v0.4.4
📦 autogen
✨ 6 features🔧 12 symbols
Summary
This release introduces full serialization support for AgentChat configurations and states, enabling persistent and portable agent sessions. It also adds a new Azure AI inference client for broader model support and a rich UI for the Magentic One CLI.
Migration Steps
- To persist agent sessions, use group_chat.dump_component() to save configuration and group_chat.save_state() to save the current execution state.
- To restore sessions, use Team.load_component(config) followed by await group_chat.load_state(state).
- For Azure/GitHub models, switch to AzureAIChatCompletionClient and provide the appropriate endpoint and AzureKeyCredential.
✨ New Features
- Introduced declarative serializable configuration for AgentChat, allowing agents and teams to be converted to/from JSON.
- Added support for serializing and deserializing the state of agents and teams via save_state and load_state.
- Added AzureAIChatCompletionClient to support Azure and GitHub-hosted models like Phi-4, Mistral, and Cohere.
- Introduced RichConsole UI for Magentic One (m1) CLI for pretty-printed output using the --rich flag.
- Implemented a default in-memory store for ChatCompletionCache.
- Added description support in dump_component for improved component configuration.
🔧 Affected Symbols
autogen_agentchat.teams.RoundRobinGroupChatautogen_agentchat.base.Teamautogen_agentchat.agents.AssistantAgentautogen_ext.models.azure.AzureAIChatCompletionClientautogen_ext.models.openai.OpenAIChatCompletionClientautogen_core.models.ChatCompletionCacheautogen_agentchat.ui.Consoledump_componentload_componentsave_stateload_stateFunctionTools