Skip to content

ARCHITECTURE.md

Latest commit

 

History

History
101 lines (76 loc) · 5.15 KB

ARCHITECTURE.md

File metadata and controls

101 lines (76 loc) · 5.15 KB

System Architecture

System Architecture Diagram

This document outlines the architecture for the Microsoft AI Agentic Workshop platform. The architecture is modular and designed to support a wide variety of agent design patterns, allowing you to focus on agent implementation and experimentation without changing the core infrastructure.

High-Level Overview

The system comprises four primary layers:

  1. Front End: User-facing chat interface.
  2. Backend: Orchestrates conversation flow, session state, and acts as a bridge between the front end and agent logic.
  3. Agent Service Layer: Loads, instantiates, and operates agent implementations (e.g., single-agent, multi-agent, multi-domain patterns).
  4. Model Context Protocol (MCP) API Server: Exposes structured business operations and tools via API endpoints for agent use.

Supporting databases include:

  • SQL Database: Contains core business/transactional data (customers, subscriptions, invoices, etc.).
  • Vector Database: Enables semantic (embedding-based) retrieval over internal knowledge/documents.

Component Breakdown

1. Frontend

  • Technology: Streamlit (Python)
  • Functionality:
    • Presents an interactive chat interface.
    • Maintains a unique, persistent session for each user.
    • Displays real-time chat history.
    • Communicates with the backend using HTTP endpoints for prompt submission, response retrieval, and session management (reset/new conversations).

2. Backend

  • Technology: FastAPI (asynchronous Python)
  • Responsibilities:
    • Exposes HTTP API endpoints for frontend communication.
    • Manages sessions and conversation history in memory.
    • Connects inbound user requests with the appropriate agent instance in the Agent Service layer.
    • Forwards tool/API calls requested by agents to the MCP API server.
  • Endpoints:
    • /chat: Processes chat requests and returns agent responses.
    • /reset_session: Clears session memory and context state.
    • /history/{session_id}: Fetches conversation history.
  • Session Store: Tracks session state and chat history (in-memory; pluggable for persistent storage).

3. Agent Service Layer

  • Design: Pluggable and modular—enables selecting and loading different agent design patterns:
    • Intelligent Single Agent: One agent using tools to handle requests end-to-end.
    • Multi-Domain Agent: Coordinator agent routes requests to specialist agents (e.g., CRM, Billing, Security).
    • Collaborative Multi-Agent System: Planning/analysis agent orchestrates multiple domain experts.
  • Capabilities:
    • Tool usage via structured API calls to MCP endpoints.
    • Augmented responses using Retrieval-Augmented Generation (RAG) with the vector knowledge base.
    • Maintains both short-term (session) and, optionally, shared or long-term memory for multi-step reasoning.
  • Implementation:
    • Built using Semantic Kernel, autogen, or Azure Agent-Service frameworks for agent planning, orchestration, and tool connectivity.
    • Easily configured and swapped by changing agent module imports and startup parameters.

4. Model Context Protocol (MCP) API Server

  • Technology: FastAPI/asyncio, JSON schema validation with Pydantic.
  • Purpose: Simulates realistic enterprise APIs, exposing business and operational data through agent-friendly, structured APIs.
  • Key Endpoint Categories:
    • Customer/account lookup
    • Subscription, invoice, and payment management
    • Data usage reporting
    • Product and promotion queries
    • Support ticket management
    • Security log review and account unlock
    • Semantic search over the knowledge base
  • Endpoint Examples:
    • get_all_customers, get_customer_detail
    • get_subscription_detail, get_invoice_payments, pay_invoice
    • get_data_usage, get_promotions, get_eligible_promotions
    • search_knowledge_base
    • get_security_logs, unlock_account
    • get_customer_orders, get_support_tickets, create_support_ticket
    • get_products, get_product_detail, update_subscription
    • get_billing_summary
  • Characteristics:
    • Pydantic-model-validated requests and responses for reliable tool invocation.
    • Serves as the central integration point for all structured data and knowledge base access.

Databases

  • SQL Database: Stores structured business data, such as customer profiles, subscriptions, invoices, and order histories.
  • Vector Database: Stores embedding-based representations of knowledge documents and internal policies, enabling fast and relevant retrieval during RAG operations by agents.

Summary

This four-layer architecture ensures clear separation of concerns across user interaction, backend orchestration, agent intelligence, and enterprise data access. The platform makes it easy to experiment with a wide range of agent frameworks and design patterns, supporting robust, enterprise-grade agentic solutions without the need to modify underlying infrastructure.