rerank v1.1.3

reranker-ts

reranker-ts is a lightweight TypeScript library for re-ranking search results from retreival systems.

Installation

Install rerank using npm:

npm install rerank

Algorithms

LLM Re-Rankers

• Permutation Generation with sliding windows

Available Providers:

• ProviderGroq
• ProviderOpenAI

Example Usage:

import { LLMReranker, ProviderGroq } from "rerank";
const provider = new ProviderGroq("llama3-8b-8192", API_KEY);
const reranker = new LLMReranker(provider);

// Replace with your own list of objects to rerank.
const list = [
  { key: "bc8fe338", value: "I hate vegetables" },
  { key: "236386f2", value: "I love mangoes" },
];

const query = "I love apples";
const result = await reranker.rerank(list, "key", "value", query);
// ["236386f2", "bc8fe338"]

Reciprocal RankFusion

Combine multiple rank lists by assigning scores based on reciprocal ranks, effectively prioritizing higher-ranked items across all lists. Paper

Example Usage:

import { reciprocalRankFusion } from "rerank";

interface ImageSearchResult {
  url: string;
  width: string;
  height: string;
  score: number;
}

// assume you are searching 3 different queries
const rankedLists: ImageSearchResult[][] = await Promise.all([
  searchIndex("imageEmbeddingIndex", "person riding skateboard"),
  searchIndex("imageEmbeddingIndex", "person skating on the sidewalk"),
  searchIndex("imageIndex", "skateboard trick"),
]);

// perform reciprocal rank fusion on all of the results and specify key id, in this case "url"
// this will return a map of all urls now ranked with rrf score
const rankedURLs = reciprocalRankFusion(rankedLists, "url");

// build map of results keyed by url
const resultsMap = new Map<string, ImageSearchResult>();
rankedLists.flat().forEach((result) => {
  resultsMap.set(result.url, result);
});

// get single sorted results list from rrf
const sortedResults = Array.from(rankedUrls.keys()).map((url) => {
  return resultsMap.get(url);
});