Sora-mem is a self-hosted Python Android/Web/Lite app with long-term memory features, using FastAPI, PostgreSQL with pgvector, and a Capacitor-based Android app. It provides end-to-end sources for backend, frontend, Chrome extension, and deployment guidance.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
Sora-mem is a self-hosted, long-term memory AI companion and memory assistant with a full stack including backend FastAPI, frontend React, Capacitor Android app, and a Chrome sidebar extension. It supports multimodal conversation with memory retrieval, memory consolidation, reading mode, automatic diaries, and an artifact canvas. The project is distributed as self-contained frontend and backend code, with deployment references for nginx/systemd and PostgreSQL 16 + pgvector.
How it works
Backend stack: Python 3.11+; FastAPI; PostgreSQL 16 + pgvector; SQLAlchemy 2; APScheduler. Frontend: React 19, Vite, TypeScript. LLM interface is OpenAI-compatible ( SiliconFlow by default) and can be swapped to other OpenAI-compatible providers. The repository describes a long-term memory system using vector retrieval with decay and automatic integration/merging. Interactions include chat with memory-aware context handling and multimodal image understanding. The architecture includes a backend API, a frontend SPA, Android Capacitor app, and a Chrome extension for reading current pages.
Getting started
Local development steps provided for backend, frontend, Chrome extension, and Android app:
Backend
cd backend
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
# Database setup steps
createdb sora_db
psql sora_db -c "CREATE EXTENSION IF NOT EXISTS vector;"
# Create first admin user
PYTHONPATH=. python scripts/create_first_user.py
# Start server
PYTHONPATH=. uvicorn app.main:app --reload --host 0.0.0.0 --port 8000
Frontend
cd frontend
npm install
cp .env.example .env.local
npm run dev
Chrome extension and Android app setup steps are described with file edits and build commands in the README, including loading the extension and building the Android app via Capacitor.
Recent releases
- latest 0: none
Traction
20 stars, 3 forks, 0 open issues
Behind the repo
Not specified in the provided FACTS
Caveats
MIT license; open-source code with self-hosting requirement; users must provide VPS, domain, and SSL as described in deployment notes. The repository structure and environment variable guidance are included in the local dev sections and deployment guide.