Gamma Performance Tuning

Overview

Optimize Gamma API integration performance. Gamma's generate-poll-retrieve pattern means most latency is in generation time (10-60s), not API call overhead. Optimize by: reducing poll overhead, parallelizing batch operations, caching results, and choosing the right generation parameters.

Prerequisites

• Working Gamma integration (see gamma-sdk-patterns)
• Understanding of async patterns
• Redis or in-memory cache (recommended)

Performance Characteristics

Instructions

Step 1: Optimize Poll Strategy

// src/gamma/smart-poll.ts
// Adaptive polling: start fast, slow down over time

export async function smartPoll(
  gamma: GammaClient,
  generationId: string,
  opts = { maxTimeMs: 180000 }
): Promise<GenerateResult> {
  const deadline = Date.now() + opts.maxTimeMs;
  let interval = 2000; // Start at 2s

  while (Date.now() < deadline) {
    const result = await gamma.poll(generationId);

    if (result.status === "completed") return result;
    if (result.status === "failed") throw new Error("Generation failed");

    // Adaptive backoff: poll faster early, slower later
    await new Promise((r) => setTimeout(r, interval));
    interval = Math.min(interval * 1.5, 10000); // Max 10s between polls
  }

  throw new Error(`Poll timeout after ${opts.maxTimeMs}ms`);
}

Step 2: Cache Static Data

// src/gamma/cache.ts
import NodeCache from "node-cache";

const cache = new NodeCache({ stdTTL: 3600 }); // 1 hour for static data

export async function getCachedThemes(gamma: GammaClient) {
  const key = "gamma:themes";
  const cached = cache.get(key);
  if (cached) return cached;

  const themes = await gamma.listThemes();
  cache.set(key, themes);
  return themes;
}

export async function getCachedFolders(gamma: GammaClient) {
  const key = "gamma:folders";
  const cached = cache.get(key);
  if (cached) return cached;

  const folders = await gamma.listFolders();
  cache.set(key, folders);
  return folders;
}

// Cache generation results (useful for showing status)
export async function cacheGenerationResult(
  generationId: string,
  result: GenerateResult
) {
  cache.set(`gamma:gen:${generationId}`, result, 86400); // 24 hours
}

Step 3: Parallel Batch Generation

// src/gamma/batch.ts
import pLimit from "p-limit";

const limit = pLimit(3); // Max 3 concurrent generations

export async function batchGenerate(
  gamma: GammaClient,
  requests: Array<{ content: string; exportAs?: string }>
): Promise<Array<{ index: number; result?: GenerateResult; error?: string }>> {
  const results = await Promise.allSettled(
    requests.map((req, index) =>
      limit(async () => {
        const { generationId } = await gamma.generate({
          content: req.content,
          outputFormat: "presentation",
          exportAs: req.exportAs,
        });
        const result = await smartPoll(gamma, generationId);
        return { index, result };
      })
    )
  );

  return results.map((r, i) => {
    if (r.status === "fulfilled") return r.value;
    return { index: i, error: (r.reason as Error).message };
  });
}

Step 4: Reduce Generation Time

// Shorter content = faster generation
// "brief" text = fewer AI-generated words per card = faster

// SLOWER: extensive text on many cards
await gamma.generate({
  content: "Comprehensive 20-card guide to machine learning...",
  outputFormat: "presentation",
  textAmount: "extensive",  // More text per card = slower
});

// FASTER: brief text, fewer implied cards
await gamma.generate({
  content: "5-card overview of ML basics: supervised, unsupervised, reinforcement, deep learning, applications",
  outputFormat: "presentation",
  textAmount: "brief",      // Less text per card = faster
});

// FASTEST: preserve mode (no AI text generation)
await gamma.generate({
  content: "Your pre-written slide content here...",
  outputFormat: "presentation",
  textMode: "preserve",     // Uses your text as-is, no AI rewriting
});

Step 5: Preload Data at Startup

// src/gamma/preload.ts
// Fetch themes and folders at app startup, not per-request

let preloaded = false;

export async function preloadGammaData(gamma: GammaClient) {
  if (preloaded) return;

  const [themes, folders] = await Promise.all([
    gamma.listThemes(),
    gamma.listFolders(),
  ]);

  // Cache for the session
  cache.set("gamma:themes", themes, 0);   // No TTL (until restart)
  cache.set("gamma:folders", folders, 0);

  preloaded = true;
  console.log(`Preloaded ${themes.length} themes, ${folders.length} folders`);
}

Step 6: Connection Keep-Alive

// src/gamma/optimized-client.ts
import http from "node:http";
import https from "node:https";

// Reuse TCP connections
const agent = new https.Agent({
  keepAlive: true,
  maxSockets: 10,
  keepAliveMsecs: 60000,
});

export function createOptimizedClient(apiKey: string) {
  const base = "https://public-api.gamma.app/v1.0";
  const headers = { "X-API-KEY": apiKey, "Content-Type": "application/json" };

  async function request(method: string, path: string, body?: unknown) {
    const res = await fetch(`${base}${path}`, {
      method, headers,
      body: body ? JSON.stringify(body) : undefined,
      // @ts-ignore — agent support in Node.js
      agent,
    });
    if (!res.ok) throw new Error(`Gamma ${res.status}`);
    return res.json();
  }

  return {
    generate: (body: any) => request("POST", "/generations", body),
    poll: (id: string) => request("GET", `/generations/${id}`),
    listThemes: () => request("GET", "/themes"),
    listFolders: () => request("GET", "/folders"),
  };
}

Performance Targets

Error Handling

Resources

• Gamma API Parameters
• p-limit Documentation
• node-cache

Next Steps

Proceed to gamma-cost-tuning for credit optimization.