How to Do RAG with Mastra and ZeroEntropy

Why This Stack?

Mastra provides a complete RAG framework with vector store abstractions, metadata filtering, and flexible retrieval patterns. ZeroEntropy offers purpose-built reranking models that outperform LLM-based approaches while being 10x cheaper and faster.

This combination gives you:

• Sub-second retrieval even with large document collections
• Better relevance than basic vector similarity
• Production-ready cost structure ($2-5 per 1k queries vs $20-100 for LLM reranking)
• Clean abstractions that let you swap vector stores without rewriting code

Prerequisites

npm install @mastra/core @mastra/rag @mastra/pg @ai-sdk/openai ai

You’ll need:

• A PostgreSQL database with pgvector extension
• OpenAI API key for embeddings
• ZeroEntropy API access (sign up at dashboard.zeroentropy.dev)

Environment setup:

POSTGRES_CONNECTION_STRING=postgresql://user:pass@localhost:5432/rag_db
OPENAI_API_KEY=sk-...
ZEROENTROPY_API_KEY=ze-...  # Sign up at zeroentropy.ai

Step 1: Create the ZeroEntropy Reranker

First, implement the RelevanceScoreProvider interface for ZeroEntropy:

import type { RelevanceScoreProvider } from '@mastra/core/relevance';
import ZeroEntropy from 'zeroentropy';

export class ZeroEntropyRelevanceScorer implements RelevanceScoreProvider {
  private client: ZeroEntropy;
  private model: string;

  constructor(model?: string, apiKey?: string) {
    this.client = new ZeroEntropy({
      apiKey: apiKey || process.env.ZEROENTROPY_API_KEY || '',
    });
    this.model = model || 'zerank-1';
  }

  async getRelevanceScore(query: string, text: string): Promise<number> {
    const response = await this.client.models.rerank({
      query,
      documents: [text],
      model: this.model,
      top_n: 1,
    });

    return response.results[0]?.relevance_score ?? 0;
  }
}

How this works:

Step 2: Complete Working Example

Here’s the full implementation replacing GPT-4o-mini with ZeroEntropy:

import { openai } from '@ai-sdk/openai';
import { Mastra } from '@mastra/core/mastra';
import { Agent } from '@mastra/core/agent';
import { PgVector } from '@mastra/pg';
import { MDocument, createVectorQueryTool } from '@mastra/rag';
import { embedMany } from 'ai';
import { ZeroEntropyRelevanceScorer } from './zeroentropy-scorer';

// Create vector query tool with ZeroEntropy reranker
const vectorQueryTool = createVectorQueryTool({
  vectorStoreName: 'pgVector',
  indexName: 'embeddings',
  model: openai.embedding('text-embedding-3-small'),
  reranker: {
    provider: new ZeroEntropyRelevanceScorer('zerank-1'),
  },
});

// Create RAG agent
export const ragAgent = new Agent({
  id: 'rag-agent',
  name: 'RAG Agent',
  instructions: `You are a helpful assistant that answers questions based on the provided context. Keep your answers concise and relevant.
    Important: When asked to answer a question, please base your answer only on the context provided in the tool.
    If the context doesn't contain enough information to fully answer the question, please state that explicitly.`,
  model: openai('gpt-4o-mini'),
  tools: {
    vectorQueryTool,
  },
});

// Initialize Mastra with PgVector
const pgVector = new PgVector({
  connectionString: process.env.POSTGRES_CONNECTION_STRING!
});

Key Takeaways