@getmem/sdk

TypeScript SDK for getmem.ai Memory API.

Install

npm install @getmem/sdk

Quick Start

import { GetMem } from '@getmem/sdk';

const mem = new GetMem({
  apiKey: 'gm_live_...',
});

// Get memory context for an LLM prompt
const { context, memories, meta } = await mem.getContext('user-123', 'What do you know about me?');

console.log(context);        // ready-to-use context string for system prompt
console.log(meta.total_ms);  // request timing

Usage with OpenAI

import { GetMem } from '@getmem/sdk';
import OpenAI from 'openai';

const mem = new GetMem({ apiKey: 'gm_live_...' });
const openai = new OpenAI({ apiKey: 'sk-...' });

const userId = 'user-123';
const userMessage = 'Tell me something about my hobbies';

// 1. Get memory context
const { context } = await mem.getContext(userId, userMessage);

// 2. Call LLM with memory as context
const completion = await openai.chat.completions.create({
  model: 'gpt-4o-mini',
  messages: [
    { role: 'system', content: `You are a helpful assistant.\n\n## Memory Context\n${context}` },
    { role: 'user', content: userMessage },
  ],
});

const reply = completion.choices[0].message.content!;

// 3. Save conversation to memory
await mem.ingestConversation(userId, userMessage, reply);

API

new GetMem(config)

| Option | Type | Default | Description | |-----------|----------|--------------------------------|---------------------| | apiKey | string | required | API key | | baseUrl | string | https://memory.getmem.ai | API base URL | | timeout | number | 30000 | Request timeout, ms |

mem.getContext(userId, query)

Retrieve relevant memory context. Returns:

{
  context: string;          // ready for LLM system prompt
  memories: MemoryItem[];   // individual items with scores
  meta: ContextMeta;        // timings, token count, entities
}

mem.ingest(userId, messages, sessionId?)

Ingest raw messages into memory.

await mem.ingest('user-123', [
  { role: 'user', content: 'I like fishing', timestamp: new Date().toISOString() },
  { role: 'assistant', content: 'Great hobby!', timestamp: new Date().toISOString() },
], 'session-abc');

mem.ingestConversation(userId, userMessage, assistantMessage, sessionId?)

Convenience method for a single user+assistant exchange.

await mem.ingestConversation('user-123', 'I like fishing', 'Great hobby!');

mem.health()

Health check. Returns service status and component health.

Error handling

import { GetMemError } from '@getmem/sdk';

try {
  await mem.getContext('user-123', 'hello');
} catch (e) {
  if (e instanceof GetMemError) {
    console.error(e.status, e.body);
  }
}

Custom base URL

For local development or self-hosted instances:

const mem = new GetMem({
  apiKey: 'gm_live_...',
  baseUrl: 'http://localhost:8001',
});