Choose a Deployment#

中文｜ English ｜ Docs home

Three routes for three kinds of machines. The steps do not overlap, so pick one and follow it through. Each runs from installing uv to auto-start on boot; the code is the same, and the differences come down to system libraries, startup method and permissions.

Your machine	Take this route	What you will run into
Windows 10 / 11 (64-bit)	Windows deployment	You can skip installing Python entirely and let uv fetch it; copy `icudtl.dat` into the Python install dir; open ports 5000 / 6700 in the firewall
Linux server / cloud host	Linux deployment	Install `libfontconfig1`, `libgl1`, `libegl1` and friends; keep it alive with systemd and put Nginx in front
Android phone (Termux)	Android deployment	It runs, but there is no prebuilt `skia-python` wheel for phones so you compile it yourself; not thoroughly validated, test at your own risk

If this is your first time with the project, get /setup working and the assets downloaded on your own machine first, then move to a server or a phone.

Shared by all three routes#

Python version#

3.10 – 3.13 (64-bit), 3.13 recommended. The version and architecture are hard requirements: skia-python, Pillow and similar only publish prebuilt wheels, and a mismatch fails to install outright. 3.14 and newer will not work — the meme engine pins Pillow ^10.0.0, and 10.x predates 3.14 with no matching wheel. With uv you can just let it pull a 3.13 for you instead of wrestling with the system Python.

Manage dependencies with uv#

uv is a standalone executable. Install it one of three ways; after that, every route uses exactly the same commands.

Windows (PowerShell):

PowerShell

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

Linux / macOS:

Shell

curl -LsSf https://astral.sh/uv/install.sh | sh

If neither script works for you (or you would rather keep uv inside Python), fall back to pip:

Shell

pip install uv

On Linux / macOS the script install puts uv outside your PATH — open a new terminal, or run:

Shell

source $HOME/.local/bin/env

Confirm it is there:

Shell

uv --version

Creating the environment and installing are just these two lines, identical everywhere:

Shell

uv venv --python 3.13
uv pip install -r requirements.txt

The dependency list lives in requirements.txt at the repository root. The meme engine under vendor/meme-generator-main/ is not installed through pip itself, but its own dependencies cannot be resolved automatically (there is no pyproject.toml inside vendor/), so the list already covers them — the single command above is all you need.

Configuration and first run#

Every .env option has a default, so you can leave the file alone and configure things in the browser instead. Start the service from the project root — cache, temp and font are resolved relatively — then open http://127.0.0.1:5000/setup and let the wizard handle the environment check, asset download and basic config in one pass.

Ports#

5000 is the web workbench, 6700 is the OneBot V11 reverse WS. NapCat has to be able to reach both, so remember to open them in the firewall or security group on a server.

Connect NapCat#

A running service is only the first step. In NapCat's (NC) network settings, add a reverse WebSocket entry pointing at ws://<ILBB machine IP>:6700/onebot/v11/ws, set the message format to array, keep the Token aligned with WS_ACCESS_TOKEN in .env (leave it empty if you never set one), and open port 6700 in the firewall — only then does the bot receive messages in QQ. Full steps and connection troubleshooting are in Platform Integration.

Where to go next#

Once the bot is answering in QQ, the Command Manual covers day-to-day usage; Configuration explains what every .env option does; and the Plugin Development Guide is where to start if you want to add features.