Mem0 Node SDK (v3.0.0)

A ground-up redesign of how memories are extracted, stored, and retrieved — plus a long-overdue cleanup of the SDK surface.

Highlights

• New extraction algorithm — single-pass, ADD-only, roughly half the latency
• Multi-signal hybrid retrieval — semantic + BM25 keyword + entity matching fused into one score
• Built-in entity linking — replaces graph memory with no external store to manage
• Fully camelCase SDK — both ways — every parameter and every field on every response is camelCase, end to end

What's new

Single-pass ADD-only extraction

One LLM call per add(). No separate UPDATE/DELETE pass. The model now spends its capacity on understanding the input instead of diffing against existing memories, and agent-generated facts ("I've booked your flight for March 3rd") are captured as first-class memories for the first time. Hash-based deduplication prevents exact duplicates; ranking at retrieval time handles the rest.

Hybrid retrieval

Semantic vector similarity, BM25 keyword matching, and entity-graph boosting are normalized and fused into a single score on every result. Entities extracted from both new memories and queries boost ranking on matching results.

Entity linking (replaces graph memory)

Entities (proper nouns, quoted text, compound noun phrases) are extracted automatically during add() and stored in a parallel {collection}_entities collection inside your existing vector store. At query time, entities from the query boost ranking on matching memories. No Neo4j / Memgraph / Kuzu / Apache AGE deployment needed — graph driver support has been removed from the SDK.

Vector store enhancements

All supported vector stores now implement keyword search and batch search primitives, used transparently by the new retrieval pipeline.

Stricter, clearer validation

• Empty or whitespace-only entity IDs throw with a specific message
• threshold must be in [0, 1] — out-of-range values throw
• messages must be a string or array; null / undefined throws
• customPrompt → renamed to customInstructions

Platform client cleanup

• new MemoryClient({ apiKey }) — organizationId, projectId, organizationName, projectName removed
• All request parameters are now camelCase — userId, agentId, topK, filters — the SDK handles snake_case conversion at the wire boundary
• All response fields are now camelCase too — createdAt, updatedAt, userId, agentId, runId, memoryId, eventId, scoreBreakdown, and every other field come back in camelCase. No more mixing user_id in the response with userId in the request — one casing, everywhere.
• getAll returns the paginated envelope {count, next, previous, results} when the filter-based API is used
• add is async by default and returns {status: "PENDING", eventId} for polling
• Removed: OutputFormat enum, API_VERSION enum, enableGraph, asyncMode, outputFormat, filterMemories, batchSize, forceAddOnly, immutable, expirationDate, includes, excludes, keywordSearch

Breaking changes at a glance

add() and deleteAll() continue to accept entity IDs at the top level.

Install

npm install mem0ai@latest

Quick example

import MemoryClient from "mem0ai";

const client = new MemoryClient({ apiKey: process.env.MEM0_API_KEY });

await client.add(messages, { userId: "alice" });

const results = await client.search("what does alice like?", {
  filters: { AND: [{ user_id: "alice" }] },
  topK: 10,
});

Full migration guide: docs.mem0.ai/migration/oss-v2-to-v3