OpenBuff: Custom AI Workflow Generator (TypeScript)

Initialize the generation process by describing the workflow's goals in the OpenBuff terminal. The terminal acts as the primary interface for instructing the generator agent.

buff generate --name "CustomerSupportRouter" --description "Route emails to Slack channels based on sentiment and urgency"

Watch out: Be specific about the inputs and outputs; vague descriptions will result in generic boilerplate that requires more manual refactoring.

The generator sends the intent to Claude 3.5 Sonnet to map out the required logic gates, API integrations, and data validation steps needed for a robust implementation.

Watch out: Ensure your API key has sufficient quota for high-token output, as generating full TypeScript classes can be resource-intensive.

Based on the analysis, the agent generates a complete .ts file including imports, interfaces, and the core execution logic with built-in retry mechanisms.

import { z } from 'zod';

export const InputSchema = z.object({
  emailBody: z.string(),
  sender: z.string().email(),
});

export class WorkflowRunner {
  async execute(input: z.infer<typeof InputSchema>) {
    // Generated logic here
  }
}

Watch out: Review the generated Zod schemas to ensure they match your external data sources' actual payloads.