Databricks Forge is a community-maintained Databricks Field Engineering accelerator. It is not an official Databricks product and is not covered by Databricks Support. See Support and Notice.
Turn Unity Catalog metadata into a prioritized, financially grounded backlog of data and AI opportunities, then activate recommendations as Genie Spaces, AI/BI dashboards, metric views, SQL notebooks, and catalog documentation.
Forge runs as a Databricks App. Metadata is the default input. Optional row sampling is explicit, configurable, and never persisted.
- A Databricks workspace with Unity Catalog
- A running Serverless or Pro SQL warehouse
- Databricks CLI 0.205 or later, authenticated to the target workspace
- Node.js 20.9 or later and Python 3
- Permission to create Databricks Apps and Lakebase resources
git clone https://github.com/databricks-solutions/databricks-forge.git
cd databricks-forge
databricks auth login --host https://your-workspace.cloud.databricks.com
./deploy.shThe script selects a warehouse, checks model availability, provisions an isolated Lakebase project, binds resources, configures user authorization, and deploys the app. It prints the app URL when deployment completes.
Continue with the Quick Start to verify health, complete first-run setup, update Forge, or remove it safely.
- Use-case discovery: Generate, score, deduplicate, and rank analytics and AI opportunities from catalog metadata.
- Business value: Estimate financial impact, delivery phases, dependencies, stakeholders, and executive recommendations.
- Genie Studio: Create, assess, benchmark, and improve Databricks Genie Spaces.
- AI/BI activation: Generate validated dashboard, metric-view, SQL, and notebook recommendations.
- Estate intelligence: Analyze domains, lineage, health, maturity, governance gaps, and data-product opportunities.
- Catalog documentation: Generate and review table and column comments with bulk apply and undo.
- Ask Forge: Explore findings through a retrieval-grounded conversational assistant.
- Executive exports: Produce Excel, PowerPoint, PDF, and board-ready portfolio artifacts.
See the User Guide for the complete product walkthrough.
Forge reads schema metadata such as catalog, schema, table, and column names, types, and comments. When an operator explicitly enables data sampling, Forge reads a limited number of rows to improve SQL accuracy; sampled rows are held in memory and are not persisted.
Interactive Databricks operations use the signed-in user's on-behalf-of token where required. Background operations use the Databricks App service principal and explicit resource bindings. See Security Architecture for data flows, permissions, isolation, and threat mitigations.
| Layer | Technology |
|---|---|
| Application | Next.js 16, React 19, TypeScript |
| UI | Tailwind CSS 4, shadcn/ui |
| SQL | Databricks SQL Statement Execution API |
| AI | Databricks Model Serving |
| Persistence | Lakebase Autoscaling with Prisma |
| Deployment | Databricks Apps |
- Quick Start
- Deployment and Operations
- User Guide
- Architecture
- Security Architecture
- Engine Reference
- Support
- Security Policy
- Contributing
- Changelog
Use Node.js 20.9 or later:
npm ci
databricks auth login --host https://your-workspace.cloud.databricks.com
bash .deploy_local.sh
npm run devThe setup script provisions local Lakebase access, selects a warehouse, and
writes .env.local. It uses the Databricks CLI OAuth session; no personal
access token is required. See
Deployment and Operations for details.
Before opening a pull request:
npm run check:public
npm run format:check
npm run lint
npm run typecheck
npm test
npm run test:quality
npm run buildCommunity support is best effort through GitHub Discussions and GitHub Issues. Do not post vulnerabilities publicly; use private vulnerability reporting.
Copyright 2024-2026 Databricks, Inc. Forge is provided under the Databricks License, without warranty. See NOTICE.