Form Submit

Workflow Type: Form Submit AI Workflow

Form Submit workflows are triggered when users fill out a form. Perfect for content generation, data processing, and user-initiated tasks.

Use Case: Story Generator

Users submit a genre, description, and optional inspiration document, and the workflow generates a complete story in a Word document.

Here is how a Form Submit AI workflow looks like:

---

Step 1: Create the Story Generator Function

This function accepts form inputs and generates a story document.

from abacusai import (
    ApiClient,
    AgentInterface, 
    AgentResponse,
    WorkflowGraph,
    WorkflowGraphNode,
    WorkflowNodeInputMapping,
    WorkflowNodeInputSchema,
    WorkflowNodeInputType,
    WorkflowNodeOutputMapping,
    WorkflowNodeOutputSchema,
    WorkflowNodeOutputType,
)

def story_generator(genre_type, description, inspiration=None):
    """
    Generates a story based on inspiration document, genre, and description.

    Args:
        genre_type (str): Selected genre for the story
        description (str): Description of the story to be generated
        inspiration (Blob, optional): Uploaded document containing inspiration material

    Returns:
        AgentResponse: Contains the generated story as a Word document
    """
    # ALL imports inside the function
    from abacusai import AgentResponse, ApiClient, Blob
    from docx import Document
    import io

    client = ApiClient()
    
    # Notify user of progress (streaming messages)
    client.stream_message("Processing inspiration document...")
    
    # Extract text from uploaded document (if provided)
    inspiration_text = ""
    if inspiration:
        try:
            # Try advanced extraction with OCR
            extracted_doc_data = client.extract_document_data(
                inspiration.contents,
                document_processing_config={
                    "extract_bounding_boxes": False,
                    "ocr_mode": "AUTO",
                    "use_full_ocr": True,
                },
            )
            inspiration_text = extracted_doc_data.extracted_text
        except:
            # Fallback to basic extraction
            extracted_doc_data = client.extract_document_data(inspiration.contents)
            inspiration_text = extracted_doc_data.embedded_text
    
    client.stream_message("Generating story based on your inputs...")
    
    # Create detailed prompt for story generation
    prompt = f"""
    Create an engaging and well-structured story based on the following inputs:

    GENRE: {genre_type}
    DESCRIPTION: {description}
    INSPIRATION MATERIAL: {(inspiration_text[:5000] if inspiration_text else 'No inspiration material provided')}

    Please write a complete story that:
    - Follows the conventions of the {genre_type} genre
    - Incorporates elements from the description provided
    - Draws inspiration from the provided material (if any)
    - Has a clear beginning, middle, and end
    - Is approximately 1500-2500 words
    - Includes engaging dialogue and descriptive scenes
    """
    
    system_message = f"""You are a professional creative writer specializing in {genre_type} stories.
    Create an original, captivating story that incorporates the user's requirements while maintaining
    high literary quality. Use vivid descriptions, compelling characters, and engaging plot development
    appropriate for the {genre_type} genre."""
    
    # Generate story using LLM
    story_response = client.evaluate_prompt(
        prompt=prompt,
        system_message=system_message,
        llm_name="CLAUDE_V3_5_SONNET"  # High-quality creative writing model
    )
    generated_story = story_response.content
    
    client.stream_message("Creating Word document...")
    
    # Create Word document
    doc = Document()
    
    # Add title
    title = doc.add_heading(f"Generated {genre_type} Story", 0)
    
    # Add metadata
    doc.add_heading("Story Details", level=1)
    doc.add_paragraph(f"Genre: {genre_type}")
    doc.add_paragraph(f"Description: {description}")
    doc.add_paragraph("")
    
    # Add story content
    doc.add_heading("Story", level=1)
    story_paragraphs = generated_story.split("\n\n")
    for paragraph in story_paragraphs:
        if paragraph.strip():
            doc.add_paragraph(paragraph.strip())
    
    # Save document to bytes
    doc_buffer = io.BytesIO()
    doc.save(doc_buffer)
    doc_bytes = doc_buffer.getvalue()
    
    # Return as Blob (file download)
    return AgentResponse(
        story_document=Blob(
            doc_bytes,
            "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
            filename=f"{genre_type.lower().replace(' ', '_')}_story.docx",
        )
    )

Key Concepts:

• client.stream_message(): Send real-time updates to the user
• Blob: Wrapper for file data (documents, images, etc.)
• extract_document_data(): OCR and text extraction from PDFs/Word docs

---

Step 2: Create Node with Input Schema

For form-based workflows, you define an input_schema to create the user interface.

from abacusai import WorkflowNodeInputSchema, WorkflowNodeOutputSchema

story_generator_node = WorkflowGraphNode(
    "Story Generator",
    story_generator,
    
    # Define the input form
    input_schema=WorkflowNodeInputSchema(
        json_schema={
            "type": "object",
            "title": "Story Generator Inputs",
            "required": ["genre_type", "description"],  # Required fields
            "properties": {
                "genre_type": {
                    "enum": [  # Dropdown options
                        "Fantasy",
                        "Science Fiction",
                        "Mystery",
                        "Romance",
                        "Horror",
                        "Thriller",
                        "Adventure",
                        "Historical Fiction",
                        "Contemporary Fiction",
                        "Young Adult",
                        "Comedy",
                        "Drama",
                    ],
                    "type": "string",
                    "title": "Story Genre",
                    "description": "Select the genre for your story",
                },
                "description": {
                    "type": "string",
                    "title": "Story Description",
                    "description": "Provide a brief description of the story you want to generate",
                },
                "inspiration": {
                    "type": "string",
                    "title": "Inspiration Document",
                    "format": "data-url",  # File upload
                    "description": "Upload a PDF or Word document for inspiration (optional)",
                },
            },
        },
        
        # Define UI widgets
        ui_schema={
            "genre_type": {"ui:widget": "select"},      # Dropdown
            "description": {"ui:widget": "textarea"},   # Multi-line text
            "inspiration": {"ui:widget": "file"},       # File uploader
        },
    ),
    
    output_mappings=["story_document"],
    
    # Define the output display
    output_schema=WorkflowNodeOutputSchema(
        json_schema={
            "type": "object",
            "title": "Generated Story",
            "properties": {
                "story_document": {
                    "type": "string",
                    "title": "Story Document",
                    "format": "data-url",  # File download
                    "description": "Download your generated story",
                }
            },
        }
    ),
)

Input Schema Breakdown:

Property	Type	Widget	Purpose
genre_type	string (enum)	select	Dropdown menu
description	string	textarea	Multi-line input
inspiration	string (data-url)	file	File upload

Output Schema:

• story_document: File download with format: "data-url"

---

Step 3: Build Workflow Graph

agent_interface = AgentInterface.DEFAULT

workflow_graph = WorkflowGraph(
    nodes=[story_generator_node],
    specification_type="data_flow"
)

Note: Form submit workflows often have a single node since user input directly triggers the function.

---

Step 4: Register the Agent

from abacusai import ApiClient

client = ApiClient()

agent = client.create_agent(
    project_id='your_project_id',
    workflow_graph=workflow_graph,
    agent_interface=agent_interface,
    description="AI-powered story generator with genre selection and document inspiration",
    package_requirements=["python-docx"],  # Required for Word document creation
)

agent.wait_for_publish()
print(f"Story generator agent created: {agent.agent_id}")

Key Differences:

• agent_interface=AgentInterface.DEFAULT
• package_requirements: Include python-docx for Word document generation
• No connectors needed for this workflow