Migrate Agent Framework demos to Azure AI Foundry SDK

demos/03-agent-framework/02-basics/agentfw_basics-py-foundry/.env.copy

@@ -0,0 +1,18 @@
+# Azure AI Foundry project
+PROJECT_ENDPOINT=REPLACE_WITH_YOUR_VALUE
+MODEL_DEPLOYMENT=gpt-4o-mini
+
+# Azure OpenAI (Direct OpenAI demos)
+AZURE_OPENAI_ENDPOINT=REPLACE_WITH_YOUR_VALUE
+AZURE_OPENAI_CHAT_DEPLOYMENT_NAME=gpt-4o
+AZURE_OPENAI_API_KEY=REPLACE_WITH_YOUR_VALUE
+# If not set, scripts default to 2024-07-01-preview
+AZURE_OPENAI_API_VERSION=2025-01-01-preview
+
+# Input & Output files
+DATA_PATH=./data
+OUTPUT_PATH=./output   
+
+# Optional demo specific settings
+AZURE_AI_AGENT_ID=REPLACE_WITH_YOUR_VALUE # Used in use_existing_agent.py
+VECTOR_STORE_ID=REPLACE_WITH_YOUR_VALUE

demos/03-agent-framework/02-basics/agentfw_basics-py-foundry/.gitignore

@@ -0,0 +1,8 @@
+.env
+*.pyc
+__pycache__/
+.DS_Store
+.vscode/
+output/
+thread_history.json
+*.log

demos/03-agent-framework/02-basics/agentfw_basics-py-foundry/MIGRATION.md

@@ -0,0 +1,244 @@
+# Migration Summary: Agent Framework to Azure AI Foundry SDK
+
+## Overview
+This document summarizes the migration of Agent Framework demos to Azure AI Foundry SDK.
+
+**Source:** `demos/03-agent-framework/02-basics/agentfw_basics-py`  
+**Target:** `demos/03-agent-framework/02-basics/agentfw_basics-py-foundry`
+
+## Migration Date
+November 24, 2025
+
+## Files Migrated
+
+| Original File | Migrated File | Status | Notes |
+|--------------|---------------|--------|-------|
+| agentfw_openai_chat.py | openai_chat.py | ✅ Complete | Uses azure-ai-inference ChatCompletionsClient |
+| agentfw_create_agent.py | create_agent.py | ✅ Complete | Uses AgentsClient.create_agent() |
+| agentfw_streaming.py | streaming.py | ✅ Complete | Event-based streaming with create_run_stream() |
+| agentfw_use_existing_agent.py | use_existing_agent.py | ✅ Complete | Uses AgentsClient.get_agent() |
+| agentfw_threading.py | threading.py | ✅ Complete | Thread ID-based persistence |
+| agentfw_chat_history.py | chat_history.py | ✅ Complete | Uses list_messages() for history |
+| agentfw_long_term_memory.py | long_term_memory.py | ✅ Complete | Custom memory with file storage |
+| agentfw_middleware.py | function_calling.py | ✅ Complete | Tool definitions with timing/logging |
+| agentfw_observability.py | observability.py | ✅ Complete | OpenTelemetry with custom spans |
+| agentfw_structured_output.py | structured_output.py | ✅ Complete | JSON schema with response_format |
+| agentfw_multimodal.py | multimodal.py | ✅ Complete | File upload and attachments |
+| agents-delete-all.py | agents-delete-all.py | ✅ Complete | Uses AgentsClient.list_agents() |
+
+## Dependency Changes
+
+### Removed Dependencies
+- `agent-framework==1.0.0b251120`
+- `agent-framework-azure-ai==1.0.0b251120`
+
+### Added Dependencies
+- `azure-ai-projects>=2.0.0b2`
+- `azure-ai-agents==1.2.0b5`
+- `azure-ai-inference>=1.0.0b6`
+- `opentelemetry-api`
+- `opentelemetry-sdk`
+
+### Retained Dependencies
+- `azure-identity==1.25.1`
+- `python-dotenv==1.2.1`
+- `pdf2image==1.17.0`
+- `pillow==12.0.0`
+- `pymupdf==1.24.14`
+- `pydantic>=2.0.0`
+
+## Key Technical Changes
+
+### 1. Agent Creation
+**Before:**
+```python
+from agent_framework import ChatAgent
+from agent_framework.azure import AzureAIAgentClient
+
+agent = ChatAgent(
+    chat_client=AzureAIAgentClient(
+        project_client=project_client,
+        agent_id=agent_id,
+        async_credential=credential
+    )
+)
+```
+
+**After:**
+```python
+from azure.ai.agents.aio import AgentsClient
+
+async with AgentsClient(endpoint=endpoint, credential=credential) as agents_client:
+    agent = await agents_client.create_agent(
+        model=model,
+        name="My Agent",
+        instructions="You are helpful."
+    )
+```
+
+### 2. Streaming Responses
+**Before:**
+```python
+async for chunk in agent.run_stream(user_input):
+    if chunk.text:
+        print(chunk.text, end="")
+```
+
+**After:**
+```python
+async with await agents_client.create_run_stream(
+    thread_id=thread.id,
+    assistant_id=agent.id
+) as stream:
+    async for event in stream:
+        if event.event == "thread.message.delta":
+            if hasattr(event.data, 'delta'):
+                for content in event.data.delta.content:
+                    if hasattr(content, 'text'):
+                        print(content.text.value, end="")
+```
+
+### 3. Thread Management
+**Before:**
+```python
+thread = agent.get_new_thread()
+serialized = await thread.serialize()
+thread = await agent.deserialize_thread(thread_data)
+```
+
+**After:**
+```python
+thread = await agents_client.create_thread()
+# Save thread.id to file/database
+thread = await agents_client.get_thread(thread_id)
+```
+
+### 4. Function Calling
+**Before:**
+```python
+agent = client.create_agent(
+    tools=[get_weather, calculate],
+    middleware=[timing_middleware, function_logger_middleware]
+)
+```
+
+**After:**
+```python
+tools = [
+    {
+        "type": "function",
+        "function": {
+            "name": "get_weather",
+            "description": "Get weather",
+            "parameters": {...}
+        }
+    }
+]
+
+agent = await agents_client.create_agent(
+    model=model,
+    tools=tools
+)
+
+# Manual function execution
+if run.status == "requires_action":
+    tool_outputs = []
+    for tool_call in run.required_action.submit_tool_outputs.tool_calls:
+        result = functions[tool_call.function.name](**json.loads(tool_call.function.arguments))
+        tool_outputs.append({"tool_call_id": tool_call.id, "output": result})
+
+    run = await agents_client.submit_tool_outputs(
+        thread_id=thread.id,
+        run_id=run.id,
+        tool_outputs=tool_outputs
+    )
+```
+
+### 5. Chat Completions (Direct)
+**Before:**
+```python
+from agent_framework.azure import AzureOpenAIChatClient
+
+agent = AzureOpenAIChatClient(...).create_agent(...)
+response = await agent.run(message)
+```
+
+**After:**
+```python
+from azure.ai.inference import ChatCompletionsClient
+from azure.ai.inference.models import SystemMessage, UserMessage
+
+client = ChatCompletionsClient(endpoint=endpoint, credential=credential)
+response = client.complete(
+    model=deployment,
+    messages=[SystemMessage(...), UserMessage(...)],
+    stream=True
+)
+```
+
+## Features Not Directly Migrated
+
+### 1. Middleware System
+Agent Framework's built-in middleware decorators (`@agent_middleware`, `@function_middleware`, `@chat_middleware`) are replaced with:
+- Manual timing using `datetime`
+- Custom logging and function tracking
+- OpenTelemetry spans for observability
+
+### 2. Chat Reducers
+Agent Framework's `ChatReducer` classes for history management are replaced with:
+- Direct `list_messages()` API calls
+- Manual message count tracking
+- File-based persistence of thread IDs
+
+### 3. Thread Serialization
+Agent Framework's `thread.serialize()`/`deserialize_thread()` is replaced with:
+- Storing thread IDs in JSON files
+- Using `get_thread()` to retrieve existing threads
+- Manual state management
+
+## Benefits of Migration
+
+1. **Direct SDK Access**: Full control over Azure AI services without abstraction overhead
+2. **Official Support**: Uses Microsoft's official SDKs with comprehensive documentation
+3. **Transparency**: Clear understanding of API interactions
+4. **Flexibility**: Easy to customize and extend
+5. **Stability**: Relies on stable Azure SDK versions
+
+## Trade-offs
+
+1. **More Code**: Requires more boilerplate for common operations
+2. **Manual Management**: Thread and message management is explicit
+3. **Event Handling**: Streaming requires event-type checking
+4. **No Middleware**: Must implement custom patterns for cross-cutting concerns
+
+## Testing Status
+
+- ✅ Syntax validation completed
+- ✅ Code review completed
+- ✅ Security scan (CodeQL) completed - 0 vulnerabilities
+- ⚠️  Runtime testing requires Azure AI Foundry resources
+
+## Documentation
+
+- ✅ Comprehensive README.md created
+- ✅ Migration guide included
+- ✅ Code examples updated
+- ✅ Inline documentation added
+
+## Security Improvements
+
+1. Replaced `eval()` with AST-based safe expression evaluation
+2. Replaced `eval()` for JSON parsing with `json.loads()`
+3. All CodeQL security checks passed
+
+## Next Steps for Users
+
+1. Review the new `readme.md` in the migrated folder
+2. Update `.env` file with Azure credentials
+3. Install dependencies: `pip install -r requirements.txt`
+4. Run demos to validate functionality
+5. Adapt code patterns to your specific use cases
+
+## Conclusion
+
+The migration successfully converts all Agent Framework demos to use Azure AI Foundry SDK directly. The new implementation provides equivalent functionality with more direct control over Azure services, at the cost of additional boilerplate code. All demos maintain their core educational value while demonstrating modern Azure AI SDK usage patterns.

demos/03-agent-framework/02-basics/agentfw_basics-py-foundry/agents-delete-all.py

@@ -0,0 +1,54 @@
+"""
+Delete all agents in the Azure AI Foundry project.
+
+This utility script deletes all agents to clean up resources.
+"""
+
+import asyncio
+from azure.identity.aio import DefaultAzureCredential
+from azure.ai.agents.aio import AgentsClient
+from dotenv import load_dotenv
+import os
+
+load_dotenv()
+
+PROJECT_ENDPOINT = os.getenv("PROJECT_ENDPOINT")
+
+
+async def main():
+    print("\n" + "="*70)
+    print("DELETE ALL AGENTS")
+    print("="*70)
+
+    async with DefaultAzureCredential() as credential:
+        async with AgentsClient(
+            endpoint=PROJECT_ENDPOINT,
+            credential=credential
+        ) as agents_client:
+
+            print("\nListing all agents...")
+            agents = await agents_client.list_agents()
+
+            if not agents.data:
+                print("No agents found.")
+                return
+
+            print(f"\nFound {len(agents.data)} agent(s):")
+            for agent in agents.data:
+                print(f"  - {agent.name} (ID: {agent.id})")
+
+            confirm = input("\nDelete all agents? (yes/no): ").strip().lower()
+
+            if confirm == 'yes':
+                print("\nDeleting agents...")
+                for agent in agents.data:
+                    await agents_client.delete_agent(agent.id)
+                    print(f"  ✓ Deleted: {agent.name}")
+
+                print("\nAll agents deleted successfully!")
+            else:
+                print("\nCancelled. No agents were deleted.")
+
+
+if __name__ == "__main__":
+    asyncio.run(main())