Deploy a custom AI Agent

In this guide, we'll cover the usage of a custom deployment script to handle any special use case you might have, including (but not limited to):

• Customizing the runtime before the deployment
• Installing and connecting additional programs to your software (multi-agents architecture, in-memory or vector database...)
• Decentralizing the deployment of your agent deployed with another framework (not limited to Python, anything works)

So let's get started 🚀

⚙ Requirements

You just need to be familiar with shell scripting and Docker.

❓Understanding the deployment process

Let's first get a quick overview of the general deployment process so you can understand how to customize it.

LibertAI's agents deployment on Aleph Cloud instances is centered around a simple shell script that you can find in our repository.
Every time a deployment is triggered, a ZIP archive of the code will be sent on the instance, along with this script that has 3 main purposes:

• Making sure all the necessary tools (mainly Docker) are installed
• If an agent is already running, stopping it and cleaning its data properly
• Start the new version of the agent

This is done by receiving 3 information regarding the agent in the script's arguments:

• The Python version used in the project (to choose the right docker image)
• The package manager (to choose the right Dockerfile that will perform optimizations)
• The way the agent is triggered (to use the right entrypoint for the docker container)

✍ Customizing it

While it is great for most use cases, if you need to perform additional steps or want to deploy your agent in a different way, you just have to write your own script (and Dockerfile if applicable).

Testing your script

We don't currently have an easy way for you to test this script, you can reach out to us and we'll give you a way to reset your deployment instances on-demand to ease testing. You can also use an Ubuntu VPS to manually run and test your script.

Just remember a few things while creating your script:

• It will always receive the same arguments detailed above
• It should be able to handle all the possible scenarios, including
  • First deployment on a fresh instance (some tooling missing)
  • Deployment of a new version (older version needs to be properly stopped)
• It shouldn't emit unnecessary/unwanted logs on stderr (as those will always be reported in the CLI after deployments to help you debug)

💡 We are planning to provide a small scripts library in the future, please reach out to us if you'd like to contribute with your script!

🚀 Usage

Your script is ready to be used, or you want to test it?
Either way, to use it install our CLI if you don't already have it:

pip install libertai-client

and use the deploy command with the --deploy-script-url option:

libertai agent deploy . --deploy-script-url https://link-to-your.com/script.sh

💡 It needs to be accessible online as we'll use wget to fetch it on the deployment instance.

And just like that, your AI Agent is deployed using your custom deployment process 😎
If you'd like to go further, feel free to check other guides and to join our Telegram if you need assistance from the team.