ART Framework API Docs
    Preparing search index...

    Interface PromptContext

    Represents the contextual data gathered by Agent Logic (e.g., PESAgent) to be injected into a Mustache blueprint/template by the PromptManager.assemblePrompt method.

    Contains standard fields commonly needed for prompts, plus allows for arbitrary additional properties required by specific agent blueprints. Agent logic is responsible for populating this context appropriately before calling assemblePrompt.

    PromptContext

    interface PromptContext {
        availableTools?: (ToolSchema & { inputSchemaJson?: string })[];
        history?: { content: string; role: string; [key: string]: any }[];
        query?: string;
        systemPrompt?: string;
        toolResults?: (ToolResult & { outputJson?: string })[];
        [key: string]: any;
    }

    Indexable

    • [key: string]: any

      Allows agent patterns (like PES) to pass any other custom data needed by their specific blueprints (e.g., intent, plan).

    Index

    Properties

    availableTools?: (ToolSchema & { inputSchemaJson?: string })[]

    The schemas of the tools available for use, potentially pre-formatted for the blueprint (e.g., with inputSchemaJson pre-stringified).

    history?: { content: string; role: string; [key: string]: any }[]

    The conversation history, typically formatted as an array suitable for the blueprint (e.g., array of objects with role and content). Agent logic should pre-format this.

    While ArtStandardPrompt could be used, simpler structures might be preferred for blueprints.

    query?: string

    The user's current query or input relevant to this prompt generation step.

    systemPrompt?: string

    The system prompt string to be used (resolved by agent logic from config or defaults).

    toolResults?: (ToolResult & { outputJson?: string })[]

    The results from any tools executed in a previous step, potentially pre-formatted for the blueprint (e.g., with outputJson pre-stringified).