Skip to Content
DocumentationFeaturesAI Structured Output

AI Structured Output

Use ClaudeGenerateDocument to have the LLM generate structured data conforming to a Zod schema. The output is validated and stored as a typed document.

Define a Document

import { z } from 'zod';
import { Document } from '@loopstack/common';
 
export const FileDocumentSchema = z
  .object({
    filename: z.string(),
    description: z.string(),
    code: z.string(),
  })
  .strict();
 
export type FileDocumentType = z.infer<typeof FileDocumentSchema>;
 
@Document({
  schema: FileDocumentSchema,
  uiConfig: __dirname + '/file-document.yaml',
})
export class FileDocument {
  filename: string;
  description: string;
  code: string;
}

Workflow Example

import { ClaudeGenerateDocument, ClaudeMessageDocument } from '@loopstack/claude-module';
import { BaseWorkflow, DocumentEntity, Final, Initial, InjectTool, Transition, Workflow } from '@loopstack/common';
 
@Workflow({
  uiConfig: __dirname + '/prompt-structured-output.ui.yaml',
  schema: z.object({
    language: z.enum(['python', 'javascript', 'java', 'cpp', 'ruby', 'go', 'php']).default('python'),
  }),
})
export class PromptStructuredOutputWorkflow extends BaseWorkflow {
  @InjectTool() claudeGenerateDocument: ClaudeGenerateDocument;
 
  llmResult?: DocumentEntity<FileDocumentType>;
 
  @Initial({ to: 'ready' })
  async greeting() {
    const args = this.ctx.args as { language: string };
    await this.repository.save(
      ClaudeMessageDocument,
      {
        role: 'assistant',
        content: [{ type: 'text', text: `Creating a 'Hello, World!' script in ${args.language}...` }],
      },
      { id: 'status' },
    );
  }
 
  @Transition({ from: 'ready', to: 'prompt_executed' })
  async prompt() {
    const args = this.ctx.args as { language: string };
    const result = await this.claudeGenerateDocument.call({
      claude: { model: 'claude-sonnet-4-6' },
      response: { document: FileDocument },
      prompt: this.render(__dirname + '/templates/prompt.md', { language: args.language }),
    });
    this.llmResult = result.data as DocumentEntity<FileDocumentType>;
  }
 
  @Final({ from: 'prompt_executed' })
  async respond() {
    await this.repository.save(
      ClaudeMessageDocument,
      {
        role: 'assistant',
        content: [{ type: 'text', text: `Successfully generated: ${this.llmResult?.content?.description ?? ''}` }],
      },
      { id: 'status' },
    );
  }
}

How It Works

  1. claudeGenerateDocument calls the LLM with a structured output schema derived from the document’s Zod schema
  2. The LLM response is validated against the schema
  3. The result is returned as a DocumentEntity with typed content
  4. The document is automatically saved and displayed in the UI

Key Parameters

await this.claudeGenerateDocument.call({
  claude: { model: 'claude-sonnet-4-6' },
  response: {
    document: FileDocument, // Target document class
    id: 'unique-id', // Optional: custom document ID
  },
  prompt: 'Generate a Hello World script.',
});

Registry References

Last updated on