Changelog
Approved changes only. Pending proposals are private to maintainers.
Refresh: record post-v0.22.0 tool/model bug fixes (#4735 variadic fixed-length tuple rejection, #4728 Chat Completions streamed-turn message merge)
Refresh openai-agents-python graph for v0.22.0 + post-v0.22.0 main (BASE 2c55603..HEAD f1a806a)
Refresh openai-agents-python graph for 23da2b6..2c55603: adds v0.20.0/v0.21.0/v0.21.1 releases, the new agents.testing scripted test-double surface, model call timeouts, run-scoped sandbox working directories, and OpenAI Python 3 / HTTPX 2 compatibility.
Refresh openai-agents-python graph with post-v0.19.4 unreleased changes (2b115b6..23da2b6): retry unsafe-replay approval, durable pending input, realtime GA transcription, mount credential acks, and assorted model/mcp/voice/sandbox fixes.
Refresh OpenAI Agents SDK graph for post-v0.19.4 commits ece7b0e..2b115b6: add mount-security and tool-invocation-identity nodes; update models (gpt-5.6-luna default, raw usage), sandbox, mcp, memory, realtime, tools, HITL, guardrails.
Refresh openai-agents-python: append post-v0.19.4 (unreleased) fixes to memory, models, voice, sandbox, runner, and HITL nodes; advance sync to ece7b0e.
Refresh openai-agents-python: add v0.19.3 & v0.19.4 releases; update MCP (v1/v2 SDK support) and memory (run context to custom sessions) features; advance sync to v0.19.4+ (b47a0e4)
Refresh openai-agents-python graph for post-v0.19.2 fixes (e943ded..648c252): tool-name collision policy + tool-identity helper, and model/runner/handoff/tracing/sandbox/memory fixes.