterations": 50,
  "max_input_tokens": 0,
  "compaction": {
    "enabled": true,
    "trigger_ratio": 0.8,
    "target_ratio": 0.5,
    "summary_max_tokens": 1024
  },
  "options": {},
  "reasoning_effort": "low"
}