Promote staging to production (da36f4e)

.github/workflows/test.yml

@@ -62,16 +62,24 @@ jobs:
           grep -E '^(pytest|pytest-asyncio|pytest-xdist|respx|dirty-equals|time-machine)==' requirements-dev.lock > test-tools.txt
           # Resolve as of a day ago: a release still uploading its wheels forces a
           # source build that can fail (SQLAlchemy 2.1.0 did).
-          # google-adk<2: 2.x wraps callback exceptions and fails test_user_exception_propagates_without_chaining.
           uv pip install --exclude-newer "$(date -u -d '1 day ago' +%Y-%m-%dT%H:%M:%SZ)" \
             -e '.[aiohttp]' -r test-tools.txt \
-            openai boto3 claude-agent-sdk 'google-adk<2' google-genai langchain-core wrapt \
+            openai boto3 claude-agent-sdk google-adk google-genai langchain-core wrapt \
             azure-cognitiveservices-speech
 
       # Generated API tests need the mock server and already run in `test`.
       - name: Run integration tests
         run: .venv/bin/pytest --ignore=tests/api_resources
 
+      # google-adk 2.x runs agents as workflow nodes and moved the patch targets,
+      # so the ADK tracer also has to keep working on 1.x. Runs even when the
+      # step above fails, and resolves as of a day ago for the same reason.
+      - name: Run Google ADK tests on google-adk 1.x
+        if: ${{ !cancelled() }}
+        run: |
+          uv pip install --exclude-newer "$(date -u -d '1 day ago' +%Y-%m-%dT%H:%M:%SZ)" 'google-adk<2'
+          .venv/bin/pytest tests/test_google_adk_handoffs.py tests/test_google_adk_integration.py tests/test_google_genai_adk_dedup.py
+
   # The one check to require in rulesets, so jobs can be added or renamed freely.
   # The integrations job stays advisory until its optional dependencies are pinned.
   # always(), not !cancelled(): a skipped required check would count as passing.

examples/tracing/google-adk/google_adk_tracing.ipynb

@@ -498,11 +498,15 @@
     "\n",
     "After running the callback agent, you'll see in your Openlayer dashboard:\n",
     "\n",
-    "1. **Agent Step** (`CallbackDemoAgent`):\n",
+    "1. **Agent turn step** (`Agent turn: <app_name>`): one per `runner.run_async` call, holding\n",
+    "   the user's message, the final answer and the routing columns (`starting_agent`,\n",
+    "   `final_agent`, `handoff_count`, `handoff_path`)\n",
+    "\n",
+    "2. **Agent Step** (`CallbackDemoAgent`), inside the turn:\n",
     "   - Shows which callbacks are registered\n",
     "   - Contains all nested steps in chronological order\n",
     "\n",
-    "2. **All Callbacks as Siblings** (direct children of Agent):\n",
+    "3. **All Callbacks as Siblings** (direct children of Agent):\n",
     "   - `Callback: before_agent` - First, before any processing\n",
     "   - `Callback: before_model` - Before each LLM call\n",
     "   - `LLM Call: gemini-2.0-flash-exp` - The actual LLM invocation\n",
@@ -512,9 +516,9 @@
     "   - `Callback: after_tool` - After tool completion\n",
     "   - `Callback: after_agent` - Last, after all processing complete\n",
     "\n",
-    "3. **Token Usage** (captured on LLM Call steps):\n",
+    "4. **Token Usage** (captured on LLM Call steps):\n",
     "   - Prompt tokens\n",
-    "   - Completion tokens  \n",
+    "   - Completion tokens (including thinking tokens)\n",
     "   - Total tokens\n"
    ]
   },
@@ -538,21 +542,24 @@
     "\n",
     "The traces will show the hierarchy of operations in chronological order:\n",
     "```\n",
-    "Agent: CallbackDemoAgent\n",
-    "├── Callback: before_agent (CallbackDemoAgent)\n",
-    "├── Callback: before_model (CallbackDemoAgent)\n",
-    "├── LLM Call: gemini-2.0-flash-exp\n",
-    "├── Callback: after_model (CallbackDemoAgent)\n",
-    "├── Callback: before_tool (CallbackDemoAgent)\n",
-    "├── Tool: get_current_time\n",
-    "├── Callback: after_tool (CallbackDemoAgent)\n",
-    "├── Callback: before_model (CallbackDemoAgent)\n",
-    "├── LLM Call: gemini-2.0-flash-exp\n",
-    "├── Callback: after_model (CallbackDemoAgent)\n",
-    "└── Callback: after_agent (CallbackDemoAgent)\n",
+    "Agent turn: <app_name>\n",
+    "└── Agent: CallbackDemoAgent\n",
+    "    ├── Callback: before_agent (CallbackDemoAgent)\n",
+    "    ├── Callback: before_model (CallbackDemoAgent)\n",
+    "    ├── LLM Call: gemini-2.0-flash-exp\n",
+    "    ├── Callback: after_model (CallbackDemoAgent)\n",
+    "    ├── Callback: before_tool (CallbackDemoAgent)\n",
+    "    ├── Tool: get_current_time\n",
+    "    ├── Callback: after_tool (CallbackDemoAgent)\n",
+    "    ├── Callback: before_model (CallbackDemoAgent)\n",
+    "    ├── LLM Call: gemini-2.0-flash-exp\n",
+    "    ├── Callback: after_model (CallbackDemoAgent)\n",
+    "    └── Callback: after_agent (CallbackDemoAgent)\n",
     "```\n",
     "\n",
-    "**Note:** All callbacks are direct children of the Agent step, appearing in chronological order alongside LLM calls and tool executions. This provides a clear timeline view of the agent execution flow.\n"
+    "**Note:** All callbacks are direct children of the Agent step, appearing in chronological order alongside LLM calls and tool executions. This provides a clear timeline view of the agent execution flow.\n",
+    "\n",
+    "With sub-agents, each transfer ADK makes is a `Handoff: <from> → <to>` step inside the transferring agent, and the agent that takes over is the next step of the turn. Agents used as tools (`AgentTool`) stay nested under their Tool step and aren't counted as handoffs.\n"
    ]
   },
   {