Skip to main content

Personal Assistant

Build an AI assistant that learns about you over time and provides personalized help.

The Problem​

Generic AI assistants:

  • Give the same advice to everyone
  • Can't remember your preferences
  • Ask for context you've already provided

The Solution​

A memory-enabled assistant that:

  • Learns your preferences and habits
  • Provides personalized recommendations
  • Gets smarter with every interaction

What It Remembers​

Personal Assistant Memory:
├── Preferences
│   ├── "Prefers morning meetings"
│   └── "Coffee: oat milk latte"
├── Context
│   ├── "Works at Acme Corp as PM"
│   └── "Main project: Q1 Launch"
└── Important Dates
    ├── "Partner's birthday: March 15"
    └── "Project deadline: April 1"

Implementation​

from openai import OpenAI
import requests

MEMORY_API = "https://api.memory.tensorheart.com/v1"
API_KEY = "mem_live_..."
openai = OpenAI()

def query_memory(user_id: str, context: str) -> list[str]:
    """Get relevant memories for this user."""
    response = requests.post(
        f"{MEMORY_API}/query",
        headers={"Authorization": f"Bearer {API_KEY}"},
        json={
            "context": context,
            "space_id": f"user_{user_id}",
            "max_memories": 8
        }
    )
    return [m["text"] for m in response.json().get("data", {}).get("memories", [])]

def personal_assistant(user_id: str, message: str) -> str:
    """A personalized AI assistant."""
    memories = query_memory(user_id, message)
    context = "\n".join(f"- {m}" for m in memories) if memories else "New user."

    system_prompt = f"""You are a helpful personal assistant.

What you know about this user:
{context}

Use this naturally in responses. Be concise and actionable."""

    response = openai.chat.completions.create(
        model="gpt-4o",
        messages=[
            {"role": "system", "content": system_prompt},
            {"role": "user", "content": message}
        ]
    )
    return response.choices[0].message.content

Example Conversations​

Day 1: Learning​

User: Hey, I need help planning my day. I have back-to-back
      meetings and a deadline Friday.

Bot: I'd suggest blocking 2-hour focus time for the deadline work.
     What time are your meetings?

User: Mostly afternoons. I'm a morning person anyway.

Bot: Perfect. I'll note that you prefer mornings for deep work.

The assistant learns: "User prefers morning focus time"

Week 2: Personalized​

User: I have a presentation tomorrow, when should I prep?

Bot: Since you do your best work in the morning, I'd block
     9-11am tomorrow for prep. That gives you focused time
     before your usual afternoon meetings.

What Makes This Work​

FeatureHow It Helps
Preference learningAdapts to your style over time
Context retrievalOnly fetches relevant info per query
Space isolationEach user's memories are private

Next Steps​