Kitsuna Data

Self-hosted one-person data platform

Table of Contents

1. About The Project
   • Main Features
   • Goals
   • Tech Stack
2. Getting Started
   • Prerequisites
   • Installation
3. Usage
4. Roadmap
5. Contact
6. Inspiration

About The Project

This is a concept for what a Rails-inspired small data platform for startups and SMEs could look like. After using a variety of end-to-end solutions like DOMO, Keboola, Mozart Data and others, I keep wishing there was something that would do the 80% of ELT + BI out-of-the-box, without the price surprises.

This project is an attempt to stitch together a set of solid and reliable open-source tools that combine into a lean platform where one data engineer can own the entire lifecycle. From ELT, to data modelling, to deploying and scaling in production.

Main Features

1. 🧪 From laptop to production in minutes - Develop locally with DuckDB, deploy with the same code. No more "it works on my machine" problems.

2. ⚡ Lightning-fast analytics on any data size - DuckDB's column-oriented design handles gigabytes of data on modest hardware. Query billions of rows in seconds.

3. 📊 Beautiful dashboards - Drag-and-drop dataviz with Metabase. Perfect for everyone - tech and non-tech alike.

4. 💸 Scale without breaking the bank - Enterprise-grade data stack for as little as $30/month. DuckDB + SQLMesh's efficiency means less compute costs than Snowflake or BigQuery.

5. 🔄 30+ ready-to-use integrations - Instant integrations with dlt for Stripe, GitHub, Salesforce, and more. Connect your SaaS tools with minimal code.

6. 🤖 Just ask your DB - Ask questions in plain English with DuckDB's MCP. Get immediate answers without writing complex queries.

7. 🔍 End-to-end data lineage - SQLMesh tracks transformations from raw to gold data. Understand exactly where metrics come from and debug easily.

Caution

The project is very much in the pre-alpha stage. This is more of an experiment and is not meant for produciton workloads.

Goals

• Local-first development for the entire stack.
• Support companies that can't afford heavy, expensive data tools or large teams.
• No "SSO tax" - all tools should be either fully free, or affordable once deployed in serious prod use case.
• No k8s, so a small data team can be self-sufficient .
• Cheap path to production and scaling.

Tech Stack

• Extract (planned): dlt
• Transform: SQLMesh
• Data Storage: DuckDB
• BI / data viz: Metabase
• Deployment: Dokploy

(back to top)

Getting Started

Prerequisites

This is an example of how to list things you need to use the software and how to install them.

• uv
• mise (recommended)
• claude (recommended)

Installation

1. Clone this repository
2. Download the DuckDB driver for Metabase:

   make download-duckdb-driver

3. Start the services:

   docker-compose up -d

4. Access Metabase at http://localhost:3000

(back to top)

Usage

TODO

(back to top)

Deployment

This project can be deployed to DigitalOcean/Hetzner/EC2 using Dokploy with the following architecture:

1. Metabase Container:

   • Dedicated hostname (e.g., metabase.yourdomain.com)
   • Access to mounted DuckDB volume

2. dlt + SQLMesh Container:

   • Combined container for data processing
   • Access to the same DuckDB volume

3. Shared Storage:

   • Used for persistent DuckDB storage

Roadmap

• Add SQLMesh
• Add MCP for DuckDB
• Add dlt
  • Implement as an example: Exploring StarCraft 2 data with Airflow, DuckDB and Streamlit | by Volker Janz | Data Engineer Things
• Add Dokku deployment configuration
  • Create a DigitalOcean box for a public demo
• Add installation docs
• Add usage docs
• Add Aider docs

(back to top)

Contact

Greg Goltsov - @gregoltsov, gregoltsov.bsky.social.

(back to top)

Inspiration

Here are some projects which inspired my thinking and this project:

• Modern Data Stack in a Box with DuckDB
• MDS-in-a-box: Monte Carlo simulation of the NBA season
• Exploring StarCraft 2 data with Airflow, DuckDB and Streamlit

(back to top)