Interface PlannerConfig

Configuration for the MissionPlanner.

interface PlannerConfig {
    branchCount: number;
    autonomy: AutonomyMode;
    providerStrategy: ProviderStrategyConfig;
    thresholds: GuardrailThresholds;
    costCap: number;
    maxAgents: number;
    maxToolForges: number;
    maxExpansions: number;
    maxDepth: number;
    reevalInterval: number;
    llmCaller: ((system, user) => Promise<string>);
    plannerLlmCaller?: ((system, user) => Promise<string>);
    plannerModel?: string;
    executionModel?: string;
}

Properties

branchCount autonomy providerStrategy thresholds costCap maxAgents maxToolForges maxExpansions maxDepth reevalInterval llmCaller plannerLlmCaller? plannerModel? executionModel?

Properties

branchCount

branchCount: number

Number of Tree of Thought branches to explore.

autonomy

autonomy: AutonomyMode

providerStrategy

providerStrategy: ProviderStrategyConfig

thresholds

thresholds: GuardrailThresholds

costCap

costCap: number

maxAgents

maxAgents: number

maxToolForges

maxToolForges: number

maxExpansions

maxExpansions: number

maxDepth

maxDepth: number

reevalInterval

reevalInterval: number

Re-evaluate graph every N completed nodes.

llmCaller

llmCaller: ((system, user) => Promise<string>)

LLM caller used by agent nodes during execution. Falls back to plannerLlmCaller if not provided separately.

Type declaration

    • (system, user): Promise<string>

    • Parameters

      • system: string
      • user: string

      Returns Promise<string>

Optional plannerLlmCaller

plannerLlmCaller?: ((system, user) => Promise<string>)

LLM caller used specifically for the Tree of Thought planning phases (decomposition, evaluation, refinement, expansion).

Use a stronger model here (e.g., Opus 4.6) for better plan quality, while execution agents can use a different model (e.g., GPT-5.4).

Defaults to llmCaller if not provided.

Type declaration

    • (system, user): Promise<string>

    • Parameters

      • system: string
      • user: string

      Returns Promise<string>

Optional plannerModel

plannerModel?: string

Human-readable label for the planner model (for logging/events).

Optional executionModel

executionModel?: string

Human-readable label for the default execution model (for logging/events).

Settings

Member Visibility

Theme

On This Page

branchCountautonomyproviderStrategythresholdscostCapmaxAgentsmaxToolForgesmaxExpansionsmaxDepthreevalIntervalllmCallerplannerLlmCallerplannerModelexecutionModel