OpenAI Python SDK v3.36.0
OpenAI Python SDK v3.36.0 adds Vertex AI integration for the Responses API, 30-minute cache-write usage reporting, compaction trigger IDs, and custom voice consent management, alongside several WebSocket and Realtime API bug fixes.
Why this signal matters
The v3.36.0 release of the OpenAI Python SDK introduces several new API features and operational fixes. Key updates include Vertex AI support for the Responses API, support for 30-minute cache-write usage token counts, compaction trigger IDs, and custom voice consent management endpoints. On the networking and typing side, it adds a native synchronous WebSocket option type. Bug fixes address WebSocket cleanup on failure, interruption of sync WebSocket recovery attempts, header preservation in legacy Beta Realtime connections, Pydantic v1 alias validation, canonical batch response acceptance in strict mode, and Vertex API root retention across response lifecycles.
Actionable summary
OpenAI Python SDK release v3.36.0 introduces Vertex AI support for the Responses API, adds API types for compaction trigger IDs and 30-minute cache-write counts, enables custom voice consent management, and implements fixes for sync WebSocket option handling, connection cancellation, and legacy Beta Realtime headers.
- Agent usefulness
- 77/100
- Confidence
- 96%
- Canonical data
- JSON + Markdown
What builders should check
- Upgrade openai to version 3.36.0 via pip install --upgrade openai==3.36.0 .
- Verify WebSocket and Realtime client logic if relying on connection headers or recovery cancellation behavior.
- Adopt newly typed fields for cache-write usage counts, compaction trigger IDs, or Vertex AI Responses API endpoints if applicable.