Perplexity API Rate Limiting and Error Handling Rules

When building applications with the Perplexity API, follow these operational rules:

1. RATE LIMITS: Respect API rate limits. Implement exponential backoff starting at 1 second, doubling up to 32 seconds maximum. Track rate limit headers in responses.

2. TIMEOUT CONFIGURATION: Set timeouts appropriately per model: sonar/sonar-pro: 30-60 seconds. sonar-reasoning-pro: 60-120 seconds. sonar-deep-research: 180-300 seconds (research can take several minutes).

3. ERROR HANDLING: Handle these error codes: 400 (bad request - check input format), 401 (invalid API key), 403 (rate limited - backoff), 429 (too many requests - queue and retry), 500 (server error - retry with backoff), 503 (service unavailable - retry after delay).

4. TOKEN MANAGEMENT: Monitor token usage per request. Set max_tokens appropriately for your use case. Track costs by multiplying input/output tokens by per-model pricing.

5. CACHING: Cache responses for identical queries within a reasonable TTL (5-15 minutes for real-time data, 1-24 hours for stable information). Use content hashing for cache keys.

6. FALLBACK: Implement model fallback chains: Try sonar-pro first, fall back to sonar on error or timeout. For deep research, fall back to pro-search if sonar-deep-research times out.

7. LOGGING: Log all API calls with: timestamp, model used, token count, latency, and response status. This enables cost tracking and performance monitoring.