Bump version: 0.30.0-beta.3 → 0.30.0-beta.4

Without this fix, if user directly use the native table to do operations
like `add_columns`, even if it is configured to use namespace db
connection, it is not really propagated through.

The fix is to bring lancedb's python binding up to date and do a similar
implementation as https://github.com/lance-format/lance/pull/5968, and
make sure the namespace is fully propagated through all the related
calls.

---------

Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>

.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.27.0-beta.2"
+current_version = "0.27.0-beta.3"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

.github/workflows/build_linux_wheel/action.yml

@@ -29,6 +29,7 @@ runs:
       if: ${{ inputs.arm-build == 'false' }}
       uses: PyO3/maturin-action@v1
       with:
+        maturin-version: "1.12.4"
         command: build
         working-directory: python
         docker-options: "-e PIP_EXTRA_INDEX_URL='https://pypi.fury.io/lance-format/ https://pypi.fury.io/lancedb/'"
@@ -44,6 +45,7 @@ runs:
       if: ${{ inputs.arm-build == 'true' }}
       uses: PyO3/maturin-action@v1
       with:
+        maturin-version: "1.12.4"
         command: build
         working-directory: python
         docker-options: "-e PIP_EXTRA_INDEX_URL='https://pypi.fury.io/lance-format/ https://pypi.fury.io/lancedb/'"

.github/workflows/build_mac_wheel/action.yml

@@ -20,6 +20,7 @@ runs:
       uses: PyO3/maturin-action@v1
       with:
         command: build
+        maturin-version: "1.12.4"
         # TODO: pass through interpreter
         args: ${{ inputs.args }}
         docker-options: "-e PIP_EXTRA_INDEX_URL='https://pypi.fury.io/lance-format/ https://pypi.fury.io/lancedb/'"

.github/workflows/build_windows_wheel/action.yml

@@ -25,6 +25,7 @@ runs:
       uses: PyO3/maturin-action@v1
       with:
         command: build
+        maturin-version: "1.12.4"
         args: ${{ inputs.args }}
         docker-options: "-e PIP_EXTRA_INDEX_URL='https://pypi.fury.io/lance-format/ https://pypi.fury.io/lancedb/'"
         working-directory: python

.github/workflows/npm-publish.yml

@@ -356,7 +356,8 @@ jobs:
           if [[ $DRY_RUN == "true" ]]; then
             ARGS="$ARGS --dry-run"
           fi
-          if [[ $GITHUB_REF =~ refs/tags/v(.*)-beta.* ]]; then
+          VERSION=$(node -p "require('./package.json').version")
+          if [[ $VERSION == *-* ]]; then
             ARGS="$ARGS --tag preview"
           fi
           npm publish $ARGS

.github/workflows/python.yml

@@ -10,6 +10,10 @@ on:
       - python/**
       - rust/**
       - .github/workflows/python.yml
+      - .github/workflows/build_linux_wheel/**
+      - .github/workflows/build_mac_wheel/**
+      - .github/workflows/build_windows_wheel/**
+      - .github/workflows/run_tests/**
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}

.github/workflows/rust.yml

@@ -100,7 +100,9 @@ jobs:
           lfs: true
       - uses: Swatinem/rust-cache@v2
       - name: Install dependencies
-        run: sudo apt install -y protobuf-compiler libssl-dev
+        run: |
+          sudo apt update
+          sudo apt install -y protobuf-compiler libssl-dev
       - uses: rui314/setup-mold@v1
       - name: Make Swap
         run: |

Cargo.toml

@@ -5,7 +5,7 @@ exclude = ["python"]
 resolver = "2"
 
 [workspace.package]
-edition = "2021"
+edition = "2024"
 authors = ["LanceDB Devs <dev@lancedb.com>"]
 license = "Apache-2.0"
 repository = "https://github.com/lancedb/lancedb"
@@ -15,20 +15,20 @@ categories = ["database-implementations"]
 rust-version = "1.91.0"
 
 [workspace.dependencies]
-lance = { "version" = "=3.1.0-beta.2", default-features = false, "tag" = "v3.1.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-core = { "version" = "=3.1.0-beta.2", "tag" = "v3.1.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-datagen = { "version" = "=3.1.0-beta.2", "tag" = "v3.1.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-file = { "version" = "=3.1.0-beta.2", "tag" = "v3.1.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-io = { "version" = "=3.1.0-beta.2", default-features = false, "tag" = "v3.1.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-index = { "version" = "=3.1.0-beta.2", "tag" = "v3.1.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-linalg = { "version" = "=3.1.0-beta.2", "tag" = "v3.1.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-namespace = { "version" = "=3.1.0-beta.2", "tag" = "v3.1.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-namespace-impls = { "version" = "=3.1.0-beta.2", default-features = false, "tag" = "v3.1.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-table = { "version" = "=3.1.0-beta.2", "tag" = "v3.1.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-testing = { "version" = "=3.1.0-beta.2", "tag" = "v3.1.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-datafusion = { "version" = "=3.1.0-beta.2", "tag" = "v3.1.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-encoding = { "version" = "=3.1.0-beta.2", "tag" = "v3.1.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-arrow = { "version" = "=3.1.0-beta.2", "tag" = "v3.1.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance = { "version" = "=3.0.0-rc.2", default-features = false, "tag" = "v3.0.0-rc.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-core = { "version" = "=3.0.0-rc.2", "tag" = "v3.0.0-rc.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-datagen = { "version" = "=3.0.0-rc.2", "tag" = "v3.0.0-rc.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-file = { "version" = "=3.0.0-rc.2", "tag" = "v3.0.0-rc.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-io = { "version" = "=3.0.0-rc.2", default-features = false, "tag" = "v3.0.0-rc.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-index = { "version" = "=3.0.0-rc.2", "tag" = "v3.0.0-rc.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-linalg = { "version" = "=3.0.0-rc.2", "tag" = "v3.0.0-rc.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-namespace = { "version" = "=3.0.0-rc.2", "tag" = "v3.0.0-rc.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-namespace-impls = { "version" = "=3.0.0-rc.2", default-features = false, "tag" = "v3.0.0-rc.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-table = { "version" = "=3.0.0-rc.2", "tag" = "v3.0.0-rc.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-testing = { "version" = "=3.0.0-rc.2", "tag" = "v3.0.0-rc.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-datafusion = { "version" = "=3.0.0-rc.2", "tag" = "v3.0.0-rc.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-encoding = { "version" = "=3.0.0-rc.2", "tag" = "v3.0.0-rc.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-arrow = { "version" = "=3.0.0-rc.2", "tag" = "v3.0.0-rc.2", "git" = "https://github.com/lance-format/lance.git" }
 ahash = "0.8"
 # Note that this one does not include pyarrow
 arrow = { version = "57.2", optional = false }
@@ -40,13 +40,15 @@ arrow-schema = "57.2"
 arrow-select = "57.2"
 arrow-cast = "57.2"
 async-trait = "0"
-datafusion = { version = "51.0", default-features = false }
-datafusion-catalog = "51.0"
-datafusion-common = { version = "51.0", default-features = false }
-datafusion-execution = "51.0"
-datafusion-expr = "51.0"
-datafusion-physical-plan = "51.0"
-datafusion-physical-expr = "51.0"
+datafusion = { version = "52.1", default-features = false }
+datafusion-catalog = "52.1"
+datafusion-common = { version = "52.1", default-features = false }
+datafusion-execution = "52.1"
+datafusion-expr = "52.1"
+datafusion-functions = "52.1"
+datafusion-physical-plan = "52.1"
+datafusion-physical-expr = "52.1"
+datafusion-sql = "52.1"
 env_logger = "0.11"
 half = { "version" = "2.7.1", default-features = false, features = [
     "num-traits",

docs/mkdocs.yml

@@ -52,14 +52,21 @@ plugins:
           options:
             docstring_style: numpy
             heading_level: 3
-            show_source: true
-            show_symbol_type_in_heading: true
             show_signature_annotations: true
             show_root_heading: true
+            show_docstring_examples: true
+            show_docstring_attributes: false
+            show_docstring_other_parameters: true
+            show_symbol_type_heading: true
+            show_labels: false
+            show_if_no_docstring: true
+            show_source: false
             members_order: source
             docstring_section_style: list
             signature_crossrefs: true
             separate_signature: true
+            filters:
+              - "!^_"
           import:
             # for cross references
             - https://arrow.apache.org/docs/objects.inv
@@ -113,7 +120,7 @@ markdown_extensions:
       emoji_index: !!python/name:material.extensions.emoji.twemoji
       emoji_generator: !!python/name:material.extensions.emoji.to_svg
   - markdown.extensions.toc:
-      toc_depth: 3
+      toc_depth: 4
       permalink: true
       permalink_title: Anchor link to this section
 

docs/src/java/java.md

@@ -14,7 +14,7 @@ Add the following dependency to your `pom.xml`:
 <dependency>
     <groupId>com.lancedb</groupId>
     <artifactId>lancedb-core</artifactId>
-    <version>0.27.0-beta.2</version>
+    <version>0.27.0-beta.3</version>
 </dependency>
 ```
 

docs/src/js/interfaces/DeleteResult.md

@@ -8,6 +8,14 @@
 
 ## Properties
 
+### numDeletedRows
+
+```ts
+numDeletedRows: number;
+```
+
+***
+
 ### version
 
 ```ts

java/lancedb-core/pom.xml

@@ -8,7 +8,7 @@
     <parent>
       <groupId>com.lancedb</groupId>
       <artifactId>lancedb-parent</artifactId>
-      <version>0.27.0-beta.2</version>
+      <version>0.27.0-beta.3</version>
       <relativePath>../pom.xml</relativePath>
     </parent>
 

java/pom.xml

@@ -6,7 +6,7 @@
 
     <groupId>com.lancedb</groupId>
     <artifactId>lancedb-parent</artifactId>
-    <version>0.27.0-beta.2</version>
+    <version>0.27.0-beta.3</version>
     <packaging>pom</packaging>
     <name>${project.artifactId}</name>
     <description>LanceDB Java SDK Parent POM</description>

nodejs/Cargo.toml

@@ -1,7 +1,7 @@
 [package]
 name = "lancedb-nodejs"
 edition.workspace = true
-version = "0.27.0-beta.2"
+version = "0.27.0-beta.3"
 license.workspace = true
 description.workspace = true
 repository.workspace = true

nodejs/__test__/table.test.ts

@@ -1697,6 +1697,65 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
       expect(results2[0].text).toBe(data[1].text);
     });
 
+    test("full text search fast search", async () => {
+      const db = await connect(tmpDir.name);
+      const data = [{ text: "hello world", vector: [0.1, 0.2, 0.3], id: 1 }];
+      const table = await db.createTable("test", data);
+      await table.createIndex("text", {
+        config: Index.fts(),
+      });
+
+      // Insert unindexed data after creating the index.
+      await table.add([{ text: "xyz", vector: [0.4, 0.5, 0.6], id: 2 }]);
+
+      const withFlatSearch = await table
+        .search("xyz", "fts")
+        .limit(10)
+        .toArray();
+      expect(withFlatSearch.length).toBeGreaterThan(0);
+
+      const fastSearchResults = await table
+        .search("xyz", "fts")
+        .fastSearch()
+        .limit(10)
+        .toArray();
+      expect(fastSearchResults.length).toBe(0);
+
+      const nearestToTextFastSearch = await table
+        .query()
+        .nearestToText("xyz")
+        .fastSearch()
+        .limit(10)
+        .toArray();
+      expect(nearestToTextFastSearch.length).toBe(0);
+
+      // fastSearch should be chainable with other methods.
+      const chainedFastSearch = await table
+        .search("xyz", "fts")
+        .fastSearch()
+        .select(["text"])
+        .limit(5)
+        .toArray();
+      expect(chainedFastSearch.length).toBe(0);
+
+      await table.optimize();
+
+      const indexedFastSearch = await table
+        .search("xyz", "fts")
+        .fastSearch()
+        .limit(10)
+        .toArray();
+      expect(indexedFastSearch.length).toBeGreaterThan(0);
+
+      const indexedNearestToTextFastSearch = await table
+        .query()
+        .nearestToText("xyz")
+        .fastSearch()
+        .limit(10)
+        .toArray();
+      expect(indexedNearestToTextFastSearch.length).toBeGreaterThan(0);
+    });
+
     test("prewarm full text search index", async () => {
       const db = await connect(tmpDir.name);
       const data = [

nodejs/npm/darwin-arm64/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-darwin-arm64",
-	"version": "0.27.0-beta.2",
+	"version": "0.27.0-beta.3",
 	"os": ["darwin"],
 	"cpu": ["arm64"],
 	"main": "lancedb.darwin-arm64.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-gnu",
-	"version": "0.27.0-beta.2",
+	"version": "0.27.0-beta.3",
 	"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": "0.27.0-beta.2",
+	"version": "0.27.0-beta.3",
 	"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": "0.27.0-beta.2",
+	"version": "0.27.0-beta.3",
 	"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": "0.27.0-beta.2",
+	"version": "0.27.0-beta.3",
 	"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": "0.27.0-beta.2",
+  "version": "0.27.0-beta.3",
   "os": [
     "win32"
   ],

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-win32-x64-msvc",
-	"version": "0.27.0-beta.2",
+	"version": "0.27.0-beta.3",
 	"os": ["win32"],
 	"cpu": ["x64"],
 	"main": "lancedb.win32-x64-msvc.node",

nodejs/package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "@lancedb/lancedb",
-  "version": "0.27.0-beta.2",
+  "version": "0.27.0-beta.3",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "@lancedb/lancedb",
-      "version": "0.27.0-beta.2",
+      "version": "0.27.0-beta.3",
       "cpu": [
         "x64",
         "arm64"

nodejs/package.json

@@ -11,7 +11,7 @@
     "ann"
   ],
   "private": false,
-  "version": "0.27.0-beta.2",
+  "version": "0.27.0-beta.3",
   "main": "dist/index.js",
   "exports": {
     ".": "./dist/index.js",

nodejs/src/connection.rs

@@ -8,10 +8,10 @@ use lancedb::database::{CreateTableMode, Database};
 use napi::bindgen_prelude::*;
 use napi_derive::*;
 
+use crate::ConnectionOptions;
 use crate::error::NapiErrorExt;
 use crate::header::JsHeaderProvider;
 use crate::table::Table;
-use crate::ConnectionOptions;
 use lancedb::connection::{ConnectBuilder, Connection as LanceDBConnection};
 
 use lancedb::ipc::{ipc_file_to_batches, ipc_file_to_schema};

nodejs/src/index.rs

@@ -3,12 +3,12 @@
 
 use std::sync::Mutex;
 
+use lancedb::index::Index as LanceDbIndex;
 use lancedb::index::scalar::{BTreeIndexBuilder, FtsIndexBuilder};
 use lancedb::index::vector::{
     IvfFlatIndexBuilder, IvfHnswPqIndexBuilder, IvfHnswSqIndexBuilder, IvfPqIndexBuilder,
     IvfRqIndexBuilder,
 };
-use lancedb::index::Index as LanceDbIndex;
 use napi_derive::napi;
 
 use crate::util::parse_distance_type;

nodejs/src/query.rs

@@ -17,8 +17,8 @@ use lancedb::query::VectorQuery as LanceDbVectorQuery;
 use napi::bindgen_prelude::*;
 use napi_derive::napi;
 
-use crate::error::convert_error;
 use crate::error::NapiErrorExt;
+use crate::error::convert_error;
 use crate::iterator::RecordBatchIterator;
 use crate::rerankers::RerankHybridCallbackArgs;
 use crate::rerankers::Reranker;
@@ -551,15 +551,12 @@ fn parse_fts_query(query: Object) -> napi::Result<FullTextSearchQuery> {
             }
         };
         let mut query = FullTextSearchQuery::new_query(query);
-        if let Some(cols) = columns {
-            if !cols.is_empty() {
-                query = query.with_columns(&cols).map_err(|e| {
-                    napi::Error::from_reason(format!(
-                        "Failed to set full text search columns: {}",
-                        e
-                    ))
-                })?;
-            }
+        if let Some(cols) = columns
+            && !cols.is_empty()
+        {
+            query = query.with_columns(&cols).map_err(|e| {
+                napi::Error::from_reason(format!("Failed to set full text search columns: {}", e))
+            })?;
         }
         Ok(query)
     } else {

nodejs/src/session.rs

@@ -95,7 +95,7 @@ impl napi::bindgen_prelude::FromNapiValue for Session {
         napi_val: napi::sys::napi_value,
     ) -> napi::Result<Self> {
         let object: napi::bindgen_prelude::ClassInstance<Self> =
-            napi::bindgen_prelude::ClassInstance::from_napi_value(env, napi_val)?;
+            unsafe { napi::bindgen_prelude::ClassInstance::from_napi_value(env, napi_val)? };
         Ok((*object).clone())
     }
 }

nodejs/src/table.rs

@@ -753,12 +753,14 @@ impl From<lancedb::table::AddResult> for AddResult {
 
 #[napi(object)]
 pub struct DeleteResult {
+    pub num_deleted_rows: i64,
     pub version: i64,
 }
 
 impl From<lancedb::table::DeleteResult> for DeleteResult {
     fn from(value: lancedb::table::DeleteResult) -> Self {
         Self {
+            num_deleted_rows: value.num_deleted_rows as i64,
             version: value.version as i64,
         }
     }

python/.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.30.0-beta.2"
+current_version = "0.30.0-beta.4"
 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 = "0.30.0-beta.2"
+version = "0.30.0-beta.4"
 edition.workspace = true
 description = "Python bindings for LanceDB"
 license.workspace = true
@@ -16,9 +16,11 @@ crate-type = ["cdylib"]
 [dependencies]
 arrow = { version = "57.2", features = ["pyarrow"] }
 async-trait = "0.1"
+bytes = "1"
 lancedb = { path = "../rust/lancedb", default-features = false }
 lance-core.workspace = true
 lance-namespace.workspace = true
+lance-namespace-impls.workspace = true
 lance-io.workspace = true
 env_logger.workspace = true
 pyo3 = { version = "0.26", features = ["extension-module", "abi3-py39"] }
@@ -28,6 +30,8 @@ pyo3-async-runtimes = { version = "0.26", features = [
 ] }
 pin-project = "1.1.5"
 futures.workspace = true
+serde = "1"
+serde_json = "1"
 snafu.workspace = true
 tokio = { version = "1.40", features = ["sync"] }
 

python/pyproject.toml

@@ -45,7 +45,7 @@ repository = "https://github.com/lancedb/lancedb"
 
 [project.optional-dependencies]
 pylance = [
-    "pylance>=1.0.0b14",
+    "pylance>=4.0.0b7",
 ]
 tests = [
     "aiohttp",
@@ -59,9 +59,9 @@ tests = [
     "polars>=0.19, <=1.3.0",
     "tantivy",
     "pyarrow-stubs",
-    "pylance>=1.0.0b14",
+    "pylance>=4.0.0b7",
     "requests",
-    "datafusion<52",
+    "datafusion>=52,<53",
 ]
 dev = [
     "ruff",

python/python/lancedb/db.py

@@ -8,7 +8,7 @@ from abc import abstractmethod
 from datetime import timedelta
 from pathlib import Path
 import sys
-from typing import TYPE_CHECKING, Dict, Iterable, List, Literal, Optional, Union
+from typing import TYPE_CHECKING, Any, Dict, Iterable, List, Literal, Optional, Union
 
 if sys.version_info >= (3, 12):
     from typing import override
@@ -1541,6 +1541,8 @@ class AsyncConnection(object):
         storage_options_provider: Optional["StorageOptionsProvider"] = None,
         index_cache_size: Optional[int] = None,
         location: Optional[str] = None,
+        namespace_client: Optional[Any] = None,
+        managed_versioning: Optional[bool] = None,
     ) -> AsyncTable:
         """Open a Lance Table in the database.
 
@@ -1573,6 +1575,9 @@ class AsyncConnection(object):
             The explicit location (URI) of the table. If provided, the table will be
             opened from this location instead of deriving it from the database URI
             and table name.
+        managed_versioning: bool, optional
+            Whether managed versioning is enabled for this table. If provided,
+            avoids a redundant describe_table call when namespace_client is set.
 
         Returns
         -------
@@ -1587,6 +1592,8 @@ class AsyncConnection(object):
             storage_options_provider=storage_options_provider,
             index_cache_size=index_cache_size,
             location=location,
+            namespace_client=namespace_client,
+            managed_versioning=managed_versioning,
         )
         return AsyncTable(table)
 

python/python/lancedb/namespace.py

@@ -12,7 +12,7 @@ from __future__ import annotations
 
 import asyncio
 import sys
-from typing import Dict, Iterable, List, Optional, Union
+from typing import Any, Dict, Iterable, List, Optional, Union
 
 if sys.version_info >= (3, 12):
     from typing import override
@@ -240,7 +240,7 @@ class LanceNamespaceDBConnection(DBConnection):
         session : Optional[Session]
             A session to use for this connection
         """
-        self._ns = namespace
+        self._namespace_client = namespace
         self.read_consistency_interval = read_consistency_interval
         self.storage_options = storage_options or {}
         self.session = session
@@ -269,7 +269,7 @@ class LanceNamespaceDBConnection(DBConnection):
         if namespace is None:
             namespace = []
         request = ListTablesRequest(id=namespace, page_token=page_token, limit=limit)
-        response = self._ns.list_tables(request)
+        response = self._namespace_client.list_tables(request)
         return response.tables if response.tables else []
 
     @override
@@ -309,7 +309,9 @@ class LanceNamespaceDBConnection(DBConnection):
             # Try to describe the table first to see if it exists
             try:
                 describe_request = DescribeTableRequest(id=table_id)
-                describe_response = self._ns.describe_table(describe_request)
+                describe_response = self._namespace_client.describe_table(
+                    describe_request
+                )
                 location = describe_response.location
                 namespace_storage_options = describe_response.storage_options
             except Exception:
@@ -323,7 +325,7 @@ class LanceNamespaceDBConnection(DBConnection):
                 location=None,
                 properties=self.storage_options if self.storage_options else None,
             )
-            declare_response = self._ns.declare_table(declare_request)
+            declare_response = self._namespace_client.declare_table(declare_request)
 
             if not declare_response.location:
                 raise ValueError(
@@ -353,7 +355,7 @@ class LanceNamespaceDBConnection(DBConnection):
         # Only create if namespace returned storage_options (not None)
         if storage_options_provider is None and namespace_storage_options is not None:
             storage_options_provider = LanceNamespaceStorageOptionsProvider(
-                namespace=self._ns,
+                namespace=self._namespace_client,
                 table_id=table_id,
             )
 
@@ -371,6 +373,7 @@ class LanceNamespaceDBConnection(DBConnection):
             storage_options=merged_storage_options,
             storage_options_provider=storage_options_provider,
             location=location,
+            namespace_client=self._namespace_client,
         )
 
         return tbl
@@ -389,7 +392,7 @@ class LanceNamespaceDBConnection(DBConnection):
             namespace = []
         table_id = namespace + [name]
         request = DescribeTableRequest(id=table_id)
-        response = self._ns.describe_table(request)
+        response = self._namespace_client.describe_table(request)
 
         # Merge storage options: self.storage_options < user options < namespace options
         merged_storage_options = dict(self.storage_options)
@@ -402,10 +405,14 @@ class LanceNamespaceDBConnection(DBConnection):
         # Only create if namespace returned storage_options (not None)
         if storage_options_provider is None and response.storage_options is not None:
             storage_options_provider = LanceNamespaceStorageOptionsProvider(
-                namespace=self._ns,
+                namespace=self._namespace_client,
                 table_id=table_id,
             )
 
+        # Pass managed_versioning to avoid redundant describe_table call in Rust.
+        # Convert None to False since we already have the answer from describe_table.
+        managed_versioning = response.managed_versioning is True
+
         return self._lance_table_from_uri(
             name,
             response.location,
@@ -413,6 +420,8 @@ class LanceNamespaceDBConnection(DBConnection):
             storage_options=merged_storage_options,
             storage_options_provider=storage_options_provider,
             index_cache_size=index_cache_size,
+            namespace_client=self._namespace_client,
+            managed_versioning=managed_versioning,
         )
 
     @override
@@ -422,7 +431,7 @@ class LanceNamespaceDBConnection(DBConnection):
             namespace = []
         table_id = namespace + [name]
         request = DropTableRequest(id=table_id)
-        self._ns.drop_table(request)
+        self._namespace_client.drop_table(request)
 
     @override
     def rename_table(
@@ -484,7 +493,7 @@ class LanceNamespaceDBConnection(DBConnection):
         request = ListNamespacesRequest(
             id=namespace, page_token=page_token, limit=limit
         )
-        response = self._ns.list_namespaces(request)
+        response = self._namespace_client.list_namespaces(request)
         return ListNamespacesResponse(
             namespaces=response.namespaces if response.namespaces else [],
             page_token=response.page_token,
@@ -520,7 +529,7 @@ class LanceNamespaceDBConnection(DBConnection):
             mode=_normalize_create_namespace_mode(mode),
             properties=properties,
         )
-        response = self._ns.create_namespace(request)
+        response = self._namespace_client.create_namespace(request)
         return CreateNamespaceResponse(
             properties=response.properties if hasattr(response, "properties") else None
         )
@@ -555,7 +564,7 @@ class LanceNamespaceDBConnection(DBConnection):
             mode=_normalize_drop_namespace_mode(mode),
             behavior=_normalize_drop_namespace_behavior(behavior),
         )
-        response = self._ns.drop_namespace(request)
+        response = self._namespace_client.drop_namespace(request)
         return DropNamespaceResponse(
             properties=(
                 response.properties if hasattr(response, "properties") else None
@@ -581,7 +590,7 @@ class LanceNamespaceDBConnection(DBConnection):
             Response containing the namespace properties.
         """
         request = DescribeNamespaceRequest(id=namespace)
-        response = self._ns.describe_namespace(request)
+        response = self._namespace_client.describe_namespace(request)
         return DescribeNamespaceResponse(
             properties=response.properties if hasattr(response, "properties") else None
         )
@@ -615,7 +624,7 @@ class LanceNamespaceDBConnection(DBConnection):
         if namespace is None:
             namespace = []
         request = ListTablesRequest(id=namespace, page_token=page_token, limit=limit)
-        response = self._ns.list_tables(request)
+        response = self._namespace_client.list_tables(request)
         return ListTablesResponse(
             tables=response.tables if response.tables else [],
             page_token=response.page_token,
@@ -630,6 +639,8 @@ class LanceNamespaceDBConnection(DBConnection):
         storage_options: Optional[Dict[str, str]] = None,
         storage_options_provider: Optional[StorageOptionsProvider] = None,
         index_cache_size: Optional[int] = None,
+        namespace_client: Optional[Any] = None,
+        managed_versioning: Optional[bool] = None,
     ) -> LanceTable:
         # Open a table directly from a URI using the location parameter
         # Note: storage_options should already be merged by the caller
@@ -643,6 +654,8 @@ class LanceNamespaceDBConnection(DBConnection):
         )
 
         # Open the table using the temporary connection with the location parameter
+        # Pass namespace_client to enable managed versioning support
+        # Pass managed_versioning to avoid redundant describe_table call
         return LanceTable.open(
             temp_conn,
             name,
@@ -651,6 +664,8 @@ class LanceNamespaceDBConnection(DBConnection):
             storage_options_provider=storage_options_provider,
             index_cache_size=index_cache_size,
             location=table_uri,
+            namespace_client=namespace_client,
+            managed_versioning=managed_versioning,
         )
 
 
@@ -685,7 +700,7 @@ class AsyncLanceNamespaceDBConnection:
         session : Optional[Session]
             A session to use for this connection
         """
-        self._ns = namespace
+        self._namespace_client = namespace
         self.read_consistency_interval = read_consistency_interval
         self.storage_options = storage_options or {}
         self.session = session
@@ -713,7 +728,7 @@ class AsyncLanceNamespaceDBConnection:
         if namespace is None:
             namespace = []
         request = ListTablesRequest(id=namespace, page_token=page_token, limit=limit)
-        response = self._ns.list_tables(request)
+        response = self._namespace_client.list_tables(request)
         return response.tables if response.tables else []
 
     async def create_table(
@@ -750,7 +765,9 @@ class AsyncLanceNamespaceDBConnection:
             # Try to describe the table first to see if it exists
             try:
                 describe_request = DescribeTableRequest(id=table_id)
-                describe_response = self._ns.describe_table(describe_request)
+                describe_response = self._namespace_client.describe_table(
+                    describe_request
+                )
                 location = describe_response.location
                 namespace_storage_options = describe_response.storage_options
             except Exception:
@@ -764,7 +781,7 @@ class AsyncLanceNamespaceDBConnection:
                 location=None,
                 properties=self.storage_options if self.storage_options else None,
             )
-            declare_response = self._ns.declare_table(declare_request)
+            declare_response = self._namespace_client.declare_table(declare_request)
 
             if not declare_response.location:
                 raise ValueError(
@@ -797,7 +814,7 @@ class AsyncLanceNamespaceDBConnection:
                 and namespace_storage_options is not None
             ):
                 provider = LanceNamespaceStorageOptionsProvider(
-                    namespace=self._ns,
+                    namespace=self._namespace_client,
                     table_id=table_id,
                 )
             else:
@@ -817,6 +834,7 @@ class AsyncLanceNamespaceDBConnection:
                 storage_options=merged_storage_options,
                 storage_options_provider=provider,
                 location=location,
+                namespace_client=self._namespace_client,
             )
 
         lance_table = await asyncio.to_thread(_create_table)
@@ -837,7 +855,7 @@ class AsyncLanceNamespaceDBConnection:
             namespace = []
         table_id = namespace + [name]
         request = DescribeTableRequest(id=table_id)
-        response = self._ns.describe_table(request)
+        response = self._namespace_client.describe_table(request)
 
         # Merge storage options: self.storage_options < user options < namespace options
         merged_storage_options = dict(self.storage_options)
@@ -849,10 +867,14 @@ class AsyncLanceNamespaceDBConnection:
         # Create a storage options provider if not provided by user
         if storage_options_provider is None and response.storage_options is not None:
             storage_options_provider = LanceNamespaceStorageOptionsProvider(
-                namespace=self._ns,
+                namespace=self._namespace_client,
                 table_id=table_id,
             )
 
+        # Capture managed_versioning from describe response.
+        # Convert None to False since we already have the answer from describe_table.
+        managed_versioning = response.managed_versioning is True
+
         # Open table in a thread
         def _open_table():
             temp_conn = LanceDBConnection(
@@ -870,6 +892,8 @@ class AsyncLanceNamespaceDBConnection:
                 storage_options_provider=storage_options_provider,
                 index_cache_size=index_cache_size,
                 location=response.location,
+                namespace_client=self._namespace_client,
+                managed_versioning=managed_versioning,
             )
 
         lance_table = await asyncio.to_thread(_open_table)
@@ -881,7 +905,7 @@ class AsyncLanceNamespaceDBConnection:
             namespace = []
         table_id = namespace + [name]
         request = DropTableRequest(id=table_id)
-        self._ns.drop_table(request)
+        self._namespace_client.drop_table(request)
 
     async def rename_table(
         self,
@@ -943,7 +967,7 @@ class AsyncLanceNamespaceDBConnection:
         request = ListNamespacesRequest(
             id=namespace, page_token=page_token, limit=limit
         )
-        response = self._ns.list_namespaces(request)
+        response = self._namespace_client.list_namespaces(request)
         return ListNamespacesResponse(
             namespaces=response.namespaces if response.namespaces else [],
             page_token=response.page_token,
@@ -978,7 +1002,7 @@ class AsyncLanceNamespaceDBConnection:
             mode=_normalize_create_namespace_mode(mode),
             properties=properties,
         )
-        response = self._ns.create_namespace(request)
+        response = self._namespace_client.create_namespace(request)
         return CreateNamespaceResponse(
             properties=response.properties if hasattr(response, "properties") else None
         )
@@ -1012,7 +1036,7 @@ class AsyncLanceNamespaceDBConnection:
             mode=_normalize_drop_namespace_mode(mode),
             behavior=_normalize_drop_namespace_behavior(behavior),
         )
-        response = self._ns.drop_namespace(request)
+        response = self._namespace_client.drop_namespace(request)
         return DropNamespaceResponse(
             properties=(
                 response.properties if hasattr(response, "properties") else None
@@ -1039,7 +1063,7 @@ class AsyncLanceNamespaceDBConnection:
             Response containing the namespace properties.
         """
         request = DescribeNamespaceRequest(id=namespace)
-        response = self._ns.describe_namespace(request)
+        response = self._namespace_client.describe_namespace(request)
         return DescribeNamespaceResponse(
             properties=response.properties if hasattr(response, "properties") else None
         )
@@ -1072,7 +1096,7 @@ class AsyncLanceNamespaceDBConnection:
         if namespace is None:
             namespace = []
         request = ListTablesRequest(id=namespace, page_token=page_token, limit=limit)
-        response = self._ns.list_tables(request)
+        response = self._namespace_client.list_tables(request)
         return ListTablesResponse(
             tables=response.tables if response.tables else [],
             page_token=response.page_token,

python/python/lancedb/query.py

@@ -606,6 +606,7 @@ class LanceQueryBuilder(ABC):
                 query,
                 ordering_field_name=ordering_field_name,
                 fts_columns=fts_columns,
+                fast_search=fast_search,
             )
 
         if isinstance(query, list):
@@ -1456,12 +1457,14 @@ class LanceFtsQueryBuilder(LanceQueryBuilder):
         query: str | FullTextQuery,
         ordering_field_name: Optional[str] = None,
         fts_columns: Optional[Union[str, List[str]]] = None,
+        fast_search: bool = None,
     ):
         super().__init__(table)
         self._query = query
         self._phrase_query = False
         self.ordering_field_name = ordering_field_name
         self._reranker = None
+        self._fast_search = fast_search
         if isinstance(fts_columns, str):
             fts_columns = [fts_columns]
         self._fts_columns = fts_columns
@@ -1483,6 +1486,19 @@ class LanceFtsQueryBuilder(LanceQueryBuilder):
         self._phrase_query = phrase_query
         return self
 
+    def fast_search(self) -> LanceFtsQueryBuilder:
+        """
+        Skip a flat search of unindexed data. This will improve
+        search performance but search results will not include unindexed data.
+
+        Returns
+        -------
+        LanceFtsQueryBuilder
+            The LanceFtsQueryBuilder object.
+        """
+        self._fast_search = True
+        return self
+
     def to_query_object(self) -> Query:
         return Query(
             columns=self._columns,
@@ -1494,6 +1510,7 @@ class LanceFtsQueryBuilder(LanceQueryBuilder):
                 query=self._query, columns=self._fts_columns
             ),
             offset=self._offset,
+            fast_search=self._fast_search,
         )
 
     def output_schema(self) -> pa.Schema:

python/python/lancedb/remote/table.py

@@ -218,8 +218,6 @@ class RemoteTable(Table):
         train: bool = True,
     ):
         """Create an index on the table.
-        Currently, the only parameters that matter are
-        the metric and the vector column name.
 
         Parameters
         ----------
@@ -250,11 +248,6 @@ class RemoteTable(Table):
         >>> table.create_index("l2", "vector") # doctest: +SKIP
         """
 
-        if num_sub_vectors is not None:
-            logging.warning(
-                "num_sub_vectors is not supported on LanceDB cloud."
-                "This parameter will be tuned automatically."
-            )
         if accelerator is not None:
             logging.warning(
                 "GPU accelerator is not yet supported on LanceDB cloud."

python/python/lancedb/table.py

@@ -1331,7 +1331,7 @@ class Table(ABC):
         1  2  [3.0, 4.0]
         2  3  [5.0, 6.0]
         >>> table.delete("x = 2")
-        DeleteResult(version=2)
+        DeleteResult(num_deleted_rows=1, version=2)
         >>> table.to_pandas()
            x      vector
         0  1  [1.0, 2.0]
@@ -1345,7 +1345,7 @@ class Table(ABC):
         >>> to_remove
         '1, 5'
         >>> table.delete(f"x IN ({to_remove})")
-        DeleteResult(version=3)
+        DeleteResult(num_deleted_rows=1, version=3)
         >>> table.to_pandas()
            x      vector
         0  3  [5.0, 6.0]
@@ -1746,6 +1746,8 @@ class LanceTable(Table):
         storage_options_provider: Optional["StorageOptionsProvider"] = None,
         index_cache_size: Optional[int] = None,
         location: Optional[str] = None,
+        namespace_client: Optional[Any] = None,
+        managed_versioning: Optional[bool] = None,
         _async: AsyncTable = None,
     ):
         if namespace is None:
@@ -1753,6 +1755,7 @@ class LanceTable(Table):
         self._conn = connection
         self._namespace = namespace
         self._location = location  # Store location for use in _dataset_path
+        self._namespace_client = namespace_client
         if _async is not None:
             self._table = _async
         else:
@@ -1764,6 +1767,8 @@ class LanceTable(Table):
                     storage_options_provider=storage_options_provider,
                     index_cache_size=index_cache_size,
                     location=location,
+                    namespace_client=namespace_client,
+                    managed_versioning=managed_versioning,
                 )
             )
 
@@ -1806,6 +1811,8 @@ class LanceTable(Table):
         storage_options_provider: Optional["StorageOptionsProvider"] = None,
         index_cache_size: Optional[int] = None,
         location: Optional[str] = None,
+        namespace_client: Optional[Any] = None,
+        managed_versioning: Optional[bool] = None,
     ):
         if namespace is None:
             namespace = []
@@ -1817,6 +1824,8 @@ class LanceTable(Table):
             storage_options_provider=storage_options_provider,
             index_cache_size=index_cache_size,
             location=location,
+            namespace_client=namespace_client,
+            managed_versioning=managed_versioning,
         )
 
         # check the dataset exists
@@ -1848,6 +1857,16 @@ class LanceTable(Table):
                 "Please install with `pip install pylance`."
             )
 
+        if self._namespace_client is not None:
+            table_id = self._namespace + [self.name]
+            return lance.dataset(
+                version=self.version,
+                storage_options=self._conn.storage_options,
+                namespace=self._namespace_client,
+                table_id=table_id,
+                **kwargs,
+            )
+
         return lance.dataset(
             self._dataset_path,
             version=self.version,
@@ -2713,6 +2732,7 @@ class LanceTable(Table):
         data_storage_version: Optional[str] = None,
         enable_v2_manifest_paths: Optional[bool] = None,
         location: Optional[str] = None,
+        namespace_client: Optional[Any] = None,
     ):
         """
         Create a new table.
@@ -2773,6 +2793,7 @@ class LanceTable(Table):
         self._conn = db
         self._namespace = namespace
         self._location = location
+        self._namespace_client = namespace_client
 
         if data_storage_version is not None:
             warnings.warn(
@@ -4215,7 +4236,7 @@ class AsyncTable:
         1  2  [3.0, 4.0]
         2  3  [5.0, 6.0]
         >>> table.delete("x = 2")
-        DeleteResult(version=2)
+        DeleteResult(num_deleted_rows=1, version=2)
         >>> table.to_pandas()
            x      vector
         0  1  [1.0, 2.0]
@@ -4229,7 +4250,7 @@ class AsyncTable:
         >>> to_remove
         '1, 5'
         >>> table.delete(f"x IN ({to_remove})")
-        DeleteResult(version=3)
+        DeleteResult(num_deleted_rows=1, version=3)
         >>> table.to_pandas()
            x      vector
         0  3  [5.0, 6.0]

python/python/lancedb/util.py

@@ -324,6 +324,16 @@ def _(value: list):
     return "[" + ", ".join(map(value_to_sql, value)) + "]"
 
 
+@value_to_sql.register(dict)
+def _(value: dict):
+    # https://datafusion.apache.org/user-guide/sql/scalar_functions.html#named-struct
+    return (
+        "named_struct("
+        + ", ".join(f"'{k}', {value_to_sql(v)}" for k, v in value.items())
+        + ")"
+    )
+
+
 @value_to_sql.register(np.ndarray)
 def _(value: np.ndarray):
     return value_to_sql(value.tolist())

python/python/tests/test_fts.py

@@ -27,6 +27,7 @@ from lancedb.query import (
     PhraseQuery,
     BooleanQuery,
     Occur,
+    LanceFtsQueryBuilder,
 )
 import numpy as np
 import pyarrow as pa
@@ -882,3 +883,109 @@ def test_fts_query_to_json():
         '"must_not":[]}}'
     )
     assert json_str == expected
+
+
+def test_fts_fast_search(table):
+    table.create_fts_index("text", use_tantivy=False)
+
+    # Insert some unindexed data
+    table.add(
+        [
+            {
+                "text": "xyz",
+                "vector": [0 for _ in range(128)],
+                "id": 101,
+                "text2": "xyz",
+                "nested": {"text": "xyz"},
+                "count": 10,
+            }
+        ]
+    )
+
+    # Without fast_search, the query object should not have fast_search set
+    builder = table.search("xyz", query_type="fts").limit(10)
+    query = builder.to_query_object()
+    assert query.fast_search is None
+
+    # With fast_search, the query object should have fast_search=True
+    builder = table.search("xyz", query_type="fts").fast_search().limit(10)
+    query = builder.to_query_object()
+    assert query.fast_search is True
+
+    # fast_search should be chainable with other methods
+    builder = (
+        table.search("xyz", query_type="fts").fast_search().select(["text"]).limit(5)
+    )
+    query = builder.to_query_object()
+    assert query.fast_search is True
+    assert query.limit == 5
+    assert query.columns == ["text"]
+
+    # fast_search should be enabled by keyword argument too
+    query = LanceFtsQueryBuilder(table, "xyz", fast_search=True).to_query_object()
+    assert query.fast_search is True
+
+    # Verify it executes without error and skips unindexed data
+    results = table.search("xyz", query_type="fts").fast_search().limit(5).to_list()
+    assert len(results) == 0
+
+    # Update index and verify it returns results
+    table.optimize()
+    results = table.search("xyz", query_type="fts").fast_search().limit(5).to_list()
+    assert len(results) > 0
+
+
+@pytest.mark.asyncio
+async def test_fts_fast_search_async(async_table):
+    await async_table.create_index("text", config=FTS())
+
+    # Insert some unindexed data
+    await async_table.add(
+        [
+            {
+                "text": "xyz",
+                "vector": [0 for _ in range(128)],
+                "id": 101,
+                "text2": "xyz",
+                "nested": {"text": "xyz"},
+                "count": 10,
+            }
+        ]
+    )
+
+    # Without fast_search, should return results
+    results = await async_table.query().nearest_to_text("xyz").limit(5).to_list()
+    assert len(results) > 0
+
+    # With fast_search, should return no results data unindexed
+    fast_results = (
+        await async_table.query()
+        .nearest_to_text("xyz")
+        .fast_search()
+        .limit(5)
+        .to_list()
+    )
+    assert len(fast_results) == 0
+
+    # Update index and verify it returns results
+    await async_table.optimize()
+
+    fast_results = (
+        await async_table.query()
+        .nearest_to_text("xyz")
+        .fast_search()
+        .limit(5)
+        .to_list()
+    )
+    assert len(fast_results) > 0
+
+    # fast_search should be chainable with other methods
+    results = (
+        await async_table.query()
+        .nearest_to_text("xyz")
+        .fast_search()
+        .select(["text"])
+        .limit(5)
+        .to_list()
+    )
+    assert len(results) > 0

python/python/tests/test_table.py

@@ -326,6 +326,24 @@ def test_add_struct(mem_db: DBConnection):
     table = mem_db.create_table("test2", schema=schema)
     table.add(data)
 
+    struct_type = pa.struct(
+        [
+            ("b", pa.int64()),
+            ("a", pa.int64()),
+        ]
+    )
+    expected = pa.table(
+        {
+            "s_list": [
+                [
+                    pa.scalar({"b": 1, "a": 2}, type=struct_type),
+                    pa.scalar({"b": 4, "a": None}, type=struct_type),
+                ]
+            ],
+        }
+    )
+    assert table.to_arrow() == expected
+
 
 def test_add_subschema(mem_db: DBConnection):
     schema = pa.schema(

python/python/tests/test_util.py

@@ -121,6 +121,32 @@ def test_value_to_sql_string(tmp_path):
         assert table.to_pandas().query("search == @value")["replace"].item() == value
 
 
+def test_value_to_sql_dict():
+    # Simple flat struct
+    assert value_to_sql({"a": 1, "b": "hello"}) == "named_struct('a', 1, 'b', 'hello')"
+
+    # Nested struct
+    assert (
+        value_to_sql({"outer": {"inner": 1}})
+        == "named_struct('outer', named_struct('inner', 1))"
+    )
+
+    # List inside struct
+    assert value_to_sql({"a": [1, 2]}) == "named_struct('a', [1, 2])"
+
+    # Mixed types
+    assert (
+        value_to_sql({"name": "test", "count": 42, "rate": 3.14, "active": True})
+        == "named_struct('name', 'test', 'count', 42, 'rate', 3.14, 'active', TRUE)"
+    )
+
+    # Null value inside struct
+    assert value_to_sql({"a": None}) == "named_struct('a', NULL)"
+
+    # Empty dict
+    assert value_to_sql({}) == "named_struct()"
+
+
 def test_append_vector_columns():
     registry = EmbeddingFunctionRegistry.get_instance()
     registry.register("test")(MockTextEmbeddingFunction)

python/src/arrow.rs

@@ -10,7 +10,7 @@ use arrow::{
 use futures::stream::StreamExt;
 use lancedb::arrow::SendableRecordBatchStream;
 use pyo3::{
-    exceptions::PyStopAsyncIteration, pyclass, pymethods, Bound, Py, PyAny, PyRef, PyResult, Python,
+    Bound, Py, PyAny, PyRef, PyResult, Python, exceptions::PyStopAsyncIteration, pyclass, pymethods,
 };
 use pyo3_async_runtimes::tokio::future_into_py;
 

python/src/connection.rs

@@ -9,15 +9,16 @@ use lancedb::{
     database::{CreateTableMode, Database, ReadConsistency},
 };
 use pyo3::{
+    Bound, FromPyObject, Py, PyAny, PyRef, PyResult, Python,
     exceptions::{PyRuntimeError, PyValueError},
     pyclass, pyfunction, pymethods,
     types::{PyDict, PyDictMethods},
-    Bound, FromPyObject, Py, PyAny, PyRef, PyResult, Python,
 };
 use pyo3_async_runtimes::tokio::future_into_py;
 
 use crate::{
-    error::PythonErrorExt, storage_options::py_object_to_storage_options_provider, table::Table,
+    error::PythonErrorExt, namespace::extract_namespace_arc,
+    storage_options::py_object_to_storage_options_provider, table::Table,
 };
 
 #[pyclass]
@@ -182,7 +183,8 @@ impl Connection {
         })
     }
 
-    #[pyo3(signature = (name, namespace=vec![], storage_options = None, storage_options_provider=None, index_cache_size = None, location=None))]
+    #[allow(clippy::too_many_arguments)]
+    #[pyo3(signature = (name, namespace=vec![], storage_options = None, storage_options_provider=None, index_cache_size = None, location=None, namespace_client=None, managed_versioning=None))]
     pub fn open_table(
         self_: PyRef<'_, Self>,
         name: String,
@@ -191,11 +193,13 @@ impl Connection {
         storage_options_provider: Option<Py<PyAny>>,
         index_cache_size: Option<u32>,
         location: Option<String>,
+        namespace_client: Option<Py<PyAny>>,
+        managed_versioning: Option<bool>,
     ) -> PyResult<Bound<'_, PyAny>> {
         let inner = self_.get_inner()?.clone();
 
         let mut builder = inner.open_table(name);
-        builder = builder.namespace(namespace);
+        builder = builder.namespace(namespace.clone());
         if let Some(storage_options) = storage_options {
             builder = builder.storage_options(storage_options);
         }
@@ -209,6 +213,20 @@ impl Connection {
         if let Some(location) = location {
             builder = builder.location(location);
         }
+        // Extract namespace client from Python object if provided
+        let ns_client = if let Some(ns_obj) = namespace_client {
+            let py = self_.py();
+            Some(extract_namespace_arc(py, ns_obj)?)
+        } else {
+            None
+        };
+        if let Some(ns_client) = ns_client {
+            builder = builder.namespace_client(ns_client);
+        }
+        // Pass managed_versioning if provided to avoid redundant describe_table call
+        if let Some(enabled) = managed_versioning {
+            builder = builder.managed_versioning(enabled);
+        }
 
         future_into_py(self_.py(), async move {
             let table = builder.execute().await.infer_error()?;

python/src/error.rs

@@ -2,10 +2,10 @@
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
 use pyo3::{
+    PyErr, PyResult, Python,
     exceptions::{PyIOError, PyNotImplementedError, PyOSError, PyRuntimeError, PyValueError},
     intern,
     types::{PyAnyMethods, PyNone},
-    PyErr, PyResult, Python,
 };
 
 use lancedb::error::Error as LanceError;

python/src/index.rs

@@ -3,17 +3,17 @@
 
 use lancedb::index::vector::{IvfFlatIndexBuilder, IvfRqIndexBuilder, IvfSqIndexBuilder};
 use lancedb::index::{
+    Index as LanceDbIndex,
     scalar::{BTreeIndexBuilder, FtsIndexBuilder},
     vector::{IvfHnswPqIndexBuilder, IvfHnswSqIndexBuilder, IvfPqIndexBuilder},
-    Index as LanceDbIndex,
 };
-use pyo3::types::PyStringMethods;
 use pyo3::IntoPyObject;
+use pyo3::types::PyStringMethods;
 use pyo3::{
+    Bound, FromPyObject, PyAny, PyResult, Python,
     exceptions::{PyKeyError, PyValueError},
     intern, pyclass, pymethods,
     types::PyAnyMethods,
-    Bound, FromPyObject, PyAny, PyResult, Python,
 };
 
 use crate::util::parse_distance_type;
@@ -41,7 +41,12 @@ pub fn extract_index_params(source: &Option<Bound<'_, PyAny>>) -> PyResult<Lance
                 let inner_opts = FtsIndexBuilder::default()
                     .base_tokenizer(params.base_tokenizer)
                     .language(&params.language)
-                    .map_err(|_| PyValueError::new_err(format!("LanceDB does not support the requested language: '{}'", params.language)))?
+                    .map_err(|_| {
+                        PyValueError::new_err(format!(
+                            "LanceDB does not support the requested language: '{}'",
+                            params.language
+                        ))
+                    })?
                     .with_position(params.with_position)
                     .lower_case(params.lower_case)
                     .max_token_length(params.max_token_length)
@@ -52,7 +57,7 @@ pub fn extract_index_params(source: &Option<Bound<'_, PyAny>>) -> PyResult<Lance
                     .ngram_max_length(params.ngram_max_length)
                     .ngram_prefix_only(params.prefix_only);
                 Ok(LanceDbIndex::FTS(inner_opts))
-            },
+            }
             "IvfFlat" => {
                 let params = source.extract::<IvfFlatParams>()?;
                 let distance_type = parse_distance_type(params.distance_type)?;
@@ -64,10 +69,11 @@ pub fn extract_index_params(source: &Option<Bound<'_, PyAny>>) -> PyResult<Lance
                     ivf_flat_builder = ivf_flat_builder.num_partitions(num_partitions);
                 }
                 if let Some(target_partition_size) = params.target_partition_size {
-                    ivf_flat_builder = ivf_flat_builder.target_partition_size(target_partition_size);
+                    ivf_flat_builder =
+                        ivf_flat_builder.target_partition_size(target_partition_size);
                 }
                 Ok(LanceDbIndex::IvfFlat(ivf_flat_builder))
-            },
+            }
             "IvfPq" => {
                 let params = source.extract::<IvfPqParams>()?;
                 let distance_type = parse_distance_type(params.distance_type)?;
@@ -86,7 +92,7 @@ pub fn extract_index_params(source: &Option<Bound<'_, PyAny>>) -> PyResult<Lance
                     ivf_pq_builder = ivf_pq_builder.num_sub_vectors(num_sub_vectors);
                 }
                 Ok(LanceDbIndex::IvfPq(ivf_pq_builder))
-            },
+            }
             "IvfSq" => {
                 let params = source.extract::<IvfSqParams>()?;
                 let distance_type = parse_distance_type(params.distance_type)?;
@@ -101,7 +107,7 @@ pub fn extract_index_params(source: &Option<Bound<'_, PyAny>>) -> PyResult<Lance
                     ivf_sq_builder = ivf_sq_builder.target_partition_size(target_partition_size);
                 }
                 Ok(LanceDbIndex::IvfSq(ivf_sq_builder))
-            },
+            }
             "IvfRq" => {
                 let params = source.extract::<IvfRqParams>()?;
                 let distance_type = parse_distance_type(params.distance_type)?;
@@ -117,7 +123,7 @@ pub fn extract_index_params(source: &Option<Bound<'_, PyAny>>) -> PyResult<Lance
                     ivf_rq_builder = ivf_rq_builder.target_partition_size(target_partition_size);
                 }
                 Ok(LanceDbIndex::IvfRq(ivf_rq_builder))
-            },
+            }
             "HnswPq" => {
                 let params = source.extract::<IvfHnswPqParams>()?;
                 let distance_type = parse_distance_type(params.distance_type)?;
@@ -138,7 +144,7 @@ pub fn extract_index_params(source: &Option<Bound<'_, PyAny>>) -> PyResult<Lance
                     hnsw_pq_builder = hnsw_pq_builder.num_sub_vectors(num_sub_vectors);
                 }
                 Ok(LanceDbIndex::IvfHnswPq(hnsw_pq_builder))
-            },
+            }
             "HnswSq" => {
                 let params = source.extract::<IvfHnswSqParams>()?;
                 let distance_type = parse_distance_type(params.distance_type)?;
@@ -155,7 +161,7 @@ pub fn extract_index_params(source: &Option<Bound<'_, PyAny>>) -> PyResult<Lance
                     hnsw_sq_builder = hnsw_sq_builder.target_partition_size(target_partition_size);
                 }
                 Ok(LanceDbIndex::IvfHnswSq(hnsw_sq_builder))
-            },
+            }
             not_supported => Err(PyValueError::new_err(format!(
                 "Invalid index type '{}'.  Must be one of BTree, Bitmap, LabelList, FTS, IvfPq, IvfSq, IvfHnswPq, or IvfHnswSq",
                 not_supported

python/src/lib.rs

@@ -2,14 +2,14 @@
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
 use arrow::RecordBatchStream;
-use connection::{connect, Connection};
+use connection::{Connection, connect};
 use env_logger::Env;
 use index::IndexConfig;
 use permutation::{PyAsyncPermutationBuilder, PyPermutationReader};
 use pyo3::{
-    pymodule,
+    Bound, PyResult, Python, pymodule,
     types::{PyModule, PyModuleMethods},
-    wrap_pyfunction, Bound, PyResult, Python,
+    wrap_pyfunction,
 };
 use query::{FTSQuery, HybridQuery, Query, VectorQuery};
 use session::Session;
@@ -23,6 +23,7 @@ pub mod connection;
 pub mod error;
 pub mod header;
 pub mod index;
+pub mod namespace;
 pub mod permutation;
 pub mod query;
 pub mod session;

python/src/namespace.rs

@@ -0,0 +1,746 @@
+// SPDX-License-Identifier: Apache-2.0
+// SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+//! Namespace utilities for Python bindings
+
+use std::collections::HashMap;
+use std::sync::Arc;
+
+use async_trait::async_trait;
+use bytes::Bytes;
+use lance_namespace::LanceNamespace as LanceNamespaceTrait;
+use lance_namespace::models::*;
+use pyo3::prelude::*;
+use pyo3::types::PyDict;
+
+/// Wrapper that allows any Python object implementing LanceNamespace protocol
+/// to be used as a Rust LanceNamespace.
+///
+/// This is similar to PyLanceNamespace in lance's Python bindings - it wraps a Python
+/// object and calls back into Python when namespace methods are invoked.
+pub struct PyLanceNamespace {
+    py_namespace: Arc<Py<PyAny>>,
+    namespace_id: String,
+}
+
+impl PyLanceNamespace {
+    /// Create a new PyLanceNamespace wrapper around a Python namespace object.
+    pub fn new(_py: Python<'_>, py_namespace: &Bound<'_, PyAny>) -> PyResult<Self> {
+        let namespace_id = py_namespace
+            .call_method0("namespace_id")?
+            .extract::<String>()?;
+
+        Ok(Self {
+            py_namespace: Arc::new(py_namespace.clone().unbind()),
+            namespace_id,
+        })
+    }
+
+    /// Create an Arc<dyn LanceNamespace> from a Python namespace object.
+    pub fn create_arc(
+        py: Python<'_>,
+        py_namespace: &Bound<'_, PyAny>,
+    ) -> PyResult<Arc<dyn LanceNamespaceTrait>> {
+        let wrapper = Self::new(py, py_namespace)?;
+        Ok(Arc::new(wrapper))
+    }
+}
+
+impl std::fmt::Debug for PyLanceNamespace {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "PyLanceNamespace {{ id: {} }}", self.namespace_id)
+    }
+}
+
+/// Get or create the DictWithModelDump class in Python.
+/// This class acts like a dict but also has model_dump() method.
+/// This allows it to work with both:
+/// - depythonize (which expects a dict/Mapping)
+/// - Python code that calls .model_dump() (like DirectoryNamespace wrapper)
+fn get_dict_with_model_dump_class(py: Python<'_>) -> PyResult<Bound<'_, PyAny>> {
+    // Use a module-level cache via __builtins__
+    let builtins = py.import("builtins")?;
+    if builtins.hasattr("_DictWithModelDump")? {
+        return builtins.getattr("_DictWithModelDump");
+    }
+
+    // Create the class using exec
+    let locals = PyDict::new(py);
+    py.run(
+        c"class DictWithModelDump(dict):
+    def model_dump(self):
+        return dict(self)",
+        None,
+        Some(&locals),
+    )?;
+    let class = locals.get_item("DictWithModelDump")?.ok_or_else(|| {
+        pyo3::exceptions::PyRuntimeError::new_err("Failed to create DictWithModelDump class")
+    })?;
+
+    // Cache it
+    builtins.setattr("_DictWithModelDump", &class)?;
+    Ok(class)
+}
+
+/// Helper to call a Python namespace method with JSON serialization.
+/// For methods that take a request and return a response.
+/// Uses DictWithModelDump to pass a dict that also has model_dump() method,
+/// making it compatible with both depythonize and Python wrappers.
+async fn call_py_method<Req, Resp>(
+    py_namespace: Arc<Py<PyAny>>,
+    method_name: &'static str,
+    request: Req,
+) -> lance_core::Result<Resp>
+where
+    Req: serde::Serialize + Send + 'static,
+    Resp: serde::de::DeserializeOwned + Send + 'static,
+{
+    let request_json = serde_json::to_string(&request).map_err(|e| {
+        lance_core::Error::io(
+            format!("Failed to serialize request for {}: {}", method_name, e),
+            Default::default(),
+        )
+    })?;
+
+    let response_json = tokio::task::spawn_blocking(move || {
+        Python::attach(|py| {
+            let json_module = py.import("json")?;
+            let request_dict = json_module.call_method1("loads", (&request_json,))?;
+
+            // Wrap dict in DictWithModelDump so it works with both depythonize and .model_dump()
+            let dict_class = get_dict_with_model_dump_class(py)?;
+            let request_arg = dict_class.call1((request_dict,))?;
+
+            // Call the Python method
+            let result = py_namespace.call_method1(py, method_name, (request_arg,))?;
+
+            // Convert response to dict, then to JSON
+            // Pydantic models have model_dump() method
+            let result_dict = if result.bind(py).hasattr("model_dump")? {
+                result.call_method0(py, "model_dump")?
+            } else {
+                result
+            };
+            let response_json: String = json_module
+                .call_method1("dumps", (result_dict,))?
+                .extract()?;
+            Ok::<_, PyErr>(response_json)
+        })
+    })
+    .await
+    .map_err(|e| {
+        lance_core::Error::io(
+            format!("Task join error for {}: {}", method_name, e),
+            Default::default(),
+        )
+    })?
+    .map_err(|e: PyErr| {
+        lance_core::Error::io(
+            format!("Python error in {}: {}", method_name, e),
+            Default::default(),
+        )
+    })?;
+
+    serde_json::from_str(&response_json).map_err(|e| {
+        lance_core::Error::io(
+            format!("Failed to deserialize response from {}: {}", method_name, e),
+            Default::default(),
+        )
+    })
+}
+
+/// Helper for methods that return () on success
+async fn call_py_method_unit<Req>(
+    py_namespace: Arc<Py<PyAny>>,
+    method_name: &'static str,
+    request: Req,
+) -> lance_core::Result<()>
+where
+    Req: serde::Serialize + Send + 'static,
+{
+    let request_json = serde_json::to_string(&request).map_err(|e| {
+        lance_core::Error::io(
+            format!("Failed to serialize request for {}: {}", method_name, e),
+            Default::default(),
+        )
+    })?;
+
+    tokio::task::spawn_blocking(move || {
+        Python::attach(|py| {
+            let json_module = py.import("json")?;
+            let request_dict = json_module.call_method1("loads", (&request_json,))?;
+
+            // Wrap dict in DictWithModelDump
+            let dict_class = get_dict_with_model_dump_class(py)?;
+            let request_arg = dict_class.call1((request_dict,))?;
+
+            // Call the Python method
+            py_namespace.call_method1(py, method_name, (request_arg,))?;
+            Ok::<_, PyErr>(())
+        })
+    })
+    .await
+    .map_err(|e| {
+        lance_core::Error::io(
+            format!("Task join error for {}: {}", method_name, e),
+            Default::default(),
+        )
+    })?
+    .map_err(|e: PyErr| {
+        lance_core::Error::io(
+            format!("Python error in {}: {}", method_name, e),
+            Default::default(),
+        )
+    })
+}
+
+/// Helper for methods that return a primitive type
+async fn call_py_method_primitive<Req, Resp>(
+    py_namespace: Arc<Py<PyAny>>,
+    method_name: &'static str,
+    request: Req,
+) -> lance_core::Result<Resp>
+where
+    Req: serde::Serialize + Send + 'static,
+    Resp: for<'py> pyo3::FromPyObject<'py> + Send + 'static,
+{
+    let request_json = serde_json::to_string(&request).map_err(|e| {
+        lance_core::Error::io(
+            format!("Failed to serialize request for {}: {}", method_name, e),
+            Default::default(),
+        )
+    })?;
+
+    tokio::task::spawn_blocking(move || {
+        Python::attach(|py| {
+            let json_module = py.import("json")?;
+            let request_dict = json_module.call_method1("loads", (&request_json,))?;
+
+            // Wrap dict in DictWithModelDump
+            let dict_class = get_dict_with_model_dump_class(py)?;
+            let request_arg = dict_class.call1((request_dict,))?;
+
+            // Call the Python method
+            let result = py_namespace.call_method1(py, method_name, (request_arg,))?;
+            let value: Resp = result.extract(py)?;
+            Ok::<_, PyErr>(value)
+        })
+    })
+    .await
+    .map_err(|e| {
+        lance_core::Error::io(
+            format!("Task join error for {}: {}", method_name, e),
+            Default::default(),
+        )
+    })?
+    .map_err(|e: PyErr| {
+        lance_core::Error::io(
+            format!("Python error in {}: {}", method_name, e),
+            Default::default(),
+        )
+    })
+}
+
+/// Helper for methods that return Bytes
+async fn call_py_method_bytes<Req>(
+    py_namespace: Arc<Py<PyAny>>,
+    method_name: &'static str,
+    request: Req,
+) -> lance_core::Result<Bytes>
+where
+    Req: serde::Serialize + Send + 'static,
+{
+    let request_json = serde_json::to_string(&request).map_err(|e| {
+        lance_core::Error::io(
+            format!("Failed to serialize request for {}: {}", method_name, e),
+            Default::default(),
+        )
+    })?;
+
+    tokio::task::spawn_blocking(move || {
+        Python::attach(|py| {
+            let json_module = py.import("json")?;
+            let request_dict = json_module.call_method1("loads", (&request_json,))?;
+
+            // Wrap dict in DictWithModelDump
+            let dict_class = get_dict_with_model_dump_class(py)?;
+            let request_arg = dict_class.call1((request_dict,))?;
+
+            // Call the Python method
+            let result = py_namespace.call_method1(py, method_name, (request_arg,))?;
+            let bytes_data: Vec<u8> = result.extract(py)?;
+            Ok::<_, PyErr>(Bytes::from(bytes_data))
+        })
+    })
+    .await
+    .map_err(|e| {
+        lance_core::Error::io(
+            format!("Task join error for {}: {}", method_name, e),
+            Default::default(),
+        )
+    })?
+    .map_err(|e: PyErr| {
+        lance_core::Error::io(
+            format!("Python error in {}: {}", method_name, e),
+            Default::default(),
+        )
+    })
+}
+
+/// Helper for methods that take request + data and return a response
+async fn call_py_method_with_data<Req, Resp>(
+    py_namespace: Arc<Py<PyAny>>,
+    method_name: &'static str,
+    request: Req,
+    data: Bytes,
+) -> lance_core::Result<Resp>
+where
+    Req: serde::Serialize + Send + 'static,
+    Resp: serde::de::DeserializeOwned + Send + 'static,
+{
+    let request_json = serde_json::to_string(&request).map_err(|e| {
+        lance_core::Error::io(
+            format!("Failed to serialize request for {}: {}", method_name, e),
+            Default::default(),
+        )
+    })?;
+
+    let response_json = tokio::task::spawn_blocking(move || {
+        Python::attach(|py| {
+            let json_module = py.import("json")?;
+            let request_dict = json_module.call_method1("loads", (&request_json,))?;
+
+            // Wrap dict in DictWithModelDump
+            let dict_class = get_dict_with_model_dump_class(py)?;
+            let request_arg = dict_class.call1((request_dict,))?;
+
+            // Pass request and bytes to Python method
+            let py_bytes = pyo3::types::PyBytes::new(py, &data);
+            let result = py_namespace.call_method1(py, method_name, (request_arg, py_bytes))?;
+
+            // Convert response dict to JSON
+            let response_json: String = json_module.call_method1("dumps", (result,))?.extract()?;
+            Ok::<_, PyErr>(response_json)
+        })
+    })
+    .await
+    .map_err(|e| {
+        lance_core::Error::io(
+            format!("Task join error for {}: {}", method_name, e),
+            Default::default(),
+        )
+    })?
+    .map_err(|e: PyErr| {
+        lance_core::Error::io(
+            format!("Python error in {}: {}", method_name, e),
+            Default::default(),
+        )
+    })?;
+
+    serde_json::from_str(&response_json).map_err(|e| {
+        lance_core::Error::io(
+            format!("Failed to deserialize response from {}: {}", method_name, e),
+            Default::default(),
+        )
+    })
+}
+
+#[async_trait]
+impl LanceNamespaceTrait for PyLanceNamespace {
+    fn namespace_id(&self) -> String {
+        self.namespace_id.clone()
+    }
+
+    async fn list_namespaces(
+        &self,
+        request: ListNamespacesRequest,
+    ) -> lance_core::Result<ListNamespacesResponse> {
+        call_py_method(self.py_namespace.clone(), "list_namespaces", request).await
+    }
+
+    async fn describe_namespace(
+        &self,
+        request: DescribeNamespaceRequest,
+    ) -> lance_core::Result<DescribeNamespaceResponse> {
+        call_py_method(self.py_namespace.clone(), "describe_namespace", request).await
+    }
+
+    async fn create_namespace(
+        &self,
+        request: CreateNamespaceRequest,
+    ) -> lance_core::Result<CreateNamespaceResponse> {
+        call_py_method(self.py_namespace.clone(), "create_namespace", request).await
+    }
+
+    async fn drop_namespace(
+        &self,
+        request: DropNamespaceRequest,
+    ) -> lance_core::Result<DropNamespaceResponse> {
+        call_py_method(self.py_namespace.clone(), "drop_namespace", request).await
+    }
+
+    async fn namespace_exists(&self, request: NamespaceExistsRequest) -> lance_core::Result<()> {
+        call_py_method_unit(self.py_namespace.clone(), "namespace_exists", request).await
+    }
+
+    async fn list_tables(
+        &self,
+        request: ListTablesRequest,
+    ) -> lance_core::Result<ListTablesResponse> {
+        call_py_method(self.py_namespace.clone(), "list_tables", request).await
+    }
+
+    async fn describe_table(
+        &self,
+        request: DescribeTableRequest,
+    ) -> lance_core::Result<DescribeTableResponse> {
+        call_py_method(self.py_namespace.clone(), "describe_table", request).await
+    }
+
+    async fn register_table(
+        &self,
+        request: RegisterTableRequest,
+    ) -> lance_core::Result<RegisterTableResponse> {
+        call_py_method(self.py_namespace.clone(), "register_table", request).await
+    }
+
+    async fn table_exists(&self, request: TableExistsRequest) -> lance_core::Result<()> {
+        call_py_method_unit(self.py_namespace.clone(), "table_exists", request).await
+    }
+
+    async fn drop_table(&self, request: DropTableRequest) -> lance_core::Result<DropTableResponse> {
+        call_py_method(self.py_namespace.clone(), "drop_table", request).await
+    }
+
+    async fn deregister_table(
+        &self,
+        request: DeregisterTableRequest,
+    ) -> lance_core::Result<DeregisterTableResponse> {
+        call_py_method(self.py_namespace.clone(), "deregister_table", request).await
+    }
+
+    async fn count_table_rows(&self, request: CountTableRowsRequest) -> lance_core::Result<i64> {
+        call_py_method_primitive(self.py_namespace.clone(), "count_table_rows", request).await
+    }
+
+    async fn create_table(
+        &self,
+        request: CreateTableRequest,
+        request_data: Bytes,
+    ) -> lance_core::Result<CreateTableResponse> {
+        call_py_method_with_data(
+            self.py_namespace.clone(),
+            "create_table",
+            request,
+            request_data,
+        )
+        .await
+    }
+
+    async fn declare_table(
+        &self,
+        request: DeclareTableRequest,
+    ) -> lance_core::Result<DeclareTableResponse> {
+        call_py_method(self.py_namespace.clone(), "declare_table", request).await
+    }
+
+    async fn insert_into_table(
+        &self,
+        request: InsertIntoTableRequest,
+        request_data: Bytes,
+    ) -> lance_core::Result<InsertIntoTableResponse> {
+        call_py_method_with_data(
+            self.py_namespace.clone(),
+            "insert_into_table",
+            request,
+            request_data,
+        )
+        .await
+    }
+
+    async fn merge_insert_into_table(
+        &self,
+        request: MergeInsertIntoTableRequest,
+        request_data: Bytes,
+    ) -> lance_core::Result<MergeInsertIntoTableResponse> {
+        call_py_method_with_data(
+            self.py_namespace.clone(),
+            "merge_insert_into_table",
+            request,
+            request_data,
+        )
+        .await
+    }
+
+    async fn update_table(
+        &self,
+        request: UpdateTableRequest,
+    ) -> lance_core::Result<UpdateTableResponse> {
+        call_py_method(self.py_namespace.clone(), "update_table", request).await
+    }
+
+    async fn delete_from_table(
+        &self,
+        request: DeleteFromTableRequest,
+    ) -> lance_core::Result<DeleteFromTableResponse> {
+        call_py_method(self.py_namespace.clone(), "delete_from_table", request).await
+    }
+
+    async fn query_table(&self, request: QueryTableRequest) -> lance_core::Result<Bytes> {
+        call_py_method_bytes(self.py_namespace.clone(), "query_table", request).await
+    }
+
+    async fn create_table_index(
+        &self,
+        request: CreateTableIndexRequest,
+    ) -> lance_core::Result<CreateTableIndexResponse> {
+        call_py_method(self.py_namespace.clone(), "create_table_index", request).await
+    }
+
+    async fn list_table_indices(
+        &self,
+        request: ListTableIndicesRequest,
+    ) -> lance_core::Result<ListTableIndicesResponse> {
+        call_py_method(self.py_namespace.clone(), "list_table_indices", request).await
+    }
+
+    async fn describe_table_index_stats(
+        &self,
+        request: DescribeTableIndexStatsRequest,
+    ) -> lance_core::Result<DescribeTableIndexStatsResponse> {
+        call_py_method(
+            self.py_namespace.clone(),
+            "describe_table_index_stats",
+            request,
+        )
+        .await
+    }
+
+    async fn describe_transaction(
+        &self,
+        request: DescribeTransactionRequest,
+    ) -> lance_core::Result<DescribeTransactionResponse> {
+        call_py_method(self.py_namespace.clone(), "describe_transaction", request).await
+    }
+
+    async fn alter_transaction(
+        &self,
+        request: AlterTransactionRequest,
+    ) -> lance_core::Result<AlterTransactionResponse> {
+        call_py_method(self.py_namespace.clone(), "alter_transaction", request).await
+    }
+
+    async fn create_table_scalar_index(
+        &self,
+        request: CreateTableIndexRequest,
+    ) -> lance_core::Result<CreateTableScalarIndexResponse> {
+        call_py_method(
+            self.py_namespace.clone(),
+            "create_table_scalar_index",
+            request,
+        )
+        .await
+    }
+
+    async fn drop_table_index(
+        &self,
+        request: DropTableIndexRequest,
+    ) -> lance_core::Result<DropTableIndexResponse> {
+        call_py_method(self.py_namespace.clone(), "drop_table_index", request).await
+    }
+
+    async fn list_all_tables(
+        &self,
+        request: ListTablesRequest,
+    ) -> lance_core::Result<ListTablesResponse> {
+        call_py_method(self.py_namespace.clone(), "list_all_tables", request).await
+    }
+
+    async fn restore_table(
+        &self,
+        request: RestoreTableRequest,
+    ) -> lance_core::Result<RestoreTableResponse> {
+        call_py_method(self.py_namespace.clone(), "restore_table", request).await
+    }
+
+    async fn rename_table(
+        &self,
+        request: RenameTableRequest,
+    ) -> lance_core::Result<RenameTableResponse> {
+        call_py_method(self.py_namespace.clone(), "rename_table", request).await
+    }
+
+    async fn list_table_versions(
+        &self,
+        request: ListTableVersionsRequest,
+    ) -> lance_core::Result<ListTableVersionsResponse> {
+        call_py_method(self.py_namespace.clone(), "list_table_versions", request).await
+    }
+
+    async fn create_table_version(
+        &self,
+        request: CreateTableVersionRequest,
+    ) -> lance_core::Result<CreateTableVersionResponse> {
+        call_py_method(self.py_namespace.clone(), "create_table_version", request).await
+    }
+
+    async fn describe_table_version(
+        &self,
+        request: DescribeTableVersionRequest,
+    ) -> lance_core::Result<DescribeTableVersionResponse> {
+        call_py_method(self.py_namespace.clone(), "describe_table_version", request).await
+    }
+
+    async fn batch_delete_table_versions(
+        &self,
+        request: BatchDeleteTableVersionsRequest,
+    ) -> lance_core::Result<BatchDeleteTableVersionsResponse> {
+        call_py_method(
+            self.py_namespace.clone(),
+            "batch_delete_table_versions",
+            request,
+        )
+        .await
+    }
+
+    async fn update_table_schema_metadata(
+        &self,
+        request: UpdateTableSchemaMetadataRequest,
+    ) -> lance_core::Result<UpdateTableSchemaMetadataResponse> {
+        call_py_method(
+            self.py_namespace.clone(),
+            "update_table_schema_metadata",
+            request,
+        )
+        .await
+    }
+
+    async fn get_table_stats(
+        &self,
+        request: GetTableStatsRequest,
+    ) -> lance_core::Result<GetTableStatsResponse> {
+        call_py_method(self.py_namespace.clone(), "get_table_stats", request).await
+    }
+
+    async fn explain_table_query_plan(
+        &self,
+        request: ExplainTableQueryPlanRequest,
+    ) -> lance_core::Result<String> {
+        call_py_method_primitive(
+            self.py_namespace.clone(),
+            "explain_table_query_plan",
+            request,
+        )
+        .await
+    }
+
+    async fn analyze_table_query_plan(
+        &self,
+        request: AnalyzeTableQueryPlanRequest,
+    ) -> lance_core::Result<String> {
+        call_py_method_primitive(
+            self.py_namespace.clone(),
+            "analyze_table_query_plan",
+            request,
+        )
+        .await
+    }
+
+    async fn alter_table_add_columns(
+        &self,
+        request: AlterTableAddColumnsRequest,
+    ) -> lance_core::Result<AlterTableAddColumnsResponse> {
+        call_py_method(
+            self.py_namespace.clone(),
+            "alter_table_add_columns",
+            request,
+        )
+        .await
+    }
+
+    async fn alter_table_alter_columns(
+        &self,
+        request: AlterTableAlterColumnsRequest,
+    ) -> lance_core::Result<AlterTableAlterColumnsResponse> {
+        call_py_method(
+            self.py_namespace.clone(),
+            "alter_table_alter_columns",
+            request,
+        )
+        .await
+    }
+
+    async fn alter_table_drop_columns(
+        &self,
+        request: AlterTableDropColumnsRequest,
+    ) -> lance_core::Result<AlterTableDropColumnsResponse> {
+        call_py_method(
+            self.py_namespace.clone(),
+            "alter_table_drop_columns",
+            request,
+        )
+        .await
+    }
+
+    async fn list_table_tags(
+        &self,
+        request: ListTableTagsRequest,
+    ) -> lance_core::Result<ListTableTagsResponse> {
+        call_py_method(self.py_namespace.clone(), "list_table_tags", request).await
+    }
+
+    async fn create_table_tag(
+        &self,
+        request: CreateTableTagRequest,
+    ) -> lance_core::Result<CreateTableTagResponse> {
+        call_py_method(self.py_namespace.clone(), "create_table_tag", request).await
+    }
+
+    async fn delete_table_tag(
+        &self,
+        request: DeleteTableTagRequest,
+    ) -> lance_core::Result<DeleteTableTagResponse> {
+        call_py_method(self.py_namespace.clone(), "delete_table_tag", request).await
+    }
+
+    async fn update_table_tag(
+        &self,
+        request: UpdateTableTagRequest,
+    ) -> lance_core::Result<UpdateTableTagResponse> {
+        call_py_method(self.py_namespace.clone(), "update_table_tag", request).await
+    }
+
+    async fn get_table_tag_version(
+        &self,
+        request: GetTableTagVersionRequest,
+    ) -> lance_core::Result<GetTableTagVersionResponse> {
+        call_py_method(self.py_namespace.clone(), "get_table_tag_version", request).await
+    }
+}
+
+/// Convert Python dict to HashMap<String, String>
+#[allow(dead_code)]
+fn dict_to_hashmap(dict: &Bound<'_, PyDict>) -> PyResult<HashMap<String, String>> {
+    let mut map = HashMap::new();
+    for (key, value) in dict.iter() {
+        let key_str: String = key.extract()?;
+        let value_str: String = value.extract()?;
+        map.insert(key_str, value_str);
+    }
+    Ok(map)
+}
+
+/// Extract an Arc<dyn LanceNamespace> from a Python namespace object.
+///
+/// This function wraps any Python namespace object with PyLanceNamespace.
+/// The PyLanceNamespace wrapper uses DictWithModelDump to pass requests,
+/// which works with both:
+/// - Native namespaces (DirectoryNamespace, RestNamespace) that use depythonize (expects dict)
+/// - Custom Python implementations that call .model_dump() on the request
+pub fn extract_namespace_arc(
+    py: Python<'_>,
+    ns: Py<PyAny>,
+) -> PyResult<Arc<dyn LanceNamespaceTrait>> {
+    let ns_ref = ns.bind(py);
+    PyLanceNamespace::create_arc(py, ns_ref)
+}

python/src/permutation.rs

@@ -16,10 +16,10 @@ use lancedb::{
     query::Select,
 };
 use pyo3::{
+    Bound, PyAny, PyRef, PyRefMut, PyResult, Python,
     exceptions::PyRuntimeError,
     pyclass, pymethods,
     types::{PyAnyMethods, PyDict, PyDictMethods, PyType},
-    Bound, PyAny, PyRef, PyRefMut, PyResult, Python,
 };
 use pyo3_async_runtimes::tokio::future_into_py;
 

python/src/query.rs

@@ -4,9 +4,9 @@
 use std::sync::Arc;
 use std::time::Duration;
 
-use arrow::array::make_array;
 use arrow::array::Array;
 use arrow::array::ArrayData;
+use arrow::array::make_array;
 use arrow::pyarrow::FromPyArrow;
 use arrow::pyarrow::IntoPyArrow;
 use arrow::pyarrow::ToPyArrow;
@@ -22,23 +22,23 @@ use lancedb::query::{
     VectorQuery as LanceDbVectorQuery,
 };
 use lancedb::table::AnyQuery;
-use pyo3::prelude::{PyAnyMethods, PyDictMethods};
-use pyo3::pyfunction;
-use pyo3::pymethods;
-use pyo3::types::PyList;
-use pyo3::types::{PyDict, PyString};
 use pyo3::Bound;
 use pyo3::IntoPyObject;
 use pyo3::PyAny;
 use pyo3::PyRef;
 use pyo3::PyResult;
 use pyo3::Python;
-use pyo3::{exceptions::PyRuntimeError, FromPyObject};
+use pyo3::prelude::{PyAnyMethods, PyDictMethods};
+use pyo3::pyfunction;
+use pyo3::pymethods;
+use pyo3::types::PyList;
+use pyo3::types::{PyDict, PyString};
+use pyo3::{FromPyObject, exceptions::PyRuntimeError};
+use pyo3::{PyErr, pyclass};
 use pyo3::{
     exceptions::{PyNotImplementedError, PyValueError},
     intern,
 };
-use pyo3::{pyclass, PyErr};
 use pyo3_async_runtimes::tokio::future_into_py;
 
 use crate::util::parse_distance_type;

python/src/session.rs

@@ -4,7 +4,7 @@
 use std::sync::Arc;
 
 use lancedb::{ObjectStoreRegistry, Session as LanceSession};
-use pyo3::{pyclass, pymethods, PyResult};
+use pyo3::{PyResult, pyclass, pymethods};
 
 /// A session for managing caches and object stores across LanceDB operations.
 ///

python/src/table.rs

@@ -5,7 +5,7 @@ use std::{collections::HashMap, sync::Arc};
 use crate::{
     connection::Connection,
     error::PythonErrorExt,
-    index::{extract_index_params, IndexConfig},
+    index::{IndexConfig, extract_index_params},
     query::{Query, TakeQuery},
     table::scannable::PyScannable,
 };
@@ -19,10 +19,10 @@ use lancedb::table::{
     Table as LanceDbTable,
 };
 use pyo3::{
+    Bound, FromPyObject, PyAny, PyRef, PyResult, Python,
     exceptions::{PyKeyError, PyRuntimeError, PyValueError},
     pyclass, pymethods,
     types::{IntoPyDict, PyAnyMethods, PyDict, PyDictMethods},
-    Bound, FromPyObject, PyAny, PyRef, PyResult, Python,
 };
 use pyo3_async_runtimes::tokio::future_into_py;
 
@@ -112,19 +112,24 @@ impl From<lancedb::table::AddResult> for AddResult {
 #[pyclass(get_all)]
 #[derive(Clone, Debug)]
 pub struct DeleteResult {
+    pub num_deleted_rows: u64,
     pub version: u64,
 }
 
 #[pymethods]
 impl DeleteResult {
     pub fn __repr__(&self) -> String {
-        format!("DeleteResult(version={})", self.version)
+        format!(
+            "DeleteResult(num_deleted_rows={}, version={})",
+            self.num_deleted_rows, self.version
+        )
     }
 }
 
 impl From<lancedb::table::DeleteResult> for DeleteResult {
     fn from(result: lancedb::table::DeleteResult) -> Self {
         Self {
+            num_deleted_rows: result.num_deleted_rows,
             version: result.version,
         }
     }
@@ -537,7 +542,7 @@ impl Table {
         let inner = self_.inner_ref()?.clone();
         future_into_py(self_.py(), async move {
             let versions = inner.list_versions().await.infer_error()?;
-            let versions_as_dict = Python::attach(|py| {
+            Python::attach(|py| {
                 versions
                     .iter()
                     .map(|v| {
@@ -554,9 +559,7 @@ impl Table {
                         Ok(dict.unbind())
                     })
                     .collect::<PyResult<Vec<_>>>()
-            });
-
-            versions_as_dict
+            })
         })
     }
 

python/src/table/scannable.rs

@@ -10,11 +10,11 @@ use arrow::{
 };
 use futures::StreamExt;
 use lancedb::{
+    Error,
     arrow::{SendableRecordBatchStream, SimpleRecordBatchStream},
     data::scannable::Scannable,
-    Error,
 };
-use pyo3::{types::PyAnyMethods, FromPyObject, Py, PyAny, Python};
+use pyo3::{FromPyObject, Py, PyAny, Python, types::PyAnyMethods};
 
 /// Adapter that implements Scannable for a Python reader factory callable.
 ///
@@ -99,15 +99,15 @@ impl Scannable for PyScannable {
                         // Channel closed. Check if the task panicked — a panic
                         // drops the sender without sending an error, so without
                         // this check we'd silently return a truncated stream.
-                        if let Some(handle) = join_handle {
-                            if let Err(join_err) = handle.await {
-                                return Some((
-                                    Err(Error::Runtime {
-                                        message: format!("Reader task panicked: {}", join_err),
-                                    }),
-                                    (rx, None),
-                                ));
-                            }
+                        if let Some(handle) = join_handle
+                            && let Err(join_err) = handle.await
+                        {
+                            return Some((
+                                Err(Error::Runtime {
+                                    message: format!("Reader task panicked: {}", join_err),
+                                }),
+                                (rx, None),
+                            ));
                         }
                         None
                     }

python/src/util.rs

@@ -5,8 +5,9 @@ use std::sync::Mutex;
 
 use lancedb::DistanceType;
 use pyo3::{
+    PyResult,
     exceptions::{PyRuntimeError, PyValueError},
-    pyfunction, PyResult,
+    pyfunction,
 };
 
 /// A wrapper around a rust builder

python/uv.lock

@@ -2006,7 +2006,7 @@ requires-dist = [
     { name = "botocore", marker = "extra == 'embeddings'", specifier = ">=1.31.57" },
     { name = "cohere", marker = "extra == 'embeddings'" },
     { name = "colpali-engine", marker = "extra == 'embeddings'", specifier = ">=0.3.10" },
-    { name = "datafusion", marker = "extra == 'tests'" },
+    { name = "datafusion", marker = "extra == 'tests'", specifier = "<52" },
     { name = "deprecation" },
     { name = "duckdb", marker = "extra == 'tests'" },
     { name = "google-generativeai", marker = "extra == 'embeddings'" },
@@ -2035,7 +2035,7 @@ requires-dist = [
     { name = "pyarrow-stubs", marker = "extra == 'tests'" },
     { name = "pydantic", specifier = ">=1.10" },
     { name = "pylance", marker = "extra == 'pylance'", specifier = ">=1.0.0b14" },
-    { name = "pylance", marker = "extra == 'tests'", specifier = ">=1.0.0b14" },
+    { name = "pylance", marker = "extra == 'tests'", specifier = ">=1.0.0b14,<3.0.0" },
     { name = "pyright", marker = "extra == 'dev'" },
     { name = "pytest", marker = "extra == 'tests'" },
     { name = "pytest-asyncio", marker = "extra == 'tests'" },

rust/lancedb/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb"
-version = "0.27.0-beta.2"
+version = "0.27.0-beta.3"
 edition.workspace = true
 description = "LanceDB: A serverless, low-latency vector database for AI applications"
 license.workspace = true
@@ -25,9 +25,9 @@ datafusion-catalog.workspace = true
 datafusion-common.workspace = true
 datafusion-execution.workspace = true
 datafusion-expr.workspace = true
-datafusion-functions = "51.0"
+datafusion-functions.workspace = true
 datafusion-physical-expr.workspace = true
-datafusion-sql = "51.0"
+datafusion-sql.workspace = true
 datafusion-physical-plan.workspace = true
 datafusion.workspace = true
 object_store = { workspace = true }

rust/lancedb/examples/bedrock.rs

@@ -9,10 +9,9 @@ use aws_config::Region;
 use aws_sdk_bedrockruntime::Client;
 use futures::StreamExt;
 use lancedb::{
-    connect,
-    embeddings::{bedrock::BedrockEmbeddingFunction, EmbeddingDefinition, EmbeddingFunction},
+    Result, connect,
+    embeddings::{EmbeddingDefinition, EmbeddingFunction, bedrock::BedrockEmbeddingFunction},
     query::{ExecutableQuery, QueryBase},
-    Result,
 };
 
 #[tokio::main]

rust/lancedb/examples/full_text_search.rs

@@ -10,10 +10,10 @@ use futures::TryStreamExt;
 use lance_index::scalar::FullTextSearchQuery;
 use lancedb::connection::Connection;
 
-use lancedb::index::scalar::FtsIndexBuilder;
 use lancedb::index::Index;
+use lancedb::index::scalar::FtsIndexBuilder;
 use lancedb::query::{ExecutableQuery, QueryBase};
-use lancedb::{connect, Result, Table};
+use lancedb::{Result, Table, connect};
 use rand::random;
 
 #[tokio::main]
@@ -46,19 +46,21 @@ fn create_some_records() -> Result<Box<dyn arrow_array::RecordBatchReader + Send
         .collect::<Vec<_>>();
     let n_terms = 3;
     let batches = RecordBatchIterator::new(
-        vec![RecordBatch::try_new(
-            schema.clone(),
-            vec![
-                Arc::new(Int32Array::from_iter_values(0..TOTAL as i32)),
-                Arc::new(StringArray::from_iter_values((0..TOTAL).map(|_| {
-                    (0..n_terms)
-                        .map(|_| words[random::<u32>() as usize % words.len()])
-                        .collect::<Vec<_>>()
-                        .join(" ")
-                }))),
-            ],
-        )
-        .unwrap()]
+        vec![
+            RecordBatch::try_new(
+                schema.clone(),
+                vec![
+                    Arc::new(Int32Array::from_iter_values(0..TOTAL as i32)),
+                    Arc::new(StringArray::from_iter_values((0..TOTAL).map(|_| {
+                        (0..n_terms)
+                            .map(|_| words[random::<u32>() as usize % words.len()])
+                            .collect::<Vec<_>>()
+                            .join(" ")
+                    }))),
+                ],
+            )
+            .unwrap(),
+        ]
         .into_iter()
         .map(Ok),
         schema.clone(),

rust/lancedb/examples/hybrid_search.rs

@@ -5,16 +5,15 @@ use arrow_array::{RecordBatch, StringArray};
 use arrow_schema::{DataType, Field, Schema};
 use futures::TryStreamExt;
 use lance_index::scalar::FullTextSearchQuery;
-use lancedb::index::scalar::FtsIndexBuilder;
 use lancedb::index::Index;
+use lancedb::index::scalar::FtsIndexBuilder;
 use lancedb::{
-    connect,
+    Result, Table, connect,
     embeddings::{
-        sentence_transformers::SentenceTransformersEmbeddings, EmbeddingDefinition,
-        EmbeddingFunction,
+        EmbeddingDefinition, EmbeddingFunction,
+        sentence_transformers::SentenceTransformersEmbeddings,
     },
     query::{QueryBase, QueryExecutionOptions},
-    Result, Table,
 };
 use std::{iter::once, sync::Arc};
 

rust/lancedb/examples/ivf_pq.rs

@@ -14,10 +14,10 @@ use arrow_schema::{DataType, Field, Schema};
 use futures::TryStreamExt;
 use lancedb::connection::Connection;
 
-use lancedb::index::vector::IvfPqIndexBuilder;
 use lancedb::index::Index;
+use lancedb::index::vector::IvfPqIndexBuilder;
 use lancedb::query::{ExecutableQuery, QueryBase};
-use lancedb::{connect, DistanceType, Result, Table};
+use lancedb::{DistanceType, Result, Table, connect};
 
 #[tokio::main]
 async fn main() -> Result<()> {
@@ -51,19 +51,21 @@ fn create_some_records() -> Result<Box<dyn arrow_array::RecordBatchReader + Send
 
     // Create a RecordBatch stream.
     let batches = RecordBatchIterator::new(
-        vec![RecordBatch::try_new(
-            schema.clone(),
-            vec![
-                Arc::new(Int32Array::from_iter_values(0..TOTAL as i32)),
-                Arc::new(
-                    FixedSizeListArray::from_iter_primitive::<Float32Type, _, _>(
-                        (0..TOTAL).map(|_| Some(vec![Some(1.0); DIM])),
-                        DIM as i32,
+        vec![
+            RecordBatch::try_new(
+                schema.clone(),
+                vec![
+                    Arc::new(Int32Array::from_iter_values(0..TOTAL as i32)),
+                    Arc::new(
+                        FixedSizeListArray::from_iter_primitive::<Float32Type, _, _>(
+                            (0..TOTAL).map(|_| Some(vec![Some(1.0); DIM])),
+                            DIM as i32,
+                        ),
                     ),
-                ),
-            ],
-        )
-        .unwrap()]
+                ],
+            )
+            .unwrap(),
+        ]
         .into_iter()
         .map(Ok),
         schema.clone(),

rust/lancedb/examples/openai.rs

@@ -8,10 +8,9 @@ use std::{iter::once, sync::Arc};
 use arrow_array::{RecordBatch, StringArray};
 use futures::StreamExt;
 use lancedb::{
-    connect,
-    embeddings::{openai::OpenAIEmbeddingFunction, EmbeddingDefinition, EmbeddingFunction},
+    Result, connect,
+    embeddings::{EmbeddingDefinition, EmbeddingFunction, openai::OpenAIEmbeddingFunction},
     query::{ExecutableQuery, QueryBase},
-    Result,
 };
 
 // --8<-- [end:imports]

rust/lancedb/examples/sentence_transformers.rs

@@ -7,13 +7,12 @@ use arrow_array::{RecordBatch, StringArray};
 use arrow_schema::{DataType, Field, Schema};
 use futures::StreamExt;
 use lancedb::{
-    connect,
+    Result, connect,
     embeddings::{
-        sentence_transformers::SentenceTransformersEmbeddings, EmbeddingDefinition,
-        EmbeddingFunction,
+        EmbeddingDefinition, EmbeddingFunction,
+        sentence_transformers::SentenceTransformersEmbeddings,
     },
     query::{ExecutableQuery, QueryBase},
-    Result,
 };
 
 #[tokio::main]

rust/lancedb/examples/simple.rs

@@ -14,7 +14,7 @@ use futures::TryStreamExt;
 use lancedb::connection::Connection;
 use lancedb::index::Index;
 use lancedb::query::{ExecutableQuery, QueryBase};
-use lancedb::{connect, Result, Table as LanceDbTable};
+use lancedb::{Result, Table as LanceDbTable, connect};
 
 #[tokio::main]
 async fn main() -> Result<()> {

rust/lancedb/src/arrow.rs

@@ -12,7 +12,7 @@ use lance_datagen::{BatchCount, BatchGeneratorBuilder, RowCount};
 #[cfg(feature = "polars")]
 use {crate::polars_arrow_convertors, polars::frame::ArrowChunk, polars::prelude::DataFrame};
 
-use crate::{error::Result, Error};
+use crate::{Error, error::Result};
 
 /// An iterator of batches that also has a schema
 pub trait RecordBatchReader: Iterator<Item = Result<arrow_array::RecordBatch>> {

rust/lancedb/src/connection.rs

@@ -17,6 +17,7 @@ use lance_namespace::models::{
 #[cfg(feature = "aws")]
 use object_store::aws::AwsCredential;
 
+use crate::Table;
 use crate::connection::create_table::CreateTableBuilder;
 use crate::data::scannable::Scannable;
 use crate::database::listing::ListingDatabase;
@@ -31,7 +32,6 @@ use crate::remote::{
     client::ClientConfig,
     db::{OPT_REMOTE_API_KEY, OPT_REMOTE_HOST_OVERRIDE, OPT_REMOTE_REGION},
 };
-use crate::Table;
 use lance::io::ObjectStoreParams;
 pub use lance_encoding::version::LanceFileVersion;
 #[cfg(feature = "remote")]
@@ -136,6 +136,7 @@ impl OpenTableBuilder {
                 lance_read_params: None,
                 location: None,
                 namespace_client: None,
+                managed_versioning: None,
             },
             embedding_registry,
         }
@@ -235,6 +236,29 @@ impl OpenTableBuilder {
         self
     }
 
+    /// Set a namespace client for managed versioning support.
+    ///
+    /// When a namespace client is provided and the table has `managed_versioning` enabled,
+    /// the table will use the namespace's commit handler to notify the namespace of
+    /// version changes. This enables features like event emission for table modifications.
+    pub fn namespace_client(mut self, client: Arc<dyn lance_namespace::LanceNamespace>) -> Self {
+        self.request.namespace_client = Some(client);
+        self
+    }
+
+    /// Set whether managed versioning is enabled for this table.
+    ///
+    /// When set to `Some(true)`, the table will use namespace-managed commits.
+    /// When set to `Some(false)`, the table will use local commits even if namespace_client is set.
+    /// When set to `None` (default), the value will be fetched from the namespace if namespace_client is set.
+    ///
+    /// This is typically set when the caller has already queried the namespace and knows the
+    /// managed_versioning status, avoiding a redundant describe_table call.
+    pub fn managed_versioning(mut self, enabled: bool) -> Self {
+        self.request.managed_versioning = Some(enabled);
+        self
+    }
+
     /// Open the table
     pub async fn execute(self) -> Result<Table> {
         let table = self.parent.open_table(self.request).await?;
@@ -294,6 +318,12 @@ impl CloneTableBuilder {
         self
     }
 
+    /// Set a namespace client for managed versioning support.
+    pub fn namespace_client(mut self, client: Arc<dyn lance_namespace::LanceNamespace>) -> Self {
+        self.request.namespace_client = Some(client);
+        self
+    }
+
     /// Execute the clone operation
     pub async fn execute(self) -> Result<Table> {
         let parent = self.parent.clone();
@@ -566,8 +596,11 @@ pub struct ConnectBuilder {
 }
 
 #[cfg(feature = "remote")]
-const ENV_VARS_TO_STORAGE_OPTS: [(&str, &str); 1] =
-    [("AZURE_STORAGE_ACCOUNT_NAME", "azure_storage_account_name")];
+const ENV_VARS_TO_STORAGE_OPTS: [(&str, &str); 3] = [
+    ("AZURE_STORAGE_ACCOUNT_NAME", "azure_storage_account_name"),
+    ("AZURE_CLIENT_ID", "azure_client_id"),
+    ("AZURE_TENANT_ID", "azure_tenant_id"),
+];
 
 impl ConnectBuilder {
     /// Create a new [`ConnectOptions`] with the given database URI.
@@ -758,10 +791,10 @@ impl ConnectBuilder {
         options: &mut HashMap<String, String>,
     ) {
         for (env_key, opt_key) in env_var_to_remote_storage_option {
-            if let Ok(env_value) = std::env::var(env_key) {
-                if !options.contains_key(*opt_key) {
-                    options.insert((*opt_key).to_string(), env_value);
-                }
+            if let Ok(env_value) = std::env::var(env_key)
+                && !options.contains_key(*opt_key)
+            {
+                options.insert((*opt_key).to_string(), env_value);
             }
         }
     }
@@ -1011,14 +1044,13 @@ mod tests {
     #[cfg(feature = "remote")]
     #[test]
     fn test_apply_env_defaults() {
-        let env_key = "TEST_APPLY_ENV_DEFAULTS_ENVIRONMENT_VARIABLE_ENV_KEY";
-        let env_val = "TEST_APPLY_ENV_DEFAULTS_ENVIRONMENT_VARIABLE_ENV_VAL";
+        let env_key = "PATH";
+        let env_val = std::env::var(env_key).expect("PATH should be set in test environment");
         let opts_key = "test_apply_env_defaults_environment_variable_opts_key";
-        std::env::set_var(env_key, env_val);
 
         let mut options = HashMap::new();
         ConnectBuilder::apply_env_defaults(&[(env_key, opts_key)], &mut options);
-        assert_eq!(Some(&env_val.to_string()), options.get(opts_key));
+        assert_eq!(Some(&env_val), options.get(opts_key));
 
         options.insert(opts_key.to_string(), "EXPLICIT-VALUE".to_string());
         ConnectBuilder::apply_env_defaults(&[(env_key, opts_key)], &mut options);

rust/lancedb/src/connection/create_table.rs

@@ -6,12 +6,12 @@ use std::sync::Arc;
 use lance_io::object_store::StorageOptionsProvider;
 
 use crate::{
+    Error, Result, Table,
     connection::{merge_storage_options, set_storage_options_provider},
     data::scannable::{Scannable, WithEmbeddingsScannable},
     database::{CreateTableMode, CreateTableRequest, Database},
     embeddings::{EmbeddingDefinition, EmbeddingFunction, EmbeddingRegistry},
     table::WriteOptions,
-    Error, Result, Table,
 };
 
 pub struct CreateTableBuilder {
@@ -167,7 +167,7 @@ impl CreateTableBuilder {
 #[cfg(test)]
 mod tests {
     use arrow_array::{
-        record_batch, Array, FixedSizeListArray, Float32Array, RecordBatch, RecordBatchIterator,
+        Array, FixedSizeListArray, Float32Array, RecordBatch, RecordBatchIterator, record_batch,
     };
     use arrow_schema::{ArrowError, DataType, Field, Schema};
     use futures::TryStreamExt;
@@ -380,11 +380,12 @@ mod tests {
             .await
             .unwrap();
         let other_schema = Arc::new(Schema::new(vec![Field::new("y", DataType::Int32, false)]));
-        assert!(db
-            .create_empty_table("test", other_schema.clone())
-            .execute()
-            .await
-            .is_err()); // TODO: assert what this error is
+        assert!(
+            db.create_empty_table("test", other_schema.clone())
+                .execute()
+                .await
+                .is_err()
+        ); // TODO: assert what this error is
         let overwritten = db
             .create_empty_table("test", other_schema.clone())
             .mode(CreateTableMode::Overwrite)

rust/lancedb/src/data/inspect.rs

@@ -5,9 +5,9 @@ use std::collections::HashMap;
 
 use arrow::compute::kernels::{aggregate::bool_and, length::length};
 use arrow_array::{
+    Array, GenericListArray, OffsetSizeTrait, PrimitiveArray, RecordBatchReader,
     cast::AsArray,
     types::{ArrowPrimitiveType, Int32Type, Int64Type},
-    Array, GenericListArray, OffsetSizeTrait, PrimitiveArray, RecordBatchReader,
 };
 use arrow_ord::cmp::eq;
 use arrow_schema::DataType;
@@ -78,7 +78,7 @@ pub fn infer_vector_columns(
                 _ => {
                     return Err(Error::Schema {
                         message: format!("Column {} is not a list", col_name),
-                    })
+                    });
                 }
             } {
                 if let Some(Some(prev_dim)) = columns_to_infer.get(&col_name) {
@@ -102,8 +102,8 @@ mod tests {
     use super::*;
 
     use arrow_array::{
-        types::{Float32Type, Float64Type},
         FixedSizeListArray, Float32Array, ListArray, RecordBatch, RecordBatchIterator, StringArray,
+        types::{Float32Type, Float64Type},
     };
     use arrow_schema::{DataType, Field, Schema};
     use std::{sync::Arc, vec};

rust/lancedb/src/data/sanitize.rs

@@ -4,10 +4,10 @@
 use std::{iter::repeat_with, sync::Arc};
 
 use arrow_array::{
-    cast::AsArray,
-    types::{Float16Type, Float32Type, Float64Type, Int32Type, Int64Type},
     Array, ArrowNumericType, FixedSizeListArray, PrimitiveArray, RecordBatch, RecordBatchIterator,
     RecordBatchReader,
+    cast::AsArray,
+    types::{Float16Type, Float32Type, Float64Type, Int32Type, Int64Type},
 };
 use arrow_cast::{can_cast_types, cast};
 use arrow_schema::{ArrowError, DataType, Field, Schema};
@@ -184,7 +184,7 @@ mod tests {
     use std::sync::Arc;
 
     use arrow_array::{
-        FixedSizeListArray, Float16Array, Float32Array, Float64Array, Int32Array, Int8Array,
+        FixedSizeListArray, Float16Array, Float32Array, Float64Array, Int8Array, Int32Array,
         RecordBatch, RecordBatchIterator, StringArray,
     };
     use arrow_schema::Field;

rust/lancedb/src/data/scannable.rs

@@ -13,16 +13,16 @@ use crate::arrow::{
     SendableRecordBatchStream, SendableRecordBatchStreamExt, SimpleRecordBatchStream,
 };
 use crate::embeddings::{
-    compute_embeddings_for_batch, compute_output_schema, EmbeddingDefinition, EmbeddingFunction,
-    EmbeddingRegistry,
+    EmbeddingDefinition, EmbeddingFunction, EmbeddingRegistry, compute_embeddings_for_batch,
+    compute_output_schema,
 };
 use crate::table::{ColumnDefinition, ColumnKind, TableDefinition};
 use crate::{Error, Result};
 use arrow_array::{ArrayRef, RecordBatch, RecordBatchIterator, RecordBatchReader};
 use arrow_schema::{ArrowError, SchemaRef};
 use async_trait::async_trait;
-use futures::stream::once;
 use futures::StreamExt;
+use futures::stream::once;
 use lance_datafusion::utils::StreamingWriteSource;
 
 pub trait Scannable: Send {

rust/lancedb/src/database.rs

@@ -19,12 +19,12 @@ use std::sync::Arc;
 use std::time::Duration;
 
 use lance::dataset::ReadParams;
+use lance_namespace::LanceNamespace;
 use lance_namespace::models::{
     CreateNamespaceRequest, CreateNamespaceResponse, DescribeNamespaceRequest,
     DescribeNamespaceResponse, DropNamespaceRequest, DropNamespaceResponse, ListNamespacesRequest,
     ListNamespacesResponse, ListTablesRequest, ListTablesResponse,
 };
-use lance_namespace::LanceNamespace;
 
 use crate::data::scannable::Scannable;
 use crate::error::Result;
@@ -66,6 +66,10 @@ pub struct OpenTableRequest {
     /// Optional namespace client for server-side query execution.
     /// When set, queries will be executed on the namespace server instead of locally.
     pub namespace_client: Option<Arc<dyn LanceNamespace>>,
+    /// Whether managed versioning is enabled for this table.
+    /// When Some(true), the table will use namespace-managed commits instead of local commits.
+    /// When None and namespace_client is provided, the value will be fetched from the namespace.
+    pub managed_versioning: Option<bool>,
 }
 
 impl std::fmt::Debug for OpenTableRequest {
@@ -77,6 +81,7 @@ impl std::fmt::Debug for OpenTableRequest {
             .field("lance_read_params", &self.lance_read_params)
             .field("location", &self.location)
             .field("namespace_client", &self.namespace_client)
+            .field("managed_versioning", &self.managed_versioning)
             .finish()
     }
 }
@@ -161,6 +166,9 @@ pub struct CloneTableRequest {
     /// Whether to perform a shallow clone (true) or deep clone (false). Defaults to true.
     /// Currently only shallow clone is supported.
     pub is_shallow: bool,
+    /// Optional namespace client for managed versioning support.
+    /// When set, enables the commit handler to track table versions through the namespace.
+    pub namespace_client: Option<Arc<dyn LanceNamespace>>,
 }
 
 impl CloneTableRequest {
@@ -172,6 +180,7 @@ impl CloneTableRequest {
             source_version: None,
             source_tag: None,
             is_shallow: true,
+            namespace_client: None,
         }
     }
 }

rust/lancedb/src/database/listing.rs

@@ -8,7 +8,7 @@ use std::path::Path;
 use std::{collections::HashMap, sync::Arc};
 
 use lance::dataset::refs::Ref;
-use lance::dataset::{builder::DatasetBuilder, ReadParams, WriteMode};
+use lance::dataset::{ReadParams, WriteMode, builder::DatasetBuilder};
 use lance::io::{ObjectStore, ObjectStoreParams, WrappingObjectStore};
 use lance_datafusion::utils::StreamingWriteSource;
 use lance_encoding::version::LanceFileVersion;
@@ -669,6 +669,7 @@ impl ListingDatabase {
                     lance_read_params: None,
                     location: None,
                     namespace_client: None,
+                    managed_versioning: None,
                 };
                 let req = (callback)(req);
                 let table = self.open_table(req).await?;
@@ -869,6 +870,7 @@ impl Database for ListingDatabase {
             Some(write_params),
             self.read_consistency_interval,
             request.namespace_client,
+            false, // server_side_query_enabled - listing database doesn't support server-side queries
         )
         .await
         {
@@ -946,7 +948,9 @@ impl Database for ListingDatabase {
             self.store_wrapper.clone(),
             None,
             self.read_consistency_interval,
-            None,
+            request.namespace_client,
+            false, // server_side_query_enabled - listing database doesn't support server-side queries
+            None,  // managed_versioning - will be queried if namespace_client is provided
         )
         .await?;
 
@@ -1022,6 +1026,8 @@ impl Database for ListingDatabase {
                 Some(read_params),
                 self.read_consistency_interval,
                 request.namespace_client,
+                false, // server_side_query_enabled - listing database doesn't support server-side queries
+                request.managed_versioning, // Pass through managed_versioning from request
             )
             .await?,
         );
@@ -1097,11 +1103,11 @@ impl Database for ListingDatabase {
 #[cfg(test)]
 mod tests {
     use super::*;
+    use crate::Table;
     use crate::connection::ConnectRequest;
     use crate::data::scannable::Scannable;
     use crate::database::{CreateTableMode, CreateTableRequest};
     use crate::table::WriteOptions;
-    use crate::Table;
     use arrow_array::{Int32Array, RecordBatch, StringArray};
     use arrow_schema::{DataType, Field, Schema};
     use std::path::PathBuf;
@@ -1162,6 +1168,7 @@ mod tests {
                 source_version: None,
                 source_tag: None,
                 is_shallow: true,
+                namespace_client: None,
             })
             .await
             .unwrap();
@@ -1222,6 +1229,7 @@ mod tests {
                 source_version: None,
                 source_tag: None,
                 is_shallow: true,
+                namespace_client: None,
             })
             .await
             .unwrap();
@@ -1281,6 +1289,7 @@ mod tests {
                 source_version: None,
                 source_tag: None,
                 is_shallow: true,
+                namespace_client: None,
             })
             .await;
 
@@ -1317,6 +1326,7 @@ mod tests {
                 source_version: None,
                 source_tag: None,
                 is_shallow: false, // Request deep clone
+                namespace_client: None,
             })
             .await;
 
@@ -1357,6 +1367,7 @@ mod tests {
                 source_version: None,
                 source_tag: None,
                 is_shallow: true,
+                namespace_client: None,
             })
             .await;
 
@@ -1397,6 +1408,7 @@ mod tests {
                 source_version: None,
                 source_tag: None,
                 is_shallow: true,
+                namespace_client: None,
             })
             .await;
 
@@ -1416,6 +1428,7 @@ mod tests {
                 source_version: None,
                 source_tag: None,
                 is_shallow: true,
+                namespace_client: None,
             })
             .await;
 
@@ -1452,6 +1465,7 @@ mod tests {
                 source_version: Some(1),
                 source_tag: Some("v1.0".to_string()),
                 is_shallow: true,
+                namespace_client: None,
             })
             .await;
 
@@ -1525,6 +1539,7 @@ mod tests {
                 source_version: Some(initial_version),
                 source_tag: None,
                 is_shallow: true,
+                namespace_client: None,
             })
             .await
             .unwrap();
@@ -1603,6 +1618,7 @@ mod tests {
                 source_version: None,
                 source_tag: Some("v1.0".to_string()),
                 is_shallow: true,
+                namespace_client: None,
             })
             .await
             .unwrap();
@@ -1654,6 +1670,7 @@ mod tests {
                 source_version: None,
                 source_tag: None,
                 is_shallow: true,
+                namespace_client: None,
             })
             .await
             .unwrap();
@@ -1746,6 +1763,7 @@ mod tests {
                 source_version: None,
                 source_tag: None,
                 is_shallow: true,
+                namespace_client: None,
             })
             .await
             .unwrap();

rust/lancedb/src/database/namespace.rs

@@ -7,18 +7,20 @@ use std::collections::HashMap;
 use std::sync::Arc;
 
 use async_trait::async_trait;
+use lance::io::commit::namespace_manifest::LanceNamespaceExternalManifestStore;
 use lance_io::object_store::{ObjectStoreParams, StorageOptionsAccessor};
 use lance_namespace::{
-    models::{
-        CreateEmptyTableRequest, CreateNamespaceRequest, CreateNamespaceResponse,
-        DeclareTableRequest, DescribeNamespaceRequest, DescribeNamespaceResponse,
-        DescribeTableRequest, DropNamespaceRequest, DropNamespaceResponse, DropTableRequest,
-        ListNamespacesRequest, ListNamespacesResponse, ListTablesRequest, ListTablesResponse,
-    },
     LanceNamespace,
+    models::{
+        CreateNamespaceRequest, CreateNamespaceResponse, DeclareTableRequest,
+        DescribeNamespaceRequest, DescribeNamespaceResponse, DescribeTableRequest,
+        DropNamespaceRequest, DropNamespaceResponse, DropTableRequest, ListNamespacesRequest,
+        ListNamespacesResponse, ListTablesRequest, ListTablesResponse,
+    },
 };
 use lance_namespace_impls::ConnectBuilder;
-use log::warn;
+use lance_table::io::commit::CommitHandler;
+use lance_table::io::commit::external_manifest::ExternalManifestCommitHandler;
 
 use crate::database::ReadConsistency;
 use crate::error::{Error, Result};
@@ -206,83 +208,55 @@ impl Database for LanceNamespaceDatabase {
         let mut table_id = request.namespace.clone();
         table_id.push(request.name.clone());
 
-        // Try declare_table first, falling back to create_empty_table for backwards
-        // compatibility with older namespace clients that don't support declare_table
         let declare_request = DeclareTableRequest {
             id: Some(table_id.clone()),
             ..Default::default()
         };
 
-        let (location, initial_storage_options) =
-            match self.namespace.declare_table(declare_request).await {
-                Ok(response) => {
-                    let loc = response.location.ok_or_else(|| Error::Runtime {
-                        message: "Table location is missing from declare_table response"
-                            .to_string(),
-                    })?;
-                    // Use storage options from response, fall back to self.storage_options
-                    let opts = response
-                        .storage_options
-                        .or_else(|| Some(self.storage_options.clone()))
-                        .filter(|o| !o.is_empty());
-                    (loc, opts)
-                }
-                Err(e) => {
-                    // Check if the error is "not supported" and try create_empty_table as fallback
-                    let err_str = e.to_string().to_lowercase();
-                    if err_str.contains("not supported") || err_str.contains("not implemented") {
-                        warn!(
-                            "declare_table is not supported by the namespace client, \
-                        falling back to deprecated create_empty_table. \
-                        create_empty_table is deprecated and will be removed in Lance 3.0.0. \
-                        Please upgrade your namespace client to support declare_table."
-                        );
-                        #[allow(deprecated)]
-                        let create_empty_request = CreateEmptyTableRequest {
-                            id: Some(table_id.clone()),
-                            ..Default::default()
-                        };
+        let response = self
+            .namespace
+            .declare_table(declare_request)
+            .await
+            .map_err(|e| Error::Runtime {
+                message: format!("Failed to declare table: {}", e),
+            })?;
 
-                        #[allow(deprecated)]
-                        let create_response = self
-                            .namespace
-                            .create_empty_table(create_empty_request)
-                            .await
-                            .map_err(|e| Error::Runtime {
-                                message: format!("Failed to create empty table: {}", e),
-                            })?;
+        let location = response.location.ok_or_else(|| Error::Runtime {
+            message: "Table location is missing from declare_table response".to_string(),
+        })?;
 
-                        let loc = create_response.location.ok_or_else(|| Error::Runtime {
-                            message: "Table location is missing from create_empty_table response"
-                                .to_string(),
-                        })?;
-                        // For deprecated path, use self.storage_options
-                        let opts = if self.storage_options.is_empty() {
-                            None
-                        } else {
-                            Some(self.storage_options.clone())
-                        };
-                        (loc, opts)
-                    } else {
-                        return Err(Error::Runtime {
-                            message: format!("Failed to declare table: {}", e),
-                        });
-                    }
-                }
-            };
+        // Use storage options from response, fall back to self.storage_options
+        let initial_storage_options = response
+            .storage_options
+            .or_else(|| Some(self.storage_options.clone()))
+            .filter(|o| !o.is_empty());
 
-        let write_params = if let Some(storage_opts) = initial_storage_options {
-            let mut params = request.write_options.lance_write_params.unwrap_or_default();
+        let managed_versioning = response.managed_versioning;
+
+        // Build write params with storage options and commit handler
+        let mut params = request.write_options.lance_write_params.unwrap_or_default();
+
+        // Set up storage options if provided
+        if let Some(storage_opts) = initial_storage_options {
             let store_params = params
                 .store_params
                 .get_or_insert_with(ObjectStoreParams::default);
             store_params.storage_options_accessor = Some(Arc::new(
                 StorageOptionsAccessor::with_static_options(storage_opts),
             ));
-            Some(params)
-        } else {
-            request.write_options.lance_write_params
-        };
+        }
+
+        // Set up commit handler when managed_versioning is enabled
+        if managed_versioning == Some(true) {
+            let external_store =
+                LanceNamespaceExternalManifestStore::new(self.namespace.clone(), table_id.clone());
+            let commit_handler: Arc<dyn CommitHandler> = Arc::new(ExternalManifestCommitHandler {
+                external_manifest_store: Arc::new(external_store),
+            });
+            params.commit_handler = Some(commit_handler);
+        }
+
+        let write_params = Some(params);
 
         let native_table = NativeTable::create_from_namespace(
             self.namespace.clone(),

rust/lancedb/src/dataloader/permutation/builder.rs

@@ -11,16 +11,16 @@ use lance_core::ROW_ID;
 use lance_datafusion::exec::SessionContextExt;
 
 use crate::{
+    Error, Result, Table,
     arrow::{SendableRecordBatchStream, SendableRecordBatchStreamExt, SimpleRecordBatchStream},
     connect,
     database::{CreateTableRequest, Database},
     dataloader::permutation::{
         shuffle::{Shuffler, ShufflerConfig},
-        split::{SplitStrategy, Splitter, SPLIT_ID_COLUMN},
-        util::{rename_column, TemporaryDirectory},
+        split::{SPLIT_ID_COLUMN, SplitStrategy, Splitter},
+        util::{TemporaryDirectory, rename_column},
     },
     query::{ExecutableQuery, QueryBase, Select},
-    Error, Result, Table,
 };
 
 pub const SRC_ROW_ID_COL: &str = "row_id";

rust/lancedb/src/dataloader/permutation/reader.rs

@@ -25,8 +25,8 @@ use futures::{StreamExt, TryStreamExt};
 use lance::dataset::scanner::DatasetRecordBatchStream;
 use lance::io::RecordBatchStream;
 use lance_arrow::RecordBatchExt;
-use lance_core::error::LanceOptionExt;
 use lance_core::ROW_ID;
+use lance_core::error::LanceOptionExt;
 use std::collections::HashMap;
 use std::sync::Arc;
 
@@ -500,10 +500,10 @@ mod tests {
     use rand::seq::SliceRandom;
 
     use crate::{
+        Table,
         arrow::SendableRecordBatchStream,
         query::{ExecutableQuery, QueryBase},
-        test_utils::datagen::{virtual_table, LanceDbDatagenExt},
-        Table,
+        test_utils::datagen::{LanceDbDatagenExt, virtual_table},
     };
 
     use super::*;

rust/lancedb/src/dataloader/permutation/shuffle.rs

@@ -18,12 +18,12 @@ use lance_io::{
     scheduler::{ScanScheduler, SchedulerConfig},
     utils::CachedFileSize,
 };
-use rand::{seq::SliceRandom, Rng, RngCore};
+use rand::{Rng, RngCore, seq::SliceRandom};
 
 use crate::{
-    arrow::{SendableRecordBatchStream, SimpleRecordBatchStream},
-    dataloader::permutation::util::{non_crypto_rng, TemporaryDirectory},
     Error, Result,
+    arrow::{SendableRecordBatchStream, SimpleRecordBatchStream},
+    dataloader::permutation::util::{TemporaryDirectory, non_crypto_rng},
 };
 
 #[derive(Debug, Clone)]
@@ -281,7 +281,7 @@ mod tests {
     use datafusion_expr::col;
     use futures::TryStreamExt;
     use lance_datagen::{BatchCount, BatchGeneratorBuilder, ByteCount, RowCount, Seed};
-    use rand::{rngs::SmallRng, SeedableRng};
+    use rand::{SeedableRng, rngs::SmallRng};
 
     fn test_gen() -> BatchGeneratorBuilder {
         lance_datagen::gen_batch()

rust/lancedb/src/dataloader/permutation/split.rs

@@ -2,8 +2,8 @@
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
 use std::sync::{
-    atomic::{AtomicBool, AtomicU64, AtomicUsize, Ordering},
     Arc,
+    atomic::{AtomicBool, AtomicU64, AtomicUsize, Ordering},
 };
 
 use arrow_array::{Array, BooleanArray, RecordBatch, UInt64Array};
@@ -15,13 +15,13 @@ use lance_arrow::SchemaExt;
 use lance_core::ROW_ID;
 
 use crate::{
+    Error, Result,
     arrow::{SendableRecordBatchStream, SimpleRecordBatchStream},
     dataloader::{
         permutation::shuffle::{Shuffler, ShufflerConfig},
         permutation::util::TemporaryDirectory,
     },
     query::{Query, QueryBase, Select},
-    Error, Result,
 };
 
 pub const SPLIT_ID_COLUMN: &str = "split_id";

rust/lancedb/src/dataloader/permutation/util.rs

@@ -7,12 +7,12 @@ use arrow_array::RecordBatch;
 use arrow_schema::{Fields, Schema};
 use datafusion_execution::disk_manager::DiskManagerMode;
 use futures::TryStreamExt;
-use rand::{rngs::SmallRng, RngCore, SeedableRng};
+use rand::{RngCore, SeedableRng, rngs::SmallRng};
 use tempfile::TempDir;
 
 use crate::{
-    arrow::{SendableRecordBatchStream, SimpleRecordBatchStream},
     Error, Result,
+    arrow::{SendableRecordBatchStream, SimpleRecordBatchStream},
 };
 
 /// Directory to use for temporary files

rust/lancedb/src/embeddings.rs

@@ -23,9 +23,9 @@ use arrow_schema::{DataType, Field, SchemaBuilder, SchemaRef};
 use serde::{Deserialize, Serialize};
 
 use crate::{
+    Error,
     error::Result,
     table::{ColumnDefinition, ColumnKind, TableDefinition},
-    Error,
 };
 
 /// Trait for embedding functions

rust/lancedb/src/embeddings/bedrock.rs

@@ -8,7 +8,7 @@ use arrow::array::{AsArray, Float32Builder};
 use arrow_array::{Array, ArrayRef, FixedSizeListArray, Float32Array};
 use arrow_data::ArrayData;
 use arrow_schema::DataType;
-use serde_json::{json, Value};
+use serde_json::{Value, json};
 
 use super::EmbeddingFunction;
 use crate::{Error, Result};

rust/lancedb/src/embeddings/openai.rs

@@ -8,9 +8,9 @@ use arrow_array::{Array, ArrayRef, FixedSizeListArray, Float32Array};
 use arrow_data::ArrayData;
 use arrow_schema::DataType;
 use async_openai::{
+    Client,
     config::OpenAIConfig,
     types::{CreateEmbeddingRequest, Embedding, EmbeddingInput, EncodingFormat},
-    Client,
 };
 use tokio::{runtime::Handle, task};
 

rust/lancedb/src/embeddings/sentence_transformers.rs

@@ -7,7 +7,7 @@ use super::EmbeddingFunction;
 use arrow::{
     array::{AsArray, PrimitiveBuilder},
     datatypes::{
-        ArrowPrimitiveType, Float16Type, Float32Type, Float64Type, Int64Type, UInt32Type, UInt8Type,
+        ArrowPrimitiveType, Float16Type, Float32Type, Float64Type, Int64Type, UInt8Type, UInt32Type,
     },
 };
 use arrow_array::{Array, FixedSizeListArray, PrimitiveArray};
@@ -16,8 +16,8 @@ use arrow_schema::DataType;
 use candle_core::{CpuStorage, Device, Layout, Storage, Tensor};
 use candle_nn::VarBuilder;
 use candle_transformers::models::bert::{BertModel, DTYPE};
-use hf_hub::{api::sync::Api, Repo, RepoType};
-use tokenizers::{tokenizer::Tokenizer, PaddingParams};
+use hf_hub::{Repo, RepoType, api::sync::Api};
+use tokenizers::{PaddingParams, tokenizer::Tokenizer};
 
 /// Compute embeddings using huggingface sentence-transformers.
 pub struct SentenceTransformersEmbeddingsBuilder {
@@ -230,7 +230,7 @@ impl SentenceTransformersEmbeddings {
             Storage::Cpu(CpuStorage::BF16(_)) => {
                 return Err(crate::Error::Runtime {
                     message: "unsupported data type".to_string(),
-                })
+                });
             }
             _ => unreachable!("we already moved the tensor to the CPU device"),
         };
@@ -298,12 +298,12 @@ impl SentenceTransformersEmbeddings {
             DataType::Utf8View => {
                 return Err(crate::Error::Runtime {
                     message: "Utf8View not yet implemented".to_string(),
-                })
+                });
             }
             _ => {
                 return Err(crate::Error::Runtime {
                     message: "invalid type".to_string(),
-                })
+                });
             }
         };
 

rust/lancedb/src/expr.rs

@@ -24,7 +24,7 @@ pub use sql::expr_to_sql_string;
 use std::sync::Arc;
 
 use arrow_schema::DataType;
-use datafusion_expr::{expr_fn::cast, Expr, ScalarUDF};
+use datafusion_expr::{Expr, ScalarUDF, expr_fn::cast};
 use datafusion_functions::string::expr_fn as string_expr_fn;
 
 pub use datafusion_expr::{col, lit};

rust/lancedb/src/index.rs

@@ -9,7 +9,7 @@ use std::time::Duration;
 use vector::IvfFlatIndexBuilder;
 
 use crate::index::vector::IvfRqIndexBuilder;
-use crate::{table::BaseTable, DistanceType, Error, Result};
+use crate::{DistanceType, Error, Result, table::BaseTable};
 
 use self::{
     scalar::{BTreeIndexBuilder, BitmapIndexBuilder, LabelListIndexBuilder},

rust/lancedb/src/index/scalar.rs

@@ -27,7 +27,7 @@
 ///
 /// The btree index does not currently have any parameters though parameters such as the
 /// block size may be added in the future.
-#[derive(Default, Debug, Clone)]
+#[derive(Default, Debug, Clone, serde::Serialize)]
 pub struct BTreeIndexBuilder {}
 
 impl BTreeIndexBuilder {}
@@ -39,7 +39,7 @@ impl BTreeIndexBuilder {}
 /// This index works best for low-cardinality (i.e., less than 1000 unique values) columns,
 /// where the number of unique values is small.
 /// The bitmap stores a list of row ids where the value is present.
-#[derive(Debug, Clone, Default)]
+#[derive(Debug, Clone, Default, serde::Serialize)]
 pub struct BitmapIndexBuilder {}
 
 /// Builder for LabelList index.
@@ -48,10 +48,10 @@ pub struct BitmapIndexBuilder {}
 /// support queries with `array_contains_all` and `array_contains_any`
 /// using an underlying bitmap index.
 ///
-#[derive(Debug, Clone, Default)]
+#[derive(Debug, Clone, Default, serde::Serialize)]
 pub struct LabelListIndexBuilder {}
 
-pub use lance_index::scalar::inverted::query::*;
 pub use lance_index::scalar::FullTextSearchQuery;
 pub use lance_index::scalar::InvertedIndexParams as FtsIndexBuilder;
 pub use lance_index::scalar::InvertedIndexParams;
+pub use lance_index::scalar::inverted::query::*;

rust/lancedb/src/index/vector.rs

@@ -7,6 +7,7 @@
 //! Vector indices are only supported on fixed-size-list (tensor) columns of floating point
 //! values
 use lance::table::format::{IndexMetadata, Manifest};
+use serde::Serialize;
 
 use crate::DistanceType;
 
@@ -181,14 +182,17 @@ macro_rules! impl_hnsw_params_setter {
 /// The partitioning process is called IVF and the `num_partitions` parameter controls how many groups to create.
 ///
 /// Note that training an IVF Flat index on a large dataset is a slow operation and currently is also a memory intensive operation.
-#[derive(Debug, Clone)]
+#[derive(Debug, Clone, Serialize)]
 pub struct IvfFlatIndexBuilder {
+    #[serde(rename = "metric_type")]
     pub(crate) distance_type: DistanceType,
 
     // IVF
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) num_partitions: Option<u32>,
     pub(crate) sample_rate: u32,
     pub(crate) max_iterations: u32,
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) target_partition_size: Option<u32>,
 }
 
@@ -213,14 +217,17 @@ impl IvfFlatIndexBuilder {
 ///
 /// This index compresses vectors using scalar quantization and groups them into IVF partitions.
 /// It offers a balance between search performance and storage footprint.
-#[derive(Debug, Clone)]
+#[derive(Debug, Clone, Serialize)]
 pub struct IvfSqIndexBuilder {
+    #[serde(rename = "metric_type")]
     pub(crate) distance_type: DistanceType,
 
     // IVF
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) num_partitions: Option<u32>,
     pub(crate) sample_rate: u32,
     pub(crate) max_iterations: u32,
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) target_partition_size: Option<u32>,
 }
 
@@ -261,18 +268,23 @@ impl IvfSqIndexBuilder {
 ///
 /// Note that training an IVF PQ index on a large dataset is a slow operation and
 /// currently is also a memory intensive operation.
-#[derive(Debug, Clone)]
+#[derive(Debug, Clone, Serialize)]
 pub struct IvfPqIndexBuilder {
+    #[serde(rename = "metric_type")]
     pub(crate) distance_type: DistanceType,
 
     // IVF
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) num_partitions: Option<u32>,
     pub(crate) sample_rate: u32,
     pub(crate) max_iterations: u32,
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) target_partition_size: Option<u32>,
 
     // PQ
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) num_sub_vectors: Option<u32>,
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) num_bits: Option<u32>,
 }
 
@@ -323,14 +335,18 @@ pub(crate) fn suggested_num_sub_vectors(dim: u32) -> u32 {
 ///
 /// Note that training an IVF RQ index on a large dataset is a slow operation and
 /// currently is also a memory intensive operation.
-#[derive(Debug, Clone)]
+#[derive(Debug, Clone, Serialize)]
 pub struct IvfRqIndexBuilder {
     // IVF
+    #[serde(rename = "metric_type")]
     pub(crate) distance_type: DistanceType,
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) num_partitions: Option<u32>,
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) num_bits: Option<u32>,
     pub(crate) sample_rate: u32,
     pub(crate) max_iterations: u32,
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) target_partition_size: Option<u32>,
 }
 
@@ -365,13 +381,16 @@ impl IvfRqIndexBuilder {
 /// quickly find the closest vectors to a query vector.
 ///
 /// The PQ (product quantizer) is used to compress the vectors as the same as IVF PQ.
-#[derive(Debug, Clone)]
+#[derive(Debug, Clone, Serialize)]
 pub struct IvfHnswPqIndexBuilder {
     // IVF
+    #[serde(rename = "metric_type")]
     pub(crate) distance_type: DistanceType,
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) num_partitions: Option<u32>,
     pub(crate) sample_rate: u32,
     pub(crate) max_iterations: u32,
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) target_partition_size: Option<u32>,
 
     // HNSW
@@ -379,7 +398,9 @@ pub struct IvfHnswPqIndexBuilder {
     pub(crate) ef_construction: u32,
 
     // PQ
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) num_sub_vectors: Option<u32>,
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) num_bits: Option<u32>,
 }
 
@@ -415,13 +436,16 @@ impl IvfHnswPqIndexBuilder {
 ///
 /// The SQ (scalar quantizer) is used to compress the vectors,
 /// each vector is mapped to a 8-bit integer vector, 4x compression ratio for float32 vector.
-#[derive(Debug, Clone)]
+#[derive(Debug, Clone, Serialize)]
 pub struct IvfHnswSqIndexBuilder {
     // IVF
+    #[serde(rename = "metric_type")]
     pub(crate) distance_type: DistanceType,
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) num_partitions: Option<u32>,
     pub(crate) sample_rate: u32,
     pub(crate) max_iterations: u32,
+    #[serde(skip_serializing_if = "Option::is_none")]
     pub(crate) target_partition_size: Option<u32>,
 
     // HNSW

rust/lancedb/src/index/waiter.rs

@@ -1,9 +1,9 @@
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
+use crate::Error;
 use crate::error::Result;
 use crate::table::BaseTable;
-use crate::Error;
 use log::debug;
 use std::time::{Duration, Instant};
 use tokio::time::sleep;

rust/lancedb/src/io/object_store.rs

@@ -5,11 +5,11 @@
 
 use std::{fmt::Formatter, sync::Arc};
 
-use futures::{stream::BoxStream, TryFutureExt};
+use futures::{TryFutureExt, stream::BoxStream};
 use lance::io::WrappingObjectStore;
 use object_store::{
-    path::Path, Error, GetOptions, GetResult, ListResult, MultipartUpload, ObjectMeta, ObjectStore,
-    PutMultipartOptions, PutOptions, PutPayload, PutResult, Result, UploadPart,
+    Error, GetOptions, GetResult, ListResult, MultipartUpload, ObjectMeta, ObjectStore,
+    PutMultipartOptions, PutOptions, PutPayload, PutResult, Result, UploadPart, path::Path,
 };
 
 use async_trait::async_trait;

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

@@ -10,8 +10,9 @@ use bytes::Bytes;
 use futures::stream::BoxStream;
 use lance::io::WrappingObjectStore;
 use object_store::{
-    path::Path, GetOptions, GetResult, ListResult, MultipartUpload, ObjectMeta, ObjectStore,
+    GetOptions, GetResult, ListResult, MultipartUpload, ObjectMeta, ObjectStore,
     PutMultipartOptions, PutOptions, PutPayload, PutResult, Result as OSResult, UploadPart,
+    path::Path,
 };
 
 #[derive(Debug, Default)]

rust/lancedb/src/query.rs

@@ -5,26 +5,26 @@ use std::sync::Arc;
 use std::{future::Future, time::Duration};
 
 use arrow::compute::concat_batches;
-use arrow_array::{make_array, Array, Float16Array, Float32Array, Float64Array};
+use arrow_array::{Array, Float16Array, Float32Array, Float64Array, make_array};
 use arrow_schema::{DataType, SchemaRef};
 use datafusion_expr::Expr;
 use datafusion_physical_plan::ExecutionPlan;
-use futures::{stream, try_join, FutureExt, TryFutureExt, TryStreamExt};
+use futures::{FutureExt, TryFutureExt, TryStreamExt, stream, try_join};
 use half::f16;
-use lance::dataset::{scanner::DatasetRecordBatchStream, ROW_ID};
+use lance::dataset::{ROW_ID, scanner::DatasetRecordBatchStream};
 use lance_arrow::RecordBatchExt;
 use lance_datafusion::exec::execute_plan;
-use lance_index::scalar::inverted::SCORE_COL;
 use lance_index::scalar::FullTextSearchQuery;
+use lance_index::scalar::inverted::SCORE_COL;
 use lance_index::vector::DIST_COL;
 use lance_io::stream::RecordBatchStreamAdapter;
 
+use crate::DistanceType;
 use crate::error::{Error, Result};
 use crate::rerankers::rrf::RRFReranker;
-use crate::rerankers::{check_reranker_result, NormalizeMethod, Reranker};
+use crate::rerankers::{NormalizeMethod, Reranker, check_reranker_result};
 use crate::table::BaseTable;
 use crate::utils::TimeoutStream;
-use crate::DistanceType;
 use crate::{arrow::SendableRecordBatchStream, table::AnyQuery};
 
 mod hybrid;
@@ -161,10 +161,11 @@ impl IntoQueryVector for &dyn Array {
         if data_type != self.data_type() {
             Err(Error::InvalidInput {
                 message: format!(
-                "failed to create query vector, the input data type was {:?} but the expected data type was {:?}",
-                self.data_type(),
-                data_type
-            )})
+                    "failed to create query vector, the input data type was {:?} but the expected data type was {:?}",
+                    self.data_type(),
+                    data_type
+                ),
+            })
         } else {
             let data = self.to_data();
             Ok(make_array(data))
@@ -186,7 +187,7 @@ impl IntoQueryVector for &[f16] {
             DataType::Float32 => {
                 let arr: Vec<f32> = self.iter().map(|x| f32::from(*x)).collect();
                 Ok(Arc::new(Float32Array::from(arr)))
-            },
+            }
             DataType::Float64 => {
                 let arr: Vec<f64> = self.iter().map(|x| f64::from(*x)).collect();
                 Ok(Arc::new(Float64Array::from(arr)))
@@ -194,8 +195,7 @@ impl IntoQueryVector for &[f16] {
             _ => Err(Error::InvalidInput {
                 message: format!(
                     "failed to create query vector, the input data type was &[f16] but the embedding model \"{}\" expected data type {:?}",
-                    embedding_model_label,
-                    data_type
+                    embedding_model_label, data_type
                 ),
             }),
         }
@@ -216,7 +216,7 @@ impl IntoQueryVector for &[f32] {
             DataType::Float32 => {
                 let arr: Vec<f32> = self.to_vec();
                 Ok(Arc::new(Float32Array::from(arr)))
-            },
+            }
             DataType::Float64 => {
                 let arr: Vec<f64> = self.iter().map(|x| *x as f64).collect();
                 Ok(Arc::new(Float64Array::from(arr)))
@@ -224,8 +224,7 @@ impl IntoQueryVector for &[f32] {
             _ => Err(Error::InvalidInput {
                 message: format!(
                     "failed to create query vector, the input data type was &[f32] but the embedding model \"{}\" expected data type {:?}",
-                    embedding_model_label,
-                    data_type
+                    embedding_model_label, data_type
                 ),
             }),
         }
@@ -239,26 +238,25 @@ impl IntoQueryVector for &[f64] {
         embedding_model_label: &str,
     ) -> Result<Arc<dyn Array>> {
         match data_type {
-                DataType::Float16 => {
-                    let arr: Vec<f16> = self.iter().map(|x| f16::from_f64(*x)).collect();
-                    Ok(Arc::new(Float16Array::from(arr)))
-                }
-                DataType::Float32 => {
-                    let arr: Vec<f32> = self.iter().map(|x| *x as f32).collect();
-                    Ok(Arc::new(Float32Array::from(arr)))
-                },
-                DataType::Float64 => {
-                    let arr: Vec<f64> = self.to_vec();
-                    Ok(Arc::new(Float64Array::from(arr)))
-                }
-                _ => Err(Error::InvalidInput {
-                    message: format!(
-                        "failed to create query vector, the input data type was &[f64] but the embedding model \"{}\" expected data type {:?}",
-                        embedding_model_label,
-                        data_type
-                    ),
-                }),
+            DataType::Float16 => {
+                let arr: Vec<f16> = self.iter().map(|x| f16::from_f64(*x)).collect();
+                Ok(Arc::new(Float16Array::from(arr)))
             }
+            DataType::Float32 => {
+                let arr: Vec<f32> = self.iter().map(|x| *x as f32).collect();
+                Ok(Arc::new(Float32Array::from(arr)))
+            }
+            DataType::Float64 => {
+                let arr: Vec<f64> = self.to_vec();
+                Ok(Arc::new(Float64Array::from(arr)))
+            }
+            _ => Err(Error::InvalidInput {
+                message: format!(
+                    "failed to create query vector, the input data type was &[f64] but the embedding model \"{}\" expected data type {:?}",
+                    embedding_model_label, data_type
+                ),
+            }),
+        }
     }
 }
 
@@ -1011,13 +1009,13 @@ impl VectorQuery {
                 message: "minimum_nprobes must be greater than 0".to_string(),
             });
         }
-        if let Some(maximum_nprobes) = self.request.maximum_nprobes {
-            if minimum_nprobes > maximum_nprobes {
-                return Err(Error::InvalidInput {
-                    message: "minimum_nprobes must be less than or equal to maximum_nprobes"
-                        .to_string(),
-                });
-            }
+        if let Some(maximum_nprobes) = self.request.maximum_nprobes
+            && minimum_nprobes > maximum_nprobes
+        {
+            return Err(Error::InvalidInput {
+                message: "minimum_nprobes must be less than or equal to maximum_nprobes"
+                    .to_string(),
+            });
         }
         self.request.minimum_nprobes = minimum_nprobes;
         Ok(self)
@@ -1407,8 +1405,8 @@ mod tests {
     use super::*;
     use arrow::{array::downcast_array, compute::concat_batches, datatypes::Int32Type};
     use arrow_array::{
-        cast::AsArray, types::Float32Type, FixedSizeListArray, Float32Array, Int32Array,
-        RecordBatch, StringArray,
+        FixedSizeListArray, Float32Array, Int32Array, RecordBatch, StringArray, cast::AsArray,
+        types::Float32Type,
     };
     use arrow_schema::{DataType, Field as ArrowField, Schema as ArrowSchema};
     use futures::{StreamExt, TryStreamExt};
@@ -1416,7 +1414,7 @@ mod tests {
     use rand::seq::IndexedRandom;
     use tempfile::tempdir;
 
-    use crate::{connect, database::CreateTableMode, index::Index, Table};
+    use crate::{Table, connect, database::CreateTableMode, index::Index};
 
     #[tokio::test]
     async fn test_setters_getters() {
@@ -1754,11 +1752,13 @@ mod tests {
             .limit(1)
             .execute()
             .await;
-        assert!(error_result
-            .err()
-            .unwrap()
-            .to_string()
-            .contains("No vector column found to match with the query vector dimension: 3"));
+        assert!(
+            error_result
+                .err()
+                .unwrap()
+                .to_string()
+                .contains("No vector column found to match with the query vector dimension: 3")
+        );
     }
 
     #[tokio::test]
@@ -2010,7 +2010,7 @@ mod tests {
 
         // Sample 1 - 3 tokens for each string value
         let tokens = ["a", "b", "c", "d", "e"];
-        use rand::{rng, Rng};
+        use rand::{Rng, rng};
 
         let mut rng = rng();
         let text: StringArray = (0..nrows)

rust/lancedb/src/query/hybrid.rs

@@ -5,7 +5,7 @@ use arrow::compute::{
     kernels::numeric::{div, sub},
     max, min,
 };
-use arrow_array::{cast::downcast_array, Float32Array, RecordBatch};
+use arrow_array::{Float32Array, RecordBatch, cast::downcast_array};
 use arrow_schema::{DataType, Field, Schema, SortOptions};
 use lance::dataset::ROW_ID;
 use lance_index::{scalar::inverted::SCORE_COL, vector::DIST_COL};
@@ -253,7 +253,10 @@ mod test {
         let result = rank(batch.clone(), "bad_col", None);
         match result {
             Err(Error::InvalidInput { message }) => {
-                assert_eq!("expected column bad_col not found in rank. found columns [\"name\", \"score\"]", message);
+                assert_eq!(
+                    "expected column bad_col not found in rank. found columns [\"name\", \"score\"]",
+                    message
+                );
             }
             _ => {
                 panic!("expected invalid input error, received {:?}", result)

rust/lancedb/src/remote/client.rs

@@ -4,8 +4,8 @@
 use http::HeaderName;
 use log::debug;
 use reqwest::{
-    header::{HeaderMap, HeaderValue},
     Body, Request, RequestBuilder, Response,
+    header::{HeaderMap, HeaderValue},
 };
 use std::{collections::HashMap, future::Future, str::FromStr, sync::Arc, time::Duration};
 
@@ -446,13 +446,23 @@ impl<S: HttpSend> RestfulLanceDbClient<S> {
                 })?,
             );
         }
-        if let Some(v) = options.0.get("azure_storage_account_name") {
-            headers.insert(
-                HeaderName::from_static("x-azure-storage-account-name"),
-                HeaderValue::from_str(v).map_err(|_| Error::InvalidInput {
-                    message: format!("non-ascii storage account name '{}' provided", db_name),
-                })?,
-            );
+        // Map azure storage options to x-azure-* headers.
+        // The option key uses underscores (e.g. "azure_client_id") while the
+        // header uses hyphens (e.g. "x-azure-client-id").
+        let azure_opts: [(&str, &str); 3] = [
+            ("azure_storage_account_name", "x-azure-storage-account-name"),
+            ("azure_client_id", "x-azure-client-id"),
+            ("azure_tenant_id", "x-azure-tenant-id"),
+        ];
+        for (opt_key, header_name) in azure_opts {
+            if let Some(v) = options.0.get(opt_key) {
+                headers.insert(
+                    HeaderName::from_static(header_name),
+                    HeaderValue::from_str(v).map_err(|_| Error::InvalidInput {
+                        message: format!("non-ascii value '{}' for option '{}'", v, opt_key),
+                    })?,
+                );
+            }
         }
 
         for (key, value) in &config.extra_headers {
@@ -650,14 +660,13 @@ impl<S: HttpSend> RestfulLanceDbClient<S> {
     pub fn extract_request_id(&self, request: &mut Request) -> String {
         // Set a request id.
         // TODO: allow the user to supply this, through middleware?
-        let request_id = if let Some(request_id) = request.headers().get(REQUEST_ID_HEADER) {
+        if let Some(request_id) = request.headers().get(REQUEST_ID_HEADER) {
             request_id.to_str().unwrap().to_string()
         } else {
             let request_id = uuid::Uuid::new_v4().to_string();
             self.set_request_id(request, &request_id);
             request_id
-        };
-        request_id
+        }
     }
 
     /// Set the request ID header
@@ -1076,4 +1085,34 @@ mod tests {
             _ => panic!("Expected Runtime error"),
         }
     }
+
+    #[test]
+    fn test_default_headers_azure_opts() {
+        let mut opts = HashMap::new();
+        opts.insert(
+            "azure_storage_account_name".to_string(),
+            "myaccount".to_string(),
+        );
+        opts.insert("azure_client_id".to_string(), "my-client-id".to_string());
+        opts.insert("azure_tenant_id".to_string(), "my-tenant-id".to_string());
+        let remote_opts = RemoteOptions::new(opts);
+
+        let headers = RestfulLanceDbClient::<Sender>::default_headers(
+            "test-key",
+            "us-east-1",
+            "testdb",
+            false,
+            &remote_opts,
+            None,
+            &ClientConfig::default(),
+        )
+        .unwrap();
+
+        assert_eq!(
+            headers.get("x-azure-storage-account-name").unwrap(),
+            "myaccount"
+        );
+        assert_eq!(headers.get("x-azure-client-id").unwrap(), "my-client-id");
+        assert_eq!(headers.get("x-azure-tenant-id").unwrap(), "my-tenant-id");
+    }
 }

rust/lancedb/src/remote/db.rs

@@ -16,6 +16,7 @@ use lance_namespace::models::{
     ListNamespacesResponse, ListTablesRequest, ListTablesResponse,
 };
 
+use crate::Error;
 use crate::database::{
     CloneTableRequest, CreateTableMode, CreateTableRequest, Database, DatabaseOptions,
     OpenTableRequest, ReadConsistency, TableNamesRequest,
@@ -23,12 +24,11 @@ use crate::database::{
 use crate::error::Result;
 use crate::remote::util::stream_as_body;
 use crate::table::BaseTable;
-use crate::Error;
 
+use super::ARROW_STREAM_CONTENT_TYPE;
 use super::client::{ClientConfig, HttpSend, RequestResultExt, RestfulLanceDbClient, Sender};
 use super::table::RemoteTable;
 use super::util::parse_server_version;
-use super::ARROW_STREAM_CONTENT_TYPE;
 
 // Request structure for the remote clone table API
 #[derive(serde::Serialize)]
@@ -249,9 +249,9 @@ impl RemoteDatabase {
 #[cfg(all(test, feature = "remote"))]
 mod test_utils {
     use super::*;
+    use crate::remote::ClientConfig;
     use crate::remote::client::test_utils::MockSender;
     use crate::remote::client::test_utils::{client_with_handler, client_with_handler_and_config};
-    use crate::remote::ClientConfig;
 
     impl RemoteDatabase<MockSender> {
         pub fn new_mock<F, T>(handler: F) -> Self
@@ -464,6 +464,7 @@ impl<S: HttpSend> Database for RemoteDatabase<S> {
                             lance_read_params: None,
                             location: None,
                             namespace_client: None,
+                            managed_versioning: None,
                         };
                         let req = (callback)(req);
                         self.open_table(req).await
@@ -777,7 +778,12 @@ impl RemoteOptions {
 
 impl From<StorageOptions> for RemoteOptions {
     fn from(options: StorageOptions) -> Self {
-        let supported_opts = vec!["account_name", "azure_storage_account_name"];
+        let supported_opts = vec![
+            "account_name",
+            "azure_storage_account_name",
+            "azure_client_id",
+            "azure_tenant_id",
+        ];
         let mut filtered = HashMap::new();
         for opt in supported_opts {
             if let Some(v) = options.0.get(opt) {
@@ -799,9 +805,9 @@ mod tests {
 
     use crate::connection::ConnectBuilder;
     use crate::{
-        database::CreateTableMode,
-        remote::{ClientConfig, HeaderProvider, ARROW_STREAM_CONTENT_TYPE, JSON_CONTENT_TYPE},
         Connection, Error,
+        database::CreateTableMode,
+        remote::{ARROW_STREAM_CONTENT_TYPE, ClientConfig, HeaderProvider, JSON_CONTENT_TYPE},
     };
 
     #[test]

rust/lancedb/src/remote/retry.rs

@@ -1,8 +1,8 @@
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
-use crate::remote::RetryConfig;
 use crate::Error;
+use crate::remote::RetryConfig;
 use log::debug;
 use std::time::Duration;
 

rust/lancedb/src/remote/table.rs

@@ -6,15 +6,14 @@ pub mod insert;
 use self::insert::RemoteInsertExec;
 use crate::expr::expr_to_sql_string;
 
+use super::ARROW_STREAM_CONTENT_TYPE;
 use super::client::RequestResultExt;
 use super::client::{HttpSend, RestfulLanceDbClient, Sender};
 use super::db::ServerVersion;
-use super::ARROW_STREAM_CONTENT_TYPE;
-use crate::index::waiter::wait_for_index;
 use crate::index::Index;
 use crate::index::IndexStatistics;
+use crate::index::waiter::wait_for_index;
 use crate::query::{QueryFilter, QueryRequest, Select, VectorQueryRequest};
-use crate::table::query::create_multi_vector_plan;
 use crate::table::AddColumnsResult;
 use crate::table::AddResult;
 use crate::table::AlterColumnsResult;
@@ -23,19 +22,20 @@ use crate::table::DropColumnsResult;
 use crate::table::MergeResult;
 use crate::table::Tags;
 use crate::table::UpdateResult;
+use crate::table::query::create_multi_vector_plan;
 use crate::table::{AnyQuery, Filter, TableStatistics};
 use crate::utils::background_cache::BackgroundCache;
 use crate::utils::{supported_btree_data_type, supported_vector_data_type};
+use crate::{DistanceType, Error};
 use crate::{
     error::Result,
     index::{IndexBuilder, IndexConfig},
     query::QueryExecutionOptions,
     table::{
-        merge::MergeInsertBuilder, AddDataBuilder, BaseTable, OptimizeAction, OptimizeStats,
-        TableDefinition, UpdateBuilder,
+        AddDataBuilder, BaseTable, OptimizeAction, OptimizeStats, TableDefinition, UpdateBuilder,
+        merge::MergeInsertBuilder,
     },
 };
-use crate::{DistanceType, Error};
 use arrow_array::{RecordBatch, RecordBatchIterator, RecordBatchReader};
 use arrow_ipc::reader::FileReader;
 use arrow_schema::{DataType, SchemaRef};
@@ -50,7 +50,7 @@ use lance::arrow::json::{JsonDataType, JsonSchema};
 use lance::dataset::refs::TagContents;
 use lance::dataset::scanner::DatasetRecordBatchStream;
 use lance::dataset::{ColumnAlteration, NewColumnTransform, Version};
-use lance_datafusion::exec::{execute_plan, OneShotExec};
+use lance_datafusion::exec::{OneShotExec, execute_plan};
 use reqwest::{RequestBuilder, Response};
 use serde::{Deserialize, Serialize};
 use serde_json::Number;
@@ -612,8 +612,8 @@ impl<S: HttpSend> RemoteTable<S> {
                 message: format!(
                     "Cannot mutate table reference fixed at version {}. Call checkout_latest() to get a mutable table reference.",
                     version
-                )
-            })
+                ),
+            }),
         }
     }
 
@@ -697,10 +697,10 @@ impl<S: HttpSend> RemoteTable<S> {
             Error::Retry { status_code, .. } => *status_code,
             _ => None,
         };
-        if let Some(status_code) = status_code {
-            if Self::should_invalidate_cache_for_status(status_code) {
-                self.invalidate_schema_cache();
-            }
+        if let Some(status_code) = status_code
+            && Self::should_invalidate_cache_for_status(status_code)
+        {
+            self.invalidate_schema_cache();
         }
     }
 }
@@ -783,9 +783,9 @@ impl<S: HttpSend> std::fmt::Display for RemoteTable<S> {
 #[cfg(all(test, feature = "remote"))]
 mod test_utils {
     use super::*;
-    use crate::remote::client::test_utils::client_with_handler;
-    use crate::remote::client::test_utils::{client_with_handler_and_config, MockSender};
     use crate::remote::ClientConfig;
+    use crate::remote::client::test_utils::client_with_handler;
+    use crate::remote::client::test_utils::{MockSender, client_with_handler_and_config};
 
     impl RemoteTable<MockSender> {
         pub fn new_mock<F, T>(name: String, handler: F, version: Option<semver::Version>) -> Self
@@ -1227,7 +1227,10 @@ impl<S: HttpSend> BaseTable for RemoteTable<S> {
         let body = response.text().await.err_to_http(request_id.clone())?;
         if body.trim().is_empty() {
             // Backward compatible with old servers
-            return Ok(DeleteResult { version: 0 });
+            return Ok(DeleteResult {
+                num_deleted_rows: 0,
+                version: 0,
+            });
         }
         let delete_response: DeleteResult =
             serde_json::from_str(&body).map_err(|e| Error::Http {
@@ -1248,13 +1251,13 @@ impl<S: HttpSend> BaseTable for RemoteTable<S> {
             0 => {
                 return Err(Error::InvalidInput {
                     message: "No columns specified".into(),
-                })
+                });
             }
             1 => index.columns.pop().unwrap(),
             _ => {
                 return Err(Error::NotSupported {
                     message: "Indices over multiple columns not yet supported".into(),
-                })
+                });
             }
         };
         let mut body = serde_json::json!({
@@ -1273,73 +1276,24 @@ impl<S: HttpSend> BaseTable for RemoteTable<S> {
             );
         }
 
-        match index.index {
-            // TODO: Should we pass the actual index parameters? SaaS does not
-            // yet support them.
-            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::IvfSq(index) => {
-                body[INDEX_TYPE_KEY] = serde_json::Value::String("IVF_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::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::IvfRq(index) => {
-                body[INDEX_TYPE_KEY] = serde_json::Value::String("IVF_RQ".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::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();
-                }
-            }
+        fn to_json(params: &impl serde::Serialize) -> crate::Result<serde_json::Value> {
+            serde_json::to_value(params).map_err(|e| Error::InvalidInput {
+                message: format!("failed to serialize index params {:?}", e),
+            })
+        }
+
+        // Map each Index variant to its wire type name and serializable params.
+        // Auto is special-cased since it needs schema inspection.
+        let (index_type_str, params) = match &index.index {
+            Index::IvfFlat(p) => ("IVF_FLAT", Some(to_json(p)?)),
+            Index::IvfPq(p) => ("IVF_PQ", Some(to_json(p)?)),
+            Index::IvfSq(p) => ("IVF_SQ", Some(to_json(p)?)),
+            Index::IvfHnswSq(p) => ("IVF_HNSW_SQ", Some(to_json(p)?)),
+            Index::IvfRq(p) => ("IVF_RQ", Some(to_json(p)?)),
+            Index::BTree(p) => ("BTREE", Some(to_json(p)?)),
+            Index::Bitmap(p) => ("BITMAP", Some(to_json(p)?)),
+            Index::LabelList(p) => ("LABEL_LIST", Some(to_json(p)?)),
+            Index::FTS(p) => ("FTS", Some(to_json(p)?)),
             Index::Auto => {
                 let schema = self.schema().await?;
                 let field = schema
@@ -1348,11 +1302,11 @@ impl<S: HttpSend> BaseTable for RemoteTable<S> {
                         message: format!("Column {} not found in schema", column),
                     })?;
                 if supported_vector_data_type(field.data_type()) {
-                    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());
+                    ("IVF_PQ", None)
                 } else if supported_btree_data_type(field.data_type()) {
-                    body[INDEX_TYPE_KEY] = serde_json::Value::String("BTREE".to_string());
+                    ("BTREE", None)
                 } else {
                     return Err(Error::NotSupported {
                         message: format!(
@@ -1366,10 +1320,17 @@ impl<S: HttpSend> BaseTable for RemoteTable<S> {
             _ => {
                 return Err(Error::NotSupported {
                     message: "Index type not supported".into(),
-                })
+                });
             }
         };
 
+        body[INDEX_TYPE_KEY] = index_type_str.into();
+        if let Some(params) = params {
+            for (key, value) in params.as_object().expect("params should be a JSON object") {
+                body[key] = value.clone();
+            }
+        }
+
         let request = request.json(&body);
 
         let (request_id, response) = self.send(request, true).await?;
@@ -1810,8 +1771,8 @@ impl TryFrom<MergeInsertBuilder> for MergeInsertRequest {
 
 #[cfg(test)]
 mod tests {
-    use std::sync::atomic::{AtomicUsize, Ordering};
     use std::sync::Arc;
+    use std::sync::atomic::{AtomicUsize, Ordering};
     use std::time::Duration;
     use std::{collections::HashMap, pin::Pin};
 
@@ -1820,25 +1781,27 @@ mod tests {
     use crate::table::AddDataMode;
 
     use arrow::{array::AsArray, compute::concat_batches, datatypes::Int32Type};
-    use arrow_array::{record_batch, Int32Array, RecordBatch, RecordBatchIterator};
+    use arrow_array::{Int32Array, RecordBatch, RecordBatchIterator, record_batch};
     use arrow_schema::{DataType, Field, Schema};
     use chrono::{DateTime, Utc};
-    use futures::{future::BoxFuture, StreamExt, TryFutureExt};
+    use futures::{StreamExt, TryFutureExt, future::BoxFuture};
     use lance_index::scalar::inverted::query::MatchQuery;
     use lance_index::scalar::{FullTextSearchQuery, InvertedIndexParams};
     use reqwest::Body;
     use rstest::rstest;
     use serde_json::json;
 
-    use crate::index::vector::{IvfFlatIndexBuilder, IvfHnswSqIndexBuilder};
-    use crate::remote::db::DEFAULT_SERVER_VERSION;
+    use crate::index::vector::{
+        IvfFlatIndexBuilder, IvfHnswSqIndexBuilder, IvfRqIndexBuilder, IvfSqIndexBuilder,
+    };
     use crate::remote::JSON_CONTENT_TYPE;
+    use crate::remote::db::DEFAULT_SERVER_VERSION;
     use crate::utils::background_cache::clock;
     use crate::{
-        index::{vector::IvfPqIndexBuilder, Index, IndexStatistics, IndexType},
+        DistanceType, Error, Table,
+        index::{Index, IndexStatistics, IndexType, vector::IvfPqIndexBuilder},
         query::{ExecutableQuery, QueryBase},
         remote::ARROW_FILE_CONTENT_TYPE,
-        DistanceType, Error, Table,
     };
 
     #[tokio::test]
@@ -2067,11 +2030,13 @@ mod tests {
                     .unwrap(),
                 "/v1/table/my_table/insert/" => {
                     assert_eq!(request.method(), "POST");
-                    assert!(request
-                        .url()
-                        .query_pairs()
-                        .filter(|(k, _)| k == "mode")
-                        .all(|(_, v)| v == "append"));
+                    assert!(
+                        request
+                            .url()
+                            .query_pairs()
+                            .filter(|(k, _)| k == "mode")
+                            .all(|(_, v)| v == "append")
+                    );
                     assert_eq!(
                         request.headers().get("Content-Type").unwrap(),
                         ARROW_STREAM_CONTENT_TYPE
@@ -2992,6 +2957,8 @@ mod tests {
                 "IVF_FLAT",
                 json!({
                     "metric_type": "hamming",
+                    "sample_rate": 256,
+                    "max_iterations": 50,
                 }),
                 Index::IvfFlat(IvfFlatIndexBuilder::default().distance_type(DistanceType::Hamming)),
             ),
@@ -3000,6 +2967,8 @@ mod tests {
                 json!({
                     "metric_type": "hamming",
                     "num_partitions": 128,
+                    "sample_rate": 256,
+                    "max_iterations": 50,
                 }),
                 Index::IvfFlat(
                     IvfFlatIndexBuilder::default()
@@ -3011,6 +2980,8 @@ mod tests {
                 "IVF_PQ",
                 json!({
                     "metric_type": "l2",
+                    "sample_rate": 256,
+                    "max_iterations": 50,
                 }),
                 Index::IvfPq(Default::default()),
             ),
@@ -3020,6 +2991,8 @@ mod tests {
                     "metric_type": "cosine",
                     "num_partitions": 128,
                     "num_bits": 4,
+                    "sample_rate": 256,
+                    "max_iterations": 50,
                 }),
                 Index::IvfPq(
                     IvfPqIndexBuilder::default()
@@ -3028,10 +3001,29 @@ mod tests {
                         .num_bits(4),
                 ),
             ),
+            (
+                "IVF_PQ",
+                json!({
+                    "metric_type": "l2",
+                    "num_sub_vectors": 16,
+                    "sample_rate": 512,
+                    "max_iterations": 100,
+                }),
+                Index::IvfPq(
+                    IvfPqIndexBuilder::default()
+                        .num_sub_vectors(16)
+                        .sample_rate(512)
+                        .max_iterations(100),
+                ),
+            ),
             (
                 "IVF_HNSW_SQ",
                 json!({
                     "metric_type": "l2",
+                    "sample_rate": 256,
+                    "max_iterations": 50,
+                    "m": 20,
+                    "ef_construction": 300,
                 }),
                 Index::IvfHnswSq(Default::default()),
             ),
@@ -3040,11 +3032,65 @@ mod tests {
                 json!({
                     "metric_type": "l2",
                     "num_partitions": 128,
+                    "sample_rate": 256,
+                    "max_iterations": 50,
+                    "m": 40,
+                    "ef_construction": 500,
                 }),
                 Index::IvfHnswSq(
                     IvfHnswSqIndexBuilder::default()
                         .distance_type(DistanceType::L2)
-                        .num_partitions(128),
+                        .num_partitions(128)
+                        .num_edges(40)
+                        .ef_construction(500),
+                ),
+            ),
+            (
+                "IVF_SQ",
+                json!({
+                    "metric_type": "l2",
+                    "sample_rate": 256,
+                    "max_iterations": 50,
+                }),
+                Index::IvfSq(Default::default()),
+            ),
+            (
+                "IVF_SQ",
+                json!({
+                    "metric_type": "cosine",
+                    "num_partitions": 64,
+                    "sample_rate": 256,
+                    "max_iterations": 50,
+                }),
+                Index::IvfSq(
+                    IvfSqIndexBuilder::default()
+                        .distance_type(DistanceType::Cosine)
+                        .num_partitions(64),
+                ),
+            ),
+            (
+                "IVF_RQ",
+                json!({
+                    "metric_type": "l2",
+                    "sample_rate": 256,
+                    "max_iterations": 50,
+                }),
+                Index::IvfRq(Default::default()),
+            ),
+            (
+                "IVF_RQ",
+                json!({
+                    "metric_type": "cosine",
+                    "num_partitions": 64,
+                    "num_bits": 8,
+                    "sample_rate": 256,
+                    "max_iterations": 50,
+                }),
+                Index::IvfRq(
+                    IvfRqIndexBuilder::default()
+                        .distance_type(DistanceType::Cosine)
+                        .num_partitions(64)
+                        .num_bits(8),
                 ),
             ),
             // HNSW_PQ isn't yet supported on SaaS
@@ -3548,7 +3594,7 @@ mod tests {
     }
 
     fn _make_table_with_indices(unindexed_rows: usize) -> Table {
-        let table = Table::new_with_handler("my_table", move |request| {
+        Table::new_with_handler("my_table", move |request| {
             assert_eq!(request.method(), "POST");
 
             let response_body = match request.url().path() {
@@ -3592,8 +3638,7 @@ mod tests {
             let body = serde_json::to_string(&response_body).unwrap();
             let status = if body == "null" { 404 } else { 200 };
             http::Response::builder().status(status).body(body).unwrap()
-        });
-        table
+        })
     }
 
     #[tokio::test]
@@ -3804,8 +3849,8 @@ mod tests {
 
     #[tokio::test]
     async fn test_uri_caching() {
-        use std::sync::atomic::{AtomicUsize, Ordering};
         use std::sync::Arc;
+        use std::sync::atomic::{AtomicUsize, Ordering};
 
         let call_count = Arc::new(AtomicUsize::new(0));
         let call_count_clone = call_count.clone();

rust/lancedb/src/remote/table/insert.rs

@@ -16,12 +16,12 @@ use datafusion_physical_plan::{DisplayAs, DisplayFormatType, ExecutionPlan, Plan
 use futures::StreamExt;
 use http::header::CONTENT_TYPE;
 
+use crate::Error;
+use crate::remote::ARROW_STREAM_CONTENT_TYPE;
 use crate::remote::client::{HttpSend, RestfulLanceDbClient, Sender};
 use crate::remote::table::RemoteTable;
-use crate::remote::ARROW_STREAM_CONTENT_TYPE;
-use crate::table::datafusion::insert::COUNT_SCHEMA;
 use crate::table::AddResult;
-use crate::Error;
+use crate::table::datafusion::insert::COUNT_SCHEMA;
 
 /// ExecutionPlan for inserting data into a remote LanceDB table.
 ///
@@ -309,12 +309,12 @@ mod tests {
     use arrow_schema::{DataType, Field, Schema as ArrowSchema};
     use datafusion::prelude::SessionContext;
     use datafusion_catalog::MemTable;
-    use std::sync::atomic::{AtomicUsize, Ordering};
     use std::sync::Arc;
+    use std::sync::atomic::{AtomicUsize, Ordering};
 
+    use crate::Table;
     use crate::remote::ARROW_STREAM_CONTENT_TYPE;
     use crate::table::datafusion::BaseTableAdapter;
-    use crate::Table;
 
     fn schema_json() -> &'static str {
         r#"{"fields": [{"name": "id", "type": {"type": "int32"}, "nullable": true}]}"#

rust/lancedb/src/remote/util.rs

@@ -5,7 +5,7 @@ use arrow_ipc::CompressionType;
 use futures::{Stream, StreamExt};
 use reqwest::Response;
 
-use crate::{arrow::SendableRecordBatchStream, Result};
+use crate::{Result, arrow::SendableRecordBatchStream};
 
 use super::db::ServerVersion;
 

rust/lancedb/src/rerankers/rrf.rs

@@ -14,7 +14,7 @@ use async_trait::async_trait;
 use lance::dataset::ROW_ID;
 
 use crate::error::{Error, Result};
-use crate::rerankers::{Reranker, RELEVANCE_SCORE};
+use crate::rerankers::{RELEVANCE_SCORE, Reranker};
 
 /// Reranks the results using Reciprocal Rank Fusion(RRF) algorithm based
 /// on the scores of vector and FTS search.

rust/lancedb/src/table.rs

@@ -8,54 +8,58 @@ use arrow_schema::{DataType, Field, Schema, SchemaRef};
 use async_trait::async_trait;
 use datafusion_execution::TaskContext;
 use datafusion_expr::Expr;
-use datafusion_physical_plan::display::DisplayableExecutionPlan;
 use datafusion_physical_plan::ExecutionPlan;
-use futures::stream::FuturesUnordered;
+use datafusion_physical_plan::display::DisplayableExecutionPlan;
 use futures::StreamExt;
-use lance::dataset::builder::DatasetBuilder;
+use futures::stream::FuturesUnordered;
 pub use lance::dataset::ColumnAlteration;
 pub use lance::dataset::NewColumnTransform;
 pub use lance::dataset::ReadParams;
 pub use lance::dataset::Version;
 use lance::dataset::WriteMode;
+use lance::dataset::builder::DatasetBuilder;
 use lance::dataset::{InsertBuilder, WriteParams};
-use lance::index::vector::utils::infer_vector_dim;
 use lance::index::vector::VectorIndexParams;
+use lance::index::vector::utils::infer_vector_dim;
 use lance::io::{ObjectStoreParams, WrappingObjectStore};
 use lance_datafusion::utils::StreamingWriteSource;
+use lance_index::DatasetIndexExt;
+use lance_index::IndexType;
 use lance_index::scalar::{BuiltinIndexType, ScalarIndexParams};
 use lance_index::vector::bq::RQBuildParams;
 use lance_index::vector::hnsw::builder::HnswBuildParams;
 use lance_index::vector::ivf::IvfBuildParams;
 use lance_index::vector::pq::PQBuildParams;
 use lance_index::vector::sq::builder::SQBuildParams;
-use lance_index::DatasetIndexExt;
-use lance_index::IndexType;
 use lance_io::object_store::{LanceNamespaceStorageOptionsProvider, StorageOptionsAccessor};
 pub use query::AnyQuery;
 
+use lance::io::commit::namespace_manifest::LanceNamespaceExternalManifestStore;
 use lance_namespace::LanceNamespace;
+use lance_namespace::models::DescribeTableRequest;
 use lance_table::format::Manifest;
+use lance_table::io::commit::CommitHandler;
 use lance_table::io::commit::ManifestNamingScheme;
+use lance_table::io::commit::external_manifest::ExternalManifestCommitHandler;
 use serde::{Deserialize, Serialize};
 use std::collections::HashMap;
 use std::format;
 use std::path::Path;
 use std::sync::Arc;
 
-use crate::data::scannable::{estimate_write_partitions, PeekedScannable, Scannable};
+use crate::data::scannable::{PeekedScannable, Scannable, estimate_write_partitions};
 use crate::database::Database;
 use crate::embeddings::{EmbeddingDefinition, EmbeddingRegistry, MemoryRegistry};
 use crate::error::{Error, Result};
-use crate::index::vector::VectorIndex;
 use crate::index::IndexStatistics;
-use crate::index::{vector::suggested_num_sub_vectors, Index, IndexBuilder};
+use crate::index::vector::VectorIndex;
+use crate::index::{Index, IndexBuilder, vector::suggested_num_sub_vectors};
 use crate::index::{IndexConfig, IndexStatisticsImpl};
 use crate::query::{IntoQueryVector, Query, QueryExecutionOptions, TakeQuery, VectorQuery};
 use crate::table::datafusion::insert::InsertExec;
 use crate::utils::{
-    supported_bitmap_data_type, supported_btree_data_type, supported_fts_data_type,
-    supported_label_list_data_type, supported_vector_data_type, PatchReadParam, PatchWriteParam,
+    PatchReadParam, PatchWriteParam, supported_bitmap_data_type, supported_btree_data_type,
+    supported_fts_data_type, supported_label_list_data_type, supported_vector_data_type,
 };
 
 use self::dataset::DatasetConsistencyWrapper;
@@ -1212,10 +1216,13 @@ pub struct NativeTable {
     // This comes from the connection options. We store here so we can pass down
     // to the dataset when we recreate it (for example, in checkout_latest).
     read_consistency_interval: Option<std::time::Duration>,
-    // Optional namespace client for server-side query execution.
-    // When set, queries will be executed on the namespace server instead of locally.
-    // pub (crate) namespace_client so query.rs can access the fields
+    // Optional namespace client for namespace operations (e.g., managed versioning).
+    // pub(crate) so query.rs can access the field for server-side query execution.
     pub(crate) namespace_client: Option<Arc<dyn LanceNamespace>>,
+    // Whether to enable server-side query execution via the namespace client.
+    // When true and namespace_client is set, queries will be executed on the
+    // namespace server instead of locally.
+    pub(crate) server_side_query_enabled: bool,
 }
 
 impl std::fmt::Debug for NativeTable {
@@ -1227,6 +1234,7 @@ impl std::fmt::Debug for NativeTable {
             .field("uri", &self.uri)
             .field("read_consistency_interval", &self.read_consistency_interval)
             .field("namespace_client", &self.namespace_client)
+            .field("server_side_query_enabled", &self.server_side_query_enabled)
             .finish()
     }
 }
@@ -1263,7 +1271,7 @@ impl NativeTable {
     /// * A [NativeTable] object.
     pub async fn open(uri: &str) -> Result<Self> {
         let name = Self::get_table_name(uri)?;
-        Self::open_with_params(uri, &name, vec![], None, None, None, None).await
+        Self::open_with_params(uri, &name, vec![], None, None, None, None, false, None).await
     }
 
     /// Opens an existing Table
@@ -1273,7 +1281,10 @@ impl NativeTable {
     /// * `base_path` - The base path where the table is located
     /// * `name` The Table name
     /// * `params` The [ReadParams] to use when opening the table
-    /// * `namespace_client` - Optional namespace client for server-side query execution
+    /// * `namespace_client` - Optional namespace client for namespace operations
+    /// * `server_side_query_enabled` - Whether to enable server-side query execution
+    /// * `managed_versioning` - Whether managed versioning is enabled. If None and namespace_client
+    ///   is provided, the value will be fetched via describe_table.
     ///
     /// # Returns
     ///
@@ -1287,6 +1298,8 @@ impl NativeTable {
         params: Option<ReadParams>,
         read_consistency_interval: Option<std::time::Duration>,
         namespace_client: Option<Arc<dyn LanceNamespace>>,
+        server_side_query_enabled: bool,
+        managed_versioning: Option<bool>,
     ) -> Result<Self> {
         let params = params.unwrap_or_default();
         // patch the params if we have a write store wrapper
@@ -1295,17 +1308,54 @@ impl NativeTable {
             None => params,
         };
 
-        let dataset = DatasetBuilder::from_uri(uri)
-            .with_read_params(params)
-            .load()
-            .await
-            .map_err(|e| match e {
-                lance::Error::DatasetNotFound { .. } => Error::TableNotFound {
-                    name: name.to_string(),
-                    source: Box::new(e),
-                },
-                e => e.into(),
-            })?;
+        // Build table_id from namespace + name
+        let mut table_id = namespace.clone();
+        table_id.push(name.to_string());
+
+        // Determine if managed_versioning is enabled
+        // Use the provided value if available, otherwise query the namespace
+        let managed_versioning = match managed_versioning {
+            Some(value) => value,
+            None if namespace_client.is_some() => {
+                let ns_client = namespace_client.as_ref().unwrap();
+                let describe_request = DescribeTableRequest {
+                    id: Some(table_id.clone()),
+                    ..Default::default()
+                };
+                let response = ns_client
+                    .describe_table(describe_request)
+                    .await
+                    .map_err(|e| Error::Runtime {
+                        message: format!(
+                            "Failed to describe table via namespace client: {}. \
+                             If you don't need managed versioning, don't pass namespace_client.",
+                            e
+                        ),
+                    })?;
+                response.managed_versioning == Some(true)
+            }
+            None => false,
+        };
+
+        let mut builder = DatasetBuilder::from_uri(uri).with_read_params(params);
+
+        // Set up commit handler when managed_versioning is enabled
+        if managed_versioning && let Some(ref ns_client) = namespace_client {
+            let external_store =
+                LanceNamespaceExternalManifestStore::new(ns_client.clone(), table_id.clone());
+            let commit_handler: Arc<dyn CommitHandler> = Arc::new(ExternalManifestCommitHandler {
+                external_manifest_store: Arc::new(external_store),
+            });
+            builder = builder.with_commit_handler(commit_handler);
+        }
+
+        let dataset = builder.load().await.map_err(|e| match e {
+            lance::Error::DatasetNotFound { .. } => Error::TableNotFound {
+                name: name.to_string(),
+                source: Box::new(e),
+            },
+            e => e.into(),
+        })?;
 
         let dataset = DatasetConsistencyWrapper::new_latest(dataset, read_consistency_interval);
         let id = Self::build_id(&namespace, name);
@@ -1318,6 +1368,7 @@ impl NativeTable {
             dataset,
             read_consistency_interval,
             namespace_client,
+            server_side_query_enabled,
         })
     }
 
@@ -1421,6 +1472,7 @@ impl NativeTable {
             dataset,
             read_consistency_interval,
             namespace_client: stored_namespace_client,
+            server_side_query_enabled,
         })
     }
 
@@ -1460,7 +1512,8 @@ impl NativeTable {
     /// * `namespace` - The namespace path. When non-empty, an explicit URI must be provided.
     /// * `batches` RecordBatch to be saved in the database.
     /// * `params` - Write parameters.
-    /// * `namespace_client` - Optional namespace client for server-side query execution
+    /// * `namespace_client` - Optional namespace client for namespace operations
+    /// * `server_side_query_enabled` - Whether to enable server-side query execution
     ///
     /// # Returns
     ///
@@ -1475,6 +1528,7 @@ impl NativeTable {
         params: Option<WriteParams>,
         read_consistency_interval: Option<std::time::Duration>,
         namespace_client: Option<Arc<dyn LanceNamespace>>,
+        server_side_query_enabled: bool,
     ) -> Result<Self> {
         // Default params uses format v1.
         let params = params.unwrap_or(WriteParams {
@@ -1507,6 +1561,7 @@ impl NativeTable {
             dataset: DatasetConsistencyWrapper::new_latest(dataset, read_consistency_interval),
             read_consistency_interval,
             namespace_client,
+            server_side_query_enabled,
         })
     }
 
@@ -1520,6 +1575,7 @@ impl NativeTable {
         params: Option<WriteParams>,
         read_consistency_interval: Option<std::time::Duration>,
         namespace_client: Option<Arc<dyn LanceNamespace>>,
+        server_side_query_enabled: bool,
     ) -> Result<Self> {
         let data: Box<dyn Scannable> = Box::new(RecordBatch::new_empty(schema));
         Self::create(
@@ -1531,6 +1587,7 @@ impl NativeTable {
             params,
             read_consistency_interval,
             namespace_client,
+            server_side_query_enabled,
         )
         .await
     }
@@ -1634,6 +1691,7 @@ impl NativeTable {
             dataset: DatasetConsistencyWrapper::new_latest(dataset, read_consistency_interval),
             read_consistency_interval,
             namespace_client: stored_namespace_client,
+            server_side_query_enabled,
         })
     }
 
@@ -2555,22 +2613,21 @@ pub struct FragmentSummaryStats {
 #[cfg(test)]
 #[allow(deprecated)]
 mod tests {
-    use std::sync::atomic::{AtomicBool, Ordering};
     use std::sync::Arc;
+    use std::sync::atomic::{AtomicBool, Ordering};
     use std::time::Duration;
 
     use arrow_array::{
-        builder::{ListBuilder, StringBuilder},
         Array, BooleanArray, FixedSizeListArray, Int32Array, LargeStringArray, RecordBatch,
         RecordBatchIterator, RecordBatchReader, StringArray,
+        builder::{ListBuilder, StringBuilder},
     };
     use arrow_array::{BinaryArray, LargeBinaryArray};
     use arrow_data::ArrayDataBuilder;
     use arrow_schema::{DataType, Field, Schema};
     use futures::TryStreamExt;
-    use lance::io::{ObjectStoreParams, WrappingObjectStore};
     use lance::Dataset;
-    use rand::Rng;
+    use lance::io::{ObjectStoreParams, WrappingObjectStore};
     use tempfile::tempdir;
 
     use super::*;
@@ -2626,7 +2683,7 @@ mod tests {
             vec![Ok(batch.clone())],
             batch.schema(),
         ));
-        let table = NativeTable::create(uri, "test", vec![], reader, None, None, None, None)
+        let table = NativeTable::create(uri, "test", vec![], reader, None, None, None, None, false)
             .await
             .unwrap();
 
@@ -2777,9 +2834,8 @@ mod tests {
             false,
         )]));
 
-        let mut rng = rand::thread_rng();
         let float_arr = Float32Array::from(
-            repeat_with(|| rng.gen::<f32>())
+            repeat_with(rand::random::<f32>)
                 .take(512 * dimension as usize)
                 .collect::<Vec<f32>>(),
         );
@@ -2884,8 +2940,8 @@ mod tests {
             .await
             .unwrap();
 
-        use lance::index::vector::ivf::v2::IvfPq as LanceIvfPq;
         use lance::index::DatasetIndexInternalExt;
+        use lance::index::vector::ivf::v2::IvfPq as LanceIvfPq;
         use lance_index::metrics::NoOpMetricsCollector;
         use lance_index::vector::VectorIndex as LanceVectorIndex;
 
@@ -2933,9 +2989,8 @@ mod tests {
             false,
         )]));
 
-        let mut rng = rand::thread_rng();
         let float_arr = Float32Array::from(
-            repeat_with(|| rng.gen::<f32>())
+            repeat_with(rand::random::<f32>)
                 .take(512 * dimension as usize)
                 .collect::<Vec<f32>>(),
         );
@@ -2993,9 +3048,8 @@ mod tests {
             false,
         )]));
 
-        let mut rng = rand::thread_rng();
         let float_arr = Float32Array::from(
-            repeat_with(|| rng.gen::<f32>())
+            repeat_with(rand::random::<f32>)
                 .take(512 * dimension as usize)
                 .collect::<Vec<f32>>(),
         );
@@ -3256,16 +3310,20 @@ mod tests {
             .unwrap();
 
         // Can not create btree or bitmap index on list column
-        assert!(table
-            .create_index(&["tags"], Index::BTree(Default::default()))
-            .execute()
-            .await
-            .is_err());
-        assert!(table
-            .create_index(&["tags"], Index::Bitmap(Default::default()))
-            .execute()
-            .await
-            .is_err());
+        assert!(
+            table
+                .create_index(&["tags"], Index::BTree(Default::default()))
+                .execute()
+                .await
+                .is_err()
+        );
+        assert!(
+            table
+                .create_index(&["tags"], Index::Bitmap(Default::default()))
+                .execute()
+                .await
+                .is_err()
+        );
 
         // Create bitmap index on the "category" column
         table

rust/lancedb/src/table/add_data.rs

@@ -7,8 +7,8 @@ use arrow_schema::{DataType, Fields, Schema};
 use lance::dataset::WriteMode;
 use serde::{Deserialize, Serialize};
 
-use crate::data::scannable::scannable_with_embeddings;
 use crate::data::scannable::Scannable;
+use crate::data::scannable::scannable_with_embeddings;
 use crate::embeddings::EmbeddingRegistry;
 use crate::table::datafusion::cast::cast_to_table_schema;
 use crate::table::datafusion::reject_nan::reject_nan_vectors;
@@ -204,13 +204,14 @@ mod tests {
 
     use arrow::datatypes::Float64Type;
     use arrow_array::{
-        record_batch, FixedSizeListArray, Float32Array, Int32Array, LargeStringArray, ListArray,
-        RecordBatch, RecordBatchIterator,
+        FixedSizeListArray, Float32Array, Int32Array, LargeStringArray, ListArray, RecordBatch,
+        RecordBatchIterator, record_batch,
     };
     use arrow_schema::{ArrowError, DataType, Field, Schema};
     use futures::TryStreamExt;
     use lance::dataset::{WriteMode, WriteParams};
 
+    use crate::Error;
     use crate::arrow::{SendableRecordBatchStream, SimpleRecordBatchStream};
     use crate::connect;
     use crate::data::scannable::Scannable;
@@ -220,9 +221,8 @@ mod tests {
     use crate::query::{ExecutableQuery, QueryBase, Select};
     use crate::table::add_data::NaNVectorBehavior;
     use crate::table::{ColumnDefinition, ColumnKind, Table, TableDefinition, WriteOptions};
-    use crate::test_utils::embeddings::MockEmbed;
     use crate::test_utils::TestCustomError;
-    use crate::Error;
+    use crate::test_utils::embeddings::MockEmbed;
 
     use super::AddDataMode;
 

rust/lancedb/src/table/datafusion.rs

@@ -17,17 +17,17 @@ use async_trait::async_trait;
 use datafusion_catalog::{Session, TableProvider};
 use datafusion_common::{DataFusionError, Result as DataFusionResult, Statistics};
 use datafusion_execution::{SendableRecordBatchStream, TaskContext};
-use datafusion_expr::{dml::InsertOp, Expr, TableProviderFilterPushDown, TableType};
+use datafusion_expr::{Expr, TableProviderFilterPushDown, TableType, dml::InsertOp};
 use datafusion_physical_plan::{
-    stream::RecordBatchStreamAdapter, DisplayAs, DisplayFormatType, ExecutionPlan, PlanProperties,
+    DisplayAs, DisplayFormatType, ExecutionPlan, PlanProperties, stream::RecordBatchStreamAdapter,
 };
 use futures::{TryFutureExt, TryStreamExt};
 use lance::dataset::{WriteMode, WriteParams};
 
 use super::{AnyQuery, BaseTable};
 use crate::{
-    query::{QueryExecutionOptions, QueryFilter, QueryRequest, Select},
     Result,
+    query::{QueryExecutionOptions, QueryFilter, QueryRequest, Select},
 };
 use arrow_schema::{DataType, Field};
 use lance_index::scalar::FullTextSearchQuery;
@@ -268,7 +268,7 @@ impl TableProvider for BaseTableAdapter {
             InsertOp::Replace => {
                 return Err(DataFusionError::NotImplemented(
                     "Replace mode is not supported for LanceDB tables".to_string(),
-                ))
+                ));
             }
         };
 
@@ -300,13 +300,13 @@ pub mod tests {
     use datafusion_catalog::TableProvider;
     use datafusion_common::stats::Precision;
     use datafusion_execution::SendableRecordBatchStream;
-    use datafusion_expr::{col, lit, LogicalPlan, LogicalPlanBuilder};
+    use datafusion_expr::{LogicalPlan, LogicalPlanBuilder, col, lit};
     use futures::{StreamExt, TryStreamExt};
     use tempfile::tempdir;
 
     use crate::{
         connect,
-        index::{scalar::BTreeIndexBuilder, Index},
+        index::{Index, scalar::BTreeIndexBuilder},
         table::datafusion::BaseTableAdapter,
     };