Skip to content

TechnicalArchitecture.md

Latest commit

 

History

History
148 lines (97 loc) · 4.85 KB

TechnicalArchitecture.md

File metadata and controls

148 lines (97 loc) · 4.85 KB

Technical Architecture

This document provides a comprehensive technical overview of the Container Migration Solution Accelerator architecture, including system components, data flows, and integration patterns.

Overview

The Container Migration Solution Accelerator is built on a modern, cloud-native, queue-driven architecture that leverages artificial intelligence, multi-agent orchestration, and the Model Context Protocol (MCP) to automate container platform migrations to Azure.

High-Level Architecture

graph TB
    UI[Frontend UI] --> API[Backend API]
    API --> Q[Azure Storage Queue]

    subgraph Processor[Processor (Queue Worker)]
        QW[Queue Worker]
        WF[Agent Framework Workflow\nanalysis → design → yaml → docs]
        CA[Control API]
        PC[Process Control Store]
        QW --> WF
        CA --> PC
    end

    Q --> QW

    subgraph Tools[Tools (MCP + local tools)]
        Blob[Blob IO]
        Docs[Microsoft Docs]
        Mermaid[Mermaid]
        Datetime[Datetime]
        YamlInv[YAML Inventory]
    end

    subgraph External[External Services]
        ST[Azure Blob Storage]
        Models[Azure OpenAI / Azure AI Foundry Models]
    end

    WF --> Tools
    Blob --> ST
    Docs --> Models
Loading

Core Components (Processor)

1. Queue Worker

The processor runs as a queue-driven worker in hosted scenarios.

Responsibilities:

  • Poll Azure Storage Queue for jobs
  • Validate/deserialize request payloads
  • Execute the Agent Framework workflow
  • Persist artifacts and emit telemetry

Implementation Locations:

Operational Notes:

  • The queue worker is intentionally simple; behavior such as retries and DLQ are controlled by the queue configuration/patterns and service logic.

2. Control API + Process Control

The processor exposes a lightweight control surface for health and termination.

Implementation Locations:

3. Workflow Engine (Microsoft Agent Framework)

The migration pipeline is defined as an Agent Framework workflow built via WorkflowBuilder and executed step-by-step.

Execution Order:

  • analysis → design → yaml → documentation

Implementation Location:

4. Multi-Agent Orchestration

Steps that require multi-agent reasoning use a group chat style orchestrator.

Key Concepts:

  • Coordinator agent manages turn-taking and termination
  • Platform experts contribute source-platform-specific guidance
  • Result generator consolidates structured outputs

Implementation Locations:

5. MCP Tool Integration (Agent Framework Tools)

Tools are exposed to agents using Agent Framework tool abstractions, including MCP.

Processor MCP tools:

Technology Stack (Processor)

Core

  • Microsoft Agent Framework (workflow + orchestration)
  • Python 3.12+
  • asyncio
  • aiohttp (control API)

AI

  • Azure OpenAI / Azure AI Foundry models (project-dependent)
  • Model Context Protocol (MCP) for tool access

Azure SDKs

  • Azure Storage Queue (job intake)
  • Azure Storage Blob (artifact IO)
  • Azure Identity (auth)

Dev/Ops

  • uv (dependency management)
  • Docker (containerized execution)
  • pytest (tests)

For additional technical details, refer to: