tab

// overview

tab is a wearable ai device that acts as a personal assistant and life coach. it records conversations, processes them through ai models, and stores structured data for easy recall and decision-making support.

designed specifically for individuals with adhd — it offers insights and proactive suggestions for personal development and productivity.

• continuous capture — raspberry pi zero w records audio throughout the day
• transcription — whisper large v3 converts speech to text
• semantic indexing — conversations are embedded and stored for search
• ai chat — ask questions about your past conversations
• proactive insights — surfaces relevant memories when you need them

// hardware

wearable unit

• raspberry pi zero w (1ghz cpu, 512mb ram)
• usb microphone
• battery pack

processing server

• um790 pro workstation (amd ryzen 9 7940hs, 32gb ram)
• handles transcription, embedding, and llm inference

// pipeline

the data flow from capture to recall:

1. capture — pi records audio segments and uploads to supabase storage
2. transcribe — serverless function triggers whisper large v3
3. process — mistral extracts key information, entities, topics
4. embed — sentence-transformers/all-MiniLM-L6-v2 generates vectors
5. store — structured data + embeddings saved to supabase (postgres)
6. recall — semantic search over conversations via the web interface

// stack

• frontend — next.js web app with chat interface
• backend — supabase (postgres + auth + storage + serverless functions)
• transcription — whisper large v3
• embeddings — sentence-transformers/all-MiniLM-L6-v2
• llm — mistral (local) + openai gpt-3.5 (fallback)
• firmware — c++ on raspberry pi zero w
• deployment — docker

// setup

backend

cd backend
cp .env.example .env
# configure supabase credentials in .env
npm install
npm run dev

web app

cd app
cp .env.example .env
# configure api endpoints in .env
npm install
npm run dev

raspberry pi

# flash raspbian lite to sd card
# enable ssh and wifi
# install dependencies
sudo apt-get install python3-pip
pip3 install -r requirements.txt

# configure and start recording
cp config.example.json config.json
python3 record.py

full hardware assembly guide and detailed walkthrough on hackster.io.