🏦 Agentic Banking App with Microsoft Fabric

Last updated: 🟢 2026-04-27 🟢

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 & Contribute!	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

IMPORTANT: Before you begin, install the following:

Requirement	Notes
VS Code	Recommended; tested environment
Dev Containers extension	Required to use the included Dev Container in VS Code
Docker Desktop / Docker Engine	Required to use the included Dev Container
Azure CLI	Auth. - Windows · macOS
Node.js v18+	Frontend. Skip this install when using the Dev Container; it already includes Node.js 20.19.0
Python 3.11.9+	Backend. Skip this install when using the Dev Container; it already includes Python 3.11
ODBC Driver 18 for SQL Server	Database connectivity. Skip this install when using the Dev Container; it already includes ODBC Driver 18
Microsoft Fabric capacity	Start a free 60-day trial if needed. NOTE: Data Agent is not available in Free Trial
Azure OpenAI resource	Create one in Azure Portal - Deploy an llm model (ex. gpt 4.1) and text-embedding-ada-002 embedding model

---

Setup with dev container

Step 1 — Clone the Repo

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

Step 2 - Run the Dev Container

Open the main folder in VS Code and run Dev Containers: Rebuild and Reopen in Container in the quick access bar as shown below (Reminder that you must have Docker installed):

Running this command start the process of seting up the container, during which .devcontainer/post-create.sh creates .venv virtual environment, runs pip install -r requirements.txt and install required Python version and libraries, runs npm install, copies backend/.env.sample to backend/.env if needed, and the container forwards ports 5173, 5001, and 5002

When container setup is complete, you will see below in VS code terminal:

Press enter and you will see below, which means now you are in the container:

⚠️Facing Issues with Dev container Setup? Follow steps here instead to install dependencies on your system: Manual Setup

Step 3 — 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 4 — Preflight Check

Before deploying anything, run the preflight to verify your dev container, Azure context, region, Fabric capacity, OpenAI deployments, and tenant permissions are all set correctly. This catches 95% of issues up-front instead of discovering them halfway through a 10-minute deployment.

First, set your target region in backend/.env (single source of truth used by all scripts):

# backend/.env
AZURE_REGION="swedencentral"            # or eastus2, westeurope, etc.
AZURE_RESOURCE_GROUP="agentic-app-with-fabric-rg"

Both values come pre-populated in backend/.env.sample (copied to .env by the dev container). Edit AZURE_REGION to match where you want your Fabric capacity, Cosmos DB, and Azure OpenAI provisioned.

Then run:

# Reads AZURE_REGION + AZURE_RESOURCE_GROUP from backend/.env
python3 scripts/preflight.py

# Override the region for a single run (also persisted back to .env)
python3 scripts/preflight.py --region eastus2

# Auto-fix: register providers, create RG + Fabric capacity (F4),
# create Azure OpenAI account + deploy gpt-4o + text-embedding-ada-002,
# and write AZURE_OPENAI_KEY/ENDPOINT/DEPLOYMENT into backend/.env
python3 scripts/preflight.py --fix

# Pin a specific OpenAI resource if you have several
python3 scripts/preflight.py --openai-resource my-aoai

Exit code 0 = ready, exit code 1 = blocking issues found (suggested fixes printed).

ℹ️ Region notes:

• Cosmos DB in Fabric is unsupported in: indiawest, qatarcentral, uaecentral, austriaeast, chilecentral, southcentralus. In israelcentral it's documented as available but the rollout is delayed — preflight will warn and suggest a fallback (scripts/provision_azure_cosmos.py provisions a regular Azure Cosmos DB account instead).
• Embedding model must be text-embedding-ada-002 (the repo's vector embeddings were generated with it).
• Chat model preflight cascades through gpt-4o-mini → gpt-4o → gpt-4.1-mini → gpt-4.1 and stops at the first one that deploys in your region/quota.

---

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

Wait for the script to finish (disregard warnings if you see any), at the end, it will give you a final status of all items that were successfully deployed.

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

---

Step 6 — Finalize the Deployment

After setup_workspace.py completes, wait a minute or two and 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               # default is false
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