Agent Skills: Ideogram Events & Async Patterns

Ideogram Events & Async Patterns

Overview

Ideogram's API is synchronous -- each call blocks until the image is generated (5-15 seconds). For production applications, wrap it in async patterns: job queues for batch generation, callbacks for downstream processing, and pipelines for image post-processing. This skill covers BullMQ queue patterns, callback handlers, and asset processing pipelines.

Prerequisites

• IDEOGRAM_API_KEY configured
• Redis for BullMQ job queue
• Storage for generated images (S3, GCS, or R2)
• Understanding of Ideogram models and style types

Instructions

Step 1: Job Queue for Async Generation

import { Queue, Worker } from "bullmq";
import { writeFileSync, mkdirSync } from "fs";
import { join } from "path";

interface GenerationJob {
  prompt: string;
  style: string;
  aspectRatio: string;
  model: string;
  callbackUrl?: string;
  metadata?: Record<string, string>;
}

const connection = { host: "localhost", port: 6379 };
const imageQueue = new Queue("ideogram-generation", { connection });

// Enqueue a generation job
async function submitGeneration(job: GenerationJob) {
  return imageQueue.add("generate", job, {
    attempts: 3,
    backoff: { type: "exponential", delay: 2000 },
    removeOnComplete: 100,
    removeOnFail: 50,
  });
}

// Worker processes jobs with concurrency limit
const worker = new Worker("ideogram-generation", async (job) => {
  const { prompt, style, aspectRatio, model, callbackUrl, metadata } = job.data;

  // Call Ideogram API (synchronous, blocks 5-15s)
  const response = await fetch("https://api.ideogram.ai/generate", {
    method: "POST",
    headers: {
      "Api-Key": process.env.IDEOGRAM_API_KEY!,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      image_request: {
        prompt,
        model: model || "V_2",
        style_type: style || "AUTO",
        aspect_ratio: aspectRatio || "ASPECT_1_1",
        magic_prompt_option: "AUTO",
      },
    }),
  });

  if (response.status === 429) {
    throw new Error("Rate limited"); // BullMQ will retry with backoff
  }
  if (!response.ok) {
    throw new Error(`Ideogram API error: ${response.status}`);
  }

  const result = await response.json();
  const image = result.data[0];

  // Download immediately (URLs expire)
  const imgResp = await fetch(image.url);
  const buffer = Buffer.from(await imgResp.arrayBuffer());
  const outputDir = "./generated";
  mkdirSync(outputDir, { recursive: true });
  const filePath = join(outputDir, `${image.seed}.png`);
  writeFileSync(filePath, buffer);

  // Fire callback if provided
  if (callbackUrl) {
    await fetch(callbackUrl, {
      method: "POST",
      headers: { "Content-Type": "application/json" },
      body: JSON.stringify({
        event: "generation.completed",
        jobId: job.id,
        prompt,
        seed: image.seed,
        resolution: image.resolution,
        filePath,
        metadata,
      }),
    });
  }

  return { seed: image.seed, filePath, resolution: image.resolution };
}, {
  connection,
  concurrency: 5, // Stay under 10 in-flight limit
});

worker.on("failed", (job, err) => {
  console.error(`Job ${job?.id} failed:`, err.message);
});

Step 2: Callback Handler

import express from "express";

const app = express();
app.use(express.json());

app.post("/callbacks/ideogram", async (req, res) => {
  const { event, jobId, seed, filePath, metadata } = req.body;
  res.status(200).json({ received: true });

  switch (event) {
    case "generation.completed":
      console.log(`Image generated: seed=${seed}, path=${filePath}`);
      await processImage(filePath, metadata);
      break;
    case "generation.failed":
      console.error(`Generation failed: job=${jobId}`);
      await notifyFailure(jobId, req.body.error);
      break;
  }
});

Step 3: Batch Marketing Asset Generation

async function generateMarketingCampaign(
  campaignName: string,
  products: string[],
  formats: Array<{ name: string; aspect: string; style: string }>
) {
  const jobs = [];

  for (const product of products) {
    for (const format of formats) {
      const job = await submitGeneration({
        prompt: `${product}, professional ${format.name} design, high quality`,
        style: format.style,
        aspectRatio: format.aspect,
        model: "V_2",
        callbackUrl: "https://api.myapp.com/callbacks/ideogram",
        metadata: { campaign: campaignName, product, format: format.name },
      });
      jobs.push(job);
    }
  }

  console.log(`Submitted ${jobs.length} generation jobs for campaign: ${campaignName}`);
  return jobs.map(j => j.id);
}

// Example: Generate all assets for a product launch
await generateMarketingCampaign("Q1 Launch", [
  "Cloud analytics dashboard",
  "Mobile payment app",
], [
  { name: "social-square", aspect: "ASPECT_1_1", style: "DESIGN" },
  { name: "story-vertical", aspect: "ASPECT_9_16", style: "DESIGN" },
  { name: "blog-hero", aspect: "ASPECT_16_9", style: "REALISTIC" },
]);

Step 4: Image Post-Processing Pipeline

import sharp from "sharp";

async function processImage(filePath: string, metadata?: Record<string, string>) {
  const variants = [
    { suffix: "-og", width: 1200, height: 630 },       // Open Graph
    { suffix: "-thumb", width: 400, height: 400 },      // Thumbnail
    { suffix: "-social", width: 1080, height: 1080 },   // Instagram
  ];

  for (const variant of variants) {
    const outputPath = filePath.replace(".png", `${variant.suffix}.webp`);
    await sharp(filePath)
      .resize(variant.width, variant.height, { fit: "cover" })
      .webp({ quality: 85 })
      .toFile(outputPath);
    console.log(`Created variant: ${outputPath}`);
  }
}

Error Handling

| Issue | Cause | Solution | |-------|-------|----------| | Rate limited | Too many concurrent jobs | Set worker concurrency to 5 | | Content filtered | Prompt violates policy | Log and skip, notify reviewer | | Expired URL | Worker too slow | Download in same worker step | | Queue stalled | Redis connection lost | Configure BullMQ connection retry | | Callback fails | Downstream service down | Fire-and-forget with retry queue |

Output

• BullMQ job queue for async generation
• Callback handler for downstream processing
• Batch generation for marketing campaigns
• Post-processing pipeline with sharp

Resources

• Ideogram API Reference
• BullMQ Documentation
• sharp Image Processing

Next Steps

For performance optimization, see ideogram-performance-tuning.