fix(claude_agent_sdk): log ResultMessage metrices

[Mahhheshh (Mahhheshh)]

Capture previously dropped ResultMessage fields in _handle_result.

ref 5 (#149)

[starfolkai]

Drop into this review session: sfk devbox connect pr-225-braintrust-sdk-python-17a8 --attach

Found 2 issues:

1. Test guards use hasattr but production code filters with if v is not None -- these have different semantics. When a ResultMessage attribute exists but has value None (e.g., stop_sequence is None when no stop sequence was triggered), hasattr returns True but the production filter excludes it from metadata. The assertion assert task_span.get("metadata", {}).get("stop_sequence") is not None would then fail. All five new guards (lines 175-184) should use getattr(result_message, ..., None) is not None to match the if v is not None filter in _handle_result (line 739).

braintrust-sdk-python/py/src/braintrust/integrations/claude_agent_sdk/test_claude_agent_sdk.py

Lines 174 to 185 in 5c9b3dd

	assert task_span.get("metadata", {}).get("session_id") is not None
	if hasattr(result_message, "stop_reason"):
	assert task_span.get("metadata", {}).get("stop_reason") is not None
	if hasattr(result_message, "stop_sequence"):
	assert task_span.get("metadata", {}).get("stop_sequence") is not None
	if hasattr(result_message, "total_cost_usd"):
	assert task_span.get("metadata", {}).get("total_cost_usd") is not None
	if hasattr(result_message, "duration_ms"):
	assert task_span.get("metadata", {}).get("duration_ms") is not None
	if hasattr(result_message, "duration_api_ms"):
	assert task_span.get("metadata", {}).get("duration_api_ms") is not None
	llm_spans = [s for s in spans if s["span_attributes"]["type"] == SpanTypeAttribute.LLM]

Production filter for reference:

braintrust-sdk-python/py/src/braintrust/integrations/claude_agent_sdk/tracing.py

Lines 728 to 740 in 5c9b3dd

	result_metadata = {
	k: v
	for k, v in {
	"num_turns": getattr(message, "num_turns", None),
	"session_id": getattr(message, "session_id", None),
	"stop_reason": getattr(message, "stop_reason", None),
	"stop_sequence": getattr(message, "stop_sequence", None),
	"total_cost_usd": getattr(message, "total_cost_usd", None),
	"duration_ms": getattr(message, "duration_ms", None),
	"duration_api_ms": getattr(message, "duration_api_ms", None),
	}.items()
	if v is not None
	}

2. The stop_sequence assertion is dead code -- the cassette for test_calculator_with_multiple_operations does not include stop_sequence in the result payload, so the guard is always False and the assertion at line 177-178 never executes. This field has zero test coverage. Consider either updating the cassette to include a result with a non-null stop_sequence, or adding a targeted unit test with a synthetic ResultMessage (the existing ResultMessage stub at line 1360 also needs updating -- it only defines num_turns and session_id).

braintrust-sdk-python/py/src/braintrust/integrations/claude_agent_sdk/test_claude_agent_sdk.py

Lines 176 to 179 in 5c9b3dd

	assert task_span.get("metadata", {}).get("stop_reason") is not None
	if hasattr(result_message, "stop_sequence"):
	assert task_span.get("metadata", {}).get("stop_sequence") is not None
	if hasattr(result_message, "total_cost_usd"):

Fake ResultMessage stub that needs the new fields:

braintrust-sdk-python/py/src/braintrust/integrations/claude_agent_sdk/test_claude_agent_sdk.py

Lines 1360 to 1377 in 5c9b3dd

	class ResultMessage:
	def __init__(
	self,
	*,
	input_tokens: int = 1,
	output_tokens: int = 1,
	cache_creation_input_tokens: int = 0,
	num_turns: int = 1,
	session_id: str = "session-123",
	):
	self.usage = types.SimpleNamespace(
	input_tokens=input_tokens,
	output_tokens=output_tokens,
	cache_creation_input_tokens=cache_creation_input_tokens,
	)
	self.num_turns = num_turns
	self.session_id = session_id

[Abhijeet Prasad (AbhiPrasad)]

please rebase! CI should be fixed