I noticed an interesting update from OpenAI — they have added support for WebSocket in the API responses, and this looks like a pretty significant improvement for developers.

The essence is that WebSocket allows establishing persistent connections instead of traditional HTTP requests. This is especially useful when working with frequent tool calls — the system speeds up by about 40% for long chains with more than 20 calls. Agree, that's a pretty good performance boost.

Technically, WebSocket supports incremental data input and is compatible with Zero Data Retention (ZDR), which is important for low latency during context reconnection. That means developers get not just speed but also flexibility in managing data.

There is, however, a limitation — each connection is limited to 60 minutes. But for most workflows, that's enough. It seems OpenAI is seriously working on optimizing their API, and this will definitely make life easier for those working with complex automations and multi-step tasks.

If you're a developer working with the OpenAI API, it's worth checking the documentation — WebSocket can significantly improve your application's performance.