How to Prevent Development API Usage from Impacting Production?

Community
1

We have a production service used by customers and a separate development environment where we run tests and experiments.

Our concern is that development experiments or high-volume tests could impact production through shared organization-level rate limits or billing/usage limits.

Project-level budgets and alerts seem to be soft limits, not hard spending limits.

What is the recommended best practice to create hard separation between production and development usage?

Should we use separate OpenAI organizations for production and development, or is there another recommended way to enforce hard limits per environment/project?

2

Project hard spend limits are now supported: https://developers.openai.com/api/docs/guides/spend-limits. Enable Enforce a hard limit for the development project; enforcement can have a small delay. Separate projects still share organization-level limits, so review those too if production needs stronger isolation. Thanks for raising the use case!