[PFR] Add Genie Code support as a workshop coding assistant target

[varunrao]

Product Feature Request: Genie Code Integration

Summary

The Vibe Coding Workshop currently supports Cursor, VS Code + GitHub Copilot, and Claude Code as the AI coding assistant targets where participants paste and run generated prompts. With the launch of Genie Code (March 2026), Databricks now has its own autonomous AI partner purpose-built for data work — deeply integrated with Unity Catalog, notebooks, Lakeflow pipelines, and dashboards.

The workshop should add Genie Code as a first-class supported assistant so participants can run the entire workshop directly inside the Databricks platform without needing an external IDE.

---

1. Current State

The workshop currently assumes participants work in an external IDE (Cursor, VS Code, or a terminal with Claude Code):

Area	Current Behavior
Project setup	Instructs users to clone a Git repo and open it in Cursor or VS Code (SetUpProjectStep.tsx)
Prompt delivery	Generates prompts with one-click copy to clipboard, intended for pasting into an external AI assistant
"How to Apply" guidance	References Cursor/VS Code workflows — "Open File → Open Folder", "press Cmd/Ctrl+O", etc.
AI model config	Step B in setup walks through configuring a Claude model in the external IDE
UI design step	Section tag cursor_copilot_ui_design — hardcoded to Cursor/Copilot workflow
PRD language	"...can be used directly with AI coding assistants like Cursor, GitHub Copilot, or Claude Code" (design_prd.md)
Lakebase setup docs	"A coding assistant (Cursor, Copilot, or Claude) is enabled in that codebase" (db/lakebase/README.md)

No Genie Code path exists today. Participants who want to use Genie Code have no guided workflow, no adapted prompts, and no way to run workshop steps inside Databricks notebooks or the SQL Editor.

---

2. Desired State

Add Genie Code as a supported coding assistant target with a complete workshop experience:

2a. Assistant Selector

Add a UI element (e.g., in the workshop intro or setup step) where participants choose their coding assistant:

• Cursor (existing)
• VS Code + GitHub Copilot (existing)
• Claude Code (existing)
• Genie Code (new) — runs inside Databricks notebooks/workspace

The selected assistant should influence prompt formatting, "How to Apply" instructions, and the project setup flow for the remainder of the workshop.

2b. Genie Code-Specific Setup Flow

When Genie Code is selected, replace the "clone repo + open in IDE" setup with:

• Navigate to your Databricks workspace
• Open a new notebook or import the workshop notebook template
• Verify Genie Code is enabled (chat panel visible)
• Connect to a cluster or SQL warehouse
• Verify Unity Catalog access to the workshop catalog/schema

2c. Adapted Prompt Delivery

For Genie Code participants, prompts should be formatted for the Databricks notebook experience:

• Chat-based prompts: Formatted for the Genie Code chat panel (right-side panel in notebooks)
• Slash command prompts: Where applicable, use /slash command syntax
• Cell-based prompts: For code generation steps, format as notebook cell instructions
• Agent mode prompts: For multi-step tasks (pipeline building, dashboard creation), leverage Genie Code's agentic capabilities

2d. Genie Code-Native Workshop Steps

Some workshop sections map naturally to Genie Code's specialized surfaces:

Workshop Section	Genie Code Surface
Foundation (PRD, project setup)	Notebook chat — generate PRD, scaffold project
Databricks App (UI design)	Notebook agent mode — generate and iterate on app code
Lakebase (database setup)	Notebook chat — create tables, test queries
Lakehouse (medallion architecture)	Lakeflow Pipeline Editor — Genie Code's DE agent can build pipelines directly
Data Intelligence (dashboards, Genie spaces)	Dashboard Editor — Genie Code's dashboard agent can plan and generate dashboards
Refinement (test, iterate)	Notebook — diagnose errors, quick fix, iterate

2e. Output Capture

Participants using Genie Code should still be able to track progress in the workshop app:

• A mechanism to report step completion back to the workshop (e.g., API callback, manual "mark complete", or notebook widget integration)
• Session state saved to Lakebase just like other assistant paths

---

3. Why This Matters

• Zero external tooling: Participants can run the entire workshop inside Databricks — no IDE install, no Git clone, no local setup
• Databricks-native experience: Showcases the platform's own AI capabilities rather than relying on third-party tools
• Genie Code's unique strengths: Deep Unity Catalog integration means Genie Code understands the participant's data landscape (tables, columns, lineage, governance) — prompts can leverage this context automatically
• Agentic capabilities: Genie Code's agent mode in Lakeflow and dashboards can handle multi-step workshop tasks autonomously, which is a differentiated experience vs. copy-paste into Cursor
• Lower barrier to entry: Workshop facilitators don't need to ensure every participant has Cursor/VS Code installed and configured — just a Databricks workspace
• Alignment with product direction: Genie Code is available at no additional cost to all Databricks customers and represents the future of AI-assisted development on the platform

---

4. Proposed Implementation

Phase 1: Assistant Selection UI

• Add an assistant picker to the workshop intro / level selector
• Store selection in session state (Lakebase sessions table)
• Gate setup instructions based on selection

Phase 2: Conditional Prompt Formatting

• Update prompt templates (section_input_prompts in Lakebase, prompts_config.yaml fallback) to include Genie Code variants
• Adjust "How to Apply" instructions per assistant
• Update the cursor_copilot_ui_design section tag to be assistant-agnostic

Phase 3: Genie Code Workshop Notebook

• Create a companion notebook template that participants import into their workspace
• Notebook cells correspond to workshop steps with instructions for using Genie Code chat/agent mode
• Include checkpoint cells that report progress back to the workshop app API

Phase 4: Documentation Updates

• Update design_prd.md to list Genie Code as a supported assistant
• Update db/lakebase/README.md references
• Update SetUpProjectStep.tsx to handle the Genie Code setup path
• Add Genie Code to the workshop's introductory material

---

5. Open Questions

• Should the Genie Code path use a pre-built notebook template, or should participants create everything from scratch via Genie Code chat?
• How should step completion be tracked — manual checkbox, API callback from the notebook, or automatic detection?
• Should Genie Code prompts leverage specific Unity Catalog context (e.g., referencing the workshop's catalog/schema by name), or stay generic?
• Does the workshop need to handle Genie Code's geo availability restrictions for international workshop sessions?

---

References

• Genie Code documentation: https://docs.databricks.com/genie-code
• Genie Code blog post: https://www.databricks.com/blog/introducing-genie-code
• Genie Code press release: https://www.databricks.com/company/newsroom/press-releases/databricks-launches-genie-code-bringing-agentic-engineering-data
• Current app PRD: docs/design_prd.md
• Current setup step: src/components/SetUpProjectStep.tsx
• Current workflow sections: src/constants/workflowSections.ts
• Current prompt config: src/backend/prompts_config.yaml