mindverse/second-me

★ 15,532PythonAudience · generalComplexity · 3/5LicenseSetup · moderate

Mindmap

mindmap
  root((Second Me))
    What it is
      Personal AI
      Runs locally
      Private data
    Core ideas
      Memory modeling
      Me-Alignment
      Identity capture
    Use cases
      Private assistant
      Digital stand-in
      AI collaboration
    Setup
      Docker one-command
      Apple MLX support
      Web interface

mindmap root((Second Me)) What it is Personal AI Runs locally Private data Core ideas Memory modeling Me-Alignment Identity capture Use cases Private assistant Digital stand-in AI collaboration Setup Docker one-command Apple MLX support Web interface

Click or tap to explore — scroll the page freely

Things people build with this

USE CASE 1

Train a personal AI on your own notes and memories so it answers from your perspective instead of giving generic responses

USE CASE 2

Run a fully private AI assistant on your own hardware with no data sent to cloud services

USE CASE 3

Share your AI self with other people or apps as a digital stand-in that can represent your views

USE CASE 4

Collaborate with multiple personal AIs on ideas using the built-in multi-agent AI Space feature

Tech stack

PythonDockerMLX

Getting it running

Difficulty · moderate Time to first run · 30min

Start with a single Docker command after cloning, larger models need more RAM and Apple M-series chips benefit from MLX.

In plain English

Second Me is an open-source prototype that lets you train your own personal AI, an "AI self" that captures who you are and represents you in different contexts. The project frames itself as a response to centralized AI from large companies: instead of one massive model that everyone shares, your Second Me is trained on your own memories, runs on your own machine, and is meant to amplify rather than replace you. You feed it information about yourself and it builds an internal model of your identity using two ideas described in the project's research papers: Hierarchical Memory Modeling (a way of organizing memories at different levels of detail) and a process called the Me-Alignment Algorithm, which tries to make the AI's responses match your perspective. Once trained, the AI self can stay local for full privacy, or join a wider Second Me network where, with your permission, other people or apps can interact with it as a digital stand-in for you. Two example use cases ship with it: Roleplay, where your AI takes on different personas, and AI Space, where multiple Second Mes collaborate on ideas. The size of model you can run depends on how much memory your computer has, and Mac users with Apple M-series chips can use a library called MLX to run larger models. You would use Second Me if you want an AI assistant that reflects your own views and history rather than a generic chatbot, and if you care about keeping that data on your own hardware. It is a Python project, you can get started by cloning the repository and running a single Docker command, then opening a web interface in your browser. The full README is longer than what was provided.

Copy-paste prompts

Prompt 1

I want to set up Second Me. Walk me through cloning the repo, running the Docker command to start it, and feeding it my first set of personal memories.

Prompt 2

I have an Apple M3 MacBook. How do I configure Second Me to use MLX so I can run a larger local model than the default?

Prompt 3

I want my Second Me to join the network so colleagues can ask it questions on my behalf. What privacy controls does the project offer and how do I enable public access safely?

Prompt 4

Explain the Hierarchical Memory Modeling approach Second Me uses, what does it store, how is it organized, and how does it differ from a normal RAG setup?

Open on GitHub → Explain another repo

← mindverse on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.