Stop Burning Cash on AI Generation in Development

#webdev #ai

I’m building a startup with a very complex AI workflow — combining text generation and image generation.

Here’s the catch:

Text generation costs are negligible for one man team like mine, but not so much for larger teams.
Image generation? Not so much. Using DALL·E 3 costs about \$0.04 per image (standard size).

At first, I underestimated how fast this adds up. But when you consider:

Team size,
Frequency of commits,
Automated tests or integration runs,

…it’s easy to rack up serious costs just from developers iterating on code.

The Problem

If every commit triggers image generation in dev/test environments, you’re basically paying per debug cycle. That’s unsustainable.

The Solution: Provider Pattern

Instead of letting dev environments hit expensive APIs, use a provider pattern with a mock fallback.

In staging/production, use the real DALL·E provider.
In development, serve mock images (placeholders) to keep costs at zero.

Here’s a simplified example:

import { DallEAPIWrapper } from '@langchain/openai';
import { ENV } from '../getEnv';

type ImageServiceOptions = {
  description: string;
  size?: '256x256' | '512x512' | '1024x1024';
  quality?: 'standard' | 'hd';
  n?: number;
};

const defaultParams = {
  size: '1024x1024' as const,
  quality: 'standard' as const,
  n: 1,
  model: 'dall-e-3',
};

function buildFoodPrompt(description: string): string {
  return `
  You are professional photographer generate image for the following

  User description:
${description}
  `.trim();
}

// --- Provider Interface ---
interface ImageProvider {
  generate(options: ImageServiceOptions): Promise<string | null>;
}

// --- Providers ---
class DalleProvider implements ImageProvider {
  async generate({ description, size, quality, n }: ImageServiceOptions) {
    const dalle = new DallEAPIWrapper({
      size: size ?? defaultParams.size,
      quality: quality ?? defaultParams.quality,
      n: n ?? defaultParams.n,
      model: defaultParams.model,
      style: 'natural',
    });

    return (await dalle.invoke(buildFoodPrompt(description))) ?? null;
  }
}

class MockProvider implements ImageProvider {
  async generate({ description }: ImageServiceOptions) {
    // Could log description to mimic "generation" in dev
    console.log('Mock generation for:', description);
    return 'https://placehold.co/600x400.png';
  }
}

// --- Provider Factory ---
function getProvider(): ImageProvider {
  return ENV === 'dev' ? new MockProvider() : new DalleProvider();
}

// --- Service function using Provider Pattern ---
export async function generateMenuItemImage(
  options: ImageServiceOptions
): Promise<string | null> {
  const provider = getProvider();
  return provider.generate(options);
}