​
Pre-requisites

Ensure you’ve followed the Authentication steps to set up an organization and API key.

​
Deploying a prompt

A deployed prompt will have all of its settings saved. The prompt, the model type, model parameters, and any functions attached to it will all be stored as one deployed unit. With the deployment id you can fetch and invoke this prompt with all of its settings. If needed, you can also override any specific configurations via the SDK.

Deployed Prompts section in the Studio:

DeployedPrompts

There are two ways to deploy a prompt:

​
Option 1: From the Lab

From the Lab you can click the rocket icon on a specific prompt to deploy it. In the modal, you will give the deployed prompt a name and be able to copy the deployment-id. Don’t worry, if you miss the deployment-id, you can always find it on the studio tab.

DeployLab

​
Option 2: From the Studio

If you have already saved a prompt and want to deploy it you can find all of your saved prompts in the Studio. From there, you can select a specific prompt and click the Deploy button. Like in the Lab, you will give the deployed prompt a name and be able to copy the deployment-id.

DeployStudio

​
Using your deployed prompts

Take a look at our Cookbook for the python SDK to see how to use your deployed prompts with logging and the ability to capture feedback and metadata.

BenchmarkingMeasure Performance