API Integration

Process

1. Read the docs — find auth method, rate limits, pagination, error codes
2. Auth first — get a working authenticated request before anything else
3. Build incrementally — one endpoint at a time, verify response shape
4. Handle failures — retries with backoff, timeout handling, error parsing
5. Extract — parse response into clean data structures

Patterns

REST client

• Use requests/httpx (Python) or fetch (JS) — don't build custom HTTP
• Set timeouts on every request (10s default, 30s for large payloads)
• Retry 3x with exponential backoff (1s, 2s, 4s) on 429/500/502/503
• Log request/response on failure, not success

Rate limiting

• Check response headers: X-RateLimit-Remaining, Retry-After
• Implement token bucket or sliding window
• Never hammer an API — respect limits, your key depends on it

Error handling

• Parse error response bodies — they usually tell you what's wrong
• Map API errors to your domain errors
• Never swallow errors silently

Rules

• Never hardcode API keys — use environment variables
• Always set User-Agent header
• Cache responses when the data doesn't change frequently
• Document the API's quirks (undocumented behaviors, known bugs)