OpenAI Python SDK 3.42.0 Adds Agent Context Compaction Controls
TL;DR
- OpenAI released Python SDK 3.42.0 on October 31, 2026, adding agent context compaction settings, session review, and experimental time awareness.
- The update introduces organization external key management and keeps callable API keys scoped to each request for tighter credential handling.
- WebSocket work spans recovery and delivery reporting improvements, warmup support, and a fix that preserves refusal output and authentication in helpers.
OpenAI shipped version 3.42.0 of its Python SDK on October 31, pushing agent-focused controls deeper into client code. The headline additions, per the release notes, are "add agent context compaction settings," "add agent session review and tool settings," and "add experimental agent time awareness."
Security and credentials get attention too. The notes list "add organization external key management" and promise to "keep callable API keys scoped to each request." Real-time telephony users pick up "add SIP DTMF commands and delivery events," while WebSocket work spans "improve WebSocket recovery and delivery reporting" and "support WebSocket warmup and preserve cleanup cancellation." One bug fix is called out: "preserve refusal output and authentication in WebSocket helpers."
It is one of a run of OpenAI entries in our tracker today. For agent builders, the compaction setting is the one to read first: summarization behavior that used to be opaque is now something you set in your own call.
Originally reported by github.com
Read the original article →Original headline: OpenAI Python SDK 3.42.0 Adds Agent Context Compaction, Session Review and Organization Key Management