🚧

Note

This Playground option is currently only available within the SUPERWISE® Agent Builder Studio.

To allow users to test their application and agent configuration and “try it out” before deploying and publishing it, SUPERWISE provides a playground where users can test how their application will behave based on their current configuration, even before it is saved.

Using the UI

In this mode, you can adjust various settings and immediately observe their impact on the model's responses and performance. Perfect for tweaking and getting everything just right in an iterative, real-time environment.

Debugging the application

SUPERWISE® now enables you to trace the model’s thought process. This allows you to better understand and improve the app configuration and model settings, ensuring greater relevance and accuracy for its task. Read more here.

Some things worth knowing

Application preview

In the playground, SUPERWISE® provides a feature that allows you to preview and interact with the app using the current configurations, even before saving them. This enables you to see and experience the app as it would appear in the embedded environment, allowing you to make adjustments in real time based on the preview feedback.

Clear button

Need to start over? Click the clear button to wipe the chat slate clean, removing all prior conversations and resetting the chat for a new session.

Stop generating

Take command of the chat's response output with this feature, which lets you promptly stop the model from generating replies whenever needed.

Query the playground using the API / SDK:

We can utilize the SDK to test different configurations through the SUPERWISE application Playground.

The playground essentially enables users to run queries against an ad-hoc application configuration, as shown below.

application_response = sw.application.ask_playground(input=user_input,
                                                     llm_model=model,
                                                     prompt=app.prompt,
                                                     tools =tools,
                                                     chat_history=chat_history,
                                                      guards=[]
                                                     
                                                    )