SecondBrain

SecondBrain gives your AI memory of your real life.

What This Is

SecondBrain is an experimental memory layer for real-world relationships.

It helps you remember:

• who someone is
• where you met
• what mattered in your last conversation
• what you promised to do

Instead of treating interactions as raw transcript logs, SecondBrain stores structured memory episodes so context can be surfaced instantly when it matters.

Why Customers Care

Most tools remember documents and prompts. They do not remember your life.

SecondBrain is built for the moment right before you speak to someone and realize:

• you forgot their name
• you forgot where you met
• you forgot what you owe them

SecondBrain makes that moment recoverable by turning interactions into memory that can be searched and used in real time.

Core Product Experience

1) Recognition + retrieval

During a live session, the system identifies (or suggests) who is in front of you and immediately loads context.

2) Memory capture

Conversation text is transformed into structured fields:

• people involved
• key topics
• explicit promises
• actionable next steps

3) Proactive nudge

You get short context before asking:

• name
• where met
• one key detail
• one open loop

4) Dashboard oracle

Use dashboard + chat to ask questions like:

• what did I promise them?
• who did I meet this week?
• what follow-ups are still open?

How Recognition Works (Vectorizing + Centroid + Top-K)

Face tracking and matching is implemented as a practical experiment pipeline:

1. Face is detected from webcam frames.
2. The face is vectorized into an embedding.
3. Stored identity vectors are grouped by person.
4. A centroid vector is used as a fast representative for each person.
5. Distances are computed and top-K candidates are returned.
6. Confidence gating determines auto-resolve vs manual selection fallback.

This gives a resilient demo path:

• fast best-guess identity
• candidate fallback when uncertain
• quick manual correction loop

Local Storage

SecondBrain is local-first in this repo:

• local SQLite persistence for people + episodes
• local profile/session artifacts
• local development environment for rapid iteration

This keeps iteration fast and demo behavior deterministic.

Privacy + Scope

This repo is a hackathon experiment, not a production compliance product.

• Privacy/compliance guarantees are out of scope in this build.
• This is experimentation of thought and product direction.
• Do not treat this repository as a finalized privacy architecture.

Long-Term Vision

In the long term, SecondBrain could become a wearable context layer that supports people with memory loss by helping reconnect names, faces, and personal history in real time.

Think: lightweight Meta Glasses-style attachment that adds context for people you have met and indexes people you meet next.

Architecture Overview

Pipeline:

camera + mic -> identity resolution -> person_id -> memory retrieval -> context nudge -> transcript extraction -> memory update

Key boundaries:

• Nia: memory/retrieval layer
• Recognition system: identity resolution
• Dashboard + live UI: user-facing interaction surface

Tech Notes

• Next.js + React frontend (frontend/)
• API routes under frontend/app/api/*
• Local DB and memory utilities under frontend/lib/*
• Dashboard routes in frontend/app/dashboard/*

Quickstart

cd frontend
npm install
npm run dev

Open http://localhost:3000.

Environment

Set frontend/.env.local as needed:

• OPENAI_API_KEY
• NIA_API_KEY
• NIA_BASE_URL

Optional:

• GOOGLE_CLIENT_ID
• GOOGLE_CLIENT_SECRET
• GOOGLE_REDIRECT_URI
• TAVILY_API_KEY
• APOLLO_API_KEY
• OPENAI_MEMORY_MODEL

Scripts

From frontend/:

• npm run dev
• npm run build
• npm run start
• npm run lint

License

MIT. See LICENSE.