Deploying a container

While Launchpad is built specifically for AI workloads, it supports the deployment of any containerized image.

To start, log in to the dashboard, select a project, and click on the Launchpad menu on the sidebar, then on the Create container button.

Container size

Start by selecting the size of the container you want to deploy. We provision containers directly on bare metal, which means the resources of your containers are dedicated to you and can't be utilized by other users.

Container size selection on the Latitude.sh dashboard

Container size selection on the Latitude.sh dashboard

Image setup

Next, select the image that you want to deploy. It is not the purpose of this guide to delve into the concept of an image, but it is important to understand that the Docker image is a package that contains all the necessary definitions to run a specific application. In other words, it defines the operating system to be used, the software to be installed, and the necessary commands for execution, among others.

The image must be provided its name, followed by the version, and separated by :.

For example, nginx:latest or alpine:20231219. The version can be omitted, in which case the "latest" version will be considered.

Note that the registry is also requested in addition to the image. Launchpad supports public and private registries.

If the registry is private, the username and password will be requested. We follow the pattern defined by docker login and adopted by most registries.

If a private registry is added through the creation page, it will be saved for future use.

Deploying Ollama

Deploying Ollama

Container settings

  • Name: Cannot consist only of numbers or contain spaces and special characters.
  • Docker command: You can define an optional command that will be executed at the start of the container's execution. There is no need to add any prefixes such as bash -c.
    • Important: This command replaces the entry point defined in the image's Dockerfile.
  • HTTP ports: Separate each port by a comma. Setting an HTTP port creates an ingress for your container in the format `https://-<port>.lsh.ai/
  • TCP ports: These ports become accessible through a public IP provided after the container is provisioned.
  • Environment variables: This can serve various purposes, depending on the use case. It is common for sensitive data to be used as environment variables. Use the lock icon to turn the environment variable into a secret. The value will be hidden.

Monitoring the deployment

You can follow the deployment process from the logs provided in the UI. This is useful to see if your image was downloaded correctly and is working as expected.