supermemory/apps/docs/cookbook/personal-assistant.mdx

---
title: "Personal AI Assistant"
description: "Build an AI assistant that remembers user preferences, habits, and context across conversations"
---

Build a personal AI assistant that learns and remembers everything about the user - their preferences, habits, work context, and conversation history. This recipe shows how to create a truly personalized AI experience using Supermemory's memory tools.

## What You'll Build

A personal AI assistant that:
- **Remembers user preferences** (dietary restrictions, work schedule, communication style)
- **Learns from conversations** and improves responses over time
- **Maintains context** across multiple chat sessions
- **Provides personalized recommendations** based on user history
- **Handles multiple conversation topics** while maintaining context

## Prerequisites

- Node.js 18+ or Python 3.8+
- Supermemory API key
- OpenAI or Anthropic API key
- Basic understanding of chat applications

## Implementation

### Step 1: Project Setup

<Tabs>
  <Tab title="Next.js (TypeScript)">
    ```bash
    npx create-next-app@latest personal-ai --typescript --tailwind --eslint
    cd personal-ai
    npm install @supermemory/tools ai openai
    ```

    Create your environment variables:
    ```bash .env.local
    SUPERMEMORY_API_KEY=your_supermemory_key
    OPENAI_API_KEY=your_openai_key
    ```
  </Tab>

  <Tab title="Python">
    ```bash
    mkdir personal-ai && cd personal-ai
    python -m venv venv
    source venv/bin/activate  # On Windows: venv\Scripts\activate
    pip install supermemory openai fastapi uvicorn python-multipart
    ```

    Create your environment variables:
    ```bash .env
    SUPERMEMORY_API_KEY=your_supermemory_key
    OPENAI_API_KEY=your_openai_key
    ```
  </Tab>
</Tabs>

### Step 2: Core Assistant Logic

<Tabs>
  <Tab title="Next.js API Route">
    ```typescript app/api/chat/route.ts
    import { streamText } from 'ai'
    import { createOpenAI } from '@ai-sdk/openai'
    import { supermemoryTools } from '@supermemory/tools/ai-sdk'

    const openai = createOpenAI({
      apiKey: process.env.OPENAI_API_KEY!
    })

    export async function POST(request: Request) {
      const { messages, userId = 'default-user' } = await request.json()

      const result = await streamText({
        model: openai('gpt-5'),
        messages,
        tools: supermemoryTools(process.env.SUPERMEMORY_API_KEY!, {
          containerTags: [userId]
        }),
        system: `You are a highly personalized AI assistant. Your primary goal is to learn about the user and provide increasingly personalized help over time.

    MEMORY MANAGEMENT:
    1. When users share personal information, preferences, or context, immediately use addMemory to store it
    2. Before responding to requests, search your memories for relevant context about the user
    3. Use past conversations to inform current responses
    4. Remember user's communication style, preferences, and frequently discussed topics

    PERSONALITY:
    - Adapt your communication style to match the user's preferences
    - Reference past conversations naturally when relevant
    - Proactively offer help based on learned patterns
    - Be genuinely helpful while respecting privacy

    EXAMPLES OF WHAT TO REMEMBER:
    - Work schedule and role
    - Dietary preferences/restrictions
    - Communication preferences (formal/casual)
    - Frequent topics of interest
    - Goals and projects they're working on
    - Family/personal context they share
    - Preferred tools and workflows
    - Time zone and availability

    Always search memories before responding to provide personalized, contextual help.`
      })

      return result.toAIStreamResponse()
    }
    ```
  </Tab>

  <Tab title="Python FastAPI">
    ```python main.py
    from fastapi import FastAPI, HTTPException
    from fastapi.responses import StreamingResponse
    import openai
    from supermemory import Supermemory
    import json
    import os
    from typing import List, Dict, Any
    import asyncio

    app = FastAPI()

    openai_client = openai.AsyncOpenAI(api_key=os.getenv("OPENAI_API_KEY"))
    supermemory_client = Supermemory(api_key=os.getenv("SUPERMEMORY_API_KEY"))

    SYSTEM_PROMPT = """You are a highly personalized AI assistant. Your primary goal is to learn about the user and provide increasingly personalized help over time.

    MEMORY MANAGEMENT:
    1. When users share personal information, preferences, or context, immediately store it
    2. Before responding to requests, search for relevant context about the user
    3. Use past conversations to inform current responses
    4. Remember user's communication style, preferences, and frequently discussed topics

    PERSONALITY:
    - Adapt your communication style to match the user's preferences
    - Reference past conversations naturally when relevant
    - Proactively offer help based on learned patterns
    - Be genuinely helpful while respecting privacy

    Always search memories before responding to provide personalized, contextual help."""

    async def search_user_memories(query: str, user_id: str) -> str:
        """Search user's memories for relevant context"""
        try:
            results = supermemory_client.search.memories(
                q=query,
                container_tag=f"user_{user_id}",
                limit=5
            )

            if results.results:
                context = "\n".join([r.memory for r in results.results])
                return f"Relevant memories about the user:\n{context}"
            return "No relevant memories found."
        except Exception as e:
            return f"Error searching memories: {e}"

    async def add_user_memory(content: str, user_id: str):
        """Add new information to user's memory"""
        try:
            supermemory_client.memories.add(
                content=content,
                container_tag=f"user_{user_id}",
                metadata={"type": "personal_info", "timestamp": "auto"}
            )
        except Exception as e:
            print(f"Error adding memory: {e}")

    @app.post("/chat")
    async def chat_endpoint(data: dict):
        messages = data.get("messages", [])
        user_id = data.get("userId", "default-user")

        if not messages:
            raise HTTPException(status_code=400, detail="No messages provided")

        # Get user's last message for memory search
        user_message = messages[-1]["content"] if messages else ""

        # Search for relevant memories
        memory_context = await search_user_memories(user_message, user_id)

        # Add system message with memory context
        enhanced_messages = [
            {"role": "system", "content": f"{SYSTEM_PROMPT}\n\n{memory_context}"}
        ] + messages

        try:
            response = await openai_client.chat.completions.create(
                model="gpt-5",
                messages=enhanced_messages,
                stream=True,
                temperature=0.7
            )

            async def generate():
                full_response = ""
                async for chunk in response:
                    if chunk.choices[0].delta.content:
                        content = chunk.choices[0].delta.content
                        full_response += content
                        yield f"data: {json.dumps({'content': content})}\n\n"

                # After response is complete, analyze for memory-worthy content
                if "remember" in user_message.lower() or any(word in user_message.lower() for word in ["prefer", "like", "dislike", "work", "schedule", "diet"]):
                    await add_user_memory(user_message, user_id)

            return StreamingResponse(generate(), media_type="text/plain")

        except Exception as e:
            raise HTTPException(status_code=500, detail=str(e))

    if __name__ == "__main__":
        import uvicorn
        uvicorn.run(app, host="0.0.0.0", port=8000)
    ```
  </Tab>
</Tabs>

### Step 3: Frontend Interface

<Tabs>
  <Tab title="Next.js Chat Component">
    ```tsx app/page.tsx
    'use client'

    import { useChat } from 'ai/react'
    import { useState, useEffect } from 'react'

    export default function PersonalAssistant() {
      const [userId, setUserId] = useState('')
      const [userName, setUserName] = useState('')

      const { messages, input, handleInputChange, handleSubmit, isLoading } = useChat({
        api: '/api/chat',
        body: {
          userId
        }
      })

      // Generate or retrieve user ID
      useEffect(() => {
        const storedUserId = localStorage.getItem('personal-ai-user-id')
        const storedUserName = localStorage.getItem('personal-ai-user-name')

        if (storedUserId) {
          setUserId(storedUserId)
          setUserName(storedUserName || '')
        } else {
          const newUserId = `user_${Date.now()}_${Math.random().toString(36).substr(2, 9)}`
          localStorage.setItem('personal-ai-user-id', newUserId)
          setUserId(newUserId)
        }
      }, [])

      const handleNameSubmit = (e: React.FormEvent) => {
        e.preventDefault()
        if (userName.trim()) {
          localStorage.setItem('personal-ai-user-name', userName)
          // Send introduction message
          handleSubmit(e, {
            data: {
              content: `Hi! My name is ${userName}. I'm looking for a personal AI assistant that can learn about me and help me with various tasks.`
            }
          })
        }
      }

      return (
        <div className="flex flex-col h-screen max-w-4xl mx-auto p-4">
          {/* Header */}
          <div className="bg-gradient-to-r from-blue-500 to-purple-600 text-white p-6 rounded-lg mb-6">
            <h1 className="text-2xl font-bold">Personal AI Assistant</h1>
            <p className="text-blue-100">
              {userName ? `Hello ${userName}!` : 'Your AI that learns and remembers'}
            </p>
          </div>

          {/* Name Setup */}
          {!userName && (
            <div className="bg-white border border-gray-200 rounded-lg p-6 mb-6">
              <form onSubmit={handleNameSubmit} className="flex gap-2">
                <input
                  type="text"
                  value={userName}
                  onChange={(e) => setUserName(e.target.value)}
                  placeholder="What should I call you?"
                  className="flex-1 p-2 border border-gray-300 rounded focus:outline-none focus:ring-2 focus:ring-blue-500"
                />
                <button
                  type="submit"
                  className="px-4 py-2 bg-blue-500 text-white rounded hover:bg-blue-600 focus:outline-none focus:ring-2 focus:ring-blue-500"
                >
                  Get Started
                </button>
              </form>
            </div>
          )}

          {/* Messages */}
          <div className="flex-1 overflow-y-auto space-y-4 mb-4">
            {messages.length === 0 && userName && (
              <div className="bg-gray-50 border border-gray-200 rounded-lg p-4">
                <p className="text-gray-600">
                  Hi {userName}! I'm your personal AI assistant. I'll learn about your preferences,
                  work style, and interests as we chat. Feel free to share anything you'd like me to remember!
                </p>
                <div className="mt-3 text-sm text-gray-500">
                  <p><strong>Try saying:</strong></p>
                  <ul className="list-disc list-inside mt-1 space-y-1">
                    <li>"I work as a software engineer and prefer concise responses"</li>
                    <li>"Remember that I'm vegetarian and allergic to nuts"</li>
                    <li>"I usually work from 9-5 EST and take lunch at noon"</li>
                  </ul>
                </div>
              </div>
            )}

            {messages.map((message) => (
              <div
                key={message.id}
                className={`p-4 rounded-lg ${
                  message.role === 'user'
                    ? 'bg-blue-500 text-white ml-auto max-w-2xl'
                    : 'bg-white border border-gray-200 max-w-2xl'
                }`}
              >
                <div className="flex items-start space-x-2">
                  {message.role === 'assistant' && (
                    <div className="w-8 h-8 bg-gradient-to-r from-blue-500 to-purple-600 rounded-full flex items-center justify-center text-white text-sm font-bold">
                      AI
                    </div>
                  )}
                  <div className="flex-1">
                    <p className="whitespace-pre-wrap">{message.content}</p>
                  </div>
                </div>
              </div>
            ))}

            {isLoading && (
              <div className="bg-white border border-gray-200 rounded-lg p-4 max-w-2xl">
                <div className="flex items-center space-x-2">
                  <div className="w-8 h-8 bg-gradient-to-r from-blue-500 to-purple-600 rounded-full flex items-center justify-center text-white text-sm font-bold">
                    AI
                  </div>
                  <div className="flex space-x-1">
                    <div className="w-2 h-2 bg-gray-400 rounded-full animate-bounce"></div>
                    <div className="w-2 h-2 bg-gray-400 rounded-full animate-bounce" style={{animationDelay: '0.1s'}}></div>
                    <div className="w-2 h-2 bg-gray-400 rounded-full animate-bounce" style={{animationDelay: '0.2s'}}></div>
                  </div>
                </div>
              </div>
            )}
          </div>

          {/* Input */}
          {userName && (
            <form onSubmit={handleSubmit} className="flex gap-2">
              <input
                value={input}
                onChange={handleInputChange}
                placeholder="Tell me something about yourself, or ask for help..."
                className="flex-1 p-3 border border-gray-300 rounded-lg focus:outline-none focus:ring-2 focus:ring-blue-500"
                disabled={isLoading}
              />
              <button
                type="submit"
                disabled={isLoading || !input.trim()}
                className="px-6 py-3 bg-blue-500 text-white rounded-lg hover:bg-blue-600 focus:outline-none focus:ring-2 focus:ring-blue-500 disabled:opacity-50 disabled:cursor-not-allowed"
              >
                Send
              </button>
            </form>
          )}
        </div>
      )
    }
    ```
  </Tab>

  <Tab title="Python Streamlit">
    ```python streamlit_app.py
    import streamlit as st
    import requests
    import json
    import uuid

    st.set_page_config(page_title="Personal AI Assistant", page_icon="🤖", layout="wide")

    # Initialize session state
    if 'messages' not in st.session_state:
        st.session_state.messages = []
    if 'user_id' not in st.session_state:
        st.session_state.user_id = f"user_{uuid.uuid4().hex[:8]}"
    if 'user_name' not in st.session_state:
        st.session_state.user_name = None

    # Header
    st.title("🤖 Personal AI Assistant")
    st.markdown("*Your AI that learns and remembers*")

    # Sidebar for user info
    with st.sidebar:
        st.header("👤 User Profile")

        if not st.session_state.user_name:
            name = st.text_input("What should I call you?")
            if st.button("Get Started") and name:
                st.session_state.user_name = name
                st.session_state.messages.append({
                    "role": "user",
                    "content": f"Hi! My name is {name}. I'm looking for a personal AI assistant."
                })
                st.rerun()
        else:
            st.write(f"**Name:** {st.session_state.user_name}")
            st.write(f"**User ID:** {st.session_state.user_id[:12]}...")

            if st.button("Reset Conversation"):
                st.session_state.messages = []
                st.rerun()

        st.markdown("---")
        st.markdown("""
        ### 💡 Try saying:
        - "I work as a software engineer and prefer concise responses"
        - "Remember that I'm vegetarian"
        - "I usually work from 9-5 EST"
        """)

    # Main chat interface
    if st.session_state.user_name:
        # Display messages
        for message in st.session_state.messages:
            with st.chat_message(message["role"]):
                st.markdown(message["content"])

        # Chat input
        if prompt := st.chat_input("Tell me something about yourself, or ask for help..."):
            # Add user message
            st.session_state.messages.append({"role": "user", "content": prompt})

            with st.chat_message("user"):
                st.markdown(prompt)

            # Get AI response
            with st.chat_message("assistant"):
                with st.spinner("Thinking..."):
                    try:
                        response = requests.post(
                            "http://localhost:8000/chat",
                            json={
                                "messages": st.session_state.messages,
                                "userId": st.session_state.user_id
                            },
                            timeout=30
                        )

                        if response.status_code == 200:
                            # Handle streaming response
                            full_response = ""
                            for line in response.iter_lines():
                                if line:
                                    try:
                                        data = json.loads(line.decode('utf-8').replace('data: ', ''))
                                        if 'content' in data:
                                            full_response += data['content']
                                    except:
                                        continue

                            st.markdown(full_response)
                            st.session_state.messages.append({
                                "role": "assistant",
                                "content": full_response
                            })
                        else:
                            st.error(f"Error: {response.status_code}")
                    except Exception as e:
                        st.error(f"Connection error: {e}")

    else:
        st.info("👆 Please enter your name in the sidebar to get started!")

    # Run with: streamlit run streamlit_app.py
    ```
  </Tab>
</Tabs>

## Testing Your Assistant

### Step 4: Test Memory Formation

Try these conversation flows to test memory capabilities:

1. **Personal Preferences**:
   ```
   User: "Hi! I'm Sarah, a product manager at a tech startup. I prefer brief, actionable responses and I'm always busy with user research."

   Assistant: [Should remember name, role, communication preference]

   User: "What's a good way to prioritize features?"

   Assistant: [Should reference that you're a PM and prefer brief responses]
   ```

2. **Dietary & Lifestyle**:
   ```
   User: "Remember that I'm vegan and I work out every morning at 6 AM."

   User: "Suggest a quick breakfast for tomorrow."

   Assistant: [Should suggest vegan options that work for pre/post workout]
   ```

3. **Work Context**:
   ```
   User: "I'm working on a React project and I prefer TypeScript over JavaScript."

   User: "Help me with state management."

   Assistant: [Should suggest TypeScript-specific solutions]
   ```

### Step 5: Verify Memory Storage

Check that memories are being stored properly:

<Tabs>
  <Tab title="TypeScript">
    ```typescript scripts/check-memories.ts
    import { Supermemory } from '@supermemory/tools'

    const client = new Supermemory({
      apiKey: process.env.SUPERMEMORY_API_KEY!
    })

    async function checkUserMemories(userId: string) {
      try {
        const memories = await client.memories.list({
          containerTags: [userId],
          limit: 20,
          sort: 'updatedAt',
          order: 'desc'
        })

        console.log(`Found ${memories.memories.length} memories for ${userId}:`)
        memories.memories.forEach((memory, i) => {
          console.log(`${i + 1}. ${memory.content.substring(0, 100)}...`)
        })

        // Test search
        const searchResults = await client.search.memories({
          q: "preferences work",
          containerTag: userId,
          limit: 5
        })

        console.log('\nSearch Results:')
        searchResults.results.forEach((result, i) => {
          console.log(`${i + 1}. (${result.similarity}) ${result.memory.substring(0, 100)}...`)
        })

      } catch (error) {
        console.error('Error:', error)
      }
    }

    // Run: npx ts-node scripts/check-memories.ts USER_ID_HERE
    checkUserMemories(process.argv[2] || 'default-user')
    ```
  </Tab>

  <Tab title="Python">
    ```python check_memories.py
    from supermemory import Supermemory
    import os
    import sys

    client = Supermemory(api_key=os.getenv("SUPERMEMORY_API_KEY"))

    def check_user_memories(user_id):
        try:
            # List all memories for user
            memories = client.memories.list(
                container_tags=[user_id],
                limit=20,
                sort="updatedAt",
                order="desc"
            )

            print(f"Found {len(memories.memories)} memories for {user_id}:")
            for i, memory in enumerate(memories.memories):
                print(f"{i + 1}. {memory.content[:100]}...")

            # Test search
            search_results = client.search.memories(
                q="preferences work",
                container_tag=user_id,
                limit=5
            )

            print('\nSearch Results:')
            for i, result in enumerate(search_results.results):
                print(f"{i + 1}. ({result.similarity}) {result.memory[:100]}...")

        except Exception as error:
            print(f'Error: {error}')

    # Run: python check_memories.py USER_ID_HERE
    user_id = sys.argv[1] if len(sys.argv) > 1 else 'default-user'
    check_user_memories(user_id)
    ```
  </Tab>
</Tabs>

## Production Considerations

### Security & Privacy

1. **User Isolation**:
   ```typescript
   // Always use user-specific container tags
   const tools = supermemoryTools(apiKey, {
     containerTags: [userId]
   })
   ```

2. **Memory Encryption**:
   ```typescript
   // For sensitive data, consider client-side encryption
   const encryptedContent = encrypt(sensitiveData, userKey)
   await client.memories.add({
     content: encryptedContent,
     containerTag: userId,
     metadata: { encrypted: true }
   })
   ```

### Performance Optimization

1. **Memory Search Optimization**:
   ```typescript
   // Use appropriate thresholds for speed vs accuracy
   const quickSearch = await client.search.memories({
     q: userQuery,
     containerTag: userId,
     threshold: 0.6,     // Balanced
     rerank: false,      // Skip for speed
     limit: 3            // Fewer results
   })
   ```

2. **Caching Strategy**:
   ```typescript
   // Cache frequently accessed user context
   const userContext = await redis.get(`user_context:${userId}`)
   if (!userContext) {
     const memories = await client.search.memories({
       q: "user preferences work style",
       containerTag: userId,
       limit: 10
     })
     await redis.setex(`user_context:${userId}`, 300, JSON.stringify(memories))
   }
   ```

### Monitoring & Analytics

```typescript
// Track memory formation and retrieval
const analytics = {
  memoriesCreated: await redis.incr(`memories_created:${userId}`),
  searchesPerformed: await redis.incr(`searches:${userId}`),
  conversationLength: messages.length
}

// Log for analysis
console.log('User Interaction:', {
  userId,
  action: 'chat_response',
  memoriesFound: searchResults.results.length,
  responseTime: Date.now() - startTime,
  ...analytics
})
```

## Extensions & Customization

### 1. Add Personality Profiles

```typescript
const personalityProfiles = {
  professional: "Respond in a formal, business-appropriate tone",
  casual: "Use a friendly, conversational tone with occasional humor",
  technical: "Provide detailed technical explanations with examples",
  concise: "Keep responses brief and to the point"
}

// Add to system prompt based on user preference
const userProfile = await getUserProfile(userId)
const systemPrompt = `${basePrompt}\n\nCommunication Style: ${personalityProfiles[userProfile.style]}`
```

### 2. Smart Notifications

```typescript
// Proactive suggestions based on user patterns
const shouldSuggest = await analyzeUserPatterns(userId)
if (shouldSuggest.type === 'daily_standup') {
  return {
    message: "Based on your schedule, would you like me to help prepare for your 9 AM standup?",
    suggestedActions: ["Review yesterday's progress", "Prepare today's goals"]
  }
}
```

### 3. Multi-Modal Memory

```typescript
// Handle images and documents
if (message.attachments) {
  for (const attachment of message.attachments) {
    await client.memories.uploadFile({
      file: attachment,
      containerTag: userId,
      metadata: {
        type: 'user_shared',
        context: message.content
      }
    })
  }
}
```

## Next Steps

- **Scale to multiple users**: Add user authentication and proper isolation
- **Add voice interaction**: Integrate with speech-to-text/text-to-speech APIs
- **Mobile app**: Create React Native or Flutter mobile version
- **Integrations**: Connect to calendar, email, task management tools
- **Advanced AI features**: Add emotion detection, conversation summarization

## Troubleshooting

**Memory not persisting?**
- Check that `x-sm-user-id` header is consistent
- Verify API key has write permissions
- Ensure container tags are properly set

**Responses not personalized?**
- Increase search limit to find more relevant memories
- Lower threshold to cast a wider net
- Check that memories are being added with proper context

**Performance issues?**
- Reduce search limits for faster responses
- Implement caching for frequent searches
- Use appropriate thresholds to balance speed vs accuracy

---

*This recipe provides the foundation for a personal AI assistant. Customize it based on your specific needs and use cases.*