Install LLiMa on Modalix Early Access Kit

The flagship LLiMa demo showcases multimodal generative AI capabilities and can be installed on the Modalix DevKit with a single command.

For the Modalix Early Access Kit (EA) board, which lacks an NVMe drive, installation typically relies on NFS or an SD card. This guide walks you through the setup steps for the EA board.

Prerequisites

  • Ensure sima-cli is installed on the board.

  • Confirm the board has an active Internet connection.

  • You must have a valid Developer Portal account to download assets from docs.sima.ai.

Updating Firmware

The EA board may come pre-installed with firmware 1.6.

To upgrade to the latest 1.7.0 release, you can run sima-cli from the board

modalix:~$ sima-cli login
modalix:~$ sima-cli update 1.7.0

Installing via SD Card

The SD card shipped with the EA is preformatted, making installation straightforward:

modalix:~$ mkdir -p /media/card
modalix:~$ sudo mount /dev/mmcblk1p1 /media/card
modalix:~$ sudo chown sima /media/card
modalix:~$ cd /media/card && mkdir llima && cd llima
modalix:~$ sima-cli install -v 1.7.0 samples/llima

If you are using your own SD card, format it on a Ubuntu host for use with Modalix:

user@ubuntu-host-machine:~$ sima-cli sdcard format

Installing via NFS

While SD cards work, model loading is slower. Using a NFS provides a significant speed boost.

Important

NFS requires a more complex setup. For instructions on configuring NFS, refer to this guide.

Step 1 — Install on NFS Host
user@ubuntu-host-machine:~$ pip3 install sima-cli
user@ubuntu-host-machine:~$ cd /mnt/nfs_share && mkdir llima && cd llima
user@ubuntu-host-machine:~$ sima-cli install -v 1.7.0 samples/llima

Note

The process will fail after downloading assets when attempting to install python dependencies. This is expected — proceed to Step 2.

Step 2 — Install on the Modalix EA Board

Navigate to the NFS mount directory on the DVT board:

cd /mnt/llima/simaai-genai-demo
./install.sh

This script installs platform-specific dependencies required for Modalix. Running this script elsewhere will install x86 packages instead, which will break the setup.

Running the Application

After installation, start the application using one of the following commands:

./run.sh                      # Standard launch
./run.sh -cli                 # Command-line interface
./run.sh --ragfps <serverIP>  # Connect to RAG FPS server

For more information regarding how to setup the RAG FPS (File Processing Server), refer to this article.