Bump version: 0.24.2-beta.0 → 0.24.2-beta.1

Signed-off-by: BubbleCal <bubble-cal@outlook.com>

.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.20.1-beta.1"
+current_version = "0.21.2-beta.0"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

.github/workflows/npm-publish.yml

@@ -550,6 +550,9 @@ jobs:
           bash ci/update_lockfiles.sh
       - name: Push new commit
         uses: ad-m/github-push-action@master
+        with:
+          github_token: ${{ secrets.LANCEDB_RELEASE_TOKEN }}
+          branch: main
       - name: Notify Slack Action
         uses: ravsamhq/notify-slack-action@2.3.0
         if: ${{ always() }}

Cargo.toml

@@ -21,14 +21,16 @@ categories = ["database-implementations"]
 rust-version = "1.78.0"
 
 [workspace.dependencies]
-lance = { "version" = "=0.30.0", "features" = ["dynamodb"], tag = "v0.30.0-beta.1", git="https://github.com/lancedb/lance.git" }
-lance-io = { version = "=0.30.0", tag = "v0.30.0-beta.1", git="https://github.com/lancedb/lance.git" }
-lance-index = { version = "=0.30.0", tag = "v0.30.0-beta.1", git="https://github.com/lancedb/lance.git" }
-lance-linalg = { version = "=0.30.0", tag = "v0.30.0-beta.1", git="https://github.com/lancedb/lance.git" }
-lance-table = { version = "=0.30.0", tag = "v0.30.0-beta.1", git="https://github.com/lancedb/lance.git" }
-lance-testing = { version = "=0.30.0", tag = "v0.30.0-beta.1", git="https://github.com/lancedb/lance.git" }
-lance-datafusion = { version = "=0.30.0", tag = "v0.30.0-beta.1", git="https://github.com/lancedb/lance.git" }
-lance-encoding = { version = "=0.30.0", tag = "v0.30.0-beta.1", git="https://github.com/lancedb/lance.git" }
+lance = { "version" = "=0.31.2", "features" = [
+    "dynamodb",
+], "tag" = "v0.31.2-beta.3", "git" = "https://github.com/lancedb/lance.git" }
+lance-io = { "version" = "=0.31.2", "tag" = "v0.31.2-beta.3", "git" = "https://github.com/lancedb/lance.git" }
+lance-index = { "version" = "=0.31.2", "tag" = "v0.31.2-beta.3", "git" = "https://github.com/lancedb/lance.git" }
+lance-linalg = { "version" = "=0.31.2", "tag" = "v0.31.2-beta.3", "git" = "https://github.com/lancedb/lance.git" }
+lance-table = { "version" = "=0.31.2", "tag" = "v0.31.2-beta.3", "git" = "https://github.com/lancedb/lance.git" }
+lance-testing = { "version" = "=0.31.2", "tag" = "v0.31.2-beta.3", "git" = "https://github.com/lancedb/lance.git" }
+lance-datafusion = { "version" = "=0.31.2", "tag" = "v0.31.2-beta.3", "git" = "https://github.com/lancedb/lance.git" }
+lance-encoding = { "version" = "=0.31.2", "tag" = "v0.31.2-beta.3", "git" = "https://github.com/lancedb/lance.git" }
 # Note that this one does not include pyarrow
 arrow = { version = "55.1", optional = false }
 arrow-array = "55.1"
@@ -39,20 +41,20 @@ arrow-schema = "55.1"
 arrow-arith = "55.1"
 arrow-cast = "55.1"
 async-trait = "0"
-datafusion = { version = "47.0", default-features = false }
-datafusion-catalog = "47.0"
-datafusion-common = { version = "47.0", default-features = false }
-datafusion-execution = "47.0"
-datafusion-expr = "47.0"
-datafusion-physical-plan = "47.0"
+datafusion = { version = "48.0", default-features = false }
+datafusion-catalog = "48.0"
+datafusion-common = { version = "48.0", default-features = false }
+datafusion-execution = "48.0"
+datafusion-expr = "48.0"
+datafusion-physical-plan = "48.0"
 env_logger = "0.11"
-half = { "version" = "=2.5.0", default-features = false, features = [
+half = { "version" = "2.6.0", default-features = false, features = [
     "num-traits",
 ] }
 futures = "0"
 log = "0.4"
 moka = { version = "0.12", features = ["future"] }
-object_store = "0.11.0"
+object_store = "0.12.0"
 pin-project = "1.0.7"
 snafu = "0.8"
 url = "2"

ci/set_lance_version.py

@@ -47,10 +47,10 @@ def extract_features(line: str) -> list:
     """
     import re
 
-    match = re.search(r'"features"\s*=\s*\[(.*?)\]', line)
+    match = re.search(r'"features"\s*=\s*\[\s*(.*?)\s*\]', line, re.DOTALL)
     if match:
         features_str = match.group(1)
-        return [f.strip('"') for f in features_str.split(",")]
+        return [f.strip('"') for f in features_str.split(",") if len(f) > 0]
     return []
 
 
@@ -63,10 +63,24 @@ def update_cargo_toml(line_updater):
         lines = f.readlines()
 
     new_lines = []
+    lance_line = ""
+    is_parsing_lance_line = False
     for line in lines:
         if line.startswith("lance"):
             # Update the line using the provided function
-            new_lines.append(line_updater(line))
+            if line.strip().endswith("}"):
+                new_lines.append(line_updater(line))
+            else:
+                lance_line = line
+                is_parsing_lance_line = True
+        elif is_parsing_lance_line:
+            lance_line += line
+            if line.strip().endswith("}"):
+                new_lines.append(line_updater(lance_line))
+                lance_line = ""
+                is_parsing_lance_line = False
+            else:
+                print("doesn't end with }:", line)
         else:
             # Keep the line unchanged
             new_lines.append(line)

docs/package-lock.json

@@ -19,7 +19,7 @@
     },
     "../node": {
       "name": "vectordb",
-      "version": "0.12.0",
+      "version": "0.21.2-beta.0",
       "cpu": [
         "x64",
         "arm64"
@@ -65,11 +65,11 @@
         "uuid": "^9.0.0"
       },
       "optionalDependencies": {
-        "@lancedb/vectordb-darwin-arm64": "0.12.0",
-        "@lancedb/vectordb-darwin-x64": "0.12.0",
-        "@lancedb/vectordb-linux-arm64-gnu": "0.12.0",
-        "@lancedb/vectordb-linux-x64-gnu": "0.12.0",
-        "@lancedb/vectordb-win32-x64-msvc": "0.12.0"
+        "@lancedb/vectordb-darwin-arm64": "0.21.2-beta.0",
+        "@lancedb/vectordb-darwin-x64": "0.21.2-beta.0",
+        "@lancedb/vectordb-linux-arm64-gnu": "0.21.2-beta.0",
+        "@lancedb/vectordb-linux-x64-gnu": "0.21.2-beta.0",
+        "@lancedb/vectordb-win32-x64-msvc": "0.21.2-beta.0"
       },
       "peerDependencies": {
         "@apache-arrow/ts": "^14.0.2",

docs/src/guides/sql_querying.md

@@ -1,7 +1,9 @@
+# SQL Querying
+
 You can use DuckDB and Apache Datafusion to query your LanceDB tables using SQL.
 This guide will show how to query Lance tables them using both.
 
-We will re-use the dataset [created previously](./pandas_and_pyarrow.md):
+We will re-use the dataset [created previously](./tables.md):
 
 ```python
 import lancedb
@@ -27,15 +29,10 @@ arrow_table = table.to_lance()
 duckdb.query("SELECT * FROM arrow_table")
 ```
 
-```
-┌─────────────┬─────────┬────────┐
-│   vector    │  item   │ price  │
-│   float[]   │ varchar │ double │
-├─────────────┼─────────┼────────┤
-│ [3.1, 4.1]  │ foo     │   10.0 │
-│ [5.9, 26.5] │ bar     │   20.0 │
-└─────────────┴─────────┴────────┘
-```
+| vector      | item | price |
+| ----------- | ---- | ----- |
+| [3.1, 4.1]  | foo  | 10.0  |
+| [5.9, 26.5] | bar  | 20.0  |
 
 ## Querying a LanceDB Table with Apache Datafusion
 
@@ -57,12 +54,7 @@ Register the table created with the Datafusion session context.
     --8<-- "python/python/tests/docs/test_guide_tables.py:lance_sql_basic"
     ```
 
-```
-┌─────────────┬─────────┬────────┐
-│   vector    │  item   │ price  │
-│   float[]   │ varchar │ double │
-├─────────────┼─────────┼────────┤
-│ [3.1, 4.1]  │ foo     │   10.0 │
-│ [5.9, 26.5] │ bar     │   20.0 │
-└─────────────┴─────────┴────────┘
-```
+| vector      | item | price |
+| ----------- | ---- | ----- |
+| [3.1, 4.1]  | foo  | 10.0  |
+| [5.9, 26.5] | bar  | 20.0  |

docs/src/js/classes/MatchQuery.md

@@ -41,6 +41,7 @@ Creates an instance of MatchQuery.
     - `fuzziness`: The fuzziness level for the query (default is 0).
     - `maxExpansions`: The maximum number of terms to consider for fuzzy matching (default is 50).
     - `operator`: The logical operator to use for combining terms in the query (default is "OR").
+    - `prefixLength`: The number of beginning characters being unchanged for fuzzy matching.
 
 * **options.boost?**: `number`
 
@@ -50,6 +51,8 @@ Creates an instance of MatchQuery.
 
 * **options.operator?**: [`Operator`](../enumerations/Operator.md)
 
+* **options.prefixLength?**: `number`
+
 #### Returns
 
 [`MatchQuery`](MatchQuery.md)

docs/src/js/classes/Table.md

@@ -612,7 +612,7 @@ of the given query
 
 #### Parameters
 
-* **query**: `string` \| [`IntoVector`](../type-aliases/IntoVector.md) \| [`FullTextQuery`](../interfaces/FullTextQuery.md)
+* **query**: `string` \| [`IntoVector`](../type-aliases/IntoVector.md) \| [`MultiVector`](../type-aliases/MultiVector.md) \| [`FullTextQuery`](../interfaces/FullTextQuery.md)
     the query, a vector or string
 
 * **queryType?**: `string`
@@ -799,7 +799,7 @@ by `query`.
 
 #### Parameters
 
-* **vector**: [`IntoVector`](../type-aliases/IntoVector.md)
+* **vector**: [`IntoVector`](../type-aliases/IntoVector.md) \| [`MultiVector`](../type-aliases/MultiVector.md)
 
 #### Returns
 

docs/src/js/classes/VectorQuery.md

@@ -386,6 +386,53 @@ called then every valid row from the table will be returned.
 
 ***
 
+### maximumNprobes()
+
+```ts
+maximumNprobes(maximumNprobes): VectorQuery
+```
+
+Set the maximum number of probes used.
+
+This controls the maximum number of partitions that will be searched.  If this
+number is greater than minimumNprobes then the excess partitions will _only_ be
+searched if we have not found enough results.  This can be useful when there is
+a narrow filter to allow these queries to spend more time searching and avoid
+potential false negatives.
+
+#### Parameters
+
+* **maximumNprobes**: `number`
+
+#### Returns
+
+[`VectorQuery`](VectorQuery.md)
+
+***
+
+### minimumNprobes()
+
+```ts
+minimumNprobes(minimumNprobes): VectorQuery
+```
+
+Set the minimum number of probes used.
+
+This controls the minimum number of partitions that will be searched.  This
+parameter will impact every query against a vector index, regardless of the
+filter.  See `nprobes` for more details.  Higher values will increase recall
+but will also increase latency.
+
+#### Parameters
+
+* **minimumNprobes**: `number`
+
+#### Returns
+
+[`VectorQuery`](VectorQuery.md)
+
+***
+
 ### nprobes()
 
 ```ts
@@ -413,6 +460,10 @@ For best results we recommend tuning this parameter with a benchmark against
 your actual data to find the smallest possible value that will still give
 you the desired recall.
 
+For more fine grained control over behavior when you have a very narrow filter
+you can use `minimumNprobes` and `maximumNprobes`.  This method sets both
+the minimum and maximum to the same value.
+
 #### Parameters
 
 * **nprobes**: `number`

docs/src/js/enumerations/Occur.md

@@ -10,6 +10,7 @@ Enum representing the occurrence of terms in full-text queries.
 
 - `Must`: The term must be present in the document.
 - `Should`: The term should contribute to the document score, but is not required.
+- `MustNot`: The term must not be present in the document.
 
 ## Enumeration Members
 
@@ -21,6 +22,14 @@ Must: "MUST";
 
 ***
 
+### MustNot
+
+```ts
+MustNot: "MUST_NOT";
+```
+
+***
+
 ### Should
 
 ```ts

docs/src/js/globals.md

@@ -84,6 +84,7 @@
 - [FieldLike](type-aliases/FieldLike.md)
 - [IntoSql](type-aliases/IntoSql.md)
 - [IntoVector](type-aliases/IntoVector.md)
+- [MultiVector](type-aliases/MultiVector.md)
 - [RecordBatchLike](type-aliases/RecordBatchLike.md)
 - [SchemaLike](type-aliases/SchemaLike.md)
 - [TableLike](type-aliases/TableLike.md)

docs/src/js/interfaces/FtsOptions.md

@@ -23,7 +23,7 @@ whether to remove punctuation
 ### baseTokenizer?
 
 ```ts
-optional baseTokenizer: "raw" | "simple" | "whitespace";
+optional baseTokenizer: "raw" | "simple" | "whitespace" | "ngram";
 ```
 
 The tokenizer to use when building the index.
@@ -71,6 +71,36 @@ tokens longer than this length will be ignored
 
 ***
 
+### ngramMaxLength?
+
+```ts
+optional ngramMaxLength: number;
+```
+
+ngram max length
+
+***
+
+### ngramMinLength?
+
+```ts
+optional ngramMinLength: number;
+```
+
+ngram min length
+
+***
+
+### prefixOnly?
+
+```ts
+optional prefixOnly: boolean;
+```
+
+whether to only index the prefix of the token for ngram tokenizer
+
+***
+
 ### removeStopWords?
 
 ```ts

docs/src/js/interfaces/OptimizeOptions.md

@@ -24,10 +24,10 @@ The default is 7 days
 // Delete all versions older than 1 day
 const olderThan = new Date();
 olderThan.setDate(olderThan.getDate() - 1));
-tbl.cleanupOlderVersions(olderThan);
+tbl.optimize({cleanupOlderThan: olderThan});
 
 // Delete all versions except the current version
-tbl.cleanupOlderVersions(new Date());
+tbl.optimize({cleanupOlderThan: new Date()});
 ```
 
 ***

docs/src/js/type-aliases/MultiVector.md

@@ -0,0 +1,11 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / MultiVector
+
+# Type Alias: MultiVector
+
+```ts
+type MultiVector: IntoVector[];
+```

docs/src/notebooks/Multivector_on_LanceDB.ipynb

@@ -428,7 +428,7 @@
         "\n",
         "**Why?**  \n",
         "Embedding the UFO dataset and ingesting it into LanceDB takes **~2 hours on a T4 GPU**. To save time:  \n",
-        "- **Use the pre-prepared table with index created ** (provided below) to proceed directly to step7: search.  \n",
+        "- **Use the pre-prepared table with index created** (provided below) to proceed directly to **Step 7**: search.  \n",
         "- **Step 5a** contains the full ingestion code for reference (run it only if necessary).  \n",
         "- **Step 6** contains the details on creating the index on the multivector column"
       ]

docs/test/md_testing.py

@@ -30,7 +30,8 @@ excluded_globs = [
     "../src/rag/advanced_techniques/*.md",
     "../src/guides/scalar_index.md",
     "../src/guides/storage.md",
-    "../src/search.md"
+    "../src/search.md",
+    "../src/guides/sql_querying.md",
 ]
 
 python_prefix = "py"

java/.mvn/wrapper/maven-wrapper.properties

@@ -0,0 +1,19 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#   http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+wrapperVersion=3.3.2
+distributionType=only-script
+distributionUrl=https://repo.maven.apache.org/maven2/org/apache/maven/apache-maven/3.9.9/apache-maven-3.9.9-bin.zip

java/README.md

@@ -0,0 +1,37 @@
+# LanceDB Java SDK
+
+## Configuration and Initialization
+
+### LanceDB Cloud
+
+For LanceDB Cloud, use the simplified builder API:
+
+```java
+import com.lancedb.lance.namespace.LanceRestNamespace;
+
+// If your DB url is db://example-db, then your database here is example-db
+LanceRestNamespace namespace = LanceDBRestNamespaces.builder()
+    .apiKey("your_lancedb_cloud_api_key")
+    .database("your_database_name")
+    .build();
+```
+
+### LanceDB Enterprise
+
+For Enterprise deployments, use your VPC endpoint:
+
+```java
+LanceRestNamespace namespace = LanceDBRestNamespaces.builder()
+    .apiKey("your_lancedb_enterprise_api_key")
+    .database("your-top-dir") // Your top level folder under your cloud bucket, e.g. s3://your-bucket/your-top-dir/
+    .hostOverride("http://<vpc_endpoint_dns_name>:80")
+    .build();
+```
+
+## Development
+
+Build:
+
+```shell
+./mvnw install
+```

java/core/pom.xml

@@ -8,18 +8,24 @@
     <parent>
         <groupId>com.lancedb</groupId>
         <artifactId>lancedb-parent</artifactId>
-        <version>0.20.1-beta.1</version>
+        <version>0.21.2-beta.0</version>
         <relativePath>../pom.xml</relativePath>
     </parent>
 
     <artifactId>lancedb-core</artifactId>
-    <name>LanceDB Core</name>
+    <name>${project.artifactId}</name>
+    <description>LanceDB Core</description>
     <packaging>jar</packaging>
     <properties>
         <rust.release.build>false</rust.release.build>
     </properties>
 
     <dependencies>
+        <dependency>
+            <groupId>com.lancedb</groupId>
+            <artifactId>lance-namespace-core</artifactId>
+            <version>0.0.1</version>
+        </dependency>
         <dependency>
             <groupId>org.apache.arrow</groupId>
             <artifactId>arrow-vector</artifactId>

java/lance-namespace/pom.xml

@@ -0,0 +1,26 @@
+<?xml version="1.0" encoding="UTF-8"?>
+
+<project xmlns="http://maven.apache.org/POM/4.0.0"
+    xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
+    xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
+    <modelVersion>4.0.0</modelVersion>
+
+    <parent>
+        <groupId>com.lancedb</groupId>
+        <artifactId>lancedb-parent</artifactId>
+        <version>0.21.2-beta.0</version>
+        <relativePath>../pom.xml</relativePath>
+    </parent>
+
+    <artifactId>lancedb-lance-namespace</artifactId>
+    <name>${project.artifactId}</name>
+    <description>LanceDB Java Integration with Lance Namespace</description>
+    <packaging>jar</packaging>
+
+    <dependencies>
+        <dependency>
+            <groupId>com.lancedb</groupId>
+            <artifactId>lance-namespace-core</artifactId>
+        </dependency>
+    </dependencies>
+</project>

java/lance-namespace/src/main/java/com/lancedb/lancedb/LanceDbRestNamespaces.java

@@ -0,0 +1,146 @@
+/*
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package com.lancedb.lancedb;
+
+import com.lancedb.lance.namespace.LanceRestNamespace;
+import com.lancedb.lance.namespace.client.apache.ApiClient;
+
+import java.util.HashMap;
+import java.util.Map;
+import java.util.Optional;
+
+/** Util class to help construct a {@link LanceRestNamespace} for LanceDB. */
+public class LanceDbRestNamespaces {
+  private static final String DEFAULT_REGION = "us-east-1";
+  private static final String CLOUD_URL_PATTERN = "https://%s.%s.api.lancedb.com";
+
+  private String apiKey;
+  private String database;
+  private Optional<String> hostOverride = Optional.empty();
+  private Optional<String> region = Optional.empty();
+  private Map<String, String> additionalConfig = new HashMap<>();
+
+  private LanceDbRestNamespaces() {}
+
+  /**
+   * Create a new builder instance.
+   *
+   * @return A new LanceRestNamespaceBuilder
+   */
+  public static LanceDbRestNamespaces builder() {
+    return new LanceDbRestNamespaces();
+  }
+
+  /**
+   * Set the API key (required).
+   *
+   * @param apiKey The LanceDB API key
+   * @return This builder
+   */
+  public LanceDbRestNamespaces apiKey(String apiKey) {
+    if (apiKey == null || apiKey.trim().isEmpty()) {
+      throw new IllegalArgumentException("API key cannot be null or empty");
+    }
+    this.apiKey = apiKey;
+    return this;
+  }
+
+  /**
+   * Set the database name (required).
+   *
+   * @param database The database name
+   * @return This builder
+   */
+  public LanceDbRestNamespaces database(String database) {
+    if (database == null || database.trim().isEmpty()) {
+      throw new IllegalArgumentException("Database cannot be null or empty");
+    }
+    this.database = database;
+    return this;
+  }
+
+  /**
+   * Set a custom host override (optional). When set, this overrides the default LanceDB Cloud URL
+   * construction. Use this for LanceDB Enterprise deployments.
+   *
+   * @param hostOverride The complete base URL (e.g., "http://your-vpc-endpoint:80")
+   * @return This builder
+   */
+  public LanceDbRestNamespaces hostOverride(String hostOverride) {
+    this.hostOverride = Optional.ofNullable(hostOverride);
+    return this;
+  }
+
+  /**
+   * Set the region for LanceDB Cloud (optional). Defaults to "us-east-1" if not specified. This is
+   * ignored when hostOverride is set.
+   *
+   * @param region The AWS region (e.g., "us-east-1", "eu-west-1")
+   * @return This builder
+   */
+  public LanceDbRestNamespaces region(String region) {
+    this.region = Optional.ofNullable(region);
+    return this;
+  }
+
+  /**
+   * Add additional configuration parameters.
+   *
+   * @param key The configuration key
+   * @param value The configuration value
+   * @return This builder
+   */
+  public LanceDbRestNamespaces config(String key, String value) {
+    this.additionalConfig.put(key, value);
+    return this;
+  }
+
+  /**
+   * Build the LanceRestNamespace instance.
+   *
+   * @return A configured LanceRestNamespace
+   * @throws IllegalStateException if required parameters are missing
+   */
+  public LanceRestNamespace build() {
+    // Validate required fields
+    if (apiKey == null) {
+      throw new IllegalStateException("API key is required");
+    }
+    if (database == null) {
+      throw new IllegalStateException("Database is required");
+    }
+
+    // Build configuration map
+    Map<String, String> config = new HashMap<>(additionalConfig);
+    config.put("headers.x-lancedb-database", database);
+    config.put("headers.x-api-key", apiKey);
+
+    // Determine base URL
+    String baseUrl;
+    if (hostOverride.isPresent()) {
+      baseUrl = hostOverride.get();
+      config.put("host_override", hostOverride.get());
+    } else {
+      String effectiveRegion = region.orElse(DEFAULT_REGION);
+      baseUrl = String.format(CLOUD_URL_PATTERN, database, effectiveRegion);
+      config.put("region", effectiveRegion);
+    }
+
+    // Create and configure ApiClient
+    ApiClient apiClient = new ApiClient();
+    apiClient.setBasePath(baseUrl);
+
+    return new LanceRestNamespace(apiClient, config);
+  }
+}

java/mvnw

@@ -0,0 +1,259 @@
+#!/bin/sh
+# ----------------------------------------------------------------------------
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements.  See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership.  The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License.  You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied.  See the License for the
+# specific language governing permissions and limitations
+# under the License.
+# ----------------------------------------------------------------------------
+
+# ----------------------------------------------------------------------------
+# Apache Maven Wrapper startup batch script, version 3.3.2
+#
+# Optional ENV vars
+# -----------------
+#   JAVA_HOME - location of a JDK home dir, required when download maven via java source
+#   MVNW_REPOURL - repo url base for downloading maven distribution
+#   MVNW_USERNAME/MVNW_PASSWORD - user and password for downloading maven
+#   MVNW_VERBOSE - true: enable verbose log; debug: trace the mvnw script; others: silence the output
+# ----------------------------------------------------------------------------
+
+set -euf
+[ "${MVNW_VERBOSE-}" != debug ] || set -x
+
+# OS specific support.
+native_path() { printf %s\\n "$1"; }
+case "$(uname)" in
+CYGWIN* | MINGW*)
+  [ -z "${JAVA_HOME-}" ] || JAVA_HOME="$(cygpath --unix "$JAVA_HOME")"
+  native_path() { cygpath --path --windows "$1"; }
+  ;;
+esac
+
+# set JAVACMD and JAVACCMD
+set_java_home() {
+  # For Cygwin and MinGW, ensure paths are in Unix format before anything is touched
+  if [ -n "${JAVA_HOME-}" ]; then
+    if [ -x "$JAVA_HOME/jre/sh/java" ]; then
+      # IBM's JDK on AIX uses strange locations for the executables
+      JAVACMD="$JAVA_HOME/jre/sh/java"
+      JAVACCMD="$JAVA_HOME/jre/sh/javac"
+    else
+      JAVACMD="$JAVA_HOME/bin/java"
+      JAVACCMD="$JAVA_HOME/bin/javac"
+
+      if [ ! -x "$JAVACMD" ] || [ ! -x "$JAVACCMD" ]; then
+        echo "The JAVA_HOME environment variable is not defined correctly, so mvnw cannot run." >&2
+        echo "JAVA_HOME is set to \"$JAVA_HOME\", but \"\$JAVA_HOME/bin/java\" or \"\$JAVA_HOME/bin/javac\" does not exist." >&2
+        return 1
+      fi
+    fi
+  else
+    JAVACMD="$(
+      'set' +e
+      'unset' -f command 2>/dev/null
+      'command' -v java
+    )" || :
+    JAVACCMD="$(
+      'set' +e
+      'unset' -f command 2>/dev/null
+      'command' -v javac
+    )" || :
+
+    if [ ! -x "${JAVACMD-}" ] || [ ! -x "${JAVACCMD-}" ]; then
+      echo "The java/javac command does not exist in PATH nor is JAVA_HOME set, so mvnw cannot run." >&2
+      return 1
+    fi
+  fi
+}
+
+# hash string like Java String::hashCode
+hash_string() {
+  str="${1:-}" h=0
+  while [ -n "$str" ]; do
+    char="${str%"${str#?}"}"
+    h=$(((h * 31 + $(LC_CTYPE=C printf %d "'$char")) % 4294967296))
+    str="${str#?}"
+  done
+  printf %x\\n $h
+}
+
+verbose() { :; }
+[ "${MVNW_VERBOSE-}" != true ] || verbose() { printf %s\\n "${1-}"; }
+
+die() {
+  printf %s\\n "$1" >&2
+  exit 1
+}
+
+trim() {
+  # MWRAPPER-139:
+  #   Trims trailing and leading whitespace, carriage returns, tabs, and linefeeds.
+  #   Needed for removing poorly interpreted newline sequences when running in more
+  #   exotic environments such as mingw bash on Windows.
+  printf "%s" "${1}" | tr -d '[:space:]'
+}
+
+# parse distributionUrl and optional distributionSha256Sum, requires .mvn/wrapper/maven-wrapper.properties
+while IFS="=" read -r key value; do
+  case "${key-}" in
+  distributionUrl) distributionUrl=$(trim "${value-}") ;;
+  distributionSha256Sum) distributionSha256Sum=$(trim "${value-}") ;;
+  esac
+done <"${0%/*}/.mvn/wrapper/maven-wrapper.properties"
+[ -n "${distributionUrl-}" ] || die "cannot read distributionUrl property in ${0%/*}/.mvn/wrapper/maven-wrapper.properties"
+
+case "${distributionUrl##*/}" in
+maven-mvnd-*bin.*)
+  MVN_CMD=mvnd.sh _MVNW_REPO_PATTERN=/maven/mvnd/
+  case "${PROCESSOR_ARCHITECTURE-}${PROCESSOR_ARCHITEW6432-}:$(uname -a)" in
+  *AMD64:CYGWIN* | *AMD64:MINGW*) distributionPlatform=windows-amd64 ;;
+  :Darwin*x86_64) distributionPlatform=darwin-amd64 ;;
+  :Darwin*arm64) distributionPlatform=darwin-aarch64 ;;
+  :Linux*x86_64*) distributionPlatform=linux-amd64 ;;
+  *)
+    echo "Cannot detect native platform for mvnd on $(uname)-$(uname -m), use pure java version" >&2
+    distributionPlatform=linux-amd64
+    ;;
+  esac
+  distributionUrl="${distributionUrl%-bin.*}-$distributionPlatform.zip"
+  ;;
+maven-mvnd-*) MVN_CMD=mvnd.sh _MVNW_REPO_PATTERN=/maven/mvnd/ ;;
+*) MVN_CMD="mvn${0##*/mvnw}" _MVNW_REPO_PATTERN=/org/apache/maven/ ;;
+esac
+
+# apply MVNW_REPOURL and calculate MAVEN_HOME
+# maven home pattern: ~/.m2/wrapper/dists/{apache-maven-<version>,maven-mvnd-<version>-<platform>}/<hash>
+[ -z "${MVNW_REPOURL-}" ] || distributionUrl="$MVNW_REPOURL$_MVNW_REPO_PATTERN${distributionUrl#*"$_MVNW_REPO_PATTERN"}"
+distributionUrlName="${distributionUrl##*/}"
+distributionUrlNameMain="${distributionUrlName%.*}"
+distributionUrlNameMain="${distributionUrlNameMain%-bin}"
+MAVEN_USER_HOME="${MAVEN_USER_HOME:-${HOME}/.m2}"
+MAVEN_HOME="${MAVEN_USER_HOME}/wrapper/dists/${distributionUrlNameMain-}/$(hash_string "$distributionUrl")"
+
+exec_maven() {
+  unset MVNW_VERBOSE MVNW_USERNAME MVNW_PASSWORD MVNW_REPOURL || :
+  exec "$MAVEN_HOME/bin/$MVN_CMD" "$@" || die "cannot exec $MAVEN_HOME/bin/$MVN_CMD"
+}
+
+if [ -d "$MAVEN_HOME" ]; then
+  verbose "found existing MAVEN_HOME at $MAVEN_HOME"
+  exec_maven "$@"
+fi
+
+case "${distributionUrl-}" in
+*?-bin.zip | *?maven-mvnd-?*-?*.zip) ;;
+*) die "distributionUrl is not valid, must match *-bin.zip or maven-mvnd-*.zip, but found '${distributionUrl-}'" ;;
+esac
+
+# prepare tmp dir
+if TMP_DOWNLOAD_DIR="$(mktemp -d)" && [ -d "$TMP_DOWNLOAD_DIR" ]; then
+  clean() { rm -rf -- "$TMP_DOWNLOAD_DIR"; }
+  trap clean HUP INT TERM EXIT
+else
+  die "cannot create temp dir"
+fi
+
+mkdir -p -- "${MAVEN_HOME%/*}"
+
+# Download and Install Apache Maven
+verbose "Couldn't find MAVEN_HOME, downloading and installing it ..."
+verbose "Downloading from: $distributionUrl"
+verbose "Downloading to: $TMP_DOWNLOAD_DIR/$distributionUrlName"
+
+# select .zip or .tar.gz
+if ! command -v unzip >/dev/null; then
+  distributionUrl="${distributionUrl%.zip}.tar.gz"
+  distributionUrlName="${distributionUrl##*/}"
+fi
+
+# verbose opt
+__MVNW_QUIET_WGET=--quiet __MVNW_QUIET_CURL=--silent __MVNW_QUIET_UNZIP=-q __MVNW_QUIET_TAR=''
+[ "${MVNW_VERBOSE-}" != true ] || __MVNW_QUIET_WGET='' __MVNW_QUIET_CURL='' __MVNW_QUIET_UNZIP='' __MVNW_QUIET_TAR=v
+
+# normalize http auth
+case "${MVNW_PASSWORD:+has-password}" in
+'') MVNW_USERNAME='' MVNW_PASSWORD='' ;;
+has-password) [ -n "${MVNW_USERNAME-}" ] || MVNW_USERNAME='' MVNW_PASSWORD='' ;;
+esac
+
+if [ -z "${MVNW_USERNAME-}" ] && command -v wget >/dev/null; then
+  verbose "Found wget ... using wget"
+  wget ${__MVNW_QUIET_WGET:+"$__MVNW_QUIET_WGET"} "$distributionUrl" -O "$TMP_DOWNLOAD_DIR/$distributionUrlName" || die "wget: Failed to fetch $distributionUrl"
+elif [ -z "${MVNW_USERNAME-}" ] && command -v curl >/dev/null; then
+  verbose "Found curl ... using curl"
+  curl ${__MVNW_QUIET_CURL:+"$__MVNW_QUIET_CURL"} -f -L -o "$TMP_DOWNLOAD_DIR/$distributionUrlName" "$distributionUrl" || die "curl: Failed to fetch $distributionUrl"
+elif set_java_home; then
+  verbose "Falling back to use Java to download"
+  javaSource="$TMP_DOWNLOAD_DIR/Downloader.java"
+  targetZip="$TMP_DOWNLOAD_DIR/$distributionUrlName"
+  cat >"$javaSource" <<-END
+	public class Downloader extends java.net.Authenticator
+	{
+	  protected java.net.PasswordAuthentication getPasswordAuthentication()
+	  {
+	    return new java.net.PasswordAuthentication( System.getenv( "MVNW_USERNAME" ), System.getenv( "MVNW_PASSWORD" ).toCharArray() );
+	  }
+	  public static void main( String[] args ) throws Exception
+	  {
+	    setDefault( new Downloader() );
+	    java.nio.file.Files.copy( java.net.URI.create( args[0] ).toURL().openStream(), java.nio.file.Paths.get( args[1] ).toAbsolutePath().normalize() );
+	  }
+	}
+	END
+  # For Cygwin/MinGW, switch paths to Windows format before running javac and java
+  verbose " - Compiling Downloader.java ..."
+  "$(native_path "$JAVACCMD")" "$(native_path "$javaSource")" || die "Failed to compile Downloader.java"
+  verbose " - Running Downloader.java ..."
+  "$(native_path "$JAVACMD")" -cp "$(native_path "$TMP_DOWNLOAD_DIR")" Downloader "$distributionUrl" "$(native_path "$targetZip")"
+fi
+
+# If specified, validate the SHA-256 sum of the Maven distribution zip file
+if [ -n "${distributionSha256Sum-}" ]; then
+  distributionSha256Result=false
+  if [ "$MVN_CMD" = mvnd.sh ]; then
+    echo "Checksum validation is not supported for maven-mvnd." >&2
+    echo "Please disable validation by removing 'distributionSha256Sum' from your maven-wrapper.properties." >&2
+    exit 1
+  elif command -v sha256sum >/dev/null; then
+    if echo "$distributionSha256Sum  $TMP_DOWNLOAD_DIR/$distributionUrlName" | sha256sum -c >/dev/null 2>&1; then
+      distributionSha256Result=true
+    fi
+  elif command -v shasum >/dev/null; then
+    if echo "$distributionSha256Sum  $TMP_DOWNLOAD_DIR/$distributionUrlName" | shasum -a 256 -c >/dev/null 2>&1; then
+      distributionSha256Result=true
+    fi
+  else
+    echo "Checksum validation was requested but neither 'sha256sum' or 'shasum' are available." >&2
+    echo "Please install either command, or disable validation by removing 'distributionSha256Sum' from your maven-wrapper.properties." >&2
+    exit 1
+  fi
+  if [ $distributionSha256Result = false ]; then
+    echo "Error: Failed to validate Maven distribution SHA-256, your Maven distribution might be compromised." >&2
+    echo "If you updated your Maven version, you need to update the specified distributionSha256Sum property." >&2
+    exit 1
+  fi
+fi
+
+# unzip and move
+if command -v unzip >/dev/null; then
+  unzip ${__MVNW_QUIET_UNZIP:+"$__MVNW_QUIET_UNZIP"} "$TMP_DOWNLOAD_DIR/$distributionUrlName" -d "$TMP_DOWNLOAD_DIR" || die "failed to unzip"
+else
+  tar xzf${__MVNW_QUIET_TAR:+"$__MVNW_QUIET_TAR"} "$TMP_DOWNLOAD_DIR/$distributionUrlName" -C "$TMP_DOWNLOAD_DIR" || die "failed to untar"
+fi
+printf %s\\n "$distributionUrl" >"$TMP_DOWNLOAD_DIR/$distributionUrlNameMain/mvnw.url"
+mv -- "$TMP_DOWNLOAD_DIR/$distributionUrlNameMain" "$MAVEN_HOME" || [ -d "$MAVEN_HOME" ] || die "fail to move MAVEN_HOME"
+
+clean || :
+exec_maven "$@"

java/pom.xml

@@ -6,11 +6,10 @@
 
     <groupId>com.lancedb</groupId>
     <artifactId>lancedb-parent</artifactId>
-    <version>0.20.1-beta.1</version>
+    <version>0.21.2-beta.0</version>
     <packaging>pom</packaging>
-
-    <name>LanceDB Parent</name>
-    <description>LanceDB vector database Java API</description>
+    <name>${project.artifactId}</name>
+    <description>LanceDB Java SDK Parent POM</description>
     <url>http://lancedb.com/</url>
 
     <developers>
@@ -29,6 +28,7 @@
     <properties>
         <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
         <arrow.version>15.0.0</arrow.version>
+        <lance-namespace.verison>0.0.1</lance-namespace.verison>
         <spotless.skip>false</spotless.skip>
         <spotless.version>2.30.0</spotless.version>
         <spotless.java.googlejavaformat.version>1.7</spotless.java.googlejavaformat.version>
@@ -52,6 +52,7 @@
 
     <modules>
         <module>core</module>
+        <module>lance-namespace</module>
     </modules>
 
     <scm>
@@ -62,6 +63,11 @@
 
     <dependencyManagement>
         <dependencies>
+            <dependency>
+                <groupId>com.lancedb</groupId>
+                <artifactId>lance-namespace-core</artifactId>
+                <version>${lance-namespace.verison}</version>
+            </dependency>
             <dependency>
                 <groupId>org.apache.arrow</groupId>
                 <artifactId>arrow-vector</artifactId>

node/package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "vectordb",
-  "version": "0.20.1-beta.1",
+  "version": "0.21.2-beta.0",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "vectordb",
-      "version": "0.20.1-beta.1",
+      "version": "0.21.2-beta.0",
       "cpu": [
         "x64",
         "arm64"
@@ -52,11 +52,11 @@
         "uuid": "^9.0.0"
       },
       "optionalDependencies": {
-        "@lancedb/vectordb-darwin-arm64": "0.20.1-beta.1",
-        "@lancedb/vectordb-darwin-x64": "0.20.1-beta.1",
-        "@lancedb/vectordb-linux-arm64-gnu": "0.20.1-beta.1",
-        "@lancedb/vectordb-linux-x64-gnu": "0.20.1-beta.1",
-        "@lancedb/vectordb-win32-x64-msvc": "0.20.1-beta.1"
+        "@lancedb/vectordb-darwin-arm64": "0.21.2-beta.0",
+        "@lancedb/vectordb-darwin-x64": "0.21.2-beta.0",
+        "@lancedb/vectordb-linux-arm64-gnu": "0.21.2-beta.0",
+        "@lancedb/vectordb-linux-x64-gnu": "0.21.2-beta.0",
+        "@lancedb/vectordb-win32-x64-msvc": "0.21.2-beta.0"
       },
       "peerDependencies": {
         "@apache-arrow/ts": "^14.0.2",
@@ -327,60 +327,65 @@
       }
     },
     "node_modules/@lancedb/vectordb-darwin-arm64": {
-      "version": "0.20.1-beta.1",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-arm64/-/vectordb-darwin-arm64-0.20.1-beta.1.tgz",
-      "integrity": "sha512-DPD8gwFQz5aENYYbTFS/l3YX/rqzS6Kj2B4IZERccVFULQsdR5YwtaAfFwTMp7NSnsjWKwJAknohiMZlJr4njQ==",
+      "version": "0.21.2-beta.0",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-arm64/-/vectordb-darwin-arm64-0.21.2-beta.0.tgz",
+      "integrity": "sha512-RiYqpKuq9v8A4wFuHt1iPNFYjWJ1KgGFLJwQO4ajp9Hee84sDHq8mP0ATgMcc24hiaOUQ1lRRTULjGbHn4NIYw==",
       "cpu": [
         "arm64"
       ],
+      "license": "Apache-2.0",
       "optional": true,
       "os": [
         "darwin"
       ]
     },
     "node_modules/@lancedb/vectordb-darwin-x64": {
-      "version": "0.20.1-beta.1",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-x64/-/vectordb-darwin-x64-0.20.1-beta.1.tgz",
-      "integrity": "sha512-lTPtlRSTC08UgQW5Bv8WYhdbogAgUJ+9ejg+UE+fwP9gEsgEKXL/SHBm+9gmAlTo7LbrxJjg0CtCde/mW68UTw==",
+      "version": "0.21.2-beta.0",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-x64/-/vectordb-darwin-x64-0.21.2-beta.0.tgz",
+      "integrity": "sha512-togdP0YIjMYg/hBRMMxW434i5VB789JWU5o3hWrodbX8olEc0Txqw5Dg9CgIOldBIiCti6uTSQiTo6uldZon1w==",
       "cpu": [
         "x64"
       ],
+      "license": "Apache-2.0",
       "optional": true,
       "os": [
         "darwin"
       ]
     },
     "node_modules/@lancedb/vectordb-linux-arm64-gnu": {
-      "version": "0.20.1-beta.1",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-arm64-gnu/-/vectordb-linux-arm64-gnu-0.20.1-beta.1.tgz",
-      "integrity": "sha512-w/3O9FvwQiGegYsM21yZ0FezfOFVsW7HttYwwPzZMZaCpK3/i+LvZVSqwO4qXHHJBtHgKevonINyvVlg5487aQ==",
+      "version": "0.21.2-beta.0",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-arm64-gnu/-/vectordb-linux-arm64-gnu-0.21.2-beta.0.tgz",
+      "integrity": "sha512-ErS4IQDQVTYVATPeOj/dZXQR34eZQ5rAXm3vJdQi5K6X4zCDaIjOhpmnwzPBGT9W1idaBAoDJhtNfsFaJ6/PQQ==",
       "cpu": [
         "arm64"
       ],
+      "license": "Apache-2.0",
       "optional": true,
       "os": [
         "linux"
       ]
     },
     "node_modules/@lancedb/vectordb-linux-x64-gnu": {
-      "version": "0.20.1-beta.1",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-x64-gnu/-/vectordb-linux-x64-gnu-0.20.1-beta.1.tgz",
-      "integrity": "sha512-rq7Q6Lq9kJmBcgwplYQVJmRbyeP+xPVmXyyQfAO3IjekqeSsyjj1HoCZYqZIfBZyN5ELiSvIJB0731aKf9pr1A==",
+      "version": "0.21.2-beta.0",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-x64-gnu/-/vectordb-linux-x64-gnu-0.21.2-beta.0.tgz",
+      "integrity": "sha512-ycDpyBGbfxtnGGa/RQo5+So6dHALiem1pbYc/LDKKluUJpadtXtEwC61o6hZTcejoYjhEE8ET7vA3OCEJfMFaw==",
       "cpu": [
         "x64"
       ],
+      "license": "Apache-2.0",
       "optional": true,
       "os": [
         "linux"
       ]
     },
     "node_modules/@lancedb/vectordb-win32-x64-msvc": {
-      "version": "0.20.1-beta.1",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-win32-x64-msvc/-/vectordb-win32-x64-msvc-0.20.1-beta.1.tgz",
-      "integrity": "sha512-kHra0SEXeMKdgqi5h0igsqHcBr73hKBhEVJBa8VTv1DUv6Jvazwl4B4ueqllcyD4k3vvOTb2XzZomm7dhQ9QnA==",
+      "version": "0.21.2-beta.0",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-win32-x64-msvc/-/vectordb-win32-x64-msvc-0.21.2-beta.0.tgz",
+      "integrity": "sha512-IgVkAP/LiNIQD5P6n/9x3bgQOt5pGJarjtSF8r+ialD95QHmo6tcxrwTy/DlA+H1uI6B6h+sbN0c1KXTh1rYcg==",
       "cpu": [
         "x64"
       ],
+      "license": "Apache-2.0",
       "optional": true,
       "os": [
         "win32"

node/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vectordb",
-  "version": "0.20.1-beta.1",
+  "version": "0.21.2-beta.0",
   "description": " Serverless, low-latency vector database for AI applications",
   "private": false,
   "main": "dist/index.js",
@@ -89,10 +89,10 @@
     }
   },
   "optionalDependencies": {
-    "@lancedb/vectordb-darwin-x64": "0.20.1-beta.1",
-    "@lancedb/vectordb-darwin-arm64": "0.20.1-beta.1",
-    "@lancedb/vectordb-linux-x64-gnu": "0.20.1-beta.1",
-    "@lancedb/vectordb-linux-arm64-gnu": "0.20.1-beta.1",
-    "@lancedb/vectordb-win32-x64-msvc": "0.20.1-beta.1"
+    "@lancedb/vectordb-darwin-x64": "0.21.2-beta.0",
+    "@lancedb/vectordb-darwin-arm64": "0.21.2-beta.0",
+    "@lancedb/vectordb-linux-x64-gnu": "0.21.2-beta.0",
+    "@lancedb/vectordb-linux-arm64-gnu": "0.21.2-beta.0",
+    "@lancedb/vectordb-win32-x64-msvc": "0.21.2-beta.0"
   }
 }

nodejs/Cargo.toml

@@ -1,7 +1,7 @@
 [package]
 name = "lancedb-nodejs"
 edition.workspace = true
-version = "0.20.1-beta.1"
+version = "0.21.2-beta.0"
 license.workspace = true
 description.workspace = true
 repository.workspace = true

nodejs/__test__/arrow.test.ts

@@ -1,7 +1,7 @@
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
-import { Schema } from "apache-arrow";
+import { Bool, Field, Int32, List, Schema, Struct, Utf8 } from "apache-arrow";
 
 import * as arrow15 from "apache-arrow-15";
 import * as arrow16 from "apache-arrow-16";
@@ -11,10 +11,12 @@ import * as arrow18 from "apache-arrow-18";
 import {
   convertToTable,
   fromBufferToRecordBatch,
+  fromDataToBuffer,
   fromRecordBatchToBuffer,
   fromTableToBuffer,
   makeArrowTable,
   makeEmptyTable,
+  tableFromIPC,
 } from "../lancedb/arrow";
 import {
   EmbeddingFunction,
@@ -375,8 +377,221 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
         expect(table2.schema).toEqual(schema);
       });
 
+      it("will handle missing columns in schema alignment when using embeddings", async function () {
+        const schema = new Schema(
+          [
+            new Field("domain", new Utf8(), true),
+            new Field("name", new Utf8(), true),
+            new Field("description", new Utf8(), true),
+          ],
+          new Map([["embedding_functions", JSON.stringify([])]]),
+        );
+
+        const data = [
+          { domain: "google.com", name: "Google" },
+          { domain: "facebook.com", name: "Facebook" },
+        ];
+
+        const table = await convertToTable(data, undefined, { schema });
+
+        expect(table.numCols).toBe(3);
+        expect(table.numRows).toBe(2);
+
+        const descriptionColumn = table.getChild("description");
+        expect(descriptionColumn).toBeDefined();
+        expect(descriptionColumn?.nullCount).toBe(2);
+        expect(descriptionColumn?.toArray()).toEqual([null, null]);
+
+        expect(table.getChild("domain")?.toArray()).toEqual([
+          "google.com",
+          "facebook.com",
+        ]);
+        expect(table.getChild("name")?.toArray()).toEqual([
+          "Google",
+          "Facebook",
+        ]);
+      });
+
+      it("will handle completely missing nested struct columns", async function () {
+        const schema = new Schema(
+          [
+            new Field("id", new Utf8(), true),
+            new Field("name", new Utf8(), true),
+            new Field(
+              "metadata",
+              new Struct([
+                new Field("version", new Int32(), true),
+                new Field("author", new Utf8(), true),
+                new Field(
+                  "tags",
+                  new List(new Field("item", new Utf8(), true)),
+                  true,
+                ),
+              ]),
+              true,
+            ),
+          ],
+          new Map([["embedding_functions", JSON.stringify([])]]),
+        );
+
+        const data = [
+          { id: "doc1", name: "Document 1" },
+          { id: "doc2", name: "Document 2" },
+        ];
+
+        const table = await convertToTable(data, undefined, { schema });
+
+        expect(table.numCols).toBe(3);
+        expect(table.numRows).toBe(2);
+
+        const buf = await fromTableToBuffer(table);
+        const retrievedTable = tableFromIPC(buf);
+
+        const rows = [];
+        for (let i = 0; i < retrievedTable.numRows; i++) {
+          rows.push(retrievedTable.get(i));
+        }
+
+        expect(rows[0].metadata.version).toBe(null);
+        expect(rows[0].metadata.author).toBe(null);
+        expect(rows[0].metadata.tags).toBe(null);
+        expect(rows[0].id).toBe("doc1");
+        expect(rows[0].name).toBe("Document 1");
+      });
+
+      it("will handle partially missing nested struct fields", async function () {
+        const schema = new Schema(
+          [
+            new Field("id", new Utf8(), true),
+            new Field(
+              "metadata",
+              new Struct([
+                new Field("version", new Int32(), true),
+                new Field("author", new Utf8(), true),
+                new Field("created_at", new Utf8(), true),
+              ]),
+              true,
+            ),
+          ],
+          new Map([["embedding_functions", JSON.stringify([])]]),
+        );
+
+        const data = [
+          { id: "doc1", metadata: { version: 1, author: "Alice" } },
+          { id: "doc2", metadata: { version: 2 } },
+        ];
+
+        const table = await convertToTable(data, undefined, { schema });
+
+        expect(table.numCols).toBe(2);
+        expect(table.numRows).toBe(2);
+
+        const metadataColumn = table.getChild("metadata");
+        expect(metadataColumn).toBeDefined();
+        expect(metadataColumn?.type.toString()).toBe(
+          "Struct<{version:Int32, author:Utf8, created_at:Utf8}>",
+        );
+      });
+
+      it("will handle multiple levels of nested structures", async function () {
+        const schema = new Schema(
+          [
+            new Field("id", new Utf8(), true),
+            new Field(
+              "config",
+              new Struct([
+                new Field("database", new Utf8(), true),
+                new Field(
+                  "connection",
+                  new Struct([
+                    new Field("host", new Utf8(), true),
+                    new Field("port", new Int32(), true),
+                    new Field(
+                      "ssl",
+                      new Struct([
+                        new Field("enabled", new Bool(), true),
+                        new Field("cert_path", new Utf8(), true),
+                      ]),
+                      true,
+                    ),
+                  ]),
+                  true,
+                ),
+              ]),
+              true,
+            ),
+          ],
+          new Map([["embedding_functions", JSON.stringify([])]]),
+        );
+
+        const data = [
+          {
+            id: "config1",
+            config: {
+              database: "postgres",
+              connection: { host: "localhost" },
+            },
+          },
+          {
+            id: "config2",
+            config: { database: "mysql" },
+          },
+          {
+            id: "config3",
+          },
+        ];
+
+        const table = await convertToTable(data, undefined, { schema });
+
+        expect(table.numCols).toBe(2);
+        expect(table.numRows).toBe(3);
+
+        const configColumn = table.getChild("config");
+        expect(configColumn).toBeDefined();
+        expect(configColumn?.type.toString()).toBe(
+          "Struct<{database:Utf8, connection:Struct<{host:Utf8, port:Int32, ssl:Struct<{enabled:Bool, cert_path:Utf8}>}>}>",
+        );
+      });
+
+      it("will handle missing columns in Arrow table input when using embeddings", async function () {
+        const incompleteTable = makeArrowTable([
+          { domain: "google.com", name: "Google" },
+          { domain: "facebook.com", name: "Facebook" },
+        ]);
+
+        const schema = new Schema(
+          [
+            new Field("domain", new Utf8(), true),
+            new Field("name", new Utf8(), true),
+            new Field("description", new Utf8(), true),
+          ],
+          new Map([["embedding_functions", JSON.stringify([])]]),
+        );
+
+        const buf = await fromDataToBuffer(incompleteTable, undefined, schema);
+
+        expect(buf.byteLength).toBeGreaterThan(0);
+
+        const retrievedTable = tableFromIPC(buf);
+        expect(retrievedTable.numCols).toBe(3);
+        expect(retrievedTable.numRows).toBe(2);
+
+        const descriptionColumn = retrievedTable.getChild("description");
+        expect(descriptionColumn).toBeDefined();
+        expect(descriptionColumn?.nullCount).toBe(2);
+        expect(descriptionColumn?.toArray()).toEqual([null, null]);
+
+        expect(retrievedTable.getChild("domain")?.toArray()).toEqual([
+          "google.com",
+          "facebook.com",
+        ]);
+        expect(retrievedTable.getChild("name")?.toArray()).toEqual([
+          "Google",
+          "Facebook",
+        ]);
+      });
+
       it("should correctly retain values in nested struct fields", async function () {
-        // Define test data with nested struct
         const testData = [
           {
             id: "doc1",
@@ -400,10 +615,8 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
           },
         ];
 
-        // Create Arrow table from the data
         const table = makeArrowTable(testData);
 
-        // Verify schema has the nested struct fields
         const metadataField = table.schema.fields.find(
           (f) => f.name === "metadata",
         );
@@ -417,23 +630,17 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
           "text",
         ]);
 
-        // Convert to buffer and back (simulating storage and retrieval)
         const buf = await fromTableToBuffer(table);
         const retrievedTable = tableFromIPC(buf);
 
-        // Verify the retrieved table has the same structure
         const rows = [];
         for (let i = 0; i < retrievedTable.numRows; i++) {
           rows.push(retrievedTable.get(i));
         }
 
-        // Check values in the first row
         const firstRow = rows[0];
         expect(firstRow.id).toBe("doc1");
         expect(firstRow.vector.toJSON()).toEqual([1, 2, 3]);
-
-        // Verify metadata values are preserved (this is where the bug is)
-        expect(firstRow.metadata).toBeDefined();
         expect(firstRow.metadata.filePath).toBe("/path/to/file1.ts");
         expect(firstRow.metadata.startLine).toBe(10);
         expect(firstRow.metadata.endLine).toBe(20);

nodejs/__test__/table.test.ts

@@ -368,9 +368,9 @@ describe("merge insert", () => {
       { a: 4, b: "z" },
     ];
 
-    expect(
-      JSON.parse(JSON.stringify((await table.toArrow()).toArray())),
-    ).toEqual(expected);
+    const result = (await table.toArrow()).toArray().sort((a, b) => a.a - b.a);
+
+    expect(result.map((row) => ({ ...row }))).toEqual(expected);
   });
   test("conditional update", async () => {
     const newData = [
@@ -1650,13 +1650,25 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
       expect(resultSet.has("fob")).toBe(true);
       expect(resultSet.has("fo")).toBe(true);
       expect(resultSet.has("food")).toBe(true);
+
+      const prefixResults = await table
+        .search(
+          new MatchQuery("foo", "text", { fuzziness: 3, prefixLength: 3 }),
+        )
+        .toArray();
+      expect(prefixResults.length).toBe(2);
+      const resultSet2 = new Set(prefixResults.map((r) => r.text));
+      expect(resultSet2.has("foo")).toBe(true);
+      expect(resultSet2.has("food")).toBe(true);
     });
 
     test("full text search boolean query", async () => {
       const db = await connect(tmpDir.name);
       const data = [
-        { text: "hello world", vector: [0.1, 0.2, 0.3] },
-        { text: "goodbye world", vector: [0.4, 0.5, 0.6] },
+        { text: "The cat and dog are playing" },
+        { text: "The cat is sleeping" },
+        { text: "The dog is barking" },
+        { text: "The dog chases the cat" },
       ];
       const table = await db.createTable("test", data);
       await table.createIndex("text", {
@@ -1666,22 +1678,86 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
       const shouldResults = await table
         .search(
           new BooleanQuery([
-            [Occur.Should, new MatchQuery("hello", "text")],
-            [Occur.Should, new MatchQuery("goodbye", "text")],
+            [Occur.Should, new MatchQuery("cat", "text")],
+            [Occur.Should, new MatchQuery("dog", "text")],
           ]),
         )
         .toArray();
-      expect(shouldResults.length).toBe(2);
+      expect(shouldResults.length).toBe(4);
 
       const mustResults = await table
         .search(
           new BooleanQuery([
-            [Occur.Must, new MatchQuery("hello", "text")],
-            [Occur.Must, new MatchQuery("world", "text")],
+            [Occur.Must, new MatchQuery("cat", "text")],
+            [Occur.Must, new MatchQuery("dog", "text")],
           ]),
         )
         .toArray();
-      expect(mustResults.length).toBe(1);
+      expect(mustResults.length).toBe(2);
+
+      const mustNotResults = await table
+        .search(
+          new BooleanQuery([
+            [Occur.Must, new MatchQuery("cat", "text")],
+            [Occur.MustNot, new MatchQuery("dog", "text")],
+          ]),
+        )
+        .toArray();
+      expect(mustNotResults.length).toBe(1);
+    });
+
+    test("full text search ngram", async () => {
+      const db = await connect(tmpDir.name);
+      const data = [
+        { text: "hello world", vector: [0.1, 0.2, 0.3] },
+        { text: "lance database", vector: [0.4, 0.5, 0.6] },
+        { text: "lance is cool", vector: [0.7, 0.8, 0.9] },
+      ];
+      const table = await db.createTable("test", data);
+      await table.createIndex("text", {
+        config: Index.fts({ baseTokenizer: "ngram" }),
+      });
+
+      const results = await table.search("lan").toArray();
+      expect(results.length).toBe(2);
+      const resultSet = new Set(results.map((r) => r.text));
+      expect(resultSet.has("lance database")).toBe(true);
+      expect(resultSet.has("lance is cool")).toBe(true);
+
+      const results2 = await table.search("nce").toArray(); // spellchecker:disable-line
+      expect(results2.length).toBe(2);
+      const resultSet2 = new Set(results2.map((r) => r.text));
+      expect(resultSet2.has("lance database")).toBe(true);
+      expect(resultSet2.has("lance is cool")).toBe(true);
+
+      // the default min_ngram_length is 3, so "la" should not match
+      const results3 = await table.search("la").toArray();
+      expect(results3.length).toBe(0);
+
+      // test setting min_ngram_length and prefix_only
+      await table.createIndex("text", {
+        config: Index.fts({
+          baseTokenizer: "ngram",
+          ngramMinLength: 2,
+          prefixOnly: true,
+        }),
+        replace: true,
+      });
+
+      const results4 = await table.search("lan").toArray();
+      expect(results4.length).toBe(2);
+      const resultSet4 = new Set(results4.map((r) => r.text));
+      expect(resultSet4.has("lance database")).toBe(true);
+      expect(resultSet4.has("lance is cool")).toBe(true);
+
+      const results5 = await table.search("nce").toArray(); // spellchecker:disable-line
+      expect(results5.length).toBe(0);
+
+      const results6 = await table.search("la").toArray();
+      expect(results6.length).toBe(2);
+      const resultSet6 = new Set(results6.map((r) => r.text));
+      expect(resultSet6.has("lance database")).toBe(true);
+      expect(resultSet6.has("lance is cool")).toBe(true);
     });
 
     test.each([
@@ -1787,4 +1863,43 @@ describe("column name options", () => {
     expect(results[0].query_index).toBe(0);
     expect(results[1].query_index).toBe(1);
   });
+
+  test("index and search multivectors", async () => {
+    const db = await connect(tmpDir.name);
+    const data = [];
+    // generate 512 random multivectors
+    for (let i = 0; i < 256; i++) {
+      data.push({
+        multivector: Array.from({ length: 10 }, () =>
+          Array(2).fill(Math.random()),
+        ),
+      });
+    }
+    const table = await db.createTable("multivectors", data, {
+      schema: new Schema([
+        new Field(
+          "multivector",
+          new List(
+            new Field(
+              "item",
+              new FixedSizeList(2, new Field("item", new Float32())),
+            ),
+          ),
+        ),
+      ]),
+    });
+
+    const results = await table.search(data[0].multivector).limit(10).toArray();
+    expect(results.length).toBe(10);
+
+    await table.createIndex("multivector", {
+      config: Index.ivfPq({ numPartitions: 2, distanceType: "cosine" }),
+    });
+
+    const results2 = await table
+      .search(data[0].multivector)
+      .limit(10)
+      .toArray();
+    expect(results2.length).toBe(10);
+  });
 });

nodejs/lancedb/arrow.ts

@@ -107,6 +107,20 @@ export type IntoVector =
   | number[]
   | Promise<Float32Array | Float64Array | number[]>;
 
+export type MultiVector = IntoVector[];
+
+export function isMultiVector(value: unknown): value is MultiVector {
+  return Array.isArray(value) && isIntoVector(value[0]);
+}
+
+export function isIntoVector(value: unknown): value is IntoVector {
+  return (
+    value instanceof Float32Array ||
+    value instanceof Float64Array ||
+    (Array.isArray(value) && !Array.isArray(value[0]))
+  );
+}
+
 export function isArrowTable(value: object): value is TableLike {
   if (value instanceof ArrowTable) return true;
   return "schema" in value && "batches" in value;
@@ -839,6 +853,15 @@ async function applyEmbeddingsFromMetadata(
     const vector = makeVector(vectors, destType);
     columns[destColumn] = vector;
   }
+
+  // Add any missing columns from the schema as null vectors
+  for (const field of schema.fields) {
+    if (!(field.name in columns)) {
+      const nullValues = new Array(table.numRows).fill(null);
+      columns[field.name] = makeVector(nullValues, field.type);
+    }
+  }
+
   const newTable = new ArrowTable(columns);
   return alignTable(newTable, schema);
 }
@@ -987,7 +1010,21 @@ export async function convertToTable(
   embeddings?: EmbeddingFunctionConfig,
   makeTableOptions?: Partial<MakeArrowTableOptions>,
 ): Promise<ArrowTable> {
-  const table = makeArrowTable(data, makeTableOptions);
+  let processedData = data;
+
+  // If we have a schema with embedding metadata, we need to preprocess the data
+  // to ensure all nested fields are present
+  if (
+    makeTableOptions?.schema &&
+    makeTableOptions.schema.metadata?.has("embedding_functions")
+  ) {
+    processedData = ensureNestedFieldsExist(
+      data,
+      makeTableOptions.schema as Schema,
+    );
+  }
+
+  const table = makeArrowTable(processedData, makeTableOptions);
   return await applyEmbeddings(table, embeddings, makeTableOptions?.schema);
 }
 
@@ -1080,7 +1117,16 @@ export async function fromDataToBuffer(
     schema = sanitizeSchema(schema);
   }
   if (isArrowTable(data)) {
-    return fromTableToBuffer(sanitizeTable(data), embeddings, schema);
+    const table = sanitizeTable(data);
+    // If we have a schema with embedding functions, we need to ensure all columns exist
+    // before applying embeddings, since applyEmbeddingsFromMetadata expects all columns
+    // to be present in the table
+    if (schema && schema.metadata?.has("embedding_functions")) {
+      const alignedTable = alignTableToSchema(table, schema);
+      return fromTableToBuffer(alignedTable, embeddings, schema);
+    } else {
+      return fromTableToBuffer(table, embeddings, schema);
+    }
   } else {
     const table = await convertToTable(data, embeddings, { schema });
     return fromTableToBuffer(table);
@@ -1149,7 +1195,7 @@ function alignBatch(batch: RecordBatch, schema: Schema): RecordBatch {
     type: new Struct(schema.fields),
     length: batch.numRows,
     nullCount: batch.nullCount,
-    children: alignedChildren,
+    children: alignedChildren as unknown as ArrowData<DataType>[],
   });
   return new RecordBatch(schema, newData);
 }
@@ -1221,6 +1267,79 @@ function validateSchemaEmbeddings(
   return new Schema(fields, schema.metadata);
 }
 
+/**
+ * Ensures that all nested fields defined in the schema exist in the data,
+ * filling missing fields with null values.
+ */
+export function ensureNestedFieldsExist(
+  data: Array<Record<string, unknown>>,
+  schema: Schema,
+): Array<Record<string, unknown>> {
+  return data.map((row) => {
+    const completeRow: Record<string, unknown> = {};
+
+    for (const field of schema.fields) {
+      if (field.name in row) {
+        if (
+          field.type.constructor.name === "Struct" &&
+          row[field.name] !== null &&
+          row[field.name] !== undefined
+        ) {
+          // Handle nested struct
+          const nestedValue = row[field.name] as Record<string, unknown>;
+          completeRow[field.name] = ensureStructFieldsExist(
+            nestedValue,
+            field.type,
+          );
+        } else {
+          // Non-struct field or null struct value
+          completeRow[field.name] = row[field.name];
+        }
+      } else {
+        // Field is missing from the data - set to null
+        completeRow[field.name] = null;
+      }
+    }
+
+    return completeRow;
+  });
+}
+
+/**
+ * Recursively ensures that all fields in a struct type exist in the data,
+ * filling missing fields with null values.
+ */
+function ensureStructFieldsExist(
+  data: Record<string, unknown>,
+  structType: Struct,
+): Record<string, unknown> {
+  const completeStruct: Record<string, unknown> = {};
+
+  for (const childField of structType.children) {
+    if (childField.name in data) {
+      if (
+        childField.type.constructor.name === "Struct" &&
+        data[childField.name] !== null &&
+        data[childField.name] !== undefined
+      ) {
+        // Recursively handle nested struct
+        completeStruct[childField.name] = ensureStructFieldsExist(
+          data[childField.name] as Record<string, unknown>,
+          childField.type,
+        );
+      } else {
+        // Non-struct field or null struct value
+        completeStruct[childField.name] = data[childField.name];
+      }
+    } else {
+      // Field is missing - set to null
+      completeStruct[childField.name] = null;
+    }
+  }
+
+  return completeStruct;
+}
+
 interface JsonDataType {
   type: string;
   fields?: JsonField[];
@@ -1354,3 +1473,64 @@ function fieldToJson(field: Field): JsonField {
     metadata: field.metadata,
   };
 }
+
+function alignTableToSchema(
+  table: ArrowTable,
+  targetSchema: Schema,
+): ArrowTable {
+  const existingColumns = new Map<string, Vector>();
+
+  // Map existing columns
+  for (const field of table.schema.fields) {
+    existingColumns.set(field.name, table.getChild(field.name)!);
+  }
+
+  // Create vectors for all fields in target schema
+  const alignedColumns: Record<string, Vector> = {};
+
+  for (const field of targetSchema.fields) {
+    if (existingColumns.has(field.name)) {
+      // Column exists, use it
+      alignedColumns[field.name] = existingColumns.get(field.name)!;
+    } else {
+      // Column missing, create null vector
+      alignedColumns[field.name] = createNullVector(field, table.numRows);
+    }
+  }
+
+  // Create new table with aligned schema and columns
+  return new ArrowTable(targetSchema, alignedColumns);
+}
+
+function createNullVector(field: Field, numRows: number): Vector {
+  if (field.type.constructor.name === "Struct") {
+    // For struct types, create a struct with null fields
+    const structType = field.type as Struct;
+    const childVectors = structType.children.map((childField) =>
+      createNullVector(childField, numRows),
+    );
+
+    // Create struct data
+    const structData = makeData({
+      type: structType,
+      length: numRows,
+      nullCount: 0,
+      children: childVectors.map((v) => v.data[0]),
+    });
+
+    return arrowMakeVector(structData);
+  } else {
+    // For other types, create a vector of nulls
+    const nullBitmap = new Uint8Array(Math.ceil(numRows / 8));
+    // All bits are 0, meaning all values are null
+
+    const data = makeData({
+      type: field.type,
+      length: numRows,
+      nullCount: numRows,
+      nullBitmap,
+    });
+
+    return arrowMakeVector(data);
+  }
+}

nodejs/lancedb/index.ts

@@ -100,6 +100,7 @@ export {
   RecordBatchLike,
   DataLike,
   IntoVector,
+  MultiVector,
 } from "./arrow";
 export { IntoSql, packBits } from "./util";
 

nodejs/lancedb/indices.ts

@@ -439,7 +439,7 @@ export interface FtsOptions {
    *
    * "raw" - Raw tokenizer. This tokenizer does not split the text into tokens and indexes the entire text as a single token.
    */
-  baseTokenizer?: "simple" | "whitespace" | "raw";
+  baseTokenizer?: "simple" | "whitespace" | "raw" | "ngram";
 
   /**
    * language for stemming and stop words
@@ -472,6 +472,21 @@ export interface FtsOptions {
    * whether to remove punctuation
    */
   asciiFolding?: boolean;
+
+  /**
+   * ngram min length
+   */
+  ngramMinLength?: number;
+
+  /**
+   * ngram max length
+   */
+  ngramMaxLength?: number;
+
+  /**
+   * whether to only index the prefix of the token for ngram tokenizer
+   */
+  prefixOnly?: boolean;
 }
 
 export class Index {
@@ -608,6 +623,9 @@ export class Index {
         options?.stem,
         options?.removeStopWords,
         options?.asciiFolding,
+        options?.ngramMinLength,
+        options?.ngramMaxLength,
+        options?.prefixOnly,
       ),
     );
   }

nodejs/lancedb/query.ts

@@ -812,10 +812,12 @@ export enum Operator {
  *
  * - `Must`: The term must be present in the document.
  * - `Should`: The term should contribute to the document score, but is not required.
+ * - `MustNot`: The term must not be present in the document.
  */
 export enum Occur {
-  Must = "MUST",
   Should = "SHOULD",
+  Must = "MUST",
+  MustNot = "MUST_NOT",
 }
 
 /**
@@ -856,6 +858,7 @@ export class MatchQuery implements FullTextQuery {
    *   - `fuzziness`: The fuzziness level for the query (default is 0).
    *   - `maxExpansions`: The maximum number of terms to consider for fuzzy matching (default is 50).
    *   - `operator`: The logical operator to use for combining terms in the query (default is "OR").
+   *   - `prefixLength`: The number of beginning characters being unchanged for fuzzy matching.
    */
   constructor(
     query: string,
@@ -865,6 +868,7 @@ export class MatchQuery implements FullTextQuery {
       fuzziness?: number;
       maxExpansions?: number;
       operator?: Operator;
+      prefixLength?: number;
     },
   ) {
     let fuzziness = options?.fuzziness;
@@ -878,6 +882,7 @@ export class MatchQuery implements FullTextQuery {
       fuzziness,
       options?.maxExpansions ?? 50,
       options?.operator ?? Operator.Or,
+      options?.prefixLength ?? 0,
     );
   }
 

nodejs/lancedb/table.ts

@@ -6,9 +6,11 @@ import {
   Data,
   DataType,
   IntoVector,
+  MultiVector,
   Schema,
   dataTypeToJson,
   fromDataToBuffer,
+  isMultiVector,
   tableFromIPC,
 } from "./arrow";
 
@@ -75,10 +77,10 @@ export interface OptimizeOptions {
    * // Delete all versions older than 1 day
    * const olderThan = new Date();
    * olderThan.setDate(olderThan.getDate() - 1));
-   * tbl.cleanupOlderVersions(olderThan);
+   * tbl.optimize({cleanupOlderThan: olderThan});
    *
    * // Delete all versions except the current version
-   * tbl.cleanupOlderVersions(new Date());
+   * tbl.optimize({cleanupOlderThan: new Date()});
    */
   cleanupOlderThan: Date;
   deleteUnverified: boolean;
@@ -346,7 +348,7 @@ export abstract class Table {
    * if the query is a string and no embedding function is defined, it will be treated as a full text search query
    */
   abstract search(
-    query: string | IntoVector | FullTextQuery,
+    query: string | IntoVector | MultiVector | FullTextQuery,
     queryType?: string,
     ftsColumns?: string | string[],
   ): VectorQuery | Query;
@@ -357,7 +359,7 @@ export abstract class Table {
    * is the same thing as calling `nearestTo` on the builder returned
    * by `query`.  @see {@link Query#nearestTo} for more details.
    */
-  abstract vectorSearch(vector: IntoVector): VectorQuery;
+  abstract vectorSearch(vector: IntoVector | MultiVector): VectorQuery;
   /**
    * Add new columns with defined values.
    * @param {AddColumnsSql[]} newColumnTransforms pairs of column names and
@@ -668,7 +670,7 @@ export class LocalTable extends Table {
   }
 
   search(
-    query: string | IntoVector | FullTextQuery,
+    query: string | IntoVector | MultiVector | FullTextQuery,
     queryType: string = "auto",
     ftsColumns?: string | string[],
   ): VectorQuery | Query {
@@ -715,7 +717,15 @@ export class LocalTable extends Table {
     return this.query().nearestTo(queryPromise);
   }
 
-  vectorSearch(vector: IntoVector): VectorQuery {
+  vectorSearch(vector: IntoVector | MultiVector): VectorQuery {
+    if (isMultiVector(vector)) {
+      const query = this.query().nearestTo(vector[0]);
+      for (const v of vector.slice(1)) {
+        query.addQueryVector(v);
+      }
+      return query;
+    }
+
     return this.query().nearestTo(vector);
   }
 

nodejs/npm/darwin-arm64/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-darwin-arm64",
-	"version": "0.20.1-beta.1",
+	"version": "0.21.2-beta.0",
 	"os": ["darwin"],
 	"cpu": ["arm64"],
 	"main": "lancedb.darwin-arm64.node",

nodejs/npm/darwin-x64/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-darwin-x64",
-	"version": "0.20.1-beta.1",
+	"version": "0.21.2-beta.0",
 	"os": ["darwin"],
 	"cpu": ["x64"],
 	"main": "lancedb.darwin-x64.node",

nodejs/npm/linux-arm64-gnu/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-gnu",
-	"version": "0.20.1-beta.1",
+	"version": "0.21.2-beta.0",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-gnu.node",

nodejs/npm/linux-arm64-musl/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-musl",
-	"version": "0.20.1-beta.1",
+	"version": "0.21.2-beta.0",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-musl.node",

nodejs/npm/linux-x64-gnu/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-gnu",
-	"version": "0.20.1-beta.1",
+	"version": "0.21.2-beta.0",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-gnu.node",

nodejs/npm/linux-x64-musl/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-musl",
-	"version": "0.20.1-beta.1",
+	"version": "0.21.2-beta.0",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-musl.node",

nodejs/npm/win32-arm64-msvc/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lancedb/lancedb-win32-arm64-msvc",
-  "version": "0.20.1-beta.1",
+  "version": "0.21.2-beta.0",
   "os": [
     "win32"
   ],

nodejs/npm/win32-x64-msvc/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-win32-x64-msvc",
-	"version": "0.20.1-beta.1",
+	"version": "0.21.2-beta.0",
 	"os": ["win32"],
 	"cpu": ["x64"],
 	"main": "lancedb.win32-x64-msvc.node",

nodejs/package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "@lancedb/lancedb",
-  "version": "0.20.1-beta.1",
+  "version": "0.21.2-beta.0",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "@lancedb/lancedb",
-      "version": "0.20.1-beta.1",
+      "version": "0.21.2-beta.0",
       "cpu": [
         "x64",
         "arm64"

nodejs/package.json

@@ -11,7 +11,7 @@
     "ann"
   ],
   "private": false,
-  "version": "0.20.1-beta.1",
+  "version": "0.21.2-beta.0",
   "main": "dist/index.js",
   "exports": {
     ".": "./dist/index.js",

nodejs/src/index.rs

@@ -123,6 +123,9 @@ impl Index {
         stem: Option<bool>,
         remove_stop_words: Option<bool>,
         ascii_folding: Option<bool>,
+        ngram_min_length: Option<u32>,
+        ngram_max_length: Option<u32>,
+        prefix_only: Option<bool>,
     ) -> Self {
         let mut opts = FtsIndexBuilder::default();
         if let Some(with_position) = with_position {
@@ -149,6 +152,15 @@ impl Index {
         if let Some(ascii_folding) = ascii_folding {
             opts = opts.ascii_folding(ascii_folding);
         }
+        if let Some(ngram_min_length) = ngram_min_length {
+            opts = opts.ngram_min_length(ngram_min_length);
+        }
+        if let Some(ngram_max_length) = ngram_max_length {
+            opts = opts.ngram_max_length(ngram_max_length);
+        }
+        if let Some(prefix_only) = prefix_only {
+            opts = opts.ngram_prefix_only(prefix_only);
+        }
 
         Self {
             inner: Mutex::new(Some(LanceDbIndex::FTS(opts))),

nodejs/src/query.rs

@@ -335,6 +335,7 @@ impl JsFullTextQuery {
         fuzziness: Option<u32>,
         max_expansions: u32,
         operator: String,
+        prefix_length: u32,
     ) -> napi::Result<Self> {
         Ok(Self {
             inner: MatchQuery::new(query)
@@ -347,6 +348,7 @@ impl JsFullTextQuery {
                         napi::Error::from_reason(format!("Invalid operator: {}", e))
                     })?,
                 )
+                .with_prefix_length(prefix_length)
                 .into(),
         })
     }

python/.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.23.1-beta.2"
+current_version = "0.24.2-beta.1"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

python/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb-python"
-version = "0.23.1-beta.2"
+version = "0.24.2-beta.1"
 edition.workspace = true
 description = "Python bindings for LanceDB"
 license.workspace = true

python/pyproject.toml

@@ -85,7 +85,7 @@ embeddings = [
     "boto3>=1.28.57",
     "awscli>=1.29.57",
     "botocore>=1.31.57",
-    "ollama",
+    "ollama>=0.3.0",
     "ibm-watsonx-ai>=1.1.2",
 ]
 azure = ["adlfs>=2024.2.0"]

python/python/lancedb/embeddings/ollama.py

@@ -2,14 +2,15 @@
 # SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
 from functools import cached_property
-from typing import TYPE_CHECKING, List, Optional, Union
+from typing import TYPE_CHECKING, List, Optional, Sequence, Union
+
+import numpy as np
 
 from ..util import attempt_import_or_raise
 from .base import TextEmbeddingFunction
 from .registry import register
 
 if TYPE_CHECKING:
-    import numpy as np
     import ollama
 
 
@@ -28,23 +29,21 @@ class OllamaEmbeddings(TextEmbeddingFunction):
     keep_alive: Optional[Union[float, str]] = None
     ollama_client_kwargs: Optional[dict] = {}
 
-    def ndims(self):
+    def ndims(self) -> int:
         return len(self.generate_embeddings(["foo"])[0])
 
-    def _compute_embedding(self, text) -> Union["np.array", None]:
-        return (
-            self._ollama_client.embeddings(
-                model=self.name,
-                prompt=text,
-                options=self.options,
-                keep_alive=self.keep_alive,
-            )["embedding"]
-            or None
+    def _compute_embedding(self, text: Sequence[str]) -> Sequence[Sequence[float]]:
+        response = self._ollama_client.embed(
+            model=self.name,
+            input=text,
+            options=self.options,
+            keep_alive=self.keep_alive,
         )
+        return response.embeddings
 
     def generate_embeddings(
-        self, texts: Union[List[str], "np.ndarray"]
-    ) -> list[Union["np.array", None]]:
+        self, texts: Union[List[str], np.ndarray]
+    ) -> list[Union[np.array, None]]:
         """
         Get the embeddings for the given texts
 
@@ -54,8 +53,8 @@ class OllamaEmbeddings(TextEmbeddingFunction):
             The texts to embed
         """
         # TODO retry, rate limit, token limit
-        embeddings = [self._compute_embedding(text) for text in texts]
-        return embeddings
+        embeddings = self._compute_embedding(texts)
+        return list(embeddings)
 
     @cached_property
     def _ollama_client(self) -> "ollama.Client":

python/python/lancedb/index.py

@@ -137,6 +137,9 @@ class FTS:
     stem: bool = True
     remove_stop_words: bool = True
     ascii_folding: bool = True
+    ngram_min_length: int = 3
+    ngram_max_length: int = 3
+    prefix_only: bool = False
 
 
 @dataclass

python/python/lancedb/query.py

@@ -101,8 +101,9 @@ class FullTextOperator(str, Enum):
 
 
 class Occur(str, Enum):
-    MUST = "MUST"
     SHOULD = "SHOULD"
+    MUST = "MUST"
+    MUST_NOT = "MUST_NOT"
 
 
 @pydantic.dataclasses.dataclass
@@ -181,6 +182,9 @@ class MatchQuery(FullTextQuery):
         Can be either `AND` or `OR`.
         If `AND`, all terms in the query must match.
         If `OR`, at least one term in the query must match.
+    prefix_length : int, optional
+        The number of beginning characters being unchanged for fuzzy matching.
+        This is useful to achieve prefix matching.
     """
 
     query: str
@@ -189,6 +193,7 @@ class MatchQuery(FullTextQuery):
     fuzziness: int = pydantic.Field(0, kw_only=True)
     max_expansions: int = pydantic.Field(50, kw_only=True)
     operator: FullTextOperator = pydantic.Field(FullTextOperator.OR, kw_only=True)
+    prefix_length: int = pydantic.Field(0, kw_only=True)
 
     def query_type(self) -> FullTextQueryType:
         return FullTextQueryType.MATCH
@@ -1369,6 +1374,8 @@ class LanceVectorQueryBuilder(LanceQueryBuilder):
         if query_string is not None and not isinstance(query_string, str):
             raise ValueError("Reranking currently only supports string queries")
         self._str_query = query_string if query_string is not None else self._str_query
+        if reranker.score == "all":
+            self.with_row_id(True)
         return self
 
     def bypass_vector_index(self) -> LanceVectorQueryBuilder:
@@ -1446,10 +1453,13 @@ class LanceFtsQueryBuilder(LanceQueryBuilder):
 
         query = self._query
         if self._phrase_query:
-            raise NotImplementedError(
-                "Phrase query is not yet supported in Lance FTS. "
-                "Use tantivy-based index instead for now."
-            )
+            if isinstance(query, str):
+                if not query.startswith('"') or not query.endswith('"'):
+                    query = f'"{query}"'
+            elif isinstance(query, FullTextQuery) and not isinstance(
+                query, PhraseQuery
+            ):
+                raise TypeError("Please use PhraseQuery for phrase queries.")
         query = self.to_query_object()
         results = self._table._execute_query(query, timeout=timeout)
         results = results.read_all()
@@ -1561,6 +1571,8 @@ class LanceFtsQueryBuilder(LanceQueryBuilder):
             The LanceQueryBuilder object.
         """
         self._reranker = reranker
+        if reranker.score == "all":
+            self.with_row_id(True)
         return self
 
 
@@ -1837,6 +1849,8 @@ class LanceHybridQueryBuilder(LanceQueryBuilder):
 
         self._norm = normalize
         self._reranker = reranker
+        if reranker.score == "all":
+            self.with_row_id(True)
 
         return self
 
@@ -3034,15 +3048,21 @@ class AsyncHybridQuery(AsyncQueryBase, AsyncVectorQueryBase):
         >>> asyncio.run(doctest_example()) # doctest: +ELLIPSIS, +NORMALIZE_WHITESPACE
         Vector Search Plan:
         ProjectionExec: expr=[vector@0 as vector, text@3 as text, _distance@2 as _distance]
-            Take: columns="vector, _rowid, _distance, (text)"
-                CoalesceBatchesExec: target_batch_size=1024
-                GlobalLimitExec: skip=0, fetch=10
-                    FilterExec: _distance@2 IS NOT NULL
-                    SortExec: TopK(fetch=10), expr=[_distance@2 ASC NULLS LAST], preserve_partitioning=[false]
-                        KNNVectorDistance: metric=l2
-                        LanceScan: uri=..., projection=[vector], row_id=true, row_addr=false, ordered=false
+          Take: columns="vector, _rowid, _distance, (text)"
+            CoalesceBatchesExec: target_batch_size=1024
+              GlobalLimitExec: skip=0, fetch=10
+                FilterExec: _distance@2 IS NOT NULL
+                  SortExec: TopK(fetch=10), expr=[_distance@2 ASC NULLS LAST], preserve_partitioning=[false]
+                    KNNVectorDistance: metric=l2
+                      LanceScan: uri=..., projection=[vector], row_id=true, row_addr=false, ordered=false
+        <BLANKLINE>
         FTS Search Plan:
-        LanceScan: uri=..., projection=[vector, text], row_id=false, row_addr=false, ordered=true
+        ProjectionExec: expr=[vector@2 as vector, text@3 as text, _score@1 as _score]
+          Take: columns="_rowid, _score, (vector), (text)"
+            CoalesceBatchesExec: target_batch_size=1024
+              GlobalLimitExec: skip=0, fetch=10
+                MatchQuery: query=hello
+        <BLANKLINE>
 
         Parameters
         ----------

python/python/lancedb/remote/table.py

@@ -18,7 +18,7 @@ from lancedb._lancedb import (
     UpdateResult,
 )
 from lancedb.embeddings.base import EmbeddingFunctionConfig
-from lancedb.index import FTS, BTree, Bitmap, HnswPq, HnswSq, IvfFlat, IvfPq, LabelList
+from lancedb.index import FTS, BTree, Bitmap, HnswSq, IvfFlat, IvfPq, LabelList
 from lancedb.remote.db import LOOP
 import pyarrow as pa
 
@@ -89,7 +89,7 @@ class RemoteTable(Table):
 
     def to_pandas(self):
         """to_pandas() is not yet supported on LanceDB cloud."""
-        return NotImplementedError("to_pandas() is not yet supported on LanceDB cloud.")
+        raise NotImplementedError("to_pandas() is not yet supported on LanceDB cloud.")
 
     def checkout(self, version: Union[int, str]):
         return LOOP.run(self._table.checkout(version))
@@ -158,6 +158,9 @@ class RemoteTable(Table):
         stem: bool = True,
         remove_stop_words: bool = True,
         ascii_folding: bool = True,
+        ngram_min_length: int = 3,
+        ngram_max_length: int = 3,
+        prefix_only: bool = False,
     ):
         config = FTS(
             with_position=with_position,
@@ -168,6 +171,9 @@ class RemoteTable(Table):
             stem=stem,
             remove_stop_words=remove_stop_words,
             ascii_folding=ascii_folding,
+            ngram_min_length=ngram_min_length,
+            ngram_max_length=ngram_max_length,
+            prefix_only=prefix_only,
         )
         LOOP.run(
             self._table.create_index(
@@ -186,6 +192,8 @@ class RemoteTable(Table):
         accelerator: Optional[str] = None,
         index_type="vector",
         wait_timeout: Optional[timedelta] = None,
+        *,
+        num_bits: int = 8,
     ):
         """Create an index on the table.
         Currently, the only parameters that matter are
@@ -220,11 +228,6 @@ class RemoteTable(Table):
         >>> table.create_index("l2", "vector") # doctest: +SKIP
         """
 
-        if num_partitions is not None:
-            logging.warning(
-                "num_partitions is not supported on LanceDB cloud."
-                "This parameter will be tuned automatically."
-            )
         if num_sub_vectors is not None:
             logging.warning(
                 "num_sub_vectors is not supported on LanceDB cloud."
@@ -244,13 +247,21 @@ class RemoteTable(Table):
 
         index_type = index_type.upper()
         if index_type == "VECTOR" or index_type == "IVF_PQ":
-            config = IvfPq(distance_type=metric)
+            config = IvfPq(
+                distance_type=metric,
+                num_partitions=num_partitions,
+                num_sub_vectors=num_sub_vectors,
+                num_bits=num_bits,
+            )
         elif index_type == "IVF_HNSW_PQ":
-            config = HnswPq(distance_type=metric)
+            raise ValueError(
+                "IVF_HNSW_PQ is not supported on LanceDB cloud."
+                "Please use IVF_HNSW_SQ instead."
+            )
         elif index_type == "IVF_HNSW_SQ":
-            config = HnswSq(distance_type=metric)
+            config = HnswSq(distance_type=metric, num_partitions=num_partitions)
         elif index_type == "IVF_FLAT":
-            config = IvfFlat(distance_type=metric)
+            config = IvfFlat(distance_type=metric, num_partitions=num_partitions)
         else:
             raise ValueError(
                 f"Unknown vector index type: {index_type}. Valid options are"

python/python/lancedb/rerankers/answerdotai.py

@@ -74,9 +74,7 @@ class AnswerdotaiRerankers(Reranker):
         if self.score == "relevance":
             combined_results = self._keep_relevance_score(combined_results)
         elif self.score == "all":
-            raise NotImplementedError(
-                "Answerdotai Reranker does not support score='all' yet"
-            )
+            combined_results = self._merge_and_keep_scores(vector_results, fts_results)
         combined_results = combined_results.sort_by(
             [("_relevance_score", "descending")]
         )

python/python/lancedb/rerankers/base.py

@@ -232,6 +232,39 @@ class Reranker(ABC):
 
         return deduped_table
 
+    def _merge_and_keep_scores(self, vector_results: pa.Table, fts_results: pa.Table):
+        """
+        Merge the results from the vector and FTS search and keep the scores.
+        This op is slower than just keeping relevance score but can be useful
+        for debugging.
+        """
+        # add nulls to fts results for _distance
+        if "_distance" not in fts_results.column_names:
+            fts_results = fts_results.append_column(
+                "_distance",
+                pa.array([None] * len(fts_results), type=pa.float32()),
+            )
+        # add nulls to vector results for _score
+        if "_score" not in vector_results.column_names:
+            vector_results = vector_results.append_column(
+                "_score",
+                pa.array([None] * len(vector_results), type=pa.float32()),
+            )
+
+        # combine them and fill the scores
+        vector_results_dict = {row["_rowid"]: row for row in vector_results.to_pylist()}
+        fts_results_dict = {row["_rowid"]: row for row in fts_results.to_pylist()}
+
+        # merge them into vector_results
+        for key, value in fts_results_dict.items():
+            if key in vector_results_dict:
+                vector_results_dict[key]["_score"] = value["_score"]
+            else:
+                vector_results_dict[key] = value
+
+        combined = pa.Table.from_pylist(list(vector_results_dict.values()))
+        return combined
+
     def _keep_relevance_score(self, combined_results: pa.Table):
         if self.score == "relevance":
             if "_score" in combined_results.column_names:

python/python/lancedb/rerankers/cohere.py

@@ -92,14 +92,14 @@ class CohereReranker(Reranker):
         vector_results: pa.Table,
         fts_results: pa.Table,
     ):
-        combined_results = self.merge_results(vector_results, fts_results)
+        if self.score == "all":
+            combined_results = self._merge_and_keep_scores(vector_results, fts_results)
+        else:
+            combined_results = self.merge_results(vector_results, fts_results)
         combined_results = self._rerank(combined_results, query)
         if self.score == "relevance":
             combined_results = self._keep_relevance_score(combined_results)
-        elif self.score == "all":
-            raise NotImplementedError(
-                "return_score='all' not implemented for cohere reranker"
-            )
+
         return combined_results
 
     def rerank_vector(self, query: str, vector_results: pa.Table):

python/python/lancedb/rerankers/cross_encoder.py

@@ -81,15 +81,15 @@ class CrossEncoderReranker(Reranker):
         vector_results: pa.Table,
         fts_results: pa.Table,
     ):
-        combined_results = self.merge_results(vector_results, fts_results)
+        if self.score == "all":
+            combined_results = self._merge_and_keep_scores(vector_results, fts_results)
+        else:
+            combined_results = self.merge_results(vector_results, fts_results)
         combined_results = self._rerank(combined_results, query)
         # sort the results by _score
         if self.score == "relevance":
             combined_results = self._keep_relevance_score(combined_results)
-        elif self.score == "all":
-            raise NotImplementedError(
-                "return_score='all' not implemented for CrossEncoderReranker"
-            )
+
         combined_results = combined_results.sort_by(
             [("_relevance_score", "descending")]
         )

python/python/lancedb/rerankers/jinaai.py

@@ -97,14 +97,14 @@ class JinaReranker(Reranker):
         vector_results: pa.Table,
         fts_results: pa.Table,
     ):
-        combined_results = self.merge_results(vector_results, fts_results)
+        if self.score == "all":
+            combined_results = self._merge_and_keep_scores(vector_results, fts_results)
+        else:
+            combined_results = self.merge_results(vector_results, fts_results)
         combined_results = self._rerank(combined_results, query)
         if self.score == "relevance":
             combined_results = self._keep_relevance_score(combined_results)
-        elif self.score == "all":
-            raise NotImplementedError(
-                "return_score='all' not implemented for JinaReranker"
-            )
+
         return combined_results
 
     def rerank_vector(self, query: str, vector_results: pa.Table):

python/python/lancedb/rerankers/openai.py

@@ -88,14 +88,13 @@ class OpenaiReranker(Reranker):
         vector_results: pa.Table,
         fts_results: pa.Table,
     ):
-        combined_results = self.merge_results(vector_results, fts_results)
+        if self.score == "all":
+            combined_results = self._merge_and_keep_scores(vector_results, fts_results)
+        else:
+            combined_results = self.merge_results(vector_results, fts_results)
         combined_results = self._rerank(combined_results, query)
         if self.score == "relevance":
             combined_results = self._keep_relevance_score(combined_results)
-        elif self.score == "all":
-            raise NotImplementedError(
-                "OpenAI Reranker does not support score='all' yet"
-            )
 
         combined_results = combined_results.sort_by(
             [("_relevance_score", "descending")]

python/python/lancedb/rerankers/voyageai.py

@@ -94,14 +94,14 @@ class VoyageAIReranker(Reranker):
         vector_results: pa.Table,
         fts_results: pa.Table,
     ):
-        combined_results = self.merge_results(vector_results, fts_results)
+        if self.score == "all":
+            combined_results = self._merge_and_keep_scores(vector_results, fts_results)
+        else:
+            combined_results = self.merge_results(vector_results, fts_results)
         combined_results = self._rerank(combined_results, query)
         if self.score == "relevance":
             combined_results = self._keep_relevance_score(combined_results)
-        elif self.score == "all":
-            raise NotImplementedError(
-                "return_score='all' not implemented for voyageai reranker"
-            )
+
         return combined_results
 
     def rerank_vector(self, query: str, vector_results: pa.Table):

python/python/lancedb/table.py

@@ -827,7 +827,7 @@ class Table(ABC):
         ordering_field_names: Optional[Union[str, List[str]]] = None,
         replace: bool = False,
         writer_heap_size: Optional[int] = 1024 * 1024 * 1024,
-        use_tantivy: bool = True,
+        use_tantivy: bool = False,
         tokenizer_name: Optional[str] = None,
         with_position: bool = False,
         # tokenizer configs:
@@ -838,6 +838,9 @@ class Table(ABC):
         stem: bool = True,
         remove_stop_words: bool = True,
         ascii_folding: bool = True,
+        ngram_min_length: int = 3,
+        ngram_max_length: int = 3,
+        prefix_only: bool = False,
         wait_timeout: Optional[timedelta] = None,
     ):
         """Create a full-text search index on the table.
@@ -864,7 +867,7 @@ class Table(ABC):
             The tokenizer to use for the index. Can be "raw", "default" or the 2 letter
             language code followed by "_stem". So for english it would be "en_stem".
             For available languages see: https://docs.rs/tantivy/latest/tantivy/tokenizer/enum.Language.html
-        use_tantivy: bool, default True
+        use_tantivy: bool, default False
             If True, use the legacy full-text search implementation based on tantivy.
             If False, use the new full-text search implementation based on lance-index.
         with_position: bool, default False
@@ -877,6 +880,7 @@ class Table(ABC):
             - "simple": Splits text by whitespace and punctuation.
             - "whitespace": Split text by whitespace, but not punctuation.
             - "raw": No tokenization. The entire text is treated as a single token.
+            - "ngram": N-Gram tokenizer.
         language : str, default "English"
             The language to use for tokenization.
         max_token_length : int, default 40
@@ -894,6 +898,12 @@ class Table(ABC):
         ascii_folding : bool, default True
             Whether to fold ASCII characters. This converts accented characters to
             their ASCII equivalent. For example, "café" would be converted to "cafe".
+        ngram_min_length: int, default 3
+            The minimum length of an n-gram.
+        ngram_max_length: int, default 3
+            The maximum length of an n-gram.
+        prefix_only: bool, default False
+            Whether to only index the prefix of the token for ngram tokenizer.
         wait_timeout: timedelta, optional
             The timeout to wait if indexing is asynchronous.
         """
@@ -1970,7 +1980,7 @@ class LanceTable(Table):
         ordering_field_names: Optional[Union[str, List[str]]] = None,
         replace: bool = False,
         writer_heap_size: Optional[int] = 1024 * 1024 * 1024,
-        use_tantivy: bool = True,
+        use_tantivy: bool = False,
         tokenizer_name: Optional[str] = None,
         with_position: bool = False,
         # tokenizer configs:
@@ -1981,6 +1991,9 @@ class LanceTable(Table):
         stem: bool = True,
         remove_stop_words: bool = True,
         ascii_folding: bool = True,
+        ngram_min_length: int = 3,
+        ngram_max_length: int = 3,
+        prefix_only: bool = False,
     ):
         if not use_tantivy:
             if not isinstance(field_names, str):
@@ -1996,6 +2009,9 @@ class LanceTable(Table):
                     "stem": stem,
                     "remove_stop_words": remove_stop_words,
                     "ascii_folding": ascii_folding,
+                    "ngram_min_length": ngram_min_length,
+                    "ngram_max_length": ngram_max_length,
+                    "prefix_only": prefix_only,
                 }
             else:
                 tokenizer_configs = self.infer_tokenizer_configs(tokenizer_name)
@@ -2065,6 +2081,9 @@ class LanceTable(Table):
                 "stem": False,
                 "remove_stop_words": False,
                 "ascii_folding": False,
+                "ngram_min_length": 3,
+                "ngram_max_length": 3,
+                "prefix_only": False,
             }
         elif tokenizer_name == "raw":
             return {
@@ -2075,6 +2094,9 @@ class LanceTable(Table):
                 "stem": False,
                 "remove_stop_words": False,
                 "ascii_folding": False,
+                "ngram_min_length": 3,
+                "ngram_max_length": 3,
+                "prefix_only": False,
             }
         elif tokenizer_name == "whitespace":
             return {
@@ -2085,6 +2107,9 @@ class LanceTable(Table):
                 "stem": False,
                 "remove_stop_words": False,
                 "ascii_folding": False,
+                "ngram_min_length": 3,
+                "ngram_max_length": 3,
+                "prefix_only": False,
             }
 
         # or it's with language stemming with pattern like "en_stem"
@@ -2103,6 +2128,9 @@ class LanceTable(Table):
             "stem": True,
             "remove_stop_words": False,
             "ascii_folding": False,
+            "ngram_min_length": 3,
+            "ngram_max_length": 3,
+            "prefix_only": False,
         }
 
     def add(

python/python/lancedb/types.py

@@ -25,4 +25,4 @@ IndexType = Literal[
 ]
 
 # Tokenizer literals
-BaseTokenizerType = Literal["simple", "raw", "whitespace"]
+BaseTokenizerType = Literal["simple", "raw", "whitespace", "ngram"]

python/python/tests/docs/test_search.py

@@ -6,7 +6,7 @@ import lancedb
 
 # --8<-- [end:import-lancedb]
 # --8<-- [start:import-numpy]
-from lancedb.query import BoostQuery, MatchQuery
+from lancedb.query import BooleanQuery, BoostQuery, MatchQuery, Occur
 import numpy as np
 import pyarrow as pa
 
@@ -191,6 +191,15 @@ def test_fts_fuzzy_query():
         "food",  # 1 insertion
     }
 
+    results = table.search(
+        MatchQuery("foo", "text", fuzziness=1, prefix_length=3)
+    ).to_pandas()
+    assert len(results) == 2
+    assert set(results["text"].to_list()) == {
+        "foo",
+        "food",
+    }
+
 
 @pytest.mark.skipif(
     os.name == "nt", reason="Need to fix https://github.com/lancedb/lance/issues/3905"
@@ -240,6 +249,60 @@ def test_fts_boost_query():
     )
 
 
+@pytest.mark.skipif(
+    os.name == "nt", reason="Need to fix https://github.com/lancedb/lance/issues/3905"
+)
+def test_fts_boolean_query(tmp_path):
+    uri = tmp_path / "boolean-example"
+    db = lancedb.connect(uri)
+    table = db.create_table(
+        "my_table_fts_boolean",
+        data=[
+            {"text": "The cat and dog are playing"},
+            {"text": "The cat is sleeping"},
+            {"text": "The dog is barking"},
+            {"text": "The dog chases the cat"},
+        ],
+        mode="overwrite",
+    )
+    table.create_fts_index("text", use_tantivy=False, replace=True)
+
+    # SHOULD
+    results = table.search(
+        MatchQuery("cat", "text") | MatchQuery("dog", "text")
+    ).to_pandas()
+    assert len(results) == 4
+    assert set(results["text"].to_list()) == {
+        "The cat and dog are playing",
+        "The cat is sleeping",
+        "The dog is barking",
+        "The dog chases the cat",
+    }
+    # MUST
+    results = table.search(
+        MatchQuery("cat", "text") & MatchQuery("dog", "text")
+    ).to_pandas()
+    assert len(results) == 2
+    assert set(results["text"].to_list()) == {
+        "The cat and dog are playing",
+        "The dog chases the cat",
+    }
+
+    # MUST NOT
+    results = table.search(
+        BooleanQuery(
+            [
+                (Occur.MUST, MatchQuery("cat", "text")),
+                (Occur.MUST_NOT, MatchQuery("dog", "text")),
+            ]
+        )
+    ).to_pandas()
+    assert len(results) == 1
+    assert set(results["text"].to_list()) == {
+        "The cat is sleeping",
+    }
+
+
 @pytest.mark.skipif(
     os.name == "nt", reason="Need to fix https://github.com/lancedb/lance/issues/3905"
 )

python/python/tests/test_fts.py

@@ -669,3 +669,46 @@ def test_fts_on_list(mem_db: DBConnection):
 
     res = table.search(PhraseQuery("lance database", "text")).limit(5).to_list()
     assert len(res) == 2
+
+
+def test_fts_ngram(mem_db: DBConnection):
+    data = pa.table({"text": ["hello world", "lance database", "lance is cool"]})
+    table = mem_db.create_table("test", data=data)
+    table.create_fts_index("text", use_tantivy=False, base_tokenizer="ngram")
+
+    results = table.search("lan", query_type="fts").limit(10).to_list()
+    assert len(results) == 2
+    assert set(r["text"] for r in results) == {"lance database", "lance is cool"}
+
+    results = (
+        table.search("nce", query_type="fts").limit(10).to_list()
+    )  # spellchecker:disable-line
+    assert len(results) == 2
+    assert set(r["text"] for r in results) == {"lance database", "lance is cool"}
+
+    # the default min_ngram_length is 3, so "la" should not match
+    results = table.search("la", query_type="fts").limit(10).to_list()
+    assert len(results) == 0
+
+    # test setting min_ngram_length and prefix_only
+    table.create_fts_index(
+        "text",
+        use_tantivy=False,
+        base_tokenizer="ngram",
+        replace=True,
+        ngram_min_length=2,
+        prefix_only=True,
+    )
+
+    results = table.search("lan", query_type="fts").limit(10).to_list()
+    assert len(results) == 2
+    assert set(r["text"] for r in results) == {"lance database", "lance is cool"}
+
+    results = (
+        table.search("nce", query_type="fts").limit(10).to_list()
+    )  # spellchecker:disable-line
+    assert len(results) == 0
+
+    results = table.search("la", query_type="fts").limit(10).to_list()
+    assert len(results) == 2
+    assert set(r["text"] for r in results) == {"lance database", "lance is cool"}

python/python/tests/test_query.py

@@ -272,7 +272,9 @@ async def test_distance_range_with_new_rows_async():
     # append more rows so that execution plan would be mixed with ANN & Flat KNN
     new_data = pa.table(
         {
-            "vector": pa.FixedShapeTensorArray.from_numpy_ndarray(np.random.rand(4, 2)),
+            "vector": pa.FixedShapeTensorArray.from_numpy_ndarray(
+                np.random.rand(4, 2) + 1
+            ),
         }
     )
     await table.add(new_data)
@@ -775,6 +777,82 @@ async def test_explain_plan_async(table_async: AsyncTable):
     assert "KNN" in plan
 
 
+@pytest.mark.asyncio
+async def test_explain_plan_fts(table_async: AsyncTable):
+    """Test explain plan for FTS queries"""
+    # Create FTS index
+    from lancedb.index import FTS
+
+    await table_async.create_index("text", config=FTS())
+
+    # Test pure FTS query
+    query = await table_async.search("dog", query_type="fts", fts_columns="text")
+    plan = await query.explain_plan()
+    # Should show FTS details (issue #2465 is now fixed)
+    assert "MatchQuery: query=dog" in plan
+    assert "GlobalLimitExec" in plan  # Default limit
+
+    # Test FTS query with limit
+    query_with_limit = await table_async.search(
+        "dog", query_type="fts", fts_columns="text"
+    )
+    plan_with_limit = await query_with_limit.limit(1).explain_plan()
+    assert "MatchQuery: query=dog" in plan_with_limit
+    assert "GlobalLimitExec: skip=0, fetch=1" in plan_with_limit
+
+    # Test FTS query with offset and limit
+    query_with_offset = await table_async.search(
+        "dog", query_type="fts", fts_columns="text"
+    )
+    plan_with_offset = await query_with_offset.offset(1).limit(1).explain_plan()
+    assert "MatchQuery: query=dog" in plan_with_offset
+    assert "GlobalLimitExec: skip=1, fetch=1" in plan_with_offset
+
+
+@pytest.mark.asyncio
+async def test_explain_plan_vector_with_limit_offset(table_async: AsyncTable):
+    """Test explain plan for vector queries with limit and offset"""
+    # Test vector query with limit
+    plan_with_limit = await (
+        table_async.query().nearest_to(pa.array([1, 2])).limit(1).explain_plan()
+    )
+    assert "KNN" in plan_with_limit
+    assert "GlobalLimitExec: skip=0, fetch=1" in plan_with_limit
+
+    # Test vector query with offset and limit
+    plan_with_offset = await (
+        table_async.query()
+        .nearest_to(pa.array([1, 2]))
+        .offset(1)
+        .limit(1)
+        .explain_plan()
+    )
+    assert "KNN" in plan_with_offset
+    assert "GlobalLimitExec: skip=1, fetch=1" in plan_with_offset
+
+
+@pytest.mark.asyncio
+async def test_explain_plan_with_filters(table_async: AsyncTable):
+    """Test explain plan for queries with filters"""
+    # Test vector query with filter
+    plan_with_filter = await (
+        table_async.query().nearest_to(pa.array([1, 2])).where("id = 1").explain_plan()
+    )
+    assert "KNN" in plan_with_filter
+    assert "FilterExec" in plan_with_filter
+
+    # Test FTS query with filter
+    from lancedb.index import FTS
+
+    await table_async.create_index("text", config=FTS())
+    query_fts_filter = await table_async.search(
+        "dog", query_type="fts", fts_columns="text"
+    )
+    plan_fts_filter = await query_fts_filter.where("id = 1").explain_plan()
+    assert "MatchQuery: query=dog" in plan_fts_filter
+    assert "FilterExec: id@" in plan_fts_filter  # Should show filter details
+
+
 @pytest.mark.asyncio
 async def test_query_camelcase_async(tmp_path):
     db = await lancedb.connect_async(tmp_path)

python/python/tests/test_remote_db.py

@@ -210,6 +210,25 @@ async def test_retry_error():
         assert cause.status_code == 429
 
 
+def test_table_unimplemented_functions():
+    def handler(request):
+        if request.path == "/v1/table/test/create/?mode=create":
+            request.send_response(200)
+            request.send_header("Content-Type", "application/json")
+            request.end_headers()
+            request.wfile.write(b"{}")
+        else:
+            request.send_response(404)
+            request.end_headers()
+
+    with mock_lancedb_connection(handler) as db:
+        table = db.create_table("test", [{"id": 1}])
+        with pytest.raises(NotImplementedError):
+            table.to_arrow()
+        with pytest.raises(NotImplementedError):
+            table.to_pandas()
+
+
 def test_table_add_in_threadpool():
     def handler(request):
         if request.path == "/v1/table/test/insert/":

python/python/tests/test_rerankers.py

@@ -499,3 +499,19 @@ def test_empty_result_reranker():
             .rerank(reranker)
             .to_arrow()
         )
+
+
+@pytest.mark.parametrize("use_tantivy", [True, False])
+def test_cross_encoder_reranker_return_all(tmp_path, use_tantivy):
+    pytest.importorskip("sentence_transformers")
+    reranker = CrossEncoderReranker(return_score="all")
+    table, schema = get_test_table(tmp_path, use_tantivy)
+    query = "single player experience"
+    result = (
+        table.search(query, query_type="hybrid", vector_column_name="vector")
+        .rerank(reranker=reranker)
+        .to_arrow()
+    )
+    assert "_relevance_score" in result.column_names
+    assert "_score" in result.column_names
+    assert "_distance" in result.column_names

python/python/tests/test_s3.py

@@ -245,7 +245,7 @@ def test_s3_dynamodb_sync(s3_bucket: str, commit_table: str, monkeypatch):
         NotImplementedError,
         match="Full-text search is only supported on the local filesystem",
     ):
-        table.create_fts_index("x")
+        table.create_fts_index("x", use_tantivy=True)
 
     # make sure list tables still works
     assert db.table_names() == ["test_ddb_sync"]

python/src/index.rs

@@ -47,7 +47,10 @@ pub fn extract_index_params(source: &Option<Bound<'_, PyAny>>) -> PyResult<Lance
                     .max_token_length(params.max_token_length)
                     .remove_stop_words(params.remove_stop_words)
                     .stem(params.stem)
-                    .ascii_folding(params.ascii_folding);
+                    .ascii_folding(params.ascii_folding)
+                    .ngram_min_length(params.ngram_min_length)
+                    .ngram_max_length(params.ngram_max_length)
+                    .ngram_prefix_only(params.prefix_only);
                 Ok(LanceDbIndex::FTS(inner_opts))
             },
             "IvfFlat" => {
@@ -130,6 +133,9 @@ struct FtsParams {
     stem: bool,
     remove_stop_words: bool,
     ascii_folding: bool,
+    ngram_min_length: u32,
+    ngram_max_length: u32,
+    prefix_only: bool,
 }
 
 #[derive(FromPyObject)]

python/src/query.rs

@@ -50,8 +50,9 @@ impl FromPyObject<'_> for PyLanceDB<FtsQuery> {
                 let fuzziness = ob.getattr("fuzziness")?.extract()?;
                 let max_expansions = ob.getattr("max_expansions")?.extract()?;
                 let operator = ob.getattr("operator")?.extract::<String>()?;
+                let prefix_length = ob.getattr("prefix_length")?.extract()?;
 
-                Ok(PyLanceDB(
+                Ok(Self(
                     MatchQuery::new(query)
                         .with_column(Some(column))
                         .with_boost(boost)
@@ -60,6 +61,7 @@ impl FromPyObject<'_> for PyLanceDB<FtsQuery> {
                         .with_operator(Operator::try_from(operator.as_str()).map_err(|e| {
                             PyValueError::new_err(format!("Invalid operator: {}", e))
                         })?)
+                        .with_prefix_length(prefix_length)
                         .into(),
                 ))
             }
@@ -68,7 +70,7 @@ impl FromPyObject<'_> for PyLanceDB<FtsQuery> {
                 let column = ob.getattr("column")?.extract()?;
                 let slop = ob.getattr("slop")?.extract()?;
 
-                Ok(PyLanceDB(
+                Ok(Self(
                     PhraseQuery::new(query)
                         .with_column(Some(column))
                         .with_slop(slop)
@@ -76,10 +78,10 @@ impl FromPyObject<'_> for PyLanceDB<FtsQuery> {
                 ))
             }
             "BoostQuery" => {
-                let positive: PyLanceDB<FtsQuery> = ob.getattr("positive")?.extract()?;
-                let negative: PyLanceDB<FtsQuery> = ob.getattr("negative")?.extract()?;
+                let positive: Self = ob.getattr("positive")?.extract()?;
+                let negative: Self = ob.getattr("negative")?.extract()?;
                 let negative_boost = ob.getattr("negative_boost")?.extract()?;
-                Ok(PyLanceDB(
+                Ok(Self(
                     BoostQuery::new(positive.0, negative.0, negative_boost).into(),
                 ))
             }
@@ -101,18 +103,17 @@ impl FromPyObject<'_> for PyLanceDB<FtsQuery> {
                 let op = Operator::try_from(operator.as_str())
                     .map_err(|e| PyValueError::new_err(format!("Invalid operator: {}", e)))?;
 
-                Ok(PyLanceDB(q.with_operator(op).into()))
+                Ok(Self(q.with_operator(op).into()))
             }
             "BooleanQuery" => {
-                let queries: Vec<(String, PyLanceDB<FtsQuery>)> =
-                    ob.getattr("queries")?.extract()?;
+                let queries: Vec<(String, Self)> = ob.getattr("queries")?.extract()?;
                 let mut sub_queries = Vec::with_capacity(queries.len());
                 for (occur, q) in queries {
                     let occur = Occur::try_from(occur.as_str())
                         .map_err(|e| PyValueError::new_err(e.to_string()))?;
                     sub_queries.push((occur, q.0));
                 }
-                Ok(PyLanceDB(BooleanQuery::new(sub_queries).into()))
+                Ok(Self(BooleanQuery::new(sub_queries).into()))
             }
             name => Err(PyValueError::new_err(format!(
                 "Unsupported FTS query type: {}",
@@ -139,7 +140,8 @@ impl<'py> IntoPyObject<'py> for PyLanceDB<FtsQuery> {
                 kwargs.set_item("boost", query.boost)?;
                 kwargs.set_item("fuzziness", query.fuzziness)?;
                 kwargs.set_item("max_expansions", query.max_expansions)?;
-                kwargs.set_item("operator", operator_to_str(query.operator))?;
+                kwargs.set_item::<_, &str>("operator", query.operator.into())?;
+                kwargs.set_item("prefix_length", query.prefix_length)?;
                 namespace
                     .getattr(intern!(py, "MatchQuery"))?
                     .call((query.terms, query.column.unwrap()), Some(&kwargs))
@@ -152,8 +154,8 @@ impl<'py> IntoPyObject<'py> for PyLanceDB<FtsQuery> {
                     .call((query.terms, query.column.unwrap()), Some(&kwargs))
             }
             FtsQuery::Boost(query) => {
-                let positive = PyLanceDB(query.positive.as_ref().clone()).into_pyobject(py)?;
-                let negative = PyLanceDB(query.negative.as_ref().clone()).into_pyobject(py)?;
+                let positive = Self(query.positive.as_ref().clone()).into_pyobject(py)?;
+                let negative = Self(query.negative.as_ref().clone()).into_pyobject(py)?;
                 let kwargs = PyDict::new(py);
                 kwargs.set_item("negative_boost", query.negative_boost)?;
                 namespace
@@ -169,19 +171,25 @@ impl<'py> IntoPyObject<'py> for PyLanceDB<FtsQuery> {
                     .unzip();
                 let kwargs = PyDict::new(py);
                 kwargs.set_item("boosts", boosts)?;
-                kwargs.set_item("operator", operator_to_str(first.operator))?;
+                kwargs.set_item::<_, &str>("operator", first.operator.into())?;
                 namespace
                     .getattr(intern!(py, "MultiMatchQuery"))?
                     .call((first.terms.clone(), columns), Some(&kwargs))
             }
             FtsQuery::Boolean(query) => {
-                let mut queries = Vec::with_capacity(query.must.len() + query.should.len());
-                for q in query.must {
-                    queries.push((occur_to_str(Occur::Must), PyLanceDB(q).into_pyobject(py)?));
-                }
+                let mut queries: Vec<(&str, Bound<'py, PyAny>)> = Vec::with_capacity(
+                    query.should.len() + query.must.len() + query.must_not.len(),
+                );
                 for q in query.should {
-                    queries.push((occur_to_str(Occur::Should), PyLanceDB(q).into_pyobject(py)?));
+                    queries.push((Occur::Should.into(), Self(q).into_pyobject(py)?));
                 }
+                for q in query.must {
+                    queries.push((Occur::Must.into(), Self(q).into_pyobject(py)?));
+                }
+                for q in query.must_not {
+                    queries.push((Occur::MustNot.into(), Self(q).into_pyobject(py)?));
+                }
+
                 namespace
                     .getattr(intern!(py, "BooleanQuery"))?
                     .call1((queries,))
@@ -190,21 +198,6 @@ impl<'py> IntoPyObject<'py> for PyLanceDB<FtsQuery> {
     }
 }
 
-fn operator_to_str(op: Operator) -> &'static str {
-    match op {
-        Operator::And => "AND",
-        Operator::Or => "OR",
-    }
-}
-
-fn occur_to_str(occur: Occur) -> &'static str {
-    match occur {
-        Occur::Must => "MUST",
-        Occur::Should => "SHOULD",
-        Occur::MustNot => "MUST NOT",
-    }
-}
-
 // Python representation of query vector(s)
 #[derive(Clone)]
 pub struct PyQueryVectors(Vec<Arc<dyn Array>>);
@@ -569,7 +562,10 @@ impl FTSQuery {
     }
 
     pub fn explain_plan(self_: PyRef<'_, Self>, verbose: bool) -> PyResult<Bound<'_, PyAny>> {
-        let inner = self_.inner.clone();
+        let inner = self_
+            .inner
+            .clone()
+            .full_text_search(self_.fts_query.clone());
         future_into_py(self_.py(), async move {
             inner
                 .explain_plan(verbose)
@@ -579,7 +575,10 @@ impl FTSQuery {
     }
 
     pub fn analyze_plan(self_: PyRef<'_, Self>) -> PyResult<Bound<'_, PyAny>> {
-        let inner = self_.inner.clone();
+        let inner = self_
+            .inner
+            .clone()
+            .full_text_search(self_.fts_query.clone());
         future_into_py(self_.py(), async move {
             inner
                 .analyze_plan()

rust/ffi/node/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb-node"
-version = "0.20.1-beta.1"
+version = "0.21.2-beta.0"
 description = "Serverless, low-latency vector database for AI applications"
 license.workspace = true
 edition.workspace = true

rust/lancedb/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb"
-version = "0.20.1-beta.1"
+version = "0.21.2-beta.0"
 edition.workspace = true
 description = "LanceDB: A serverless, low-latency vector database for AI applications"
 license.workspace = true

rust/lancedb/src/catalog/listing.rs

@@ -105,7 +105,7 @@ impl ListingCatalog {
     }
 
     async fn open_path(path: &str) -> Result<Self> {
-        let (object_store, base_path) = ObjectStore::from_uri(path).await.unwrap();
+        let (object_store, base_path) = ObjectStore::from_uri(path).await?;
         if object_store.is_local() {
             Self::try_create_dir(path).context(CreateDirSnafu { path })?;
         }
@@ -216,6 +216,7 @@ impl Catalog for ListingCatalog {
             client_config: Default::default(),
             read_consistency_interval: None,
             options: Default::default(),
+            session: None,
         };
 
         // Add the db options to the connect request
@@ -243,6 +244,7 @@ impl Catalog for ListingCatalog {
             client_config: Default::default(),
             read_consistency_interval: None,
             options: Default::default(),
+            session: None,
         };
 
         // Add the db options to the connect request
@@ -312,6 +314,7 @@ mod tests {
             client_config: Default::default(),
             options: Default::default(),
             read_consistency_interval: None,
+            session: None,
         };
 
         let catalog = ListingCatalog::connect(&request).await.unwrap();
@@ -573,6 +576,7 @@ mod tests {
             client_config: Default::default(),
             options: Default::default(),
             read_consistency_interval: None,
+            session: None,
         };
 
         let catalog = ListingCatalog::connect(&request).await.unwrap();
@@ -592,6 +596,7 @@ mod tests {
             client_config: Default::default(),
             options: Default::default(),
             read_consistency_interval: None,
+            session: None,
         };
 
         let catalog = ListingCatalog::connect(&request).await.unwrap();
@@ -608,6 +613,7 @@ mod tests {
             client_config: Default::default(),
             options: Default::default(),
             read_consistency_interval: None,
+            session: None,
         };
 
         let result = ListingCatalog::connect(&request).await;

rust/lancedb/src/connection.rs

@@ -627,6 +627,12 @@ pub struct ConnectRequest {
     /// consistency only applies to read operations. Write operations are
     /// always consistent.
     pub read_consistency_interval: Option<std::time::Duration>,
+
+    /// Optional session for object stores and caching
+    ///
+    /// If provided, this session will be used instead of creating a default one.
+    /// This allows for custom configuration of object store registries, caching, etc.
+    pub session: Option<Arc<lance::session::Session>>,
 }
 
 #[derive(Debug)]
@@ -645,6 +651,7 @@ impl ConnectBuilder {
                 client_config: Default::default(),
                 read_consistency_interval: None,
                 options: HashMap::new(),
+                session: None,
             },
             embedding_registry: None,
         }
@@ -802,6 +809,20 @@ impl ConnectBuilder {
         self
     }
 
+    /// Set a custom session for object stores and caching.
+    ///
+    /// By default, a new session with default configuration will be created.
+    /// This method allows you to provide a custom session with your own
+    /// configuration for object store registries, caching, etc.
+    ///
+    /// # Arguments
+    ///
+    /// * `session` - A custom session to use for this connection
+    pub fn session(mut self, session: Arc<lance::session::Session>) -> Self {
+        self.request.session = Some(session);
+        self
+    }
+
     #[cfg(feature = "remote")]
     fn execute_remote(self) -> Result<Connection> {
         use crate::remote::db::RemoteDatabaseOptions;
@@ -884,6 +905,7 @@ impl CatalogConnectBuilder {
                 client_config: Default::default(),
                 read_consistency_interval: None,
                 options: HashMap::new(),
+                session: None,
             },
         }
     }

rust/lancedb/src/database/listing.rs

@@ -8,7 +8,7 @@ use std::path::Path;
 use std::{collections::HashMap, sync::Arc};
 
 use lance::dataset::{ReadParams, WriteMode};
-use lance::io::{ObjectStore, ObjectStoreParams, ObjectStoreRegistry, WrappingObjectStore};
+use lance::io::{ObjectStore, ObjectStoreParams, WrappingObjectStore};
 use lance_datafusion::utils::StreamingWriteSource;
 use lance_encoding::version::LanceFileVersion;
 use lance_table::io::commit::commit_handler_from_url;
@@ -217,6 +217,9 @@ pub struct ListingDatabase {
 
     // Options for tables created by this connection
     new_table_config: NewTableConfig,
+
+    // Session for object stores and caching
+    session: Arc<lance::session::Session>,
 }
 
 impl std::fmt::Display for ListingDatabase {
@@ -262,6 +265,7 @@ impl ListingDatabase {
                     uri,
                     request.read_consistency_interval,
                     options.new_table_config,
+                    request.session.clone(),
                 )
                 .await
             }
@@ -313,13 +317,20 @@ impl ListingDatabase {
 
                 let plain_uri = url.to_string();
 
-                let registry = Arc::new(ObjectStoreRegistry::default());
+                let session = request
+                    .session
+                    .clone()
+                    .unwrap_or_else(|| Arc::new(lance::session::Session::default()));
                 let os_params = ObjectStoreParams {
                     storage_options: Some(options.storage_options.clone()),
                     ..Default::default()
                 };
-                let (object_store, base_path) =
-                    ObjectStore::from_uri_and_params(registry, &plain_uri, &os_params).await?;
+                let (object_store, base_path) = ObjectStore::from_uri_and_params(
+                    session.store_registry(),
+                    &plain_uri,
+                    &os_params,
+                )
+                .await?;
                 if object_store.is_local() {
                     Self::try_create_dir(&plain_uri).context(CreateDirSnafu { path: plain_uri })?;
                 }
@@ -342,6 +353,7 @@ impl ListingDatabase {
                     read_consistency_interval: request.read_consistency_interval,
                     storage_options: options.storage_options,
                     new_table_config: options.new_table_config,
+                    session,
                 })
             }
             Err(_) => {
@@ -349,6 +361,7 @@ impl ListingDatabase {
                     uri,
                     request.read_consistency_interval,
                     options.new_table_config,
+                    request.session.clone(),
                 )
                 .await
             }
@@ -359,8 +372,15 @@ impl ListingDatabase {
         path: &str,
         read_consistency_interval: Option<std::time::Duration>,
         new_table_config: NewTableConfig,
+        session: Option<Arc<lance::session::Session>>,
     ) -> Result<Self> {
-        let (object_store, base_path) = ObjectStore::from_uri(path).await?;
+        let session = session.unwrap_or_else(|| Arc::new(lance::session::Session::default()));
+        let (object_store, base_path) = ObjectStore::from_uri_and_params(
+            session.store_registry(),
+            path,
+            &ObjectStoreParams::default(),
+        )
+        .await?;
         if object_store.is_local() {
             Self::try_create_dir(path).context(CreateDirSnafu { path })?;
         }
@@ -374,6 +394,7 @@ impl ListingDatabase {
             read_consistency_interval,
             storage_options: HashMap::new(),
             new_table_config,
+            session,
         })
     }
 
@@ -441,6 +462,128 @@ impl ListingDatabase {
         }
         Ok(())
     }
+
+    /// Inherit storage options from the connection into the target map
+    fn inherit_storage_options(&self, target: &mut HashMap<String, String>) {
+        for (key, value) in self.storage_options.iter() {
+            if !target.contains_key(key) {
+                target.insert(key.clone(), value.clone());
+            }
+        }
+    }
+
+    /// Extract storage option overrides from the request
+    fn extract_storage_overrides(
+        &self,
+        request: &CreateTableRequest,
+    ) -> Result<(Option<LanceFileVersion>, Option<bool>)> {
+        let storage_options = request
+            .write_options
+            .lance_write_params
+            .as_ref()
+            .and_then(|p| p.store_params.as_ref())
+            .and_then(|sp| sp.storage_options.as_ref());
+
+        let storage_version_override = storage_options
+            .and_then(|opts| opts.get(OPT_NEW_TABLE_STORAGE_VERSION))
+            .map(|s| s.parse::<LanceFileVersion>())
+            .transpose()?;
+
+        let v2_manifest_override = storage_options
+            .and_then(|opts| opts.get(OPT_NEW_TABLE_V2_MANIFEST_PATHS))
+            .map(|s| s.parse::<bool>())
+            .transpose()
+            .map_err(|_| Error::InvalidInput {
+                message: "enable_v2_manifest_paths must be a boolean".to_string(),
+            })?;
+
+        Ok((storage_version_override, v2_manifest_override))
+    }
+
+    /// Prepare write parameters for table creation
+    fn prepare_write_params(
+        &self,
+        request: &CreateTableRequest,
+        storage_version_override: Option<LanceFileVersion>,
+        v2_manifest_override: Option<bool>,
+    ) -> lance::dataset::WriteParams {
+        let mut write_params = request
+            .write_options
+            .lance_write_params
+            .clone()
+            .unwrap_or_default();
+
+        // Only modify the storage options if we actually have something to
+        // inherit. There is a difference between storage_options=None and
+        // storage_options=Some({}). Using storage_options=None will cause the
+        // connection's session store registry to be used. Supplying Some({})
+        // will cause a new connection to be created, and that connection will
+        // be dropped from the cache when python GCs the table object, which
+        // confounds reuse across tables.
+        if !self.storage_options.is_empty() {
+            let storage_options = write_params
+                .store_params
+                .get_or_insert_with(Default::default)
+                .storage_options
+                .get_or_insert_with(Default::default);
+            self.inherit_storage_options(storage_options);
+        }
+
+        write_params.data_storage_version = self
+            .new_table_config
+            .data_storage_version
+            .or(storage_version_override);
+
+        if let Some(enable_v2_manifest_paths) = self
+            .new_table_config
+            .enable_v2_manifest_paths
+            .or(v2_manifest_override)
+        {
+            write_params.enable_v2_manifest_paths = enable_v2_manifest_paths;
+        }
+
+        if matches!(&request.mode, CreateTableMode::Overwrite) {
+            write_params.mode = WriteMode::Overwrite;
+        }
+
+        write_params.session = Some(self.session.clone());
+
+        write_params
+    }
+
+    /// Handle the case where table already exists based on the create mode
+    async fn handle_table_exists(
+        &self,
+        table_name: &str,
+        mode: CreateTableMode,
+        data_schema: &arrow_schema::Schema,
+    ) -> Result<Arc<dyn BaseTable>> {
+        match mode {
+            CreateTableMode::Create => Err(Error::TableAlreadyExists {
+                name: table_name.to_string(),
+            }),
+            CreateTableMode::ExistOk(callback) => {
+                let req = OpenTableRequest {
+                    name: table_name.to_string(),
+                    index_cache_size: None,
+                    lance_read_params: None,
+                };
+                let req = (callback)(req);
+                let table = self.open_table(req).await?;
+
+                let table_schema = table.schema().await?;
+
+                if table_schema.as_ref() != data_schema {
+                    return Err(Error::Schema {
+                        message: "Provided schema does not match existing table schema".to_string(),
+                    });
+                }
+
+                Ok(table)
+            }
+            CreateTableMode::Overwrite => unreachable!(),
+        }
+    }
 }
 
 #[async_trait::async_trait]
@@ -475,50 +618,14 @@ impl Database for ListingDatabase {
         Ok(f)
     }
 
-    async fn create_table(&self, mut request: CreateTableRequest) -> Result<Arc<dyn BaseTable>> {
+    async fn create_table(&self, request: CreateTableRequest) -> Result<Arc<dyn BaseTable>> {
         let table_uri = self.table_uri(&request.name)?;
-        // Inherit storage options from the connection
-        let storage_options = request
-            .write_options
-            .lance_write_params
-            .get_or_insert_with(Default::default)
-            .store_params
-            .get_or_insert_with(Default::default)
-            .storage_options
-            .get_or_insert_with(Default::default);
-        for (key, value) in self.storage_options.iter() {
-            if !storage_options.contains_key(key) {
-                storage_options.insert(key.clone(), value.clone());
-            }
-        }
 
-        let storage_options = storage_options.clone();
+        let (storage_version_override, v2_manifest_override) =
+            self.extract_storage_overrides(&request)?;
 
-        let mut write_params = request.write_options.lance_write_params.unwrap_or_default();
-
-        if let Some(storage_version) = &self.new_table_config.data_storage_version {
-            write_params.data_storage_version = Some(*storage_version);
-        } else {
-            // Allow the user to override the storage version via storage options (backwards compatibility)
-            if let Some(data_storage_version) = storage_options.get(OPT_NEW_TABLE_STORAGE_VERSION) {
-                write_params.data_storage_version = Some(data_storage_version.parse()?);
-            }
-        }
-        if let Some(enable_v2_manifest_paths) = self.new_table_config.enable_v2_manifest_paths {
-            write_params.enable_v2_manifest_paths = enable_v2_manifest_paths;
-        } else {
-            // Allow the user to override the storage version via storage options (backwards compatibility)
-            if let Some(enable_v2_manifest_paths) = storage_options
-                .get(OPT_NEW_TABLE_V2_MANIFEST_PATHS)
-                .map(|s| s.parse::<bool>().unwrap())
-            {
-                write_params.enable_v2_manifest_paths = enable_v2_manifest_paths;
-            }
-        }
-
-        if matches!(&request.mode, CreateTableMode::Overwrite) {
-            write_params.mode = WriteMode::Overwrite;
-        }
+        let write_params =
+            self.prepare_write_params(&request, storage_version_override, v2_manifest_override);
 
         let data_schema = request.data.arrow_schema();
 
@@ -533,30 +640,10 @@ impl Database for ListingDatabase {
         .await
         {
             Ok(table) => Ok(Arc::new(table)),
-            Err(Error::TableAlreadyExists { name }) => match request.mode {
-                CreateTableMode::Create => Err(Error::TableAlreadyExists { name }),
-                CreateTableMode::ExistOk(callback) => {
-                    let req = OpenTableRequest {
-                        name: request.name.clone(),
-                        index_cache_size: None,
-                        lance_read_params: None,
-                    };
-                    let req = (callback)(req);
-                    let table = self.open_table(req).await?;
-
-                    let table_schema = table.schema().await?;
-
-                    if table_schema != data_schema {
-                        return Err(Error::Schema {
-                            message: "Provided schema does not match existing table schema"
-                                .to_string(),
-                        });
-                    }
-
-                    Ok(table)
-                }
-                CreateTableMode::Overwrite => unreachable!(),
-            },
+            Err(Error::TableAlreadyExists { .. }) => {
+                self.handle_table_exists(&request.name, request.mode, &data_schema)
+                    .await
+            }
             Err(err) => Err(err),
         }
     }
@@ -564,18 +651,22 @@ impl Database for ListingDatabase {
     async fn open_table(&self, mut request: OpenTableRequest) -> Result<Arc<dyn BaseTable>> {
         let table_uri = self.table_uri(&request.name)?;
 
-        // Inherit storage options from the connection
-        let storage_options = request
-            .lance_read_params
-            .get_or_insert_with(Default::default)
-            .store_options
-            .get_or_insert_with(Default::default)
-            .storage_options
-            .get_or_insert_with(Default::default);
-        for (key, value) in self.storage_options.iter() {
-            if !storage_options.contains_key(key) {
-                storage_options.insert(key.clone(), value.clone());
-            }
+        // Only modify the storage options if we actually have something to
+        // inherit. There is a difference between storage_options=None and
+        // storage_options=Some({}). Using storage_options=None will cause the
+        // connection's session store registry to be used. Supplying Some({})
+        // will cause a new connection to be created, and that connection will
+        // be dropped from the cache when python GCs the table object, which
+        // confounds reuse across tables.
+        if !self.storage_options.is_empty() {
+            let storage_options = request
+                .lance_read_params
+                .get_or_insert_with(Default::default)
+                .store_options
+                .get_or_insert_with(Default::default)
+                .storage_options
+                .get_or_insert_with(Default::default);
+            self.inherit_storage_options(storage_options);
         }
 
         // Some ReadParams are exposed in the OpenTableBuilder, but we also
@@ -584,13 +675,14 @@ impl Database for ListingDatabase {
         // If we have a user provided ReadParams use that
         // If we don't then start with the default ReadParams and customize it with
         // the options from the OpenTableBuilder
-        let read_params = request.lance_read_params.unwrap_or_else(|| {
+        let mut read_params = request.lance_read_params.unwrap_or_else(|| {
             let mut default_params = ReadParams::default();
             if let Some(index_cache_size) = request.index_cache_size {
                 default_params.index_cache_size = index_cache_size as usize;
             }
             default_params
         });
+        read_params.session(self.session.clone());
 
         let native_table = Arc::new(
             NativeTable::open_with_params(

rust/lancedb/src/io/object_store.rs

@@ -107,7 +107,7 @@ impl ObjectStore for MirroringObjectStore {
         self.primary.delete(location).await
     }
 
-    fn list(&self, prefix: Option<&Path>) -> BoxStream<'_, Result<ObjectMeta>> {
+    fn list(&self, prefix: Option<&Path>) -> BoxStream<'static, Result<ObjectMeta>> {
         self.primary.list(prefix)
     }
 

rust/lancedb/src/io/object_store/io_tracking.rs

@@ -119,7 +119,7 @@ impl ObjectStore for IoTrackingStore {
         let result = self.target.get(location).await;
         if let Ok(result) = &result {
             let num_bytes = result.range.end - result.range.start;
-            self.record_read(num_bytes as u64);
+            self.record_read(num_bytes);
         }
         result
     }
@@ -128,12 +128,12 @@ impl ObjectStore for IoTrackingStore {
         let result = self.target.get_opts(location, options).await;
         if let Ok(result) = &result {
             let num_bytes = result.range.end - result.range.start;
-            self.record_read(num_bytes as u64);
+            self.record_read(num_bytes);
         }
         result
     }
 
-    async fn get_range(&self, location: &Path, range: std::ops::Range<usize>) -> OSResult<Bytes> {
+    async fn get_range(&self, location: &Path, range: std::ops::Range<u64>) -> OSResult<Bytes> {
         let result = self.target.get_range(location, range).await;
         if let Ok(result) = &result {
             self.record_read(result.len() as u64);
@@ -144,7 +144,7 @@ impl ObjectStore for IoTrackingStore {
     async fn get_ranges(
         &self,
         location: &Path,
-        ranges: &[std::ops::Range<usize>],
+        ranges: &[std::ops::Range<u64>],
     ) -> OSResult<Vec<Bytes>> {
         let result = self.target.get_ranges(location, ranges).await;
         if let Ok(result) = &result {
@@ -170,7 +170,7 @@ impl ObjectStore for IoTrackingStore {
         self.target.delete_stream(locations)
     }
 
-    fn list(&self, prefix: Option<&Path>) -> BoxStream<'_, OSResult<ObjectMeta>> {
+    fn list(&self, prefix: Option<&Path>) -> BoxStream<'static, OSResult<ObjectMeta>> {
         self.record_read(0);
         self.target.list(prefix)
     }
@@ -179,7 +179,7 @@ impl ObjectStore for IoTrackingStore {
         &self,
         prefix: Option<&Path>,
         offset: &Path,
-    ) -> BoxStream<'_, OSResult<ObjectMeta>> {
+    ) -> BoxStream<'static, OSResult<ObjectMeta>> {
         self.record_read(0);
         self.target.list_with_offset(prefix, offset)
     }

rust/lancedb/src/remote/table.rs

@@ -57,6 +57,8 @@ use crate::{
 };
 
 const REQUEST_TIMEOUT_HEADER: HeaderName = HeaderName::from_static("x-request-timeout-ms");
+const METRIC_TYPE_KEY: &str = "metric_type";
+const INDEX_TYPE_KEY: &str = "index_type";
 
 pub struct RemoteTags<'a, S: HttpSend = Sender> {
     inner: &'a RemoteTable<S>,
@@ -997,23 +999,53 @@ impl<S: HttpSend> BaseTable for RemoteTable<S> {
             "column": column
         });
 
-        let (index_type, distance_type) = match index.index {
+        match index.index {
             // TODO: Should we pass the actual index parameters? SaaS does not
             // yet support them.
-            Index::IvfFlat(index) => ("IVF_FLAT", Some(index.distance_type)),
-            Index::IvfPq(index) => ("IVF_PQ", Some(index.distance_type)),
-            Index::IvfHnswSq(index) => ("IVF_HNSW_SQ", Some(index.distance_type)),
-            Index::BTree(_) => ("BTREE", None),
-            Index::Bitmap(_) => ("BITMAP", None),
-            Index::LabelList(_) => ("LABEL_LIST", None),
+            Index::IvfFlat(index) => {
+                body[INDEX_TYPE_KEY] = serde_json::Value::String("IVF_FLAT".to_string());
+                body[METRIC_TYPE_KEY] =
+                    serde_json::Value::String(index.distance_type.to_string().to_lowercase());
+                if let Some(num_partitions) = index.num_partitions {
+                    body["num_partitions"] = serde_json::Value::Number(num_partitions.into());
+                }
+            }
+            Index::IvfPq(index) => {
+                body[INDEX_TYPE_KEY] = serde_json::Value::String("IVF_PQ".to_string());
+                body[METRIC_TYPE_KEY] =
+                    serde_json::Value::String(index.distance_type.to_string().to_lowercase());
+                if let Some(num_partitions) = index.num_partitions {
+                    body["num_partitions"] = serde_json::Value::Number(num_partitions.into());
+                }
+                if let Some(num_bits) = index.num_bits {
+                    body["num_bits"] = serde_json::Value::Number(num_bits.into());
+                }
+            }
+            Index::IvfHnswSq(index) => {
+                body[INDEX_TYPE_KEY] = serde_json::Value::String("IVF_HNSW_SQ".to_string());
+                body[METRIC_TYPE_KEY] =
+                    serde_json::Value::String(index.distance_type.to_string().to_lowercase());
+                if let Some(num_partitions) = index.num_partitions {
+                    body["num_partitions"] = serde_json::Value::Number(num_partitions.into());
+                }
+            }
+            Index::BTree(_) => {
+                body[INDEX_TYPE_KEY] = serde_json::Value::String("BTREE".to_string());
+            }
+            Index::Bitmap(_) => {
+                body[INDEX_TYPE_KEY] = serde_json::Value::String("BITMAP".to_string());
+            }
+            Index::LabelList(_) => {
+                body[INDEX_TYPE_KEY] = serde_json::Value::String("LABEL_LIST".to_string());
+            }
             Index::FTS(fts) => {
+                body[INDEX_TYPE_KEY] = serde_json::Value::String("FTS".to_string());
                 let params = serde_json::to_value(&fts).map_err(|e| Error::InvalidInput {
                     message: format!("failed to serialize FTS index params {:?}", e),
                 })?;
                 for (key, value) in params.as_object().unwrap() {
                     body[key] = value.clone();
                 }
-                ("FTS", None)
             }
             Index::Auto => {
                 let schema = self.schema().await?;
@@ -1023,9 +1055,11 @@ impl<S: HttpSend> BaseTable for RemoteTable<S> {
                         message: format!("Column {} not found in schema", column),
                     })?;
                 if supported_vector_data_type(field.data_type()) {
-                    ("IVF_PQ", Some(DistanceType::L2))
+                    body[INDEX_TYPE_KEY] = serde_json::Value::String("IVF_PQ".to_string());
+                    body[METRIC_TYPE_KEY] =
+                        serde_json::Value::String(DistanceType::L2.to_string().to_lowercase());
                 } else if supported_btree_data_type(field.data_type()) {
-                    ("BTREE", None)
+                    body[INDEX_TYPE_KEY] = serde_json::Value::String("BTREE".to_string());
                 } else {
                     return Err(Error::NotSupported {
                         message: format!(
@@ -1042,12 +1076,6 @@ impl<S: HttpSend> BaseTable for RemoteTable<S> {
                 })
             }
         };
-        body["index_type"] = serde_json::Value::String(index_type.into());
-        if let Some(distance_type) = distance_type {
-            // Phalanx expects this to be lowercase right now.
-            body["metric_type"] =
-                serde_json::Value::String(distance_type.to_string().to_lowercase());
-        }
 
         let request = request.json(&body);
 
@@ -1429,11 +1457,12 @@ mod tests {
     use chrono::{DateTime, Utc};
     use futures::{future::BoxFuture, StreamExt, TryFutureExt};
     use lance_index::scalar::inverted::query::MatchQuery;
-    use lance_index::scalar::FullTextSearchQuery;
+    use lance_index::scalar::{FullTextSearchQuery, InvertedIndexParams};
     use reqwest::Body;
     use rstest::rstest;
+    use serde_json::json;
 
-    use crate::index::vector::IvfFlatIndexBuilder;
+    use crate::index::vector::{IvfFlatIndexBuilder, IvfHnswSqIndexBuilder};
     use crate::remote::db::DEFAULT_SERVER_VERSION;
     use crate::remote::JSON_CONTENT_TYPE;
     use crate::{
@@ -2433,29 +2462,79 @@ mod tests {
         let cases = [
             (
                 "IVF_FLAT",
-                Some("hamming"),
+                json!({
+                    "metric_type": "hamming",
+                }),
                 Index::IvfFlat(IvfFlatIndexBuilder::default().distance_type(DistanceType::Hamming)),
             ),
-            ("IVF_PQ", Some("l2"), Index::IvfPq(Default::default())),
+            (
+                "IVF_FLAT",
+                json!({
+                    "metric_type": "hamming",
+                    "num_partitions": 128,
+                }),
+                Index::IvfFlat(
+                    IvfFlatIndexBuilder::default()
+                        .distance_type(DistanceType::Hamming)
+                        .num_partitions(128),
+                ),
+            ),
             (
                 "IVF_PQ",
-                Some("cosine"),
-                Index::IvfPq(IvfPqIndexBuilder::default().distance_type(DistanceType::Cosine)),
+                json!({
+                    "metric_type": "l2",
+                }),
+                Index::IvfPq(Default::default()),
+            ),
+            (
+                "IVF_PQ",
+                json!({
+                    "metric_type": "cosine",
+                    "num_partitions": 128,
+                    "num_bits": 4,
+                }),
+                Index::IvfPq(
+                    IvfPqIndexBuilder::default()
+                        .distance_type(DistanceType::Cosine)
+                        .num_partitions(128)
+                        .num_bits(4),
+                ),
             ),
             (
                 "IVF_HNSW_SQ",
-                Some("l2"),
+                json!({
+                    "metric_type": "l2",
+                }),
                 Index::IvfHnswSq(Default::default()),
             ),
+            (
+                "IVF_HNSW_SQ",
+                json!({
+                    "metric_type": "l2",
+                    "num_partitions": 128,
+                }),
+                Index::IvfHnswSq(
+                    IvfHnswSqIndexBuilder::default()
+                        .distance_type(DistanceType::L2)
+                        .num_partitions(128),
+                ),
+            ),
             // HNSW_PQ isn't yet supported on SaaS
-            ("BTREE", None, Index::BTree(Default::default())),
-            ("BITMAP", None, Index::Bitmap(Default::default())),
-            ("LABEL_LIST", None, Index::LabelList(Default::default())),
-            ("FTS", None, Index::FTS(Default::default())),
+            ("BTREE", json!({}), Index::BTree(Default::default())),
+            ("BITMAP", json!({}), Index::Bitmap(Default::default())),
+            (
+                "LABEL_LIST",
+                json!({}),
+                Index::LabelList(Default::default()),
+            ),
+            (
+                "FTS",
+                serde_json::to_value(InvertedIndexParams::default()).unwrap(),
+                Index::FTS(Default::default()),
+            ),
         ];
 
-        for (index_type, distance_type, index) in cases {
-            let params = index.clone();
+        for (index_type, expected_body, index) in cases {
             let table = Table::new_with_handler("my_table", move |request| {
                 assert_eq!(request.method(), "POST");
                 assert_eq!(request.url().path(), "/v1/table/my_table/create_index/");
@@ -2465,19 +2544,9 @@ mod tests {
                 );
                 let body = request.body().unwrap().as_bytes().unwrap();
                 let body: serde_json::Value = serde_json::from_slice(body).unwrap();
-                let mut expected_body = serde_json::json!({
-                    "column": "a",
-                    "index_type": index_type,
-                });
-                if let Some(distance_type) = distance_type {
-                    expected_body["metric_type"] = distance_type.to_lowercase().into();
-                }
-                if let Index::FTS(fts) = &params {
-                    let params = serde_json::to_value(fts).unwrap();
-                    for (key, value) in params.as_object().unwrap() {
-                        expected_body[key] = value.clone();
-                    }
-                }
+                let mut expected_body = expected_body.clone();
+                expected_body["column"] = "a".into();
+                expected_body[INDEX_TYPE_KEY] = index_type.into();
 
                 assert_eq!(body, expected_body);
 

rust/lancedb/src/table/datafusion.rs

@@ -392,9 +392,18 @@ pub mod tests {
                 } else {
                     expected_line.trim()
                 };
-                assert_eq!(&actual_trimmed[..expected_trimmed.len()], expected_trimmed);
+                assert_eq!(
+                    &actual_trimmed[..expected_trimmed.len()],
+                    expected_trimmed,
+                    "\nactual:\n{physical_plan}\nexpected:\n{expected}"
+                );
             }
-            assert_eq!(lines_checked, expected.lines().count());
+            assert_eq!(
+                lines_checked,
+                expected.lines().count(),
+                "\nlines_checked:\n{lines_checked}\nexpected:\n{}",
+                expected.lines().count()
+            );
         }
     }
 
@@ -477,9 +486,9 @@ pub mod tests {
         TestFixture::check_plan(
             plan,
             "MetadataEraserExec
-             RepartitionExec:...
              CoalesceBatchesExec:...
              FilterExec: i@0 >= 5
+             RepartitionExec:...
              ProjectionExec:...
              LanceScan:...",
         )

rust/lancedb/src/table/dataset.rs

@@ -129,7 +129,9 @@ impl DatasetRef {
                 dataset: ref mut ds,
                 ..
             } => {
-                *ds = dataset;
+                if dataset.manifest().version > ds.manifest().version {
+                    *ds = dataset;
+                }
             }
             _ => unreachable!("Dataset should be in latest mode at this point"),
         }

rust/lancedb/tests/object_store_test.rs

@@ -281,6 +281,46 @@ async fn test_encryption() -> Result<()> {
     Ok(())
 }
 
+#[tokio::test]
+async fn test_table_storage_options_override() -> Result<()> {
+    // Test that table-level storage options override connection-level options
+    let bucket = S3Bucket::new("test-override").await;
+    let key1 = KMSKey::new().await;
+    let key2 = KMSKey::new().await;
+
+    let uri = format!("s3://{}", bucket.0);
+
+    // Create connection with key1 encryption
+    let db = lancedb::connect(&uri)
+        .storage_options(CONFIG.iter().cloned())
+        .storage_option("aws_server_side_encryption", "aws:kms")
+        .storage_option("aws_sse_kms_key_id", &key1.0)
+        .execute()
+        .await?;
+
+    // Create table overriding with key2 encryption
+    let data = test_data();
+    let data = RecordBatchIterator::new(vec![Ok(data.clone())], data.schema());
+    let _table = db
+        .create_table("test_override", data)
+        .storage_option("aws_sse_kms_key_id", &key2.0)
+        .execute()
+        .await?;
+
+    // Verify objects are encrypted with key2, not key1
+    validate_objects_encrypted(&bucket.0, "test_override", &key2.0).await;
+
+    // Also test that a table created without override uses connection settings
+    let data = test_data();
+    let data = RecordBatchIterator::new(vec![Ok(data.clone())], data.schema());
+    let _table2 = db.create_table("test_inherit", data).execute().await?;
+
+    // Verify this table uses key1 from connection
+    validate_objects_encrypted(&bucket.0, "test_inherit", &key1.0).await;
+
+    Ok(())
+}
+
 struct DynamoDBCommitTable(String);
 
 impl DynamoDBCommitTable {