output
    Preparing search index...

    Type Alias Prompt

    Configuration for LLM prompt generation.

    const prompt: Prompt = {
      name: 'summarizePrompt',
      fileDir: '/app/prompts',
      variables: {},
      config: {
        provider: 'anthropic',
        model: 'claude-opus-4-1',
        temperature: 0.7,
        maxOutputTokens: 2048,
        maxSteps: 10,
        skills: []
      },
      messages: [{ role: 'user', content: 'Summarize this document.' }],
      instructions: null
    };
    type Prompt = {
        config: {
            aspectRatio?: `${number}:${number}`;
            frequencyPenalty?: number;
            maxImagesPerCall?: number;
            maxOutputTokens?: number;
            maxSteps: number;
            maxTokens?: number;
            messageOptions?: Record<string, Record<string, Record<string, unknown>>>;
            model: string;
            n?: number;
            presencePenalty?: number;
            provider: string;
            providerOptions?: Record<string, unknown>;
            seed?: number;
            size?: `${number}x${number}`;
            skills: string[];
            stopSequences?: string[];
            temperature?: number;
            tools?: Record<string, Record<string, unknown>>;
            topK?: number;
            topP?: number;
        };
        fileDir: string;
        instructions: string
        | null;
        messages: PromptMessage[];
        name: string;
        variables: PromptVariables;
    }
    Index

    Properties

    config fileDir instructions messages name variables

    Properties

    config: {
        aspectRatio?: `${number}:${number}`;
        frequencyPenalty?: number;
        maxImagesPerCall?: number;
        maxOutputTokens?: number;
        maxSteps: number;
        maxTokens?: number;
        messageOptions?: Record<string, Record<string, Record<string, unknown>>>;
        model: string;
        n?: number;
        presencePenalty?: number;
        provider: string;
        providerOptions?: Record<string, unknown>;
        seed?: number;
        size?: `${number}x${number}`;
        skills: string[];
        stopSequences?: string[];
        temperature?: number;
        tools?: Record<string, Record<string, unknown>>;
        topK?: number;
        topP?: number;
    }

    General configuration for the LLM

    Type Declaration

    • OptionalaspectRatio?: `${number}:${number}`

      Image aspect ratio, for example 16:9

    • OptionalfrequencyPenalty?: number

      Penalizes tokens according to how often they appear.

    • OptionalmaxImagesPerCall?: number

      Maximum images to request per provider call

    • OptionalmaxOutputTokens?: number

      Maximum number of tokens in the response.

    • maxSteps: number

      Tool-loop iterations when stopWhen is omitted. Always a positive integer after load (default 10).

    • OptionalmaxTokens?: number

      Maximum number of tokens in the response.

      Use maxOutputTokens.

    • OptionalmessageOptions?: Record<string, Record<string, Record<string, unknown>>>

      Named, reusable per-message providerOptions sets, referenced from message blocks via the options="<name>" attribute. Each value is a provider-namespaced options object, e.g. { anthropic: { cacheControl: { type: 'ephemeral' } } }.

    • model: string

      Model name/identifier

    • Optionaln?: number

      Number of images to generate

    • OptionalpresencePenalty?: number

      Penalizes tokens that already appear in the prompt or generated output.

    • provider: string

      LLM provider.

      Built-in: 'anthropic', 'openai', 'azure', 'amazon-bedrock', 'google-vertex', 'perplexity'. Legacy aliases 'bedrock' and 'vertex' are deprecated but still accepted in prompts and normalized to their canonical names with a warning. Custom providers registered via registerProvider are also accepted.

    • OptionalproviderOptions?: Record<string, unknown>

      Provider-specific options

    • Optionalseed?: number

      Random seed for deterministic text or image generation when supported

    • Optionalsize?: `${number}x${number}`

      Image size, for example 1024x1024

    • skills: string[]

      Skill file or directory paths relative to the prompt file. Always a string[] after load ([] if unset).

    • OptionalstopSequences?: string[]

      Sequences that stop text generation when produced.

    • Optionaltemperature?: number

      Generation temperature. Supported range and behavior vary by provider.

    • Optionaltools?: Record<string, Record<string, unknown>>

      Provider-specific tools with configuration.

      tools:
        googleSearch:
          mode: MODE_DYNAMIC
          dynamicThreshold: 0.8

      tools:
        webSearch:
          searchContextSize: high
          filters:
            allowedDomains: [wikipedia.org]
    • OptionaltopK?: number

      Limits sampling to the top K token choices.

    • OptionaltopP?: number

      Nucleus sampling value. Usually set instead of temperature.

    fileDir: string

    Base directory for resolving relative prompt resources such as skills.

    instructions: string | null

    The whole trimmed prompt body when its first meaningful token is plain text. Always set after loadPrompt: a non-empty string with messages: [] for instruction mode, or null when the body starts with markup and is parsed into messages.

    messages: PromptMessage[]

    Array of messages in the conversation

    name: string

    Prompt name used in tracing and errors.

    variables: PromptVariables

    Interpolation values associated with the prompt. Defaults to {} when validated.

    Settings

    Member Visibility

    On This Page

    Properties