Release 0.8.0

[github-actions]

Release readiness review (v0.7.0 -> TARGET 93cffe1)

This is a release readiness report done by $final-release-review skill.

Diff

v0.7.0...93cffe1

Release call:

🟢 GREEN LIGHT TO SHIP Major refactor and new features look well-covered by tests, with no explicit breaking-change flags beyond the 0.8.0 bump.

Scope summary:

• 199 files changed (+30451/-5987); key areas touched: runtime refactor (src/agents/run_internal), run state/persistence, HITL/approvals, MCP, memory sessions, realtime, tests, docs/examples.

Risk assessment (ordered by impact):

1. Runner refactor and new run_internal orchestration

   • Risk: 🟡 MODERATE. Large structural changes in the core run loop can introduce subtle behavioral drift between streamed/non-streamed paths or error handling.
   • Files: src/agents/run.py, src/agents/run_internal/run_loop.py, src/agents/run_internal/turn_resolution.py, src/agents/run_internal/tool_execution.py, src/agents/result.py, src/agents/items.py, tests/test_agent_runner.py, tests/test_agent_runner_streamed.py

2. RunState + session persistence changes

   • Risk: 🟡 MODERATE. New serialization/persistence paths may affect resume/rewind behavior or compatibility for persisted sessions.
   • Files: src/agents/run_state.py, src/agents/run_internal/session_persistence.py, src/agents/memory/session_settings.py, src/agents/memory/openai_conversations_session.py, tests/test_run_state.py, tests/test_run_impl_resume_paths.py

3. HITL/approvals and tool flow adjustments

   • Risk: 🟡 MODERATE. Approval gating and tool execution logic changes can alter tool lifecycle or user approval semantics, especially in MCP flows.
   • Files: src/agents/run_internal/approvals.py, src/agents/util/_approvals.py, src/agents/tool.py, src/agents/items.py, src/agents/mcp/server.py, tests/test_hitl_error_scenarios.py, tests/mcp/test_mcp_approval.py

4. Realtime and MCP behavior tweaks

   • Risk: 🟢 LOW. Focused fixes and additions (timeouts, tracing, meta resolvers) are likely safe but should be watched for edge cases in streaming or retry behavior.
   • Files: src/agents/realtime/openai_realtime.py, src/agents/realtime/session.py, src/agents/mcp/util.py, tests/realtime/test_session.py, tests/mcp/test_mcp_util.py

Notes:

• BASE_TAG v0.7.0, TARGET 93cffe1f54a56f36fde411da9ce9755a8f210f34.
• No local verification run here; assuming CI and $code-change-verification have already passed for this target.