fix: address all PR review comments

- resource.py: Use raw IP for host.ip, add service.instance.id for <ip>-<pid>
- distro/__init__.py: Use setdefault so user-provided values take precedence
- hermes_agent/helpers.py: Handle schemeless URLs in _extract_server_info,
  return (None, None) when hostname is falsy; use agent_name for agent_id
- vita/utils.py: Remove unused _infer_server_address function
- claw-eval/wrappers.py: Change default fallback from 'openai' to 'unknown',
  add Google/Gemini provider detection, remove dead if-guard
- agentscope/patch.py: Add comment explaining tool_type='function' assumption
- dashscope/generation.py: Add comment about hardcoded server_address
- tox-loongsuite.ini: Add loongsuite-distro test environment
- test_resource.py/test_configurator.py: Update tests for new behavior

Change-Id: I9128e7bb1db9bd10b1529a40c5f450e61dcc2aa8
Co-developed-by: Qoder <noreply@qoder.com>

Author: 123liuziming

instrumentation-loongsuite/loongsuite-instrumentation-agentscope/src/opentelemetry/instrumentation/agentscope/patch.py

@@ -401,6 +401,8 @@ async def wrap_tool_call(wrapped, instance, args, kwargs, handler):
     matched_skill = _match_skill_for_tool(instance, tool_args)
 
     # Create invocation object with all tool data
+    # NOTE: tool_type is set to "function" as agentscope currently only
+    # supports function-type tools. Update when other types are supported.
     invocation = ExecuteToolInvocation(
         tool_name=tool_name,
         tool_type="function",

instrumentation-loongsuite/loongsuite-instrumentation-claw-eval/src/opentelemetry/instrumentation/claw_eval/internal/wrappers.py

@@ -79,6 +79,8 @@ def _infer_provider_name(provider: Any) -> str:
         return "anthropic"
     if "dashscope" in cls_name:
         return "dashscope"
+    if "gemini" in cls_name or "google" in cls_name:
+        return "google"
     # Infer from model_id
     model_id = str(getattr(provider, "model_id", "") or "")
     if model_id.startswith(("gpt-", "o1-", "o3-", "chatgpt-")):
@@ -87,7 +89,9 @@ def _infer_provider_name(provider: Any) -> str:
         return "anthropic"
     if model_id.startswith(("qwen",)):
         return "dashscope"
-    return "openai"
+    if model_id.startswith(("gemini",)):
+        return "google"
+    return "unknown"
 
 
 # ---------------------------------------------------------------------------
@@ -544,8 +548,7 @@ def __call__(self, wrapped, instance, args, kwargs):
 
             # Set gen_ai.provider.name (Required attribute)
             _provider_name = _infer_provider_name(provider)
-            if _provider_name:
-                span.set_attribute(GenAI.GEN_AI_PROVIDER_NAME, _provider_name)
+            span.set_attribute(GenAI.GEN_AI_PROVIDER_NAME, _provider_name)
 
             model_id = ""
             if provider is not None:

instrumentation-loongsuite/loongsuite-instrumentation-dashscope/src/opentelemetry/instrumentation/dashscope/utils/generation.py

@@ -469,6 +469,8 @@ def _create_invocation_from_generation(
 
     invocation = LLMInvocation(request_model=request_model)
     invocation.provider = "dashscope"
+    # DashScope SDK always targets dashscope.aliyuncs.com; if custom endpoint
+    # support is added in the future, extract from instance/env instead.
     invocation.server_address = "dashscope.aliyuncs.com"
     invocation.server_port = 443
     invocation.input_messages = _extract_input_messages(kwargs)

instrumentation-loongsuite/loongsuite-instrumentation-hermes-agent/src/opentelemetry/instrumentation/hermes_agent/helpers.py

@@ -571,7 +571,7 @@ def create_agent_invocation(
     invocation = InvokeAgentInvocation(
         provider=_HERMES_AGENT_SYSTEM,
         agent_name="Hermes",
-        agent_id="hermes-agent",
+        agent_id="Hermes",
         conversation_id=getattr(instance, "session_id", None),
         request_model=getattr(instance, "model", None),
         input_messages=[
@@ -616,11 +616,16 @@ def _extract_server_info(instance: Any) -> tuple[str | None, int | None]:
     if not base_url:
         return None, None
     base_url_str = str(base_url).rstrip("/")
+    # Handle schemeless URLs by prepending "//" for urlparse.
+    if "://" not in base_url_str:
+        base_url_str = "//" + base_url_str
     try:
         from urllib.parse import urlparse  # noqa: PLC0415
 
         parsed = urlparse(base_url_str)
         address = parsed.hostname
+        if not address:
+            return None, None
         port = parsed.port
         if port is None:
             port = 443 if parsed.scheme == "https" else 80

instrumentation-loongsuite/loongsuite-instrumentation-vita/src/opentelemetry/instrumentation/vita/utils.py

@@ -149,20 +149,6 @@ def _infer_provider(model_name: str) -> str:
     return "unknown"
 
 
-def _infer_server_address(model: str) -> tuple[str | None, int | None]:
-    """Infer server address from model name for vita."""
-    model_lower = model.lower() if model else ""
-    if "qwen" in model_lower or "dashscope" in model_lower:
-        return "dashscope.aliyuncs.com", 443
-    if "gpt" in model_lower or "openai" in model_lower:
-        return "api.openai.com", 443
-    if "claude" in model_lower or "anthropic" in model_lower:
-        return "api.anthropic.com", 443
-    if "deepseek" in model_lower:
-        return "api.deepseek.com", 443
-    return None, None
-
-
 def _get_tool_definitions(
     tools: Any,
 ) -> Optional[List[FunctionToolDefinition]]:

loongsuite-distro/src/loongsuite/distro/__init__.py

@@ -40,9 +40,10 @@ def _configure(self, **kwargs: Any) -> None:
         resource_attributes = dict(
             kwargs.pop("resource_attributes", None) or {}
         )
-        resource_attributes.update(
-            LoongSuiteResourceDetector().detect().attributes
-        )
+        # Use setdefault so user-provided values take precedence over detector.
+        detected = LoongSuiteResourceDetector().detect().attributes
+        for k, v in detected.items():
+            resource_attributes.setdefault(k, v)
         super()._configure(resource_attributes=resource_attributes, **kwargs)
 
 

loongsuite-distro/src/loongsuite/distro/resource.py

@@ -22,6 +22,7 @@
 
 # Resource attribute keys contributed by LoongSuite.
 HOST_IP = "host.ip"
+SERVICE_INSTANCE_ID = "service.instance.id"
 GEN_AI_INSTRUMENTATION_SDK_NAME = "gen_ai.instrumentation.sdk.name"
 
 # Fixed value identifying the GenAI instrumentation SDK shipped with LoongSuite.
@@ -67,14 +68,18 @@ class LoongSuiteResourceDetector(ResourceDetector):
 
     Contributes the following attributes to the resource:
 
-    * ``host.ip`` formatted as ``<ip>-<pid>`` (e.g. ``127.0.0.1-1``).
+    * ``host.ip`` — the raw host IP address (e.g. ``127.0.0.1``).
+    * ``service.instance.id`` — ``<ip>-<pid>`` uniquely identifying the process
+      instance (e.g. ``127.0.0.1-1``).
     * ``gen_ai.instrumentation.sdk.name`` set to ``loongsuite-genai-utils``.
     """
 
     def detect(self) -> Resource:
+        host_ip = _get_host_ip()
         return Resource(
             {
-                HOST_IP: _get_host_ip_with_pid(),
+                HOST_IP: host_ip,
+                SERVICE_INSTANCE_ID: f"{host_ip}-{os.getpid()}",
                 GEN_AI_INSTRUMENTATION_SDK_NAME: _GEN_AI_INSTRUMENTATION_SDK_NAME_VALUE,
             }
         )

loongsuite-distro/tests/test_configurator.py

@@ -19,6 +19,7 @@
 from loongsuite.distro.resource import (
     GEN_AI_INSTRUMENTATION_SDK_NAME,
     HOST_IP,
+    SERVICE_INSTANCE_ID,
 )
 
 
@@ -30,20 +31,27 @@ def test_configure_injects_loongsuite_attributes(self, mock_init):
         mock_init.assert_called_once()
         resource_attributes = mock_init.call_args.kwargs["resource_attributes"]
         self.assertIn(HOST_IP, resource_attributes)
+        self.assertIn(SERVICE_INSTANCE_ID, resource_attributes)
         self.assertEqual(
             resource_attributes[GEN_AI_INSTRUMENTATION_SDK_NAME],
             "loongsuite-genai-utils",
         )
 
     @mock.patch("opentelemetry.sdk._configuration._initialize_components")
     def test_configure_preserves_existing_attributes(self, mock_init):
+        """User-provided values take precedence over detector values."""
         LoongSuiteConfigurator().configure(
-            resource_attributes={"service.name": "my-service"}
+            resource_attributes={
+                "service.name": "my-service",
+                HOST_IP: "custom-ip",
+            }
         )
 
         resource_attributes = mock_init.call_args.kwargs["resource_attributes"]
         self.assertEqual(resource_attributes["service.name"], "my-service")
-        self.assertIn(HOST_IP, resource_attributes)
+        # User-provided host.ip should NOT be overwritten by detector
+        self.assertEqual(resource_attributes[HOST_IP], "custom-ip")
+        self.assertIn(SERVICE_INSTANCE_ID, resource_attributes)
         self.assertIn(GEN_AI_INSTRUMENTATION_SDK_NAME, resource_attributes)
 
     @mock.patch("opentelemetry.sdk._configuration._initialize_components")

loongsuite-distro/tests/test_resource.py

@@ -18,6 +18,7 @@
 from loongsuite.distro.resource import (
     GEN_AI_INSTRUMENTATION_SDK_NAME,
     HOST_IP,
+    SERVICE_INSTANCE_ID,
     LoongSuiteResourceDetector,
     _get_host_ip,
     _get_host_ip_with_pid,
@@ -37,8 +38,15 @@ def test_detect_returns_resource(self):
     def test_detect_contains_expected_keys(self):
         attributes = LoongSuiteResourceDetector().detect().attributes
         self.assertIn(HOST_IP, attributes)
+        self.assertIn(SERVICE_INSTANCE_ID, attributes)
         self.assertIn(GEN_AI_INSTRUMENTATION_SDK_NAME, attributes)
 
+    def test_host_ip_is_raw_ip(self):
+        """host.ip should contain a raw IP address without PID suffix."""
+        attributes = LoongSuiteResourceDetector().detect().attributes
+        # Raw IP should not contain a dash-PID suffix
+        self.assertNotIn("-", attributes[HOST_IP].rsplit(".", 1)[-1])
+
     def test_gen_ai_instrumentation_sdk_name_value(self):
         attributes = LoongSuiteResourceDetector().detect().attributes
         self.assertEqual(
@@ -50,9 +58,12 @@ def test_gen_ai_instrumentation_sdk_name_value(self):
     @mock.patch(
         "loongsuite.distro.resource._get_host_ip", return_value="127.0.0.1"
     )
-    def test_host_ip_format_is_ip_dash_pid(self, _mock_ip, _mock_pid):
+    def test_service_instance_id_format_is_ip_dash_pid(
+        self, _mock_ip, _mock_pid
+    ):
         attributes = LoongSuiteResourceDetector().detect().attributes
-        self.assertEqual(attributes[HOST_IP], "127.0.0.1-1")
+        self.assertEqual(attributes[SERVICE_INSTANCE_ID], "127.0.0.1-1")
+        self.assertEqual(attributes[HOST_IP], "127.0.0.1")
 
     @mock.patch("loongsuite.distro.resource.os.getpid", return_value=42)
     @mock.patch(

tox-loongsuite.ini

@@ -126,6 +126,9 @@ envlist =
     py3{10,11,12,13}-test-loongsuite-instrumentation-wildtool
     lint-loongsuite-instrumentation-wildtool
 
+    ; loongsuite-distro
+    py3{9,10,11,12,13}-test-loongsuite-distro
+
 [testenv]
 test_deps =
   opentelemetry-api@{env:CORE_REPO}\#egg=opentelemetry-api&subdirectory=opentelemetry-api
@@ -234,6 +237,9 @@ deps =
   wildtool: {[testenv]test_deps}
   wildtool: -r {toxinidir}/instrumentation-loongsuite/loongsuite-instrumentation-wildtool/test-requirements.txt
 
+  loongsuite-distro: {[testenv]test_deps}
+  loongsuite-distro: {toxinidir}/loongsuite-distro
+
   ; FIXME: add coverage testing
 allowlist_externals =
   sh
@@ -334,6 +340,8 @@ commands =
   test-loongsuite-instrumentation-wildtool: pytest {toxinidir}/instrumentation-loongsuite/loongsuite-instrumentation-wildtool/tests {posargs}
   lint-loongsuite-instrumentation-wildtool: python -m ruff check {toxinidir}/instrumentation-loongsuite/loongsuite-instrumentation-wildtool
 
+  test-loongsuite-distro: pytest {toxinidir}/loongsuite-distro/tests {posargs}
+
   ; TODO: add coverage commands
   ; coverage: {toxinidir}/scripts/coverage.sh