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

SalesLoft Performance Tuning

Overview

Optimize SalesLoft REST API v2 performance. Key bottlenecks: deep pagination (cost multiplier), no batch endpoints, and per-minute rate limits. Solutions: caching, incremental sync, and pagination-aware request planning.

Latency Benchmarks

Instructions

Step 1: Response Caching

import { LRUCache } from 'lru-cache';

const cache = new LRUCache<string, any>({ max: 5000, ttl: 60_000 });

async function cachedGet<T>(endpoint: string, params?: Record<string, any>): Promise<T> {
  const key = `${endpoint}:${JSON.stringify(params || {})}`;
  const hit = cache.get(key);
  if (hit) return hit as T;

  const { data } = await api.get(endpoint, { params });
  cache.set(key, data);
  return data;
}

// Cache people lookups (frequent during cadence enrollment)
const person = await cachedGet('/people.json', { email_addresses: ['alex@co.com'] });

Step 2: Incremental Sync with updated_at

// Only fetch records changed since last sync
async function incrementalSync(lastSyncTime: string) {
  const updated: any[] = [];
  let page = 1;

  while (true) {
    const { data } = await api.get('/people.json', {
      params: {
        updated_at: { gt: lastSyncTime }, // ISO 8601
        per_page: 100,
        page,
        sort_by: 'updated_at',
        sort_direction: 'ASC',
      },
    });
    updated.push(...data.data);
    if (page >= data.metadata.paging.total_pages) break;
    page++;
  }

  return { updated, newSyncTime: new Date().toISOString() };
}

Step 3: Avoid Deep Pagination Cost

// Deep pages cost 3-30x. Instead of paginating all 25k records,
// use updated_at filter to get incremental changes
function shouldUseIncremental(totalCount: number): boolean {
  // If total records > 1000, incremental sync is more efficient
  // Full pagination of 250 pages = 910 cost points vs.
  // incremental of last 50 changes = 1 page = 1 point
  return totalCount > 1000;
}

Step 4: Connection Pooling

import { Agent } from 'https';

const agent = new Agent({
  keepAlive: true,
  maxSockets: 10,     // Max concurrent connections
  maxFreeSockets: 5,  // Keep idle connections alive
  timeout: 30_000,
});

const api = axios.create({
  baseURL: 'https://api.salesloft.com/v2',
  headers: { Authorization: `Bearer ${process.env.SALESLOFT_API_KEY}` },
  httpsAgent: agent,
});

Step 5: Parallel Safe Reads

// Parallelize independent reads (each costs 1 point)
const [people, cadences, activities] = await Promise.all([
  api.get('/people.json', { params: { per_page: 100 } }),
  api.get('/cadences.json', { params: { per_page: 50 } }),
  api.get('/activities/emails.json', { params: { per_page: 100 } }),
]);
// 3 points total, ~120ms parallel vs ~360ms sequential

Error Handling

Resources

• SalesLoft Rate Limits
• LRU Cache

Next Steps

For cost optimization, see salesloft-cost-tuning.