Port to Responses API (and remove GitHub Models)

.env-sample

@@ -1,25 +1,23 @@
 # API Host Selection
-# Options: github, azure, ollama, openai
-API_HOST=github
-
-# GitHub Models Configuration
-GITHUB_TOKEN=your_github_token_here
-GITHUB_MODEL=gpt-4o
+# Options: azure, ollama, openai
+API_HOST=azure
 
 # Azure OpenAI Configuration
 AZURE_OPENAI_ENDPOINT=https://your-resource.openai.azure.com/
 AZURE_OPENAI_CHAT_DEPLOYMENT=your-deployment-name
-AZURE_OPENAI_VERSION=2024-02-15-preview
 
 # Ollama Configuration
-OLLAMA_MODEL=llama3.1
+OLLAMA_MODEL=qwen3.5:9b
 OLLAMA_ENDPOINT=http://localhost:11434/v1
-OLLAMA_API_KEY=ollama
+OLLAMA_API_KEY=no-key-needed
 
 # OpenAI Configuration (default if API_HOST not set)
-OPENAI_MODEL=gpt-4o-mini
+OPENAI_MODEL=gpt-5.2
 OPENAI_API_KEY=your_openai_api_key_here
 
+# GitHub MCP Configuration (for Langchain+GitHub example)
+GITHUB_TOKEN=your_github_token_here
+
 # OpenTelemetry Configuration (for Aspire Dashboard)
 # Uncomment to enable tracing, metrics, and logs export
 # OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317

README.md

@@ -68,9 +68,8 @@ If you're not using one of the above options, then you'll need to:
    ```
 
 6. Edit `.env` with your API credentials. Choose one of the following providers by setting `API_HOST`:
-   - `github` - GitHub Models (requires `GITHUB_TOKEN`)
    - `azure` - Azure OpenAI (requires Azure credentials)
-   - `ollama` - Local Ollama instance
+   - `ollama` - Local Ollama instance with Responses API support
    - `openai` - OpenAI API (requires `OPENAI_API_KEY`)
 
 ## Run local MCP servers
@@ -280,7 +279,7 @@ Pricing varies per region and usage, so it isn't possible to predict exact costs
 
 You can try the [Azure pricing calculator](https://azure.com/e/3987c81282c84410b491d28094030c9a) for the resources:
 
-- **Azure OpenAI Service**: S0 tier, GPT-4o-mini model. Pricing is based on token count. [Pricing](https://azure.microsoft.com/pricing/details/cognitive-services/openai-service/)
+- **Azure OpenAI Service**: S0 tier, GPT-5.2 model. Pricing is based on token count. [Pricing](https://azure.microsoft.com/pricing/details/cognitive-services/openai-service/)
 - **Azure Container Apps**: Consumption tier. [Pricing](https://azure.microsoft.com/pricing/details/container-apps/)
 - **Azure Container Registry**: Standard tier. [Pricing](https://azure.microsoft.com/pricing/details/container-registry/)
 - **Azure Cosmos DB**: Serverless tier. [Pricing](https://azure.microsoft.com/pricing/details/cosmos-db/)

agents/agentframework_http.py

@@ -4,7 +4,7 @@
 from datetime import datetime
 
 from agent_framework import Agent, MCPStreamableHTTPTool
-from agent_framework.openai import OpenAIChatClient
+from agent_framework.openai import OpenAIResponsesClient
 from azure.identity.aio import DefaultAzureCredential, get_bearer_token_provider
 from dotenv import load_dotenv
 from rich import print
@@ -24,33 +24,29 @@
 MCP_SERVER_URL = os.getenv("MCP_SERVER_URL", "http://localhost:8000/mcp/")
 
 # Configure chat client based on API_HOST
-API_HOST = os.getenv("API_HOST", "github")
+API_HOST = os.getenv("API_HOST", "azure")
 async_credential = None
 
 if API_HOST == "azure":
     async_credential = DefaultAzureCredential()
     token_provider = get_bearer_token_provider(async_credential, "https://cognitiveservices.azure.com/.default")
-    client = OpenAIChatClient(
+    client = OpenAIResponsesClient(
         base_url=f"{os.environ['AZURE_OPENAI_ENDPOINT']}/openai/v1/",
         api_key=token_provider,
         model_id=os.environ["AZURE_OPENAI_CHAT_DEPLOYMENT"],
     )
-elif API_HOST == "github":
-    client = OpenAIChatClient(
-        base_url="https://models.github.ai/inference",
-        api_key=os.environ["GITHUB_TOKEN"],
-        model_id=os.getenv("GITHUB_MODEL", "openai/gpt-4.1-mini"),
-    )
 elif API_HOST == "ollama":
-    client = OpenAIChatClient(
+    client = OpenAIResponsesClient(
         base_url=os.environ.get("OLLAMA_ENDPOINT", "http://localhost:11434/v1"),
-        api_key="none",
-        model_id=os.environ.get("OLLAMA_MODEL", "llama3.1:latest"),
+        api_key=os.getenv("OLLAMA_API_KEY", "no-key-needed"),
+        model_id=os.environ.get("OLLAMA_MODEL", "qwen3.5:9b"),
     )
-else:
-    client = OpenAIChatClient(
-        api_key=os.environ.get("OPENAI_API_KEY"), model_id=os.environ.get("OPENAI_MODEL", "gpt-4.1-mini")
+elif API_HOST == "openai":
+    client = OpenAIResponsesClient(
+        api_key=os.environ.get("OPENAI_API_KEY"), model_id=os.environ.get("OPENAI_MODEL", "gpt-5.2")
     )
+else:
+    raise ValueError(f"Unsupported API_HOST '{API_HOST}'. Use one of: azure, ollama, openai.")
 
 
 # --- Main Agent Logic ---

agents/agentframework_learn.py

@@ -3,7 +3,7 @@
 import os
 
 from agent_framework import Agent, MCPStreamableHTTPTool
-from agent_framework.openai import OpenAIChatClient
+from agent_framework.openai import OpenAIResponsesClient
 from azure.identity.aio import DefaultAzureCredential, get_bearer_token_provider
 from dotenv import load_dotenv
 from rich import print
@@ -18,32 +18,28 @@
 load_dotenv(override=True)
 
 # Configure chat client based on API_HOST
-API_HOST = os.getenv("API_HOST", "github")
+API_HOST = os.getenv("API_HOST", "azure")
 async_credential = None
 if API_HOST == "azure":
     async_credential = DefaultAzureCredential()
     token_provider = get_bearer_token_provider(async_credential, "https://cognitiveservices.azure.com/.default")
-    client = OpenAIChatClient(
+    client = OpenAIResponsesClient(
         base_url=f"{os.environ['AZURE_OPENAI_ENDPOINT']}/openai/v1/",
         api_key=token_provider,
         model_id=os.environ["AZURE_OPENAI_CHAT_DEPLOYMENT"],
     )
-elif API_HOST == "github":
-    client = OpenAIChatClient(
-        base_url="https://models.github.ai/inference",
-        api_key=os.environ["GITHUB_TOKEN"],
-        model_id=os.getenv("GITHUB_MODEL", "openai/gpt-4.1-mini"),
-    )
 elif API_HOST == "ollama":
-    client = OpenAIChatClient(
+    client = OpenAIResponsesClient(
         base_url=os.environ.get("OLLAMA_ENDPOINT", "http://localhost:11434/v1"),
-        api_key="none",
-        model_id=os.environ.get("OLLAMA_MODEL", "llama3.1:latest"),
+        api_key=os.getenv("OLLAMA_API_KEY", "no-key-needed"),
+        model_id=os.environ.get("OLLAMA_MODEL", "qwen3.5:9b"),
     )
-else:
-    client = OpenAIChatClient(
-        api_key=os.environ.get("OPENAI_API_KEY"), model_id=os.environ.get("OPENAI_MODEL", "gpt-4.1-mini")
+elif API_HOST == "openai":
+    client = OpenAIResponsesClient(
+        api_key=os.environ.get("OPENAI_API_KEY"), model_id=os.environ.get("OPENAI_MODEL", "gpt-5.2")
     )
+else:
+    raise ValueError(f"Unsupported API_HOST '{API_HOST}'. Use one of: azure, ollama, openai.")
 
 
 async def http_mcp_example() -> None:

agents/langchainv1_github.py

@@ -1,7 +1,8 @@
 """LangChain MCP Tool Filtering Example
 
 Demonstrates how to filter MCP tools to create safe, focused agents.
-Shows filtering for read-only research agent using GitHub MCP server.
+Shows filtering for a read-only research agent using the GitHub MCP server
+with a Responses-compatible model host.
 """
 
 import asyncio
@@ -21,7 +22,7 @@
 load_dotenv(override=True)
 
 # Configure model
-API_HOST = os.getenv("API_HOST", "github")
+API_HOST = os.getenv("API_HOST", "azure")
 if API_HOST == "azure":
     token_provider = azure.identity.get_bearer_token_provider(
         azure.identity.DefaultAzureCredential(),
@@ -31,19 +32,23 @@
         model=os.environ.get("AZURE_OPENAI_CHAT_DEPLOYMENT"),
         base_url=os.environ["AZURE_OPENAI_ENDPOINT"] + "/openai/v1/",
         api_key=token_provider,
-    )
-elif API_HOST == "github":
-    model = ChatOpenAI(
-        model=os.getenv("GITHUB_MODEL", "gpt-4o"),
-        base_url="https://models.inference.ai.azure.com",
-        api_key=SecretStr(os.environ["GITHUB_TOKEN"]),
+        use_responses_api=True,
     )
 elif API_HOST == "ollama":
     model = ChatOpenAI(
-        model=os.environ.get("OLLAMA_MODEL", "llama3.1"),
+        model=os.environ.get("OLLAMA_MODEL", "qwen3.5:9b"),
         base_url=os.environ.get("OLLAMA_ENDPOINT", "http://localhost:11434/v1"),
-        api_key=SecretStr("none"),
+        api_key=SecretStr(os.getenv("OLLAMA_API_KEY", "no-key-needed")),
+        use_responses_api=True,
+    )
+elif API_HOST == "openai":
+    model = ChatOpenAI(
+        model=os.getenv("OPENAI_MODEL", "gpt-5.2"),
+        api_key=SecretStr(os.environ["OPENAI_API_KEY"]),
+        use_responses_api=True,
     )
+else:
+    raise ValueError(f"Unsupported API_HOST '{API_HOST}'. Use one of: azure, ollama, openai.")
 
 console = Console()
 
@@ -90,12 +95,12 @@ async def main():
         prompt="You help users research GitHub repositories. Search and analyze information.",
     )
 
-    query = "Find 5 popular Python MCP server repositories and describe in a bulleted list."
+    query = "Make a list of last 5 issues from the 'PrefectHQ/FastMCP' repository that discuss auth."
     rprint(f"[bold]Query:[/bold] {query}\n")
 
     try:
         result = await agent.ainvoke({"messages": [HumanMessage(content=query)]})
-        rprint(f"[bold green]Result:[/bold green]\n{result['messages'][-1].content}\n")
+        rprint(f"[bold green]Result:[/bold green]\n{result['messages'][-1].text}\n")
     except Exception as e:
         rprint(f"[bold red]Error:[/bold red] {str(e)}\n")
 

agents/langchainv1_http.py

@@ -24,7 +24,7 @@
 MCP_SERVER_URL = os.getenv("MCP_SERVER_URL", "http://localhost:8000/mcp/")
 
 # Configure language model based on API_HOST
-API_HOST = os.getenv("API_HOST", "github")
+API_HOST = os.getenv("API_HOST", "azure")
 
 if API_HOST == "azure":
     token_provider = azure.identity.get_bearer_token_provider(
@@ -34,21 +34,23 @@
         model=os.environ.get("AZURE_OPENAI_CHAT_DEPLOYMENT"),
         base_url=os.environ["AZURE_OPENAI_ENDPOINT"] + "/openai/v1/",
         api_key=token_provider,
-    )
-elif API_HOST == "github":
-    base_model = ChatOpenAI(
-        model=os.getenv("GITHUB_MODEL", "gpt-4o"),
-        base_url="https://models.inference.ai.azure.com",
-        api_key=SecretStr(os.environ["GITHUB_TOKEN"]),
+        use_responses_api=True,
     )
 elif API_HOST == "ollama":
     base_model = ChatOpenAI(
-        model=os.environ.get("OLLAMA_MODEL", "llama3.1"),
+        model=os.environ.get("OLLAMA_MODEL", "qwen3.5:9b"),
         base_url=os.environ.get("OLLAMA_ENDPOINT", "http://localhost:11434/v1"),
-        api_key=SecretStr(os.environ["OLLAMA_API_KEY"]),
+        api_key=SecretStr(os.getenv("OLLAMA_API_KEY", "no-key-needed")),
+        use_responses_api=True,
+    )
+elif API_HOST == "openai":
+    base_model = ChatOpenAI(
+        model=os.getenv("OPENAI_MODEL", "gpt-5.2"),
+        api_key=SecretStr(os.environ["OPENAI_API_KEY"]),
+        use_responses_api=True,
     )
 else:
-    base_model = ChatOpenAI(model=os.getenv("OPENAI_MODEL", "gpt-4o-mini"))
+    raise ValueError(f"Unsupported API_HOST '{API_HOST}'. Use one of: azure, ollama, openai.")
 
 
 async def run_agent() -> None:
@@ -79,7 +81,7 @@ async def run_agent() -> None:
     )
 
     # Display result
-    final_response = response["messages"][-1].content
+    final_response = response["messages"][-1].text
     print(final_response)
 
 

infra/main.bicep

@@ -112,8 +112,8 @@ resource resourceGroup 'Microsoft.Resources/resourceGroups@2021-04-01' = {
 
 var prefix = '${name}-${resourceToken}'
 
-var openAiDeploymentName = 'gpt-4o-mini'
-var openAiModelName = 'gpt-4o-mini'
+var openAiDeploymentName = 'gpt-5.2'
+var openAiModelName = 'gpt-5.2'
 
 // Cosmos DB configuration
 var cosmosDbDatabaseName = 'expenses-database'
@@ -150,7 +150,7 @@ module openAi 'br/public:avm/res/cognitive-services/account:0.7.2' = {
         model: {
           format: 'OpenAI'
           name: openAiModelName
-          version: '2024-07-18'
+          version: '2025-12-11'
         }
         sku: {
           name: 'GlobalStandard'

pyproject.toml

@@ -16,7 +16,7 @@ dependencies = [
     "langchain-openai>=1.0.1",
     "langchain-mcp-adapters>=0.1.11",
     "azure-ai-agents>=1.1.0",
-    "agent-framework>=1.0.0rc5",
+    "agent-framework-core==1.0.0rc5",
     "azure-cosmos>=4.9.0",
     "azure-monitor-opentelemetry>=1.8.3",
     "opentelemetry-instrumentation-starlette>=0.60b0",

spanish/README.md

@@ -67,9 +67,8 @@ Si no usas una de las opciones anteriores, necesitas:
    ```
 
 6. Edita `.env` con tus credenciales de API. Elige uno de los siguientes proveedores definiendo `API_HOST`:
-   - `github` - GitHub Models (requiere `GITHUB_TOKEN`)
    - `azure` - Azure OpenAI (requiere credenciales de Azure)
-   - `ollama` - Instancia local de Ollama
+   - `ollama` - Instancia local de Ollama con soporte para Responses API
    - `openai` - OpenAI API (requiere `OPENAI_API_KEY`)
 
 ## Correr servidores MCP locales
@@ -279,7 +278,7 @@ Los precios varían por región y uso, así que no es posible predecir costos ex
 
 Puedes usar la [calculadora de precios de Azure](https://azure.com/e/3987c81282c84410b491d28094030c9a) para estos recursos:
 
-- **Azure OpenAI Service**: nivel S0, modelo GPT-4o-mini. El precio se basa en tokens. [Precios](https://azure.microsoft.com/pricing/details/cognitive-services/openai-service/)
+- **Azure OpenAI Service**: nivel S0, modelo GPT-5.2. El precio se basa en tokens. [Precios](https://azure.microsoft.com/pricing/details/cognitive-services/openai-service/)
 - **Azure Container Apps**: nivel de consumo. [Precios](https://azure.microsoft.com/pricing/details/container-apps/)
 - **Azure Container Registry**: nivel estándar. [Precios](https://azure.microsoft.com/pricing/details/container-registry/)
 - **Azure Cosmos DB**: nivel serverless. [Precios](https://azure.microsoft.com/pricing/details/cosmos-db/)