Imported from jeremylongshore/tons-of-skills-marketplace (plugins/saas-packs/adobe-pack/skills/adobe-performance-tuning/SKILL.md). Install upstream with npx skills add jeremylongshore/tons-of-skills-marketplace --skill adobe-performance-tuning. Copyright stays with the author (MIT).

Adobe Performance Tuning

Overview

Optimize Adobe API performance across Firefly Services, PDF Services, and Photoshop APIs. Key bottlenecks include IMS token generation, async job polling overhead, and cold-start latency on serverless platforms.

Prerequisites

• Adobe SDK installed and functional
• Understanding of which APIs your app uses most
• Redis or in-memory cache available (optional)
• Performance monitoring in place

Latency Benchmarks (Real-World)

Instructions

Optimization 1: Cache IMS Access Tokens (Biggest Win)

The IMS token endpoint returns tokens valid for 24 hours. Never re-generate per request:

// WRONG: generates new token every call (adds 200-500ms each time)
async function makeRequest() {
  const token = await getAccessToken(); // hits IMS every time
}

// RIGHT: cache token and only refresh when expiring
let tokenCache: { token: string; expiresAt: number } | null = null;

async function getCachedToken(): Promise<string> {
  if (tokenCache && tokenCache.expiresAt > Date.now() + 300_000) {
    return tokenCache.token; // Cache hit — 0ms
  }
  const res = await fetch('https://ims-na1.adobelogin.com/ims/token/v3', {
    method: 'POST',
    headers: { 'Content-Type': 'application/x-www-form-urlencoded' },
    body: new URLSearchParams({
      client_id: process.env.ADOBE_CLIENT_ID!,
      client_secret: process.env.ADOBE_CLIENT_SECRET!,
      grant_type: 'client_credentials',
      scope: process.env.ADOBE_SCOPES!,
    }),
  });
  const data = await res.json();
  tokenCache = { token: data.access_token, expiresAt: Date.now() + data.expires_in * 1000 };
  return tokenCache.token;
}

Optimization 2: Parallel Async Job Submission

Firefly and Photoshop APIs are async — submit all jobs first, then poll all:

// SLOW: sequential (total = sum of all job times)
for (const prompt of prompts) {
  const result = await generateImageSync(prompt); // 5-20s each
}

// FAST: parallel submit + parallel poll (total = max job time)
async function batchFireflyGenerate(prompts: string[]) {
  const token = await getCachedToken();

  // 1. Submit all jobs simultaneously
  const jobSubmissions = await Promise.all(
    prompts.map(prompt =>
      fetch('https://firefly-api.adobe.io/v3/images/generate-async', {
        method: 'POST',
        headers: {
          'Authorization': `Bearer ${token}`,
          'x-api-key': process.env.ADOBE_CLIENT_ID!,
          'Content-Type': 'application/json',
        },
        body: JSON.stringify({ prompt, n: 1, size: { width: 1024, height: 1024 } }),
      }).then(r => r.json())
    )
  );

  // 2. Poll all jobs in parallel
  const results = await Promise.all(
    jobSubmissions.map(job => pollUntilDone(job.statusUrl, token))
  );

  return results;
}

Optimization 3: Response Caching for Repeated Operations

import { LRUCache } from 'lru-cache';

// Cache PDF extraction results (same PDF = same output)
const extractionCache = new LRUCache<string, any>({
  max: 100,
  ttl: 3600_000, // 1 hour
});

async function cachedPdfExtract(pdfHash: string, pdfPath: string) {
  const cached = extractionCache.get(pdfHash);
  if (cached) {
    console.log('PDF extraction cache hit');
    return cached;
  }

  const result = await extractPdfContent(pdfPath);
  extractionCache.set(pdfHash, result);
  return result;
}

Optimization 4: Connection Keep-Alive

import { Agent } from 'https';

// Reuse TCP connections to Adobe endpoints
const adobeAgent = new Agent({
  keepAlive: true,
  maxSockets: 10,
  maxFreeSockets: 5,
  timeout: 60_000,
});

// Use with node-fetch or undici
const response = await fetch(url, {
  // @ts-ignore — agent option supported by node-fetch
  agent: adobeAgent,
  headers: { ... },
});

Optimization 5: Smart Polling Intervals

// Adaptive polling: start fast, slow down over time
async function adaptivePoll(statusUrl: string, token: string) {
  const intervals = [1000, 2000, 3000, 5000, 5000, 10000]; // ms
  let attempt = 0;

  while (true) {
    const res = await fetch(statusUrl, {
      headers: {
        'Authorization': `Bearer ${token}`,
        'x-api-key': process.env.ADOBE_CLIENT_ID!,
      },
    });
    const status = await res.json();

    if (status.status === 'succeeded') return status;
    if (status.status === 'failed') throw new Error(status.error?.message);

    const delay = intervals[Math.min(attempt, intervals.length - 1)];
    await new Promise(r => setTimeout(r, delay));
    attempt++;
  }
}

Output

• IMS token cached for 24h (eliminates 200-500ms per request)
• Parallel job submission for batch operations
• LRU response caching for repeated extractions
• Connection keep-alive reducing TLS handshake overhead
• Adaptive polling reducing unnecessary API calls

Error Handling

Examples

Start with the smallest applicable command or code example already provided in this guide, using a non-production Adobe environment and credentials. Confirm the documented response or validation result before applying the pattern to production.

Resources

• Firefly Async API Guide
• PDF Services Quickstart
• LRU Cache npm

Next Steps

For cost optimization, see adobe-cost-tuning.