DuckDB notebook

DuckDB has changed how analysts work with local data. It is fast, embeddable, and speaks SQL over CSVs, Parquet files, JSON, and remote data sources. A DuckDB notebook wraps that engine in an interactive document where queries, results, and narrative live together. Instead of switching between a SQL client and a Python script, you write a query, see the result, and iterate in one place.

Why DuckDB in a notebook

Traditional data notebooks are built around Python. That is fine for machine learning, but it is overkill for the everyday work of filtering, joining, and aggregating data. DuckDB gives you columnar analytics performance without the overhead of a separate server. It runs in-process, so you can query a multi-gigabyte CSV with the same ease as a small table.

A DuckDB notebook makes that power accessible. You write SQL cells next to markdown and charts. The notebook tracks the lineage between queries so that when one table changes, downstream queries update automatically. For analysts who think in SQL, this is the most natural way to explore data.

SpurLab's DuckDB notebook

SpurLab treats DuckDB as a first-class app. You can create a DuckDB notebook from the app gallery, import local files, attach remote databases, and connect APIs. Every query is a cell in the reactive graph. If you replace a CSV with a newer version, every dependent table, chart, and SQL summary refreshes.

The notebook is also agent-aware. When you hit a complex query, you can ask @worker:claude to rewrite it, explain it, or turn it into a parameterized view. The agent sees the schema and sample rows, so its suggestions are grounded in your actual data. You stay in the SQL-first mindset while the agent handles the boilerplate.

Because SpurLab is local-first, your DuckDB files live on your machine. You can move notebooks between workspaces, share them with colleagues, or keep them private. There is no cloud runtime to babysit and no surprise egress bill.

From exploration to automation

A DuckDB notebook in SpurLab is not just for exploration. You can turn a notebook into a scheduled report, a dashboard widget, or an agent-triggered pipeline. When a query result crosses a threshold, an agent can be notified and asked to draft a summary or open a task. The notebook becomes a living component of your personal software stack.

FAQ

Can I query files outside the notebook workspace?

Yes. You can attach any file or database SpurLab has permission to access. The notebook tracks those attachments as dependencies and re-runs cells when the files change.

Does SpurLab install DuckDB for me?

Yes. When you add a DuckDB notebook app, SpurLab bundles the DuckDB engine and keeps it updated as part of the local runtime.

Can agents create new tables or views?

Agents can propose SQL changes and data transformations. You review every proposal as a bead before it is applied to your notebook.

Is my query history saved?

Yes. Every query, result, and agent edit is part of the notebook bead history, so you can reproduce any past state.