beclab/Olares

Olares – a self‑hosted personal‑cloud OS for AI agents

What it is – Olares is an open‑source platform that turns your own machines into a private cloud for running AI agents and large language models. It bundles a Kubernetes‑based stack (K3s, Kubesphere, etc.) with a set of ready‑made system apps (Files, Vault, Market, Dashboard, Control Hub) and a web UI that you can reach from any browser. The goal is to give individuals or small teams the same convenience of a public‑cloud AI service while keeping all data and compute on hardware you control.

Core components

Component	Role
olaresd (daemon)	Core system process that orchestrates services, storage, networking and GPU resources.
olares‑cli	Command‑line tool with Agent Skills that lets AI agents (e.g., Lares, OpenClaw) issue commands to the platform.
Apps directory	Built‑in SaaS‑style applications (Files, Vault, Market, etc.) that run as containers on the cluster.
Infrastructure layer	Handles compute (GPU pooling, time‑/memory‑slicing), storage (JuiceFS, MinIO, Seafile), and networking (Envoy, Tailscale/HeadScale VPN).
Platform layer	Cloud‑native services such as databases, message queues, and operators that support the higher‑level apps.

Key features (as described in the README)

One‑click local AI – Install open‑source AI apps and models from the built‑in Olares Market with a single click.
GPU pooling & slicing – Share accelerators across nodes, with options for exclusive, time‑sliced, or memory‑sliced usage.
Unified file & storage management – Access local disks, synced folders, cloud buckets, SMB/NFS shares via the Files app; configurable backups are built‑in.
Private networking & access controls – Automatic VPN, reverse‑proxy, and HTTPS endpoints; no need to expose individual ports manually.
Any‑time, any‑where access – Web UI works on desktop, phone, or tablet.
System apps suite – Files, Vault (secret store), Market (app catalog), Dashboard, Control Hub, plus AI‑focused agents like Lares.
Agent‑driven operations – AI agents can control the platform through the CLI’s Agent Skills (e.g., “Check this device's configuration”).

Getting started

Hardware requirements – 4 CPU cores, ≥8 GB RAM, ≥150 GB SSD (HDD not supported). A GPU is optional but enables local model acceleration.
Installation – Run the one‑liner on Ubuntu 22.04‑25.04 or Debian 12/13:
```
curl -fsSL https://olares.sh | bash -
```
– Alternatively use the ISO image for a fresh install or Docker‑Compose for a container‑only deployment.
First AI task – Follow the AI onboarding guide to set up the Router, the Lares assistant, and a local model, then start a conversation (e.g., “Check this device's configuration”).

Typical use cases

Personal AI assistant for research, file handling, and automation.
Running generative AI locally (chat, image/video generation) without sending data to third‑party services.
Home‑automation and media streaming from a self‑hosted hub.
Developing, testing, and deploying agentic applications in isolated environments.
Local audio processing: transcription, translation, text‑to‑speech.
Collaborative self‑hosted workspace for families or small teams (documents, project management, communication).
Centralised personal data store with sync, backup, and secure access.

Architecture overview

Olares mirrors the public‑cloud stack (IaaS → PaaS → SaaS) with open‑source equivalents:

IaaS layer – K3s/Kubesphere provides the Kubernetes cluster that abstracts compute, storage, and networking.
PaaS layer – Framework services (databases, message queues, operators) run on the cluster to support higher‑level apps.
SaaS layer – System apps (Files, Vault, Market, Dashboard) expose user‑facing functionality via a web UI and APIs. A diagram in the repo visualises how components such as JuiceFS, MinIO, Envoy, Tailscale, and LangChain fit into these layers.

Community & contribution

Discord & Forum – Primary places for support, deployment tips, and workflow discussions.
GitHub – Issues for bugs/feature requests, PRs for core code, docs, or new market apps.
CLI Agent Skills – Developers can extend the olares-cli so AI agents can manage the platform programmatically.
License – AGPL‑3.0, encouraging open collaboration while protecting downstream users.

TL;DR

Olares is a Kubernetes‑powered personal cloud OS that lets you run AI agents and LLMs on your own hardware. It provides a unified UI, GPU pooling, private networking, and a marketplace of ready‑to‑install AI apps, making self‑hosted generative AI accessible to individuals and small teams.