Release 0.8.0

[github-actions]

Release readiness review (v0.7.0 -> TARGET ef2fc15)

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

Diff

v0.7.0...ef2fc15

Release call:

🟢 GREEN LIGHT TO SHIP Large runtime refactor and new HITL/MCP features, but no concrete breaking regressions found in diff; proceed with targeted validation.

Scope summary:

• 243 files changed (+33,221/-6,072); key areas touched: src/agents/run.py, src/agents/run_internal/*, src/agents/run_state.py, MCP/memory/realtime modules, extensive new/updated tests, examples, and docs.

Risk assessment (ordered by impact):

1. Runtime orchestration refactor (run loop + state)

   • Risk: 🟡 MODERATE. Changes in core execution and state serialization could regress streaming, resume, or guardrail behavior for some workflows.
   • Evidence: Removal of src/agents/_run_impl.py and large rewrite into src/agents/run_internal/* plus major edits to src/agents/run.py and new src/agents/run_state.py.
   • Files: src/agents/run.py, src/agents/run_internal/run_loop.py, src/agents/run_internal/turn_resolution.py, src/agents/run_state.py
   • Action: Run targeted runtime tests: uv run pytest -k "run_state or run_impl_resume_paths or agent_runner_streamed or run_context_approvals"; pass when all selected tests succeed.

2. Human-in-the-loop approvals and session persistence

   • Risk: 🟡 MODERATE. New approval flows and persistence paths could change tool execution gating or resume semantics in edge cases.
   • Evidence: New approvals modules and HITL-focused tests/examples (e.g., src/agents/run_internal/approvals.py, tests/test_hitl_*, examples/agent_patterns/human_in_the_loop*).
   • Files: src/agents/run_internal/approvals.py, src/agents/util/_approvals.py, tests/test_hitl_error_scenarios.py
   • Action: Run HITL-focused tests: uv run pytest -k "hitl or approvals"; pass when all selected tests succeed.

3. MCP server/util behavior changes

   • Risk: 🟡 MODERATE. MCP server/util changes and tool timeout handling could affect integrations relying on MCP tool lifecycle or retries.
   • Evidence: Large edits in MCP server/util plus new MCP tests and examples.
   • Files: src/agents/mcp/server.py, src/agents/mcp/util.py, tests/mcp/test_mcp_util.py
   • Action: Run MCP suite: uv run pytest -k "mcp"; pass when all selected tests succeed.

Notes:

• BASE tag determined locally: v0.7.0; TARGET is current HEAD ef2fc15e28deb9a999b6eec7b2472df75b2ba2eb.
• Local verification not run in this review; assume CI coverage unless otherwise indicated.