> ## Documentation Index > Fetch the complete documentation index at: https://xchangeapi.com/docs/llms.txt > Use this file to discover all available pages before exploring further. # Developer Best Practices > Best practices for integrating with xchangeapi.com REST and WebSocket APIs To ensure high performance, reliability, and security of your financial applications, we recommend following these industry-standard best practices when integrating with the xchangeapi.com REST and WebSocket APIs. ## 1. Authentication & Security Treat your API keys as sensitive credentials. * **Environment Variables**: Never hardcode your API keys in your source code. Use environment variables (e.g., `.env` files) or secrets management services (AWS Secrets Manager, HashiCorp Vault). * **Server-Side Execution**: For web applications, always proxy requests through a backend server. Exposing API keys in frontend (client-side) code allows third parties to hijack your quota. * **Key Rotation**: Regularly rotate your API keys and immediately revoke any keys that may have been compromised. ## 2. Optimizing API Requests Efficiently managing your request volume reduces latency and prevents rate-limiting issues. * **Use Batch Endpoints**: Whenever possible, use batch requests to fetch data for multiple symbols in a single HTTP call rather than iterating through individual tickers. * **Filtering and Pagination**: Use the `limit` and `offset` (or cursor-based pagination) parameters to fetch only the data you need. * **Conditional Requests**: Utilize E-Tag or Last-Modified headers to avoid re-downloading data that hasn't changed since your last request. ## 3. Resilience and Error Handling Financial markets are volatile; your integration should not be. * **Implement Exponential Backoff**: If you receive a 429 Too Many Requests or 5xx Server Error, do not immediately retry. Use an exponential backoff strategy (retrying after 1s, 2s, 4s, etc.) to allow the system to recover. * **Graceful Degradation**: Design your application to handle missing data points or delayed updates. Use stale data with a warning flag if real-time data is momentarily unavailable. * **Timeouts**: Set reasonable client-side timeouts (e.g., 5-10 seconds for REST calls) to prevent your application threads from hanging indefinitely. ## 4. Working with Real-time Data (WebSockets) For high-frequency trading or live dashboards, WebSockets provide the lowest latency. * **Connection Heartbeats**: Implement a heartbeat (ping/pong) mechanism to detect "zombie" connections that are technically open but no longer receiving data. * **Message Buffering**: Ensure your application can process incoming messages faster than the stream's throughput. If your processing logic is heavy, use a message queue (like RabbitMQ or Redis) to decouple data ingestion from processing. * **Reconnection Logic**: Implement automatic reconnection with a jittered backoff to avoid a "thundering herd" effect on the server during network flickers. ## 5. Data Integrity and Validation * **Precision Handling**: Financial values should never be handled as floating-point numbers due to rounding errors. Use decimal-specific libraries (e.g., `Decimal` in Python, `BigDecimal` in Java) for all calculations. * **Timestamp Synchronization**: All xchangeapi.com timestamps are in UTC. Ensure your local environment is synchronized with a reliable NTP server to accurately track market events. * **Schema Validation**: Use tools like JSON Schema or Protobuf to validate incoming API responses, ensuring your downstream logic doesn't break if a non-critical field is added to the payload. ## 6. Caching Strategy Reduce costs and improve UI responsiveness. * **Cache Static Data**: Symbols, exchange lists, and historical OHLC data that is older than 24 hours should be cached locally. * **Cache-Control Headers**: Always respect Cache-Control headers returned by our API to ensure you are not serving outdated financial information. **Pro-Tip for Quants** If you are performing backtesting, we recommend downloading historical datasets as CSV rather than making millions of individual REST requests. This is significantly more performant for large-scale vector operations in libraries like Pandas or NumPy.