Performance

Keep-Alive: Persistent connection support. If you initialize AiClient = Client() once at the beginning and reuse it throughout, subsequent requests are processed faster thanks to TCP connection reuse.

Local Address: Added the local_address option, which in some cases speeds up DNS resolution and connection setup.

Automatic Retries: In case of temporary network failures or server errors (e.g., 5xx responses), requests are automatically retried several times, improving reliability.

Pre-warm: Introduced a mechanism that warms up the API in advance before main requests, reducing the latency of the first response.

RETRY STREAM: If an error occurs during streaming (stream=True), such as an SSL failure, the client automatically reconnects and retries the request without stopping the main code execution.

RESUME: Thanks to this feature, the AI continues generating text from where it left off. While there may be occasional hiccups, the output remains coherent and readable.