@gptp/core
Version:
Library to supercharge your use of large language models
39 lines (38 loc) • 1.14 kB
TypeScript
import { string_model_name } from '.././types/typeAliases';
/**
* Prompt result is the simplest concept of execution.
* It is the result of executing one prompt _(NOT a template)_.
*
* @see https://github.com/webgptorg/ptp#prompt-result
*/
export type PromptResult = PromptCompletionResult | PromptChatResult;
/**
* Prompt completion result
* It contains only the following text NOT the whole completion
*/
export type PromptCompletionResult = PromptCommonResult;
/**
* Prompt chat result
*/
export interface PromptChatResult extends PromptCommonResult {
}
export interface PromptCommonResult {
/**
* Exact text response from the model
*/
readonly content: string;
/**
* Name of the model used to generate the response
*/
readonly model: string_model_name;
/**
* Raw response from the model
*/
readonly rawResponse: object;
}
/**
* TODO: [🧠] Should here be link to the prompt?
* TODO: [🧠] Maybe type raw properly - not onject but OpenAI.result.whatever
* TODO: [🧠] Maybe remove redundant raw.choices.text
* TODO: Log raw even if prompt failed - log the raw error
*/