Quick Start

Get Recall running in your application in under 5 minutes.

Prerequisites

Before you begin, make sure you have:

Python 3.8+ or Node.js 16+
Redis installed locally or a Redis Cloud instance
A Mem0 API key (get one free)

Installation

Bash

1pip install recall-memory

Bash

1npm install @recall/client

Bash

1yarn add @recall/client

Bash

1pnpm add @recall/client

Basic Setup

1. Start Redis

If you don't have Redis running locally:

Bash

1docker run -d -p 6379:6379 redis:alpine

Bash

1brew services start redis

Bash

1sudo systemctl start redis

2. Set Environment Variables

Create a .env file in your project root:

ENV

1REDIS_URL=redis://localhost:6379
2MEM0_API_KEY=your_mem0_api_key_here
3RECALL_ENV=development

3. Initialize the Client

Python

1from recall import RecallClient
2from dotenv import load_dotenv
3import os
4
5# Load environment variables
6load_dotenv()
7
8# Initialize client
9client = RecallClient(
10    redis_url=os.getenv("REDIS_URL"),
11    mem0_api_key=os.getenv("MEM0_API_KEY"),
12    environment=os.getenv("RECALL_ENV", "development")
13)
14
15# Test the connection
16health = client.health_check()
17print(f"Recall status: {health['status']}")

TypeScript

1import { RecallClient } from "@recall/client";
2import dotenv from "dotenv";
3
4// Load environment variables
5dotenv.config();
6
7// Initialize client
8const client = new RecallClient({
9  redisUrl: process.env.REDIS_URL,
10  mem0ApiKey: process.env.MEM0_API_KEY,
11  environment: process.env.RECALL_ENV || "development",
12});
13
14// Test the connection
15const health = await client.healthCheck();
16console.log(`Recall status: ${health.status}`);

JavaScript

1const { RecallClient } = require("@recall/client");
2require("dotenv").config();
3
4// Initialize client
5const client = new RecallClient({
6  redisUrl: process.env.REDIS_URL,
7  mem0ApiKey: process.env.MEM0_API_KEY,
8  environment: process.env.RECALL_ENV || "development",
9});
10
11// Test the connection
12client.healthCheck().then((health) => {
13  console.log(`Recall status: ${health.status}`);
14});

Your First Memory

Let's create, retrieve, and search memories:

Python

1# Store a memory
2memory = client.add(
3    content="User prefers concise responses and technical details",
4    user_id="user_123",
5    priority="high",
6    metadata={
7        "category": "preferences",
8        "learned_from": "conversation"
9    }
10)
11print(f"Memory stored with ID: {memory['id']}")
12
13# Search memories
14results = client.search(
15    query="user communication preferences",
16    user_id="user_123",
17    limit=5
18)
19
20for memory in results:
21    print(f"- {memory['content']} (relevance: {memory['score']})")
22
23# Get all memories for a user
24all_memories = client.get_all(user_id="user_123")
25print(f"Total memories: {len(all_memories)}")

TypeScript

1// Store a memory
2const memory = await client.add({
3  content: "User prefers concise responses and technical details",
4  userId: "user_123",
5  priority: "high",
6  metadata: {
7    category: "preferences",
8    learnedFrom: "conversation",
9  },
10});
11console.log(`Memory stored with ID: ${memory.id}`);
12
13// Search memories
14const results = await client.search({
15  query: "user communication preferences",
16  userId: "user_123",
17  limit: 5,
18});
19
20results.forEach((memory) => {
21  console.log(`- ${memory.content} (relevance: ${memory.score})`);
22});
23
24// Get all memories for a user
25const allMemories = await client.getAll({ userId: "user_123" });
26console.log(`Total memories: ${allMemories.length}`);

Common Patterns

Conversation Memory

Store and retrieve conversation context:

Python

1# Store conversation turn
2client.add(
3    content=f"User asked about {topic}. Provided detailed explanation.",
4    user_id=user_id,
5    priority="medium",
6    metadata={
7        "type": "conversation",
8        "session_id": session_id,
9        "timestamp": datetime.now().isoformat()
10    }
11)
12
13# Retrieve recent conversation context
14context = client.search(
15    query="recent conversations",
16    user_id=user_id,
17    filters={"type": "conversation"},
18    limit=10
19)

TypeScript

1// Store conversation turn
2await client.add({
3  content: `User asked about ${topic}. Provided detailed explanation.`,
4  userId: userId,
5  priority: "medium",
6  metadata: {
7    type: "conversation",
8    sessionId: sessionId,
9    timestamp: new Date().toISOString(),
10  },
11});
12
13// Retrieve recent conversation context
14const context = await client.search({
15  query: "recent conversations",
16  userId: userId,
17  filters: { type: "conversation" },
18  limit: 10,
19});

User Preferences

Track and apply user preferences:

Python

1# Store preference
2client.add(
3    content="Prefers email notifications over SMS",
4    user_id=user_id,
5    priority="high",
6    metadata={"type": "preference", "category": "notifications"}
7)
8
9# Check preferences before action
10prefs = client.search(
11    query="notification preferences",
12    user_id=user_id,
13    filters={"type": "preference"}
14)

TypeScript

1// Store preference
2await client.add({
3  content: "Prefers email notifications over SMS",
4  userId: userId,
5  priority: "high",
6  metadata: { type: "preference", category: "notifications" },
7});
8
9// Check preferences before action
10const prefs = await client.search({
11  query: "notification preferences",
12  userId: userId,
13  filters: { type: "preference" },
14});

Performance Tips

1. Use Priority Levels

Set appropriate priority levels to optimize cache usage:

critical: Always in cache, never evicted
high: Preferentially cached, rarely evicted
medium: Cached when accessed, normal eviction
low: Minimal caching, first to evict

2. Batch Operations

Use batch methods for better performance:

Python

1# Add multiple memories at once
2memories = [
3    {"content": "Fact 1", "user_id": "user_123"},
4    {"content": "Fact 2", "user_id": "user_123"},
5    {"content": "Fact 3", "user_id": "user_123"}
6]
7client.add_batch(memories)

TypeScript

1// Add multiple memories at once
2const memories = [
3  { content: "Fact 1", userId: "user_123" },
4  { content: "Fact 2", userId: "user_123" },
5  { content: "Fact 3", userId: "user_123" },
6];
7await client.addBatch(memories);

3. Use Async Operations

For non-critical memories, use async mode:

Python

1# Fire-and-forget for non-critical memories
2client.add(
3    content="Background information",
4    user_id="user_123",
5    priority="low",
6    async_mode=True  # Don't wait for cloud sync
7)

TypeScript

1// Fire-and-forget for non-critical memories
2await client.add({
3  content: "Background information",
4  userId: "user_123",
5  priority: "low",
6  asyncMode: true, // Don't wait for cloud sync
7});

Next Steps

Now that you have Recall running:

Explore the API Reference for all available methods
Check out Examples for real-world use cases
Learn about Advanced Features like custom caching strategies
Read the Best Practices guide for production deployments

Quick Start

Prerequisites

Installation

Basic Setup

1. Start Redis

2. Set Environment Variables

3. Initialize the Client

Your First Memory

Common Patterns

Conversation Memory

User Preferences

Performance Tips

1. Use Priority Levels

2. Batch Operations

3. Use Async Operations

Next Steps

Need Help?