@axflow/models/dist/ollama/generation.d.mts

Zero-dependency, modular SDK for building robust natural language applications

type OllamaModelOptions = {
    top_k?: number;
    top_p?: number;
    stop?: string;
    temperature?: number;
    repeat_penalty?: number;
    repeat_last_n?: number;
    num_threads?: number;
    num_gpu?: number;
    num_gqa?: number;
    num_ctx?: number;
    mirostat?: number;
    mirostat_eta?: number;
    mirostat_tau?: number;
    tfs_z?: number;
};

declare namespace OllamaGenerationTypes {
    type Request = {
        model: string;
        prompt: string;
        system?: string;
        template?: string;
        context?: string;
        options?: OllamaModelOptions;
    };
    type RequestOptions = {
        apiUrl?: string;
        fetch?: typeof fetch;
        headers?: Record<string, string>;
        signal?: AbortSignal;
    };
    type Chunk = {
        model: string;
        created_at: string;
        response: string;
        done: boolean;
        context?: Array<number>;
        total_duration?: number;
        load_duration?: number;
        prompt_eval_count?: number;
        prompt_eval_duration?: number;
        eval_count?: number;
        eval_duration?: number;
    };
}
/**
 * Stream a generation against an ollama serving endpoint. Return a stream of bytes.
 * Docs: https://github.com/jmorganca/ollama/blob/main/docs/api.md
 *
 * @param request The request body containing the model, prompt, and options.
 * @param options
 * @param options.apiurl The ollama serving url. defaults to http://127.0.0.1:11343.
 * @param options.fetch The fetch implementation to use. defaults to globalthis.fetch.
 * @param options.headers Optionally add additional http headers to the request.
 * @param options.signal An abortsignal that can be used to abort the fetch request.
 * @returns A stream of bytes directly from the API.
 */
declare function streamBytes(request: OllamaGenerationTypes.Request, options: OllamaGenerationTypes.RequestOptions): Promise<ReadableStream<Uint8Array>>;
/**
 * Stream a generation against an ollama serving endpoint, return javascript objects.
 *
 * Example chunk:
 *   {
 *     token: { id: 11, text: ' and', logprob: -0.00002193451, special: false },
 *     generated_text: null,
 *     details: null
 *   }
 *
 * @param request The request body containing the model, prompt, and options.
 * @param options
 * @param options.apiurl The ollama serving url. defaults to http://127.0.0.1:11343.
 * @param options.fetch The fetch implementation to use. defaults to globalthis.fetch.
 * @param options.headers Optionally add additional http headers to the request.
 * @param options.signal An abortsignal that can be used to abort the fetch request.
 * @returns A stream of objects representing each chunk from the api.
 */
declare function stream(request: OllamaGenerationTypes.Request, options: OllamaGenerationTypes.RequestOptions): Promise<ReadableStream<OllamaGenerationTypes.Chunk>>;
/**
 * Stream a generation against an ollama serving endpoint, return only the text tokens.
 *
 * @param request The request body containing the model, prompt, and options.
 * @param options
 * @param options.apiurl The ollama serving url. defaults to http://127.0.0.1:11343.
 * @param options.fetch The fetch implementation to use. defaults to globalthis.fetch.
 * @param options.headers Optionally add additional http headers to the request.
 * @param options.signal An abortsignal that can be used to abort the fetch request.
 * @returns A stream of tokens from the API.
 */
declare function streamTokens(request: OllamaGenerationTypes.Request, options: OllamaGenerationTypes.RequestOptions): Promise<ReadableStream<string>>;
/**
 * An object that encapsulates methods for calling the HF inference API.
 */
declare class OllamaGeneration {
    static streamBytes: typeof streamBytes;
    static stream: typeof stream;
    static streamTokens: typeof streamTokens;
}

export { OllamaGeneration, OllamaGenerationTypes };