Download

Install Mission LLM on your own infrastructure.

Mission LLM runs as one container on a lab workstation, a school server or a cluster, beside a local model runtime. Pick the method that fits your environment. The commands below come from the project's deployment documentation.

Public release images are not published yet. Build the image from source today, push it to your own registry, and replace REGISTRY in the commands below.

Available today
Server
Docker container
A lab workstation, a server in your rack, or Kubernetes, Helm and OpenShift.
Get started
In development
Windows
Desktop app
A native desktop app for Windows, in development.
In development
macOS
Desktop app
A native desktop app for macOS, in development.
In development
Linux
Desktop app
A native desktop app for Linux, in development.

Until the desktop apps ship, build and run Mission LLM in Docker on your own computer. Checksums and signatures will be published with the first release.

From one container to a cluster.

Install options

Build from source

Build the image yourself from the repository. This is the supported path today.

Terminal

git clone https://github.com/jordan-pesavento/mission-llm.git mission-llm
cd mission-llm
docker build -f docker/Dockerfile -t mission-llm:latest .

Then run it with the Docker command, using mission-llm:latest as the image name. Or build and run with Compose from the repository:

Terminal

touch server/storage/missionllm.db
cd docker
cp .env.example .env
docker-compose up -d --build

Then open http://localhost:3001

The build downloads packages from the internet. For a disconnected site, build on a connected machine and move the image across with your approved transfer process.

Docker

One container and one storage folder. Your data persists across image updates.

Linux or macOS

export STORAGE_LOCATION=$HOME/missionllm && \
mkdir -p $STORAGE_LOCATION && \
touch "$STORAGE_LOCATION/.env" && \
docker run -d --rm -p 3001:3001 \
--cap-add SYS_ADMIN \
-v ${STORAGE_LOCATION}:/app/server/storage \
-v ${STORAGE_LOCATION}/.env:/app/server/.env \
-e STORAGE_DIR="/app/server/storage" \
REGISTRY/mission-llm:latest

Windows (PowerShell)

$env:STORAGE_LOCATION="$HOME\Documents\missionllm"; `
If(!(Test-Path $env:STORAGE_LOCATION)) {New-Item $env:STORAGE_LOCATION -ItemType Directory}; `
If(!(Test-Path "$env:STORAGE_LOCATION\.env")) {New-Item "$env:STORAGE_LOCATION\.env" -ItemType File}; `
docker run -d --rm -p 3001:3001 `
--cap-add SYS_ADMIN `
-v "$env:STORAGE_LOCATION`:/app/server/storage" `
-v "$env:STORAGE_LOCATION\.env:/app/server/.env" `
-e STORAGE_DIR="/app/server/storage" `
REGISTRY/mission-llm:latest;

Then open http://localhost:3001

To turn off telemetry from the first start, add -e DISABLE_TELEMETRY="true" before the image name.
Model runtime on the same machine: in Mission LLM, use http://host.docker.internal:PORT instead of localhost. On Linux, also add --add-host=host.docker.internal:host-gateway to the run command.

Docker Compose

The same container as a managed service that restarts automatically.

docker-compose.yml

services:
  missionllm:
    image: REGISTRY/mission-llm:latest
    container_name: missionllm
    ports:
      - "3001:3001"
    cap_add:
      - SYS_ADMIN
    extra_hosts:
      - "host.docker.internal:host-gateway"
    environment:
      - STORAGE_DIR=/app/server/storage
      - JWT_SECRET=REPLACE_WITH_A_LONG_RANDOM_STRING
      - DISABLE_TELEMETRY=true
      - LLM_PROVIDER=ollama
      - OLLAMA_BASE_PATH=http://host.docker.internal:11434
      - OLLAMA_MODEL_PREF=llama3.1:8b
      - OLLAMA_MODEL_TOKEN_LIMIT=4096
      - EMBEDDING_ENGINE=ollama
      - EMBEDDING_BASE_PATH=http://host.docker.internal:11434
      - EMBEDDING_MODEL_PREF=nomic-embed-text:latest
      - EMBEDDING_MODEL_MAX_CHUNK_LENGTH=8192
      - VECTOR_DB=lancedb
      - WHISPER_PROVIDER=local
      - TTS_PROVIDER=native
      - PASSWORDMINCHAR=8
    volumes:
      - missionllm_storage:/app/server/storage
    restart: always
 
volumes:
  missionllm_storage:
    driver: local
    driver_opts:
      type: none
      o: bind
      device: /path/on/local/disk

Terminal

docker compose up -d

Then open http://localhost:3001

Any setting from docker/.env.example can go in the environment list.

Kubernetes

An example manifest with a persistent volume, deployment, service and ingress.

1
Set image to the image in your registry.
2
Replace the volume placeholder with storage that fits your cluster. The example uses an AWS EBS volume.

Apply it:

Terminal

kubectl apply -f cloud-deployments/k8/manifest.yaml

Helm

A chart with ConfigMap settings, Secret references, persistent storage, and Ingress and Gateway API support. Telemetry is off by default in the chart.

values-secret.yaml

image:
  repository: REGISTRY/mission-llm
  tag: "latest"
 
service:
  type: ClusterIP
  port: 3001
 
envFrom:
  - secretRef:
      name: missionllm-secrets
 
persistentVolume:
  size: 16Gi

Terminal

cd cloud-deployments/helm/charts
helm install my-missionllm ./missionllm -f values-secret.yaml

Keep API keys and other secrets in Kubernetes Secrets. The chart's config values render to a ConfigMap, which is not encrypted.

OpenShift

A dedicated image for restricted security context constraints, arbitrary user IDs and GID 0. The template is community-maintained.

Terminal

docker build -f cloud-deployments/openshift/Dockerfile -t REGISTRY/missionllm:openshift .
docker push REGISTRY/missionllm:openshift
oc new-project missionllm
oc new-app REGISTRY/missionllm:openshift
oc expose svc/missionllm --port=3001
oc set env deployment/missionllm \
  STORAGE_DIR=/app/server/storage \
  JWT_SECRET=$(openssl rand -hex 32)

Cloud templates

AWS CloudFormation: One EC2 instance running the container.
Google Cloud Deployment Manager: One Compute Engine VM.
DigitalOcean Terraform: One Droplet.

Each template pulls the image you name and serves HTTP on port 3001 without TLS. Add TLS and a password before you expose it.

For evaluation

A Hugging Face Spaces Dockerfile runs the image you name in a Space. Keep the Space private until you set a password or turn on multi-user mode.

Without containers

Running directly on Node.js 18 or later with Yarn 1.x is documented for reference only and is not a supported deployment method.

What the server needs.

System requirements

Item	Requirement
Memory	2 GB of RAM or more.
Disk	10 GB minimum recommended. Storage grows with your documents, vectors and models.
CPU architecture	x86-64 (amd64) or ARM64.
Container runtime	Docker. Use 18.03 or later on Windows and macOS, or 20.10 or later on Linux, so the container can reach model runtimes on the host.
Language model	A local runtime such as Ollama or LM Studio, or an account with a cloud provider you approve.
Local model hardware	Depends on the model and runtime you choose.
Kubernetes installs	A cluster with persistent volume support. Helm 3 for the chart.

After you install

Open the app
In a browser, open http://localhost:3001 or your server's address on port 3001.
Choose a model
In onboarding, choose your model provider. Documents are embedded with the built-in embedder and stored in the built-in LanceDB store by default. You can change both in Settings.
Set up access
Choose “Just me” to set a password, or “My team” to turn on multi-user mode for a class or a team, before anyone else can reach the server.
Turn off telemetry
Confirm DISABLE_TELEMETRY is set to "true" in the server environment.
Add your documents
Create a workspace for each course, cohort or team and add its syllabus, readings, manuals or policies.

Install Mission LLM on your own infrastructure.

Platforms

Server

Windows

macOS

Linux

Install options

Build from source

Docker

Docker Compose

Kubernetes

Helm

OpenShift

Cloud templates

For evaluation

Without containers

System requirements

After you install

Open the app

Choose a model

Set up access

Turn off telemetry

Add your documents