together-ai-sdk/lib/core/client.types.d.ts

A typescript SDK for the Together AI API

import { TogetherChatModel, TogetherCodeModel, TogetherEmbeddingModel, TogetherImageModel, TogetherLanguageModel, TogetherModerationModel } from './models.types';
export interface ChatMessage {
    /**
     * The message content.
     */
    content: string;
    /**
     * The message author.
     */
    role: 'user' | 'system' | 'assistant';
}
export interface TokenUsage {
    /**
     * The number of tokens used for the prompt.
     */
    promptTokens: number;
    /**
     * The number of tokens used for the completion.
     */
    completionTokens: number;
    /**
     * The total number of tokens used.
     */
    totalTokens: number;
}
export interface TogetherChatParams {
    /**
     * The name of the model to query.
     */
    model: TogetherChatModel;
    /**
     * A list of messages comprising the conversation so far.
     */
    messages: ChatMessage[];
    /**
     * The maximum number of tokens to generate.
     * @default 512
     */
    maxTokens?: number;
    /**
     * A list of string sequences that will truncate (stop) inference text output.
     * For example, "" will stop generation as soon as the model generates the given token.
     * @default []
     */
    stop?: string[];
    /**
     * A decimal number that determines the degree of randomness in the response.
     * A value of 1 will always yield the same output.
     * A temperature less than 1 favors more correctness and is appropriate for question answering or summarization.
     * A value greater than 1 introduces more randomness in the output.
     * @default 0.7
     */
    temperature?: number;
    /**
     * The nucleus parameter is used to dynamically adjust the number of choices for each predicted token based on the cumulative probabilities.
     * It specifies a probability threshold, below which all less likely tokens are filtered out.
     * This technique helps to maintain diversity and generate more fluent and natural-sounding text.
     * @default 0.7
     */
    topP?: number;
    /**
     * Used to limit the number of choices for the next predicted word or token.
     * It specifies the maximum number of tokens to consider at each step, based on their probability of occurrence.
     * This technique helps to speed up the generation process
     * and can improve the quality of the generated text by focusing on the most likely options.
     * @default 50
     */
    topK?: number;
    /**
     * A number that controls the diversity of generated text by reducing the likelihood of repeated sequences.
     * Higher values decrease repetition.
     * @default 1
     */
    repetitionPenalty?: number;
    /**
     * How many completions to generate for each prompt.
     * @default 1
     */
    n?: number;
    /**
     * If undefined, the function will hold until the entire response is generated until returning.
     * If not undefined, the response will be streamed to this function as it is received.
     * The callback is async so that it can be awaited for processing each event in order.
     * If you have no need for order, then you can simply not return a promise.
     * @param partialCompletion - a partial completion of the response, or 'done' when the response is complete.
     */
    streamCallback?: (partialCompletion: 'done' | TogetherChatStreamedResponse) => Promise<void> | void;
}
export interface TogetherChatResponse {
    /**
     * A unique identifier for the response.
     */
    id: string;
    /**
     * The list of completions for the prompt.
     * The number of completions is determined by the `n` parameter.
     */
    choices: Array<{
        /**
         * Why the response was finished.
         */
        finishReason: string;
        /**
         * The index of the choice (corresponding to the `n` parameter)
         */
        index: number;
        logprobs: null;
        /**
         * The message completed from the LLM.
         */
        message: ChatMessage;
    }>;
    /**
     * The number of tokens used for the prompt & completion.
     */
    usage: TokenUsage;
    /**
     * The timestamp of when the response was created.
     */
    created: number;
    /**
     * Which model was used to generate the response.
     */
    model: TogetherChatModel;
    /**
     * Which type of object this is.
     */
    object: 'chat.completion';
}
export interface TogetherChatStreamedResponse {
    /**
     * A unique identifier for the response.
     */
    id: string;
    /**
     * The timestamp of when the response was created.
     */
    created: number;
    /**
     * Which type of object this is.
     */
    object: 'chat.completion.chunk';
    /**
     * The partial completion of the response for each choice.
     */
    choices: Array<{
        /**
         * The index of the choice.
         */
        index: number;
        /**
         * The partial completion of the response.
         */
        delta: {
            content: string;
        };
    }>;
    /**
     * Information about which token was generated.
     */
    token: {
        /**
         * The unique identifier for the token.
         */
        id: number;
        /**
         * The text version of the token.
         */
        text: string;
        /**
         * The log probability of the token.
         */
        logprob: number;
        /**
         * If the token is special.
         */
        special: boolean;
    };
    /**
     * The reason for finishing the response.
     * Undefined if the response is not finished.
     */
    finishReason?: string;
    /**
     * The complete generated text.
     * Undefined if the response is not finished.
     */
    generatedText?: string;
    stats: null;
    /**
     * The usage information to be provided in the final event.
     * Undefined if the response is not finished.
     */
    usage?: TokenUsage;
}
export type TogetherLanguageParams = Omit<TogetherChatParams, 'messages' | 'model' | 'streamCallback'> & {
    /**
     * The name of the model to query.
     */
    model: TogetherLanguageModel;
    /**
     * A string providing context for the model to complete.
     */
    prompt: string;
    /**
     * If undefined, the function will hold until the entire response is generated until returning.
     * If not undefined, the response will be streamed to this function as it is received.
     * The callback is async so that it can be awaited for processing each event in order.
     * If you have no need for order, then you can simply not return a promise.
     * @param partialCompletion - a partial completion of the response, or 'done' when the response is complete.
     */
    streamCallback?: (partialCompletion: 'done' | TogetherLanguageStreamedResponse) => Promise<void> | void;
};
export type TogetherLanguageResponse = Omit<TogetherChatResponse, 'choices' | 'object' | 'model'> & {
    /**
     * The name of the model to query.
     */
    model: TogetherLanguageModel;
    /**
     * Which type of object this is.
     */
    object: 'language.completion';
    /**
     * The list of completions for the prompt.
     * The number of completions is determined by the `n` parameter.
     */
    choices: Array<{
        /**
         * The completion of the prompt
         */
        text: string;
    }>;
};
export type TogetherLanguageStreamedResponse = Omit<TogetherChatStreamedResponse, 'stats' | 'choices' | 'object' | 'created'> & {
    /**
     * The choices coming back from the model.
     */
    choices: Array<{
        text: string;
    }>;
};
export type TogetherCodeParams = Omit<TogetherLanguageParams, 'model'> & {
    /**
     * The name of the model to query.
     */
    model: TogetherCodeModel;
};
export type TogetherCodeResponse = TogetherLanguageResponse;
export interface TogetherEmbeddingParams {
    /**
     * The name of the model to query.
     */
    model: TogetherEmbeddingModel;
    /**
     * The string to embed.
     */
    input: string;
}
export interface TogetherEmbeddingResponse {
    test: string;
}
export interface TogetherImageParams {
    /**
     * The name of the model to query.
     */
    model: TogetherImageModel;
    /**
     * The prompt for the image generator
     */
    prompt: string;
    /**
     * The prompt for what not to include in the image
     * @default nothing
     */
    negativePrompt?: string;
    /**
     * Width of the image to be generated
     * @default 1024
     */
    width?: number;
    /**
     * Height of the image to be generated
     * @default 1024
     */
    height?: number;
    /**
     * Number of steps to perform iterations on
     * @default 20
     */
    steps?: number;
    /**
     * A seed to produce deterministic results
     * @default randomly generated
     */
    seed?: number;
    /**
     * Number of images to generate
     * @default 1
     */
    n?: number;
}
export type TogetherImageResponse = Pick<TogetherInferenceResponse, 'output'>;
export interface TogetherInferenceParams {
    model: TogetherChatModel | TogetherLanguageModel | TogetherCodeModel | TogetherImageModel | TogetherModerationModel | TogetherEmbeddingModel;
    prompt?: string;
    negativePrompt?: string;
    n?: number;
    requestType: string;
    messages?: ChatMessage[];
    maxTokens?: number;
    promptFormatString?: string;
    repetitionPenalty?: number;
    stop?: string[];
    temperature?: number;
    topP?: number;
    topK?: number;
    seed?: number;
    width?: number;
    height?: number;
    steps?: number;
}
export interface TogetherInferenceResponse {
    args: TogetherInferenceParams;
    model: TogetherInferenceParams['model'];
    modelOwner: string;
    numReturns: number;
    output: {
        requestId: string;
        resultType: string;
        choices: Array<{
            imageBase64: string;
        }>;
    };
    prompt: string[];
    status: string;
    subjobs: unknown[];
    tags: {};
}