Best Practices

Follow these recommendations to get the most out of routing.run.

• Handle 429 responses with exponential backoff.
• Set max_tokens to avoid runaway generation costs.
• Always check the finish_reason field to know if the response was truncated.

• Pick the smallest model that meets your quality bar.
• Use shorter prompts where possible to reduce token usage.
• Monitor usage on the dashboard to track spending.

• Rotate API keys you have not used recently.
• Use separate keys for each environment.
• Never commit API keys to version control.