@deltamemory/ai-sdk

DeltaMemory integration for Vercel AI SDK. Provides memory tools that AI agents can use to recall past context and store important information.

Installation

npm install @deltamemory/ai-sdk
# or
yarn add @deltamemory/ai-sdk
# or
pnpm add @deltamemory/ai-sdk

The deltamemory client is included as a dependency, so you don't need to install it separately.

Quick Start

import { generateText } from 'ai';
import { openai } from '@ai-sdk/openai';
import { deltaMemoryTools, DeltaMemory } from '@deltamemory/ai-sdk';

const client = new DeltaMemory({
  apiKey: process.env.DELTAMEMORY_API_KEY,  // From dashboard
  baseUrl: process.env.DELTAMEMORY_URL      // Your assigned endpoint
});

const { text } = await generateText({
  model: openai('gpt-4'),
  messages: [
    {
      role: 'system',
      content: 'You are a helpful assistant with memory. Use memory tools to recall context and store important information.'
    },
    {
      role: 'user',
      content: 'What are my preferences?'
    }
  ],
  tools: {
    ...deltaMemoryTools(client, { userId: 'user-123' })
  },
  maxToolRoundtrips: 5
});

console.log(text);

API

deltaMemoryTools(client, config)

Creates memory tools for Vercel AI SDK.

Parameters:

• client: DeltaMemory - DeltaMemory client instance
• config: DeltaMemoryToolsConfig - Configuration object
  • userId: string - User identifier for memory isolation (required)
  • collection?: string - Optional collection name
  • metadata?: Record<string, string> - Optional metadata for all operations

Returns: Object with recallMemory and storeMemory tools

Tools

recallMemory

Search past conversations and user context.

When to use:

• User references past events
• User asks about preferences or history
• Context would improve the response

Parameters:

• query: string - What to search for
• limit?: number - Max results (default: 5)

Returns:

• profiles - Structured user facts
• events - Timeline entries
• context - Pre-formatted context string
• memoryCount - Number of memories found

storeMemory

Store important information for future reference.

When to use:

• User shares preferences
• User asks to remember something
• Important facts emerge

Parameters:

• content: string - Information to remember
• importance?: 'low' | 'medium' | 'high' - Importance level

Returns:

• memoryIds - IDs of stored memories
• extractedFacts - Automatically extracted facts
• extractedConcepts - Automatically extracted concepts

Examples

Basic Chat

import { generateText } from 'ai';
import { openai } from '@ai-sdk/openai';
import { deltaMemoryTools, DeltaMemory } from '@deltamemory/ai-sdk';

const client = new DeltaMemory({
  apiKey: process.env.DELTAMEMORY_API_KEY,
  baseUrl: process.env.DELTAMEMORY_URL
});

async function chat(userId: string, message: string) {
  const { text } = await generateText({
    model: openai('gpt-4'),
    messages: [{ role: 'user', content: message }],
    tools: {
      ...deltaMemoryTools(client, { userId })
    }
  });
  return text;
}

// First message - stores preference
await chat('user-123', 'I prefer TypeScript over JavaScript');

// Second message - recalls preference
await chat('user-123', 'What programming language should I use?');

Streaming

import { streamText } from 'ai';
import { openai } from '@ai-sdk/openai';
import { deltaMemoryTools, DeltaMemory } from '@deltamemory/ai-sdk';

const client = new DeltaMemory({
  apiKey: process.env.DELTAMEMORY_API_KEY,
  baseUrl: process.env.DELTAMEMORY_URL
});

async function chatStream(userId: string, message: string) {
  const { textStream } = await streamText({
    model: openai('gpt-4'),
    messages: [{ role: 'user', content: message }],
    tools: {
      ...deltaMemoryTools(client, { userId })
    }
  });

  for await (const chunk of textStream) {
    process.stdout.write(chunk);
  }
}

Multi-User

import { deltaMemoryTools, DeltaMemory } from '@deltamemory/ai-sdk';

const client = new DeltaMemory({
  apiKey: process.env.DELTAMEMORY_API_KEY,
  baseUrl: process.env.DELTAMEMORY_URL
});

async function chat(userId: string, message: string) {
  const { text } = await generateText({
    model: openai('gpt-4'),
    messages: [{ role: 'user', content: message }],
    tools: {
      ...deltaMemoryTools(client, {
        userId,
        collectionPrefix: 'myapp',  // Collection: 'myapp-user-123'
        metadata: { source: 'chat' }
      })
    }
  });
  return text;
}

Documentation

For complete documentation, visit deltamemory.ai/docs

License

Proprietary - See LICENSE file