DuckDB Ecosystem Newsletter – March 2026

Google Street View in 2026

TL;DR: Mark details an efficient DuckDB workflow for processing 7.1 million geospatial points from JSON to a spatially-sorted, ZSTD-compressed Parquet dataset for Google Street View coverage analysis.

Mark implemented a geospatial data pipeline using DuckDB, leveraging its JSON, Parquet, and Spatial extensions, alongside community H3 (Hexagonal hierarchical geospatial indexing system) and Lindel extensions, on a 5.7 GHz AMD Ryzen 9 workstation.

He ingested 131 JSON files (647 MB) by dynamically inserting data using READ_JSON and UNNEST(customCoordinates), transforming coordinates into ST_POINT geometries, then exported them to an 85 MB Parquet file optimized via HILBERT_ENCODE sorting and ZSTD compression.

Polyglot: Rust/Wasm-powered SQL transpiler for more than 30 SQL dialects

TL;DR: Polyglot is a Rust/Wasm-powered SQL transpiler supporting over 32 dialects, including DuckDB.

Available as a Rust crate, TypeScript/WASM SDK, and Python package, it handles transpilation between different SQL dialects. For example, a MySQL query like SELECT IFNULL(a, b) can be transpiled to PostgreSQL's SELECT COALESCE(a, b).

Noteworthy: it's built on top of more than 8.5k fixtures from SQLGlot, and the author estimates it took around 7B tokens to build from scratch using Claude Code Max.

duckdb-vscode: A DuckDB "studio" extension for VS Code

TL;DR: The duckdb-vscode extension integrates DuckDB into VS Code, leveraging @duckdb/node-api for direct querying of various data formats and database management within the IDE.

The extension allows SQL execution against in-memory or persistent .duckdb files, and remote sources like S3 or Postgres via explicit ATTACH statements. An implementation detail involves creating temporary tables for efficient server-side pagination, sorting, and filtering of results, with DuckDB automatically spilling to disk for large datasets.

Query Snowflake Directly from DuckDB

TL;DR: A new community Snowflake extension enables direct querying of Snowflake tables from DuckDB using the ADBC driver.

Connect via CREATE SECRET, then ATTACH the Snowflake database with enable_pushdown true to offload filters and projections. Results can be materialized into local DuckDB tables, enabling joins between Snowflake data and local files in a single query. Code on GitHub.

ducklake-hetzner: DuckLake on Hetzner for under 10 euros a month

TL;DR: A budget DuckLake lakehouse deployment on Hetzner Cloud using PostgreSQL for metadata and S3-compatible storage, orchestrated via OpenTofu and PyInfra.

This setup costs under €15/month for a CX33 VPS (4 vCPU, 8GB RAM) and object storage. Key technical implementations include OpenTofu for infrastructure provisioning and PyInfra for PostgreSQL 16 server configuration. Still in early stages, but a great starting point for cost-conscious lakehouse deployments.

duckdb-netquack: DuckDB extension for parsing domains, URIs, and paths

TL;DR: Arash's Netquack extension for DuckDB delivers performance enhancements for URI/domain parsing and network utility functions.

Netquack provides a suite of intuitive functions to handle all your network tasks efficiently. For example, SELECT extract_domain('brain.ssp.sh') AS domain; returns ssp.sh. Other functions extract the path, hostname, protocol, and query string of a URL. More advanced use cases include Base64 encoding and URL validation: SELECT is_valid_url('motherduck.com') AS valid;.

Shaper: Visualize and share your data. All in SQL. Powered by DuckDB.

TL;DR: Open-source, DuckDB-powered platform for SQL-driven dashboards — described on HackerNews as "a DuckDB-based Metabase alternative."

Built with Go and TypeScript, it leverages DuckDB's analytical power to allow users to construct interactive dashboards purely through SQL queries. Shaper supports querying across various data sources using DuckDB. Quickstart with docker run --rm -it -p5454:5454 taleshape/shaper, providing an immediate way to explore its capabilities. Try the live demo, watch the YouTube demo, or read the docs.

Duck, Dive, and Answer

TL;DR: MotherDuck has released Dives, a new feature enabling AI agents to build shareable, real-time data visualizations from composable SQL, leveraging the Remote MCP Server with LLMs.

The MotherDuck Remote MCP Server, when used with LLMs such as Claude, Gemini, and ChatGPT and provided with contextual schema, has demonstrated over 95% functional correctness in text-to-SQL tasks. "Dives" are intentionally not called dashboards — they go beyond dashboards because they can be anything you can do with code. Check out practical Dives examples by Jacob. Note: Dives are in public preview.

DuckDB Developer Meeting #1

TL;DR: Slides and videos of the first DuckDB developer meeting in 2026 are out.

Highlights include the DuckDB C API and extension template for developers (part 1, part 2), Sam on the "past, present, and future" of DuckDB extensions, Lotte on "Storage and encryption in DuckDB", Denis introducing "DuckPL" — a new procedural language being integrated — and Philip presenting "GizmoEdge".

SwanLake: An Arrow Flight SQL Datalake Service Built on DuckDB + DuckLake

TL;DR: Wang's SwanLake is a Rust-based Arrow Flight SQL server wrapping DuckDB and DuckLake, making DuckDB a deployable, observable analytics service.

It manages isolated DuckDB connections per session, preloads extensions (ducklake, httpfs, aws, postgres), and supports bootstrap SQL via SWANLAKE_DUCKLAKE_INIT_SQL. Built-in observability includes session counts, query latency percentiles (p95/p99), and error history. Initial TPC-H benchmarks show local Postgres storage at ~10.4 req/s vs. S3-backed at ~4.9 req/s. Code on GitHub.

Upcoming Events

Agents That Build Tables, Not Just Query Them

2026-03-17. h: 16:30. Online

A livestream on how the new query_rw tool in the MotherDuck MCP server gives agents write access to your warehouse, so they can save derived tables and intermediate results — not just read from them.

What's New in DuckDB 1.5

2026-03-19. h: 16:00. Online

Alex and Jacob walk through the biggest changes in DuckDB 1.5, demo key features, and take questions live. A must-watch if you want to stay current with the latest release.

MotherDuck + DuckDB Meetup — San Francisco

2026-03-26. h: 18:00. San Francisco, CA, USA

The inaugural SF meetup for DuckDB & MotherDuck users. Till Döhmen and Jacob Matson will talk about SQL evals and what AI means for infrastructure providers, plus special guests sharing real-world stories of using DuckDB at scale.