Playground
An interactive LLM testing environment for prompts, models, datasets, and tool calls — no code required.
Playground
The Playground is an interactive LLM testing environment built into the dashboard. Use it to write and iterate prompts, compare models side by side, run datasets in bulk, score outputs with evaluators, and test tool/function calling — all without writing any code.
How to Access
Navigate to your project in the dashboard and click Playground in the left sidebar.
Window state (messages, model selection, variables) persists across page refreshes so you can pick up where you left off.
Key Use Cases
Prompt Testing
Write prompts with Handlebars variables, load from the prompt library, run multi-turn conversations, and save back as new versions.
Model Comparison
Open up to 5 windows, configure each with a different model or parameters, and run them all simultaneously to compare outputs.
Dataset Runs
Connect a dataset to your variables and run batch execution across every row to systematically test your configuration.
Inline Evaluations
Score outputs with your project's evaluators directly in the Playground without creating a full experiment.
Tool & Function Calling
Define tool schemas, test model tool-call decisions, and validate structured output formats.
Assist
Let the Assist AI agent read your prompts, outputs, and scores and propose prompt edits you apply as a diff.