@langchain/core
Version:
Core LangChain.js abstractions and schemas
62 lines (61 loc) • 1.83 kB
TypeScript
import { type BaseMessage, type BaseMessageChunk } from "./messages/base.js";
export declare const RUN_KEY = "__run";
/**
* Output of a single generation.
*/
export interface Generation {
/**
* Generated text output
*/
text: string;
/**
* Raw generation info response from the provider.
* May include things like reason for finishing (e.g. in {@link OpenAI})
*/
generationInfo?: Record<string, any>;
}
export type GenerationChunkFields = {
text: string;
generationInfo?: Record<string, any>;
};
/**
* Chunk of a single generation. Used for streaming.
*/
export declare class GenerationChunk implements Generation {
text: string;
generationInfo?: Record<string, any>;
constructor(fields: GenerationChunkFields);
concat(chunk: GenerationChunk): GenerationChunk;
}
/**
* Contains all relevant information returned by an LLM.
*/
export type LLMResult = {
/**
* List of the things generated. Each input could have multiple {@link Generation | generations}, hence this is a list of lists.
*/
generations: Generation[][];
/**
* Dictionary of arbitrary LLM-provider specific output.
*/
llmOutput?: Record<string, any>;
/**
* Dictionary of run metadata
*/
[RUN_KEY]?: Record<string, any>;
};
export interface ChatGeneration extends Generation {
message: BaseMessage;
}
export type ChatGenerationChunkFields = GenerationChunkFields & {
message: BaseMessageChunk;
};
export declare class ChatGenerationChunk extends GenerationChunk implements ChatGeneration {
message: BaseMessageChunk;
constructor(fields: ChatGenerationChunkFields);
concat(chunk: ChatGenerationChunk): ChatGenerationChunk;
}
export interface ChatResult {
generations: ChatGeneration[];
llmOutput?: Record<string, any>;
}