💬 Frequently Asked Questions (FAQ)¶

🏠 Home > 📚 Docs > 💬 FAQ

Last Updated: 2026-04-27 | Version: 3.0 Status: ✅ Final | Maintainer: Documentation Team

📑 Table of Contents¶

• 🎯 General Questions
• 📋 Prerequisites & Setup
• 🚀 Deployment Questions
• 🎲 Data Generation
• 📖 Tutorial Questions
• 📊 Power BI & Reporting
• 🔐 Security & Compliance
• 🔧 Troubleshooting
• 💰 Cost & Licensing
• 🐳 Docker & Dev Containers
• 🏗️ Architecture Deep Dive
• ⚡ Performance Tuning
• 🤖 MLOps & AI
• 🔄 Migrations
• 🛠️ Dev Experience
• 📜 Compliance Frameworks

🎯 General Questions¶

What is this POC?¶

This is a production-ready proof-of-concept environment for Microsoft Fabric, purpose-built for the casino and gaming industry. It demonstrates:

• ✅ Complete medallion architecture (Bronze/Silver/Gold)
• ✅ Real-time slot machine telemetry
• ✅ Player 360 analytics
• ✅ Regulatory compliance (NIGC MICS, FinCEN BSA)
• ✅ Direct Lake Power BI dashboards
• ✅ Data governance with Microsoft Purview

Key Value: Provides a working reference implementation that can be customized for your specific casino operations.

Who is this POC for?¶

What data domains are covered?¶

How long does it take to complete the full POC?¶

3-Day Workshop Format: - Day 1 (8 hours): Medallion foundation (Bronze + Silver) - Day 2 (8 hours): Gold layer + Real-time analytics - Day 3 (8 hours): Power BI + Governance

Self-Paced Learning: 2-4 weeks (2-3 hours per week)

See the 3-Day POC Agenda for detailed schedules.

Can I use this for non-gaming industries?¶

Yes! While this POC is casino-focused, the architecture patterns apply to many industries:

See the tutorial modules (14–37) for vertical-specific examples.

📋 Prerequisites & Setup¶

What Azure resources do I need?¶

az provider register --namespace Microsoft.Fabric
az provider register --namespace Microsoft.Purview
az provider register --namespace Microsoft.Storage
az provider register --namespace Microsoft.KeyVault

What is the minimum Fabric capacity SKU?¶

Why F64 for POC? - Sufficient compute for parallel data processing - Can handle real-time streaming workloads - Supports multiple concurrent users

Can I start smaller? Yes, but expect slower performance with F2/F4. Good for learning tutorials but not representative of production performance.

See Cost Estimation Guide for detailed pricing.

Do I need to install anything locally?¶

Quick Answer: No, if using Dev Container or Codespaces.

Local Installation Option:

Easiest Setup: Use GitHub Codespaces (zero installation) or VS Code Dev Container.

See Prerequisites - Dev Container Setup.

How do I enable Fabric in my tenant?¶

What permissions do I need?¶

🚀 Deployment Questions¶

What are the deployment options?¶

See Deployment Guide for detailed instructions.

How do I deploy to Azure?¶

# 1. Login to Azure
az login
az account set --subscription "<your-subscription-id>"

# 2. Deploy infrastructure
az deployment sub create \
  --location eastus2 \
  --template-file infra/main.bicep \
  --parameters infra/environments/dev/dev.bicepparam

# 3. Verify deployment
./scripts/verify-deployment.sh

Common deployment errors?¶

az provider register --namespace Microsoft.Fabric
az provider show --namespace Microsoft.Fabric --query "registrationState"

az role assignment list --assignee "$(az ad signed-in-user show --query id -o tsv)"

PURVIEW_ACCOUNT_NAME=pv-fabricpoc-<your-unique-id>

How do I verify deployment succeeded?¶

./scripts/verify-deployment.sh

# List deployed resources
az resource list --resource-group "rg-fabric-poc-dev" --output table

How do I delete everything?¶

⚠️ Warning: This is irreversible. Ensure you have backups.

# Delete resource group (removes all resources)
az group delete --name "rg-fabric-poc-dev" --yes --no-wait

# Remove lock first if resources are locked
az lock delete --name "CanNotDelete" --resource-group "rg-fabric-poc-dev"

See Deployment Guide - Cleanup for details.

🎲 Data Generation¶

How do I generate sample data?¶

# Quick demo dataset (7 days, small)
docker-compose run --rm demo-generator

# Full dataset (30 days, production-like)
docker-compose run --rm data-generator

# Custom parameters
docker-compose run --rm data-generator --slots 100000 --players 5000 --days 14

# Install dependencies
pip install -r requirements.txt

# Generate all data
python data_generation/generate.py --all --days 30

# Generate specific domains
python data_generation/generate.py --slots 50000 --players 2000

# Sample data is already in the repository
ls sample-data/bronze/

# 10,000 slot events, 500 players, etc.

What data volumes are generated by default?¶

Customization:

# Scale up for larger POCs
docker-compose run --rm data-generator --all --days 90

# Scale down for quick testing
docker-compose run --rm demo-generator  # 7 days, smaller volumes

Is the generated data realistic?¶

Yes! The data generators include:

✅ Realistic distributions based on industry patterns ✅ Referential integrity (Player IDs match across tables) ✅ Compliance logic (CTR $10K threshold, W-2G $1,200) ✅ Time-series patterns (hourly/daily seasonality) ✅ PII protection (hashed SSN, masked credit cards)

Example realistic patterns: - Slot machine hold percentage: 8-12% - Player loyalty tiers: Bronze (60%), Silver (30%), Gold (8%), Platinum (2%) - Peak gaming hours: 7pm-2am (weekends higher) - CTR generation: ~0.5% of transactions

See Data Generation - Data Quality Features.

How do I customize the generated data?¶

# Adjust volumes
--slots 1000000        # 1 million slot events
--players 50000        # 50,000 player profiles
--days 90              # 90 days of history

# Adjust format
--format csv           # Output as CSV instead of Parquet
--format json          # Output as JSON

# Control randomness
--seed 12345           # Use specific seed for reproducibility

from generators import SlotMachineGenerator

# Custom generator configuration
generator = SlotMachineGenerator(
    num_machines=1000,
    denominations=[0.01, 0.25, 0.50, 1.00, 5.00],
    hold_pct_mean=0.10,  # 10% hold
    seed=42
)

df = generator.generate(100000)

How do I stream data to Event Hub?¶

EVENTHUB_CONNECTION_STRING="Endpoint=sb://..." \
EVENTHUB_NAME="slot-telemetry" \
STREAMING_RATE=10 \
docker-compose up streaming-generator

from generators import SlotMachineGenerator
from streaming import EventHubStreamer

# Configure streamer
streamer = EventHubStreamer(
    connection_string=os.getenv("EVENTHUB_CONNECTION_STRING"),
    eventhub_name="slot-telemetry"
)

# Stream events
generator = SlotMachineGenerator()
for event in generator.generate_stream(events_per_second=10):
    streamer.send(event)

📖 Tutorial Questions¶

What's the learning path?¶

graph LR
    T00[00-Setup] --> T01[01-Bronze]
    T01 --> T02[02-Silver]
    T02 --> T03[03-Gold]
    T03 --> T04[04-Real-Time]
    T04 --> T05[05-Power BI]
    T05 --> T06[06-Pipelines]
    T06 --> T07[07-Governance]
    T07 --> T08[08-Mirroring]
    T08 --> T09[09-AI/ML]

Recommended Path: 1. 🟢 Foundation (00-01): Environment setup, Bronze layer 2. 🟡 Core (02-03): Silver and Gold layers 3. 🟠 Advanced (04-05): Real-time analytics, Power BI 4. 🔴 Enterprise (06-09): Pipelines, governance, AI/ML

See Tutorials README for complete learning path.

Can I skip tutorials?¶

Not Recommended. Each tutorial builds on the previous one:

Minimum POC: Complete tutorials 00-03 + 05 (Power BI).

Tutorial 00: Environment setup issues?¶

Tutorial 01: Bronze layer issues?¶

# Check file path
display(dbutils.fs.ls("Files/bronze/"))

# Verify Parquet format
df = spark.read.format("parquet").load("Files/bronze/slot_telemetry.parquet")
display(df.limit(10))

df = spark.read.format("parquet") \
    .option("mergeSchema", "true") \
    .load("Files/bronze/")

Tutorial 05: Direct Lake not working?¶

-- Optimize Delta table for Direct Lake
OPTIMIZE gold_slot_performance
USING (VORDER);

📊 Power BI & Reporting¶

What is Direct Lake mode?¶

Direct Lake is a revolutionary data connectivity mode that combines the best of Import and DirectQuery:

How it works: - Queries execute directly against Delta tables in OneLake - Uses V-Order optimization for sub-second performance - No data import or scheduled refresh required - Automatic fallback to DirectQuery if needed

See Tutorial 05: Direct Lake & Power BI.

Do I need a Power BI license?¶

How often does Direct Lake refresh?¶

Answer: It doesn't! That's the beauty of Direct Lake.

Traditional Import Mode: - Requires scheduled refresh (e.g., every 8 hours) - Data is stale between refreshes - Consumes refresh capacity

Direct Lake: - ✅ Always queries the latest data - ✅ No refresh schedule needed - ✅ Updates appear immediately when Delta tables update

Exception: If you have calculated tables or composite models, those components might need refresh.

What reports are included?¶

Location: reports/report-definitions/

See Reports README for import instructions.

Can I customize the reports?¶

Absolutely! The reports are provided as starting templates.

Customization Options: 1. Edit in Power BI Desktop: - Open .pbip files - Modify visuals, add pages - Adjust DAX measures

1. Create New Reports:
2. Connect to existing semantic model
3. Build custom visuals

4. Apply your branding

5. Add Custom DAX:

   // Example: Average Daily Theo per Player
   Avg Daily Theo = 
   DIVIDE(
       SUM(gold_player_360[total_theo]),
       DISTINCTCOUNT(gold_player_360[player_id])
   )
   

Best Practice: Copy template first, then customize.

How do I implement Row-Level Security (RLS)?¶

[PropertyID] = LOOKUPVALUE(
    UserPropertyMapping[PropertyID],
    UserPropertyMapping[UserEmail],
    USERPRINCIPALNAME()
)

🔐 Security & Compliance¶

What compliance frameworks are covered?¶

How is PII protected?¶

How are secrets managed?¶

Never commit secrets to Git! This repository has multiple protections:

✅ .gitignore: Blocks common secret files ✅ Pre-commit hook: Scans for high-risk patterns ✅ .env.sample: Provides template without secrets ✅ Sample data: All PII masked

Best Practices: 1. Use .env files locally (gitignored) 2. Store production secrets in Azure Key Vault 3. Use managed identities for Azure authentication 4. Enable the pre-commit hook:

git config core.hooksPath .githooks

If you accidentally commit a secret: 1. Assume it's compromised - rotate immediately 2. Remove from Git history with BFG Repo-Cleaner 3. Report per your security policy

See Security Guide - Repository Security.

What network security options are available?¶

🔧 Troubleshooting¶

Notebook fails with "Capacity not available"¶

Cause: Fabric capacity is paused or inactive.

Fix: 1. Go to Fabric Admin Portal 2. Navigate to Capacity settings 3. Ensure capacity status is Active 4. If paused, click Resume

Prevention: Disable auto-pause for POC demos.

"Cannot connect to Lakehouse" error¶

# In notebook
%run ./_notebook_config

# Should show lakehouse tables
display(spark.catalog.listTables())

Delta table "file not found" errors¶

Cause: Delta transaction log corruption or incomplete writes.

Fix:

# Repair Delta table
from delta.tables import DeltaTable

DeltaTable.forPath(spark, "Tables/bronze_slot_telemetry").generate("symlink_format_manifest")

# Or vacuum old files
DeltaTable.forPath(spark, "Tables/bronze_slot_telemetry").vacuum()

Prevention: Use proper DataFrame writes with checkpoints.

Power BI reports show "Unable to connect"¶

Data generator produces "invalid schema" warnings¶

Cause: Schema mismatch between generator and expected Bronze schema.

Fix:

# Update to latest generator code
git pull origin main

# Or specify schema version
python generate.py --all --schema-version 1.1

Workaround: Use mergeSchema when reading:

df = spark.read.format("parquet") \
    .option("mergeSchema", "true") \
    .load("Files/bronze/")

"Out of memory" errors during data processing¶

# Bad: Load entire dataset
df = spark.read.format("delta").load("Tables/bronze_slot_telemetry")

# Good: Filter first
df = spark.read.format("delta").load("Tables/bronze_slot_telemetry") \
    .filter("event_date >= '2024-01-01'")

df.repartition(10).write.format("delta").save("Tables/silver_slot_clean")

💰 Cost & Licensing¶

What will this POC cost me?¶

Quick Estimates:

Cost Breakdown (1-month F64 24/7): - Fabric Capacity: ~\(8,500 (80%) - ADLS Gen2 Storage: ~\)500 (5%) - Purview: ~\(800 (8%) - Other services: ~\)700 (7%)

Major Cost Driver: Fabric capacity (75-80% of total cost).

See Cost Estimation Guide for detailed scenarios.

How can I reduce costs?¶

# Pause capacity when not in use
az fabric capacity pause --name "fabric-casino-poc"

# Resume when needed
az fabric capacity resume --name "fabric-casino-poc"

# Move cold data to Cool tier after 30 days
az storage blob lifecycle-management set \
  --policy @lifecycle-policy.json \
  --account-name $STORAGE_ACCOUNT_NAME

# Automated pause/resume schedule
# Run via Azure Automation or GitHub Actions
Start-FabricCapacity -Time "08:00" -Days "Mon-Fri"
Stop-FabricCapacity -Time "18:00" -Days "Mon-Fri"

Do I need to pay for Power BI separately?¶

Short Answer: Maybe, depends on your usage.

Licensing Options:

POC Recommendation: 1. Use Power BI Pro trial (60 days free) 2. Publish to Fabric workspace 3. Viewers can use free licenses

Note: Fabric capacity is already included in your F64 cost estimate - it provides Power BI Premium features.

What about Azure free tier/credits?¶

Azure Free Tier: - ❌ Microsoft Fabric not included in free tier - ✅ Some supporting services (Storage, Key Vault) have free allowances

Azure Credits: - ✅ Can use Fabric with Azure credits (students, startups) - ✅ Visual Studio subscriptions include monthly credits

Free Trials: - Power BI Pro: 60-day trial - Fabric capacity: Trial available via Microsoft - Azure subscription: $200 credit for 30 days (new customers)

POC on a Budget: - Use F2 capacity (\(265/month or ~\)9/day) - Pause when not in use - Limit to 3-day POC demo ($27-35 total)

🐳 Docker & Dev Containers¶

What's the difference between Docker and Dev Container?¶

Docker: Run generators, validate data, stream to Event Hub Dev Container: Complete development setup with all tools pre-installed

How do I use Docker for data generation?¶

# Quick demo (7 days, small dataset)
docker-compose run --rm demo-generator

# Full dataset (30 days, production-like volumes)
docker-compose run --rm data-generator

# Custom parameters
docker-compose run --rm data-generator --all --days 14 --format csv

# Specific data domains
docker-compose run --rm data-generator --slots 100000 --players 5000

Output: ./output directory

See Docker Support in main README.

How do I use Dev Containers?¶

Option 1: VS Code Local 1. Install Docker Desktop 2. Install Dev Containers extension 3. Open repository in VS Code 4. Click "Reopen in Container" when prompted

Option 2: GitHub Codespaces (Zero Installation) 1. Go to repository on GitHub 2. Click Code > Codespaces tab 3. Click Create codespace on main 4. Wait ~2 minutes for environment to build

What You Get: - ✅ Python 3.11 with all dependencies - ✅ Azure CLI + Bicep - ✅ PowerShell 7 - ✅ Git configured - ✅ All VS Code extensions pre-installed

Docker commands are failing?¶

# Docker Compose V2 (preferred)
docker compose run --rm data-generator

# Docker Compose V1 (legacy)
docker-compose run --rm data-generator

# Change port in docker-compose.yml
ports:
  - "8081:8080"  # Use 8081 instead

# Or stop conflicting service
docker ps  # Find conflicting container
docker stop <container-id>

# Rebuild without cache
docker-compose build --no-cache data-generator

# Or pull latest image
docker-compose pull

🏗️ Architecture Deep Dive¶

Why Lakehouse instead of Warehouse for this POC?¶

The POC chose Lakehouse as the primary store for three reasons: (1) the diverse data formats across 9 industry verticals (Parquet, CSV, JSON) favor schema-on-read flexibility; (2) the PySpark-first notebook workflow aligns naturally with Lakehouse's Spark engine; and (3) Direct Lake mode provides zero-copy Power BI connectivity without Import refresh schedules. Warehouse is the better choice for T-SQL-heavy teams or migrations from Synapse Dedicated SQL Pool.

See: DECISION_TREES.md | Lakehouse/Warehouse/SQL DB Decision Guide

What goes in each medallion layer?¶

The key principle: Bronze is append-only (never modify source records), Silver deduplicates and validates (MERGE upserts), Gold aggregates for consumption (overwrite on refresh).

See: Medallion Architecture Deep Dive

How should I design workspaces?¶

The recommended pattern for this POC is a per-environment workspace layout:

Each workspace contains three Lakehouses (lh_bronze, lh_silver, lh_gold), one Warehouse (for T-SQL consumers), and one Eventhouse (for real-time). For multi-tenant scenarios, see Multi-Tenant Workspace Architecture.

See: Workspace Naming

When should I use shortcuts vs. copying data?¶

Use shortcuts when you want to query data in-place without storage duplication (e.g., referencing ADLS Gen2 landing zones or cross-workspace tables). Use copy (pipeline Copy Activity) when you need to transform data during ingestion, the source requires a data gateway, or you want full control over the data lifecycle in OneLake. Shortcuts are free (no storage cost); copies consume storage.

See: DECISION_TREES.md | Shortcut Transformations Notebook

What is Workspace Identity and when do I need it?¶

Workspace Identity is a managed identity scoped to a Fabric workspace. It enables credential-free authentication to Azure resources (Storage, Key Vault, Purview) from notebooks and pipelines -- no service principal secrets to rotate. Use it whenever your notebooks access Azure resources. The POC deploys it via infra/modules/security/workspace-identity.bicep.

See: OneLake Security | Workspace Identity Module

⚡ Performance Tuning¶

What is V-Order and do I need it?¶

V-Order is a write-time optimization for Parquet files that dramatically improves Direct Lake query performance. It reorders data within row groups for optimal column compression and scan efficiency. You need it on every Gold table that feeds a Power BI semantic model via Direct Lake. Enable it with:

spark.conf.set("spark.sql.parquet.vorder.enabled", "true")

Or apply retroactively: OPTIMIZE gold_table_name USING VORDER.

See: Direct Lake | Performance & Parallelism

How should I partition large tables?¶

Partition by the most common filter column (typically a date column). For this POC, Bronze and Silver tables partition by event_date for efficient time-range queries. Rules of thumb:

• Partition size target: 256 MB - 1 GB per partition
• Do not over-partition: Avoid partitioning by high-cardinality columns (player_id) -- too many small files
• Combine with Z-Order: OPTIMIZE table ZORDER BY (property_id) within each partition for multi-column filtering

See: Performance & Parallelism | Medallion Deep Dive

What Spark settings should I tune first?¶

For POC-scale data (~700K-1M records per table), the most impactful settings are:

See: CHEAT_SHEETS.md | Spark Notebooks Best Practices

How do I prevent Direct Lake fallback to DirectQuery?¶

Direct Lake falls back to DirectQuery when: (1) the model contains calculated tables; (2) column cardinality exceeds guardrails; (3) the query uses unsupported DAX patterns. To prevent fallback:

1. Move all calculated tables into Gold notebooks (materialize as Delta tables)
2. Pre-aggregate high-cardinality columns in Gold layer
3. Monitor fallback using Power BI Performance Analyzer
4. Keep Gold tables V-Order optimized

See: Direct Lake | CHEAT_SHEETS.md

🤖 MLOps & AI¶

What ML models does this POC include?¶

The POC includes three ML notebooks:

All models use MLflow for experiment tracking and model registry.

See: ML Notebooks | AutoML Model Endpoints

How does model versioning work in Fabric?¶

Fabric uses MLflow's model registry natively. Models are logged during training with mlflow.log_model(), registered in the workspace model registry, and versioned automatically. Fabric's ML model item provides a UI for version comparison, stage transitions (Staging/Production), and deployment to endpoints.

See: AutoML Model Endpoints

Can I use AI Functions in notebooks?¶

Yes. Fabric AI Functions (ai_summarize, ai_classify, ai_translate, etc.) are available in Spark SQL for inline LLM-powered transformations. The POC demonstrates compliance-aware usage in 17_gold_ai_functions_compliance.py, including token cost estimation and PII guardrails.

See: AI Copilot Configuration | AI Functions Notebook

What about Data Agents?¶

Data Agents are autonomous AI-powered analytics assistants that can answer natural language questions about your data. They run inside Fabric workspaces with governed access to Lakehouses and Warehouses. The POC documents configuration patterns but does not deploy a live agent (requires tenant admin enablement).

See: Data Agents | Fabric IQ

🔄 Migrations¶

How do I migrate from Synapse Analytics?¶

The migration path depends on your current Synapse component:

See: Migration Patterns | Tutorial 13: Migration Planning

How do I migrate from Databricks?¶

Key differences to address:

1. Runtime: Replace dbutils with mssparkutils (file system, credentials, notebook orchestration)
2. Unity Catalog: Map to Fabric OneLake + Purview for governance
3. Delta Lake: Fully compatible -- Delta tables work as-is in OneLake
4. MLflow: Supported natively in Fabric
5. Notebook format: Databricks notebook source format imports directly

The POC notebooks already use the Databricks notebook format with # COMMAND ---------- separators. Phase 11 remediation ensured all dbutils references were replaced with mssparkutils.

See: Migration Patterns

How do I migrate from Snowflake?¶

Use Fabric Mirroring for continuous replication from Snowflake into OneLake (Delta format). This provides near-real-time sync without building custom ETL. Alternatively, use Snowflake's COPY INTO to export to ADLS Gen2, then create Lakehouse shortcuts to the exported data.

See: Mirroring | Tutorial 24: Snowflake to Fabric

What about Teradata and IBM DB2?¶

Both are covered in the POC:

• Teradata: Tutorial 10 covers TPT export patterns and migration planning
• IBM DB2: Streaming notebook 04_ibm_db2_cdc.py demonstrates CDC from DB2 z/OS and LUW with EBCDIC handling

For both, the typical pattern is: set up an on-premises Data Gateway, configure a pipeline Copy Activity, and land data in the Bronze Lakehouse.

See: Tutorial 10: Teradata Migration | IBM DB2 CDC Notebook

🛠️ Dev Experience¶

Can I develop notebooks locally?¶

Yes, but with caveats. Notebooks use the Databricks notebook format (.py files with # COMMAND ---------- separators) and can be edited in any IDE. However, mssparkutils and spark are only available inside Fabric. The POC includes a _get_arg shim at the top of every notebook so code can run in both Fabric and local pytest:

try:
    from notebookutils import mssparkutils
except ImportError:
    mssparkutils = None

The 612 unit tests in validation/unit_tests/ validate notebook logic locally without a Fabric session.

See: Testing Strategies

How does Git integration work with Fabric?¶

Fabric workspaces can connect to Azure DevOps or GitHub repos. Each Fabric item (notebook, pipeline, semantic model) is serialized as a JSON/YAML/Python file and synced bi-directionally. Best practice: establish a one-way flow (edit in IDE, push to Git, sync to Fabric) to avoid merge conflicts.

See: Git Integration | fabric-cicd Deployment

What CI/CD tool should I use?¶

The POC uses two complementary approaches:

fabric-cicd is the Microsoft-recommended tool for deploying Fabric workspace items. It handles notebook uploads, pipeline definitions, and semantic model refreshes.

See: fabric-cicd Deployment | Tutorial 12: CI/CD DevOps

How do I run tests?¶

# All 612 unit tests
pytest validation/unit_tests/ -v

# By category
pytest validation/unit_tests/test_generators.py -v      # Casino (30 tests)
pytest validation/unit_tests/federal/ -v                 # Federal (54 tests)
pytest validation/unit_tests/streaming/ -v               # Streaming (20 tests)
pytest validation/unit_tests/analytics/ -v               # Analytics (30 tests)

# Data quality (Great Expectations)
great_expectations checkpoint run bronze_checkpoint

See: Testing Strategies

📜 Compliance Frameworks¶

What compliance frameworks does this POC address?¶

See: Security | SQL Audit Logs | CMK

How are CTR and SAR reports generated?¶

Currency Transaction Reports (CTR): Any cash transaction >= $10,000 triggers automatic CTR flagging in the Bronze compliance notebook (04_bronze_compliance.py). The Silver layer validates amounts and deadlines. The Gold layer (03_gold_compliance_reporting.py) produces FinCEN-ready reports.

Suspicious Activity Reports (SAR): The Silver layer detects structuring patterns -- multiple transactions between \(8,000-\)9,999 by the same individual within 24 hours. The fraud detection ML model (02_ml_fraud_detection.py) provides additional anomaly scoring.

See: Compliance Reporting Notebook

How is HIPAA compliance handled?¶

Tribal Healthcare notebooks implement HIPAA safeguards:

1. PHI Masking: Silver layer (07_silver_tribal_health.py) masks protected health information
2. Audit Logging: Every data access is logged with user ID, timestamp, and data accessed
3. FHIR R4 Mapping: Data mapped to standardized FHIR R4 format for interoperability
4. 42 CFR Part 2: Substance abuse treatment records have additional access restrictions
5. Retention: Log Analytics configured for >= 6 years (HIPAA requirement) via log-analytics.bicep

See: Tribal Health Analytics | Tutorial 30: Tribal Healthcare

What encryption options are available?¶

See: Customer-Managed Keys | Network Security

📚 Additional Resources¶

Where can I learn more about Microsoft Fabric?¶

Official Documentation: - Microsoft Fabric Documentation - Fabric Architecture Center - Fabric Pricing

Community Resources: - Microsoft Fabric Blog - Microsoft Fabric Community - Power BI Community

Training: - Microsoft Learn: Fabric Learning Path - Data Engineering with Fabric

Where do I report issues or contribute?¶

GitHub Repository: - 🐛 Report bugs: Open an issue - 💡 Feature requests: Start a discussion - 🔧 Pull requests: See Contributing Guide

Before Opening an Issue: 1. Check existing issues for duplicates 2. Review FAQ (this document) 3. Include reproduction steps 4. Provide error messages and logs

How do I stay updated?¶

Watch the Repository: - Click Watch on GitHub - Choose notification preferences

Release Notes: - Check CHANGELOG.md for version history - Subscribe to releases on GitHub

Social Media: - Follow Microsoft Fabric on Twitter - Join the LinkedIn Fabric community

🎯 Quick Reference¶

Essential Links¶

Command Cheat Sheet¶

# Deployment
az deployment sub create --location eastus2 --template-file infra/main.bicep --parameters infra/environments/dev/dev.bicepparam

# Data Generation
docker-compose run --rm data-generator --all --days 30

# Verify
./scripts/verify-deployment.sh

# Cleanup
az group delete --name "rg-fabric-poc-dev" --yes

Support¶

Need help? Try these resources in order:

1. 📖 Check this FAQ
2. 🔍 Search existing issues
3. 💬 Ask in GitHub Discussions
4. 🐛 Open a new issue

⬆️ Back to top

📖 Documentation maintained by: Microsoft Fabric POC Team 🔗 Repository: Suppercharge_Microsoft_Fabric 🔄 Last Updated: 2025-01-21