fix: avoid nil session dereference in OpenAI local mode

[anxkhn]

What

The OpenAI agent executor crashes on every request when the A2A server is run
in local mode.

KAgentApp.build_local() constructs the executor with session_factory=None
(python/packages/kagent-openai/src/kagent/openai/_a2a.py), so execute()
leaves session as None. _stream_agent_events then built the
SessionContext with session.session_id unconditionally:

session_context = SessionContext(session_id=session.session_id)

That dereference raises AttributeError: 'NoneType' object has no attribute 'session_id' before the agent runs, so no local-mode request ever reaches the
model.

Fix

Use the persisted session id when a session exists. In local mode, check
context.session_id first and then fall back to context.context_id:

session_id = session.session_id if session else getattr(context, "session_id", None) or context.context_id
session_context = SessionContext(session_id=session_id)

This matches the identifier derivation in execute(), keeping metadata and the
SessionContext passed to Runner.run_streamed consistent.

Testing

Added python/packages/kagent-openai/tests/test_agent_executor.py, which builds
the executor with session_factory=None and drives execute():

• one test asserts the run reaches Runner.run_streamed (before the fix it
  fails with the AttributeError above),
• one test asserts the SessionContext id falls back to context.context_id,
• one test asserts context.session_id takes precedence when provided.

cd python
uv run pytest packages/kagent-openai/tests/test_agent_executor.py

[Copilot]

Pull request overview

Fixes a crash in the OpenAI A2A agent executor when running in local mode (where session_factory=None) by avoiding dereferencing session.session_id when session is None, and adds regression tests to ensure local-mode execution reaches the runner and uses a stable session identifier.

Changes:

• Update _stream_agent_events() to derive a SessionContext.session_id even when session is None.
• Add async tests covering execute() behavior when session_factory=None, including validating the SessionContext id fallback.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
python/packages/kagent-openai/src/kagent/openai/_agent_executor.py	Avoids None session dereference by computing a fallback session id for SessionContext.
python/packages/kagent-openai/tests/test_agent_executor.py	Adds regression tests for local-mode execution and SessionContext session id fallback behavior.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[anxkhn]

Pushed a revision addressing the latest review/CI feedback.