Snowflake Semantic Tools (SST)

Build, validate, and deploy Snowflake Semantic Views from your dbt project.

What is SST?

SST helps you build Snowflake Semantic Views—a standardized semantic layer that lives in Snowflake and powers AI and BI tools—all from within your dbt projects.

Why Semantic Views matter:

• Cortex Analyst & Agents — Semantic views give Snowflake's AI the context it needs to accurately answer natural language questions about your data
• BI Tools — Sigma, Tableau, and other tools can consume semantic views for consistent metrics and definitions across your organization
• Single source of truth — Define metrics, relationships, and business logic once in Snowflake, use everywhere

What SST does:

• Define semantics as code — Metrics, relationships, filters, verified queries as YAML in your dbt project
• Deploy to Snowflake — Generate native SEMANTIC VIEW objects from your definitions
• Enrich automatically — Pull column types, samples, and metadata from Snowflake schemas (including dbt sources)
• Validate before deploy — 50+ validation rules catch errors before they reach Snowflake
• Compile locally — Build a manifest for offline tooling, diffing, and CI/CD
• Diff before deploy — Preview exactly what will change before touching Snowflake
• Incremental deploys — Only regenerate views affected by your changes (--only-modified)
• Version control everything — Your semantic layer lives in git alongside your dbt models

Quick Start

Installation

pip install snowflake-semantic-tools

Setup

cd your-dbt-project
sst init  # Interactive setup wizard

The wizard will:

• Detect your dbt project and profile
• Create sst_config.yml with defaults
• Set up the semantic models directory
• Generate example files

Basic Usage

# Enrich models with metadata from Snowflake
sst enrich --models customers,orders

# Validate your semantic models (offline, no Snowflake needed)
sst validate

# Preview what will change
sst diff

# Deploy to Snowflake (validates first, then generates)
sst generate --all

# Only deploy views affected by recent changes
sst generate --all --only-modified

Commands

Key Features

Semantic Modeling

• Define metrics, relationships, filters, verified queries, and custom instructions as YAML
• Compose semantic views from multiple tables with join relationships
• Template metrics with {{ metric('name') }} references

Enrichment

• Auto-populate YAML with column types, sample values, synonyms, and enums
• Enrich dbt sources (--include-sources, --sources-only, --source raw.orders)
• AI-powered descriptions and synonym generation via Cortex

Validation

• 50+ rules covering references, types, expressions, and structure
• SQL syntax validation against Snowflake
• Schema verification with fuzzy column matching
• Cross-table metric column validation

Deployment

• Incremental deploys with --only-modified (detects YAML + SQL changes)
• Dry-run mode with SQL file output (--dry-run --output-dir)
• Diff before deploy to preview additions, removals, and modifications
• Compile to local manifest for CI/CD pipelines

Documentation

See the docs/ directory for comprehensive documentation:

• Documentation Index — Complete documentation navigation
• Getting Started — Installation and first steps
• CLI Reference — All commands and options
• Semantic Models Guide — Writing metrics and relationships
• Validation Rules — Complete list of all validation checks
• Authentication — Snowflake connection setup
• Configuration Reference — sst_config.yml options

Requirements

• Python 3.10–3.11
• Snowflake account
• dbt project (dbt Core or dbt Cloud CLI)

Development Setup

git clone https://github.com/WhoopInc/snowflake-semantic-tools.git
cd snowflake-semantic-tools

poetry install --with dev
pre-commit install

sst --version
pytest tests/unit/

See CONTRIBUTING.md for detailed development guidelines.

Contributing

We welcome contributions! Please see CONTRIBUTING.md for:

• How to report issues
• Development setup instructions
• Code style guidelines
• Pull request process

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.