Connect and discover
The Electron main process used an existing Snowflake CLI profile, discovered the Chinook database and schema, and listed supported tables without exposing credentials to the renderer.
ERD Tool / Snowflake Data Modeling
A Snowflake modeler built by an autonomous AI employee.
Reverse engineer a live Snowflake schema into an editable ERD, arrange the model, save it as a native desktop project, reopen it offline, and generate deterministic Snowflake DDL.
11live Snowflake tables imported
64columns modeled
11relationships preserved
115automated checks passing
Professional ElevenLabs narration · burned captions · live packaged-app proof
Download captions
The acceptance run
Not a mock-up. A packaged Mac app connected to live Snowflake.
Reverse engineer
The app imported 11 tables, 64 columns, primary keys, and 11 relationships, then converted them into drawDB’s editable diagram model.
Save and reopen
The native ERD project saved without passwords, keys, sessions, or tokens and reopened in a fresh packaged process with the model intact.
Forward engineer
The reopened model generated deterministic Snowflake DDL containing all 11 table definitions and all 11 informational foreign keys.
Current Mac MVP
A real desktop modeling workflow.
Live Snowflake reverse engineering
Choose a CLI profile or bounded manual connection, then select a database, schema, and table set.
Editable logical and physical model
Work with tables, columns, data types, keys, relationships, comments, and diagram layout.
Automatic layout
Use ELK-based layout to turn imported metadata into a navigable starting point.
Native project files
Save credential-free ERD documents and reopen them offline in a fresh desktop session.
Snowflake DDL
Generate stable database, schema, table, primary-key, and foreign-key DDL from the canonical model.
Desktop security boundary
Node access stays out of the renderer; connections and filesystem operations remain behind narrow IPC contracts.
The autonomous employee story
I assigned the mission. The employee delivered working software.
The first proof reverse engineered SQLite and forward engineered the structures into Snowflake. The subsequent mission turned the open-source drawDB web app into an installable Electron application with Snowflake modeling support. The AI employee worked the backlog, classified failures across product and orchestration layers, repaired root causes, and reran the stack until the Mac acceptance path completed cleanly.
Release boundary
Mac proof complete. Distribution work remains.
The current Apple Silicon package is unsigned and intended for private preview. Signing, notarization, Linux validation, and Windows validation are future release work. No public installer is offered yet.
Interested in the product or the employee that built it?
Let’s compare notes.
Discuss a Snowflake modeling use case, request a private Mac preview, or talk about autonomous AI employees and the orchestration infrastructure behind this build.
Evidence boundary: The video and acceptance figures show Lee Harrington’s live Snowflake Chinook schema running through the packaged Apple Silicon application. They are product-development proof, not a client deployment or measured customer outcome.