fix: suppress repetitive A2A experimental mode warnings

[optimus-fulcria]

Summary

• Suppresses the repetitive [EXPERIMENTAL] RemoteA2aAgent UserWarning that floods agent logs during A2A operations
• Adds warnings.filterwarnings() in both types.py and _agent_executor.py to filter upstream ADK experimental warnings at module import time
• Uses a regex pattern (\[EXPERIMENTAL\].*A2A) to catch all A2A-related experimental warnings

Fixes #1379

Test plan

• Deploy agents with A2A tools and verify logs no longer contain experimental mode warnings
• Verify A2A functionality still works correctly (warnings are suppressed, not the functionality)

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Suppresses repetitive upstream ADK A2A “experimental mode” warnings that spam logs during A2A operations (fixes #1379).

Changes:

• Adds warnings.filterwarnings() in kagent.adk.types to ignore A2A-related [EXPERIMENTAL] warnings.
• Adds the same warning filter in kagent.adk._agent_executor prior to importing upstream A2A executor components.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.

File	Description
python/packages/kagent-adk/src/kagent/adk/types.py	Installs a module-import-time warnings filter intended to suppress ADK A2A experimental warnings.
python/packages/kagent-adk/src/kagent/adk/_agent_executor.py	Installs the same warnings filter before importing upstream google.adk.a2a executor types.

---

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

[Copilot]

warnings.filterwarnings() runs at module import time and mutates the global warnings filter for any consumer that imports kagent.adk.types (including via kagent.adk.AgentConfig). This can unexpectedly suppress warnings outside of A2A usage and makes behavior hard to opt out of. Consider scoping the filter (e.g., category=UserWarning + module= for the upstream ADK A2A modules) and/or gating it behind an explicit opt-in (env var / config) or moving it into A2A-only entrypoints instead of a widely imported types module.

[Copilot]

Using action="ignore" fully hides the upstream experimental warning. If the goal is to stop log flooding while still informing users, action="once" (or default + a de-dup strategy) would prevent repetition without completely silencing the warning.

Suggested change

	warnings.filterwarnings("ignore", message=r"\[EXPERIMENTAL\].*A2A")
	warnings.filterwarnings("once", message=r"\[EXPERIMENTAL\].*A2A")

[Copilot]

This installs the same global warnings filter as kagent.adk.types. Because warnings.filterwarnings() prepends by default, importing both modules will add duplicate entries to warnings.filters (minor overhead and harder to reason about). Consider centralizing this in a single helper that is called once, or ensure one module is the single place that configures warning suppression.

[Copilot]

There’s no regression test covering this new behavior (warning suppression). Since types.py is heavily imported in tests, adding a small unit test that asserts the targeted [EXPERIMENTAL] ... A2A UserWarning is not emitted repeatedly (or that the expected filter is installed) would help prevent reintroducing log spam or accidentally over-broad suppression.

Suggested change

	warnings.filterwarnings("ignore", message=r"\[EXPERIMENTAL\].*A2A")
	warnings.filterwarnings(
	"ignore",
	message=r"\[EXPERIMENTAL\].*A2A",
	category=UserWarning,
	module=r"^google\.adk\.agents\.remote_a2a_agent$",
	)