Compliance Compass — Build AI Agents with Microsoft Foundry, Foundry Toolkit for VS Code & GitHub Copilot
Audience: Developers, Compliance Engineers, Tech Leads, Architects Duration: ~90 minutes (7 modules)
Each participant needs the following accounts and tools on their machine:
| Requirement | Details |
|---|---|
| Azure Subscription | Active subscription with permission to create resources |
| GitHub Copilot | Individual, Business, or Enterprise plan |
| Visual Studio Code | Latest version — Download |
| Python | 3.10 or higher — Download |
| Azure CLI | 2.60+ and authenticated (az login) — Install |
| Git | Any recent version — Install |
| Docker Desktop | (Optional — Module VII only) — Install |
VS Code Extensions required: GitHub Copilot, GitHub Copilot Chat, Foundry Toolkit for VS Code (Microsoft), Python (Microsoft).
Full installation and verification steps are in Module I: Prerequisites and Environment Setup.
The following Azure resources must be provisioned inside each participant's subscription prior to the workshop. All tiers listed are free or low-cost for workshop use.
| Azure Resource | Purpose | Tier / SKU |
|---|---|---|
| Azure AI Foundry Hub + Project | Hosts GPT-4o (reasoning) and the embedding model; manages the agent | Standard — no per-hub charge |
| GPT-4o Deployment | Language model for compliance reasoning and report generation | Standard deployment (pay-per-use) |
| text-embedding-ada-002 Deployment | Embeds knowledge base documents for semantic search | Standard deployment (pay-per-use) |
| Azure AI Search | Indexes the 12 compliance documents; provides RAG retrieval | Free tier (up to 50 MB / 3 indexes) |
| Azure Storage Account + Blob Container | Stores the 12 Markdown knowledge base documents | LRS Standard (minimal cost; < 1 MB of documents) |
Tip
Module II walks participants through provisioning each of these resources step-by-step, with both portal and Azure CLI instructions. If resources are pre-provisioned by the workshop facilitator, participants can skip directly to the indexing step in Module II.
Compliance Compass — a Retrieval-Augmented Generation (RAG) agent that helps compliance and risk teams evaluate regulatory risks for vendors, data transfers, and cross-border operations.
The agent is powered by:
- Microsoft AI Foundry (GPT-4o + embedding model)
- Azure AI Search (knowledge retrieval from 12 compliance documents)
- Foundry Toolkit for VS Code (visual agent design)
- GitHub Copilot (code generation, debugging, UI creation, containerization)
Azure = Brain (Models + Search)
Foundry Toolkit for VS Code = Agent Designer
GitHub Copilot = Builder + Debugger + Extender
flowchart LR
subgraph F1["Phase 1: Foundation"]
I[" Module I\nPrerequisites"] --> II[" Module II\nAzure Resources"] --> III[" Module III\nAgent Design"] --> IV[" Module IV\nExport & Debug"]
end
subgraph F2["Phase 2: Extend & Deploy"]
V[" Module V\nInteractive Input"] --> VI[" Module VI\nWeb UI"] --> VII[" Module VII\nDocker & Testing\n(Optional)"]
end
IV --> V
style VII fill:#f5f5f5,stroke:#aaaaaa
| # | Module | Description | Duration |
|---|---|---|---|
| I | Prerequisites & Environment Setup | Install tools, configure Azure CLI, sign in to Copilot & Foundry Toolkit for VS Code | 10 min |
| II | Provisioning Azure Resources | Create AI Foundry hub, deploy models, set up Blob Storage & AI Search | 15 min |
| III | Designing the Agent in Foundry Toolkit for VS Code | Build the compliance agent visually with instructions & Azure AI Search tool | 10 min |
| IV | Exporting Code & Debugging with Copilot | Export to Python, encounter real Azure SDK error, debug with Copilot | 15 min |
| V | Adding Interactive Input | Transform hardcoded queries into interactive CLI with Copilot | 10 min |
| VI | Building a Professional Web UI | Generate a complete Streamlit chat UI with Copilot Agent mode | 15 min |
| VII | Containerization & Testing (Optional) | Dockerize the app (optional), run quality tests, refine agent instructions | 15 min |
| Feature | Where Used |
|---|---|
| Agent Mode | Modules IV–VII: Debugging, code generation, UI creation, Docker |
| Ask Mode | Module IV: Understanding generated code |
| Structured Prompting | Module IV: Error diagnosis with constraints |
| Code Extension | Module V: Adding interactive input loop |
| Full App Generation | Module VI: Complete Streamlit UI from a single prompt |
| DevOps Generation | Module VII: Dockerfile and test suite creation |
flowchart LR
U[" User Query"] -->|submit| UI[" Streamlit Web UI"]
UI -->|query| A[" Azure AI Foundry\nAgent (GPT-4o)"]
A -->|search| S[" Azure AI Search\nRAG Retrieval"]
S -->|retrieve| B[" Blob Storage\n12 KB Docs"]
B -.->|chunks| S
S -.->|context| A
A -.->|report| UI
UI -.->|display| U
12 compliance documents covering:
- RBI — Data localization, cross-border transactions, vendor onboarding
- GDPR — Article 44 transfers, Schrems II, Standard Contractual Clauses
- DPDP Act — India's Digital Personal Data Protection Act 2023
- SEBI — Insider trading compliance
- Export Controls — US/EU/India technology restrictions
- Frameworks — Risk scoring, incident response, DPA templates
See kb_markdown/ for all documents.
-
Clone this repository:
git clone https://github.com/ADKWDHQWQHDQI/Agent-using-GHCP.git cd Agent-using-GHCP -
Open in VS Code:
code . -
Start with the Introduction or jump directly to Module I.
Agent-using-GHCP/
├── README.md ← This file
├── lab/
│ └── instructions/
│ ├── 00_Introduction.md
│ ├── 01_Prerequisites.md
│ ├── 02_Azure_Resources.md
│ ├── 03_Agent_Design.md
│ ├── 04_Code_Export_Debug.md
│ ├── 05_Interactive_Input.md
│ ├── 06_Web_UI.md
│ └── 07_Containerize_Test.md
├── kb_markdown/ ← 12 compliance Knowledge Base documents
├── kb_source_pdfs/ ← Source reference index
└── idea.txt ← Original concept document
This is a workshop repository. For feedback or improvements, open an issue or pull request.
This workshop is for educational purposes. All compliance documents are sourced from public regulatory sources. No proprietary or confidential data is included.