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

AppFolio Performance Tuning

Overview

AppFolio's property management API handles bulk tenant queries, property portfolio pagination, and work order batch processing. Large portfolios with thousands of units generate heavy read traffic on listing endpoints. Optimizing cache lifetimes for slow-changing property data, batching work order updates, and pooling HTTP connections reduces API call volume by 60-80% and cuts dashboard load times from seconds to sub-second.

Prerequisites

• A measured baseline for latency, call volume, error rate, payload size, cache hit behavior, and data freshness for the specific permitted endpoint.
• Endpoint-specific rate/concurrency limits, a request budget, and a data policy that excludes tenant, payment, and raw response payloads from generic caches.
• Synthetic fixtures and a rollback feature flag for validating performance changes without altering production read/write semantics.

Instructions

1. Start with the smallest safe read and capture a baseline before changing caching, concurrency, pagination, or connection settings.
2. Cache only minimized data under a bounded entry/byte policy and invalidate on known writes; show stale age to callers where decisions need freshness.
3. Respect the smallest endpoint limit, preserve cursors, and stop parallel batches before they turn a rate-limit signal into a retry storm.
4. Promote only when performance improves without changing result completeness, authorization, PII boundaries, or write/idempotency behavior; roll back on any correctness regression.

Caching Strategy

const cache = new Map<string, { data: unknown; expiry: number }>();
const MAX_CACHE_ENTRIES = 1_000;
const TTL = { properties: 300_000, tenants: 120_000, units: 300_000, workOrders: 60_000 };

async function cached(key: string, ttlKey: keyof typeof TTL, fn: () => Promise<any>) {
  const entry = cache.get(key);
  if (entry && entry.expiry > Date.now()) return entry.data;
  const data = await fn();
  if (!cache.has(key) && cache.size >= MAX_CACHE_ENTRIES) cache.delete(cache.keys().next().value!);
  cache.set(key, { data, expiry: Date.now() + TTL[ttlKey] });
  return data;
}

Batch Operations

async function batchWorkOrders(client: any, ids: string[], batchSize = 25) {
  const results = [];
  for (let i = 0; i < ids.length; i += batchSize) {
    const batch = ids.slice(i, i + batchSize);
    const res = await Promise.all(batch.map(id => client.http.get(`/work_orders/${id}`)));
    results.push(...res.map(r => r.data));
    if (i + batchSize < ids.length) await new Promise(r => setTimeout(r, 200));
  }
  return results;
}

Connection Pooling

import { Agent } from 'https';
const agent = new Agent({ keepAlive: true, maxSockets: 10, maxFreeSockets: 5, timeout: 30_000 });
// Pass to axios/fetch: { httpsAgent: agent }

Rate Limit Management

async function withRateLimit(fn: () => Promise<any>): Promise<any> {
  const res = await fn();
  const remaining = parseInt(res.headers['x-ratelimit-remaining'] || '100');
  if (remaining < 5) {
    const retryAfter = parseInt(res.headers['retry-after'] || '2') * 1000;
    await new Promise(r => setTimeout(r, retryAfter));
  }
  return res;
}

Monitoring

const metrics = { apiCalls: 0, cacheHits: 0, errors: 0, totalLatency: 0 };
function track(startMs: number, hit: boolean, error?: boolean) {
  metrics.apiCalls++; metrics.totalLatency += Date.now() - startMs;
  if (hit) metrics.cacheHits++; if (error) metrics.errors++;
}
// Log: avg latency, cache hit rate, error rate per minute

Performance Checklist

• Cache property and unit listings with 5-min TTL
• Use incremental sync via last_modified timestamps
• Batch work order updates in groups of 25
• Enable HTTP keep-alive with connection pooling
• Parse rate limit headers and back off proactively
• Parallelize independent dashboard queries with Promise.all
• Monitor cache hit ratio (target > 70%)
• Set request timeouts to 30s to avoid hung connections

Error Handling

Output

• A baseline-to-candidate comparison of latency, call volume, cache hit rate, error rate, and result completeness
• A bounded/minimized cache and endpoint-specific concurrency policy
• A rollout or rollback decision with a freshness and correctness receipt

Examples

For a property-dashboard regression, benchmark one synthetic portfolio page, then enable a bounded property-summary cache behind a feature flag. Compare p95 latency, request count, cache hits, and returned IDs before and after the change. Confirm that an authorized write invalidates the affected entry and that a full rate-limit response pauses new work. If the candidate yields stale, partial, unauthorized, or differently ordered results, disable the flag and reconcile before trying another optimization.

Resources

• AppFolio Stack APIs
• AppFolio Engineering Blog

Next Steps

See appfolio-reference-architecture.