🏦 Agentic Banking App with Microsoft Fabric

Last updated: 🟢 2026-04-16 🟢

An interactive banking demo that shows how databases power OLTP, OLAP, and AI workloads side-by-side, all wired together through Microsoft Fabric.

🚀 Try the live app

---

Table of Contents

🚀 Setup Guide	Get the app running locally
⚗️ Explore Fabric Workloads	RTI, Power BI, Data Agent walkthroughs
🧑‍💻 Learn More	Architecture, agents, embeddings, contributing

---

What This App Does

• Transactions (OLTP) — real-time writes and reads against a Fabric SQL Database
• Analytics (OLAP) — an analytics dashboard with charts and summaries of spending habits. This represents an OLAP workload, running complex, aggregate queries over a large dataset.
• AI Agents — multi-agent LangGraph system with a coordinator, support, account, Fabric Data Agent (optional), and visualization agent
• Generative UI — agents create personalized interactive visualizations on the fly
• Real-time Monitoring — app usage and content safety data streamed to Fabric Eventhouse via Eventstream

---

🚀 Setup Guide

Prerequisites

Before you begin, install the following:

Requirement	Notes
Node.js v18+	Frontend
Python 3.11.9+	Backend
Azure CLI	Auth — Windows · macOS
ODBC Driver 18 for SQL Server	Database connectivity
Microsoft Fabric capacity	Start a free 60-day trial if needed
Azure OpenAI resource	Create one in Azure Portal

Recommended: VS Code (tested environment)

---

Quick Setup Video

Step 1 — Clone the Repo

git clone https://github.com/Azure-Samples/agentic-app-with-fabric.git
cd agentic-app-with-fabric

---

Step 2 — Install Python Dependencies

# Create and activate a virtual environment
# Windows: 
python -m venv venv
.\venv\Scripts\activate

# macOS / Linux: 
python3 -m venv venv
source venv/bin/activate

# Install packages
pip install -r requirements.txt

---

Step 3 — Install Frontend Dependencies

npm install

---

Step 4 — Log in to Azure

az login

Use your Microsoft Fabric account credentials. Watch for browser pop-ups.

⚠️ You must repeat az login any time you restart the backend.

---

Step 5 — Deploy the Fabric Workspace

This single command creates the workspace, deploys all Fabric artifacts, creates SQL tables, and populates backend/.env with connection strings automatically.

# Windows:
python scripts/setup_workspace.py --workspace-name "AgenticBankingApp-{yourinistials}"

# Mac:
python3 scripts/setup_workspace.py --workspace-name "AgenticBankingApp-{yourinistials}"

The script will prompt you to select a Fabric capacity, then create a workspace named AgenticBankingApp-{yourinistials}, or any other name to make the workspace name unique.

What gets deployed:

Artifact	Type
agentic_app_db	SQL Database
agentic_cosmos_db	Cosmos DB
agentic_lake	Lakehouse
banking_semantic_model	Semantic Model
Agentic_Insights	Power BI Report
Banking_DataAgent	Data Agent
agentic_eventhouse	Eventhouse + KQL Database
agentic_stream	Eventstream
ContentSafetyMonitoring	KQL Dashboard
QA_Evaluation_Notebook	Notebook

💡 Alternative deployment: prefer Git integration? See Deploy via Git Integration.

---

Step 6 — Finalize the Deployment

After setup_workspace.py completes, run:

# Windows:
python scripts/finalize_views_and_report.py

# Mac:
python3 scripts/finalize_views_and_report.py 

This finalizes the Lakehouse SQL views, patches the Semantic Model, and deploys the Power BI Report. The workspace ID is read automatically from the previous step — no argument needed.

When done, your workspace lineage should look like this:

---

Step 7 — Configure Environment Variables

setup_workspace.py already auto-populated some of backend/.env. You only need to fill in the Azure OpenAI values, Cosmos DB Endpoint and EventHub connection details.

Open backend/.env and set (copy and rename .env.sample template file):

Azure OpenAI (required)

AZURE_OPENAI_KEY= Your API key for the Azure OpenAI service. You can find this in the Azure Portal by navigating to your Azure OpenAI resource and selecting Keys and Endpoint.
AZURE_OPENAI_ENDPOINT="https://<your-resource>.openai.azure.com/"
AZURE_OPENAI_DEPLOYMENT="<your chat model name, e.g. gpt-4o-mini>"
AZURE_OPENAI_EMBEDDING_DEPLOYMENT="text-embedding-ada-002"

⚠️ The embedding deployment must be text-embedding-ada-002 — embeddings in the repo were generated with that model.

Microsoft Fabric Cosmos DB

COSMOS_DB_ENDPOINT= You can find this in your Fabric workspace by navigating to the Cosmos DB artifact in your workspace, clicking the "settings" -> "Connection" tab and copy the endpoint string.

EventHub Connection (required for real-time monitoring)

In your Fabric workspace, open the agentic_stream Eventstream → click CustomEndpoint → SAS Key Authentication tab:

Copy the two values into .env:

FABRIC_EVENT_HUB_NAME="<Event hub name>"
FABRIC_EVENT_HUB_PRIMARY_KEY="<Connection string-primary key>"
NOTE: first click on the eye button near "Connection string-primary key" to reveal the value, then copy the value

Auto-populated Variables (no action needed)

The following were written to .env automatically by the deployment script:

FABRIC_SQL_CONNECTION_URL_AGENTIC   # SQL Database connection string
FABRIC_DATA_AGENT_SERVER_URL        # Data Agent MCP endpoint
FABRIC_DATA_AGENT_TOOL_NAME         # Data Agent tool name
USE_FABRIC_DATA_AGENT               # Set to "true"
COSMOS_DB_DATABASE_NAME             # Set to "agentic_cosmos_db"

---

Step 8 — Run the App

Open two terminal windows (both with the virtual environment activated and after az login):

Terminal 1 — Backend

cd backend

# Windows
python launcher.py

# Mac
python3 launcher.py

Starts two services:

• Banking API → http://127.0.0.1:5001
• Agent Analytics → http://127.0.0.1:5002

Terminal 2 — Frontend

npm run dev

Opens the app at → http://localhost:5173

---

Next Steps

⚗️ Explore Fabric Workloads	Set up real-time monitoring, Power BI analytics, and the Data Agent
🧑‍💻 Learn More	Understand the architecture, multi-agent design, and how to contribute