AI fields generate dynamic content using AI models during flow execution. Instead of static text, you provide prompts that are resolved at runtime with context and variables.

​
How it works

  1. Define prompt: Use AIField.prompt() with your generation prompt
  2. Runtime resolution: AI generates content based on prompt and flow context
  3. Variable access: Use {{variable_name}} to include flow variables
  4. Context aware: AI has access to conversation history and user data

​
Basic usage

from kapso.builder.ai.field import AIField

# Three equivalent ways to create AI fields:
AIField(prompt="Generate a greeting")          # Recommended
AIField.prompt("Generate a greeting")          # Also works
AIField("Generate a greeting")                 # Positional argument

# Example usage
node = SendTextNode(
    id="greeting",
    whatsapp_config_id="config_123",
    message=AIField(prompt="Generate a warm, professional greeting"),
    provider_model_name="claude-3-5-sonnet"
)

​
Supported nodes

SendTextNode 
SendTextNode(
    message=AIField(prompt="Your prompt here")
)
SendTemplateNode 
SendTemplateNode(
    parameters=AIField(prompt="Generate template parameters as JSON")
)
SendInteractiveNode 
SendInteractiveNode(
    header_text=AIField(prompt="Generate header text"),
    body_text=AIField(prompt="Generate body content"),
    footer_text=AIField(prompt="Generate footer"),
    buttons=[
        Button(text=AIField(prompt="Generate button text"))
    ]
)

​
Variable interpolation

Access flow variables in your prompts: 
# Using flow variables
node = SendTextNode(
    id="personalized_greeting",
    message=AIField(prompt=
        "Generate a personalized greeting for {{customer_name}} who is interested in {{product_category}}"
    ),
    provider_model_name="gpt-4"
)
Available variables:
  • Flow variables: {{variable_name}}
  • User context: {{phone_number}}, {{conversation_id}}
  • System info: {{flow_id}}, {{current_time}}

​
Examples

Dynamic support response 
support_response = SendTextNode(
    id="support_help",
    message=AIField(prompt=
        "Based on the user's issue about {{issue_type}}, provide a helpful solution. "
        "Keep it concise and actionable."
    ),
    provider_model_name="claude-3-5-sonnet"
)
Template with AI parameters 
# AI generates template parameters
confirmation = SendTemplateNode(
    id="order_confirmation",
    template_id="order_template_123",
    parameters=AIField(prompt=
        "Generate order confirmation parameters for customer {{customer_name}} "
        "with order total {{order_total}}. Return as JSON: "
        "{\"customer_name\": \"...\", \"order_total\": \"...\", \"estimated_delivery\": \"...\"}"
    ),
    provider_model_name="gpt-4"
)
Interactive buttons 
# Dynamic button options
interactive = SendInteractiveNode(
    id="product_options",
    header_text="Product Categories",
    body_text=AIField(prompt=
        "Generate engaging body text about our {{featured_category}} products"
    ),
    buttons=[
        Button(
            id="option_1",
            text=AIField(prompt="Generate button text for {{category_1}}")
        ),
        Button(
            id="option_2", 
            text=AIField(prompt="Generate button text for {{category_2}}")
        )
    ],
    provider_model_name="claude-3-5-sonnet"
)
Conditional content 
# Content varies based on user data
response = SendTextNode(
    id="status_update",
    message=AIField(prompt=
        "Generate a status update message. "
        "User type: {{user_type}}, Account status: {{account_status}}. "
        "If premium user, mention exclusive benefits. "
        "If basic user, suggest upgrade options."
    ),
    provider_model_name="gpt-4"
)

​
Requirements

  • AI model required: Must specify provider_model_name when using AI fields
  • Credits consumption: AI field resolution consumes processing credits
  • Runtime resolution: Fields are resolved during flow execution, not at build time