Lemonade environment (Ubuntu 24.04)

This page installs Lemonade Server on Ubuntu 24.04: OpenAI-compatible API plus Web UI. GPU uses llama.cpp; NPU uses FastFlowLM. Model pull lives in 01-Deploy.

Official docs: Ubuntu install · Linux NPU / FLM

Driver and IOMMU details: FastFlowLM environment. Lemonade only calls an already-installed flm.

---

Validated baseline (2026-09-16)

Item	Value
Package	lemonade-server 11.9.0~24.04
CLI	lemonade version 11.9.0
Server	/usr/bin/lemond, default 13305
OpenAI	http://127.0.0.1:13305/api/v1 (some clients also try /v1)
GPU backend	llamacpp:vulkan (on kernel 6.17, llamacpp:rocm reported Linux kernel missing support)
NPU backend	this tutorial recommends native FastFlowLM on 8219 (see below)
systemd	lemond.service, working directory /var/lib/lemonade

---

1. Install the server

sudo add-apt-repository -y ppa:lemonade-team/stable
sudo apt update
sudo apt install -y lemonade-server

lemonade --version
lemonade status

You should see lemonade version 11.9.0 (or newer). When the server is up: Server is running on port 13305. Open http://127.0.0.1:13305.

APT usually enables lemond.service (user lemonade, ProtectHome=yes). To stop:

sudo systemctl stop lemond
sudo systemctl disable --now lemond   # optional: no autostart

⚠️ The unit name is lemond, not lemonade-server.

---

2. Install inference backends

lemonade backends
lemonade backends install llamacpp:vulkan   # skip if already installed
lemonade backends

llamacpp:rocm needs a kernel that meets gfx1151 Linux. Otherwise use Vulkan.

About flm:npu:

• lemonade backends install flm:npu downloads FLM from GitHub and often times out in mainland China.
• Even with a system flm, lemond.service runs as lemonade with ProtectHome=yes, so it cannot read ~/.config/flm.
• This tutorial therefore runs NPU via FastFlowLM + flm serve on 8219.
• If you run lemond in the foreground as your login user (no ProtectHome) and successfully install flm:npu, you can try lemonade load <flm-tag>.

Do not source /opt/rocm on Lemonade’s NPU path.

---

3. Health check

lemonade --version
lemonade status
lemonade backends
ss -tlnp | grep 13305

Optional: bash src/npu/scripts/check_lemonade.sh. For NPU also run bash src/npu/scripts/check_flm_npu.sh after unsetting ROCm paths.

---

4. Common commands

Command	Role
lemonade backends	installed / available backends
lemonade status	server on 13305, loaded model
lemonade list	catalog; --downloaded for local only
lemonade pull NAME	download (--source modelscope in China)
lemonade load NAME	load into the server
lemonade unload	unload
lemonade run NAME	load + open Web UI
lemonade chat	terminal REPL

GPU backend pin:

lemonade load Gemma-4-E4B-it-GGUF --llamacpp vulkan

---

5. Ports in this tutorial

Port	Use
13305	Lemonade Server
8219	native flm serve (exclusive with Lemonade NPU)

Next: Gemma 4 Lemonade GPU · Gemma 4 Lemonade NPU · NPU troubleshooting