> ## Documentation Index
> Fetch the complete documentation index at: https://koreai-content-gov.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Prompt Studio Guide

This guide covers how to access Prompt Studio, create and test prompts, and use the canvas options to optimize your results.

***

## Access Prompt Studio

1. Log in and select **Prompt Studio** from the modules list.
2. On the dashboard, view prompts in three tabs: **All prompts**, **My prompts**, **Shared prompts**.
3. Click **New prompt**, enter a name, and click **Proceed**.
4. Choose a creation method: **Generate a prompt**, **Start from scratch**, or **Prompt library**.

## Create Prompts

### Generate a Prompt

Expand a basic instruction into a detailed, comprehensive prompt using AI.

1. Click **Generate a prompt**.
2. Select a model and enter your instruction.
3. Review the AI-generated prompt.
4. Click **Proceed** to copy it to the canvas for customization.

<Note>Only OpenAI and Anthropic models support prompt generation. You can also access this feature directly on the canvas via the **Generate Prompt** button.</Note>

### Start from Scratch

Opens a blank canvas where you can manually write prompts, add variables, select models, and generate outputs with full control.

### Prompt Library

Access 65+ pre-built templates for common use cases like code generation, summarization, and Q\&A.

1. Click **Prompt library**.
2. Browse **My templates** (your saved templates) or **All templates** (built-in options).
3. Select a template and click **Use template** to copy it to the canvas.

<Note>Templates are read-only. Edit only after copying to the canvas.</Note>

## Use Prompt Canvas

### Add Prompts

| Field | Purpose |
| - | - |
| **System prompt** | Define the model's role and behavior (optional—use toggle to enable/disable) |
| **Prompt** | Provide task instructions for the model |
| **Response JSON schema** | Define structured output format (optional—if unsupported, schema is included with the prompt) |

Supported JSON schema types: String, Boolean, Number, Integer, Object, Array, Enum, anyOf.

### Add Variables

Variables let you test prompts with multiple values simultaneously.

**Syntax:** Use `{{variable_name}}` in your prompt.

1. Add variables in the Prompt field using double curly braces (for example, `{{product}}`).
2. When variables are added, a **Variables** column appears.
3. Assign values to each variable. Click **Add an empty row** for multiple test cases.

Variables are substituted with their values during execution, generating outputs for all rows at once.

### Select Models

Compare up to 5 models to identify the best fit for your use case.

1. Click the **Select Model** field.
2. Choose a model and connection.
3. Add additional models in columns to the right.
4. Click the settings icon to adjust parameters (temperature, top k, top p, max tokens).

### Generate Output

1. Click **Run** to generate outputs.
2. Review results in the model columns, including: response text, token count (input + output), and response time.

<Note>Maximum 10 rows can be generated simultaneously.</Note>

## Canvas Options

### Model Column Options

| Icon | Feature | Description |
| - | - | - |
| ⋮⋮ | Rearrange | Drag to reorder model columns |
| ✕ | Remove Model | Remove a model from comparison |
| ★ | Bookmark | Set as preferred model (required when committing if none selected) |
| ⚙ | Model Settings | Adjust temperature, top k, top p, max tokens (defaults: top p=1, top k=5, temperature=1, max tokens=256) |
| ▶ | Play | Regenerate output for the entire column |
| ⏱ | Avg Response Time | View average generation time |
| 🔢 | Avg Tokens | View mean input/output token usage |
| 📋 | Copy | Copy output to clipboard |
| { } | View JSON | View request/response in JSON format |
| ↻ | Regenerate | Regenerate output for a single cell |

### Top Toolbar Options

| Icon | Feature | Description |
| - | - | - |
| 📚 | Prompt Library | Browse \~70 templates |
| 🔗 | Prompt API | Share prompts via version-specific API endpoints (cURL, Python, Node.js) |
| 💾 | Save to Library | Save current prompt as a reusable template |
| 📝 | Draft History | Capture and restore canvas states |
| 📤 | Export CSV | Export canvas data (inputs, outputs, metadata) for analysis |
| 👥 | Share | Share prompts for collaboration |
| 🕐 | Versions | View, compare, and restore prompt versions |
| ✓ | Commit | Save current prompt as a new version (V1, V2, etc.) |
| ▶ | Run | Execute prompts and generate outputs |

## Quick Reference

| Task | Action |
| - | - |
| Create prompt quickly | Use **Generate a prompt** with AI expansion |
| Full control over prompt | Use **Start from scratch** |
| Use proven templates | Browse **Prompt library** |
| Test multiple scenarios | Add variables with `{{variable}}` syntax |
| Compare model performance | Add up to 5 models side by side |
| Fine-tune outputs | Adjust model settings (temperature, tokens) |
| Save for reuse | Commit version or save to library |
| Share with team | Use Share or Prompt API |
| Analyze results | Export as CSV |