Playground

The Split Window displays the prompt editor, input values, and model response side by side, allowing you to test and compare prompt behavior efficiently.

The Save as Prompt option saves the current prompt configuration for future reuse.

Saved prompts can be reused, modified, and executed whenever required.

The Run All option executes the configured prompt using the selected model settings, input values, and prompt configuration.

The Reset Playground option clears the configured prompts, input values, settings, and generated responses, restoring the Playground to its default state.

Temperature controls the randomness of the generated response.

Lower values produce more consistent responses, while higher values generate more diverse outputs.

Max Tokens defines the maximum number of tokens that the model can generate in a response.

Top P controls the diversity of token selection during response generation.

Max Reasoning Tokens controls the reasoning budget used before generating a response for supported models.

Additional Options allows you to provide advanced, model-specific parameters as a JSON object.

Usage

1. Enable Additional Options.
2. Enter a valid JSON object.
3. Execute the prompt.

Example

{
  "frequency_penalty": 0.5,
  "presence_penalty": 0.2,
  "stop": ["\n\n"]
}

Use this option only when model-specific parameters are required.

The API Key option overrides the default API key configured for the project.

If left blank, the project default API key is used during prompt execution.

Tools allow the model to invoke predefined functions during prompt execution.

Each tool is defined using a JSON schema that specifies its inputs and enables the model to interact with external functions during execution.

Tool Definition Includes

• Tool name
• Description of the tool purpose
• Input parameters
• JSON schema definition

Configured tools can be tested within the Playground before being reused across the project.

Structured Output ensures that the model response follows a predefined JSON schema, making responses consistent and easier for downstream applications to process.

Schema Defines

• Required fields
• Data types
• Validation constraints

Prompt Variables allow dynamic values to be supplied during execution, enabling the same prompt to be reused with different inputs.

Variables are defined using double curly braces.

Examples

{{user_input}}

{{question}}

{{context}}

Variable Format

{{variable_name}}

Messages define the role of each interaction within a prompt and determine how instructions, user inputs, and generated responses are handled.

Variables allow dynamic input values to be injected into prompts during execution.

Message placeholders allow chat messages and conversation history to be inserted into prompts dynamically.

Each message placeholder contains one or more messages in JSON format.

[
  {
    "role": "user",
    "content": "..."
  }
]

Message Placeholders Can Inject

• Previous conversation history
• System instructions
• User messages
• Assistant responses
• External conversation data

Usage

1. Create a message placeholder.
2. Assign a descriptive name.
3. Provide one or more messages in JSON format.
4. Execute the prompt.