About Queria

What is Queria?

Queria is a service that publishes data as a database, not as files or an API.

For the people using the data, there is no preprocessing. Open a page and you can start writing SQL immediately, and Python, BI tools and AI agents reach the same data directly. For the people publishing it, most often a local government, what they publish becomes usable as it is and sits alongside data from other publishers.

We are starting with open data from Japan, and over time we want to cover data from anywhere in the world.

The main features and how to use them are covered in the guide.

Why we are building this

We want to create a world where data is something anyone can casually touch. That is the feeling at the very root of Queria.

When more people get to work with data, they gain small opportunities to understand the world a little more deeply. You notice that "huh, so that's how it actually is," or the background of a news story you had only vaguely heard about suddenly clicks once you see the numbers. As those small realizations accumulate, they become the raw material for understanding the world correctly. We hope that working with data brings more of those moments to more people.

That said, data analysis has long belonged to a limited few. Set up a server, provision large storage, learn SQL, study statistics and visualization — it takes money, skills, and time. So inevitably it remained the pursuit of a small group of people.

In the last few years, those assumptions have changed dramatically.

• DuckLake lets the metadata live in the same object storage as the data, which has drastically lowered the cost of hosting
• DuckDB and DuckDB WASM let all the computation needed for analysis run entirely inside the browser
• With the arrival of LLMs, you can increasingly ask questions of data in natural language, without knowing SQL or statistics

The walls that used to block analysis are disappearing one after another. A world where anyone can casually touch data is, technically, already within reach.

Open data itself, however, still has a different set of problems. Published data comes in inconsistent formats, with no standardization of character encodings, column names, or units. The moment you try to use it, you burn hours on preprocessing — and people all over the world keep writing the same preprocessing in their own environments. Reinventing the same wheel everywhere is wasteful, and more fundamentally, the very step of "download, then preprocess" pushes away data that should be easy to touch.

This problem actually gets worse in an era where AI helps with analysis. It is not as simple as "hand the data to an LLM and it will answer everything cleverly." When data is messy and its meaning is ambiguous, AI returns off-target answers or simply makes things up (hallucinates). For AI to answer accurately — and for people to verify those answers — "preparing the data" and "describing its meaning" matter more than ever. Behind the flashy technology, it is the unglamorous work of organizing data that counts.

What Queria wants to do is take on all of that preparation. Put cleanly organized data, with its meaning properly attached, in a place everyone can access in the same way. We believe that alone transforms "open data that is published" into "open data that is actually usable."

What we care about

Delivering data in a usable form

Open data is inconsistent at the moment it is published. We align formats, standardize column names and units, absorb gaps and inconsistencies, and make datasets joinable through shared keys. On top of that, we prepare metadata — the meaning, units, and aggregation methods of tables and columns — as a set. Our job is to make sure users never have to write this "preparation" over and over again.

Cleanly organized data is not just a starting point for human analysis; it directly improves the accuracy of answers when you hand the data to AI. With well-maintained metadata, AI interprets the data more precisely, and off-target answers and hallucinations become less likely. The Analyze with Claude Code guide shows what this feels like in practice.

Keeping AI answers verifiable by humans

Queria is built around accessing data with SQL. This is so that when you delegate analysis to AI, a human can trace the basis of its answers with their own eyes.

With systems that call data through an API and return only results, it is hard to trace why a number came out the way it did. What happens inside is a black box, and you have no material for deciding whether to trust the answer AI returned. With SQL, on the other hand, you can read the WHERE clauses, the JOINs, and the aggregation logic and judge whether they are sound. You can read the SQL that AI wrote, convince yourself "right, it really is aggregating under these conditions," and then accept the answer.

We want people to be able to check AI's answers with their own eyes rather than swallowing them whole. That is why we consistently choose SQL as the foundation for data access.

Use it from the tools you love

Everyone has their own favorite tools for data analysis. Some people write SQL in the browser, others work in Python or Jupyter, and others build dashboards in Tableau or Power BI. Queria publishes its data as a database rather than as files — specifically, a DuckLake catalog — so you can reference it directly from any environment that can connect to DuckDB. We do not decide which tools you should use. See the connection guide for details.

Staying open to everyone

Open data only means something when it is openly accessible. We do not charge individuals for working with open data, and our source code is public. Data analysis is not the exclusive domain of engineers and data scientists — policy officers in local government, retail developers, researchers, journalists — we believe it matters that people in all kinds of roles can explore data according to their own interests.

We will build separate mechanisms to sustain the ongoing work of maintaining the data, but we will keep protecting a state where anyone can reach the data directly.

For open data publishers

We also have a message for those who publish open data — local governments, national agencies, public institutions, and companies.

Existing catalog platforms like CKAN have played a major role as entry points that make data easier to find. At the same time, when the approach is to line up individual CSV and Excel files, connecting datasets together is left to preprocessing in each user's environment. The data is public, yet every use still means another round of downloading and reshaping — and that situation rarely changes.

Publishing as a database, the way Queria does, means the publisher organizes the data once and users can aggregate it directly — from the browser, from Python, or from BI tools. Aggregations that span multiple datasets are written the same way, so publishers themselves can create a state where users work with the data as-is, without format conversion in between.

There are two more things in it for the publisher. One is that we can measure how much the data is used and report it back. Without any visible response, it is hard to argue internally for keeping the work going, or to judge what publishing is worth. The other is that the data sits alongside data from other publishers. Municipality codes and coordinates line up across the whole catalog, so uses emerge that would never have come from one organization's data on its own.

Today Queria collects and organizes external open data, but we believe the cleanest form is for data to come out well-organized from the publisher itself. If your organization is rethinking how it publishes data, we are happy to think together about organizing data and designing and building publishing pipelines.

If Queria has caught your interest, feel free to reach out through the contact form. We would love to build the next form of open data together.