How microsoft/openaiworkshop works
Project type and purpose
• This is a full-stack “Agentic AI Workshop” application that lets users build, compare, and run single- and multi-agent solutions using Microsoft Azure AI Agent Service, Semantic Kernel, or Autogen.
• The README and ARCHITECTURE.md emphasize a FastAPI backend, a Streamlit frontend, a pluggable agent framework, a small backend microservice (MCP), and a SQLite knowledge base.
File structure
• Top-level directories and files:
• agentic_ai/agents – all agent implementations (base, Azure Agent Service, Autogen, Semantic Kernel; single & multi-agent)
• agentic_ai/applications – the application entrypoints (frontend.py = Streamlit, backend.py = FastAPI, run scripts, .env sample, requirements)
• agentic_ai/backend_services – an MCP microservice (Python + Dockerfile + SQLite DB + data files + deployment script)
• Various markdowns (SETUP.md, ARCHITECTURE.md, SCENARIO.md, DATA.md) describing environment, architecture, scenario, and data.
• Configuration artifacts: .env, Dockerfile.mcp, deploy_azure.sh, run_application.sh.
README for additional insights
• Confirms the 3 agent platforms, the full-stack design, session-based chat, environment configuration, and the existence of RESTful endpoints (chat, reset, history).
Elements
• Clear labels on each component (e.g., “Streamlit Frontend,” “FastAPI Backend,” “Agent Manager,” “Autogen Agent,” “Azure AI Service,” “MCP Microservice,” “SQLite DB”)
– Agent implementation plugins (shaded group)
Final recommendations
• Be detailed: break the agent layer into its subcomponents (single vs. multi, each platform).