Feat/semconv by 123liuziming · Pull Request #224 · alibaba/loongsuite-python

123liuziming · 2026-06-21T07:39:02Z

Description

Please include a summary of the change and which issue is fixed. Please also include relevant motivation and context. List any dependencies that are required for this change.

Fixes # (issue)

Type of change

Please delete options that are not relevant.

Bug fix (non-breaking change which fixes an issue)
New feature (non-breaking change which adds functionality)
Breaking change (fix or feature that would cause existing functionality to not work as expected)
This change requires a documentation update

How Has This Been Tested?

Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce. Please also list any relevant details for your test configuration

Test A

Does This PR Require a Core Repo Change?

Yes. - Link to PR:
No.

Checklist:

See contributing.md for styleguide, changelog guidelines, and more.

Followed the style guidelines of this project
Changelogs have been updated
Unit tests have been added
Documentation has been updated

Add a LoongSuiteResourceDetector that contributes two resource attributes to the agent: * host.ip: the local host IP combined with the process id, formatted as <ip>-<pid> (e.g. 127.0.0.1-1). * gen_ai.instrumentation.sdk.name: set to "loongsuite-genai-utils". The detector is wired into LoongSuiteConfigurator so the attributes are always present on the SDK resource. Includes unit tests for the detector (ip-pid format, fixed sdk name, fallback on socket error) and the configurator wiring. Change-Id: I7fdfd07ffa8e2021e82c1d85d7d6b527c2cf8170 Co-developed-by: Claude <noreply@anthropic.com> Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Add missing GenAI semantic convention attributes across instrumentation plugins: - execute_tool spans: add tool_type="function" (Recommended) - invoke_agent spans: add agent_id (Conditionally Required) - inference spans: add server_address/server_port (Conditionally Required) - inference spans: add response_id extraction (Recommended) Plugins modified: - hermes-agent: agent_id, server_address/port from base_url, tool_type - vita: agent_id, tool_type, response_id, server_address inference - langchain: agent_id, tool_type - dashscope: server_address/port - widesearch: agent_id - agentscope: tool_type - qwen-agent: agent_id, tool_type - minisweagent: agent_id - google-adk: agent_id, tool_type - claude-agent-sdk: tool_type Change-Id: I21fc87489630d2ac6a0d9df22ef70dde60c57fff Co-developed-by: Qoder <noreply@qoder.com>

Copilot

Pull request overview

Adds/aligns GenAI semantic-convention fields across LoongSuite distro + multiple LoongSuite instrumentations, including new resource attributes and additional invocation metadata.

Changes:

Introduces LoongSuiteResourceDetector and wires it into LoongSuiteConfigurator to inject LoongSuite resource attributes.
Adds/propagates GenAI semconv fields (agent_id, tool_type, response_id, and some server_address/server_port) across several instrumentations.
Adds initial unit tests for the new LoongSuite distro resource/configurator behavior.

Reviewed changes

Copilot reviewed 17 out of 17 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
loongsuite-distro/src/loongsuite/distro/resource.py	Adds a resource detector that provides `host.ip` and `gen_ai.instrumentation.sdk.name`.
loongsuite-distro/src/loongsuite/distro/init.py	Updates configurator to inject LoongSuite resource attributes during SDK configuration.
loongsuite-distro/tests/test_resource.py	Adds unit tests for the resource detector and host IP detection helpers.
loongsuite-distro/tests/test_configurator.py	Adds unit tests asserting injected resource attributes are forwarded into SDK initialization.
loongsuite-distro/tests/init.py	Adds package marker for distro tests.
instrumentation-loongsuite/loongsuite-instrumentation-widesearch/src/opentelemetry/instrumentation/widesearch/utils.py	Populates `agent_id` for agent invocations.
instrumentation-loongsuite/loongsuite-instrumentation-vita/src/opentelemetry/instrumentation/vita/utils.py	Adds server address inference helper (currently unused).
instrumentation-loongsuite/loongsuite-instrumentation-vita/src/opentelemetry/instrumentation/vita/patch.py	Adds `agent_id`, `response_id`, and `tool_type` for Vita spans/invocations.
instrumentation-loongsuite/loongsuite-instrumentation-qwen-agent/src/opentelemetry/instrumentation/qwen_agent/utils.py	Populates `agent_id` and `tool_type` in invocations.
instrumentation-loongsuite/loongsuite-instrumentation-minisweagent/src/opentelemetry/instrumentation/minisweagent/internal/agent_wrappers.py	Populates `agent_id` for agent invocations.
instrumentation-loongsuite/loongsuite-instrumentation-langchain/src/opentelemetry/instrumentation/langchain/internal/_tracer.py	Adds `agent_id` and `tool_type` for LangChain invocations.
instrumentation-loongsuite/loongsuite-instrumentation-hermes-agent/src/opentelemetry/instrumentation/hermes_agent/helpers.py	Adds `agent_id`, infers server info from `base_url`, and sets `tool_type`.
instrumentation-loongsuite/loongsuite-instrumentation-google-adk/src/opentelemetry/instrumentation/google_adk/internal/_plugin.py	Adds `agent_id` and `tool_type` for Google ADK invocations.
instrumentation-loongsuite/loongsuite-instrumentation-dashscope/src/opentelemetry/instrumentation/dashscope/utils/generation.py	Sets `server_address/server_port` on DashScope LLM invocations.
instrumentation-loongsuite/loongsuite-instrumentation-claude-agent-sdk/src/opentelemetry/instrumentation/claude_agent_sdk/patch.py	Populates `tool_type` for tool spans.
instrumentation-loongsuite/loongsuite-instrumentation-agentscope/tests/test_skill_detection.py	Updates tests to include `tool_type`.
instrumentation-loongsuite/loongsuite-instrumentation-agentscope/src/opentelemetry/instrumentation/agentscope/patch.py	Populates `tool_type` for tool invocations.

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

+    def detect(self) -> Resource:
+        return Resource(
+            {
+                HOST_IP: _get_host_ip_with_pid(),
+                GEN_AI_INSTRUMENTATION_SDK_NAME: _GEN_AI_INSTRUMENTATION_SDK_NAME_VALUE,
+            }


+def _extract_server_info(instance: Any) -> tuple[str | None, int | None]:
+    """Extract server address and port from instance base_url."""
+    base_url = getattr(instance, "base_url", None)
+    if not base_url:
+        return None, None
+    base_url_str = str(base_url).rstrip("/")
+    try:
+        from urllib.parse import urlparse
+
+        parsed = urlparse(base_url_str)
+        address = parsed.hostname
+        port = parsed.port
+        if port is None:
+            port = 443 if parsed.scheme == "https" else 80
+        return address, port
+    except Exception:
+        return None, None
+


+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
+


ralf0131

Summary

PR #224 propagates GenAI semantic-convention fields (tool_type, agent_id, server_address/server_port, response_id) across multiple LoongSuite instrumentations and introduces a LoongSuiteResourceDetector that injects host.ip and gen_ai.instrumentation.sdk.name resource attributes. Overall direction is good — aligning with OTel GenAI semconv is the right move. The review below focuses on additional findings beyond what Copilot already flagged (host.ip format, unused _infer_server_address, schemeless-URL handling in _extract_server_info).

CLA is signed ✅.

Findings

[Warning] __init__.py:40 — resource detector silently overrides user-provided host.ip/gen_ai.instrumentation.sdk.name via dict.update(); consider setdefault so caller values win.
[Warning] hermes_agent/helpers.py:574 — agent_id="hermes-agent" is hardcoded for all instances; inconsistent with other instrumentations and defeats the uniqueness purpose of agent.id.
[Info] dashscope/utils/generation.py:471 — server_address/server_port hardcoded; inconsistent with the dynamic extraction added for hermes-agent in this same PR.
[Info] agentscope/patch.py:406 — tool_type hardcoded to "function" across all instrumentations; consider deriving from framework metadata where available.

Additional Notes

agent_id always mirrors agent_name: across every instrumentation in this PR, agent_id is set to agent_name (or a derivative), so both attributes always carry the same value. The GenAI semconv distinguishes agent.id (unique instance identifier) from agent.name (display name). If no truly unique per-instance ID is available right now, that's acceptable, but worth a follow-up to see whether session IDs or similar can provide real uniqueness.
PR description: The description body is still the unfilled GitHub template (no summary, no test instructions, checklist items unchecked). No changelog files appear to have been modified. Please fill in the description and update changelogs per the project's contributing guidelines.

Automated review by github-manager-bot

ralf0131 · 2026-06-21T08:39:43Z

    """

+    def _configure(self, **kwargs: Any) -> None:
+        resource_attributes = dict(kwargs.pop("resource_attributes", None) or {})


[Warning] The dict.update() call gives the resource detector precedence over any user-provided resource_attributes. If a user explicitly configures host.ip or gen_ai.instrumentation.sdk.name, those values are silently overwritten by the detector. Consider merging the detector's attributes first (or using setdefault-style logic) so caller-provided values take precedence.

resource_attributes = dict(kwargs.pop("resource_attributes", None) or {}) # User values set first, detector only fills gaps: detected = LoongSuiteResourceDetector().detect().attributes for k, v in detected.items(): resource_attributes.setdefault(k, v)

ralf0131 · 2026-06-21T08:39:43Z

    invocation = InvokeAgentInvocation(
        provider=_HERMES_AGENT_SYSTEM,
        agent_name="Hermes",
+        agent_id="hermes-agent",


[Warning] agent_id is hardcoded to "hermes-agent" for every instance. Other instrumentations in this PR use agent_name (which at least varies per agent). The GenAI semconv agent.id is intended to uniquely identify an agent instance — a constant string here means all concurrent hermes-agent instances share the same ID, making the attribute non-discriminating. Consider deriving from a session/instance identifier, or at minimum use agent_name for consistency.

ralf0131 · 2026-06-21T08:39:43Z

@@ -469,6 +469,8 @@ def _create_invocation_from_generation(

    invocation = LLMInvocation(request_model=request_model)
    invocation.provider = "dashscope"


[Info] server_address/server_port are hardcoded to "dashscope.aliyuncs.com"/443. This is inconsistent with the hermes-agent instrumentation in this same PR, which dynamically extracts server info from base_url. Users pointing DashScope-compatible SDKs at custom or regional endpoints won't see their actual endpoint reflected. Consider extracting from instance.base_url (or os.environ for DASHSCOPE_* base URL overrides) if available.

ralf0131 · 2026-06-21T08:39:43Z

    # Create invocation object with all tool data
    invocation = ExecuteToolInvocation(
        tool_name=tool_name,
+        tool_type="function",


[Info] tool_type is unconditionally set to "function" here and across all other instrumentations in this PR. Some agent frameworks support non-function tool types (e.g., code_interpreter, retrieval, builtin). If the framework exposes the actual tool type, consider deriving it rather than hardcoding. If "function" is the only supported type today, a comment documenting this assumption would help future maintainers.

123liuziming and others added 2 commits May 29, 2026 22:46

Copilot AI review requested due to automatic review settings June 21, 2026 07:39

github-actions Bot assigned 123liuziming, Cirilla-zmh and ralf0131 Jun 21, 2026

github-actions Bot requested review from Cirilla-zmh and ralf0131 June 21, 2026 07:39

Copilot started reviewing on behalf of 123liuziming June 21, 2026 07:39 View session

Copilot AI reviewed Jun 21, 2026

View reviewed changes

ralf0131 reviewed Jun 21, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Feat/semconv#224

Feat/semconv#224
123liuziming wants to merge 2 commits into
mainfrom
feat/semconv

123liuziming commented Jun 21, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

ralf0131 left a comment

Uh oh!

ralf0131 Jun 21, 2026

Uh oh!

ralf0131 Jun 21, 2026

Uh oh!

ralf0131 Jun 21, 2026

Uh oh!

ralf0131 Jun 21, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

		@@ -469,6 +469,8 @@ def _create_invocation_from_generation(

		invocation = LLMInvocation(request_model=request_model)
		invocation.provider = "dashscope"

Conversation

123liuziming commented Jun 21, 2026

Description

Type of change

How Has This Been Tested?

Does This PR Require a Core Repo Change?

Checklist:

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

ralf0131 left a comment

Choose a reason for hiding this comment

Summary

Findings

Additional Notes

Uh oh!

ralf0131 Jun 21, 2026

Choose a reason for hiding this comment

Uh oh!

ralf0131 Jun 21, 2026

Choose a reason for hiding this comment

Uh oh!

ralf0131 Jun 21, 2026

Choose a reason for hiding this comment

Uh oh!

ralf0131 Jun 21, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants