The project's README from github.com/QuipNetwork/quip-node-manager. As one markdown file: README.md.
Desktop application for running and monitoring Quip network nodes. Supports Docker and native execution modes on macOS, Linux, and Windows.
macOS / Linux:
curl -fsSL https://gitlab.com/quip.network/quip-node-manager/-/raw/main/scripts/install.sh | sh
Windows (PowerShell):
irm https://gitlab.com/quip.network/quip-node-manager/-/raw/main/scripts/install.ps1 | iex
Download the latest release from the Releases page.
Download the .dmg, open it, and drag the app to /Applications.
The recommended format is AppImage (works on any distro):
chmod +x quip-node-manager-linux-x86_64.AppImage
./quip-node-manager-linux-x86_64.AppImage
A .deb package is also available for Debian/Ubuntu:
sudo dpkg -i quip-node-manager-linux-x86_64.deb
Download the .exe and run it. Windows SmartScreen may show a warning because the binary is not yet code-signed.
Click More info, then Run anyway.
host.docker.internal[dwave] config sectiondocker compose logs -f <node> in a collapsible drawer; switches to data/node.log once the node writes to itdocker-compose v1 is not supported)This repo vendors the compose stack via a git submodule. After cloning, fetch the submodule at the commit locked by this repo:
make fetch-submodules
(Equivalent to git submodule update --init --recursive, or clone with
git clone --recurse-submodules to do this in one step.)
bun install # Install JS dependencies
bun run dev # Launch development build
bun run build # Production build for current platform
cd src-tauri
cargo check # Type-check Rust code
cargo clippy # Lint
The app also supports a terminal UI mode:
quip-node-manager --cli
The terminal UI starts and stops the stack, runs the pre-flight checklist, streams logs, and edits the node configuration. Configuration covers the storage directory, ports, run mode, update channel, CPU and GPU settings, and the D-Wave keys. The dashboard, the health panel, TLS settings, and in-app updates stay in the desktop app.
src/ -- vanilla HTML/CSS/JS with Tauri IPC (withGlobalTauri: true). Dashboard tab embeds the running dashboard container in an iframe.src-tauri/src/ -- Rust + Tauri v2 commands. compose.rs drives the stack via docker compose; stack_assets.rs stages the bundled compose files into ~/quip-data/.vendor/nodes.quip.network/ -- git submodule tracking the upstream Docker Compose setup (miner + validator + dashboard + postgres + caddy)..env generated from settings on every Start.~/quip-data/ (configurable at first boot) holds app settings, runtime config, secrets, native binaries, and the staged compose files.Native mode note (macOS): the node's REST API binds to
127.0.0.1:20100. Docker Desktop's vpnkit forwards container traffic fromhost.docker.internalthrough to the host's loopback, so the REST port is not exposed to the LAN.
Node Manager rewrites ~/quip-data/docker-compose.yml from its embedded copy on every Start and Apply, discarding whatever you edited there. Put changes in ~/quip-data/docker-compose.override.yml instead. Staging never touches that file, and compose merges it over the base.
services:
quip-validator:
environment:
RUST_LOG: debug
Compose loads an override file automatically only when it discovers the compose file itself. Node Manager always passes -f, so it adds the override explicitly when the file exists, after the base file so the override wins.
Two things to watch:
command: replaces the whole list rather than merging it. To change one validator flag, copy the entire command: block and edit the line you want.~/quip-data/docker-compose.yml, not from vendor/nodes.quip.network/. Staging adds flags the vendored file doesn't contain, such as --public-addr.The compose file caps the miner container at 16 GiB, which stops a runaway round from triggering a host-wide out-of-memory kill. On a dedicated miner with more RAM, raise it by setting miner_mem_limit_gb in ~/quip-data/app-settings.json:
{ "miner_mem_limit_gb": 48 }
Set it there rather than in .env, which Node Manager regenerates on every Start. There's no UI control for this yet.
Validator pruning isn't recommended. The validator runs
--state-pruning=archive --blocks-pruning=archive, keeping every state trie and block body from genesis. Size the disk for that. Pruning reclaims disk, but it breaks the dashboard: the descriptor worker scans from genesis and fails withState already discardedonce it reads past the pruning window. Pruning also likely reduces your point awards on the SNAG platform. Only prune if you accept losing the dashboard.
See AGENTS.md for detailed architecture documentation.
Copyright (c) Postquant Labs