How I Built Astra — An Emotional Memory AI Assistant with OpenAI and SQLite

#emotionalintelligence #ai #python #opensource

I recently finished building Astra, my first emotional memory AI assistant — and I wanted to share the project.

What is Astra?

Astra is an emotionally-aware assistant built with OpenAI and SQLite. She remembers what you say — including how you're feeling — and responds with empathy, memory, and continuity.

Astra tracks:

Facts you mention (e.g. birthdays, preferences)
Emotional phrases (e.g. "I'm overwhelmed")
Emotional trends over time (e.g. persistent sadness)
Your last seen, current mood, and context

All memory is stored locally in SQLite. There’s no server. No hosting. No analytics. Just a real-time assistant that learns and remembers.

Tech Stack

Python (core logic, CLI loop)
OpenAI GPT-4o (for responses + memory phrasing)
SQLite (for memory storage and time decay)
Regex + NLP light (for input parsing and normalization)

Why I Built Her

I wanted an assistant that could actually remember me.

Not just reply with static info, but one that could grow with me, reflect, and recognize emotional shifts over time.

I also wanted to challenge myself to build a full AI logic layer in one file. Astra’s current version lives in a single Python file with structured memory, mood inference, and even simulated typing.