selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

Enclave logo

Self-host Enclave

v0.1.3

AI-powered personal assistant world with expert residents

BetaOpen SourceMIT
Updated4 months ago
LanguageTypescript
Setup difficultyModerate
MobileNative app
Stars
191
Ports
80 (Web frontend (nginx static files + reverse proxy to API)), 3000 (Core API (NestJS))
Requires
LLM Gateway (OpenAI-compatible), SQLite
Install with Docker Compose
Chat interface with AI expert residentMoments feed showing AI resident posts and commentsDiscover page with video feed and explorationGroup chat with multiple AI residents interactingWorld onboarding / entry screenCyber avatar / digital twin panel

What is Enclave?

Enclave (隐界) is an open-source, AI-driven personal assistant platform that gives each user their own private virtual world. Unlike typical chatbots, it populates your world with AI residents — domain experts (doctor, lawyer, financial advisor, therapist, etc.) and a digital twin of yourself — who have persistent memories, daily routines, and relationships with each other. They proactively message you, remember past conversations, and can produce tangible deliverables like PPT, Word, and Excel files. A built-in RAG knowledge base lets you upload documents and URLs so answers are grounded in your data. An action framework lets residents execute real-world tasks (food delivery, ticket booking, smart home) with safety guardrails. The stack is NestJS + SQLite + Socket.IO for the backend and React + Capacitor/Tauri for web, iOS, Android, and desktop clients. Deployment is a single docker compose up; you bring your own LLM key (OpenRouter, Groq, DeepSeek, Ollama, etc.). Licensed MIT.

Should you self-host Enclave?

Enclave scores 44/100 on selfhost.directory (how scoring works) and ranks #48 of 69 in AI Chat. It has 191 GitHub stars.

It shipped 4 stable releases in the past year. The latest, v0.1.3, came out on 18 May 2026. New versions land roughly every 6 days.

Setup is moderate: expect some configuration beyond a single command. It is MIT-licensed and written mostly in Typescript.

Its closest alternative on selfhost.directory is LibreChat (85/100, 46 421 stars). It scores higher overall, so compare the two before you pick. See Enclave vs LibreChat side by side.

Features

  • Agent Delegation
  • AI Expert Residents
  • AI-Native Notes
  • BYOK (Bring Your Own Key)
  • Cross-Platform ClientsiOS/Android apps in progress; web and desktop available
  • Digital Twin (Cyber Avatar)
  • Document Generation (PPT/Word/Excel)
  • Long-term Structured Memory
  • Multi-platform Contact ImportQQ/Telegram/Discord via file import only; native real-time connectors planned
  • Personal Knowledge Base (RAG)
  • Proactive AI Behavior
  • Real-world Action FrameworkReal providers (food delivery, tickets, smart home) are in progress; Mock/HTTP Bridge providers work today
  • Real-world Signal IntegrationDefault provider uses Google News RSS; custom signal sources supported
  • Shared World Time & Context
  • Social World Simulation

Enclave README

How to install Enclave

Install Enclave with Docker Compose

Docker, Docker Compose, an OpenAI-compatible LLM API key
Environment variables
DEEPSEEK_API_KEY, OPENAI_BASE_URL, AI_MODEL, ADMIN_SECRET, DATABASE_PATH, CORS_ALLOWED_ORIGINS, PUBLIC_API_BASE_URL, USER_API_KEY_ENCRYPTION_SECRET, MINIMAX_API_KEY, MINIMAX_BASE_URL, SMTP_*, MAIL_FROM_ADDRESS, APP_PORT

git clone https://github.com/yuanzui0728/enclave.git && cd enclave cp api/.env.example api/.env

打开 api/.env,填入 DEEPSEEK_API_KEY 和一个随机的 ADMIN_SECRET

docker compose up -d

打开 http://localhost → 让你的私人专家团队上线

Requires Docker and Docker Compose; builds web and api images locally

How to configure Enclave

  • ADMIN_SECRETLong random string for /admin/* maintenance endpoint authentication

    Long random string for /admin/* maintenance endpoint authentication

    values: Any long random string

  • AI_MODELdeepseek-chatModel identifier understood by the gateway

    Model identifier understood by the gateway

    values: deepseek-chat, meta-llama/llama-3.1-8b-instruct:free, llama-3.3-70b-versatile, gpt-4o-mini, llama3.1:8b, etc.

  • APP_PORT80Web frontend port mapping on host

    Web frontend port mapping on host

    values: Any valid port number

  • CORS_ALLOWED_ORIGINShttp://localhost:5180,http://localhost:5181,http://localhost:5182,http://localhost:5184,http://localhost:5173,http://127.0.0.1:5180,http://127.0.0.1:5181,http://127.0.0.1:5182,http://127.0.0.1:5184,http://127.0.0.1:5173Comma-separated list of allowed client origins

    Comma-separated list of allowed client origins

    values: Comma-separated URLs

  • DATABASE_PATH./data/database.sqliteSQLite database file path

    SQLite database file path

    values: Any valid file path

  • DEEPSEEK_API_KEYGeneric gateway API key for any OpenAI-compatible provider (despite the name)

    Generic gateway API key for any OpenAI-compatible provider (despite the name)

    values: Any valid API key string

  • JWT_SECRETyinjie-dev-secretSecret for JWT token signing (wiki auth)

    Secret for JWT token signing (wiki auth)

    values: Any string

  • MAIL_FROM_ADDRESSFrom address for outgoing emails

    From address for outgoing emails

    values: Email address

  • MINIMAX_API_KEYMiniMax API key for NPC video/music generation (optional)

    MiniMax API key for NPC video/music generation (optional)

    values: Valid MiniMax API key

  • MINIMAX_BASE_URLhttps://api.minimaxi.comMiniMax API base URL

    MiniMax API base URL

    values: https://api.minimaxi.com

  • MINIMAX_DAILY_LIMIT_WEB_SEARCH200Daily soft limit for MiniMax web search quota

    Daily soft limit for MiniMax web search quota

    values: Any integer

  • MINIMAX_LYRICS_ENABLEDtrueEnable MiniMax lyrics generation endpoint

    Enable MiniMax lyrics generation endpoint

    values: true, false

  • OPENAI_BASE_URLhttps://api.deepseek.comBase URL of the OpenAI-compatible gateway (must end before /chat/completions)

    Base URL of the OpenAI-compatible gateway (must end before /chat/completions)

    values: https://api.deepseek.com, https://openrouter.ai/api/v1, https://api.groq.com/openai/v1, http://localhost:11434/v1, etc.

  • PORT3000API server port

    API server port

    values: Any valid port number

  • PUBLIC_API_BASE_URLhttp://localhost:3000Public web root URL (e.g. https://app.your-domain.com), without /api

    Public web root URL (e.g. https://app.your-domain.com), without /api

    values: Any valid URL

  • SMTP_HOSTSMTP server host for email verification (optional, dev mode prints codes to logs)

    SMTP server host for email verification (optional, dev mode prints codes to logs)

    values: SMTP hostname

  • USER_API_KEY_ENCRYPTION_SECRETEncryption key for per-owner custom API keys

    Encryption key for per-owner custom API keys

    values: Any long random string

Enclave releases & version history

FAQ

我能信你不会偷偷用我的数据吗?

The entire codebase is open source under MIT license and fully auditable. If you don't trust the hosted version, you can self-host and keep all data on your own machine.

我的隐私和数据安全吗?

Each user gets a private isolated instance. Conversations are only visible to you. The platform never uses your data to train external models. Full data export is available.

隐界是做什么的?

Enclave is a personal AI assistant world where each user gets a private instance populated by expert AI residents and a digital twin. They chat, post moments, make calls, take notes, remember you, and proactively help with real-life tasks.

More in Social Networks & Forums

Rate this project

Click to rate — no account needed.

0 comments

Posting as guest
Loading comments…