Demo development is crucial for businesses to showcase their AI capabilities and win new customers. Through rapid prototyping and professional presentation tools, businesses can transform weeks of development into minutes of setup, dramatically accelerating sales cycles and proof-of-concept delivery.

In this Quickstart, we will build a comprehensive demo development platform called "Cortex AI Demo Framework". This demonstrates how to use Snowflake Cortex AI functions to create synthetic data, build interactive analytics, deploy search capabilities, and generate complete demonstration environments.

This Quickstart showcases the complete Cortex AI Demo Framework with:

In this step, you'll create the Snowflake database objects and prepare for framework deployment.

If you have Workspaces:

1. In Snowsight, click Projects, then Workspaces in the left navigation
2. Click + Add new to create a new Workspace
3. Click SQL File to create a new SQL file
4. Copy the setup script from setup.sql and paste it into your SQL file, then run it

If you have Worksheets:

1. In Snowsight, click Projects, then Worksheets in the left navigation
2. Click + in the top-right corner to open a new Worksheet
3. Copy the setup script from setup.sql and paste it into your worksheet, then run it

The setup script creates:

Download these framework files from the GitHub repository:

1. In Snowsight, change your role to cortex_ai_demo_data_scientist

2. Navigate to Catalog → Database Explorer → AI_FRAMEWORK_DB → APPS → Stages

Upload all framework files to the single AI_FRAMEWORK_APPS stage:

3. Click on AI_FRAMEWORK_APPS stage, then click Enable Directory Table and upload all 7 files:
   • 01_ai_framework_synthetic_data_generator.py
   • 02_ai_framework_structured_tables.py
   • 03_ai_framework_sql_to_yaml_converter.py
   • 04_ai_framework_snow_demo.py
   • 05_ai_framework_snow_viz_yaml_wizard.py
   • 06_ai_framework_snow_viz.py
   • environment.yml

1. Import into Snowflake:

   • Navigate to Projects → Notebooks in Snowsight
   • Click the down arrow next to + Notebook and select Import .ipynb file
   • Choose cortex_ai_demo_framework_setup.ipynb from your downloads

2. Configure the notebook settings:

   • Role: Select cortex_ai_demo_data_scientist
   • Database: Select AI_FRAMEWORK_DB
   • Schema: Select BRONZE_LAYER
   • Query Warehouse: Select cortex_ai_demo_wh
   • Notebook Warehouse: Select cortex_ai_demo_wh

3. Click Create to import the notebook

The notebook creates all 6 Streamlit applications using the single stage approach with automatic environment.yml detection for the latest Streamlit version.

1. Go to Projects → Notebooks in Snowsight
2. Click on CORTEX_FRAMEWORK_DEMO Notebook to open it
3. Click Run all to execute all cells in the notebook at once

What the notebook does:

The notebook processes sample data and deploys the complete framework application suite.

1. Navigate to Projects → Streamlit in Snowsight
2. You'll see 6 framework applications deployed

Creates realistic AI-powered datasets using Cortex LLMs. Saves raw JSON to BRONZE_LAYER tables.

Transforms raw JSON into clean, structured database tables. Outputs analytics-ready data to SILVER_LAYER.

Converts SQL queries into interactive demo configurations for Snow Demo (App 4).

Runs interactive SQL-driven presentations with live visualizations and AI experimentation.

Guided dashboard configuration creator. Generates YAML files for Snow Viz (App 6).

Renders advanced interactive dashboards with multi-tab analytics and AI integration.

Next: Page 5 shows which apps to use based on your role and goals.

The framework supports 4 different user personas. Find your role below to see which apps you need and in what order.

Who You Are:

What You'll Build: A complete analytics pipeline from data generation to interactive dashboards

Apps You'll Use: Synthetic Data Generator → Structured Tables → YAML Wizard → Snow Viz

Time Required: ~25 minutes

Your Workflow:

1. Synthetic Data Generator: Generate synthetic data
2. Structured Tables: Transform JSON to structured table
3. YAML Wizard: Create dashboard configuration
4. Snow Viz: View your interactive dashboard

What You'll Get:

Who You Are:

What You'll Build: Interactive SQL-driven presentations with live query execution and AI experimentation

Apps You'll Use: Synthetic Data Generator → Structured Tables → SQL to YAML Converter → Snow Demo

Time Required: ~30 minutes

Your Workflow:

1. Synthetic Data Generator: Generate synthetic data for demos
2. Structured Tables: Create structured table
3. SQL to YAML Converter: Write SQL queries and convert to demo format
4. Snow Demo: Run interactive SQL presentation

What You'll Get:

Who You Are:

What You'll Build: Clean, structured datasets for export to external tools (notebooks, ML pipelines, BI tools)

Apps You'll Use: Synthetic Data Generator → Structured Tables

Time Required: ~15 minutes

Your Workflow:

1. Synthetic Data Generator: Generate synthetic data
2. Structured Tables: Transform to structured table
3. Export data via CSV, Python/Snowpark, or direct BI tool connections

What You'll Get:

Who You Are:

What You'll Do: View and interact with dashboards created by your data team (no setup required)

Apps You'll Use: Snow Viz only (after colleague completes setup)

Time Required: ~5 minutes

Prerequisites: A colleague must first complete Synthetic Data Generator → Structured Tables → YAML Wizard to create the dashboard. Once that's done, you can view and explore it.

Your Workflow:

1. Snow Viz: Open app and select dashboard
2. Explore tabs with different visualization types
3. Use AI Assistant to ask questions in plain English
4. Export data to CSV for further analysis

What You Can Do:

Ready to get started? Jump to the pages for your persona:

Or read all app instructions (Pages 6-11) to understand the full framework capabilities.

Purpose: Create realistic AI-powered datasets for any business scenario using Cortex LLMs
Dependencies: None (START HERE)
Output: Raw JSON data saved to BRONZE_LAYER tables

All Personas start here! This is the foundation of the framework.

Navigate to Projects → Streamlit → SYNTHETIC_DATA_GENERATOR

Left Sidebar - Top Section:

For first-time use, leave "Load Configuration" as Create New. If you have saved configurations, select one from dropdown and click 📁 Load Configuration.

Left Sidebar:

Enter your company name and topic/domain:

Other Examples:

Left Sidebar - Fields Section:

Enter your fields (one per line):

Tips:

Left Sidebar:

Set your batch configuration using the sliders:

Why smaller batches?

Left Sidebar:

Configure the Cortex LLM settings:

Left Sidebar:

Keep "High-Performance Mode" checked for best results!

Left Sidebar:

Check the following options:

1. Click "Generate Default Prompts" → Review/edit prompts if needed
2. Click "🎲 Generate Synthetic Data" → Wait ~2-3 minutes
3. Watch progress: Batch 1/10... 10/10

Expected Output:

Verification Steps:

1. Go to Snowsight → Data → Databases → AI_FRAMEWORK_DB → BRONZE_LAYER
2. Find your table (e.g., GENERATED_DATA)
3. Click to view:
   • Should see 10 rows (one per batch)
   • Each row has MESSAGES column with JSON array
   • Check META columns for generation metadata

Data Quality Check:

Expected: 10 batches, 100 total records

Bottom of Main Panel:

Enter a configuration name and click 💾 Save Configuration:

Save your configuration to reuse later with different batch sizes or models!

Start small: Test with 10 records × 10 batches first
Use mistral-large2: Best accuracy across all scenarios
Name tables descriptively: Include company/topic in table name
Save configurations: Reuse settings for consistent results
Check data quality: Verify first batch before generating more
Use appropriate temperature: Low for factual, high for creative

For All Personas: → Continue to Page 7 (App 2 - Structured Tables) to transform your data from BRONZE_LAYER to SILVER_LAYER

Your data is now in raw JSON format. App 2 will clean and structure it into proper database columns!

Purpose: Transform raw JSON data into clean, structured database tables
Dependencies: Requires data from App 1
Output: Analytics-ready data in SILVER_LAYER tables

Navigate to Projects → Streamlit → STRUCTURED_TABLES

Main Panel - Left Column:

Select source table with synthetic data from the dropdown (e.g., GENERATED_DATA).

The dropdown shows all tables from BRONZE_LAYER that contain a MESSAGES column (generated by Synthetic Data Generator).

Main Panel - Right Column:

Enter name for structured table (e.g., GENERATED_DATA_STRUCTURED).

The app auto-fills this by adding _STRUCTURED to your source table name. You can customize it if needed.

Main Panel - Filter Section:

Select the company and topic you used when generating data in Step 1 from the dropdowns (e.g., Acme Corp and Customer Orders).

These dropdowns populate automatically from your source table's metadata (_meta_company_name and _meta_topic columns).

Auto-generated after selection:

What to look for:

Sample of Cleaned Data section:

This shows your raw BRONZE_LAYER data with JSON arrays in the MESSAGES column.

Auto-detected fields:

The app automatically detects field names from your JSON structure and shows how they'll appear as SQL column names (uppercase).

Verify all your expected fields are detected!

Bottom Section:

Configuration name:

1. Optional: Edit the configuration name if you want to save settings
2. Click "🔄 Transform Data" button

Progress indicator:

This process:

Expected time: 30 seconds to 2 minutes depending on data volume

Expected Output:

What happened:

Verification Steps:

1. Go to Snowsight → Data → Databases → AI_FRAMEWORK_DB → SILVER_LAYER
2. Find your table (e.g., GENERATED_DATA_STRUCTURED)
3. Click to view data
4. Verify:
   • Row count matches expected (e.g., 100 individual records)
   • All columns are present
   • Data looks clean and realistic

Data Quality Check:

If you clicked "💾 Save Configuration" before transforming, your settings are saved for reuse:

Load it next time from the configuration dropdown!

1. Cleans LLM Artifacts:

2. Flattens JSON Arrays:

3. Creates Proper SQL Table:

After transformation, your data is ready for:

Structured tables work with:

Export structured data via:

For Persona 1 (Full-Stack Developer): → Continue to Page 10 (YAML Wizard) to create dashboard configurations

For Persona 2 (SQL Demo Creator): → Continue to Page 8 (SQL to YAML Converter) to create demo flows

For Persona 3 (Data Preparation): → Export your data from SILVER_LAYER or continue to other apps

For All Personas: Your data is now in clean, structured format in SILVER_LAYER - ready for analytics, dashboards, demos, or export!

Purpose: Convert SQL queries into interactive demo configurations for Snow Demo
Dependencies: Requires tables from App 1 or 2
Output: YAML files for FRAMEWORK_YAML_STAGE

Navigate to Projects → Streamlit → SQL_TO_YAML_CONVERTER

Main Panel - Input SQL Worksheet Section:

Select Paste SQL to enter your queries directly, or Upload File to upload a .sql or .txt file.

SQL Input Text Area:

Replace the placeholder SQL with your actual queries from the structured tables you created:

Tips:

Demo Metadata Section (Two Columns):

Left Column:

Topic:

Sub-topic:

Tertiary Topic:

Title:

Right Column:

Logo URL: (optional - leave blank)

Owner:

Database: (leave blank to auto-detect)

Schema: (leave blank to auto-detect)

Overview Description:

Tips:

Expandable Advanced Options Section:

Default settings work for most cases. Only change if you have specific requirements.

Bottom of Input Section:

Click the blue [Parse SQL Worksheet] button

What happens:

1. App analyzes your SQL queries
2. Detects Cortex AI functions automatically
3. Suggests visualizations based on query patterns:
   • GROUP BY → Bar Chart
   • SELECT * → Table
   • Cortex functions → Interactive AI steps
4. Comments out unsupported commands (USE statements)
5. Generates YAML configuration

Processing time: ~5-10 seconds

Results Section - Tab 1 (Summary):

This shows what the app detected in your SQL and how it will be presented in Snow Demo.

Results Section - Tab 2 (Parsed Blocks):

Verify all your steps are correctly parsed and visualization types make sense.

Results Section - Tab 3 (Generated YAML):

Shows the complete YAML configuration that will be used by Snow Demo. This includes:

You don't need to edit this manually - it's automatically generated!

Results Section - Tab 4 (Download & Export):

Configuration Name:

Option 1: Save to Database (Recommended)

Option 2: Download YAML File

SQL Analysis:

Visualization Suggestions:

Interactive AI Steps:

YAML Generation:

Write clear SQL comments: Use -- Step X: format for step detection
Include Cortex AI: Add CORTEX.COMPLETE() for interactive demos
Mix query types: Combine SELECT, GROUP BY, and AI functions
Use descriptive metadata: Clear titles and topics help viewers understand
Test queries first: Run SQL in worksheet before converting

For Persona 2 (SQL Demo Creator):

1. Upload your YAML to Snowflake Stage (see upload instructions in Snow Demo section below)
2. Continue to Page 9 (Snow Demo) to run your interactive presentation

Your SQL queries are now a professional, interactive demo ready for presentations!

Purpose: Run interactive SQL-driven presentations with live visualizations
Dependencies: Requires YAML configs from App 3 (uploaded to FRAMEWORK_YAML_STAGE)
Output: Live demo orchestration with charts and AI experimentation

Before using Snow Demo, upload your YAML file to Snowflake:

1. Navigate to Data → Databases → AI_FRAMEWORK_DB → CONFIGS → Stages → FRAMEWORK_YAML_STAGE
2. Click "+ Files" button
3. Select your downloaded YAML file
4. In path field, enter: /analytics/ (or choose: sales_demo, customer_insights, etc.)
5. Click "Upload"

Navigate to Projects → Streamlit → SNOW_DEMO

Left Sidebar: Select the project directory where you uploaded your YAML file (e.g., analytics)

Left Sidebar: Select your YAML configuration file from the dropdown

Left Sidebar: Review the auto-displayed demo metadata, then click [Run Demo]

Main Panel: Each SQL step appears as a section with:

Tips: Change Display Options dropdown to switch visualizations on-the-fly

If your SQL includes SNOWFLAKE.CORTEX.COMPLETE() calls, you'll see an interactive panel where you can:

Prepare ahead: Test demo flow before presentations
Use talk tracks: Add presenter notes in YAML for guidance
Practice transitions: Know when to switch visualizations
Engage audience: Ask for prompt suggestions during AI steps
Keep queries fast: Use LIMIT clause for demo data

For Persona 2 (SQL Demo Creator):

Your demo is complete! You can:

Return to Page 5 to explore other workflows or continue to Page 12 for cleanup instructions.

Purpose: Create dashboard configurations through guided interface
Dependencies: Requires tables from App 1 or 2
Output: YAML files for VISUALIZATION_YAML_STAGE

Navigate to Projects → Streamlit → YAML_WIZARD

Main Panel - Top Section:

Schema Selection Guide:

Select your structured table from the previous steps.

Configure Dimensions, Metrics, Time Column Section:

Dimensions (Left Column):

Check 2-5 key categorical fields you want to analyze.

Time Column (Right Column):

Select your date/timestamp field for time-series analysis.

Metrics (Below Columns):

Check 3-7 key metrics you want to calculate. The app automatically creates aggregation functions.

Tips:

Click the "Dimensions" tab

For each dimension, you can customize:

What to customize:

IMPORTANT: After editing, click "Apply All Dimension Changes" button at the bottom!

Click the "Metrics" tab

For each metric, you can customize:

What to customize:

IMPORTANT: After editing, click "Apply All Metric Changes" button at the bottom!

Click the "Generate" tab, then enter:

App Name:

Description:

YAML Filename:

Click "Generate Customized YAML" → Generates 8 tabs (Overview, Product/Category, VS, Top N, Self Service, Search, AI Assistant, Raw Data)

Click "Download YAML" button

Optional: Click "Save to AI_FRAMEWORK_DB.CONFIGS" to save your customizations for later editing

Upload your YAML file to Snowflake:

1. Navigate to Data → Databases → AI_FRAMEWORK_DB → CONFIGS → Stages → VISUALIZATION_YAML_STAGE
2. Click "+ Files" button
3. Select your downloaded YAML file
4. In path field, enter: /customer_orders/ (or your project name)
5. Click "Upload"

What You Created:

Why Two Saves?:

These messages are NORMAL for first-time use:

Ignore this - Search services are advanced/optional

Ignore this - Normal until you save your first config

Start simple: Pick 2-3 dimensions and 3-5 metrics for first try
Use clear labels: "Product Category" is better than "PRODUCT_CATEGORY"
Format metrics: Use currency for money, percent for rates
Save your work: Both download AND save to database
Test in Snow Viz: Verify dashboard works as expected

For Persona 1 (Full-Stack Developer):

You now have a dashboard configuration file! Next steps:

1. Upload YAML to stage (instructions above)
2. Continue to Page 11 (Snow Viz) to view your interactive dashboard

Your data is now ready for visual analytics with 8 interactive dashboard tabs!

Purpose: Render advanced interactive dashboards from YAML configurations
Dependencies: Requires YAML configs from App 5 (uploaded to VISUALIZATION_YAML_STAGE)
Output: Multi-tab analytics dashboards with AI integration

Navigate to Projects → Streamlit → SNOW_VIZ

Left Sidebar - Configuration Source:

Select Stage (recommended - loads from VISUALIZATION_YAML_STAGE)

Left Sidebar - After selecting Stage:

1. Select Project: Choose the directory where you uploaded your YAML
2. Select YAML File: Choose your dashboard configuration file

The dashboard will automatically load.

Left Sidebar - Navigation Section:

Each tab provides different analytical views of your data.

Main Panel - Overview Tab:

Time Controls (Top):

Key Metrics Cards:

Metric cards are interactive - click to select which metric to visualize below.

Visualizations:

Left Side: Time Series Chart

Right Side: Ranked Grid

Dimension Analysis:

Shows detailed breakdown by selected dimension with:

Side-by-Side Comparison:

Comparison Table:

Perfect for comparing products, customers, or any dimension values.

Leaderboard Analysis:

Shows ranked list with:

Custom Analysis:

Build custom reports by selecting any combination of dimensions and metrics.

Results show in interactive data table with:

Powered by Cortex Analyst:

Type your question in natural language:

Click [Ask Analyst] → Select view option (Grid, Bar, or Line chart)

AI Narrative (Optional): Generate AI analysis by selecting a model, adjusting temperature, and clicking [Generate Analysis]

Example Questions:

Semantic Search (if configured):

Enter your search query:

Click [Search] → Shows relevant records based on semantic similarity

Note: Requires Cortex Search service to be configured. If not set up, this tab will show a setup message.

Shows complete dataset in table format with sortable columns and CSV export option.

Use this tab to:

Interactive Elements:

Time Controls:

AI Integration:

Explore systematically: Start with Overview, then drill into specific tabs
Use AI Assistant: Natural language queries are powerful and intuitive
Compare entities: VS tab helps identify top performers
Export insights: Share findings via CSV export
Adjust time windows: Find the right time range for your analysis

For Persona 1 (Full-Stack Developer):

Your complete analytics pipeline is built! You've created:

Share your dashboard with business users and stakeholders!

For Persona 4 (Dashboard Consumer):

You now have an interactive analytics dashboard! You can:

Return to Page 5 to explore other workflows or continue to Page 12 for cleanup instructions.

When you're ready to remove all the resources created during this quickstart:

1. Open the setup.sql script
2. Scroll to the bottom to find the "TEARDOWN SCRIPT" section
3. Uncomment the teardown statements
4. Run the freshly uncommented script to remove all databases, warehouses, roles, and objects

This will clean up all framework components while preserving any other work in your Snowflake account.

Congratulations! You've successfully built the complete Cortex AI Demo Framework using Snowflake Cortex AI!