Transforming Manufacturing Intelligence with Snowflake Intelligence: A Complete Solution

🏭 Snowflake Intelligence Demo - Vehicle Manufacturing

📋 Table of Contents

Overview
Key Features
Quick Start Guide
Data Types Supported
Intelligence Agents
File Structure
Usage Examples
Prerequisites
Setup Instructions
Verification
Demo Guide
Troubleshooting
Contributing
License

🎯 Overview

This repository contains a comprehensive Snowflake Intelligence demonstration specifically designed for Vehicle Manufacturing use cases. The demo showcases how Snowflake Intel…

🏭 Snowflake Intelligence Demo - Vehicle Manufacturing

🎯 Overview

This repository contains a comprehensive Snowflake Intelligence demonstration specifically designed for Vehicle Manufacturing use cases. The demo showcases how Snowflake Intelligence can break down data silos, provide real-time visibility, and enable scalable AI adoption across manufacturing operations.

What is Snowflake Intelligence?

Snowflake Intelligence is Snowflake’s integrated AI platform that enables organizations to:

🤖 Build AI Agents - Create specialized AI agents that understand your business domain
📊 Semantic Views - Transform raw data into business-friendly semantic models
🔍 Cortex Search - Enable semantic search over unstructured text data
💬 Natural Language Queries - Ask questions in plain English, get intelligent answers

Demo Scope

This demo demonstrates:

✅ Structured Data - Traditional relational tables (supply chain, production, inventory)
✅ Semi-Structured Data - JSON/VARIANT columns (connected vehicles, IoT sensors, product configurations)
✅ Unstructured Data - Free-form text (maintenance logs, quality reports, engineering docs)
✅ Cross-Type Analysis - Correlating data across all three data types
✅ Natural Language Processing - Semantic search and intelligent document analysis
✅ Data Visualization - Tables, charts, and graphs generated from queries

✨ Key Features

🏗️ Unified Architecture

Single Semantic View - manufacturing_operations consolidates all structured and semi-structured data
Unified Cortex Search - manufacturing_documents_search enables semantic search across all unstructured documents
Intelligent Agents - Four specialized agents for different manufacturing domains

📊 Comprehensive Data Coverage

50+ Records per Table - Realistic vehicle manufacturing data (Toyota Camry, Honda Accord, Ford F-150, Tesla Model 3, BMW 3 Series, etc.)
Real-World Scenarios - Data reflects actual manufacturing operations and challenges
Multiple Data Formats - Structured, semi-structured (JSON), and unstructured (text)

🤖 Specialized Intelligence Agents

Supply Chain Agent - Supplier risk, inventory management, logistics
Production Agent - Quality management, efficiency optimization, predictive maintenance
Connected Products Agent - Telematics, fleet management, vehicle diagnostics
Operations Agent - Cross-functional insights and strategic recommendations

🔍 Advanced Capabilities

Natural Language Queries - Ask questions in plain English
Semantic Search - Find relevant information across unstructured documents
Data Visualization - Automatic generation of tables, charts, and graphs
Cross-Type Analysis - Correlate data across structured, semi-structured, and unstructured sources

Intelligence Agents Location

All agents are created in the SNOWFLAKE_INTELLIGENCE.AGENTS schema:

SNOWFLAKE_INTELLIGENCE.AGENTS.supply_chain_agent
SNOWFLAKE_INTELLIGENCE.AGENTS.production_agent
SNOWFLAKE_INTELLIGENCE.AGENTS.connected_products_agent
SNOWFLAKE_INTELLIGENCE.AGENTS.manufacturing_operations_agent

🚀 Quick Start Guide

Prerequisites

✅ Snowflake account with Intelligence enabled
✅ Role with ACCOUNTADMIN or equivalent privileges
✅ SNOWFLAKE.CORTEX_USER database role granted
✅ Access to create databases, schemas, tables, and agents

Step-by-Step Setup

Step 1: Setup Infrastructure

-- Run the setup script to create databases, schemas, tables, and warehouse
@01_setup_intelligence.sql

What this does:

Creates databases: MANUFACTURING_DEMO and SNOWFLAKE_INTELLIGENCE
Creates schemas: DATA, SEMANTIC (in MANUFACTURING_DEMO), and AGENTS (in SNOWFLAKE_INTELLIGENCE)
Creates all required table structures:
Structured tables: supply_chain, production, inventory
Semi-structured tables: connected_products, iot_sensors, supplier_documents, product_configurations (with VARIANT columns for JSON data)
Unstructured tables: maintenance_logs, quality_reports, supplier_communications, engineering_docs, incident_reports (with TEXT columns)
Creates warehouse: INTEL_WH (X-SMALL, auto-suspend 60s)
Inserts minimal sample data (3-5 records per table) for initial setup
Grants USAGE privileges on databases, schemas, and warehouse to PUBLIC role
Grants SELECT privileges on all tables to PUBLIC role

Step 2: Load Vehicle Manufacturing Data

-- Load comprehensive vehicle manufacturing data (50+ records per table)
@01b_insert_vehicle_data.sql

What this does:

Inserts 50+ realistic vehicle manufacturing records per table
Structured data: Supply chain orders (Bosch, Continental, Magna, ZF, etc.), production batches, inventory levels
Semi-structured data: Connected vehicle telemetry (VEH-001 through VEH-050+), IoT sensor readings, supplier documents (contracts, invoices, certificates), product configurations (PROD-001, PROD-002, etc.)
Unstructured data: Maintenance logs, quality reports, supplier communications (emails, meeting notes, contracts), engineering docs (design specs, test plans, change requests), incident reports
Uses realistic vehicle data: Toyota Camry, Honda Accord, Ford F-150, Tesla Model 3, BMW 3 Series, etc.
Includes realistic supplier names, VINs, locations, telemetry data, and manufacturing scenarios

Step 3: Create Semantic Views & Cortex Search

-- Create unified semantic view for structured/semi-structured data
@02_create_semantic_views.sql

-- Create unified Cortex Search service for unstructured data
@02a_create_cortex_search.sql

What this does:

Semantic View (02_create_semantic_views.sql): Creates unified manufacturing_operations semantic view in MANUFACTURING_DEMO.SEMANTIC schema

Combines all structured and semi-structured tables with PRIMARY KEY and UNIQUE constraints
Defines relationships between tables (supply chain ↔ inventory, production ↔ quality reports, production ↔ maintenance logs, etc.)
Creates dimensions: supplier names, risk categories (Low/Medium/High), delivery status (On Time/Early/In Progress/Pending), quality ratings (Excellent/Good/Acceptable/Needs Improvement), stock status (Reorder Needed/Low Stock/Adequate), alert status, driver names, etc.
Creates metrics: averages (delivery days, quality scores, sensor values), totals (quantity, cost, energy consumption), counts (orders, batches, telemetry records)
Grants SELECT privilege on semantic view to PUBLIC role

Cortex Search (02a_create_cortex_search.sql): Creates unified manufacturing_documents_search Cortex Search service in MANUFACTURING_DEMO.SEMANTIC schema

Enables change tracking on all unstructured tables (required for Cortex Search)
Creates warehouse: CORTEX_SEARCH_WH (X-SMALL, auto-suspend 60s)
Combines all unstructured text data using UNION ALL: maintenance logs, quality reports, supplier communications, engineering docs, incident reports
Uses composite PRIMARY KEY (document_type, document_id) with attributes for filtering
Configures embedding model: snowflake-arctic-embed-l-v2.0
Sets TARGET_LAG to 1 day for refresh frequency
Grants USAGE privilege on Cortex Search service to PUBLIC role

Step 4: Create Intelligence Agents

-- Create specialized Intelligence agents
@03_create_agents.sql

What this does:

Creates 4 specialized Intelligence agents in SNOWFLAKE_INTELLIGENCE.AGENTS schema:
supply_chain_agent - Supplier risk, inventory, logistics
production_agent - Quality, efficiency, maintenance
connected_products_agent - Telematics, fleet management
manufacturing_operations_agent - Cross-functional insights
Configures each agent with FROM SPECIFICATION syntax (YAML format)
Sets up tools: Analyst1 (cortex_analyst_text_to_sql) for semantic view queries, Search1 (cortex_search) for unstructured data search
Configures tool_resources: semantic_view and cortex_search service references
Sets orchestration model: claude-4-sonnet with budget constraints
Defines agent instructions, sample questions, and profiles
Grants USAGE privileges on all agents to PUBLIC role

You’re all set! After running these 4 steps, you can start using the Intelligence agents in Snowflake UI.

📊 Data Types Supported

Structured Data

Traditional relational tables with fixed schemas:

Table	Description	Records
`supply_chain`	Supplier orders, deliveries, risk scores	50+
`production`	Production batches, quality scores, efficiency metrics	50+
`inventory`	Stock levels, reorder points, warehouse locations	50+

Semi-Structured Data

JSON/VARIANT columns with nested data:

Table	Description	Records
`connected_products`	Vehicle telemetry, sensor data, trip metadata	50+
`iot_sensors`	IoT sensor readings, machine states	50+
`supplier_documents`	Supplier contracts, invoices, compliance docs	50+
`product_configurations`	Vehicle configurations, specifications, BOMs	50+

Unstructured Data

Free-form text stored in TEXT columns:

Table	Description	Records
`maintenance_logs`	Maintenance activities, issues, actions taken	50+
`quality_reports`	Quality inspections, defects, root cause analysis	50+
`supplier_communications`	Emails, meeting notes, supplier discussions	50+
`engineering_docs`	Engineering specifications, design notes, test procedures	50+
`incident_reports`	Safety incidents, equipment failures, investigations	50+

🤖 Intelligence Agents

1. Supply Chain Agent

Location: SNOWFLAKE_INTELLIGENCE.AGENTS.supply_chain_agent

Capabilities:

Supplier risk assessment and management
Inventory optimization and visibility
Demand forecasting and planning
Transportation and logistics analysis
Spend analytics and cost optimization

Example Queries:

“What suppliers have high risk scores? Show me the details and suggest mitigation strategies.”
“Create a bar chart showing total spend by supplier region.”
“Search supplier communications for discussions about delivery delays.”

2. Production Agent

Location: SNOWFLAKE_INTELLIGENCE.AGENTS.production_agent

Capabilities:

Shopfloor visibility and real-time monitoring
Production efficiency and throughput optimization
Product yield and quality management
Predictive maintenance recommendations
Energy consumption and sustainability analysis

Example Queries:

“Show me production efficiency by production line with a line chart showing trends.”
“Extract IoT sensor data and show which machines have alerts.”
“Search maintenance logs for recurring issues and create a summary table.”

3. Connected Products Agent

Location: SNOWFLAKE_INTELLIGENCE.AGENTS.connected_products_agent

Capabilities:

Telematics and sensor data analysis
Predictive maintenance for vehicles/products
Fleet management and optimization
Driver behavior analysis
Electric vehicle optimization

Example Queries:

“Extract sensor data from connected vehicles and show vehicles with alerts in a table.”
“Create a line chart of battery health over time for all vehicles.”
“Find engineering documents related to brake system design.”

4. Manufacturing Operations Agent

Location: SNOWFLAKE_INTELLIGENCE.AGENTS.manufacturing_operations_agent

Capabilities:

End-to-end visibility across all operations
Cross-functional analysis and optimization
Strategic recommendations
Correlation analysis across data types
Comprehensive dashboard generation

Example Queries:

“Provide a comprehensive analysis combining supplier risk, production quality, and vehicle performance with supporting charts.”
“Correlate production batch data with IoT sensor readings and quality reports.”
“Create a unified dashboard showing all manufacturing operations metrics.”

📁 File Structure

KPMG_event/
│
├── 01_setup_intelligence.sql          # Infrastructure setup (Intelligence object, database, schemas, tables)
├── 01b_insert_vehicle_data.sql        # Vehicle manufacturing data insertion (50+ records per table)
├── 02_create_semantic_views.sql       # Unified semantic view creation with relationships
├── 02a_create_cortex_search.sql       # Unified Cortex Search service for unstructured data
├── 03_create_agents.sql               # Intelligence agents creation with tools and configurations
│
├── README.md                           # This file - comprehensive guide
├── DEMO_GUIDE.md                       # Detailed demo guide with sample queries
├── reference.MD                        # Manufacturing use cases reference
│
└── .gitignore                          # Git ignore file

File Descriptions

File	Purpose	Key Components
`01_setup_intelligence.sql`	Infrastructure setup	Creates databases (`MANUFACTURING_DEMO`, `SNOWFLAKE_INTELLIGENCE`), schemas (`DATA`, `SEMANTIC`, `AGENTS`), all table structures (11 tables: 3 structured, 4 semi-structured, 5 unstructured), warehouse (`INTEL_WH`), inserts minimal sample data, grants privileges
`01b_insert_vehicle_data.sql`	Data population	Inserts 50+ realistic vehicle manufacturing records per table covering structured (supply chain orders, production batches, inventory), semi-structured (connected vehicle telemetry, IoT sensors, supplier documents, product configs), and unstructured data (maintenance logs, quality reports, communications, engineering docs, incident reports)
`02_create_semantic_views.sql`	Semantic layer	Creates unified `manufacturing_operations` semantic view in `MANUFACTURING_DEMO.SEMANTIC` with table relationships, business-friendly dimensions (risk categories, quality ratings, stock status, alert status, etc.), and aggregate metrics (averages, totals, counts) across all structured and semi-structured tables
`02a_create_cortex_search.sql`	Search service	Enables change tracking on unstructured tables, creates `CORTEX_SEARCH_WH` warehouse, creates unified `manufacturing_documents_search` Cortex Search service in `MANUFACTURING_DEMO.SEMANTIC` combining all unstructured text data (maintenance logs, quality reports, supplier communications, engineering docs, incident reports) with embedding model `snowflake-arctic-embed-l-v2.0`
`03_create_agents.sql`	AI agents	Creates 4 specialized Intelligence agents in `SNOWFLAKE_INTELLIGENCE.AGENTS` schema (`supply_chain_agent`, `production_agent`, `connected_products_agent`, `manufacturing_operations_agent`) using FROM SPECIFICATION with YAML format, configures tools (Analyst1 for semantic view, Search1 for Cortex Search), tool_resources, orchestration model `claude-4-sonnet`, instructions, and grants USAGE privileges
`DEMO_GUIDE.md`	Demo guide	Comprehensive guide with 50+ simple natural language sample queries organized by agent and data type

💡 Usage Examples

Using Agents in Snowflake UI

Open Snowflake UI → Navigate to Intelligence Chat 1.

Select an Agent:

SNOWFLAKE_INTELLIGENCE.AGENTS.supply_chain_agent
SNOWFLAKE_INTELLIGENCE.AGENTS.production_agent
SNOWFLAKE_INTELLIGENCE.AGENTS.connected_products_agent
SNOWFLAKE_INTELLIGENCE.AGENTS.manufacturing_operations_agent

Ask Questions in Natural Language:

Structured Data Query:

Create a table showing supplier performance metrics including total orders,
average delivery time, total spend, and risk scores. Sort by risk score descending.

Semi-Structured Data Query:

Extract and analyze sensor data from connected vehicles. Show me a table with
vehicle ID, sensor types, sensor values, thresholds, and alert status for
vehicles with sensor values exceeding thresholds.

Unstructured Data Query (NLP):

Search maintenance logs for any mentions of bearing failures or bearing
replacements. What machines are affected and what actions were taken?

Combined Data Type Query:

Correlate production batch data with IoT sensor readings and quality reports.
Show me which production batches had sensor alerts and how that affected
quality scores. Include charts showing the correlation.

Example Outputs

Tables - Structured data tables with sorting, filtering, and aggregation
Charts - Bar charts, line charts, pie charts, histograms, scatter plots
NLP Results - Relevant document excerpts with semantic search
Combined Analysis - Cross-type correlations and insights

🔧 Prerequisites

Required Privileges

ACCOUNTADMIN role or equivalent
CREATE INTELLIGENCE privilege
CREATE DATABASE privilege
CREATE SCHEMA privilege
CREATE TABLE privilege
CREATE SEMANTIC VIEW privilege
CREATE CORTEX SEARCH SERVICE privilege
CREATE AGENT privilege

Required Database Roles

SNOWFLAKE.CORTEX_USER - Required for Cortex Search and Intelligence features

Grant Required Roles

-- Grant CORTEX_USER role to your role
GRANT DATABASE ROLE SNOWFLAKE.CORTEX_USER TO ROLE <your_role>;

📝 Setup Instructions

Complete Setup Process

Connect to Snowflake using your preferred SQL client or Snowsight 1.

Run Setup Scripts in Order:

-- Step 1: Infrastructure
@01_setup_intelligence.sql

-- Step 2: Data
@01b_insert_vehicle_data.sql

-- Step 3: Semantic Layer & Search
@02_create_semantic_views.sql
@02a_create_cortex_search.sql

-- Step 4: Agents
@03_create_agents.sql

Wait for Cortex Search Indexing:

Cortex Search service indexing may take a few minutes
Check status: DESCRIBE CORTEX SEARCH SERVICE MANUFACTURING_DEMO.SEMANTIC.manufacturing_documents_search;
Wait until INDEXING_STATE shows RUNNING

Start Using Agents:

Navigate to Intelligence Chat in Snowflake UI
Select an agent from SNOWFLAKE_INTELLIGENCE.AGENTS schema
Start asking questions in natural language!

✅ Verification

Manual Verification Steps

After running all setup scripts, verify your setup with these commands:

Check Database and Schemas:

SHOW DATABASES LIKE 'MANUFACTURING_DEMO';
SHOW SCHEMAS IN DATABASE MANUFACTURING_DEMO;
SHOW DATABASES LIKE 'SNOWFLAKE_INTELLIGENCE';
SHOW SCHEMAS IN DATABASE SNOWFLAKE_INTELLIGENCE;

Check Tables:

SHOW TABLES IN SCHEMA MANUFACTURING_DEMO.DATA;
SELECT COUNT(*) FROM MANUFACTURING_DEMO.DATA.supply_chain;
SELECT COUNT(*) FROM MANUFACTURING_DEMO.DATA.production;
SELECT COUNT(*) FROM MANUFACTURING_DEMO.DATA.inventory;

Check Semantic View:

SHOW SEMANTIC VIEWS IN SCHEMA MANUFACTURING_DEMO.SEMANTIC;

Check Cortex Search Service:

DESCRIBE CORTEX SEARCH SERVICE MANUFACTURING_DEMO.SEMANTIC.manufacturing_documents_search;
-- Verify INDEXING_STATE shows 'RUNNING'

Check Agents:

SHOW AGENTS IN SCHEMA SNOWFLAKE_INTELLIGENCE.AGENTS;

Expected Results

✅ Database MANUFACTURING_DEMO with schemas DATA and SEMANTIC
✅ Database SNOWFLAKE_INTELLIGENCE with schema AGENTS
✅ All 11 tables created (3 structured, 4 semi-structured, 5 unstructured)
✅ All tables populated with 50+ records each (after running 01b_insert_vehicle_data.sql)
✅ Semantic view manufacturing_operations created in MANUFACTURING_DEMO.SEMANTIC with relationships, dimensions, and metrics
✅ Cortex Search service manufacturing_documents_search created in MANUFACTURING_DEMO.SEMANTIC and indexed (INDEXING_STATE = 'RUNNING')
✅ 4 agents created in SNOWFLAKE_INTELLIGENCE.AGENTS schema with proper tool configurations (Analyst1, Search1)

📖 Demo Guide

For comprehensive demo guidance, sample queries, and presentation tips, see DEMO_GUIDE.md.

The demo guide includes:

✅ 50+ sample queries organized by agent
✅ Queries for structured, semi-structured, and unstructured data
✅ Examples generating tables, charts, and graphs
✅ NLP search examples
✅ Cross-type analysis queries
✅ Demo flow recommendations
✅ Troubleshooting guide

🐛 Troubleshooting

Common Issues

Issue: Agents don’t have enough context

Solution:

Ensure semantic view is created: SHOW SEMANTIC VIEWS IN SCHEMA MANUFACTURING_DEMO.SEMANTIC;
Ensure Cortex Search service is indexed: DESCRIBE CORTEX SEARCH SERVICE manufacturing_documents_search;
Check agent tool_resources configuration in 03_create_agents.sql
Verify agents have USAGE privileges

Issue: Queries return generic responses

Solution:

Make queries more specific to Vehicle Manufacturing domain
Reference specific data elements (suppliers, parts, production lines, vehicle models)
Ask for analysis rather than just data retrieval
Specify if you want tables or charts

Issue: Performance is slow

Solution:

Ensure warehouses are running: SHOW WAREHOUSES;
Check Cortex Search service indexing state
Verify semantic view is optimized
Consider using smaller datasets for demo

Issue: Cortex Search not working

Solution:

Verify CORTEX_SEARCH_WH warehouse exists and is running
Check INDEXING_STATE of the service
Ensure SNOWFLAKE.CORTEX_USER role is granted
Wait for initial indexing to complete (may take several minutes)

🤝 Contributing

This is a demonstration repository. For questions or suggestions:

Review the code and documentation
Test the setup in your Snowflake environment
Provide feedback or suggestions

📄 License

This demonstration is provided as-is for educational and demonstration purposes.

🏭 Snowflake Intelligence Demo - Vehicle Manufacturing

📋 Table of Contents

🎯 Overview

🏭 Snowflake Intelligence Demo - Vehicle Manufacturing

📋 Table of Contents

🎯 Overview

What is Snowflake Intelligence?

Demo Scope

✨ Key Features

🏗️ Unified Architecture

📊 Comprehensive Data Coverage

🤖 Specialized Intelligence Agents

🔍 Advanced Capabilities

Intelligence Agents Location

🚀 Quick Start Guide

Prerequisites

Step-by-Step Setup

Step 1: Setup Infrastructure

Step 2: Load Vehicle Manufacturing Data

Step 3: Create Semantic Views & Cortex Search

Step 4: Create Intelligence Agents

📊 Data Types Supported

Structured Data

Semi-Structured Data

Unstructured Data

🤖 Intelligence Agents

1. Supply Chain Agent

2. Production Agent

3. Connected Products Agent

4. Manufacturing Operations Agent

📁 File Structure

File Descriptions

💡 Usage Examples

Using Agents in Snowflake UI

Example Outputs

🔧 Prerequisites

Required Privileges

Required Database Roles

Grant Required Roles

📝 Setup Instructions

Complete Setup Process

✅ Verification

Manual Verification Steps

Expected Results

📖 Demo Guide

🐛 Troubleshooting

Common Issues

Issue: Agents don’t have enough context

Issue: Queries return generic responses

Issue: Performance is slow

Issue: Cortex Search not working

🤝 Contributing

📄 License

Similar Posts