Microsoft Agent Framework Lab

1.0.0b251024 · active · verified Thu Apr 16

agent-framework-lab provides experimental modules as part of the Microsoft Agent Framework, focusing on cutting-edge features, benchmarking, reinforcement learning, and research initiatives. It is currently in a beta development stage (1.0.0b251024) and is designed for developers building, learning to build, or exploring AI agents using modern frameworks and tooling.

Common errors

```
ModuleNotFoundError: No module named 'agent_framework.azure'
```
cause The Azure-specific implementations, especially for `AzureAIProjectAgentProvider`, have been moved or refactored in version 1.0.0 and are no longer directly under `agent_framework.azure`.

fix For Azure AI Foundry, connect directly to a Foundry-hosted agent using `FoundryAgent`. For other Azure-related components, refer to the updated documentation for new import paths and classes in `agent-framework-foundry` or other specific Azure packages.
```
TypeError: Message constructor got an unexpected keyword argument 'text'
```
cause The `text` parameter in the `Message` constructor was deprecated and then removed in later versions (specifically, around the 1.0.0 release).

fix Update your code to pass the message content as the positional `content` argument, which expects a list of content items (e.g., `Message('user', [TextContent('Your message')])` or simply `Message('user', 'Your message')` if using string for simple text content).
```
ImportError: cannot import name 'BaseContextProvider' from 'agent_framework.core'
```
cause Deprecated aliases like `BaseContextProvider` and `BaseHistoryProvider` were removed from `agent_framework.core` in version 1.0.0.

fix Refer to the updated `agent-framework-core` documentation for the correct, non-aliased class names or replacement patterns for context and history providers.

Warnings

breaking The 1.0.0 release of Microsoft Agent Framework introduced significant architectural changes, particularly around provider-leading client design, the extraction of OpenAI code from core, and standardized naming. Previous beta versions are not directly compatible.
Fix: Consult the official migration guides (e.g., from Semantic Kernel or AutoGen) and update code to use the new API patterns, especially for `agent-framework-core`, `agent-framework-openai`, and `agent-framework-foundry`.
gotcha `agent-framework-lab` contains experimental modules. Features and APIs within this sub-package are subject to rapid change, may be unstable, and are not guaranteed for long-term support.
Fix: Use these modules with caution in production environments. Pin exact versions of `agent-framework-lab` and be prepared for API changes in future updates. Monitor the GitHub repository for updates to the 'Labs directory'.
gotcha The Agent Framework does not automatically load environment variables from `.env` files. API keys and other configurations must be explicitly loaded or set in the environment.
Fix: Manually load `.env` files at the start of your application using `dotenv.load_dotenv()` or set environment variables directly in your shell/IDE before running your application.
breaking Prior to 1.0.0, there were deprecated kwargs compatibility paths, core dependencies were heavier, and OpenAI provider implementation was bundled. These have been removed or refactored.
Fix: Review the `agent-framework` changelog and upgrade guides. Ensure `agent-framework-core` and provider packages like `agent-framework-openai` are updated to 1.0.0 or later and adapt your code to the new, leaner core and separate provider packages.

Install

pip install agent-framework-lab Install `agent-framework-lab`
pip install agent-framework Install the complete Agent Framework (includes lab modules)

Imports

Agent
```
from agent_framework.core import Agent
```
Common core agent class
Message
```
from agent_framework import Message
```
For creating messages in conversations
OpenAIChatClient
```
from agent_framework.openai import OpenAIChatClient
```
Example of an LLM client, requires `agent-framework-openai` package
some_lab_module
```
from agent_framework.lab import some_lab_module
```
Example import for an experimental module from the `lab` sub-package. Specific modules will vary.

Quickstart

This quickstart demonstrates creating a basic agent using the core Microsoft Agent Framework and an OpenAI chat client. Modules from `agent-framework-lab` would typically extend or be integrated into such agents and workflows, for example, by providing specialized tools, evaluation components, or experimental agent behaviors.

import os
from agent_framework import Message
from agent_framework.openai import OpenAIChatClient
from agent_framework.core import Agent

# Ensure your OpenAI API key is set as an environment variable
# Or, if using Azure OpenAI, set AZURE_OPENAI_ENDPOINT, AZURE_OPENAI_API_KEY, etc.
# Agent Framework does not automatically load .env files; use `load_dotenv()` if needed.
OPENAI_API_KEY = os.environ.get('OPENAI_API_KEY', '')

if not OPENAI_API_KEY:
    print("Warning: OPENAI_API_KEY environment variable not set. This example may fail.")

async def main():
    client = OpenAIChatClient(api_key=OPENAI_API_KEY)
    
    # Create a simple agent
    agent = Agent(
        name="MyFirstAgent",
        instructions="You are a helpful assistant.",
        client=client
    )
    
    # Send a message to the agent
    response = await agent.run(
        messages=[Message("user", "Hello, how are you?")]
    )
    
    print(f"Agent: {response.content[0].text}")

if __name__ == "__main__":
    import asyncio
    asyncio.run(main())

view raw JSON →