Add QwenPaw 2.x runtime support

README.md

@@ -23,7 +23,7 @@ All instrumentations use [opentelemetry-util-genai](./util/opentelemetry-util-ge
 | [opentelemetry-instrumentation-genai-openai-agents](./instrumentation/opentelemetry-instrumentation-genai-openai-agents) | openai-agents >= 0.3.3, < 1 | [1.3b0](https://pypi.org/project/opentelemetry-instrumentation-genai-openai-agents/) |
 | [opentelemetry-instrumentation-genai-portkey](./instrumentation/opentelemetry-instrumentation-genai-portkey) | portkey-ai >= 1.0.0, < 3 | [1.3b0](https://pypi.org/project/opentelemetry-instrumentation-genai-portkey/) |
 | [opentelemetry-instrumentation-genai-qwen-agent](./instrumentation/opentelemetry-instrumentation-genai-qwen-agent) | qwen-agent >= 0.0.20, < 1 | [1.3b0](https://pypi.org/project/opentelemetry-instrumentation-genai-qwen-agent/) |
-| [opentelemetry-instrumentation-genai-qwenpaw](./instrumentation/opentelemetry-instrumentation-genai-qwenpaw) | qwenpaw >= 1.1.0, < 2.0.0; python_version < '3.14' | [1.3b0](https://pypi.org/project/opentelemetry-instrumentation-genai-qwenpaw/) |
+| [opentelemetry-instrumentation-genai-qwenpaw](./instrumentation/opentelemetry-instrumentation-genai-qwenpaw) | qwenpaw >= 1.1.0, < 3.0.0; python_version < '3.14' | [1.3b0](https://pypi.org/project/opentelemetry-instrumentation-genai-qwenpaw/) |
 | [opentelemetry-instrumentation-genai-smolagents](./instrumentation/opentelemetry-instrumentation-genai-smolagents) | smolagents >= 1.24.0, < 2 | [1.2b0](https://pypi.org/project/opentelemetry-instrumentation-genai-smolagents/) |
 | [opentelemetry-instrumentation-google-genai](./instrumentation/opentelemetry-instrumentation-google-genai) | google-genai >= 1.32.0, <3 | [1.3b0](https://pypi.org/project/opentelemetry-instrumentation-google-genai/) |
 

instrumentation/opentelemetry-instrumentation-genai-qwenpaw/.changelog/994.added

@@ -0,0 +1 @@
+Support QwenPaw 2.x runtime turns while retaining compatibility with 1.x.

instrumentation/opentelemetry-instrumentation-genai-qwenpaw/README.rst

@@ -9,7 +9,7 @@ OpenTelemetry QwenPaw Instrumentation
 This library traces user turns handled by `QwenPaw
 <https://github.com/agentscope-ai/QwenPaw>`_, a personal assistant
 application built on AgentScope. Each turn that goes through
-``AgentRunner.query_handler`` produces one ``invoke_agent`` span following
+``AgentRunner.query_handler`` (1.x) or ``Runtime.run`` (2.x) produces one ``invoke_agent`` span following
 the OpenTelemetry GenAI semantic conventions, carrying the agent display
 name (``gen_ai.agent.name``, when already cached by the runner), the session id
 (``gen_ai.conversation.id``), and — when content capture is enabled — the
@@ -40,11 +40,15 @@ under the ``invoke_agent`` span.
 Supported versions
 ------------------
 
-QwenPaw 1.1.x only. The 2.x `runtime architecture upgrade
-<https://github.com/agentscope-ai/QwenPaw/blob/main/website/public/blog/runtime-architecture-upgrade.en.md>`_
-refactored away ``AgentRunner.query_handler``, the seam this package
-patches, so the dependency is capped at ``< 2.0.0`` and 2.x support will
-be added separately against the new runtime entry point.
+QwenPaw >= 1.1.0, < 3.0.0. Version 1.x uses the agent runner; version
+2.x uses the runtime request entry point and captures text, reasoning,
+tool calls/results, and image/audio/video/file content from request messages and
+streamed SSE envelopes. Request normalization and span creation remain lazy;
+SDK-generated session IDs propagate to nested GenAI invocations.
+Version 2.x support requires ``opentelemetry-util-genai`` 1.4b0 or newer.
+QwenPaw 2.x requires Python >= 3.11,
+< 3.14. Both entry points produce one internal ``invoke_agent`` span per
+turn; model and tool execution remain owned by the underlying libraries.
 
 Installation
 ------------
@@ -113,6 +117,33 @@ Conformance
 Semantic-convention conformance scenarios live in `tests/conformance
 <tests/conformance>`_.
 
+The same ``QwenPawInstrumentor`` selects the entry point for the installed
+QwenPaw major version. The 1.x runner instrumentation remains supported.
+Tests cover both the oldest and latest releases of each runtime generation:
+
+.. list-table::
+   :header-rows: 1
+
+   * - QwenPaw
+     - Entry point
+     - Conformance tox environment
+   * - 1.1.0
+     - ``AgentRunner.query_handler``
+     - ``py311-test-instrumentation-genai-qwenpaw-oldest-conformance``
+   * - Latest 1.x
+     - ``AgentRunner.query_handler``
+     - ``py313-test-instrumentation-genai-qwenpaw-v1-conformance``
+   * - 2.0.0
+     - ``Runtime.run``
+     - ``py311-test-instrumentation-genai-qwenpaw-v2-conformance``
+   * - Latest 2.x
+     - ``Runtime.run``
+     - ``py313-test-instrumentation-genai-qwenpaw-conformance``
+
+Each version runs successful and failed turns through Weaver live-check.
+The 2.x scenarios additionally cover reasoning, tool calls/results, and
+multimodal request messages.
+
 References
 ----------
 

instrumentation/opentelemetry-instrumentation-genai-qwenpaw/examples/manual/requirements.txt

@@ -1,4 +1,4 @@
-qwenpaw>=1.1.12.post3,<2.0.0
+qwenpaw>=2.0,<3.0
 
 opentelemetry-sdk~=1.45.2
 opentelemetry-exporter-otlp-proto-grpc~=1.45.2

instrumentation/opentelemetry-instrumentation-genai-qwenpaw/examples/zero-code/requirements.txt

@@ -1,4 +1,4 @@
-qwenpaw>=1.1.12.post3,<2.0.0
+qwenpaw>=2.0,<3.0
 
 opentelemetry-sdk~=1.45.2
 opentelemetry-distro~=0.64b0

instrumentation/opentelemetry-instrumentation-genai-qwenpaw/pyproject.toml

@@ -32,7 +32,7 @@ dependencies = [
 [project.optional-dependencies]
 # qwenpaw itself requires Python <3.14, so the marker keeps the extra
 # installable (as a no-op) on newer interpreters used for tooling.
-instruments = ["qwenpaw >= 1.1.0, < 2.0.0; python_version < '3.14'"]
+instruments = ["qwenpaw >= 1.1.0, < 3.0.0; python_version < '3.14'"]
 
 [project.entry-points.opentelemetry_instrumentor]
 qwenpaw = "opentelemetry.instrumentation.genai.qwenpaw:QwenPawInstrumentor"

instrumentation/opentelemetry-instrumentation-genai-qwenpaw/src/opentelemetry/instrumentation/genai/qwenpaw/__init__.py

@@ -8,7 +8,8 @@
 Instrumentation for `QwenPaw <https://github.com/agentscope-ai/QwenPaw>`_,
 a personal assistant application built on AgentScope.
 
-Each user turn handled by ``AgentRunner.query_handler`` is traced as one
+Each user turn handled by ``AgentRunner.query_handler`` (1.x) or
+``Runtime.run`` (2.x) is traced as one
 ``invoke_agent`` span following the OpenTelemetry GenAI semantic
 conventions.
 
@@ -51,6 +52,7 @@
 import logging
 from collections.abc import Collection
 from importlib import import_module
+from importlib.util import find_spec
 from typing import Any
 
 from wrapt import wrap_function_wrapper
@@ -62,19 +64,30 @@
 
 from .package import _instruments
 from .patch import make_query_handler_wrapper
+from .runtime import make_runtime_wrapper, observe_context_build
 from .version import __version__
 
 logger = logging.getLogger(__name__)
 
 __all__ = ["QwenPawInstrumentor"]
 
 _RUNNER_MODULE = "qwenpaw.app.runner.runner"
+_RUNTIME_MODULE = "qwenpaw.runtime.runtime"
+
+
+def _runtime_v2() -> bool:
+    try:
+        return find_spec(_RUNTIME_MODULE) is not None
+    except ModuleNotFoundError as error:
+        if error.name not in ("qwenpaw", "qwenpaw.runtime"):
+            raise
+        return False
 
 
 class QwenPawInstrumentor(BaseInstrumentor):
     """An instrumentor for the QwenPaw application runner.
 
-    Traces ``AgentRunner.query_handler`` as an ``invoke_agent`` span and
+    Traces the 1.x agent runner or 2.x runtime as an ``invoke_agent`` span and
     optionally captures the turn's input/output messages. Model and tool
     calls are delegated to AgentScope and are not instrumented here.
     """
@@ -83,7 +96,7 @@ def instrumentation_dependencies(self) -> Collection[str]:
         return _instruments
 
     def _instrument(self, **kwargs: Any) -> None:
-        """Enable the ``AgentRunner.query_handler`` instrumentation.
+        """Enable instrumentation for the installed QwenPaw runtime.
 
         Args:
             **kwargs: Optional arguments
@@ -102,20 +115,28 @@ def _instrument(self, **kwargs: Any) -> None:
             completion_hook=kwargs.get("completion_hook")
             or load_completion_hook(),
         )
-        wrap_function_wrapper(
-            _RUNNER_MODULE,
-            "AgentRunner.query_handler",
-            make_query_handler_wrapper(handler),
-        )
-        logger.debug(
-            "Instrumented %s.AgentRunner.query_handler", _RUNNER_MODULE
-        )
+        if _runtime_v2():
+            wrap_function_wrapper(
+                _RUNTIME_MODULE,
+                "Runtime._build_context",
+                observe_context_build,
+            )
+            wrap_function_wrapper(
+                _RUNTIME_MODULE, "Runtime.run", make_runtime_wrapper(handler)
+            )
+        else:
+            wrap_function_wrapper(
+                _RUNNER_MODULE,
+                "AgentRunner.query_handler",
+                make_query_handler_wrapper(handler),
+            )
 
     def _uninstrument(self, **kwargs: Any) -> None:
-        """Disable the ``AgentRunner.query_handler`` instrumentation."""
+        """Disable instrumentation for the installed QwenPaw runtime."""
         del kwargs
-        runner_module = import_module(_RUNNER_MODULE)
-        unwrap(runner_module.AgentRunner, "query_handler")
-        logger.debug(
-            "Uninstrumented %s.AgentRunner.query_handler", _RUNNER_MODULE
-        )
+        if _runtime_v2():
+            runtime = import_module(_RUNTIME_MODULE).Runtime
+            unwrap(runtime, "run")
+            unwrap(runtime, "_build_context")
+        else:
+            unwrap(import_module(_RUNNER_MODULE).AgentRunner, "query_handler")

instrumentation/opentelemetry-instrumentation-genai-qwenpaw/src/opentelemetry/instrumentation/genai/qwenpaw/package.py

@@ -1,6 +1,6 @@
 # Copyright The OpenTelemetry Authors
 # SPDX-License-Identifier: Apache-2.0
 
-_instruments = ("qwenpaw >= 1.1.0, < 2.0.0; python_version < '3.14'",)
+_instruments = ("qwenpaw >= 1.1.0, < 3.0.0; python_version < '3.14'",)
 
 _supports_metrics = True