Integration Migration April 13, 2026

Migrate from Weaviate to REM Labs

Weaviate gives you vector search with GraphQL. REM Labs gives you AI memory -- vector search plus full-text retrieval, entity graphs, temporal decay, and neural reranking. This guide covers exporting your Weaviate collections and importing them into REM Labs.

What You Gain

No schema management -- REM handles indexing automatically; no class definitions needed
Multi-signal fusion -- vector + FTS5 + entity graph retrieval in a single query
Memory semantics -- temporal decay, consolidation, namespaces, and Dream Engine
Simpler API -- remember() and recall() instead of GraphQL mutations

Step 1: Export from Weaviate

Use the Weaviate client to export all objects from a collection:

import weaviate from "weaviate-client";
import fs from "fs";

const client = await weaviate.connectToLocal();

const collection = client.collections.get("Document");
const exported = [];

for await (const item of collection.iterator()) {
  exported.push({
    id: item.uuid,
    properties: item.properties,
    text: item.properties.content || item.properties.text || ""
  });
}

fs.writeFileSync("weaviate-export.json", JSON.stringify(exported, null, 2));
console.log(`Exported ${exported.length} objects`);

Step 2: Import into REM Labs

import { RemClient } from "@remlabs/sdk";
import fs from "fs";

const rem = new RemClient({ apiKey: process.env.REMLABS_API_KEY });
const data = JSON.parse(fs.readFileSync("weaviate-export.json", "utf-8"));

let count = 0;
for (const obj of data) {
  if (!obj.text) continue;

  await rem.remember({
    content: obj.text,
    namespace: "weaviate-import",
    metadata: {
      weaviate_id: obj.id,
      ...obj.properties
    },
    tags: obj.properties.category ? [obj.properties.category] : []
  });

  count++;
  if (count % 50 === 0) console.log(`${count}/${data.length}`);
}

console.log(`Imported ${count} memories`);

Step 3: Replace Query Code

Weaviate queries use GraphQL with nearText or hybrid operators. REM uses a single recall method that runs all retrieval strategies automatically:

// Before (Weaviate GraphQL)
const result = await client.graphql.get()
  .withClassName("Document")
  .withNearText({ concepts: ["refund policy"] })
  .withLimit(5)
  .withFields("content category _additional { distance }")
  .do();

// After (REM Labs)
const results = await rem.recall({
  query: "refund policy",
  namespace: "weaviate-import",
  limit: 5
});

// Results include content, score, metadata, and tags
results.forEach(m => console.log(m.score, m.content.slice(0, 80)));

Step 4: Verify Results

Run side-by-side comparisons on your most common queries to confirm REM returns equivalent or better results:

const testQueries = [
  "How do I reset my password?",
  "What are the pricing tiers?",
  "When was the last API outage?"
];

for (const q of testQueries) {
  const results = await rem.recall({ query: q, limit: 3 });
  console.log(`\nQuery: "${q}"`);
  results.forEach((m, i) => {
    console.log(`  ${i + 1}. [${m.score.toFixed(3)}] ${m.content.slice(0, 60)}...`);
  });
}

Hybrid search built in: Unlike Weaviate where you must choose between nearText, BM25, or hybrid mode per query, REM Labs always runs multi-signal fusion. Every recall query gets vector + full-text + entity graph retrieval automatically.

Upgrade from vector search to AI memory

Free tier. No GraphQL. Multi-signal fusion on every query.

Get Started