How to deploy? Confused about the different options

[andwaal-esmart]

Hi,

Could someone help me how to deploy an agent and the role of the different tools available? ADK CLI/ GCloud CLI/ Runner?

A quick summarization or the options I can find in the docs:

• Using the ADK CLI
• • In this case you only point to the agent location, no more code required
• Using GCloud CLI
• • In this case one are using a helper function get_fast_api_app which also supports setting up a db for state?

In addition to the two above, this example from the tutorial goes through how to setup Runner/Session/State.

So I guess my question is; If I want to host the agent (with our without UI), what is required? Is it enough to do ADK/GCloud approach? What is the role of the Runner in this and when is the Runner needed?

[boyangsvl]

The ADK CLI will be the easiest way to deploy to Cloud Run, if that's where you want to deploy. You don't have to worry about Runners when using the ADK CLI as it will create a default runner for you.

[kanlanc]

Hey @boyangsvl ,

Im finding out that making my own runners is helping the code run (and also is giving more control that im looking for)

That considered, I am trying to understand what are my deployment options.

For now, I have a fastapi server and the agent.py file that has custom runnenrs and everything.

For testing, I am just doing

if name == "main":
asyncio.run.....

using python but unsure how to expose this to the fastapi and then deploy it

Any help would be hugely appreciated!