Filter A365 exports to only include genAI spans ADO#37660292 (#247)

* Filter A365 exports to only include genAI spans

* Remove InferenceOperationType.TEXT_COMPLETION, InferenceOperationType.GENERATE_CONTENT from the list

* Addressing comments

* minor change

* Fix test failure with format

* Fix test failure

* Formatting

---------

Co-authored-by: Jianbiao Lu <jianblu@microsoft.com>

Author: alexlu4250

libraries/microsoft-agents-a365-observability-core/microsoft_agents_a365/observability/core/exporters/agent365_exporter.py

@@ -27,7 +27,7 @@
     hex_trace_id,
     kind_name,
     parse_retry_after,
-    partition_by_identity,
+    filter_and_partition_by_identity,
     status_name,
     truncate_span,
 )
@@ -80,10 +80,10 @@ def export(self, spans: Sequence[ReadableSpan]) -> SpanExportResult:
             return SpanExportResult.FAILURE
 
         try:
-            groups = partition_by_identity(spans)
+            groups = filter_and_partition_by_identity(spans)
             if not groups:
-                # No spans with identity; treat as success
-                logger.info("No spans with tenant/agent identity found; nothing exported.")
+                # No eligible genAI spans to export after filtering/partitioning; treat as success
+                logger.info("No eligible genAI spans to export; nothing exported.")
                 return SpanExportResult.SUCCESS
 
             # Log number of groups and total span count

libraries/microsoft-agents-a365-observability-core/microsoft_agents_a365/observability/core/exporters/utils.py

@@ -14,16 +14,35 @@
 from opentelemetry.trace import SpanKind, StatusCode
 
 from ..constants import (
+    CHAT_OPERATION_NAME,
     ENABLE_A365_OBSERVABILITY_EXPORTER,
+    EXECUTE_TOOL_OPERATION_NAME,
     GEN_AI_AGENT_ID_KEY,
+    GEN_AI_OPERATION_NAME_KEY,
+    INVOKE_AGENT_OPERATION_NAME,
+    OUTPUT_MESSAGES_OPERATION_NAME,
     TENANT_ID_KEY,
 )
+from ..inference_operation_type import InferenceOperationType
 
 logger = logging.getLogger(__name__)
 
 # Maximum allowed span size in bytes (250KB)
 MAX_SPAN_SIZE_BYTES = 250 * 1024
 
+# Operation names that identify a span as eligible for export to the Agent 365
+# observability ingest service. Only spans whose gen_ai.operation.name matches
+# one of these values are included; all other spans are filtered out.
+GEN_AI_OPERATION_NAMES: frozenset[str] = frozenset(
+    {
+        INVOKE_AGENT_OPERATION_NAME,
+        EXECUTE_TOOL_OPERATION_NAME,
+        OUTPUT_MESSAGES_OPERATION_NAME,
+        CHAT_OPERATION_NAME,
+        InferenceOperationType.CHAT.value,
+    }
+)
+
 
 def hex_trace_id(value: int) -> str:
     # 128-bit -> 32 hex chars
@@ -127,22 +146,44 @@ def truncate_span(span_dict: dict[str, Any]) -> dict[str, Any]:
         return span_dict
 
 
-def partition_by_identity(
+def filter_and_partition_by_identity(
     spans: Sequence[ReadableSpan],
 ) -> dict[tuple[str, str], list[ReadableSpan]]:
     """
-    Extract (tenantId, agentId). Prefer attributes; if you also stamp baggage
-    into attributes via a processor, they'll be here already.
+    Filter export-eligible spans and partition them by (tenantId, agentId).
+
+    Only spans whose ``gen_ai.operation.name`` is in
+    ``GEN_AI_OPERATION_NAMES`` are included; non-genAI spans (e.g. HTTP, DB)
+    and spans with other operation names are filtered out. Spans without
+    both tenant and agent identity are also skipped.
     """
     groups: dict[tuple[str, str], list[ReadableSpan]] = {}
+    non_gen_ai_count = 0
+    missing_identity_count = 0
     for sp in spans:
         attrs = sp.attributes or {}
+        operation_name = as_str(attrs.get(GEN_AI_OPERATION_NAME_KEY))
+        if not operation_name or operation_name not in GEN_AI_OPERATION_NAMES:
+            non_gen_ai_count += 1
+            continue
         tenant = as_str(attrs.get(TENANT_ID_KEY))
         agent = as_str(attrs.get(GEN_AI_AGENT_ID_KEY))
         if not tenant or not agent:
+            missing_identity_count += 1
             continue
         key = (tenant, agent)
         groups.setdefault(key, []).append(sp)
+
+    if non_gen_ai_count > 0:
+        logger.debug(
+            "[Agent365Exporter] %d spans without an eligible gen_ai.operation.name filtered out",
+            non_gen_ai_count,
+        )
+    if missing_identity_count > 0:
+        logger.debug(
+            "[Agent365Exporter] %d spans skipped due to missing tenant or agent ID",
+            missing_identity_count,
+        )
     return groups
 
 

tests/observability/core/exporters/test_payload_chunking.py

@@ -10,7 +10,9 @@
 from unittest.mock import Mock, patch
 
 from microsoft_agents_a365.observability.core.constants import (
+    CHAT_OPERATION_NAME,
     GEN_AI_AGENT_ID_KEY,
+    GEN_AI_OPERATION_NAME_KEY,
     TENANT_ID_KEY,
 )
 from microsoft_agents_a365.observability.core.exporters.agent365_exporter import (
@@ -165,6 +167,7 @@ def _make_span(self, span_id: int, attribute_size: int) -> ReadableSpan:
         mock_span.attributes = {
             TENANT_ID_KEY: "tenant-1",
             GEN_AI_AGENT_ID_KEY: "agent-1",
+            GEN_AI_OPERATION_NAME_KEY: CHAT_OPERATION_NAME,
             "payload": "x" * attribute_size,
         }
         mock_span.events = []

tests/observability/core/test_agent365_exporter.py

@@ -6,7 +6,12 @@
 import unittest
 from unittest.mock import Mock, patch
 
-from microsoft_agents_a365.observability.core.constants import GEN_AI_AGENT_ID_KEY, TENANT_ID_KEY
+from microsoft_agents_a365.observability.core.constants import (
+    GEN_AI_AGENT_ID_KEY,
+    GEN_AI_OPERATION_NAME_KEY,
+    INVOKE_AGENT_OPERATION_NAME,
+    TENANT_ID_KEY,
+)
 from microsoft_agents_a365.observability.core.exporters.agent365_exporter import (
     DEFAULT_ENDPOINT_URL,
     _Agent365Exporter,
@@ -54,6 +59,7 @@ def _create_mock_span(
         scope_version: str = "1.0.0",
         tenant_id: str = "test-tenant-123",
         agent_id: str = "test-agent-456",
+        operation_name: str | None = INVOKE_AGENT_OPERATION_NAME,
     ) -> ReadableSpan:
         """Create a mock ReadableSpan for testing."""
         mock_span = Mock(spec=ReadableSpan)
@@ -78,13 +84,15 @@ def _create_mock_span(
         mock_span.kind = Mock()
         mock_span.kind.name = "INTERNAL"
 
-        # Add identity attributes for partition_by_identity to work
+        # Add identity attributes for filter_and_partition_by_identity to work
         span_attributes = attributes or {}
         if tenant_id and agent_id:
             span_attributes.update({
                 TENANT_ID_KEY: tenant_id,
                 GEN_AI_AGENT_ID_KEY: agent_id,
             })
+        if operation_name is not None and GEN_AI_OPERATION_NAME_KEY not in span_attributes:
+            span_attributes[GEN_AI_OPERATION_NAME_KEY] = operation_name
 
         mock_span.attributes = span_attributes
         mock_span.events = []
@@ -350,10 +358,8 @@ def test_export_error_logging(self, mock_logger):
         # Verify export succeeded (no identity spans are treated as success)
         self.assertEqual(result, SpanExportResult.SUCCESS)
 
-        # Verify info log for no identity
-        mock_logger.info.assert_called_with(
-            "No spans with tenant/agent identity found; nothing exported."
-        )
+        # Verify info log for no eligible spans
+        mock_logger.info.assert_called_with("No eligible genAI spans to export; nothing exported.")
 
     def test_exporter_is_internal(self):
         """Test that _Agent365Exporter is marked as internal/private.
@@ -657,6 +663,97 @@ def test_export_no_fallback_when_default_succeeds(self):
             self.assertEqual(result, SpanExportResult.SUCCESS)
             mock_post.assert_called_once()
 
+    def test_export_filters_out_non_genai_spans(self):
+        """Spans without a known gen_ai.operation.name are filtered out."""
+        # Arrange: one genAI span and two non-genAI spans (no/unknown operation name)
+        genai_span = self._create_mock_span("genai_span", trace_id=1, span_id=2)
+        no_op_span = self._create_mock_span("http_span", trace_id=3, span_id=4, operation_name=None)
+        unknown_op_span = self._create_mock_span(
+            "db_span", trace_id=5, span_id=6, operation_name="some_random_op"
+        )
+
+        with patch.object(self.exporter, "_post_with_retries", return_value=True) as mock_post:
+            # Act
+            result = self.exporter.export([genai_span, no_op_span, unknown_op_span])
+
+            # Assert: only the genAI span is exported
+            self.assertEqual(result, SpanExportResult.SUCCESS)
+            mock_post.assert_called_once()
+            _, body, _ = mock_post.call_args[0]
+            request_data = json.loads(body)
+            spans_out = request_data["resourceSpans"][0]["scopeSpans"][0]["spans"]
+            self.assertEqual(len(spans_out), 1)
+            self.assertEqual(spans_out[0]["name"], "genai_span")
+
+    def test_export_filters_out_only_non_genai_spans_returns_success(self):
+        """When all spans are filtered out, export returns SUCCESS without HTTP call."""
+        # Arrange
+        spans = [
+            self._create_mock_span("http_span", operation_name=None),
+            self._create_mock_span("db_span", operation_name="other"),
+        ]
+
+        with patch.object(self.exporter, "_post_with_retries", return_value=True) as mock_post:
+            # Act
+            result = self.exporter.export(spans)
+
+            # Assert
+            self.assertEqual(result, SpanExportResult.SUCCESS)
+            mock_post.assert_not_called()
+
+    def test_export_includes_inference_operation_type_chat_spans(self):
+        """Spans with InferenceOperationType.CHAT value ('Chat') are kept without normalization."""
+        # Arrange — server accepts 'Chat' via case-insensitive matching
+        chat_span = self._create_mock_span(
+            "chat_span", trace_id=1, span_id=2, operation_name="Chat"
+        )
+
+        with patch.object(self.exporter, "_post_with_retries", return_value=True) as mock_post:
+            result = self.exporter.export([chat_span])
+
+            self.assertEqual(result, SpanExportResult.SUCCESS)
+            mock_post.assert_called_once()
+            _, body, _ = mock_post.call_args[0]
+            request_data = json.loads(body)
+            spans_out = request_data["resourceSpans"][0]["scopeSpans"][0]["spans"]
+            self.assertEqual(len(spans_out), 1)
+            # Value is preserved as-is; no normalization
+            self.assertEqual(spans_out[0]["attributes"][GEN_AI_OPERATION_NAME_KEY], "Chat")
+
+    def test_export_filters_out_unsupported_inference_operation_types(self):
+        """Spans with TextCompletion / GenerateContent are filtered out."""
+        text_completion_span = self._create_mock_span(
+            "text_completion_span", trace_id=3, span_id=4, operation_name="TextCompletion"
+        )
+        generate_content_span = self._create_mock_span(
+            "generate_content_span", trace_id=5, span_id=6, operation_name="GenerateContent"
+        )
+
+        with patch.object(self.exporter, "_post_with_retries", return_value=True) as mock_post:
+            result = self.exporter.export([text_completion_span, generate_content_span])
+
+            # Both are filtered out — nothing to export
+            self.assertEqual(result, SpanExportResult.SUCCESS)
+            mock_post.assert_not_called()
+
+    def test_export_does_not_normalize_canonical_operation_names(self):
+        """invoke_agent / execute_tool / output_messages / chat are not rewritten."""
+        cases = ["invoke_agent", "execute_tool", "output_messages", "chat"]
+        for op in cases:
+            with self.subTest(operation_name=op):
+                span = self._create_mock_span(
+                    f"{op}_span", trace_id=1, span_id=2, operation_name=op
+                )
+                with patch.object(
+                    self.exporter, "_post_with_retries", return_value=True
+                ) as mock_post:
+                    result = self.exporter.export([span])
+                    self.assertEqual(result, SpanExportResult.SUCCESS)
+                    _, body, _ = mock_post.call_args[0]
+                    request_data = json.loads(body)
+                    span_out = request_data["resourceSpans"][0]["scopeSpans"][0]["spans"][0]
+                    self.assertEqual(span_out["attributes"][GEN_AI_OPERATION_NAME_KEY], op)
+
 
 if __name__ == "__main__":
     unittest.main()