Merge pull request #944 from MicrosoftDocs/main

Refresh Agent Framework PR 4818 Python docs (#936)

Author: eavanvalkenburg

agent-framework/AGENTS.md

@@ -205,6 +205,7 @@ Every docs page maps to sample files in both repos:
 | `agents/tools/local-mcp-tools.md` | `02-agents/tools/local_mcp_tools.py` | `02-agents/tools/LocalMcpTools.cs` |
 | `agents/tools/tool-approval.md` | `02-agents/tools/tool_approval.py` | `02-agents/tools/ToolApproval.cs` |
 | `agents/middleware/*.md` | `02-agents/middleware/<matching>.py` | `02-agents/middleware/<matching>.cs` |
+| `agents/providers/foundry-local.md` | `02-agents/providers/foundry/foundry_local_agent.py` | N/A |
 | `agents/providers/*.md` | `02-agents/providers/<matching>.py` | `02-agents/providers/<matching>.cs` |
 | `agents/conversations/*.md` | `02-agents/conversations/<matching>.py` | `02-agents/conversations/<matching>.cs` |
 | `workflows/<pattern>.md` | `03-workflows/<pattern>/<matching>.py` | `03-workflows/<pattern>/<matching>.cs` |

agent-framework/TOC.yml

@@ -102,6 +102,8 @@ items:
       href: agents/providers/openai.md
     - name: Microsoft Foundry
       href: agents/providers/microsoft-foundry.md
+    - name: Foundry Local
+      href: agents/providers/foundry-local.md
     - name: Anthropic
       href: agents/providers/anthropic.md
     - name: Ollama

agent-framework/agents/agent-pipeline.md

@@ -190,12 +190,12 @@ The `Agent` class accepts any client that implements `SupportsChatGetResponse`.
 
 ```python
 from agent_framework import Agent
-from agent_framework.azure import AzureOpenAIResponsesClient
+from agent_framework.foundry import FoundryChatClient
 
-client = AzureOpenAIResponsesClient(
+client = FoundryChatClient(
     credential=credential,
     project_endpoint=endpoint,
-    deployment_name=model,
+    model=model,
 )
 
 agent = Agent(client=client, instructions="You are helpful.")

agent-framework/agents/background-responses.md

@@ -172,12 +172,12 @@ For non-streaming scenarios, when you initially run an agent with `background=Tr
 ```python
 import asyncio
 from agent_framework import Agent
-from agent_framework.openai import OpenAIResponsesClient
+from agent_framework.openai import OpenAIChatClient
 
 agent = Agent(
     name="researcher",
     instructions="You are a helpful research assistant.",
-    client=OpenAIResponsesClient(model_id="o3"),
+    client=OpenAIChatClient(model="o3"),
 )
 
 session = agent.create_session()

agent-framework/agents/conversations/compaction.md

@@ -59,7 +59,7 @@ Compaction solves these problems by selectively removing, collapsing, or summari
 
 Compaction applies only to agents that manage their own conversation history in memory. Agents that rely on service-managed context or conversation state do not benefit from compaction because the service already handles context management. Examples of service-managed agents include:
 
-- **Foundry Agents** — context is managed server-side by the Azure AI Foundry service.
+- **Foundry Agents** — context is managed server-side by the Microsoft Foundry service.
 - **Responses API with store enabled** (the default) — conversation state is stored and managed by the OpenAI service.
 - **Copilot Studio agents** — conversation context is maintained by the Copilot Studio service.
 

agent-framework/agents/index.md

@@ -101,7 +101,7 @@ When using Foundry, Azure OpenAI, OpenAI services, or Anthropic services, you ha
 | [Azure OpenAI](/azure/ai-foundry/openai/overview) <sup>1</sup> | Azure OpenAI SDK <sup>2</sup> | [Azure.AI.OpenAI](https://www.nuget.org/packages/Azure.AI.OpenAI) | https://&lt;resource&gt;.openai.azure.com/ |
 | [Azure OpenAI](/azure/ai-foundry/openai/overview) <sup>1</sup> | OpenAI SDK | [OpenAI](https://www.nuget.org/packages/OpenAI) | https://&lt;resource&gt;.openai.azure.com/openai/v1/ |
 | OpenAI | OpenAI SDK | [OpenAI](https://www.nuget.org/packages/OpenAI) | No url required |
-| [Azure AI Foundry Anthropic](/azure/ai-foundry/foundry-models/how-to/use-foundry-models-claude) | Anthropic Foundry SDK | [Anthropic.Foundry](https://www.nuget.org/packages/Anthropic.Foundry) | Resource name required |
+| [Microsoft Foundry Anthropic](/azure/ai-foundry/foundry-models/how-to/use-foundry-models-claude) | Anthropic Foundry SDK | [Anthropic.Foundry](https://www.nuget.org/packages/Anthropic.Foundry) | Resource name required |
 | Anthropic | Anthropic SDK | [Anthropic](https://www.nuget.org/packages/Anthropic) | No url or resource name required |
 
 1. [Upgrading from Azure OpenAI to Foundry](/azure/ai-foundry/how-to/upgrade-azure-openai)
@@ -169,7 +169,7 @@ AIAgent agent = await persistentAgentsClient.CreateAIAgentAsync(
     name: "Joker");
 ```
 
-### Using the Azure AI Foundry Anthropic SDK
+### Using the Foundry Anthropic SDK
 
 The resource is the subdomain name / first name coming before '.services.ai.azure.com' in the endpoint Uri.
 
@@ -217,29 +217,37 @@ To create one of these agents, simply construct an `Agent` using the chat client
 ```python
 import os
 from agent_framework import Agent
-from agent_framework.azure import AzureOpenAIResponsesClient
+from agent_framework.foundry import FoundryChatClient
 from azure.identity.aio import DefaultAzureCredential
 
-Agent(
-    client=AzureOpenAIResponsesClient(credential=DefaultAzureCredential(), project_endpoint=os.getenv("AZURE_AI_PROJECT_ENDPOINT"), deployment_name=os.getenv("AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME")),
-    instructions="You are a helpful assistant"
-) as agent
+agent = Agent(
+    client=FoundryChatClient(
+        credential=DefaultAzureCredential(),
+        project_endpoint=os.getenv("AZURE_AI_PROJECT_ENDPOINT"),
+        model=os.getenv("AZURE_AI_MODEL_DEPLOYMENT_NAME"),
+    ),
+    instructions="You are a helpful assistant",
+)
 response = await agent.run("Hello!")
 ```
 
 Alternatively, you can use the convenience method on the chat client:
 
 ```python
-from agent_framework.azure import AzureOpenAIResponsesClient
+from agent_framework.foundry import FoundryChatClient
 from azure.identity.aio import DefaultAzureCredential
 
-agent = AzureOpenAIResponsesClient(credential=DefaultAzureCredential(), project_endpoint=os.getenv("AZURE_AI_PROJECT_ENDPOINT"), deployment_name=os.getenv("AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME")).as_agent(
+agent = FoundryChatClient(
+    credential=DefaultAzureCredential(),
+    project_endpoint=os.getenv("AZURE_AI_PROJECT_ENDPOINT"),
+    model=os.getenv("AZURE_AI_MODEL_DEPLOYMENT_NAME"),
+).as_agent(
     instructions="You are a helpful assistant"
 )
 ```
 
 > [!NOTE]
-> This example shows using the AzureOpenAIResponsesClient, but the same pattern applies to any chat client that implements `SupportsChatGetResponse`, see [providers overview](./providers/index.md) for more details on other clients.
+> This example shows using the FoundryChatClient, but the same pattern applies to any chat client that implements `SupportsChatGetResponse`, see [providers overview](./providers/index.md) for more details on other clients.
 
 For detailed examples, see the agent-specific documentation sections below.
 
@@ -293,17 +301,17 @@ You can provide function tools to agents for enhanced capabilities:
 import os
 from typing import Annotated
 from azure.identity.aio import DefaultAzureCredential
-from agent_framework.azure import AzureOpenAIResponsesClient
+from agent_framework.foundry import FoundryChatClient
 
 def get_weather(location: Annotated[str, "The location to get the weather for."]) -> str:
     """Get the weather for a given location."""
     return f"The weather in {location} is sunny with a high of 25°C."
 
 async with DefaultAzureCredential() as credential:
-    agent = AzureOpenAIResponsesClient(
+    agent = FoundryChatClient(
         credential=credential,
         project_endpoint=os.getenv("AZURE_AI_PROJECT_ENDPOINT"),
-        deployment_name=os.getenv("AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME"),
+        model=os.getenv("AZURE_AI_MODEL_DEPLOYMENT_NAME"),
     ).as_agent(
         instructions="You are a helpful weather assistant.",
         tools=get_weather,

agent-framework/agents/middleware/result-overrides.md

@@ -94,7 +94,7 @@ from agent_framework import (
     ResponseStream,
     tool,
 )
-from agent_framework.openai import OpenAIResponsesClient
+from agent_framework.openai import OpenAIChatClient
 from pydantic import Field
 
 """
@@ -260,7 +260,7 @@ async def main() -> None:
 
     # For authentication, run `az login` command in terminal or replace AzureCliCredential with preferred
     # authentication option.
-    agent = OpenAIResponsesClient(
+    agent = OpenAIChatClient(
         middleware=[validate_weather_middleware, weather_override_middleware],
     ).as_agent(
         name="WeatherAgent",
@@ -316,7 +316,7 @@ from agent_framework import (
     ResponseStream,
     tool,
 )
-from agent_framework.openai import OpenAIResponsesClient
+from agent_framework.openai import OpenAIChatClient
 from pydantic import Field
 
 """
@@ -482,7 +482,7 @@ async def main() -> None:
 
     # For authentication, run `az login` command in terminal or replace AzureCliCredential with preferred
     # authentication option.
-    agent = OpenAIResponsesClient(
+    agent = OpenAIChatClient(
         middleware=[validate_weather_middleware, weather_override_middleware],
     ).as_agent(
         name="WeatherAgent",

agent-framework/agents/middleware/runtime-context.md

@@ -49,7 +49,7 @@ Use the smallest surface that fits the data. This keeps tool inputs explicit and
 from typing import Annotated
 
 from agent_framework import FunctionInvocationContext, tool
-from agent_framework.openai import OpenAIResponsesClient
+from agent_framework.openai import OpenAIChatClient
 
 
 @tool(approval_mode="never_require")
@@ -62,7 +62,7 @@ def send_email(
     return f"Queued email for {address} from {user_id} ({tenant})"
 
 
-agent = OpenAIResponsesClient().as_agent(
+agent = OpenAIChatClient().as_agent(
     name="Notifier",
     instructions="Send email updates.",
     tools=[send_email],
@@ -93,7 +93,7 @@ Function middleware uses the same `FunctionInvocationContext` object that tools
 from collections.abc import Awaitable, Callable
 
 from agent_framework import FunctionInvocationContext
-from agent_framework.openai import OpenAIResponsesClient
+from agent_framework.openai import OpenAIChatClient
 
 
 async def enrich_tool_runtime_context(
@@ -105,7 +105,7 @@ async def enrich_tool_runtime_context(
     await call_next()
 
 
-agent = OpenAIResponsesClient().as_agent(
+agent = OpenAIChatClient().as_agent(
     name="Notifier",
     instructions="Send email updates.",
     tools=[send_email],
@@ -121,7 +121,7 @@ The middleware contract uses `call_next()` with no arguments. Mutate `context.kw
 from typing import Annotated
 
 from agent_framework import FunctionInvocationContext, tool
-from agent_framework.openai import OpenAIResponsesClient
+from agent_framework.openai import OpenAIChatClient
 
 
 @tool(approval_mode="never_require")
@@ -136,7 +136,7 @@ def remember_topic(
     return f"Stored {topic!r} in session state."
 
 
-agent = OpenAIResponsesClient().as_agent(
+agent = OpenAIChatClient().as_agent(
     name="MemoryAgent",
     instructions="Remember important topics.",
     tools=[remember_topic],
@@ -155,7 +155,7 @@ When an agent is exposed as a tool via `as_tool()`, runtime function kwargs alre
 
 ```python
 from agent_framework import FunctionInvocationContext, tool
-from agent_framework.openai import OpenAIResponsesClient
+from agent_framework.openai import OpenAIChatClient
 
 
 @tool(description="Store findings for later steps.")
@@ -164,7 +164,7 @@ def store_findings(findings: str, ctx: FunctionInvocationContext) -> None:
         ctx.session.state["findings"] = findings
 
 
-client = OpenAIResponsesClient()
+client = OpenAIChatClient()
 
 research_agent = client.as_agent(
     name="ResearchAgent",

agent-framework/agents/multimodal.md

@@ -62,10 +62,16 @@ First, create an agent that is able to analyze images.
 
 ```python
 import asyncio
-from agent_framework.azure import AzureOpenAIChatClient
+import os
+from agent_framework.openai import OpenAIChatCompletionClient
 from azure.identity import AzureCliCredential
 
-agent = AzureOpenAIChatClient(credential=AzureCliCredential()).as_agent(
+agent = OpenAIChatCompletionClient(
+    model=os.environ["AZURE_OPENAI_DEPLOYMENT_NAME"],
+    azure_endpoint=os.environ["AZURE_OPENAI_ENDPOINT"],
+    api_version=os.getenv("AZURE_OPENAI_API_VERSION"),
+    credential=AzureCliCredential(),
+).as_agent(
     name="VisionAgent",
     instructions="You are a helpful agent that can analyze images"
 )