🍜 SQLop - Your AI-Powered Data Slop Kitchen

"Because sometimes you just need a big bowl of synthetic data slop!"

Welcome to SQLop - where we cook up synthetic data and serve it with a side of natural language queries. No Michelin stars here, just good honest data slop that gets the job done.

What's Cooking?

SQLop is your friendly neighborhood AI chef that:

1. Generates Data Slop 🥘 - Feed it a SQL schema, get back mountains of realistic synthetic data
2. Serves Data on Demand 🍲 - Ask questions in plain English, get answers with pretty charts

Perfect for when you need test data but don't want to spend hours writing INSERT statements. We're talking 1000+ rows of data faster than you can say "connection pooling".

The Menu

🥘 Slop Generator (Phase 1)

The data generation kitchen where the magic happens:

• Upload your schema - Drop in a .sql, .ddl, or .txt file with your table definitions
• Tell us what you want - "Make it realistic", "More variety", "Heavy on the Italian names"
• Tweak the recipe - Adjust temperature, rows per table, and other nerdy parameters
• Preview your slop - Check out what we cooked before you commit
• Refine it - Not spicy enough? Tell us to regenerate with different vibes
• Take it to go - Download as CSV or ZIP, automatically stored in PostgreSQL

Respects all your constraints:

• Primary keys? ✅ Unique every time
• Foreign keys? ✅ All relationships intact
• Data types? ✅ Dates look like dates, numbers like numbers
• NOT NULL? ✅ No nulls where you don't want 'em

🍲 Chat with Slop (Phase 2)

Your conversational data server:

• Ask in plain English - "How many orders last month?" → SQL magic happens
• See the recipe - We show you the SQL we generated
• Pretty pictures - Auto-generated charts when it makes sense
• Safe and sound - Guardrails prevent SQL injection and other nastiness
• Keep chatting - We remember the conversation so you can ask follow-ups

Tech Stack (aka Our Kitchen Equipment)

• Chef AI: Gemini 2.0 Flash (via Google Vertex AI)
• Front of House: Streamlit
• The Pantry: PostgreSQL
• Plating: Seaborn + Matplotlib + Pandas
• Kitchen Tools: Docker
• Future: Langfuse for keeping an eye on the kitchen

Quick Start (Get Cooking!)

1. Set Up Your Kitchen

# Clone the repo
cd sqlop

# Fire up the stove (virtual environment)
python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate

# Get your ingredients
pip install -r requirements.txt

2. Configure Your Recipe Book

# Copy the template
cp .env.example .env

# Edit with your Google Cloud credentials
# See SETUP.md for detailed instructions

3. Start the Database (Our Pantry)

docker-compose up -d

4. Open the Kitchen

streamlit run src/app.py

Boom! 💥 Head to http://localhost:8501 and start slopping!

Sample Schemas

Try these out:

• 📚 Library Management - Books, authors, members, loans
• 🍕 Restaurant System - Menus, orders, customers, reservations
• 🏢 Company Database - Employees, departments, projects, salaries

Project Structure (The Floor Plan)

sqlop/
├── src/
│   ├── app.py                 # Main kitchen (Streamlit UI)
│   ├── utils/
│   │   ├── config.py          # Recipe configurations
│   │   ├── db.py              # Pantry management (PostgreSQL)
│   │   └── gemini_client.py   # The AI chef interface
│   └── tools/
│       ├── ddl_parser.py      # Menu reader (DDL parser)
│       ├── data_generator.py  # Slop maker
│       ├── nl2sql.py          # Translation service
│       ├── guardrails.py      # Health inspector
│       └── visualizer.py      # Food photographer
├── data/                      # Takeout containers
├── docker-compose.yml         # Pantry setup
├── requirements.txt           # Shopping list
└── SETUP.md                   # Detailed recipes

How to Use

Making Some Slop (Data Generation)

1. Click on "Slop Generator" tab
2. Upload a DDL file or paste your schema
3. (Optional) Add instructions like "make it more realistic" or "lots of variety"
4. Adjust the sliders if you're feeling fancy
5. Hit "Cook It Up!" and watch the magic
6. Preview your data - looking good?
7. Want changes? Use the refine box: "regenerate customers with more diverse names"
8. Download your slop as CSV or grab it all as a ZIP

Chatting with Your Slop (Queries)

1. Switch to "Chat with Slop" tab
2. Ask away: "Show me top 10 customers by order value"
3. We'll show you the SQL we generated
4. See your results in a nice table
5. If it's chart-worthy, we'll throw in a visualization
6. Keep asking questions - we remember what you're talking about

Development Roadmap

See .claude/PLAN.md for the full implementation checklist and .claude/STATUS.md for current progress.

Current Status: 🚧 Building MVP (Phase 1)

• ✅ UI Framework & Styling
• ✅ Database Layer
• ✅ Docker Setup
• 🚧 Data Generation Engine
• ⏳ Chat Interface
• ⏳ Observability (Langfuse)

The Philosophy

We believe in:

• Keep It Simple - No over-engineering, just working code
• Get It Done - MVP first, fancy features later
• Have Fun - Life's too short for boring READMEs
• Learn by Doing - This is a practice project, mistakes are features

Why "Slop"?

Because not every dataset needs to be a five-star meal. Sometimes you just need a heaping bowl of functional test data, and that's okay. Embrace the slop. Love the slop. Be the slop.

License

Educational project. Made with 🍜 and ☕.

---

Ready to start slopping? Check out SETUP.md for detailed setup instructions!