DuckData

Run SQL against the data in your Obsidian notes — inline CSV, markdown tables, vault files (CSV / TSV / JSON / Parquet), or remote URLs — and render the result as a table or ECharts visualization inside a code block. Powered by DuckDB-WASM.

Desktop only. The plugin loads the DuckDB-WASM worker and .wasm module from a CDN (configurable in settings); queries then run entirely on your machine — no data leaves your vault.

Quick start

Inline CSV:

```duckdata
type: table
data: |
  date,amount
  2026-06-01,100
  2026-06-02,120
sql: |
  SELECT date, SUM(amount::INT) AS total
  FROM data
  WHERE amount::INT > 100
  GROUP BY date
```

Reference a table elsewhere via datasource (quote values that begin with YAML reserved characters such as #, @, *, &, !):

<!-- duckdata:table id=sales -->

| date       | amount |
| ---------- | ------ |
| 2026-06-01 | 100    |
| 2026-06-02 | 120    |

```duckdata
type: table
datasource: "#sales"
sql: SELECT * FROM data WHERE amount::INT > 100
```

datasource accepts:

Form	Meaning
#id	Anchor <!-- duckdata:table id=ID --> above a table in the current note
path/to/note.md#id	Anchor in another vault note
path/to/file.{csv,tsv,json,ndjson,parquet}	File in the vault
https://example.com/x.parquet	Remote file; format inferred from URL extension

Markdown-embedded sources (inline data and anchor tables) are CSV only. File and URL sources support any DuckDB-readable format (.csv, .tsv, .json, .ndjson / .jsonl, .parquet).

The loaded dataset is exposed as a query-scoped CTE named data — your SQL is rewritten to WITH data AS (<reader>) <your sql> (or spliced into your own WITH clause). Each block is fully isolated, so multiple blocks in the same note can run in parallel without interfering.

CSV / TSV cells are read with ALL_VARCHAR=TRUE, i.e. every column is text. Cast explicitly when you need numeric / date semantics, e.g. amount::INT > 100 or CAST(date AS DATE) >= DATE '2026-06-01'.

type supports table (read-only data table) and chart (ECharts visualization).

Charts

type: chart renders the result via Apache ECharts. A small DSL covers common cases; otherwise any field is passed through as a raw ECharts option (deep-merged with the dataset of query rows).

Casting matters: CSV cells are strings — CAST(... AS INT/DOUBLE) numeric columns in your SQL so axes and tooltips behave.

DSL example (bar chart grouped by a category column):

```duckdata
type: chart
data: |
  month,product,sales
  2026-01,A,100
  2026-01,B,80
  2026-02,A,140
  2026-02,B,90
sql: |
  SELECT month, product, sales::INT AS sales FROM data
chart:
  kind: bar
  x: month
  y: sales
  series: product
  stack: true
  height: 320
```

Supported kind: bar, line, area, scatter, pie. DSL fields: kind, x, y (string or array), series (group-by column), stack, smooth (line), horizontal (swap axes), donut (pie), height (px, default 300). Anything else is merged into the underlying ECharts option, so you can set title, legend, tooltip, visualMap, etc.

Raw passthrough example (any chart kind ECharts supports):

```duckdata
type: chart
data: |
  name,value
  apple,10
  pear,30
  plum,18
sql: SELECT name, value::INT AS value FROM data
chart:
  tooltip: {}
  series:
    - type: pie
      radius: ['40%', '70%']
      encode: { itemName: name, value: value }
  height: 280
```

Registered chart kinds in this build: all 22 built-in ECharts series — bar, line, scatter, effectScatter, candlestick, boxplot, heatmap, pictorialBar, lines, pie, funnel, gauge, tree, treemap, sunburst, graph, sankey, radar, parallel, map, themeRiver, custom.

Settings

Open Settings → Community plugins → DuckData to configure:

Setting	Purpose
CDN mirror	jsDelivr (default), unpkg, or a custom URL (e.g. internal mirror). Supports {version} placeholder.
DuckDB memory limit	Applied as PRAGMA memory_limit. 0 keeps the engine default.
Query timeout (ms)	Aborts a query that runs longer than this. 0 disables.
Reload DuckDB engine	Disposes the running instance so CDN / memory changes take effect on the next block.

Development

See DEVELOPMENT.md for build, link-to-vault, and release instructions.