Skip to content

databricks-solutions/vibe-coding-workshop-template

BranchesTags

Repository files navigation

Vibe Coding Workshop Template

Build end-to-end data products on Databricks with AI-assisted development

This template is a complete data product builder powered by vibe coding. Start from a raw customer schema or existing data, and build your way through the full Databricks stack — Databricks Apps with Lakebase, medallion architecture (Bronze, Silver, Gold), semantic layer, Genie Spaces, ML pipelines, and GenAI agents — all guided by 55 agent skills and your AI coding assistant.

What is Vibe Coding?

Vibe Coding is an AI-assisted development approach where you collaborate with AI tools (like Cursor, GitHub Copilot, Claude Code, Windsurf, or similar) to rapidly build, iterate, and deploy production-quality data products. Instead of writing every line from scratch, you describe what you want and let the AI handle the implementation — guided by structured agent skills that encode best practices.

Quick Start

Workshop participants: See PRE-REQUISITES.md for the full setup checklist (workspace access, CLI, IDE, and authentication).

Prerequisites

Tool Required Installation
Databricks Workspace Yes Access to a Databricks workspace with Unity Catalog
Databricks CLI >= 0.295.0 Yes curl -fsSL https://raw.githubusercontent.com/databricks/setup-cli/main/install.sh | sh
Node.js v22+ Yes (Path A) nodejs.org — required by AppKit
AI-Powered IDE Yes Cursor, Windsurf, VS Code + Copilot, or similar
Python 3.10+ Yes python.org
Git Yes git-scm.com

Clone the Template

git clone https://github.com/databricks-solutions/vibe-coding-workshop-template.git my-project
cd my-project

Choose Your Starting Point

Path A: Build and Deploy a Databricks App

Build a full-stack TypeScript app on Databricks AppKit, guided by 7 agent skills:

  1. Read the workshop guide: apps_lakebase/Instructions.md
  2. Open your AI coding assistant and prompt:
I want to build a Databricks App. Read @apps_lakebase/skills/01-appkit-scaffold/SKILL.md and scaffold a new AppKit project.
  1. Follow the 5-phase workflow in Instructions.md:
    • Phase 1: Scaffold + build UI from a PRD, test locally (uses 01-appkit-scaffold and 02-appkit-build skills)
    • Phase 2: Deploy to Databricks Apps with mock data (uses 03-appkit-deploy skill)
    • Phase 3: Setup Lakebase project (uses 00-appkit-navigator skill)
    • Phase 4: Wire Lakebase backend locally (uses 04-appkit-plugin-add and 05-appkit-lakebase-wiring skills)
    • Phase 4b (optional): Wire Model Serving / Agent endpoint (uses 04-appkit-plugin-add and 06-appkit-serving-wiring skills)
    • Phase 5: Deploy + E2E test with Lakebase (uses 03-appkit-deploy skill)

Path B: Build an End-to-End Data Pipeline

Take a raw schema CSV through the full medallion architecture to production AI agents — one prompt per stage:

  1. Drop your schema CSV into data_product_accelerator/context/
  2. Open your AI coding assistant and prompt:
I have a customer schema at @data_product_accelerator/context/Wanderbricks_Schema.csv.
Please design the Gold layer using @data_product_accelerator/skills/gold/00-gold-layer-design/SKILL.md
  1. Follow the 9-stage pipeline guide — one prompt per stage, one new conversation per stage.

Both paths work together. Build your data pipeline first, then deploy a Databricks App on top of it — or start with the app and add data products incrementally.

Project Structure

vibe-coding-workshop-template/
│
├── README.md                       # This file
├── QUICKSTART.md                   # Quick-start guide with two pathways
├── AGENTS.md                       # AI assistant routing (universal entry point)
├── PRE-REQUISITES.md               # Workshop prerequisites checklist
├── CONTRIBUTING.md                 # Contribution guidelines
├── LICENSE.md                      # License
├── SECURITY.md                     # Security policy
├── env.example                     # Environment variable template
│
├── apps_lakebase/                  # Databricks AppKit Workshop (7 agent skills)
│   ├── Instructions.md             #   Comprehensive workshop guide
│   ├── prompts/                    #   Numbered prompt files for each workshop step
│   └── skills/                     #   Agent skills for the full app lifecycle
│       ├── 00-appkit-navigator/    #     Entry-point navigator (read first)
│       ├── 01-appkit-scaffold/     #     Scaffold new AppKit projects (+ agent skills install)
│       ├── 02-appkit-build/        #     Build UI + backend from a PRD
│       ├── 03-appkit-deploy/       #     Deploy to Databricks Apps
│       ├── 04-appkit-plugin-add/   #     Add plugins (Lakebase, Analytics, Genie, Files, Serving)
│       ├── 05-appkit-lakebase-wiring/ #  Wire Lakebase DDL, API routes, frontend hooks
│       └── 06-appkit-serving-wiring/  #  Wire Model Serving / Agent endpoint to frontend
│
├── presentations/                  # Workshop slide deck (Marp, HTML, PDF, PPTX)
│
├── data_product_accelerator/       # 55 Agent Skills for End-to-End Data Products
│   ├── AGENTS.md                   #   Detailed skill routing table
│   ├── QUICKSTART.md               #   One-prompt-per-stage guide (9 stages)
│   ├── README.md                   #   Accelerator overview
│   ├── context/                    #   Schema CSV inputs (starting point)
│   ├── skills/                     #   55 skills across 12 domains
│   │   ├── admin/                  #     Skill creation, auditing (4)
│   │   ├── bronze/                 #     Bronze layer + Faker data (2)
│   │   ├── common/                 #     Cross-cutting shared skills (8)
│   │   ├── exploration/            #     Ad-hoc notebooks (1)
│   │   ├── genai-agents/           #     GenAI agent patterns (9)
│   │   ├── gold/                   #     Gold design, implementation, workers (14)
│   │   ├── ml/                     #     MLflow pipelines (1)
│   │   ├── monitoring/             #     Monitors, dashboards, alerts (5)
│   │   ├── planning/               #     Project planning (1)
│   │   ├── semantic-layer/         #     Metric Views, TVFs, Genie, optimization (10)
│   │   ├── silver/                 #     DLT pipelines, DQ rules (3)
│   │   └── skill-navigator/        #     Master routing system (1)
│   └── docs/                       #   Framework design documentation
│
└── agentic-framework/              # Multi-Agent Build Framework
    ├── agents/                     #   Agent prompts for building multi-agent systems
    │   ├── prd-analyzer.md         #     Parse PRDs, map to agent capabilities
    │   ├── skill-scaffolder.md     #     Create new Agent Skills (SKILL.md)
    │   ├── tool-builder.md         #     Build runtime Python tools
    │   ├── agent-tester.md         #     Configure agent behavior tests
    │   ├── agent-ui-wiring-prompt.md #   Guide agent-to-UI wiring
    │   ├── multi-agent-build-prompt.md # Orchestrator build with Foundation Models
    │   ├── databricks-deployer.md  #     Deployment guidance
    │   └── prd-template.md         #     PRD template
    └── skills/
        └── foundation-model-agent-loop/
            └── SKILL.md            #   Tool-calling loop with Foundation Models

How It All Fits Together

This template supports a unified workflow from raw data to production data products:

Raw Schema CSV or Existing Data
  │
  ├─► Gold Design         — dimensional model, ERDs, YAML schemas
  ├─► Bronze Layer         — source tables + test data (Faker)
  ├─► Silver Layer         — DLT pipelines + data quality expectations
  ├─► Gold Layer           — tables, MERGE scripts, FK constraints
  ├─► Semantic Layer       — Metric Views, TVFs, Genie Spaces
  ├─► Observability        — Lakehouse Monitors, AI/BI Dashboards, SQL Alerts
  ├─► ML Pipelines         — MLflow experiments, training, inference
  ├─► GenAI Agents         — ResponsesAgent, evaluation, deployment
  │
  └─► Databricks App       — AppKit (full-stack TypeScript), deployed on Databricks Apps

Each stage is driven by a single prompt to your AI coding assistant. The 55 agent skills in data_product_accelerator/ encode production-tested patterns so you get governed, high-quality output at every step.

Data Product Accelerator (55 Agent Skills)

The data_product_accelerator/ directory contains 55 agent skills organized by domain that guide your AI assistant through the entire pipeline:

Domain Skills Focus
Gold 14 Dimensional modeling, design workers, pipeline workers, ERDs, MERGE scripts
Semantic Layer 5 Metric Views, TVFs, Genie Spaces, export/import API
GenAI Agents 10 ResponsesAgent, evaluation, deployment, simple agent scaffold
Common 8 Asset Bundles, naming, constraints, imports
Monitoring 5 Lakehouse Monitors, dashboards, SQL alerts
Admin 4 Skill creation, auditing, docs
Silver 3 DLT pipelines, expectations, DQX
Bronze 2 Bronze tables, Faker data generation
ML 1 MLflow pipelines
Planning 1 Project planning
Exploration 1 Ad-hoc notebooks
Skill Navigator 1 Master routing system

See data_product_accelerator/QUICKSTART.md for the step-by-step 9-stage guide.

Databricks AppKit Workshop (7 Agent Skills)

The apps_lakebase/ directory contains 7 agent skills and a comprehensive workshop guide for building full-stack TypeScript apps on Databricks AppKit. The app is not pre-built — it gets scaffolded at runtime via databricks apps init and built iteratively with your AI coding assistant.

What gets built:

  • Full-stack TypeScript app (React + Tailwind CSS frontend, AppKit backend)
  • SQL Warehouse integration for analytics queries
  • Lakebase (managed PostgreSQL) persistence (wired in phases 3-5)
  • Deployed to Databricks Apps with hot reload for local dev

Workshop Skills

Skill Purpose
00-appkit-navigator Entry-point navigator — routes tasks to the correct skill
01-appkit-scaffold Scaffold new AppKit projects with plugins (analytics, lakebase, genie, files)
02-appkit-build Build UI and backend from a PRD — components, queries, type generation
03-appkit-deploy Deploy to Databricks Apps, validate configuration
04-appkit-plugin-add Add plugins to an existing AppKit project
05-appkit-lakebase-wiring Wire Lakebase DDL, Express API routes, frontend hooks, mock fallback
06-appkit-serving-wiring Wire Model Serving / Agent endpoint to frontend (streaming, invoke)

Local Development (after scaffolding)

Service URL
App + API http://localhost:8000
Health Check http://localhost:8000/health

Start the dev server from your scaffolded app directory with npm run dev.

Agentic Framework

The agentic-framework/ directory provides prompts and patterns for building multi-agent systems with Databricks Foundation Models.

Agent prompts (in agentic-framework/agents/):

Agent Purpose
prd-analyzer Parse PRDs, map requirements to agent capabilities
skill-scaffolder Create new Agent Skills (SKILL.md) for any domain
tool-builder Build runtime Python tools for agents
agent-tester Configure tests for agent behavior
agent-ui-wiring-prompt Guide agent-to-UI integration
multi-agent-build-prompt Build multi-agent orchestrators with Foundation Models
databricks-deployer Deployment guidance for agents
prd-template PRD template for agent projects

Foundation Model Agent Loop (in agentic-framework/skills/foundation-model-agent-loop/):

  • Pattern for tool-calling loops with Databricks Foundation Models (e.g., databricks-meta-llama-3-3-70b-instruct)
  • Supports function calling without custom model deployment

How Deployment Works

After scaffolding your AppKit app, the 5-phase workflow progresses from mock data to a fully wired Lakebase backend:

┌─────────────────────────────────────────────────────────────────────┐
│                    AppKit Workshop Phases                            │
├─────────────────────────────────────────────────────────────────────┤
│                                                                     │
│  Phase 1: SCAFFOLD + BUILD     Phase 2: DEPLOY (mock data)         │
│  ────────────────────────      ──────────────────────────           │
│  databricks apps init          npm run build                        │
│  Build UI from PRD             databricks apps deploy --profile <P> │
│  npm run dev (localhost:8000)  Verify at Databricks Apps URL        │
│                                                                     │
│  Phase 3: SETUP LAKEBASE       Phase 4: WIRE LAKEBASE              │
│  ──────────────────────        ─────────────────────               │
│  Create Lakebase project       Add Lakebase plugin (skill 04)      │
│  Configure endpoint + compute  DDL, API routes, frontend (skill 05)│
│  Record host in state file     Test locally with mock fallback      │
│                                                                     │
│  Phase 5: DEPLOY + E2E TEST                                        │
│  ────────────────────────                                           │
│  databricks apps deploy (with Lakebase config)                      │
│  Verify live data end-to-end                                        │
│                                                                     │
└─────────────────────────────────────────────────────────────────────┘

Key Files in the Generated App

File Purpose
app.yaml Defines how Databricks starts your app
databricks.yml Databricks bundle configuration
server/server.ts AppKit backend entry point
client/src/ React + Tailwind CSS frontend
package.json Node.js dependencies

See the 03-appkit-deploy skill for the full deployment workflow.

Authentication

Configure a Databricks CLI profile to authenticate:

databricks auth login --host https://your-workspace.cloud.databricks.com

Verify it works:

databricks current-user me

To use a named profile (useful when working with multiple workspaces):

databricks auth login --host https://your-workspace.cloud.databricks.com --profile myprofile
databricks current-user me --profile myprofile

All skills and CLI commands accept a --profile flag to target a specific workspace.

Customizing Your App

After scaffolding, your generated AppKit app is a full-stack TypeScript project. Customize it using standard AppKit patterns:

Adding Backend Routes

Edit server/server.ts in your generated app directory to add tRPC routes or custom endpoints. See the 02-appkit-build skill for patterns.

Adding Plugins

Use the 04-appkit-plugin-add skill to add capabilities:

Read @apps_lakebase/skills/04-appkit-plugin-add/SKILL.md and add the Lakebase plugin to my app.

Available plugins: analytics, lakebase, genie, files

Adding Dependencies

npm install your-package

Consulting AppKit Docs

npx @databricks/appkit docs              # documentation index
npx @databricks/appkit docs "<query>"    # search for a specific topic

Troubleshooting

Check Databricks CLI

databricks --version          # Should be >= 0.295.0
databricks current-user me    # Verify authentication
databricks auth profiles      # List configured profiles

Authentication failed

databricks auth login --host https://your-workspace.cloud.databricks.com

Port 8000 in use

lsof -ti:8000 | xargs kill -9 2>/dev/null || true
npm run dev

View Deployed App Logs

databricks apps get <APP_NAME> --profile <PROFILE>

Local Testing (after scaffolding)

cd <your-app-directory>
npm run dev
# Open http://localhost:8000

Resources

Using This Template

This is a Git template repository. To use it:

  1. Click "Use this template" on GitHub, or clone directly
  2. Choose your starting point:
    • Build a Databricks App: Follow the AppKit workshop guide — scaffold, build, deploy, and wire Lakebase
    • Build a data product: Drop a schema CSV in data_product_accelerator/context/ and follow the 9-stage guide
    • Build agents: Use the prompts in agentic-framework/agents/ to scaffold multi-agent systems
  3. Iterate with your AI coding assistant — the agent skills handle the patterns

Ready to build? Let's go!

git clone https://github.com/databricks-solutions/vibe-coding-workshop-template.git my-project
cd my-project

About

No description, website, or topics provided.

Resources

Readme

License

View license

Contributing

Contributing

Security policy

Security policy
Activity
Custom properties

Stars

2 stars

Watchers

0 watching

Forks

2 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages