3.0.6

Added support for Python 3.14 (with-GIL).

Added the timeout and netTimeout/net_timeout parameters (supporting datetime.timedelta) to Session.connect and DBConnection.connect to configure the timeout for waiting for server responses, as well as network timeouts for TCP connection establishment, data writing, and keep-alive checks, respectively. The legacy readTimeout and writeTimeout parameters are retained for backward compatibility.

Added the usePublicName parameter to Session.connect to support connecting via the node's publicName in cluster environments.

Added sequence numbers to script and function requests in high availability mode or when auto-reconnect is enabled, enabling the server to distinguish between initial and retry requests.

Optimized the failover process: removed the redundant 1-second delay after successfully connecting to a candidate node.

Fixed an issue where disconnection duringrun and upload could cause failover to an incorrect node, and reconnection logs were missing after a successful reconnect.

Fixed an issue where high-availability connections failed to continue attempting node failover when the cluster returned<UnknownLeader>.

Fixed the node matching and <NotLeader> redirection issues when usePublicName was enabled in high availability connections.

The default value of keepAliveTime in Session is changed from 30 seconds to unspecified. When unspecified, the global default network timeout (currently 3 seconds) is applied.

Session.setTimeout now sets the global default network timeout (net_timeout), rather than mapping to the Linux TCP_USER_TIMEOUT.

Added support for uploading pyarrow.Table objects via the PROTOCOL_ARROW protocol.

Added support for pandas 3.0.