WebsiteLegalPlatform

Prompt

A Prompt in Lleverage is a concise set of instructions in natural language that guides AI models to perform specific tasks within your workflow. Prompts are the core mechanic that drives the features you're building in your automations.

Key Features

  • Natural language instructions for directing AI behavior

  • Multiple model options to optimize for balance, speed, or intelligence

  • Support for text, images, and files as input for processing

  • Variable integration for dynamic workflow data

  • Advanced configuration through Prompt Studio

How to Add a Prompt Action

  1. Open the Add Action menu using one of three methods:

    • Click the "Add Action" button in the top left corner

    • Click on a connection circle on an existing action card

    • Click and drag from one action to create a connection

  2. Select "AI" from the categories or find "Prompt" in the "All" tab

  3. Click on "Prompt" to add it to your canvas

💡 If you add multiple prompts, they will be automatically numbered (Prompt 1, Prompt 2, etc.)

How to Configure the Prompt Action Card

  1. Locate the Prompt card on your canvas which includes:

    • Header with the action name (which can be renamed)

    • Variable chip in the top left corner for referencing outputs

    • Controls for expanding/collapsing, running the prompt, and additional options

    • Connection points for inputs and outputs

    • Natural language description that updates automatically

  2. Configure basic settings directly on the card:

    • Model selection: Choose from Balanced (default), Fastest, or Smartest options

    • Instructions: Add up to three instruction blocks in the main card view

  3. Set up your instructions:

    • Type your prompt text in the instructions area

    • Specify message roles (user or assistant) if desired

    • Generate content by clicking the star icon in the bottom left

    • Add media references using the image or clip icons

⚠️ Different models have different capabilities and speed. The model you choose may affect whether text, image, audio, and PDF processing are available.

Basic Prompt Action

How to Add Media and Files to Prompts

  1. Click the image icon or clip icon in the instruction area

  2. A link field will appear below the instruction

  3. Enter a variable reference, image URL, or file reference in this field

  4. Delete media attachments by clicking the delete icon if needed

Adding an image within a Prompt Action

Improved File Support

Gamechanger alert! Prompt node attachments can now include files of different types across all models! Using MarkItDown we support the conversion into markdown from:

  • PDF - Full text extraction and formatting

  • PowerPoint - Slide content and notes

  • Word - Document text and formatting

  • Excel - Spreadsheet data and structure

  • Images - EXIF metadata and OCR text extraction

  • Audio - EXIF metadata and speech transcription

  • HTML - Web page content

  • Text-based formats - CSV, JSON, XML

  • ZIP files - Iterates over contents

  • YouTube URLs - Video transcripts and metadata

  • EPubs - E-book content

💡 Files are automatically converted to markdown format for optimal AI processing, preserving structure and content while making it accessible to the language models.

How to Write Effective Prompts

  1. Be specific with your instructions to ensure relevant responses

  2. Provide clear context to help the model produce accurate results

  3. Structure your prompt logically

  4. Consider using:

    • Dynamic variables with the '@' button or typing with {{ brackets }} syntax for real-time data insertion

    • Parameters, traits, and constraints to guide the AI response

    • Different message roles to structure conversations

💡 You can reorder multiple instructions using the grab handles on the right side of each instruction box.

How to Access Prompt Studio (formerly Prompt IDE)

  1. Add a Prompt action to your workflow canvas

  2. Click on "Prompt Studio" at the bottom of the Prompt card

  3. The full-screen Prompt Studio interface will open

💡 You can move back and forth between Prompt Studio and your workflow for the most effective experience.

Prompt Studio
Prompt Node Footer and link to Prompt Studio

How to Use Prompt Studio

In the expanded Prompt Studio view, you can:

  • Select your model from expanded options

  • Configure advanced model settings (temperature, top P, top K, max tokens)

  • Set output format (text or JSON with optional schema)

  • Resize columns for better workspace management

  • Add and reorder multiple instructions using grab handles

Advanced configuration includes:

  • Output settings: Choose between text or JSON format with schema options

  • Column resizing: Adjust the workspace to focus on different aspects of prompt building

  • Expanded instruction view: More space for complex prompt engineering

⚠️ Advanced model settings like temperature and token limits should only be adjusted if you understand their impact on AI behavior

How to Configure Advanced Model Settings

  1. In Prompt Studio, locate the cog button on the right side of the model selection

  2. Click the cog button to open advanced configuration options

  3. Configure the following parameters as needed:

    • Temperature: Controls randomness in responses

    • Top P: Nucleus sampling parameter

    • Top K: Top-k sampling parameter

    • Frequency Penalty: Reduces repetition of tokens

    • Presence Penalty: Encourages topic diversity

    • Max Tokens: Maximum response length

How to Set Output Format

  1. In the advanced settings, locate the "Format" section

  2. Choose between two options:

    • Text: Output will be returned as plain text

    • JSON: Output will be structured as JSON data

For JSON Output:

  1. Select "JSON" from the format options

  2. Toggle "JSON Schema" to enable schema definition

  3. Enter your desired JSON schema in the provided field

  4. This ensures the AI returns data in the exact structure you specify

💡 JSON output with schema is particularly useful when you need structured data for subsequent workflow actions

How to Test Prompts in Prompt Studio

  1. Locate the "Test with mock data" option in the top right corner

  2. This allows you to test the prompt in isolation from the rest of the workflow

  3. Fill in any required variables with test values

  4. Click "Run" to see the AI's response

  5. Make adjustments based on the results

💡 Testing with mock data lets you isolate and perfect your prompt before integrating it with the full workflow

How to Create and Compare Variants

  1. In Prompt Studio, click the chevron next to your variant name in the left column

  2. Select "Add new variant" to create a new version

  3. Configure the new variant with different settings or instructions

To compare variants:

  1. Click "Compare mode"

  2. View variants side-by-side in vertical format

  3. See how each responds to the same input

  4. Create new variants directly from comparison view

  5. Duplicate variants as needed

  6. Activate your preferred variant by selecting it and returning to edit mode

💡 Creating variants allows you to keep your previous configurations while testing new approaches

How to Save and Exit

  1. All changes are automatically saved when you close Prompt Studio

  2. Click the X or close button to return to the workflow canvas

  3. Your prompt configurations will be processed and integrated with your workflow

Types of Prompts Supported

  • Chat for conversational flows

  • Completions for text analysis, text generation, or data processing

Last updated