Lighthouse by Tulapi.ai — Setup & Documentation Guide

1. Overview

Lighthouse is a Snowflake Native Application that turns your SNOWFLAKE.ACCOUNT_USAGE metadata into an operational observability dashboard. It surfaces warehouse credit consumption, query performance, storage footprint, and AI-assisted optimization recommendations — all without any data ever leaving your Snowflake account.

Lighthouse is a 100% in-account application, delivered as a Snowflake Native App with a Streamlit-in-Snowflake (SiS) UI.

2. Architecture

Lighthouse consists of three components running securely inside your account:

Component	Location	Purpose
Setup script	`setup.sql`	Runs on install/upgrade. Creates schemas and application roles.
Data mart	`core` schema	~30 pre-aggregated tables populated from `ACCOUNT_USAGE`.
Streamlit app	`ui.lighthouse`	The Python/Streamlit UI rendered in Snowsight.

Data Flow

Metadata Source: SNOWFLAKE.ACCOUNT_USAGE.*
Ingestion: Read via IMPORTED PRIVILEGES into the core.* mart tables.
Visualization: ui.lighthouse reads the mart via a Snowpark session.

3. Prerequisites

To install Lighthouse, the user must hold ACCOUNTADMIN or a role with permissions to:

Install Native Apps from the Marketplace.
Grant IMPORTED PRIVILEGES ON SNOWFLAKE DB to an application.
Access a warehouse (XS is sufficient) for initial data population.