Imported from jeremylongshore/tons-of-skills-marketplace (skills/.curated/bamboohr-performance-tuning/SKILL.md). Install upstream with npx skills add jeremylongshore/tons-of-skills-marketplace --skill bamboohr-performance-tuning. Copyright stays with the author (MIT).

BambooHR Performance Tuning

Overview

Optimize BambooHR API performance through request reduction, caching, incremental sync, and connection pooling. The biggest wins come from eliminating N+1 query patterns using custom reports and the changed-since endpoint.

Prerequisites

• BambooHR API client configured
• Redis or in-memory cache available (optional)
• Performance monitoring in place

Instructions

Step 1: Eliminate N+1 Queries with Custom Reports

The single biggest performance improvement: use POST /reports/custom instead of individual employee GETs.

// BAD: 501 API calls for 500 employees
const dir = await client.getDirectory();                      // 1 call
for (const emp of dir.employees) {
  await client.getEmployee(emp.id, ['salary', 'hireDate']);   // 500 calls
}

// GOOD: 1 API call for all employees with all needed fields
const report = await client.customReport([
  'firstName', 'lastName', 'department', 'jobTitle',
  'hireDate', 'workEmail', 'status', 'location',
  'supervisor', 'employeeNumber',
]);
// 1 call, returns all employees with all fields

Performance impact: 500x reduction in API calls. Custom reports return all active employees in one request.

Step 2: Incremental Sync with Changed-Since

import { readFileSync, writeFileSync } from 'fs';

const LAST_SYNC_FILE = '.bamboohr-last-sync';

async function incrementalSync(client: BambooHRClient): Promise<string[]> {
  // Read last sync timestamp
  let lastSync: string;
  try {
    lastSync = readFileSync(LAST_SYNC_FILE, 'utf-8').trim();
  } catch {
    lastSync = new Date(Date.now() - 24 * 60 * 60 * 1000).toISOString(); // Default: 24h ago
  }

  // GET /employees/changed/?since=... — returns only changed employee IDs
  const changed = await client.request<{
    employees: Record<string, { id: string; lastChanged: string }>;
  }>('GET', `/employees/changed/?since=${lastSync}`);

  const changedIds = Object.keys(changed.employees || {});
  console.log(`${changedIds.length} employees changed since ${lastSync}`);

  if (changedIds.length === 0) return [];

  // Fetch only changed employees' details
  // For large sets, use custom report with filter; for small sets, individual GETs
  if (changedIds.length > 20) {
    // Bulk: use custom report (returns all, then filter client-side)
    const report = await client.customReport([
      'firstName', 'lastName', 'department', 'status',
    ]);
    const changedData = report.employees.filter(e =>
      changedIds.includes(e.id?.toString()),
    );
    // Process changedData...
  } else {
    // Small set: individual GETs are fine
    for (const id of changedIds) {
      const emp = await client.getEmployee(id, ['firstName', 'lastName', 'department', 'status']);
      // Process emp...
    }
  }

  // Save sync timestamp
  writeFileSync(LAST_SYNC_FILE, new Date().toISOString());
  return changedIds;
}

Also available for table data:

// GET /employees/changed/tables/{tableName}?since=...
const changedJobs = await client.request<any>(
  'GET', `/employees/changed/tables/jobInfo?since=${lastSync}`,
);
// Returns { employees: { "123": { lastChanged: "..." }, ... } }

Step 3: Response Caching

import { LRUCache } from 'lru-cache';

// BambooHR directory data changes infrequently — cache aggressively
const cache = new LRUCache<string, any>({
  max: 500,
  ttl: 5 * 60 * 1000, // 5 minutes for directory data
});

async function cachedRequest<T>(
  key: string,
  fetcher: () => Promise<T>,
  ttlMs?: number,
): Promise<T> {
  const cached = cache.get(key) as T | undefined;
  if (cached) {
    console.log(`Cache hit: ${key}`);
    return cached;
  }

  const result = await fetcher();
  cache.set(key, result, { ttl: ttlMs });
  return result;
}

// Usage
const directory = await cachedRequest(
  'directory',
  () => client.getDirectory(),
  5 * 60 * 1000, // Cache for 5 min
);

// Single employee — shorter cache
const employee = await cachedRequest(
  `employee:${id}`,
  () => client.getEmployee(id, fields),
  60 * 1000, // Cache for 1 min
);

Redis caching for multi-instance deployments:

import Redis from 'ioredis';

const redis = new Redis(process.env.REDIS_URL);

async function redisCached<T>(
  key: string,
  fetcher: () => Promise<T>,
  ttlSec = 300,
): Promise<T> {
  const cached = await redis.get(`bamboohr:${key}`);
  if (cached) return JSON.parse(cached);

  const result = await fetcher();
  await redis.setex(`bamboohr:${key}`, ttlSec, JSON.stringify(result));
  return result;
}

// Invalidate on webhook
async function invalidateCache(employeeId: string) {
  await redis.del(`bamboohr:employee:${employeeId}`);
  await redis.del('bamboohr:directory'); // Directory includes this employee
}

Step 4: Connection Pooling

import { Agent } from 'https';

// Reuse TCP connections for BambooHR API calls
const keepAliveAgent = new Agent({
  keepAlive: true,
  maxSockets: 5,        // Max 5 parallel connections
  maxFreeSockets: 2,
  timeout: 30_000,
  keepAliveMsecs: 10_000,
});

// Pass to fetch via undici or node-fetch
// For native fetch in Node 20+, connection pooling is automatic

Step 5: Request Batching with DataLoader

import DataLoader from 'dataloader';

// Batch individual employee GETs into a custom report
const employeeLoader = new DataLoader<string, Record<string, string>>(
  async (ids) => {
    // One custom report instead of N individual GETs
    const report = await client.customReport([
      'id', 'firstName', 'lastName', 'department', 'jobTitle',
    ]);

    const byId = new Map(report.employees.map(e => [e.id, e]));
    return ids.map(id => byId.get(id) || new Error(`Employee ${id} not found`));
  },
  {
    maxBatchSize: 100,
    batchScheduleFn: cb => setTimeout(cb, 50), // Batch window: 50ms
    cache: true,
  },
);

// Usage — automatically batched into one API call
const [emp1, emp2, emp3] = await Promise.all([
  employeeLoader.load('1'),
  employeeLoader.load('2'),
  employeeLoader.load('3'),
]);

Step 6: Performance Monitoring

class BambooHRMetrics {
  private requests: { duration: number; status: number; endpoint: string }[] = [];

  record(endpoint: string, status: number, durationMs: number) {
    this.requests.push({ duration: durationMs, status, endpoint });

    // Keep last 1000 requests
    if (this.requests.length > 1000) this.requests.shift();
  }

  summary() {
    const durations = this.requests.map(r => r.duration).sort((a, b) => a - b);
    const errors = this.requests.filter(r => r.status >= 400);

    return {
      totalRequests: this.requests.length,
      errorRate: (errors.length / Math.max(this.requests.length, 1) * 100).toFixed(1) + '%',
      p50: durations[Math.floor(durations.length * 0.5)] || 0,
      p95: durations[Math.floor(durations.length * 0.95)] || 0,
      p99: durations[Math.floor(durations.length * 0.99)] || 0,
      topEndpoints: this.topEndpoints(),
    };
  }

  private topEndpoints() {
    const counts = new Map<string, number>();
    for (const r of this.requests) {
      counts.set(r.endpoint, (counts.get(r.endpoint) || 0) + 1);
    }
    return [...counts.entries()].sort((a, b) => b[1] - a[1]).slice(0, 5);
  }
}

Output

• N+1 queries eliminated via custom reports (500x reduction)
• Incremental sync using changed-since endpoint
• Multi-tier caching (LRU in-memory + Redis)
• Connection pooling with keep-alive
• DataLoader-based request batching
• Performance metrics with p50/p95/p99

Performance Reference

Examples

Benchmark a sandbox or synthetic workload using the minimum approved fields, establish a latency and freshness baseline, and alter one cache, queue, or batch variable behind a rollback flag. Encrypt cached HR data, expire it under policy, and disable the optimization if reconciliation or authorization evidence diverges.

Error Handling

Resources

• BambooHR API Technical Overview
• DataLoader Documentation
• LRU Cache Documentation

Next Steps

For cost optimization, see bamboohr-cost-tuning.