feat(ag2): add AG2 framework integration with Amazon Bedrock AgentCore

[faridun-ag2]

Summary

• Adds AG2 (formerly AutoGen) as a new agentic framework integration under 03-integrations/agentic-frameworks/ag2/
• AG2 has native Amazon Bedrock support (LLMConfig(api_type="bedrock")), no OpenAI API key required — AWS credentials from IAM role or environment are sufficient
• Two examples included: single-agent tool use and multi-agent GroupChat workflow

Changes

• ag2_agent_hello_world.py — single AssistantAgent backed by Claude 3 Sonnet on Bedrock, with a weather tool registered via AG2's @register_for_llm / @register_for_execution decorator pattern
• ag2_multi_agent_example.py — two-agent GroupChat workflow: researcher gathers information using a tool, analyst synthesizes findings into a structured summary
• requirements.txt — ag2[bedrock]>=0.11.0, bedrock-agentcore, bedrock-agentcore-starter-toolkit
• README.md — setup guide, code walkthrough, AG2 vs Microsoft AutoGen comparison table

Notes

AG2 (ag2 on PyPI) and Microsoft AutoGen (autogen-agentchat) are separate projects with incompatible APIs. This example uses AG2 imports (from autogen.agentchat import AssistantAgent), not Microsoft AutoGen imports (from autogen_agentchat.agents import AssistantAgent).

Test plan

• 24 unit/integration tests pass locally without AWS credentials: pytest test_ag2_agents.py -v
• ag2_agent_hello_world.py tested locally with real Bedrock credentials
• ag2_multi_agent_example.py tested locally with real Bedrock credentials
• agentcore invoke -l '{"prompt": "what is the weather in NYC?"}' returns {"result": "..."}

[faridun-ag2]

Hey @manuwaik !

We'd love to get your eyes on this PR when you have a moment.

This adds an AG2 integration example — AG2 (https://github.com/ag2ai/ag2) has native Amazon Bedrock support built in, which makes it a natural fit for AWS-native deployments. No OpenAI API key required — just IAM credentials.

Two examples included:

• ag2_agent_hello_world.py — single agent with tool use, mirrors the existing autogen/ example structure
• ag2_multi_agent_example.py — multi-agent GroupChat workflow (researcher + analyst), showcasing AG2's orchestration capabilities

Happy to address any feedback!