Lab Interface Guide

The Lab interface is divided into three main sections: Configuration, Input, and Output.

1. Configuration Panel

Located on the right sidebar, this panel controls the LLM execution parameters.

• Model Selection: Choose from any provider configured in your organization (e.g., gpt-4o, claude-3-5-sonnet).
• Temperature: Control randomness (0.0 to 2.0).
  • Recommendation: Use 0.0 for structured tasks (extraction, code) and 0.7+ for creative writing.
• Max Tokens: Limit the response length to manage costs and prevent runway loops.
• Response Mode: Toggle between Text and JSON Mode (force valid JSON output).

2. Input Section

The central editor where you define your prompt scenario.

System & User Prompts

• System Prompt: Define the AI’s persona, constraints, and core logic.
• User Prompt: The specific task or query for this execution.

Variables

Inject dynamic data into your prompts using handlebars syntax {{variable_name}}.

• Define variable values in the Variables tab below the editor.
• Variables can be strings, numbers, or JSON objects.

Attachments

Upload images or text files to test multimodal capabilities (Vision/RAG) if the selected model supports it.

3. Output Section

Analyzing the results of your execution.

• Completion: The raw text or JSON returned by the model.
• Latency: Time to first token (TTFT) and total generation time.
• Token Usage: Breakdown of Input, Output, and Total tokens for cost estimation.
• Reasoning: If using reasoning models (e.g., o1), the internal thought process is displayed here.