AIGenerateDocument

The AIGenerateDocument tool uses AI to generate structured documents based on your prompts. It combines the power of large language models with document validation to create properly formatted, type-safe documents automatically.

Documents generated with AIGenerateDocument are validated against your document schema and displayed as interactive forms in Loopstack Studio.

Basic Usage

Simple Document Generation


transitions:
  - id: generate_summary
    from: start
    to: summary_created
    call:
      - tool: AIGenerateDocument
        args:
          prompt: |
            Create meeting notes for a team standup on January 15th
            {{ rawNotes }}
          responseDocument: MeetingNotesDocument

Use separate messages


transitions:
  - id: generate_with_context
    from: processing
    to: complete
    call:
      - tool: AIGenerateDocument
        args:
          messages:
            - role: system
              content: "You are a professional meeting notes assistant"
            - role: user
              content: "Summarize the key decisions from today's meeting"
            - role: user
              content: "Notes: {{ rawNotes }}"
          responseDocument: MeetingNotesDocument

Use Workflow Message History


transitions:
  - id: generate_with_context
    from: processing
    to: complete
    call:
      - tool: CreateChatMessage
        args:
          role: user
          content: |
            Create meeting notes for a team standup on January 15th
            {{ rawNotes }}
      - tool: AIGenerateDocument
        args:
          responseDocument: MeetingNotesDocument
          messagesSearchTag: message                  # search for all messages in the workflow's document history

Configuration Options

LLM Settings


transitions:
  - id: generate_with_model
    from: start
    to: generated
    call:
      - tool: AIGenerateDocument
        args:
          llm:
            provider: openai
            model: gpt-4
          prompt: "Generate a project task list"
          responseDocument: TaskDocument

Parameters

prompt: A string prompt for the AI (use this for simple requests)
messages: An array of messages for more complex conversations
messagesSearchTag: Reference previously tagged messages in the workflow
responseDocument: The document class to generate (required)
llm: Optional LLM configuration
- model: The AI model to use
- provider: The AI provider (e.g., “openai”, “anthropic”)