Reranking Module

Reranking is a crucial step in the recall pipeline. After retrieving a set of potentially relevant documents (e.g. via keyword search or basic vector search), a reranker uses a more powerful semantic model to re-order those documents by their true relevance to the query.

Core Capabilities

The core reranking capabilities, contract interfaces, and deterministic fallback reranker are built directly into @tekbreed/tekmemo.

Import

import { createDeterministicFallbackReranker, createVoyageReranker } from "@tekbreed/tekmemo";

Quick start with Tekmemo

The Tekmemo class exposes reranking utilities through memo.rerank:

import { Tekmemo } from "@tekbreed/tekmemo";

const memo = new Tekmemo({ rootDir: "./.tekmemo", projectId: "my-app" });

// Sort results by rerank score
const sorted = memo.rerank.sort(results);

// Trim to top K
const top3 = memo.rerank.applyTopK(sorted, 3);

// Get a deterministic fallback reranker
const fallback = memo.rerank.createFallback();

API Reference

Tekmemo.rerank

Method	Purpose
memo.rerank.sort(results, key?)	Sorts rerank results by score.
memo.rerank.applyTopK(results, topK)	Trims results to top K.
memo.rerank.createFallback()	Creates a deterministic fallback reranker (keyword-based).

Reranker contract

Method	Purpose
reranker.rerank(input)	Scores and sorts documents by relevance.
createDeterministicFallbackReranker()	Creates a local reranker that sorts based on keyword presence.

VoyageAI Integration

For production reranking with VoyageAI's specialized models:

import { Tekmemo } from "@tekbreed/tekmemo";
import { createVoyageReranker } from "@tekbreed/tekmemo-adapter-voyage";

const reranker = createVoyageReranker({
  apiKey: process.env.VOYAGE_API_KEY,
  model: "voyage-rerank-2",
});

const results = await reranker.rerank({
  query: "How do I handle sync conflicts?",
  documents: [
    "Conflict resolution policy: keep-cloud...",
    "Sync push sends local events to the cloud...",
    "Memory records should be small and explicit.",
  ],
  topK: 1,
});

console.log(`Top match: ${results[0].document}`);

Use Cases

• Quality Improvement: Use a cheaper, faster search for the first pass, then rerank to ensure top results are the most relevant.
• Context Window Optimization: Ensure only the absolute best information is injected into an agent's context window.
• Hybrid Recall: Combine local keyword results and cloud vector results, then rerank the union for a unified set of context hits.