Skip to main content
The Memory API provides a powerful interface for storing, searching, and managing user memories using mem0 with Supabase vector store. The API includes automatic memory type classification using LLM-based categorization.

​
Base URL

http://localhost:8000

​
Memory Types

The API supports five distinct memory types for intelligent categorization:
Temporary states and current activities happening right now or today that will change soon.Examples:
  • “I’m currently working on the dashboard redesign”
  • “Right now I’m debugging the auth flow”
  • “I need to finish this today”
Key indicators: “currently”, “right now”, “today”, “at the moment”, “working on”
Permanent personal facts, preferences, identity, and habits that persist over time.Examples:
  • “I prefer dark mode in all applications”
  • “My name is John and I’m a software engineer”
  • “I like pizza with extra cheese”
Key indicators: General preferences, identity statements, lasting characteristics
Past events with specific time context - things that already happened.Examples:
  • “Yesterday I had a meeting with the design team”
  • “Last week I went to the React conference”
  • “I met John at the tech summit in 2024”
Key indicators: “yesterday”, “last week”, “last month”, past tense events
General knowledge or facts about the world (not personal preferences).Examples:
  • “Python uses indentation for code blocks”
  • “The capital of France is Paris”
  • “React is a JavaScript library for building UIs”
Key indicators: Objective facts, definitions, general truths
How-to knowledge, step-by-step processes, and instructions.Examples:
  • “To deploy, run npm build then npm start”
  • “The recipe requires boiling water first”
  • “First authenticate, then call the API endpoint”
Key indicators: “to do X, first…”, “steps to…”, instructions

​
Auto-Classification

By default, the API automatically classifies memories using GPT-4.1-nano-2025-04-14. You can disable this by setting auto_classify: false in your requests, or override it by providing a memory_type in the metadata.

​
Available Endpoints

Add Memory

Add new memories from conversations with automatic classification

Add Image Memory

Add image-based memories with visual context

Search Memory

Search memories with semantic similarity and filtering

Get All Memories

Retrieve all memories for a user

Get Memory

Get a specific memory by ID

Update Memory

Update existing memory content

Delete Memory

Delete specific memories by ID

Memory History

View change history for a memory

​
Authentication

Currently, the API does not require authentication. In production environments, ensure you implement proper authentication middleware.

​
Configuration

The API is configured with:
  • LLM Provider: OpenAI (gpt-4.1-nano-2025-04-14)
  • Vector Store: Supabase with HNSW index and cosine distance
  • Graph Store: Neo4j (optional)
  • CORS: Enabled for all origins

​
Health Check

curl http://localhost:8000/
Response: 
{
  "status": "ok",
  "message": "Memory API is running"
}

Build docs developers (and LLMs) love

Get started for freeTalk to us