Merge branch 'main' into fix/litellm-content-none-parts

Author: rohityan

.github/.release-please-manifest.json

@@ -1,3 +1,3 @@
 {
-  ".": "2.3.0"
+  ".": "2.4.0"
 }

.github/release-please-config.json

@@ -57,5 +57,5 @@
       ]
     }
   },
-  "last-release-sha": "0cb4c814928f579bfbac9b9e1f95669e4304e089"
+  "last-release-sha": "44d747ed5eaf543b5b8d22e0088f8a7c7eeee846"
 }

.github/workflows/release-update-adk-web.yaml

@@ -37,6 +37,8 @@ jobs:
     steps:
     - name: Checkout repository
       uses: actions/checkout@v6
+      with:
+        persist-credentials: false
 
     - name: Fetch and unzip frontend assets
       run: |

CHANGELOG.md

@@ -167,7 +167,7 @@ def __call__(self, messages: list[dict[str, Any]]) -> dict[str, Any]:
     Returns:
       A dictionary containing rating information including score.
     """
-    env = jinja2.Environment()
+    env = jinja2.Environment(autoescape=True)
     env.globals['user_input'] = (
         messages[0].get('parts', [{}])[0].get('text', '') if messages else ''
     )

contributing/samples/integrations/gepa/rater_lib.py

@@ -0,0 +1,79 @@
+# Copyright 2026 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import sys
+from unittest import mock
+
+# Mock the retry module before importing rater_lib
+mock_retry_module = mock.MagicMock()
+
+
+def mock_retry_decorator(*args, **kwargs):
+  def decorator(func):
+    return func
+
+  return decorator
+
+
+mock_retry_module.retry = mock_retry_decorator
+sys.modules["retry"] = mock_retry_module
+
+from gepa import rater_lib
+
+
+def test_rater_escapes_html_inputs_to_prevent_xss():
+  """Rater escapes HTML tags in user and model inputs to prevent XSS.
+
+  Setup: Mock genai.Client to return a dummy rating response.
+  Act: Call Rater with messages containing HTML tags.
+  Assert: Verify that the rendered prompt template contains escaped HTML.
+  """
+  # Arrange
+  with mock.patch("google.genai.Client") as mock_client_cls:
+    mock_client = mock_client_cls.return_value
+    mock_generate = mock_client.models.generate_content
+    mock_generate.return_value.text = (
+        "Property: The agent fulfilled the user's primary request.\n"
+        "Evidence: mock evidence\n"
+        "Rationale: mock rationale\n"
+        "Verdict: yes"
+    )
+
+    template_path = (
+        "contributing/samples/integrations/gepa/rubric_validation_template.txt"
+    )
+    rater = rater_lib.Rater(
+        tool_declarations="[]", validation_template_path=template_path
+    )
+
+    messages = [
+        {
+            "role": "user",
+            "parts": [{"text": "<script>alert('XSS')</script>"}],
+        },
+        {
+            "role": "model",
+            "parts": [{"text": "Hello <img src=x onerror=alert(1)>"}],
+        },
+    ]
+
+    # Act
+    rater(messages)
+
+    # Assert
+    assert mock_generate.called
+    call_args = mock_generate.call_args
+    contents = call_args.kwargs["contents"]
+
+    assert "&lt;script&gt;alert(&#39;XSS&#39;)&lt;/script&gt;" in contents
+    assert "Hello &lt;img src=x onerror=alert(1)&gt;" in contents

contributing/samples/integrations/gepa/rater_lib_test.py

@@ -155,12 +155,12 @@ Verdict: no
   </available_tools>
 
   <main_prompt>
-  {{user_input}}
+  {{user_input|e}}
   </main_prompt>
 </user_prompt>
 
 <responses>
-{{model_response}}
+{{model_response|e}}
 </responses>
 
 <properties>

contributing/samples/integrations/gepa/rubric_validation_template.txt

@@ -0,0 +1,47 @@
+# Managed Agent
+
+## Overview
+
+This sample runs a `ManagedAgent` configured with the built-in `google_search`
+tool. Given an open-ended request, the server-side harness autonomously issues
+many searches and synthesizes the result in a single turn.
+
+## Sample Inputs
+
+- `Compare the current flagship smartphones from Apple, Samsung, and Google. For each, find its launch price, display size, and main rear camera resolution, then recommend the best value for someone who mostly takes photos.`
+
+  The showcase input. A single question the harness answers by fanning out into a
+  dozen-odd searches. It does broad discovery first, then targeted per-model spec
+  and price lookups, self-correcting when it hits a stale model, before composing
+  a comparison table and a reasoned recommendation.
+
+- `Which of those would you pick for shooting video instead?`
+
+  A follow-up turn that reuses the recovered remote sandbox and the previous
+  interaction, continuing the same research thread. This demonstrates multi-turn
+  chaining.
+
+## Graph
+
+```mermaid
+graph LR
+    User -->|message| ManagedAgent
+    ManagedAgent -->|interactions.create| ManagedAgentsAPI
+    ManagedAgentsAPI -->|server-side research loop: google_search ×N| ManagedAgentsAPI
+    ManagedAgentsAPI -->|streamed events| ManagedAgent
+    ManagedAgent -->|answer| User
+```
+
+## How To
+
+- **Create the agent**: instantiate `ManagedAgent` with an `agent_id`, an
+  `environment` spec, and a list of server-side `tools`. No `model` is set; the
+  model is part of the managed agent on the server.
+- **Provision a sandbox**: `environment={'type': 'remote'}` requests a fresh
+  remote sandbox. The resulting environment id is stored on emitted events, so
+  subsequent turns automatically recover and reuse it.
+- **Multi-turn chaining**: the agent recovers the `previous_interaction_id` from
+  the session events, so follow-up turns continue the same interaction without
+  any extra wiring.
+- **Drive it**: a `ManagedAgent` is a `BaseAgent`, so a standard `Runner` runs
+  it just like any other agent.

contributing/samples/managed_agent/basic/README.md

@@ -0,0 +1,15 @@
+# Copyright 2026 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from . import agent

contributing/samples/managed_agent/basic/__init__.py

@@ -0,0 +1,49 @@
+# Copyright 2026 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""A ManagedAgent backed by the Managed Agents API (interactions.create).
+
+``ManagedAgent`` calls the Managed Agents API directly from its run loop instead
+of running a local model loop. It currently supports server-side tools only
+(ADK built-in tools and raw ``google.genai.types.Tool`` configs); here we wire
+up ``google_search``, which runs entirely on the server.
+
+A fresh remote sandbox is provisioned via ``environment={'type': 'remote'}``;
+the environment id is recovered from prior events so multi-turn conversations
+reuse the same sandbox.
+
+Run with ``adk web`` / ``adk run contributing/samples/managed_agent/basic``. See
+the README for the required environment / auth setup.
+"""
+
+import os
+
+from google.adk.agents import ManagedAgent
+from google.adk.tools import google_search
+
+# The Managed Agent id served by the Managed Agents API. Override with the
+# MANAGED_AGENT_ID environment variable if your project has access to a
+# different agent.
+_DEFAULT_AGENT_ID = 'antigravity-preview-05-2026'
+
+root_agent = ManagedAgent(
+    name='managed_search_agent',
+    agent_id=os.environ.get('MANAGED_AGENT_ID', _DEFAULT_AGENT_ID),
+    # Provision a remote sandbox for the agent. The environment id is recovered
+    # from prior events, so follow-up turns reuse the same sandbox.
+    environment={'type': 'remote'},
+    # Only server-side tools are supported today. google_search is an ADK
+    # built-in tool that executes on the server.
+    tools=[google_search],
+)