feat(models): add Azure OpenAI Responses API support

README.md

@@ -42,6 +42,8 @@ services: Integrations with external services, like databases, vector stores, or
 
 tools: Standalone tools that can be used by agents.
 
+models: Model adapters for third-party model APIs.
+
 deployment: Scripts and configurations to help you deploy your ADK agents to various platforms.
 
 # We Welcome Your Contributions!

contributing/samples/models/hello_world_azure_openai_responses/.gitignore

@@ -0,0 +1,2 @@
+.adk/
+__pycache__/

contributing/samples/models/hello_world_azure_openai_responses/README.md

@@ -0,0 +1,142 @@
+# Hello World with Azure OpenAI Responses
+
+## Overview
+
+This sample is a manual end-to-end test for the community Azure OpenAI
+Responses API model client. It validates that ADK can run an agent with
+`AzureOpenAIResponsesLlm`, execute a custom function tool, preserve multi-turn
+context through the runner, and directly chain Responses API calls with
+`previous_response_id`.
+
+## Sample Inputs
+
+- `Reply with exactly: AZURE_RESPONSES_TEXT_OK`
+
+  Verifies basic text generation through `Runner.run_async`.
+
+- `Use the weather tool for Tokyo and report the result.`
+
+  Verifies Responses API function calling and function result handling.
+
+- `What manual E2E code phrase did I ask you to remember?`
+
+  Verifies multi-turn runner context after a previous turn stores the phrase
+  `cobalt otter`.
+
+- `Using the previous response context, what direct Responses code phrase did I ask you to remember?`
+
+  Verifies direct model-level `previous_response_id` chaining after a prior
+  Responses API call stores the phrase `amber swan`.
+
+## Graph
+
+```mermaid
+graph TD
+    User --> Runner
+    Runner --> Agent[azure_openai_responses_agent]
+    Agent --> Model[AzureOpenAIResponsesLlm]
+    Agent --> Tool[get_current_weather]
+    Model --> Responses[Azure OpenAI Responses API]
+```
+
+## How To
+
+Create a `.env` file in this directory or export the variables in your shell:
+
+```bash
+export AZURE_OPENAI_API_KEY="your-api-key"
+export AZURE_OPENAI_ENDPOINT="https://your-resource.openai.azure.com"
+export AZURE_OPENAI_RESPONSES_DEPLOYMENT="your-model-deployment"
+```
+
+Install this repository with the OpenAI extra before running the sample:
+
+```bash
+uv sync --extra openai
+```
+
+Run the manual E2E test from this sample directory:
+
+```bash
+cd contributing/samples/models/hello_world_azure_openai_responses
+python main.py
+```
+
+The test prints the runner setup, tool calls, tool responses, interaction IDs,
+and pass/fail status for each scenario.
+
+Expected output shape:
+
+```text
+Azure OpenAI Responses manual E2E
+Endpoint: https://your-resource.openai.azure.com
+Deployment: your-model-deployment
+
+========================================================================
+TEST 1: Runner Basic Text Generation
+========================================================================
+>> User: Reply with exactly: AZURE_RESPONSES_TEXT_OK
+<< Agent: AZURE_RESPONSES_TEXT_OK
+PASSED: Runner basic text generation works
+
+========================================================================
+TEST 2: Runner Function Calling
+========================================================================
+>> User: Use the weather tool for Tokyo and report the result.
+   [Tool Call] get_current_weather({'city': 'Tokyo'})
+   [Tool Result] get_current_weather: {'city': 'Tokyo', ...}
+<< Agent: Tokyo ... 68 ... Partly Cloudy
+PASSED: Runner function calling works
+
+========================================================================
+TEST 3: Runner Multi-Turn Context
+========================================================================
+PASSED: Runner multi-turn context works
+
+========================================================================
+TEST 4: Direct Responses previous_response_id Chaining
+========================================================================
+   First response id: resp_...
+<< Direct model: amber swan
+   Second response id: resp_...
+PASSED: Direct previous_response_id chaining works
+
+========================================================================
+ALL MANUAL E2E TESTS PASSED
+```
+
+## Manual E2E PR Evidence
+
+For PR validation, include the exact command and relevant console output:
+
+```bash
+cd contributing/samples/models/hello_world_azure_openai_responses
+python main.py
+```
+
+Highlight these log sections in the PR description:
+
+- `TEST 1: Runner Basic Text Generation`
+- `TEST 2: Runner Function Calling`, including `[Tool Call]` and `[Tool Result]`
+- `TEST 3: Runner Multi-Turn Context`
+- `TEST 4: Direct Responses previous_response_id Chaining`, including response IDs
+- `ALL MANUAL E2E TESTS PASSED`
+
+## Code Structure
+
+```text
+hello_world_azure_openai_responses/
+├── __init__.py
+├── agent.py
+├── main.py
+└── README.md
+```
+
+## Notes
+
+- This is a manual E2E sample and requires a live Azure OpenAI deployment that
+  supports the Responses API.
+- `AZURE_OPENAI_RESPONSES_DEPLOYMENT` must be the Azure deployment name, not
+  necessarily the base model name.
+- The sample sets `store=True` on `AzureOpenAIResponsesLlm` so the direct
+  `previous_response_id` check can chain server-side Responses API state.

contributing/samples/models/hello_world_azure_openai_responses/__init__.py

@@ -0,0 +1,19 @@
+# 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
+
+__all__ = [
+    'agent',
+]

contributing/samples/models/hello_world_azure_openai_responses/agent.py

@@ -0,0 +1,85 @@
+# 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.
+
+"""Azure OpenAI Responses API sample agent."""
+
+from __future__ import annotations
+
+import os
+
+from google.adk.agents.llm_agent import Agent
+from google.genai import types
+from google.genai.types import GenerateContentConfig
+
+from google.adk_community.models.openai_responses import AzureOpenAIResponsesLlm
+
+
+def get_current_weather(city: str) -> dict:
+  """Get deterministic weather for a city.
+
+  Args:
+    city: The city to look up.
+
+  Returns:
+    A dictionary containing weather information for the requested city.
+  """
+  weather_data = {
+      'london': {'temperature_f': 59, 'condition': 'Cloudy'},
+      'paris': {'temperature_f': 64, 'condition': 'Rainy'},
+      'san francisco': {'temperature_f': 70, 'condition': 'Sunny'},
+      'tokyo': {'temperature_f': 68, 'condition': 'Partly Cloudy'},
+  }
+  data = weather_data.get(
+      city.lower(), {'temperature_f': 72, 'condition': 'Unknown'}
+  )
+  return {
+      'city': city,
+      'temperature_f': data['temperature_f'],
+      'condition': data['condition'],
+  }
+
+
+def get_azure_openai_responses_model() -> AzureOpenAIResponsesLlm:
+  """Builds the Azure OpenAI Responses model from environment variables."""
+  return AzureOpenAIResponsesLlm(
+      model=os.environ.get('AZURE_OPENAI_RESPONSES_DEPLOYMENT', 'gpt-5.5'),
+      azure_endpoint=os.environ.get('AZURE_OPENAI_ENDPOINT'),
+      api_key=os.environ.get('AZURE_OPENAI_API_KEY'),
+      store=True,
+      reasoning={'effort': 'medium', 'summary': 'concise'},
+  )
+
+
+root_agent = Agent(
+    model=get_azure_openai_responses_model(),
+    name='azure_openai_responses_agent',
+    description=(
+        'Manual E2E sample agent for the Azure OpenAI Responses API community'
+        ' model.'
+    ),
+    instruction="""
+You are a concise test assistant for ADK manual E2E validation.
+
+Rules:
+- For exact-response prompts, return only the requested text.
+- When the user asks about weather, call get_current_weather.
+- When reporting weather, include the city, temperature_f, and condition from
+  the tool result.
+- If the user asks what they told you earlier, answer from conversation context.
+""",
+    tools=[get_current_weather],
+    generate_content_config=GenerateContentConfig(
+        max_output_tokens=11512,
+    ),
+)