python-v0.4.6
📦 autogenView on GitHub →
✨ 8 features🐛 9 fixes🔧 9 symbols
Summary
This release introduces built-in support for MCP and HTTP tools, enhances MagenticOne and SelectorGroupChat for smaller/text-only models, and simplifies Gemini model integration.
Migration Steps
- To use MCP tools, install dependencies: pip install mcp-server-fetch autogen-ext[mcp].
- To use HTTP tools, install dependencies: pip install autogen-ext[http-tool].
- Update Gemini client implementations to simplify initialization by removing redundant model_info and base_url parameters.
✨ New Features
- Added Model Context Protocol (MCP) tool support for accessing external tool ecosystems.
- Introduced HttpTool for querying remote HTTP-based API endpoints via declarative schemas.
- Added support for text-only models in MagenticOne (M1) agents.
- Added YAML configuration file support for the m1 CLI tool.
- Enhanced SelectorGroupChat to support smaller models (e.g., Llama 13B) and models without 'name' field support.
- Improved Gemini model support in OpenAIChatCompletionClient, removing the requirement for model_info and base_url.
- Added support for o3-mini model family.
- Added a new FastAPI integration sample for chat applications and human-in-the-loop workflows.
🐛 Bug Fixes
- Fixed reading string arguments from the m1 CLI.
- Fixed summarize_page functionality for text-only contexts and unknown models.
- Changed empty chunk handling from erroring out to issuing a warning.
- Added state management for OpenAI assistants.
- Fixed streaming token mode issues during function calls.
- Excluded agent events from MaxMessageTermination condition counts.
- Removed Semantic Kernel (SK) tool adapter plugin name and updated stream tool call processing.
- Updated SK kernel from tool to use method.
- Fixed selector prompt documentation and removed unnecessary validation checks.
🔧 Affected Symbols
autogen_ext.tools.mcp.mcp_server_toolsautogen_ext.tools.mcp.StdioServerParamsautogen_ext.tools.http.HttpToolautogen_agentchat.agents.AssistantAgentautogen_agentchat.group_chat.SelectorGroupChatautogen_ext.models.openai.OpenAIChatCompletionClientautogen_core.models.UserMessageautogen_agentchat.conditions.MaxMessageTerminationWebSurfer