feat: Macae rfp agent framework #571
Draft
+5,142
−4,060
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Replaced semantic-kernel imports and dependencies with agent-framework in config, models, and utils. Updated pyproject.toml to add agent-framework and remove semantic-kernel. This refactor aligns the codebase with the new agent-framework library for agent and kernel functionality.
Migrate from semantic-kernel to agent-framework
fix: Dev v3 to main
Adds a comprehensive AVM post-deployment guide and refactors post-deployment scripts (PowerShell and Bash) to support both azd and ARM/Bicep deployments, including improved Azure subscription selection and output parsing. Updates FastAPI app to use a lifespan context for agent cleanup on shutdown. Refactors CosmosDB and config code for clarity and style. Updates workflow files for more flexible test and deployment handling. Introduces new frontend components and hooks for plan cancellation dialog and alert, and updates related styles and service logic.
Added agent-framework-core, agent-framework-azure-ai, agent-framework-copilotstudio, agent-framework-devui, agent-framework-mem0, and agent-framework-redis as dependencies from the Microsoft agent-framework repository. Updated azure-ai-agents to version 1.2.0b5 and upgraded aiohttp to 3.13.1.
Macae rfp af 101725
Replaced references to agent_framework with semantic_kernel in app_config.py and utils_kernel.py. Updated models in messages_kernel.py to use KernelBaseModel and Field from semantic_kernel.kernel_pydantic for consistency with the new package structure.
Macae rfp af 101725
Deleted the websocket_streaming.py module and removed the create_kernel method from AppConfig, as the system now uses Azure AI Agent Project pattern. Added placeholder files for messages_af.py and utils_af.py to support future Azure Foundry integration.
fix: dev-v3 to main branch merge
Introduces new Pydantic-based data models in messages_af.py to replace KernelBaseModel from semantic_kernel. All original model names and structures are preserved to support incremental migration from the previous framework.
Deleted src/backend/common/utils/check_deployments.py as it is no longer needed. This script was used to check Azure AI Foundry model deployments and their statuses.
Introduces the main FastAPI router for agent framework v3, including endpoints for team management, plan creation, approval, agent messaging, and configuration upload. Adds callback handlers for agent responses and streaming, global debug access, and service skeletons for agents, base API, foundry, MCP, and team management. These files establish the backend structure for multi-agent orchestration and extensible service integration.
Introduces common lifecycle management for agents, new agent configuration models, and refactors ProxyAgent and ReasoningAgentTemplate to use agent_framework primitives. Removes Semantic Kernel dependencies, adds Azure AI Search integration, and streamlines agent creation and invocation logic for improved maintainability and extensibility.
Updated import paths from 'v3' to 'af' for consistency and modularity. Improved error handling, logging, and tool/resource collection logic in FoundryAgentTemplate. Cleaned up unused Bing references and clarified code interpreter and MCP tool handling. These changes enhance maintainability and agent lifecycle management.
Changed imports from 'agent_framework.azure' to 'agent_framework_azure_ai' for AzureAIAgentClient in foundry_agent.py and reasoning_agent.py. Also removed an unused import in lifecycle.py to clean up dependencies.
Migrates orchestration logic and related services from semantic-kernel (v3) to agent_framework, updating imports, manager classes, and workflow construction. Updates HumanApprovalMagenticManager and OrchestrationManager to use agent_framework APIs, adapts callback handling, and ensures compatibility with new message and agent structures. Cleans up legacy code and improves maintainability for future agent_framework-based enhancements.
fix: Macae rfp af 101725
Deleted the entire v3 API router, callbacks, and common service modules including agent, base API, foundry, and MCP services, as well as related configuration, models, and orchestration files. This change removes the v3 backend implementation, likely in preparation for a major refactor or migration.
Renamed src/backend/app_kernel.py to src/backend/app.py. Updated import paths to reference 'af' modules instead of 'v3' and 'messages_af' instead of 'messages_kernel' to reflect new module structure.
Updated Dockerfiles, app.py, and tests to reference 'app' instead of 'app_kernel'. This change standardizes the application entry point and module imports, improving consistency across the backend codebase.
Macae rfp af 101725
Replaced references to deprecated Semantic Kernel and v3 modules with agent_framework (af) equivalents throughout backend, database, and test files. Removed obsolete messages_kernel and utils_kernel files, updated import paths, and fixed minor typos. Updated documentation and README to reflect new entry points and usage.
Replaced Semantic Kernel dependencies and terminology with agent_framework equivalents across backend modules, including response handlers, configuration, agent factories, orchestration, and utility functions. Updated imports, class names, comments, and logic to reflect agent_framework usage. Adjusted test files and MCP authentication to use agent_framework constructs. This refactor enables the system to operate with agent_framework as the core agent orchestration and messaging library.
Macae rfp af 101725
Deleted src/mcp_server/auth.py, which contained MCP authentication and HostedMCPTool setup logic for the employee onboarding system. This change may indicate a refactor, deprecation, or migration of authentication and plugin management functionality.
Reformatted code in human_approval_manager.py and orchestration_manager.py for better readability, including consistent indentation, line breaks, and argument formatting. No functional changes were made; this refactor aims to enhance maintainability and clarity.
Macae rfp af 101725
Moved MagenticAgentFactory import to module level in orchestration_manager.py to avoid circular imports. Cleaned up unused agent instance cache and improved RAI verdict handling in utils_af.py by checking for 'TRUE' substring and simplifying response logic.
Replaces AIProjectClient with AzureAIAgentClient in agent lifecycle, updates FoundryAgentTemplate to use ChatAgent, and improves compatibility checks and error handling. Also updates RAI agent creation to use config-based deployment name and corrects logic for verdict handling.
Added try/except blocks for chat client and manager instantiation in OrchestrationManager to log and raise errors. Also fixed import formatting and parameter names for AzureOpenAIChatClient. Minor formatting update in pyproject.toml dependencies.
The assignment of 'current_user_id' to kwargs has been commented out, possibly to prevent its propagation downstream. The user ID is still stored in the instance variable 'self.current_user_id'.
Macae rfp af 101725
Replaces Azure AI Search and code interpreter tool logic with HostedFileSearchTool and HostedCodeInterpreterTool. Updates lifecycle to use async MCP tool preparation and simplifies tool/resource collection. Removes legacy compatibility and diagnostics code, streamlining agent initialization and invocation for improved maintainability.
Refactored ReasoningAgentTemplate to use direct AzureAIAgentClient and integrate Azure AI Search via a new ReasoningSearch helper. Improved initialization, search augmentation, and resource management. Updated reasoning_search.py for better logging, error handling, and framework-agnostic usage. Added detailed docstrings and clarified configuration and usage patterns.
Replaces AzureOpenAIChatClient with AzureAIAgentClient for orchestration, updates callback signatures to match new agent framework, and enhances logging throughout the orchestration process. Improves error handling, updates workflow execution to use run_stream, and ensures compatibility with legacy WebSocket updates. Cleans up legacy shims and streamlines agent/participant management.
Macae rfp af 101725
Replaces legacy callback attachment with unified event-based handling for Magentic workflow events. Removes internal callback adapter methods and updates workflow construction to use .on_event() for streaming and final agent responses, improving maintainability and aligning with agent_framework best practices.
Refactor orchestration to use event-based callbacks
Eliminated all '# noqa: BLE001' comments from exception handling blocks across multiple modules. This change improves code cleanliness and removes unnecessary linter suppression comments.
Replaces direct client assignment with instantiation of a ChatAgent, passing relevant parameters such as instructions, name, description, tools, and model configuration. This change aligns ReasoningAgentTemplate with the foundry_agent pattern and prepares for enhanced tool and chat handling.
ProxyAgent now extends agent_framework's BaseAgent and implements the AgentProtocol with run and run_stream methods, providing improved compliance and streaming support. The orchestration_manager now extracts the inner ChatAgent from wrapper templates, ensuring correct agent registration for both wrapped and direct BaseAgent instances.
Renamed backend/af to backend/v4, updating all references from v3 to v4 across backend, infra, and documentation. Updated GitHub Actions workflows to use v4 versions and latest action versions. Changed default image tags and parameters from latest_v3 to latest_v4 in Bicep and parameter files. Updated documentation and scripts to reference v4 paths. Applied minor formatting and consistency improvements in infrastructure code.
Macae rfp af 101725
Enhanced ProxyAgent to provide more informative logging, generate consistent response and message IDs, and yield explicit text and usage updates for clarification requests. Updated message extraction to use the .text property of ChatMessage for accurate content retrieval. Also fixed agent_response_callback to properly extract text from ChatMessage objects.
Macae rfp af 101725
Bump agent-framework packages to version 1.0.0b251028 and update aiohttp from 3.13.1 to 3.13.2 in the lock file.
Refactored the workflow event loop to handle additional event types, including orchestrator messages, agent streaming updates, agent final messages, and final result events. Improved error handling and logging for each event type. Commented out unused MagenticCallbackMode references.
Added print statements for better visibility of agent creation failures in magentic_agent_factory and orchestration_manager. Enhanced error handling in orchestration_manager to log and raise exceptions during agent creation and orchestration initialization. Updated human_approval_manager to ensure correct progress ledger handling when max rounds are exceeded, and removed unused imports. Temporarily disabled file search tool addition in foundry_agent for debugging.
Macae rfp af 101725
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Purpose
Does this introduce a breaking change?
How to Test
What to Check
Verify that the following are valid
Other Information