What is StarNion?
Table of contents
Overview
StarNion is a fully self-hosted personal AI assistant platform. All data and AI interactions run on your own infrastructure โ no data is ever sent to external servers beyond the LLM API calls you authorize.
It gives you complete control over your personal information and data sovereignty, while retaining the convenience of cloud AI services.
Core Concepts
Personal AI Agent
StarNionโs AI agent is more than a simple chatbot. It is an intelligent agent that processes complex tasks step by step using the Vercel AI SDK v5 with a multi-LLM backend.
Multi-provider LLM : Anthropic Claude, Google Gemini, OpenAI, GLM (Z.AI), Ollama
Skill system : 24+ built-in skills โ finance, diary, goals, wellness, search, and more
Conversation context management with RAG-based memory
Tool Calls : agent executes skill functions autonomously
Personas : custom AI personalities configurable per context
Privacy-first
Your data = stored only on your infrastructure
All conversation history is stored in your own PostgreSQL database
Only the minimum necessary text is sent when making LLM API calls
Files, images, and audio are kept in your own MinIO storage
No third-party analytics or tracking code
Self-hosted
Running on your server = complete control
The entire stack launches with a single starnion dev command
Can be run anywhere โ cloud, on-premises, or a home server
Local AI models supported via Ollama (no internet required)
Full control over data backup and migration
Key Features
Feature
Description
Multi-LLM
Anthropic Claude ยท Gemini ยท OpenAI ยท GLM ยท Ollama
Web UI
Next.js 16 web interface with 24+ feature pages
Telegram Channel
AI chat via Telegram bot
i18n
4-language UI (Korean ยท English ยท Japanese ยท Chinese)
Finance Tracker
Record and query expenses in natural language
Budget Management
Monthly budgets with spending alerts
Diary
AI-assisted journal writing with mood tracking
Goal Management
Set, track, and check in on personal goals
D-Day
Countdown timers for important dates
Memos
Quick notes with tag filtering
Garden & Wellness
Data visualization garden + mood/wellness check-ins
Reports & Statistics
Automated periodic summaries and charts
Personas
Per-context AI personality configuration
Skills Management
Enable/disable AI tool skills per provider
Web Search
AI-powered real-time web search
AI Memory
Semantic (RAG) memory across all your data
File Management
Documents, images, audio upload & analysis
Notification Center
Per-user cron jobs (budget alerts, daily summaries, etc.)
Usage Analytics
LLM token usage and cost tracking
Docker Support
starnion dev or starnion docker up
Real-time Streaming
SSE/gRPC-based response streaming
Architecture
โโโโโโโโโโโโโโโโโโโโโโโโ โโโโโโโโโโโโโโโโโโโโโโโโ
โ Web UI (Next.js) โ โ Telegram Bot โ
โ localhost:3893 โ โ (polling) โ
โโโโโโโโโโโโฌโโโโโโโโโโโโ โโโโโโโโโโโโฌโโโโโโโโโโโโโ
โ โ
โผ โผ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ Go Gateway :8080 โ
โ REST API ยท WebSocket ยท Cron Scheduler โ
โ โ gRPC (streaming) โ
โโโโโโโโโโโโโโโโโโโโผโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โผ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ TypeScript Agent :50051 โ
โ AI SDK v5 ยท Multi-LLM ยท Skills โ
โ Streaming SSE ยท Tool Calls ยท RAG Memory โ
โโโโโโโโโโโโโโโโโโโโฌโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โผ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ PostgreSQL 16 + pgvector (HNSW) โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ
โผ
โโโโโโโโโโโโโโโโโโโ
โ MinIO (S3) โ images ยท audio ยท documents
โโโโโโโโโโโโโโโโโโโ
Component Descriptions
Component
Role
Tech Stack
Web UI
Web interface + Auth
Next.js 16 ยท React 19 ยท TypeScript ยท NextAuth v5
Gateway
REST API ยท WebSocket ยท Telegram ยท Cron
Go 1.22+ ยท Echo v4
Agent
AI engine ยท gRPC server ยท Skill execution
TypeScript ยท AI SDK v5 ยท gRPC
PostgreSQL
Main database ยท vector search
PostgreSQL 16 + pgvector
MinIO
File storage
MinIO (S3-compatible)
CLI
Service management ยท Setup wizard
Go
Data Flow
User Message
โ
โผ
Web UI (Next.js) or Telegram Bot
โ HTTP / WebSocket
โผ
Gateway (Go)
โ gRPC streaming
โผ
Agent (TypeScript / AI SDK v5)
โ
โโโโถ LLM API (Anthropic / Gemini / OpenAI / Ollama)
โโโโถ Skill execution (finance, diary, goals, search, ...)
โโโโถ PostgreSQL (conversation storage / vector search)
โโโโถ MinIO (file access)
โ
โผ
Streaming response โ Gateway โ UI
Tech Stack
Backend
Technology
Version
Purpose
Go
1.22+
Gateway API server + CLI
TypeScript / Node.js
20+
AI agent engine
Vercel AI SDK
v5
LLM communication and streaming
gRPC
-
Gateway โ Agent communication
PostgreSQL
16+
Database
pgvector
-
Vector embeddings / semantic search
MinIO
latest
S3-compatible file storage
robfig/cron
v3
Per-user cron scheduler
Frontend
Technology
Version
Purpose
Next.js
16
Web framework
React
19
UI library
TypeScript
5+
Type safety
NextAuth
v5
Authentication
Tailwind CSS
4
Styling
shadcn/ui
-
UI component library
next-intl
4
Internationalization (ko/en/ja/zh)
Infrastructure
Technology
Version
Purpose
Docker
24+
Containerization
Docker Compose
v2
Orchestration
GoReleaser
v2
CLI release automation
Who is it for?
Individual Users
Those who want full control over their AI conversations without entrusting them to cloud services
Those who want to manage personal journals, memos, goals, and finances together with AI
Those who want to access their personal AI assistant anywhere via Telegram
Developers / Technical Users
Developers who want to build and extend their own AI platform
Those who want to add custom functionality through the Skill system
Those studying AI agent architecture
Not the Right Fit
Those who want an immediately usable cloud service with no server management โ Recommend ChatGPT, Claude.ai, etc.
Cases requiring hundreds of concurrent users โ Additional scaling work will be needed
Next Steps