@suluk/agents - v0.1.6
    Preparing search index...

    Interface AgentContextLoad

    interface AgentContextLoad {
        agent: string;
        instructionsTokens: number;
        instructionsMeasured: boolean;
        residentToolTokens: number;
        overheadTokens: number;
        totalTokens: number;
        coldTailTokens: number;
        tools: ToolContextCost[];
        subAgentCount: number;
        minWindowRequired: number;
        maxRounds?: number;
        thinkingBudget?: number;
        peakTokens: number;
        modelFit: ModelFit[];
        budget?: number;
        modelWindow?: number;
        target?: number;
        utilization?: number;
    }
    Index

    Properties

    agent instructionsTokens instructionsMeasured residentToolTokens overheadTokens totalTokens coldTailTokens tools subAgentCount minWindowRequired maxRounds? thinkingBudget? peakTokens modelFit budget? modelWindow? target? utilization?

    Properties

    agent: string
    instructionsTokens: number
    instructionsMeasured: boolean
    residentToolTokens: number
    overheadTokens: number
    totalTokens: number
    coldTailTokens: number
    tools: ToolContextCost[]
    subAgentCount: number
    minWindowRequired: number

    the minimum context window a model needs to run this agent (= the multi-round PEAK load).

    maxRounds?: number

    within-agent thinking cap (C029), if declared.

    thinkingBudget?: number
    peakTokens: number

    worst-case load accounting for thinking round-accretion (= totalTokens when no thinking). Fit checks use THIS.

    modelFit: ModelFit[]

    which DECLARED models are expected to work (window ≥ load) and which can't hold it.

    budget?: number
    modelWindow?: number

    the smallest declared model window (the binding window constraint), if any model is known.

    target?: number
    utilization?: number

    Settings

    Member Visibility

    On This Page

    Properties