Skip to Content
DocsPlaygroundInterface

Playground Interface

The Playground interface is designed to emulate a real-world chat application while providing powerful developer tools.

Header Toolbar

The top toolbar controls the execution environment:

  • Model Selector: Dropdown to select the LLM (e.g., gpt-4o, claude-3-5-sonnet). Only models with configured API keys are shown.
  • Parameters:
    • Temperature: Controls randomness (0.0 to 2.0). Lower values are more deterministic.
    • Max Tokens: Limits the response length.
  • Save: Persists the current prompt state.
  • Run: Executes the prompt.

Main Editor

The central area is where you craft your prompt. It supports:

  • System Message: The core instruction for the AI.
  • User Input: The simulated user query.
  • Syntax Highlighting: For readability.
  • Injections: Auto-complete for {{ variable }} and [[ prompt_path ]].

Variables Panel (Left)

The Variables section allows you to define dynamic inputs for your prompt.

  • Variables: Key-value pairs matching your {{ variable }} placeholders.
  • Tree View: Navigate your prompt library to reference other prompts via [[ injection ]].

Results Panel (Bottom)

After running a prompt, the results panel slides up to show:

  • Output: The raw text response from the model.
  • Metrics:
    • Latency: Time to first token and total time.
    • Tokens: Input and output token usage.
    • Cost: Estimated cost of the run (if available).
  • JSON View: Inspect the full raw JSON response from the provider, which is useful for debugging tool calls.
Last updated on
OverviewExecution