Five complete, working implementations showing how to build data pipelines, ML-in-SQL, semantic layers, and AI analytics—without cloud dependencies.
Not tutorials or docs—actual production-quality code you can run, study, and fork.
See how DLT, dbt, DuckDB, Rust extensions, and MCP integration actually fit together.
Proven patterns for local-first architecture, tested and documented. Just fork and adapt.
Open source • MIT Licensed • Archived as reference implementations (Nov 2025)
Whether you're building similar tools or learning modern data engineering, these implementations save you months of research.
Stop piecing together scattered docs. See complete working examples of:
Skip tutorials. Learn from production-quality code that shows:
Each project builds on the foundation to create a complete local-first analytics platform
The core framework providing local-first data pipelines with DLT (ingestion), dbt (transformation), and DuckDB (analytics). Everything else builds on this foundation.
A DuckDB extension adding ML/AI capabilities. Run zero-shot predictions, generate embeddings, and get feature importance—all in SQL, no separate ML infrastructure needed.
Visualizes dbt semantic layers with interactive lineage graphs. Understand how your metrics, dimensions, and entities connect. Built with Tauri and React Flow.
Ask questions in natural language, get answers based on real query results with statistical rigor. Execution-first approach prevents AI fabrication with confidence intervals and significance testing.
Connects everything to AI assistants via MCP (Model Context Protocol). Query your data through Claude Desktop or ChatGPT Desktop with automatic dbt model syncing and semantic layer integration.
A complete stack for building local-first analytics tools. Start with raw data, transform it, analyze it with ML, visualize relationships, and query it conversationally—all without cloud dependencies.
Each project is a complete, production-quality reference implementation. Fork any or all to build your own local-first data tools.
A developer sandbox framework for local-first data pipeline development using DLT, DuckDB, and dbt. It provides a complete local-first environment for prototyping, learning, and developing data solutions before deploying to production systems.
A local-first semantic layer for AI-powered analytics, providing a "Snowflake Cortex for Local-First Databases." It allows you to run powerful, zero-shot tabular predictions directly in your database with simple SQL.
A local-first application for visualizing and exploring dbt semantic layers. It connects directly to your dbt project and Snowflake account to provide a real-time, interactive lineage graph of your metrics, dimensions, and entities.
An AI-powered data analyst with a semantic layer, statistical rigor, and natural language insights. It allows you to ask questions in natural language and get answers based on real query results, not AI guesses.
A local-first agentic analytics platform that extends `sbdk-dev` to enable natural language queries against your data through AI assistants like Claude Desktop and ChatGPT Desktop via the Model Context Protocol (MCP).
These projects are reference implementations showing how to build local-first data tools. Here's how to get started:
Start with SBDK.dev for the foundation, or choose any project that matches your needs. Each works standalone or as part of the ecosystem.
Fork the repository, read the README, explore the code. Each project includes comprehensive documentation and examples.
These are reference implementations—take what works, modify what doesn't, and build your own tools. All projects are MIT licensed for maximum flexibility.
Built something cool? Share it! Open an issue on the original repo to showcase your fork or derivative work.
These are complete, stable reference implementations—not active products. They're archived because they're done: production-quality code demonstrating proven patterns. Perfect for forking, learning, or adapting for your own projects.