[igalshilman]

The agent extras move to the frameworks' latest releases, and the packages they pull in to theirs:

	from	to
google-adk	2.14.0	2.14.1
pydantic-ai-slim / pydantic-ai-harness	2.66.0 / 0.66.0	2.68.0 / 0.68.0
langchain	1.4.9	1.5.0
langgraph	1.2.20	1.2.21
langchain-core	1.7.6	1.8.0
openai	3.45.0	3.50.0
mcp (via openai-agents)	1.23.2	2.3.0

The OpenAI Agents SDK is already at its latest (0.24.0). Each extra's minimum is the version tested. google-genai stays at 2.34.0: ADK caps it below 3.

• The harness's ConversationSearch now defaults to the conversation's scope and warns unless given one: history_source's example passes scope="conversation", which finds the agent's turns.
• basedpyright's typeshed deprecates Iterator return types on context managers: the two in restate.agents return Generators.
• uv.lock is now lockfile revision 5 (current uv).

[pullfrog]

✅ No new issues found.

Reviewed changes

I reviewed the dependency bump for the agent extras, plus the code and test changes it needed.

• Extras minimums and uv.lock: the extras now require the tested framework versions, and the lock moves to revision 5. uv lock --check passes. mcp 2.3.0 comes in through openai-agents (it allows mcp<3), and no repo code imports mcp directly.
• ConversationSearch scope: harness 0.68 changed the default scope from all to conversation and now warns when scope isn't given. The conversation scope only works if the run has a conversation_id. The integration already passes conversation_id_of(ctx) to the agent run, and TurnsHistorySource puts the same id on each turn's run, so passing scope="conversation" in the docstring example and the test still finds the turns.
• Context-manager return types: transaction and running are now annotated as AsyncGenerator / Generator. Nothing changes at runtime.

Checked locally after syncing the venv to the PR's lock:

• uv run pyright python/: 0 errors
• ruff check: clean
• pytest tests/* -m "not server": 436 passed

The Docker-harness tests can't run here, so CI covers those.

  ^{｜ View workflow run ｜ Using claude-opus-5-5 ｜ 𝕏}

[github-actions]

Test Results

  8 files  ±0    8 suites  ±0   3m 48s ⏱️ +12s
 71 tests ±0   71 ✅ ±0  0 💤 ±0  0 ❌ ±0 
322 runs  ±0  322 ✅ ±0  0 💤 ±0  0 ❌ ±0 

Results for commit e84854a. ± Comparison against base commit cdda80d.