Go AI Monolith - Data Assistant

This is a monolithic web application built with Go (Golang) that leverages Google Gemini AI to assist with database operations. It provides an intuitive interface for generating massive amounts of synthetic data from schema definitions and querying that data using natural language.

Features

1. Data Generation

• Schema Parsing: Upload any PostgreSQL .ddl file. The system automatically creates the tables in the database.
• AI-Powered Generation: Uses Gemini (gemini-1.5-flash) to generate context-aware INSERT statements based on your schema.
• Customizable: Adjust Temperature (creativity) and Max Tokens to control the variety and volume of generated data.
• Real-time Preview: View a sample of the generated data immediately.

2. Talk to your Data

• Natural Language Queries: Ask questions like "Show me the top 5 customers by spending" or "List all orders from yesterday".
• Automatic SQL: The AI converts your questions into safe, read-only SQL queries (SELECT only).
• Visualization: Ask for charts (e.g., "Show a bar chart of sales by region") to automatically render visualizations using Chart.js.

3. Export

• Download Data: Export your tables as CSV files or download the entire database as a ZIP archive.

Prerequisites

• Docker & Docker Compose
• A Google Gemini API Key (Get one at aistudio.google.com)

Quick Start

1. Clone the repository:

   git clone <repository-url>
   cd genai

2. Set your API Key: You can export it as an environment variable or create a .env file (if you add support for it). For Docker Compose, passing it directly works best:

   export GEMINI_API_KEY="your_actual_api_key_here"

3. Run with Docker Compose: This command builds the Go application and starts the PostgreSQL database.

   docker-compose up --build

4. Access the App: Open your browser and navigate to: http://localhost:4000

Configuration

The application is configured via environment variables.

Variable	Description	Default (in Docker)
GEMINI_API_KEY	Required. Your Google AI API Key.	None
GEMINI_MODEL	Gemini model to use.	gemini-2.0-flash
DATABASE_URL	Connection string for PostgreSQL.	postgres://user:password@db:5432/genai?sslmode=disable
PORT	Port for the web server.	4000

Development Workflow

Rebuilding After Code Changes

When you make changes to the Go source code, you need to rebuild the Docker container:

# Set your environment variables
export GEMINI_API_KEY="your_actual_api_key_here"
export GEMINI_MODEL="gemini-2.0-flash"

# Stop, remove, and rebuild the app container
docker compose stop app && docker compose rm -f app && docker compose up --build -d app

Note: Changes to ui/html/index.html only require a browser refresh, not a rebuild.

Available Gemini Models

To see which models are available with your API key, run:

GEMINI_API_KEY="your_key" go run cmd/list_models/main.go

Common models for the free tier:

• gemini-2.0-flash (recommended, default)
• gemini-2.5-flash
• gemini-flash-latest

Project Structure

├── cmd/
│   └── web/
│       └── main.go       # Application entry point & HTTP handlers
├── internal/
│   ├── database/         # Database connection & safety logic
│   └── gemini/           # AI Client & System Instructions
├── ui/
│   └── html/
│       └── index.html    # Single-page UI (Tailwind + Vanilla JS)
├── Dockerfile            # Multi-stage build for Go
├── docker-compose.yml    # App + Postgres orchestration
└── go.mod                # Go dependencies

Security Note

• Prompt Injection: The system uses a restricted regex blocklist (DROP, DELETE, UPDATE, TRUNCATE) to prevent destructive queries.
• System Instructions: The AI is instructed via strictly scoped system prompts to only perform "Read" operations in the analysis mode.
• Environment: It is recommended to run this in a development or sandboxed environment.