node-llama-cpp
Version:
Run AI models locally on your machine with node.js bindings for llama.cpp. Enforce a JSON schema on the model output on the generation level
21 lines (20 loc) • 993 B
TypeScript
import { ChatHistoryItem, Tokenizer } from "../types.js";
import { ChatWrapper } from "../ChatWrapper.js";
export declare function findCharacterRemovalCountToFitChatHistoryInContext({ compressChatHistory, chatHistory, tokensCountToFit, tokenizer, chatWrapper, initialCharactersRemovalCount, estimatedCharactersPerToken, maxDecompressionAttempts, failedCompressionErrorMessage }: {
compressChatHistory(options: {
chatHistory: readonly ChatHistoryItem[];
charactersToRemove: number;
estimatedCharactersPerToken: number;
}): ChatHistoryItem[] | Promise<ChatHistoryItem[]>;
chatHistory: ChatHistoryItem[];
tokensCountToFit: number;
tokenizer: Tokenizer;
chatWrapper: ChatWrapper;
initialCharactersRemovalCount?: number;
estimatedCharactersPerToken?: number;
maxDecompressionAttempts?: number;
failedCompressionErrorMessage?: string;
}): Promise<{
removedCharactersCount: number;
compressedChatHistory: ChatHistoryItem[];
}>;