<aside>

Description

The Prompt Builder is a visual tool for creating and managing AI prompts without code. It supports data connections from Salesforce via SOQL, JavaScript data transformation, response schema configuration, and prompt testing.

</aside>

What you can do

• Build AI prompts – Create prompts for AI to follow when responding to users
• Connect your data – Pull information from Salesforce records to make responses more relevant
• Transform data with code – Use JavaScript to process and enhance your data before sending it to AI
• Test your work – Try out your AI prompts before making them live
• Reuse your creations – Save successful prompts to use across different parts of your system

Step 1: Getting Started

1. Open the Prompt Builder from the SimpleTranslate platform.
2. Choose your starting point:
   • Select Create New Prompt to start fresh.
   • Select an existing prompt to edit or copy.

Step 2: Set up your AI model

1. Select your AI engine (the type of AI you want to use).
2. Choose a specific model from the dropdown.

Step 3: Understanding system parameters

When your prompt is called from Salesforce, these system parameters are available automatically:

• data – Contains all the data sent from Salesforce.
• recordId – The ID of the current record (when sent from a Record page).
• objectApiName – The API name of the current object (when sent from a Record page).

You can use these in your prompts like $data, $recordId, or $objectApiName.

For testing, you can fill in these parameters manually in the Prompt Builder. When the prompt is called from Salesforce, your manual values are overwritten with the actual Salesforce data.

Step 4: Create your variables (optional but powerful)

Variables let you insert extra dynamic information into your prompts.

1. In the left panel, select Add Variable.
2. Choose a variable type:
   • Text – Simple information like names or IDs (constant values).
   • JSON – Structured data from records (constant values).
   • SOQL Query – Pull specific data from Salesforce.
3. Name your variable (for example, customer_name or recent_cases).
4. Set up the variable:
   • For Text, just provide a name and value.
   • For SOQL (requires org credential setup at https://www.simpletranslate.io/dashboard/credentials):
     • Select your org credential in the Choose your Target Credential dropdown.
     • Write a query such as:
       • SELECT Name, Status FROM Case WHERE AccountId = '$recordId'

Step 5: Add JavaScript scripts (optional – for simple data processing)

JavaScript scripts let you transform and enhance data before it reaches the AI.

Note: Only synchronous functions are supported.

1. In the left panel, select Add Script.
2. Write JavaScript with an execute() function that will be called.
3. Access your data inside execute():
   • Use data to access Salesforce data.
   • Use recordId and objectApiName for record context.
   • Use any variables you created by name (for example, accounts, recent_cases).

Simple example script

function execute() {
  // Get just the account names from the accounts variable
  const accountNames = accounts.map((acc) => acc.Name);

  // Count total cases
  const totalCases = recent_cases.length;

  // Return the processed data
  return {
    names: accountNames,
    caseCount: totalCases,
  };
}

Use script results in your prompts like script_name.names or script_name.caseCount.

Step 6: Write your prompt messages

1. Add a system message (tells the AI how to behave).
   • Example: “You are a helpful customer service assistant. Always be polite and professional.”
2. Add user messages (what the user will ask).
   • Example: “Please help me with my recent order issue.”
3. Add variables to your messages:
   • Drag variables from the left panel into the message text.
   • They appear as $variable_name.
   • Example: “Hello $customer_name, I see you have $case_count open cases.”
   • Use system data: “Based on this record data: $data.”
   • Use script results: “Account names: myScript.names.”

Step 7: Configure response format

1. Choose response type:
   • Text – Regular conversational responses.
   • JSON – Structured data output (for integration with other systems).
2. If using JSON, define your schema (the structure you want), for example:

{
  "name": "salesforce_account",
  "strict": false,
  "schema": {
    "type": "object",
    "properties": {
      "Salutation": {
        "type": "string",
        "readOnly": true
      },
      "FirstName": {
        "type": "string",
        "description": "html",
        "readOnly": true
      },
      "LastName": {
        "type": "string",
        "readOnly": true
      }
    },
    "required": []
  }
}

To learn more about JSON Schema, see <https://json-schema.org/overview/what-is-jsonschema>.

<aside>

</aside>

Step 8: Test your prompt

1. Fill in test values for your variables in the left panel.
2. Select Test Prompt.
3. Review the response in the right panel.
4. Check token usage to understand costs.
5. Adjust and test again until you are satisfied.

Step 9: Save and activate