Prompt Management

Configure and save prompts to be re-used later.

The prompt library is a collection of saved prompts to be re-used within modules of orq.ai. This is useful to have pre-saved prompt configurations when setting up new Playgrounds, Experiments or Deployments.

To access your prompt library:

  • Head to the orq.ai panel
  • Select the Prompt Library page from the Data section

Adding a prompt to the library

To add a new prompt to the library, select Create > Prompt. You will then be asked to enter a Title, Project and Description.

Once validated, you will be taken to the following Prompt configuration screen:

Your prompt configuration, this will be saved and reused within other modules of **orq.ai**.

Your prompt configuration, this will be saved and reused within other modules of orq.ai.

You can enter here the following information that can be used to filter and search for your prompt:

  • Language supported by the given prompt
  • Use Cases for the current prompt, helping understand whether the prompt is used for Agents, Chatbots, Multi-modal, or more.

Optionally, you can predefine here a modelfrom your Model Garden that will be queried using your prompt. You will be able to define also all model parametersneeded.

Finally, here you can define the Messages sent to the model, you can define any number of messages of any role. These messages will be copied and re-used when using a library prompt.

Once ready to use your prompt, select Save changes

Using a prompt from the library

Once your prompt is ready to use, you can see it in the main page of the Prompt Library.

To use the prompt, select the ... next to the prompt name, the following menu opens:

Here you can decide to:

  • Run the prompt in a Playground, this will create a new Playground using the prompt configuration and optional model, ready to be tested.
  • Run the prompt in an Experiments, this will create a new experiment and add the optional model as a model to be tested.
  • Create a new Deployment, the prompt will be used and optional model will be used as primary model for the main variantof the Deployment.