diff --git a/CHANGELOG.md b/CHANGELOG.md
index 046e1f756..0d22d3299 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -7,21 +7,14 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
## [Unreleased]
-### Changed
-- **Breaking:** `ArrayStats` has one component per field of Rust's array stats, so `minExact`, `maxExact`, `isSorted`, `isStrictSorted`, `isConstant`, `uncompressedSizeInBytes` and `nanCount` are read from the file and `trueCount` is gone (a bool column's true count is its `sum`). ([#581](https://github.com/dfa1/vortex-java/pull/581))
-
-### Fixed
-- Files written by vortex-java mark their array min/max as exact, which Rust read as inexact bounds. ([#581](https://github.com/dfa1/vortex-java/pull/581))
-- `VortexReader.columnStats()` and SQL `MIN`/`MAX` no longer report a truncated (inexact) bound as a column's minimum or maximum. ([#581](https://github.com/dfa1/vortex-java/pull/581))
-
-## [0.18.0] — 2026-10-19
+## [0.18.0] — 2026-10-20
### Added
-- `Compute.matcher` returns a `RowSelection`, whose `nextSelected(long)` jumps to the next selected row instead of testing each one. ([#580](https://github.com/dfa1/vortex-java/pull/580))
-- `Compute.matcher(RowFilter)` gives the row test of each chunk of one scan, evaluating a dictionary the chunks share once. ([#576](https://github.com/dfa1/vortex-java/pull/576))
- `vortex-jdbc-driver` is a `jdbc:vortex:
` JDBC driver in one self-contained jar, so IntelliJ, DBeaver and other SQL tools can query a directory of `.vortex` files. ([#566](https://github.com/dfa1/vortex-java/issues/566))
- `VortexSchemaFactory` builds the Calcite schema from a `jdbc:calcite:model=...` URL, so SQL tools can query a directory of `.vortex` files without Java code. ([#566](https://github.com/dfa1/vortex-java/issues/566))
- `VortexCalcite.connect(Path)` and `VortexCalciteOptions` expose every `.vortex` file in a directory as a SQL table. ([#566](https://github.com/dfa1/vortex-java/issues/566))
+- `Compute.matcher` returns a `RowSelection`, whose `nextSelected(long)` jumps to the next selected row instead of testing each one. ([#580](https://github.com/dfa1/vortex-java/pull/580))
+- `Compute.matcher(RowFilter)` gives the row test of each chunk of one scan, evaluating a dictionary the chunks share once. ([#576](https://github.com/dfa1/vortex-java/pull/576))
- `DecimalArray.getUnscaledLong(long)` reads a decimal's unscaled value without building a `BigDecimal`. ([#563](https://github.com/dfa1/vortex-java/pull/563))
- `Compute.groupedAggregate(Chunk, ColumnName, RowFilter, ColumnName)` folds an aggregate per value of a dictionary-encoded string, boolean or integer column. ([#555](https://github.com/dfa1/vortex-java/pull/555))
- `Chunk.slice(offset, length)` gives a zero-copy view of a row range. ([#550](https://github.com/dfa1/vortex-java/pull/550))
@@ -32,6 +25,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
- Custom extension types: register an `ExtensionDecoder` on `ReadRegistry` and an `ExtensionEncoder` on `WriteRegistry`, with a `validate` hook on both sides, and `Chunk.as` decodes any registered type. ([#530](https://github.com/dfa1/vortex-java/issues/530))
### Changed
+- **Breaking:** `ArrayStats` has one component per field of Rust's array stats, so `minExact`, `maxExact`, `isSorted`, `isStrictSorted`, `isConstant`, `uncompressedSizeInBytes` and `nanCount` are read from the file and `trueCount` is gone (a bool column's true count is its `sum`). ([#581](https://github.com/dfa1/vortex-java/pull/581))
- SQL `WHERE` on a dictionary-encoded string column visits only the matching rows, up to a third faster for a selective `=` or `IN`. ([#580](https://github.com/dfa1/vortex-java/pull/580))
- SQL `WHERE col = 'x'` on a dictionary-encoded string column is about twice as fast, and skips the column's rows entirely when `x` is not in the dictionary. ([#576](https://github.com/dfa1/vortex-java/pull/576))
- SQL `SELECT DISTINCT` of one dictionary-encoded string column with `ORDER BY ... LIMIT` is about a third faster: it ranks the dictionary values the rows use instead of every row. ([#574](https://github.com/dfa1/vortex-java/pull/574))
@@ -65,6 +59,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
- **Breaking:** `ExtensionId` is a sealed `WellKnown`/`Custom` type like `EncodingId`, so third-party extension ids can be named, and `ExtensionId.parse` returns the id instead of an `Optional`. ([#530](https://github.com/dfa1/vortex-java/issues/530))
### Fixed
+- Files written by vortex-java mark their array min/max as exact, which Rust read as inexact bounds. ([#581](https://github.com/dfa1/vortex-java/pull/581))
+- `VortexReader.columnStats()` and SQL `MIN`/`MAX` no longer report a truncated (inexact) bound as a column's minimum or maximum. ([#581](https://github.com/dfa1/vortex-java/pull/581))
- A negative code in a nullable dictionary-encoded column fails with `VortexException` instead of a raw `IndexOutOfBoundsException`. ([#576](https://github.com/dfa1/vortex-java/pull/576))
- The CLI's terminal UI runs on JDK 28, whose arenas no longer over-align `allocate(size)`. ([8858286](https://github.com/dfa1/vortex-java/commit/8858286a1addb5ffca289c42e6106c7395564a57))
- vortex-jni no longer aborts the JVM when it filters a string column vortex-java compressed with FSST, whose symbol table could hold two symbols Rust hashes to one slot, and FSST columns are about 5% smaller. ([#573](https://github.com/dfa1/vortex-java/pull/573))
diff --git a/README.md b/README.md
index 0d686eb82..cf92b776e 100644
--- a/README.md
+++ b/README.md
@@ -17,6 +17,7 @@ and against 172 real-world [Raincloud](docs/compatibility.md#real-world-conforma
- JVM analytics engines and OLAP systems
- Anyone who wants mmap-backed, zero-copy columnar reads without native-library management
+- Anyone who wants to query `.vortex` files with SQL from IntelliJ, DBeaver or another JDBC tool
- **Small footprint.** The Rust reference's JNI bindings (`vortex-jni`) are a 75 MB jar
(native binaries for four platforms) that pulls in Arrow, Netty, Guava and RoaringBitmap.
vortex-java's reader is ~800 KB of plain Java (`vortex-reader` + `vortex-core` +
@@ -90,6 +91,27 @@ try (var ch = FileChannel.open(Path.of("data/example.vortex"),
> `IllegalArgumentException` immediately. Missing columns surface as
> `IllegalStateException` when the lambda returns.
+### Query with SQL from IntelliJ or DBeaver
+
+`vortex-jdbc-driver` is a read-only JDBC driver in one self-contained jar, so SQL tools can browse and
+query `.vortex` files with no code. Download `vortex-jdbc-driver-0.18.0-all.jar` from
+[Maven Central](https://central.sonatype.com/artifact/io.github.dfa1.vortex/vortex-jdbc-driver), add
+it as a driver with class `io.github.dfa1.vortex.jdbc.driver.VortexDriver`, and connect to a directory:
+
+```
+jdbc:vortex:/data/vortex
+```
+
+Every `*.vortex` file in it is a table of schema `vtx`, named after the file:
+
+```sql
+select symbol, count(*), max(high) from vtx.ohlc group by symbol order by 2 desc;
+```
+
+The tool must run on JDK 25 or later. Setup steps for IntelliJ and DBeaver, URL options and
+troubleshooting are in the
+[how-to](docs/how-to.md#query-vortex-files-from-an-sql-tool-intellij-dbeaver).
+
For more examples — projection, filtering, custom encodings, and the CLI — see
the [tutorial](docs/tutorial.md).
diff --git a/docs/how-to.md b/docs/how-to.md
index 74b8e7ea3..d509df861 100644
--- a/docs/how-to.md
+++ b/docs/how-to.md
@@ -643,6 +643,11 @@ select close, open from vtx.ohlc where `date` > 20000 limit 100;
other reserved words such as `close`, `open` and `value` work unquoted.
- **Filters and `LIMIT` are cheap.** A `WHERE` on a clustered column skips whole chunks using the
files' zone maps, and the data grid only reads the page it shows.
+- **Low-cardinality string columns filter fastest.** A `WHERE symbol = 'AAPL'` (or `IN`, `<>`, a
+ range) on a dictionary-encoded column is decided once per distinct value, then only the matching
+ rows are read; a value missing from the dictionary returns at once, without reading the rows.
+- **`ORDER BY ... LIMIT` does not sort the table.** The first rows are kept while scanning, so a
+ top-10 over millions of rows costs about one pass over the sort column.
- **A sort or grouping that returns thousands of values is handed to Calcite's own engine** instead
of being answered from the file's statistics. The result is the same, only slower. A huge result in
the grid is still paged.