Sampler Hub

InstallSampler Hub

Sampler Hub is a small self-hosted server that keeps your Opera devices in sync. Install it on a computer you own, pair your phone, desktop and browser, and the hub handles the rest:

• Sync: conversations, settings and AI providers move between your devices.
• Profiles: one hub for a home or flat share. Each profile is sealed end to end; the hub stores it but cannot read it.
• Shared AI: Ollama, LM Studio and llama.cpp on the hub’s computer can be used from every paired device, even outside your home network.
• Local models: the hub can download and run llama.cpp models itself, choosing the GPU backend for the machine.

There is no account and no Opera server in the middle.

Install

Windows

Download sampler-hub-setup-windows-x86_64.exe from the latest release and run it. It installs for your user in %LOCALAPPDATA%\MaestroAI-Company\Sampler-Hub and starts with your session. A tray icon shows the pairing code and offers updates.

macOS (Apple silicon)

Download sampler-hub-macos-aarch64.dmg from the latest release, or install from a terminal:

curl -fsSL https://github.com/MaestroAI-Company/sampler-hub/releases/latest/download/install.sh | sh

Linux (x86_64 and arm64)

curl -fsSL https://github.com/MaestroAI-Company/sampler-hub/releases/latest/download/install.sh | sh

The installer verifies every file against the release SHA256SUMS, puts sampler-hub in ~/.local/bin (change it with SAMPLER_HUB_DIR), and registers a systemd user service (or a launchd agent on macOS) so the hub starts with your session.

Pair a device

On the hub’s computer:

sampler-hub pair

The command prints a QR code and an 8-character code. Both are valid for a single pairing.

• Phone: scan the QR code from Settings → Sampler Hub → Pair a hub.
• Desktop or browser: type the code in the same place.

On the same network, Opera also finds the hub under Nearby hubs. You still have to enter the code; being on the network grants nothing.

Profiles

A paired device picks a profile:

• New person: create a profile with a name and password. The first profile manages the hub; later profiles are members.
• Existing profile: pick it and enter its password, for example after reinstalling Opera.
• Another device of yours: choose Add one of my devices, then compare the six digits shown on both screens. The key moves between devices without a password.

The profile key is created on the device and never reaches the hub in clear. The hub keeps a copy locked by the password, and five wrong passwords lock the profile for a growing wait. A forgotten password with no device left cannot be reset.

Manage the hub from Opera

Admin profiles can manage the hub from Opera’s hub page:

• choose what each device syncs;
• rename or remove paired devices;
• turn local network discovery and provider sharing on or off;
• install hub updates;
• download, run and delete llama.cpp models.

Devices not seen for 30 days are marked inactive. Reinstalling Opera on the same device replaces its old entry.

Local models with llama.cpp

The first model you download from Opera makes the hub install the official llama-server build for the machine: CUDA, Vulkan or Metal when available, CPU otherwise. Every download is checked against the digest published on GitHub. One model is loaded at a time and served to all paired devices as the llama.cpp provider.

Data directories

Build from source

You need a recent stable Rust toolchain.

cargo build --release -p sampler-hub
cargo build --release -p sampler-hub --no-default-features
cargo test --workspace

The second command builds the headless hub without the tray app. See the Sampler Hub CLI reference for commands and service management.