Skip to content

jas77409/vxneo

Repository files navigation

Neo — Personal Companion Intelligence

License: MIT Platform Status Android

Neo is an open-source personal companion intelligence platform that builds a living memory of your life — your goals, habits, relationships, health, and work. Unlike generic AI chatbots, Neo gets smarter about you with every conversation.

Live at vxneo.com · Android Beta · VXNeo Labs


What Makes Neo Different

  • 9-Dimensional Memory — Goals, habits, health, relationships, work, events, emotions, preferences, and context stored in a Neo4j graph that grows with every conversation
  • Fully Private — Self-hostable on your own server. Your data never leaves your infrastructure
  • Multi-Model AI — Switch between Claude, DeepSeek, Qwen, Mistral, Kimi, and Llama on the fly
  • Proactive Intelligence — Neo surfaces reminders, notices patterns, and reaches out when something matters
  • Mobile + Web + Telegram — Android app live, iOS coming soon, full Telegram bot interface

Architecture

Neo Platform ├── companion-api (FastAPI, port 8000) Core AI agent + memory graph ├── vxneo-platform (FastAPI, port 3000) Auth, billing, push notifications ├── Neo4j Memory graph database ├── Qdrant Vector similarity search ├── Redis Session + FCM token storage └── Telegram Bot Conversational interface


Quick Start (Self-Host)

Prerequisites

  • Docker + Docker Compose
  • Python 3.12+
  • Anthropic API key (or any supported model key)

Setup

git clone https://github.com/jas77409/vxneo.git
cd vxneo
cp .env.example .env
nano .env
docker compose up -d
uvicorn api:app --host 0.0.0.0 --port 8000

Environment Variables

# Required
ANTHROPIC_API_KEY=your_key

# Open models (optional)
HF_API_KEY=your_huggingface_key

# Integrations (optional)
TELEGRAM_BOT_TOKEN=your_token
NOTION_TOKEN=your_token
GITHUB_TOKEN=your_token

# Infrastructure
REDIS_URL=redis://localhost:6379
NEO4J_URI=bolt://localhost:7687
NEO4J_PASSWORD=your_password
QDRANT_URL=http://localhost:6333

Memory Architecture

Neo uses a 9-dimensional memory graph stored in Neo4j with vector search via Qdrant:

Dimension Description
Goals Short and long-term objectives
Habits Recurring patterns and daily routines
Health Physical and mental wellbeing
Relationships People, organisations, and connections
Work Projects, tasks, and professional context
Events Calendar, deadlines, and appointments
Emotions Emotional state and patterns over time
Preferences Likes, dislikes, and communication style
Context Environmental and situational awareness

Each memory node has a salience score that decays over time and is boosted by repeated references — so the most relevant memories always surface first.


Multi-Model Support

Neo routes conversations to the best available model based on task type:

Key Model Provider
claude Claude Sonnet (default) Anthropic
deepseek DeepSeek R1 HuggingFace
qwen Qwen 2.5 HuggingFace
kimi Kimi HuggingFace
mistral Mistral 7B HuggingFace
llama Llama 3.3 70B HuggingFace

Switch models via Telegram with /model or via the API at POST /models/set.


Integrations

Integration Status Capabilities
Notion ✅ Live Read, write, search pages and databases
Microsoft Calendar ✅ Live Event awareness, upcoming appointments
Google Calendar ✅ Live Upcoming events, reminders
GitHub ✅ Live Repo monitoring, issue management, shell tools
Telegram ✅ Live Full conversational interface, /model command
Firebase FCM ✅ Live Push notifications (Android)
Stripe ✅ Live Subscription billing (CAD $5/month)
Trello 🔄 Soon Task and board management

API Endpoints

Auth

Method Endpoint Description
POST /auth/signup Create account
POST /auth/login Login
GET /me Get profile

Chat

Method Endpoint Description
POST /neo/ask Send message

Memory

Method Endpoint Description
GET /memory/recent Get recent memories
GET /memory/search?q= Search memories
POST /memory/add Add memory

Models

Method Endpoint Description
GET /models List available models
POST /models/set Switch active model

Vault

Method Endpoint Description
GET /vault List vault entries
POST /vault/add Add vault entry

Notifications

Method Endpoint Description
POST /device/register Register FCM token
POST /notify Send push notification

Mobile App

The Neo Android app is built with Flutter.

Features

  • Login / signup
  • Chat with Neo with typing indicator
  • Model switcher — tap model badge in header
  • Memory tab — search and browse your memory graph
  • Vault tab — secure notes
  • Profile tab — tier, settings, sign out
  • Push notifications via Firebase FCM

Roadmap

  • Core memory graph (Neo4j + Qdrant)
  • Multi-model routing (6 models)
  • Telegram bot interface
  • Android mobile app (Flutter)
  • Firebase push notifications
  • Notion integration
  • GitHub tools + shell access
  • Microsoft + Google Calendar
  • Stripe billing
  • ULANZI pixel desk display service
  • Trello integration
  • iOS app
  • One-click Docker deploy script
  • Public API documentation site
  • Plugin system for custom integrations

Contributing

We welcome contributions of all kinds — from fixing typos to building new integrations.

git clone https://github.com/jas77409/vxneo.git
cd vxneo
pip install -r requirements.txt
cp .env.example .env
python3 agent/agent.py

Please open an issue before submitting a large PR. See CONTRIBUTING.md for full guidelines.

Good First Issues

  • Add a new memory dimension
  • Improve salience scoring algorithm
  • Add WhatsApp integration
  • Write tests for memory_manager.py
  • Improve self-hosting documentation

Active Deployments

Deployment Location Description
vxneo.com Hetzner, Helsinki Hosted platform
PROAS Persönliche Assistenz Vienna, Austria AI companion for disability support services

Self-Hosting Infrastructure

Neo runs on a single VPS comfortably:

  • Recommended: 4GB RAM, 2 vCPU (Hetzner CPX22 or equivalent)
  • Services: Neo4j, Qdrant, Redis via Docker Compose
  • Reverse proxy: nginx with Let's Encrypt SSL
  • Monitoring: UptimeRobot + daily backup cron

License

MIT License — see LICENSE

Free to use, modify, and self-host. Commercial use permitted.


Built By

VXNeo Labs Inc. Ontario, Canada

vxneolabs.com · contact@vxneolabs.com · X @vxneolabs


Neo is open source software. Your data is yours. We build in public.

About

Neo — Personal Companion Intelligence. Open-source AI companion with 9-dimensional memory, 6 mentor modes, and self-improving loops.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

No contributors