Bump version: 0.25.2 → 0.25.3-beta.0

I'm working on a lancedb version of pytorch data loading (and hopefully
addressing https://github.com/lancedb/lance/issues/3727).

However, rather than rely on pytorch for everything I'm moving some of
the things that pytorch does into rust. This gives us more control over
data loading (e.g. using shards or a hash-based split) and it allows
permutations to be persistent. In particular I hope to be able to:

* Create a persistent permutation
* This permutation can handle splits, filtering, shuffling, and sharding
* Create a rust data loader that can read a permutation (one or more
splits), or a subset of a permutation (for DDP)
* Create a python data loader that delegates to the rust data loader

Eventually create integrations for other data loading libraries,
including rust & node

.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.22.1-beta.0"
+current_version = "0.22.2"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

.github/actions/create-failure-issue/action.yml

@@ -0,0 +1,45 @@
+name: Create Failure Issue
+description: Creates a GitHub issue if any jobs in the workflow failed
+
+inputs:
+  job-results:
+    description: 'JSON string of job results from needs context'
+    required: true
+  workflow-name:
+    description: 'Name of the workflow'
+    required: true
+
+runs:
+  using: composite
+  steps:
+    - name: Check for failures and create issue
+      shell: bash
+      env:
+        JOB_RESULTS: ${{ inputs.job-results }}
+        WORKFLOW_NAME: ${{ inputs.workflow-name }}
+        RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+        GH_TOKEN: ${{ github.token }}
+      run: |
+        # Check if any job failed
+        if echo "$JOB_RESULTS" | jq -e 'to_entries | any(.value.result == "failure")' > /dev/null; then
+          echo "Detected job failures, creating issue..."
+
+          # Extract failed job names
+          FAILED_JOBS=$(echo "$JOB_RESULTS" | jq -r 'to_entries | map(select(.value.result == "failure")) | map(.key) | join(", ")')
+
+          # Create issue with workflow name, failed jobs, and run URL
+          gh issue create \
+            --title "$WORKFLOW_NAME Failed ($FAILED_JOBS)" \
+            --body "The workflow **$WORKFLOW_NAME** failed during execution.
+
+        **Failed jobs:** $FAILED_JOBS
+
+        **Run URL:** $RUN_URL
+
+        Please investigate the failed jobs and address any issues." \
+            --label "ci"
+
+          echo "Issue created successfully"
+        else
+          echo "No job failures detected, skipping issue creation"
+        fi

.github/workflows/cargo-publish.yml

@@ -38,3 +38,17 @@ jobs:
       - name: Publish the package
         run: |
           cargo publish -p lancedb --all-features --token ${{ steps.auth.outputs.token }}
+  report-failure:
+    name: Report Workflow Failure
+    runs-on: ubuntu-latest
+    needs: [build]
+    if: always() && (github.event_name == 'release' || github.event_name == 'workflow_dispatch')
+    permissions:
+      contents: read
+      issues: write
+    steps:
+      - uses: actions/checkout@v4
+      - uses: ./.github/actions/create-failure-issue
+        with:
+          job-results: ${{ toJSON(needs) }}
+          workflow-name: ${{ github.workflow }}

.github/workflows/docs.yml

@@ -56,8 +56,9 @@ jobs:
         with:
           node-version: 20
           cache: 'npm'
+          cache-dependency-path: docs/package-lock.json
       - name: Install node dependencies
-        working-directory: node
+        working-directory: nodejs
         run: |
           sudo apt update
           sudo apt install -y protobuf-compiler libssl-dev

.github/workflows/docs_test.yml

@@ -24,7 +24,8 @@ env:
 jobs:
   test-python:
     name: Test doc python code
-    runs-on: ubuntu-24.04
+    runs-on: warp-ubuntu-2204-x64-8x
+    timeout-minutes: 60
     steps:
     - name: Checkout
       uses: actions/checkout@v4
@@ -48,7 +49,6 @@ jobs:
       uses: swatinem/rust-cache@v2
     - name: Build Python
       working-directory: docs/test
-      timeout-minutes: 60
       run:
         python -m pip install --extra-index-url https://pypi.fury.io/lancedb/ -r requirements.txt
     - name: Create test files

.github/workflows/java-publish.yml

@@ -43,7 +43,6 @@ jobs:
       - uses: Swatinem/rust-cache@v2
       - uses: actions-rust-lang/setup-rust-toolchain@v1
         with:
-          toolchain: "1.81.0"
           cache-workspaces: "./java/core/lancedb-jni"
           # Disable full debug symbol generation to speed up CI build and keep memory down
           # "1" means line tables only, which is useful for panic tracebacks.
@@ -112,3 +111,17 @@ jobs:
         env:
           SONATYPE_USER: ${{ secrets.SONATYPE_USER }}
           SONATYPE_TOKEN: ${{ secrets.SONATYPE_TOKEN }}
+  report-failure:
+    name: Report Workflow Failure
+    runs-on: ubuntu-latest
+    needs: [linux-arm64, linux-x86, macos-arm64]
+    if: always() && (github.event_name == 'release' || github.event_name == 'workflow_dispatch')
+    permissions:
+      contents: read
+      issues: write
+    steps:
+      - uses: actions/checkout@v4
+      - uses: ./.github/actions/create-failure-issue
+        with:
+          job-results: ${{ toJSON(needs) }}
+          workflow-name: ${{ github.workflow }}

.github/workflows/nodejs.yml

@@ -6,6 +6,7 @@ on:
       - main
   pull_request:
     paths:
+      - Cargo.toml
       - nodejs/**
       - .github/workflows/nodejs.yml
       - docker-compose.yml
@@ -116,7 +117,7 @@ jobs:
         set -e
         npm ci
         npm run docs
-        if ! git diff --exit-code -- . ':(exclude)Cargo.lock'; then
+        if ! git diff --exit-code -- ../ ':(exclude)Cargo.lock'; then
           echo "Docs need to be updated"
           echo "Run 'npm run docs', fix any warnings, and commit the changes."
           exit 1

.github/workflows/npm-publish.yml

@@ -365,3 +365,17 @@ jobs:
             ARGS="$ARGS --tag preview"
           fi
           npm publish $ARGS
+  report-failure:
+    name: Report Workflow Failure
+    runs-on: ubuntu-latest
+    needs: [build-lancedb, test-lancedb, publish]
+    if: always() && (github.event_name == 'release' || github.event_name == 'workflow_dispatch')
+    permissions:
+      contents: read
+      issues: write
+    steps:
+      - uses: actions/checkout@v4
+      - uses: ./.github/actions/create-failure-issue
+        with:
+          job-results: ${{ toJSON(needs) }}
+          workflow-name: ${{ github.workflow }}

.github/workflows/pypi-publish.yml

@@ -56,7 +56,7 @@ jobs:
           pypi_token: ${{ secrets.LANCEDB_PYPI_API_TOKEN }}
           fury_token: ${{ secrets.FURY_TOKEN }}
   mac:
-    timeout-minutes: 60
+    timeout-minutes: 90
     runs-on: ${{ matrix.config.runner }}
     strategy:
       matrix:
@@ -64,7 +64,7 @@ jobs:
           - target: x86_64-apple-darwin
             runner: macos-13
           - target: aarch64-apple-darwin
-            runner: macos-14
+            runner: warp-macos-14-arm64-6x
     env:
       MACOSX_DEPLOYMENT_TARGET: 10.15
     steps:
@@ -173,3 +173,17 @@ jobs:
           generate_release_notes: false
           name: Python LanceDB v${{ steps.extract_version.outputs.version }}
           body: ${{ steps.python_release_notes.outputs.changelog }}
+  report-failure:
+    name: Report Workflow Failure
+    runs-on: ubuntu-latest
+    needs: [linux, mac, windows]
+    permissions:
+      contents: read
+      issues: write
+    if: always() && (github.event_name == 'release' || github.event_name == 'workflow_dispatch')
+    steps:
+      - uses: actions/checkout@v4
+      - uses: ./.github/actions/create-failure-issue
+        with:
+          job-results: ${{ toJSON(needs) }}
+          workflow-name: ${{ github.workflow }}

.github/workflows/python.yml

@@ -6,6 +6,7 @@ on:
       - main
   pull_request:
     paths:
+      - Cargo.toml
       - python/**
       - .github/workflows/python.yml
 

.github/workflows/rust.yml

@@ -96,6 +96,7 @@ jobs:
       # Need up-to-date compilers for kernels
       CC: clang-18
       CXX: clang++-18
+      GH_TOKEN: ${{ secrets.SOPHON_READ_TOKEN }}
     steps:
       - uses: actions/checkout@v4
         with:
@@ -117,15 +118,17 @@ jobs:
           sudo chmod 600 /swapfile
           sudo mkswap /swapfile
           sudo swapon /swapfile
-      - name: Start S3 integration test environment
-        working-directory: .
-        run: docker compose up --detach --wait
       - name: Build
         run: cargo build --all-features --tests --locked --examples
-      - name: Run tests
+      - name: Run feature tests
-        run: cargo test --all-features --locked
+        run: make -C ./lancedb feature-tests
       - name: Run examples
         run: cargo run --example simple --locked
+      - name: Run remote tests
+        # Running this requires access to secrets, so skip if this is
+        # a PR from a fork.
+        if: github.event_name != 'pull_request' || !github.event.pull_request.head.repo.fork
+        run: make -C ./lancedb remote-tests
 
   macos:
     timeout-minutes: 30

.github/workflows/trigger-vectordb-recipes.yml

@@ -1,26 +0,0 @@
-name: Trigger vectordb-recipers workflow
-on:
-  push:
-    branches: [ main ]
-  pull_request:
-    paths:
-      - .github/workflows/trigger-vectordb-recipes.yml
-  workflow_dispatch:
-
-jobs:
-  build:
-    runs-on: ubuntu-latest
-
-    steps:
-      - name: Trigger vectordb-recipes workflow
-        uses: actions/github-script@v6
-        with:
-          github-token: ${{ secrets.VECTORDB_RECIPES_ACTION_TOKEN }}
-          script: |
-            const result = await github.rest.actions.createWorkflowDispatch({
-                owner: 'lancedb',
-                repo: 'vectordb-recipes',
-                workflow_id: 'examples-test.yml',
-                ref: 'main'
-            });
-            console.log(result);

Cargo.toml

@@ -15,30 +15,34 @@ categories = ["database-implementations"]
 rust-version = "1.78.0"
 
 [workspace.dependencies]
-lance = { "version" = "=0.35.0", default-features = false, "features" = ["dynamodb"], "tag" = "v0.35.0-beta.4", "git" = "https://github.com/lancedb/lance.git" }
+lance = { "version" = "=0.38.2", default-features = false, "features" = ["dynamodb"], "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
-lance-io = { "version" = "=0.35.0", default-features = false, "tag" = "v0.35.0-beta.4", "git" = "https://github.com/lancedb/lance.git" }
+lance-core = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
-lance-index = { "version" = "=0.35.0", "tag" = "v0.35.0-beta.4", "git" = "https://github.com/lancedb/lance.git" }
+lance-datagen = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
-lance-linalg = { "version" = "=0.35.0", "tag" = "v0.35.0-beta.4", "git" = "https://github.com/lancedb/lance.git" }
+lance-file = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
-lance-table = { "version" = "=0.35.0", "tag" = "v0.35.0-beta.4", "git" = "https://github.com/lancedb/lance.git" }
+lance-io = { "version" = "=0.38.2", default-features = false, "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
-lance-testing = { "version" = "=0.35.0", "tag" = "v0.35.0-beta.4", "git" = "https://github.com/lancedb/lance.git" }
+lance-index = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
-lance-datafusion = { "version" = "=0.35.0", "tag" = "v0.35.0-beta.4", "git" = "https://github.com/lancedb/lance.git" }
+lance-linalg = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
-lance-encoding = { "version" = "=0.35.0", "tag" = "v0.35.0-beta.4", "git" = "https://github.com/lancedb/lance.git" }
+lance-table = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
+lance-testing = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
+lance-datafusion = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
+lance-encoding = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
+lance-namespace = "0.0.18"
+ahash = "0.8"
 # Note that this one does not include pyarrow
-arrow = { version = "55.1", optional = false }
+arrow = { version = "56.2", optional = false }
-arrow-array = "55.1"
+arrow-array = "56.2"
-arrow-data = "55.1"
+arrow-data = "56.2"
-arrow-ipc = "55.1"
+arrow-ipc = "56.2"
-arrow-ord = "55.1"
+arrow-ord = "56.2"
-arrow-schema = "55.1"
+arrow-schema = "56.2"
-arrow-arith = "55.1"
+arrow-cast = "56.2"
-arrow-cast = "55.1"
 async-trait = "0"
-datafusion = { version = "49.0", default-features = false }
+datafusion = { version = "50.1", default-features = false }
-datafusion-catalog = "49.0"
+datafusion-catalog = "50.1"
-datafusion-common = { version = "49.0", default-features = false }
+datafusion-common = { version = "50.1", default-features = false }
-datafusion-execution = "49.0"
+datafusion-execution = "50.1"
-datafusion-expr = "49.0"
+datafusion-expr = "50.1"
-datafusion-physical-plan = "49.0"
+datafusion-physical-plan = "50.1"
 env_logger = "0.11"
 half = { "version" = "2.6.0", default-features = false, features = [
     "num-traits",
@@ -48,18 +52,26 @@ log = "0.4"
 moka = { version = "0.12", features = ["future"] }
 object_store = "0.12.0"
 pin-project = "1.0.7"
+rand = "0.9"
 snafu = "0.8"
 url = "2"
 num-traits = "0.2"
-rand = "0.9"
 regex = "1.10"
 lazy_static = "1"
 semver = "1.0.25"
 crunchy = "0.2.4"
-# Temporary pins to work around downstream issues
+chrono = "0.4"
-# https://github.com/apache/arrow-rs/commit/2fddf85afcd20110ce783ed5b4cdeb82293da30b
-chrono = "=0.4.41"
-# https://github.com/RustCrypto/formats/issues/1684
-base64ct = "=1.6.0"
 # Workaround for: https://github.com/Lokathor/bytemuck/issues/306
 bytemuck_derive = ">=1.8.1, <1.9.0"
+
+# This is only needed when we reference preview releases of lance
+# Force to use the same lance version as the rest of the project to avoid duplicate dependencies
+[patch.crates-io]
+lance = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
+lance-io = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
+lance-index = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
+lance-linalg = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
+lance-table = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
+lance-testing = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
+lance-datafusion = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }
+lance-encoding = { "version" = "=0.38.2", "tag" = "v0.38.3-beta.2", "git" = "https://github.com/lancedb/lance.git" }

ci/create_lancedb_test_connection.sh

@@ -0,0 +1,4 @@
+#!/usr/bin/env bash
+
+export RUST_LOG=info
+exec ./lancedb server --port 0 --sql-port 0  --data-dir "${1}"

ci/run_with_docker_compose.sh

@@ -0,0 +1,18 @@
+#!/usr/bin/env bash
+
+#
+# A script for running the given command together with a docker compose environment.
+#
+
+# Bring down the docker setup once the command is done running.
+tear_down() {
+    docker compose -p fixture down
+}
+trap tear_down EXIT
+
+set +xe
+
+# Clean up any existing docker setup and bring up a new one.
+docker compose -p fixture up --detach --wait || exit 1
+
+"${@}"

ci/run_with_test_connection.sh

@@ -0,0 +1,68 @@
+#!/usr/bin/env bash
+
+#
+# A script for running the given command together with the lancedb cli.
+#
+
+die() {
+    echo $?
+    exit 1
+}
+
+check_command_exists() {
+    command="${1}"
+    which ${command} &> /dev/null || \
+        die "Unable to locate command: ${command}. Did you install it?"
+}
+
+if [[ ! -e ./lancedb ]]; then
+    if [[ -v SOPHON_READ_TOKEN ]]; then
+        INPUT="lancedb-linux-x64"
+        gh release \
+            --repo lancedb/lancedb \
+            download ci-support-binaries \
+            --pattern "${INPUT}" \
+            || die "failed to fetch cli."
+        check_command_exists openssl
+        openssl enc -aes-256-cbc \
+            -d -pbkdf2 \
+            -pass "env:SOPHON_READ_TOKEN" \
+            -in "${INPUT}" \
+            -out ./lancedb-linux-x64.tar.gz \
+            || die "openssl failed"
+        TARGET="${INPUT}.tar.gz"
+    else
+        ARCH="x64"
+        if [[ $OSTYPE == 'darwin'* ]]; then
+            UNAME=$(uname -m)
+            if [[ $UNAME == 'arm64' ]]; then
+                ARCH='arm64'
+            fi
+            OSTYPE="macos"
+        elif [[ $OSTYPE == 'linux'* ]]; then
+            if [[ $UNAME == 'aarch64' ]]; then
+                ARCH='arm64'
+            fi
+            OSTYPE="linux"
+        else
+            die "unknown OSTYPE: $OSTYPE"
+        fi
+
+        check_command_exists gh
+        TARGET="lancedb-${OSTYPE}-${ARCH}.tar.gz"
+        gh release \
+            --repo lancedb/sophon \
+            download lancedb-cli-v0.0.3 \
+            --pattern "${TARGET}" \
+            || die "failed to fetch cli."
+    fi
+
+    check_command_exists tar
+    tar xvf "${TARGET}" || die "tar failed."
+    [[ -e ./lancedb ]] || die "failed to extract lancedb."
+fi
+
+SCRIPT_DIR=$(dirname "$(readlink -f "$0")")
+export CREATE_LANCEDB_TEST_CONNECTION_SCRIPT="${SCRIPT_DIR}/create_lancedb_test_connection.sh"
+
+"${@}"

ci/set_lance_version.py

@@ -1,4 +1,5 @@
 import argparse
+import re
 import sys
 import json
 
@@ -18,8 +19,12 @@ def run_command(command: str) -> str:
 
 def get_latest_stable_version() -> str:
     version_line = run_command("cargo info lance | grep '^version:'")
-    version = version_line.split(" ")[1].strip()
+    # Example output: "version: 0.35.0 (latest 0.37.0)"
-    return version
+    match = re.search(r'\(latest ([0-9.]+)\)', version_line)
+    if match:
+        return match.group(1)
+    # Fallback: use the first version after 'version:'
+    return version_line.split("version:")[1].split()[0].strip()
 
 
 def get_latest_preview_version() -> str:
@@ -112,7 +117,7 @@ def update_cargo_toml(line_updater):
     lance_line = ""
     is_parsing_lance_line = False
     for line in lines:
-        if line.startswith("lance"):
+        if line.startswith("lance") and not line.startswith("lance-namespace"):
             # Check if this is a single-line or multi-line entry
             # Single-line entries either:
             # 1. End with } (complete inline table)

docs/mkdocs.yml

@@ -70,6 +70,23 @@ plugins:
   - mkdocs-jupyter
   - render_swagger:
       allow_arbitrary_locations: true
+  - redirects:
+      redirect_maps:
+        # Redirect the home page and other top-level markdown files. This enables maximum SEO benefit
+        # other sub-pages are handled by the ingected js in overrides/partials/header.html
+        'index.md': 'https://lancedb.com/docs/'
+        'guides/tables.md': 'https://lancedb.com/docs/tables/'
+        'ann_indexes.md': 'https://lancedb.com/docs/indexing/'
+        'basic.md': 'https://lancedb.com/docs/quickstart/'
+        'faq.md': 'https://lancedb.com/docs/faq/'
+        'embeddings/understanding_embeddings.md': 'https://lancedb.com/docs/embedding/'
+        'integrations.md': 'https://lancedb.com/docs/integrations/'
+        'examples.md': 'https://lancedb.com/docs/tutorials/'
+        'concepts/vector_search.md': 'https://lancedb.com/docs/search/vector-search/'
+        'troubleshooting.md': 'https://lancedb.com/docs/troubleshooting/'
+        'guides/storage.md': 'https://lancedb.com/docs/storage/integrations'
+
+
 
 markdown_extensions:
   - admonition

docs/overrides/partials/header.html

@@ -19,7 +19,13 @@
   FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
   IN THE SOFTWARE.
 -->
-
+<div id="deprecation-banner" style="background-color: #f8d7da; color: #721c24; padding: 1em; text-align: center;">                                                           
+    <p style="margin: 0; font-size: 1.1em;">                                                                                                           
+        <strong>This documentation site is deprecated.</strong>                                                                           
+        Please visit our new documentation site at <a href="https://lancedb.com/docs" style="color: #721c24; text-decoration: underline;">
+          lancedb.com/docs</a> for the latest information.                                                                                        
+    </p>                                                                                                                                               
+  </div>  
 {% set class = "md-header" %}
 {% if "navigation.tabs.sticky" in features %}
   {% set class = class ~ " md-header--shadow md-header--lifted" %}
@@ -150,9 +156,9 @@
 
     <div style="margin-left: 10px; margin-right: 5px;">
         <a href="https://discord.com/invite/zMM32dvNtd" target="_blank" rel="noopener noreferrer">
-            <svg fill="#FFFFFF" xmlns="http://www.w3.org/2000/svg"  viewBox="0 0 50 50" width="25px" height="25px"><path d="M 41.625 10.769531 C 37.644531 7.566406 31.347656 7.023438 31.078125 7.003906 C 30.660156 6.96875 30.261719 7.203125 30.089844 7.589844 C 30.074219 7.613281 29.9375 7.929688 29.785156 8.421875 C 32.417969 8.867188 35.652344 9.761719 38.578125 11.578125 C 39.046875 11.867188 39.191406 12.484375 38.902344 12.953125 C 38.710938 13.261719 38.386719 13.429688 38.050781 13.429688 C 37.871094 13.429688 37.6875 13.378906 37.523438 13.277344 C 32.492188 10.15625 26.210938 10 25 10 C 23.789063 10 17.503906 10.15625 12.476563 13.277344 C 12.007813 13.570313 11.390625 13.425781 11.101563 12.957031 C 10.808594 12.484375 10.953125 11.871094 11.421875 11.578125 C 14.347656 9.765625 17.582031 8.867188 20.214844 8.425781 C 20.0625 7.929688 19.925781 7.617188 19.914063 7.589844 C 19.738281 7.203125 19.34375 6.960938 18.921875 7.003906 C 18.652344 7.023438 12.355469 7.566406 8.320313 10.8125 C 6.214844 12.761719 2 24.152344 2 34 C 2 34.175781 2.046875 34.34375 2.132813 34.496094 C 5.039063 39.605469 12.972656 40.941406 14.78125 41 C 14.789063 41 14.800781 41 14.8125 41 C 15.132813 41 15.433594 40.847656 15.621094 40.589844 L 17.449219 38.074219 C 12.515625 36.800781 9.996094 34.636719 9.851563 34.507813 C 9.4375 34.144531 9.398438 33.511719 9.765625 33.097656 C 10.128906 32.683594 10.761719 32.644531 11.175781 33.007813 C 11.234375 33.0625 15.875 37 25 37 C 34.140625 37 38.78125 33.046875 38.828125 33.007813 C 39.242188 32.648438 39.871094 32.683594 40.238281 33.101563 C 40.601563 33.515625 40.5625 34.144531 40.148438 34.507813 C 40.003906 34.636719 37.484375 36.800781 32.550781 38.074219 L 34.378906 40.589844 C 34.566406 40.847656 34.867188 41 35.1875 41 C 35.199219 41 35.210938 41 35.21875 41 C 37.027344 40.941406 44.960938 39.605469 47.867188 34.496094 C 47.953125 34.34375 48 34.175781 48 34 C 48 24.152344 43.785156 12.761719 41.625 10.769531 Z M 18.5 30 C 16.566406 30 15 28.210938 15 26 C 15 23.789063 16.566406 22 18.5 22 C 20.433594 22 22 23.789063 22 26 C 22 28.210938 20.433594 30 18.5 30 Z M 31.5 30 C 29.566406 30 28 28.210938 28 26 C 28 23.789063 29.566406 22 31.5 22 C 33.433594 22 35 23.789063 35 26 C 35 28.210938 33.433594 30 31.5 30 Z"/></svg>
+          <svg fill="#FFFFFF" xmlns="http://www.w3.org/2000/svg"  viewBox="0 0 50 50" width="25px" height="25px"><path d="M 41.625 10.769531 C 37.644531 7.566406 31.347656 7.023438 31.078125 7.003906 C 30.660156 6.96875 30.261719 7.203125 30.089844 7.589844 C 30.074219 7.613281 29.9375 7.929688 29.785156 8.421875 C 32.417969 8.867188 35.652344 9.761719 38.578125 11.578125 C 39.046875 11.867188 39.191406 12.484375 38.902344 12.953125 C 38.710938 13.261719 38.386719 13.429688 38.050781 13.429688 C 37.871094 13.429688 37.6875 13.378906 37.523438 13.277344 C 32.492188 10.15625 26.210938 10 25 10 C 23.789063 10 17.503906 10.15625 12.476563 13.277344 C 12.007813 13.570313 11.390625 13.425781 11.101563 12.957031 C 10.808594 12.484375 10.953125 11.871094 11.421875 11.578125 C 14.347656 9.765625 17.582031 8.867188 20.214844 8.425781 C 20.0625 7.929688 19.925781 7.617188 19.914063 7.589844 C 19.738281 7.203125 19.34375 6.960938 18.921875 7.003906 C 18.652344 7.023438 12.355469 7.566406 8.320313 10.8125 C 6.214844 12.761719 2 24.152344 2 34 C 2 34.175781 2.046875 34.34375 2.132813 34.496094 C 5.039063 39.605469 12.972656 40.941406 14.78125 41 C 14.789063 41 14.800781 41 14.8125 41 C 15.132813 41 15.433594 40.847656 15.621094 40.589844 L 17.449219 38.074219 C 12.515625 36.800781 9.996094 34.636719 9.851563 34.507813 C 9.4375 34.144531 9.398438 33.511719 9.765625 33.097656 C 10.128906 32.683594 10.761719 32.644531 11.175781 33.007813 C 11.234375 33.0625 15.875 37 25 37 C 34.140625 37 38.78125 33.046875 38.828125 33.007813 C 39.242188 32.648438 39.871094 32.683594 40.238281 33.101563 C 40.601563 33.515625 40.5625 34.144531 40.148438 34.507813 C 40.003906 34.636719 37.484375 36.800781 32.550781 38.074219 L 34.378906 40.589844 C 34.566406 40.847656 34.867188 41 35.1875 41 C 35.199219 41 35.210938 41 35.21875 41 C 37.027344 40.941406 44.960938 39.605469 47.867188 34.496094 C 47.953125 34.34375 48 34.175781 48 34 C 48 24.152344 43.785156 12.761719 41.625 10.769531 Z M 18.5 30 C 16.566406 30 15 28.210938 15 26 C 15 23.789063 16.566406 22 18.5 22 C 20.433594 22 22 23.789063 22 26 C 22 28.210938 20.433594 30 18.5 30 Z M 31.5 30 C 29.566406 30 28 28.210938 28 26 C 28 23.789063 29.566406 22 31.5 22 C 33.433594 22 35 23.789063 35 26 C 35 28.210938 33.433594 30 31.5 30 Z"/></svg>
-        </a>
+      </a>
-    </div>
+  </div>
     <div style="margin-left: 5px; margin-right: 5px;">
         <a href="https://twitter.com/lancedb" target="_blank" rel="noopener noreferrer">
             <svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" viewBox="0,0,256,256" width="25px" height="25px" fill-rule="nonzero"><g fill-opacity="0" fill="#ffffff" fill-rule="nonzero" stroke="none" stroke-width="1" stroke-linecap="butt" stroke-linejoin="miter" stroke-miterlimit="10" stroke-dasharray="" stroke-dashoffset="0" font-family="none" font-weight="none" font-size="none" text-anchor="none" style="mix-blend-mode: normal"><path d="M0,256v-256h256v256z" id="bgRectangle"></path></g><g fill="#ffffff" fill-rule="nonzero" stroke="none" stroke-width="1" stroke-linecap="butt" stroke-linejoin="miter" stroke-miterlimit="10" stroke-dasharray="" stroke-dashoffset="0" font-family="none" font-weight="none" font-size="none" text-anchor="none" style="mix-blend-mode: normal"><g transform="scale(4,4)"><path d="M57,17.114c-1.32,1.973 -2.991,3.707 -4.916,5.097c0.018,0.423 0.028,0.847 0.028,1.274c0,13.013 -9.902,28.018 -28.016,28.018c-5.562,0 -12.81,-1.948 -15.095,-4.423c0.772,0.092 1.556,0.138 2.35,0.138c4.615,0 8.861,-1.575 12.23,-4.216c-4.309,-0.079 -7.946,-2.928 -9.199,-6.84c1.96,0.308 4.447,-0.17 4.447,-0.17c0,0 -7.7,-1.322 -7.899,-9.779c2.226,1.291 4.46,1.231 4.46,1.231c0,0 -4.441,-2.734 -4.379,-8.195c0.037,-3.221 1.331,-4.953 1.331,-4.953c8.414,10.361 20.298,10.29 20.298,10.29c0,0 -0.255,-1.471 -0.255,-2.243c0,-5.437 4.408,-9.847 9.847,-9.847c2.832,0 5.391,1.196 7.187,3.111c2.245,-0.443 4.353,-1.263 6.255,-2.391c-0.859,3.44 -4.329,5.448 -4.329,5.448c0,0 2.969,-0.329 5.655,-1.55z"></path></g></g></svg>
@@ -173,4 +179,77 @@
       {% include "partials/tabs.html" %}
     {% endif %}
   {% endif %}
-</header>
+</header>
+
+<script>
+  (function() {
+    function checkPathAndRedirect() {
+      var banner = document.getElementById('deprecation-banner');
+
+      if (document.querySelector('meta[http-equiv="refresh"]')) {
+        return; // The redirects plugin is already handling this page.
+      }
+
+      var currentPath = window.location.pathname;
+
+      var cleanPath = currentPath.endsWith('/') && currentPath.length > 1
+        ? currentPath.slice(0, -1)
+        : currentPath;
+
+      // These are the ONLY paths that should remain on the old site
+      var apiPaths = [
+        '/lancedb/python', 
+        '/lancedb/javascript', 
+        '/lancedb/js',
+        '/lancedb/api_reference'
+      ];
+      
+      var isApiPage = apiPaths.some(function(apiPath) {
+        return cleanPath.startsWith(apiPath);
+      });
+
+      if (isApiPage) {
+        if (banner) {
+          banner.style.display = 'none';
+        }
+      } else {
+        if (banner) {
+          banner.style.display = 'block';
+        }
+        
+        // Add noindex meta tag to prevent indexing of old docs for seo
+        var noindexMeta = document.createElement('meta');
+        noindexMeta.setAttribute('name', 'robots');
+        noindexMeta.setAttribute('content', 'noindex, follow');
+        document.head.appendChild(noindexMeta);
+
+        // Add canonical link to point to the new docs to reward new site for seo
+        var canonicalLink = document.createElement('link');
+        canonicalLink.setAttribute('rel', 'canonical');
+        canonicalLink.setAttribute('href', 'https://lancedb.com/docs');
+        document.head.appendChild(canonicalLink);
+        
+        window.location.replace('https://lancedb.com/docs');
+      }
+    }
+
+    // Run the check only if doc is ready. This makes sure we catch the initial load
+    // and redirect.
+    if (document.readyState === 'loading') {
+      document.addEventListener('DOMContentLoaded', checkPathAndRedirect);
+    } else {
+      checkPathAndRedirect();
+    }
+
+    // Use an interval to handle subsequent navigation clicks.
+    var lastPath = window.location.pathname;
+    setInterval(function() {
+      if (window.location.pathname !== lastPath) {
+        lastPath = window.location.pathname;
+        checkPathAndRedirect();
+      }
+    }, 2000); // keeping it 2 second to make it easy for user to understand
+              // what's happening
+
+  })();
+</script>

docs/requirements.txt

@@ -5,3 +5,4 @@ mkdocstrings[python]==0.25.2
 griffe
 mkdocs-render-swagger-plugin
 pydantic
+mkdocs-redirects

docs/src/js/classes/Connection.md

@@ -25,6 +25,51 @@ the underlying connection has been closed.
 
 ## Methods
 
+### cloneTable()
+
+```ts
+abstract cloneTable(
+   targetTableName,
+   sourceUri,
+   options?): Promise<Table>
+```
+
+Clone a table from a source table.
+
+A shallow clone creates a new table that shares the underlying data files
+with the source table but has its own independent manifest. This allows
+both the source and cloned tables to evolve independently while initially
+sharing the same data, deletion, and index files.
+
+#### Parameters
+
+* **targetTableName**: `string`
+    The name of the target table to create.
+
+* **sourceUri**: `string`
+    The URI of the source table to clone from.
+
+* **options?**
+    Clone options.
+
+* **options.isShallow?**: `boolean`
+    Whether to perform a shallow clone (defaults to true).
+
+* **options.sourceTag?**: `string`
+    The tag of the source table to clone.
+
+* **options.sourceVersion?**: `number`
+    The version of the source table to clone.
+
+* **options.targetNamespace?**: `string`[]
+    The namespace for the target table (defaults to root namespace).
+
+#### Returns
+
+`Promise`&lt;[`Table`](Table.md)&gt;
+
+***
+
 ### close()
 
 ```ts

docs/src/js/classes/Index.md

@@ -194,6 +194,37 @@ currently is also a memory intensive operation.
 
 ***
 
+### ivfRq()
+
+```ts
+static ivfRq(options?): Index
+```
+
+Create an IvfRq index
+
+IVF-RQ (RabitQ Quantization) compresses vectors using RabitQ quantization
+and organizes them into IVF partitions.
+
+The compression scheme is called RabitQ quantization. Each dimension is quantized into a small number of bits.
+The parameters `num_bits` and `num_partitions` control this process, providing a tradeoff
+between index size (and thus search speed) and index accuracy.
+
+The partitioning process is called IVF and the `num_partitions` parameter controls how
+many groups to create.
+
+Note that training an IVF RQ index on a large dataset is a slow operation and
+currently is also a memory intensive operation.
+
+#### Parameters
+
+* **options?**: `Partial`<[`IvfRqOptions`](../interfaces/IvfRqOptions.md)>
+
+#### Returns
+
+[`Index`](Index.md)
+
+***
+
 ### labelList()
 
 ```ts

docs/src/js/classes/MergeInsertBuilder.md

@@ -52,6 +52,30 @@ the merge result
 
 ***
 
+### useIndex()
+
+```ts
+useIndex(useIndex): MergeInsertBuilder
+```
+
+Controls whether to use indexes for the merge operation.
+
+When set to `true` (the default), the operation will use an index if available
+on the join key for improved performance. When set to `false`, it forces a full
+table scan even if an index exists. This can be useful for benchmarking or when
+the query optimizer chooses a suboptimal path.
+
+#### Parameters
+
+* **useIndex**: `boolean`
+    Whether to use indices for the merge operation. Defaults to `true`.
+
+#### Returns
+
+[`MergeInsertBuilder`](MergeInsertBuilder.md)
+
+***
+
 ### whenMatchedUpdateAll()
 
 ```ts

docs/src/js/classes/PermutationBuilder.md

@@ -0,0 +1,220 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / PermutationBuilder
+
+# Class: PermutationBuilder
+
+A PermutationBuilder for creating data permutations with splits, shuffling, and filtering.
+
+This class provides a TypeScript wrapper around the native Rust PermutationBuilder,
+offering methods to configure data splits, shuffling, and filtering before executing
+the permutation to create a new table.
+
+## Methods
+
+### execute()
+
+```ts
+execute(): Promise<Table>
+```
+
+Execute the permutation and create the destination table.
+
+#### Returns
+
+`Promise`&lt;[`Table`](Table.md)&gt;
+
+A Promise that resolves to the new Table instance
+
+#### Example
+
+```ts
+const permutationTable = await builder.execute();
+console.log(`Created table: ${permutationTable.name}`);
+```
+
+***
+
+### filter()
+
+```ts
+filter(filter): PermutationBuilder
+```
+
+Configure filtering for the permutation.
+
+#### Parameters
+
+* **filter**: `string`
+    SQL filter expression
+
+#### Returns
+
+[`PermutationBuilder`](PermutationBuilder.md)
+
+A new PermutationBuilder instance
+
+#### Example
+
+```ts
+builder.filter("age > 18 AND status = 'active'");
+```
+
+***
+
+### shuffle()
+
+```ts
+shuffle(options): PermutationBuilder
+```
+
+Configure shuffling for the permutation.
+
+#### Parameters
+
+* **options**: [`ShuffleOptions`](../interfaces/ShuffleOptions.md)
+    Configuration for shuffling
+
+#### Returns
+
+[`PermutationBuilder`](PermutationBuilder.md)
+
+A new PermutationBuilder instance
+
+#### Example
+
+```ts
+// Basic shuffle
+builder.shuffle({ seed: 42 });
+
+// Shuffle with clump size
+builder.shuffle({ seed: 42, clumpSize: 10 });
+```
+
+***
+
+### splitCalculated()
+
+```ts
+splitCalculated(calculation): PermutationBuilder
+```
+
+Configure calculated splits for the permutation.
+
+#### Parameters
+
+* **calculation**: `string`
+    SQL expression for calculating splits
+
+#### Returns
+
+[`PermutationBuilder`](PermutationBuilder.md)
+
+A new PermutationBuilder instance
+
+#### Example
+
+```ts
+builder.splitCalculated("user_id % 3");
+```
+
+***
+
+### splitHash()
+
+```ts
+splitHash(options): PermutationBuilder
+```
+
+Configure hash-based splits for the permutation.
+
+#### Parameters
+
+* **options**: [`SplitHashOptions`](../interfaces/SplitHashOptions.md)
+    Configuration for hash-based splitting
+
+#### Returns
+
+[`PermutationBuilder`](PermutationBuilder.md)
+
+A new PermutationBuilder instance
+
+#### Example
+
+```ts
+builder.splitHash({
+  columns: ["user_id"],
+  splitWeights: [70, 30],
+  discardWeight: 0
+});
+```
+
+***
+
+### splitRandom()
+
+```ts
+splitRandom(options): PermutationBuilder
+```
+
+Configure random splits for the permutation.
+
+#### Parameters
+
+* **options**: [`SplitRandomOptions`](../interfaces/SplitRandomOptions.md)
+    Configuration for random splitting
+
+#### Returns
+
+[`PermutationBuilder`](PermutationBuilder.md)
+
+A new PermutationBuilder instance
+
+#### Example
+
+```ts
+// Split by ratios
+builder.splitRandom({ ratios: [0.7, 0.3], seed: 42 });
+
+// Split by counts
+builder.splitRandom({ counts: [1000, 500], seed: 42 });
+
+// Split with fixed size
+builder.splitRandom({ fixed: 100, seed: 42 });
+```
+
+***
+
+### splitSequential()
+
+```ts
+splitSequential(options): PermutationBuilder
+```
+
+Configure sequential splits for the permutation.
+
+#### Parameters
+
+* **options**: [`SplitSequentialOptions`](../interfaces/SplitSequentialOptions.md)
+    Configuration for sequential splitting
+
+#### Returns
+
+[`PermutationBuilder`](PermutationBuilder.md)
+
+A new PermutationBuilder instance
+
+#### Example
+
+```ts
+// Split by ratios
+builder.splitSequential({ ratios: [0.8, 0.2] });
+
+// Split by counts
+builder.splitSequential({ counts: [800, 200] });
+
+// Split with fixed size
+builder.splitSequential({ fixed: 1000 });
+```

docs/src/js/functions/makeArrowTable.md

@@ -13,7 +13,7 @@ function makeArrowTable(
    metadata?): ArrowTable
 ```
 
-An enhanced version of the makeTable function from Apache Arrow
+An enhanced version of the apache-arrow makeTable function from Apache Arrow
 that supports nested fields and embeddings columns.
 
 (typically you do not need to call this function.  It will be called automatically

docs/src/js/functions/permutationBuilder.md

@@ -0,0 +1,37 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / permutationBuilder
+
+# Function: permutationBuilder()
+
+```ts
+function permutationBuilder(table, destTableName): PermutationBuilder
+```
+
+Create a permutation builder for the given table.
+
+## Parameters
+
+* **table**: [`Table`](../classes/Table.md)
+    The source table to create a permutation from
+
+* **destTableName**: `string`
+    The name for the destination permutation table
+
+## Returns
+
+[`PermutationBuilder`](../classes/PermutationBuilder.md)
+
+A PermutationBuilder instance
+
+## Example
+
+```ts
+const builder = permutationBuilder(sourceTable, "training_data")
+  .splitRandom({ ratios: [0.8, 0.2], seed: 42 })
+  .shuffle({ seed: 123 });
+
+const trainingTable = await builder.execute();
+```

docs/src/js/globals.md

@@ -28,6 +28,7 @@
 - [MultiMatchQuery](classes/MultiMatchQuery.md)
 - [NativeJsHeaderProvider](classes/NativeJsHeaderProvider.md)
 - [OAuthHeaderProvider](classes/OAuthHeaderProvider.md)
+- [PermutationBuilder](classes/PermutationBuilder.md)
 - [PhraseQuery](classes/PhraseQuery.md)
 - [Query](classes/Query.md)
 - [QueryBase](classes/QueryBase.md)
@@ -68,6 +69,7 @@
 - [IndexStatistics](interfaces/IndexStatistics.md)
 - [IvfFlatOptions](interfaces/IvfFlatOptions.md)
 - [IvfPqOptions](interfaces/IvfPqOptions.md)
+- [IvfRqOptions](interfaces/IvfRqOptions.md)
 - [MergeResult](interfaces/MergeResult.md)
 - [OpenTableOptions](interfaces/OpenTableOptions.md)
 - [OptimizeOptions](interfaces/OptimizeOptions.md)
@@ -75,9 +77,14 @@
 - [QueryExecutionOptions](interfaces/QueryExecutionOptions.md)
 - [RemovalStats](interfaces/RemovalStats.md)
 - [RetryConfig](interfaces/RetryConfig.md)
+- [ShuffleOptions](interfaces/ShuffleOptions.md)
+- [SplitHashOptions](interfaces/SplitHashOptions.md)
+- [SplitRandomOptions](interfaces/SplitRandomOptions.md)
+- [SplitSequentialOptions](interfaces/SplitSequentialOptions.md)
 - [TableNamesOptions](interfaces/TableNamesOptions.md)
 - [TableStatistics](interfaces/TableStatistics.md)
 - [TimeoutConfig](interfaces/TimeoutConfig.md)
+- [TlsConfig](interfaces/TlsConfig.md)
 - [TokenResponse](interfaces/TokenResponse.md)
 - [UpdateOptions](interfaces/UpdateOptions.md)
 - [UpdateResult](interfaces/UpdateResult.md)
@@ -101,3 +108,4 @@
 - [connect](functions/connect.md)
 - [makeArrowTable](functions/makeArrowTable.md)
 - [packBits](functions/packBits.md)
+- [permutationBuilder](functions/permutationBuilder.md)

docs/src/js/interfaces/ClientConfig.md

@@ -40,6 +40,14 @@ optional timeoutConfig: TimeoutConfig;
 
 ***
 
+### tlsConfig?
+
+```ts
+optional tlsConfig: TlsConfig;
+```
+
+***
+
 ### userAgent?
 
 ```ts

docs/src/js/interfaces/ShuffleOptions.md

@@ -0,0 +1,23 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / ShuffleOptions
+
+# Interface: ShuffleOptions
+
+## Properties
+
+### clumpSize?
+
+```ts
+optional clumpSize: number;
+```
+
+***
+
+### seed?
+
+```ts
+optional seed: number;
+```

docs/src/js/interfaces/SplitHashOptions.md

@@ -0,0 +1,31 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / SplitHashOptions
+
+# Interface: SplitHashOptions
+
+## Properties
+
+### columns
+
+```ts
+columns: string[];
+```
+
+***
+
+### discardWeight?
+
+```ts
+optional discardWeight: number;
+```
+
+***
+
+### splitWeights
+
+```ts
+splitWeights: number[];
+```

docs/src/js/interfaces/SplitRandomOptions.md

@@ -0,0 +1,39 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / SplitRandomOptions
+
+# Interface: SplitRandomOptions
+
+## Properties
+
+### counts?
+
+```ts
+optional counts: number[];
+```
+
+***
+
+### fixed?
+
+```ts
+optional fixed: number;
+```
+
+***
+
+### ratios?
+
+```ts
+optional ratios: number[];
+```
+
+***
+
+### seed?
+
+```ts
+optional seed: number;
+```

docs/src/js/interfaces/SplitSequentialOptions.md

@@ -0,0 +1,31 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / SplitSequentialOptions
+
+# Interface: SplitSequentialOptions
+
+## Properties
+
+### counts?
+
+```ts
+optional counts: number[];
+```
+
+***
+
+### fixed?
+
+```ts
+optional fixed: number;
+```
+
+***
+
+### ratios?
+
+```ts
+optional ratios: number[];
+```

docs/src/js/interfaces/TlsConfig.md

@@ -0,0 +1,49 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / TlsConfig
+
+# Interface: TlsConfig
+
+TLS/mTLS configuration for the remote HTTP client.
+
+## Properties
+
+### assertHostname?
+
+```ts
+optional assertHostname: boolean;
+```
+
+Whether to verify the hostname in the server's certificate.
+
+***
+
+### certFile?
+
+```ts
+optional certFile: string;
+```
+
+Path to the client certificate file (PEM format) for mTLS authentication.
+
+***
+
+### keyFile?
+
+```ts
+optional keyFile: string;
+```
+
+Path to the client private key file (PEM format) for mTLS authentication.
+
+***
+
+### sslCaCert?
+
+```ts
+optional sslCaCert: string;
+```
+
+Path to the CA certificate file (PEM format) for server verification.

java/core/lancedb-jni/src/ffi.rs

@@ -16,6 +16,7 @@ pub trait JNIEnvExt {
     fn get_integers(&mut self, obj: &JObject) -> Result<Vec<i32>>;
 
     /// Get strings from Java List<String> object.
+    #[allow(dead_code)]
     fn get_strings(&mut self, obj: &JObject) -> Result<Vec<String>>;
 
     /// Get strings from Java String[] object.

java/core/lancedb-jni/src/traits.rs

@@ -6,6 +6,7 @@ use jni::JNIEnv;
 
 use crate::Result;
 
+#[allow(dead_code)]
 pub trait FromJObject<T> {
     fn extract(&self) -> Result<T>;
 }
@@ -39,6 +40,7 @@ impl FromJObject<f64> for JObject<'_> {
     }
 }
 
+#[allow(dead_code)]
 pub trait FromJString {
     fn extract(&self, env: &mut JNIEnv) -> Result<String>;
 }
@@ -66,6 +68,7 @@ pub trait JMapExt {
     fn get_f64(&self, env: &mut JNIEnv, key: &str) -> Result<Option<f64>>;
 }
 
+#[allow(dead_code)]
 fn get_map_value<T>(env: &mut JNIEnv, map: &JMap, key: &str) -> Result<Option<T>>
 where
     for<'a> JObject<'a>: FromJObject<T>,

java/core/pom.xml

@@ -8,7 +8,7 @@
     <parent>
         <groupId>com.lancedb</groupId>
         <artifactId>lancedb-parent</artifactId>
-        <version>0.22.1-beta.0</version>
+        <version>0.22.2-final.0</version>
         <relativePath>../pom.xml</relativePath>
     </parent>
 

java/lance-namespace/pom.xml

@@ -8,7 +8,7 @@
     <parent>
         <groupId>com.lancedb</groupId>
         <artifactId>lancedb-parent</artifactId>
-        <version>0.22.1-beta.0</version>
+        <version>0.22.2-final.0</version>
         <relativePath>../pom.xml</relativePath>
     </parent>
 

java/pom.xml

@@ -6,7 +6,7 @@
 
     <groupId>com.lancedb</groupId>
     <artifactId>lancedb-parent</artifactId>
-    <version>0.22.1-beta.0</version>
+    <version>0.22.2-final.0</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 = "0.22.1-beta.0"
+version = "0.22.2"
 license.workspace = true
 description.workspace = true
 repository.workspace = true

nodejs/__test__/arrow.test.ts

@@ -1,17 +1,5 @@
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
-
-import {
-  Bool,
-  Field,
-  Int32,
-  List,
-  Schema,
-  Struct,
-  Uint8,
-  Utf8,
-} from "apache-arrow";
-
 import * as arrow15 from "apache-arrow-15";
 import * as arrow16 from "apache-arrow-16";
 import * as arrow17 from "apache-arrow-17";
@@ -25,11 +13,9 @@ import {
   fromTableToBuffer,
   makeArrowTable,
   makeEmptyTable,
-  tableFromIPC,
 } from "../lancedb/arrow";
 import {
   EmbeddingFunction,
-  FieldOptions,
   FunctionOptions,
 } from "../lancedb/embedding/embedding_function";
 import { EmbeddingFunctionConfig } from "../lancedb/embedding/registry";
@@ -1008,5 +994,64 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
         expect(result).toEqual(null);
       });
     });
+
+    describe("boolean null handling", function () {
+      it("should handle null values in nullable boolean fields", () => {
+        const { makeArrowTable } = require("../lancedb/arrow");
+        const schema = new Schema([new Field("test", new arrow.Bool(), true)]);
+
+        // Test with all null values
+        const data = [{ test: null }];
+        const table = makeArrowTable(data, { schema });
+
+        expect(table.numRows).toBe(1);
+        expect(table.schema.names).toEqual(["test"]);
+        expect(table.getChild("test")!.get(0)).toBeNull();
+      });
+
+      it("should handle mixed null and non-null boolean values", () => {
+        const { makeArrowTable } = require("../lancedb/arrow");
+        const schema = new Schema([new Field("test", new Bool(), true)]);
+
+        // Test with mixed values
+        const data = [{ test: true }, { test: null }, { test: false }];
+        const table = makeArrowTable(data, { schema });
+
+        expect(table.numRows).toBe(3);
+        expect(table.getChild("test")!.get(0)).toBe(true);
+        expect(table.getChild("test")!.get(1)).toBeNull();
+        expect(table.getChild("test")!.get(2)).toBe(false);
+      });
+    });
+
+    // Test for the undefined values bug fix
+    describe("undefined values handling", () => {
+      it("should handle mixed undefined and actual values", () => {
+        const schema = new Schema([
+          new Field("text", new Utf8(), true), // nullable
+          new Field("number", new Int32(), true), // nullable
+          new Field("bool", new Bool(), true), // nullable
+        ]);
+
+        const data = [
+          { text: undefined, number: 42, bool: true },
+          { text: "hello", number: undefined, bool: false },
+          { text: "world", number: 123, bool: undefined },
+        ];
+        const table = makeArrowTable(data, { schema });
+
+        const result = table.toArray();
+        expect(result).toHaveLength(3);
+        expect(result[0].text).toBe(null);
+        expect(result[0].number).toBe(42);
+        expect(result[0].bool).toBe(true);
+        expect(result[1].text).toBe("hello");
+        expect(result[1].number).toBe(null);
+        expect(result[1].bool).toBe(false);
+        expect(result[2].text).toBe("world");
+        expect(result[2].number).toBe(123);
+        expect(result[2].bool).toBe(null);
+      });
+    });
   },
 );

nodejs/__test__/connection.test.ts

@@ -203,3 +203,106 @@ describe("given a connection", () => {
     });
   });
 });
+
+describe("clone table functionality", () => {
+  let tmpDir: tmp.DirResult;
+  let db: Connection;
+  beforeEach(async () => {
+    tmpDir = tmp.dirSync({ unsafeCleanup: true });
+    db = await connect(tmpDir.name);
+  });
+  afterEach(() => tmpDir.removeCallback());
+
+  it("should clone a table with latest version (default behavior)", async () => {
+    // Create source table with some data
+    const data = [
+      { id: 1, text: "hello", vector: [1.0, 2.0] },
+      { id: 2, text: "world", vector: [3.0, 4.0] },
+    ];
+    const sourceTable = await db.createTable("source", data);
+
+    // Add more data to create a new version
+    const moreData = [{ id: 3, text: "test", vector: [5.0, 6.0] }];
+    await sourceTable.add(moreData);
+
+    // Clone the table (should get latest version with 3 rows)
+    const sourceUri = `${tmpDir.name}/source.lance`;
+    const clonedTable = await db.cloneTable("cloned", sourceUri);
+
+    // Verify cloned table has all 3 rows
+    expect(await clonedTable.countRows()).toBe(3);
+    expect((await db.tableNames()).includes("cloned")).toBe(true);
+  });
+
+  it("should clone a table from a specific version", async () => {
+    // Create source table with initial data
+    const data = [
+      { id: 1, text: "hello", vector: [1.0, 2.0] },
+      { id: 2, text: "world", vector: [3.0, 4.0] },
+    ];
+    const sourceTable = await db.createTable("source", data);
+
+    // Get the initial version
+    const initialVersion = await sourceTable.version();
+
+    // Add more data to create a new version
+    const moreData = [{ id: 3, text: "test", vector: [5.0, 6.0] }];
+    await sourceTable.add(moreData);
+
+    // Verify source now has 3 rows
+    expect(await sourceTable.countRows()).toBe(3);
+
+    // Clone from the initial version (should have only 2 rows)
+    const sourceUri = `${tmpDir.name}/source.lance`;
+    const clonedTable = await db.cloneTable("cloned", sourceUri, {
+      sourceVersion: initialVersion,
+    });
+
+    // Verify cloned table has only the initial 2 rows
+    expect(await clonedTable.countRows()).toBe(2);
+  });
+
+  it("should clone a table from a tagged version", async () => {
+    // Create source table with initial data
+    const data = [
+      { id: 1, text: "hello", vector: [1.0, 2.0] },
+      { id: 2, text: "world", vector: [3.0, 4.0] },
+    ];
+    const sourceTable = await db.createTable("source", data);
+
+    // Create a tag for the current version
+    const tags = await sourceTable.tags();
+    await tags.create("v1.0", await sourceTable.version());
+
+    // Add more data after the tag
+    const moreData = [{ id: 3, text: "test", vector: [5.0, 6.0] }];
+    await sourceTable.add(moreData);
+
+    // Verify source now has 3 rows
+    expect(await sourceTable.countRows()).toBe(3);
+
+    // Clone from the tagged version (should have only 2 rows)
+    const sourceUri = `${tmpDir.name}/source.lance`;
+    const clonedTable = await db.cloneTable("cloned", sourceUri, {
+      sourceTag: "v1.0",
+    });
+
+    // Verify cloned table has only the tagged version's 2 rows
+    expect(await clonedTable.countRows()).toBe(2);
+  });
+
+  it("should fail when attempting deep clone", async () => {
+    // Create source table with some data
+    const data = [
+      { id: 1, text: "hello", vector: [1.0, 2.0] },
+      { id: 2, text: "world", vector: [3.0, 4.0] },
+    ];
+    await db.createTable("source", data);
+
+    // Try to create a deep clone (should fail)
+    const sourceUri = `${tmpDir.name}/source.lance`;
+    await expect(
+      db.cloneTable("cloned", sourceUri, { isShallow: false }),
+    ).rejects.toThrow("Deep clone is not yet implemented");
+  });
+});

nodejs/__test__/embedding.test.ts

@@ -256,6 +256,60 @@ describe("embedding functions", () => {
     expect(actual).toHaveProperty("text");
   });
 
+  it("should handle undefined vector field with embedding function correctly", async () => {
+    @register("undefined_test")
+    class MockEmbeddingFunction extends EmbeddingFunction<string> {
+      ndims() {
+        return 3;
+      }
+      embeddingDataType(): Float {
+        return new Float32();
+      }
+      async computeQueryEmbeddings(_data: string) {
+        return [1, 2, 3];
+      }
+      async computeSourceEmbeddings(data: string[]) {
+        return Array.from({ length: data.length }).fill([
+          1, 2, 3,
+        ]) as number[][];
+      }
+    }
+    const func = getRegistry()
+      .get<MockEmbeddingFunction>("undefined_test")!
+      .create();
+    const schema = new Schema([
+      new Field("text", new Utf8(), true),
+      new Field(
+        "vector",
+        new FixedSizeList(3, new Field("item", new Float32(), true)),
+        true,
+      ),
+    ]);
+
+    const db = await connect(tmpDir.name);
+    const table = await db.createEmptyTable("test_undefined", schema, {
+      embeddingFunction: {
+        function: func,
+        sourceColumn: "text",
+        vectorColumn: "vector",
+      },
+    });
+
+    // Test that undefined, null, and omitted vector fields all work
+    await table.add([{ text: "test1", vector: undefined }]);
+    await table.add([{ text: "test2", vector: null }]);
+    await table.add([{ text: "test3" }]);
+
+    const rows = await table.query().toArray();
+    expect(rows.length).toBe(3);
+
+    // All rows should have vectors computed by the embedding function
+    for (const row of rows) {
+      expect(row.vector).toBeDefined();
+      expect(JSON.parse(JSON.stringify(row.vector))).toEqual([1, 2, 3]);
+    }
+  });
+
   test.each([new Float16(), new Float32(), new Float64()])(
     "should be able to provide manual embeddings with multiple float datatype",
     async (floatType) => {

nodejs/__test__/permutation.test.ts

@@ -0,0 +1,234 @@
+// SPDX-License-Identifier: Apache-2.0
+// SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+import * as tmp from "tmp";
+import { Table, connect, permutationBuilder } from "../lancedb";
+import { makeArrowTable } from "../lancedb/arrow";
+
+describe("PermutationBuilder", () => {
+  let tmpDir: tmp.DirResult;
+  let table: Table;
+
+  beforeEach(async () => {
+    tmpDir = tmp.dirSync({ unsafeCleanup: true });
+    const db = await connect(tmpDir.name);
+
+    // Create test data
+    const data = makeArrowTable(
+      [
+        { id: 1, value: 10 },
+        { id: 2, value: 20 },
+        { id: 3, value: 30 },
+        { id: 4, value: 40 },
+        { id: 5, value: 50 },
+        { id: 6, value: 60 },
+        { id: 7, value: 70 },
+        { id: 8, value: 80 },
+        { id: 9, value: 90 },
+        { id: 10, value: 100 },
+      ],
+      { vectorColumns: {} },
+    );
+
+    table = await db.createTable("test_table", data);
+  });
+
+  afterEach(() => {
+    tmpDir.removeCallback();
+  });
+
+  test("should create permutation builder", () => {
+    const builder = permutationBuilder(table, "permutation_table");
+    expect(builder).toBeDefined();
+  });
+
+  test("should execute basic permutation", async () => {
+    const builder = permutationBuilder(table, "permutation_table");
+    const permutationTable = await builder.execute();
+
+    expect(permutationTable).toBeDefined();
+    expect(permutationTable.name).toBe("permutation_table");
+
+    const rowCount = await permutationTable.countRows();
+    expect(rowCount).toBe(10);
+  });
+
+  test("should create permutation with random splits", async () => {
+    const builder = permutationBuilder(table, "permutation_table").splitRandom({
+      ratios: [1.0],
+      seed: 42,
+    });
+
+    const permutationTable = await builder.execute();
+    const rowCount = await permutationTable.countRows();
+    expect(rowCount).toBe(10);
+  });
+
+  test("should create permutation with percentage splits", async () => {
+    const builder = permutationBuilder(table, "permutation_table").splitRandom({
+      ratios: [0.3, 0.7],
+      seed: 42,
+    });
+
+    const permutationTable = await builder.execute();
+    const rowCount = await permutationTable.countRows();
+    expect(rowCount).toBe(10);
+
+    // Check split distribution
+    const split0Count = await permutationTable.countRows("split_id = 0");
+    const split1Count = await permutationTable.countRows("split_id = 1");
+
+    expect(split0Count).toBeGreaterThan(0);
+    expect(split1Count).toBeGreaterThan(0);
+    expect(split0Count + split1Count).toBe(10);
+  });
+
+  test("should create permutation with count splits", async () => {
+    const builder = permutationBuilder(table, "permutation_table").splitRandom({
+      counts: [3, 7],
+      seed: 42,
+    });
+
+    const permutationTable = await builder.execute();
+    const rowCount = await permutationTable.countRows();
+    expect(rowCount).toBe(10);
+
+    // Check split distribution
+    const split0Count = await permutationTable.countRows("split_id = 0");
+    const split1Count = await permutationTable.countRows("split_id = 1");
+
+    expect(split0Count).toBe(3);
+    expect(split1Count).toBe(7);
+  });
+
+  test("should create permutation with hash splits", async () => {
+    const builder = permutationBuilder(table, "permutation_table").splitHash({
+      columns: ["id"],
+      splitWeights: [50, 50],
+      discardWeight: 0,
+    });
+
+    const permutationTable = await builder.execute();
+    const rowCount = await permutationTable.countRows();
+    expect(rowCount).toBe(10);
+
+    // Check that splits exist
+    const split0Count = await permutationTable.countRows("split_id = 0");
+    const split1Count = await permutationTable.countRows("split_id = 1");
+
+    expect(split0Count).toBeGreaterThan(0);
+    expect(split1Count).toBeGreaterThan(0);
+    expect(split0Count + split1Count).toBe(10);
+  });
+
+  test("should create permutation with sequential splits", async () => {
+    const builder = permutationBuilder(
+      table,
+      "permutation_table",
+    ).splitSequential({ ratios: [0.5, 0.5] });
+
+    const permutationTable = await builder.execute();
+    const rowCount = await permutationTable.countRows();
+    expect(rowCount).toBe(10);
+
+    // Check split distribution - sequential should give exactly 5 and 5
+    const split0Count = await permutationTable.countRows("split_id = 0");
+    const split1Count = await permutationTable.countRows("split_id = 1");
+
+    expect(split0Count).toBe(5);
+    expect(split1Count).toBe(5);
+  });
+
+  test("should create permutation with calculated splits", async () => {
+    const builder = permutationBuilder(
+      table,
+      "permutation_table",
+    ).splitCalculated("id % 2");
+
+    const permutationTable = await builder.execute();
+    const rowCount = await permutationTable.countRows();
+    expect(rowCount).toBe(10);
+
+    // Check split distribution
+    const split0Count = await permutationTable.countRows("split_id = 0");
+    const split1Count = await permutationTable.countRows("split_id = 1");
+
+    expect(split0Count).toBeGreaterThan(0);
+    expect(split1Count).toBeGreaterThan(0);
+    expect(split0Count + split1Count).toBe(10);
+  });
+
+  test("should create permutation with shuffle", async () => {
+    const builder = permutationBuilder(table, "permutation_table").shuffle({
+      seed: 42,
+    });
+
+    const permutationTable = await builder.execute();
+    const rowCount = await permutationTable.countRows();
+    expect(rowCount).toBe(10);
+  });
+
+  test("should create permutation with shuffle and clump size", async () => {
+    const builder = permutationBuilder(table, "permutation_table").shuffle({
+      seed: 42,
+      clumpSize: 2,
+    });
+
+    const permutationTable = await builder.execute();
+    const rowCount = await permutationTable.countRows();
+    expect(rowCount).toBe(10);
+  });
+
+  test("should create permutation with filter", async () => {
+    const builder = permutationBuilder(table, "permutation_table").filter(
+      "value > 50",
+    );
+
+    const permutationTable = await builder.execute();
+    const rowCount = await permutationTable.countRows();
+    expect(rowCount).toBe(5); // Values 60, 70, 80, 90, 100
+  });
+
+  test("should chain multiple operations", async () => {
+    const builder = permutationBuilder(table, "permutation_table")
+      .filter("value <= 80")
+      .splitRandom({ ratios: [0.5, 0.5], seed: 42 })
+      .shuffle({ seed: 123 });
+
+    const permutationTable = await builder.execute();
+    const rowCount = await permutationTable.countRows();
+    expect(rowCount).toBe(8); // Values 10, 20, 30, 40, 50, 60, 70, 80
+
+    // Check split distribution
+    const split0Count = await permutationTable.countRows("split_id = 0");
+    const split1Count = await permutationTable.countRows("split_id = 1");
+
+    expect(split0Count).toBeGreaterThan(0);
+    expect(split1Count).toBeGreaterThan(0);
+    expect(split0Count + split1Count).toBe(8);
+  });
+
+  test("should throw error for invalid split arguments", () => {
+    const builder = permutationBuilder(table, "permutation_table");
+
+    // Test no arguments provided
+    expect(() => builder.splitRandom({})).toThrow(
+      "Exactly one of 'ratios', 'counts', or 'fixed' must be provided",
+    );
+
+    // Test multiple arguments provided
+    expect(() =>
+      builder.splitRandom({ ratios: [0.5, 0.5], counts: [3, 7], seed: 42 }),
+    ).toThrow("Exactly one of 'ratios', 'counts', or 'fixed' must be provided");
+  });
+
+  test("should throw error when builder is consumed", async () => {
+    const builder = permutationBuilder(table, "permutation_table");
+
+    // Execute once
+    await builder.execute();
+
+    // Should throw error on second execution
+    await expect(builder.execute()).rejects.toThrow("Builder already consumed");
+  });
+});

nodejs/__test__/remote.test.ts

@@ -7,7 +7,6 @@ import {
   ClientConfig,
   Connection,
   ConnectionOptions,
-  NativeJsHeaderProvider,
   TlsConfig,
   connect,
 } from "../lancedb";

nodejs/__test__/sanitize.test.ts

@@ -0,0 +1,184 @@
+// SPDX-License-Identifier: Apache-2.0
+// SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+import * as arrow from "../lancedb/arrow";
+import { sanitizeField, sanitizeType } from "../lancedb/sanitize";
+
+describe("sanitize", function () {
+  describe("sanitizeType function", function () {
+    it("should handle type objects", function () {
+      const type = new arrow.Int32();
+      const result = sanitizeType(type);
+
+      expect(result.typeId).toBe(arrow.Type.Int);
+      expect((result as arrow.Int).bitWidth).toBe(32);
+      expect((result as arrow.Int).isSigned).toBe(true);
+
+      const floatType = {
+        typeId: 3, // Type.Float = 3
+        precision: 2,
+        toString: () => "Float",
+        isFloat: true,
+        isFixedWidth: true,
+      };
+
+      const floatResult = sanitizeType(floatType);
+      expect(floatResult).toBeInstanceOf(arrow.DataType);
+      expect(floatResult.typeId).toBe(arrow.Type.Float);
+
+      const floatResult2 = sanitizeType({ ...floatType, typeId: () => 3 });
+      expect(floatResult2).toBeInstanceOf(arrow.DataType);
+      expect(floatResult2.typeId).toBe(arrow.Type.Float);
+    });
+
+    const allTypeNameTestCases = [
+      ["null", new arrow.Null()],
+      ["binary", new arrow.Binary()],
+      ["utf8", new arrow.Utf8()],
+      ["bool", new arrow.Bool()],
+      ["int8", new arrow.Int8()],
+      ["int16", new arrow.Int16()],
+      ["int32", new arrow.Int32()],
+      ["int64", new arrow.Int64()],
+      ["uint8", new arrow.Uint8()],
+      ["uint16", new arrow.Uint16()],
+      ["uint32", new arrow.Uint32()],
+      ["uint64", new arrow.Uint64()],
+      ["float16", new arrow.Float16()],
+      ["float32", new arrow.Float32()],
+      ["float64", new arrow.Float64()],
+      ["datemillisecond", new arrow.DateMillisecond()],
+      ["dateday", new arrow.DateDay()],
+      ["timenanosecond", new arrow.TimeNanosecond()],
+      ["timemicrosecond", new arrow.TimeMicrosecond()],
+      ["timemillisecond", new arrow.TimeMillisecond()],
+      ["timesecond", new arrow.TimeSecond()],
+      ["intervaldaytime", new arrow.IntervalDayTime()],
+      ["intervalyearmonth", new arrow.IntervalYearMonth()],
+      ["durationnanosecond", new arrow.DurationNanosecond()],
+      ["durationmicrosecond", new arrow.DurationMicrosecond()],
+      ["durationmillisecond", new arrow.DurationMillisecond()],
+      ["durationsecond", new arrow.DurationSecond()],
+    ] as const;
+
+    it.each(allTypeNameTestCases)(
+      'should map type name "%s" to %s',
+      function (name, expected) {
+        const result = sanitizeType(name);
+        expect(result).toBeInstanceOf(expected.constructor);
+      },
+    );
+
+    const caseVariationTestCases = [
+      ["NULL", new arrow.Null()],
+      ["Utf8", new arrow.Utf8()],
+      ["FLOAT32", new arrow.Float32()],
+      ["DaTedAy", new arrow.DateDay()],
+    ] as const;
+
+    it.each(caseVariationTestCases)(
+      'should be case insensitive for type name "%s" mapped to %s',
+      function (name, expected) {
+        const result = sanitizeType(name);
+        expect(result).toBeInstanceOf(expected.constructor);
+      },
+    );
+
+    it("should throw error for unrecognized type name", function () {
+      expect(() => sanitizeType("invalid_type")).toThrow(
+        "Unrecognized type name in schema: invalid_type",
+      );
+    });
+  });
+
+  describe("sanitizeField function", function () {
+    it("should handle field with string type name", function () {
+      const field = sanitizeField({
+        name: "string_field",
+        type: "utf8",
+        nullable: true,
+        metadata: new Map([["key", "value"]]),
+      });
+
+      expect(field).toBeInstanceOf(arrow.Field);
+      expect(field.name).toBe("string_field");
+      expect(field.type).toBeInstanceOf(arrow.Utf8);
+      expect(field.nullable).toBe(true);
+      expect(field.metadata?.get("key")).toBe("value");
+    });
+
+    it("should handle field with type object", function () {
+      const floatType = {
+        typeId: 3, // Float
+        precision: 32,
+      };
+
+      const field = sanitizeField({
+        name: "float_field",
+        type: floatType,
+        nullable: false,
+      });
+
+      expect(field).toBeInstanceOf(arrow.Field);
+      expect(field.name).toBe("float_field");
+      expect(field.type).toBeInstanceOf(arrow.DataType);
+      expect(field.type.typeId).toBe(arrow.Type.Float);
+      expect((field.type as arrow.Float64).precision).toBe(32);
+      expect(field.nullable).toBe(false);
+    });
+
+    it("should handle field with direct Type instance", function () {
+      const field = sanitizeField({
+        name: "bool_field",
+        type: new arrow.Bool(),
+        nullable: true,
+      });
+
+      expect(field).toBeInstanceOf(arrow.Field);
+      expect(field.name).toBe("bool_field");
+      expect(field.type).toBeInstanceOf(arrow.Bool);
+      expect(field.nullable).toBe(true);
+    });
+
+    it("should throw error for invalid field object", function () {
+      expect(() =>
+        sanitizeField({
+          type: "int32",
+          nullable: true,
+        }),
+      ).toThrow(
+        "The field passed in is missing a `type`/`name`/`nullable` property",
+      );
+
+      // Invalid type
+      expect(() =>
+        sanitizeField({
+          name: "invalid",
+          type: { invalid: true },
+          nullable: true,
+        }),
+      ).toThrow("Expected a Type to have a typeId property");
+
+      // Invalid nullable
+      expect(() =>
+        sanitizeField({
+          name: "invalid_nullable",
+          type: "int32",
+          nullable: "not a boolean",
+        }),
+      ).toThrow("The field passed in had a non-boolean `nullable` property");
+    });
+
+    it("should report error for invalid type name", function () {
+      expect(() =>
+        sanitizeField({
+          name: "invalid_field",
+          type: "invalid_type",
+          nullable: true,
+        }),
+      ).toThrow(
+        "Unable to sanitize type for field: invalid_field due to error: Error: Unrecognized type name in schema: invalid_type",
+      );
+    });
+  });
+});

nodejs/__test__/table.test.ts

@@ -10,7 +10,13 @@ import * as arrow16 from "apache-arrow-16";
 import * as arrow17 from "apache-arrow-17";
 import * as arrow18 from "apache-arrow-18";
 
-import { MatchQuery, PhraseQuery, Table, connect } from "../lancedb";
+import {
+  Connection,
+  MatchQuery,
+  PhraseQuery,
+  Table,
+  connect,
+} from "../lancedb";
 import {
   Table as ArrowTable,
   Field,
@@ -21,6 +27,8 @@ import {
   Int64,
   List,
   Schema,
+  SchemaLike,
+  Type,
   Uint8,
   Utf8,
   makeArrowTable,
@@ -39,7 +47,6 @@ import {
   Operator,
   instanceOfFullTextQuery,
 } from "../lancedb/query";
-import exp = require("constants");
 
 describe.each([arrow15, arrow16, arrow17, arrow18])(
   "Given a table",
@@ -212,8 +219,7 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
       },
     );
 
-    // TODO: https://github.com/lancedb/lancedb/issues/1832
+    it("should be able to omit nullable fields", async () => {
-    it.skip("should be able to omit nullable fields", async () => {
       const db = await connect(tmpDir.name);
       const schema = new arrow.Schema([
         new arrow.Field(
@@ -237,23 +243,36 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
       await table.add([data3]);
 
       let res = await table.query().limit(10).toArray();
-      const resVector = res.map((r) => r.get("vector").toArray());
+      const resVector = res.map((r) =>
+        r.vector ? Array.from(r.vector) : null,
+      );
       expect(resVector).toEqual([null, data2.vector, data3.vector]);
-      const resItem = res.map((r) => r.get("item").toArray());
+      const resItem = res.map((r) => r.item);
       expect(resItem).toEqual(["foo", null, "bar"]);
-      const resPrice = res.map((r) => r.get("price").toArray());
+      const resPrice = res.map((r) => r.price);
       expect(resPrice).toEqual([10.0, 2.0, 3.0]);
 
       const data4 = { item: "foo" };
       // We can't omit a column if it's not nullable
-      await expect(table.add([data4])).rejects.toThrow("Invalid user input");
+      await expect(table.add([data4])).rejects.toThrow(
+        "Append with different schema",
+      );
 
       // But we can alter columns to make them nullable
       await table.alterColumns([{ path: "price", nullable: true }]);
       await table.add([data4]);
 
-      res = (await table.query().limit(10).toArray()).map((r) => r.toJSON());
+      res = (await table.query().limit(10).toArray()).map((r) => ({
-      expect(res).toEqual([data1, data2, data3, data4]);
+        ...r.toJSON(),
+        vector: r.vector ? Array.from(r.vector) : null,
+      }));
+      // Rust fills missing nullable fields with null
+      expect(res).toEqual([
+        { ...data1, vector: null },
+        { ...data2, item: null },
+        data3,
+        { ...data4, price: null, vector: null },
+      ]);
     });
 
     it("should be able to insert nullable data for non-nullable fields", async () => {
@@ -331,6 +350,43 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
       const table = await db.createTable("my_table", data);
       expect(await table.countRows()).toEqual(2);
     });
+
+    it("should allow undefined and omitted nullable vector fields", async () => {
+      // Test for the bug: can't pass undefined or omit vector column
+      const db = await connect("memory://");
+      const schema = new arrow.Schema([
+        new arrow.Field("id", new arrow.Int32(), true),
+        new arrow.Field(
+          "vector",
+          new arrow.FixedSizeList(
+            32,
+            new arrow.Field("item", new arrow.Float32(), true),
+          ),
+          true, // nullable = true
+        ),
+      ]);
+      const table = await db.createEmptyTable("test_table", schema);
+
+      // Should not throw error for undefined value
+      await table.add([{ id: 0, vector: undefined }]);
+
+      // Should not throw error for omitted field
+      await table.add([{ id: 1 }]);
+
+      // Should still work for null
+      await table.add([{ id: 2, vector: null }]);
+
+      // Should still work for actual vector
+      const testVector = new Array(32).fill(0.5);
+      await table.add([{ id: 3, vector: testVector }]);
+      expect(await table.countRows()).toEqual(4);
+
+      const res = await table.query().limit(10).toArray();
+      const resVector = res.map((r) =>
+        r.vector ? Array.from(r.vector) : null,
+      );
+      expect(resVector).toEqual([null, null, null, testVector]);
+    });
   },
 );
 
@@ -488,6 +544,32 @@ describe("merge insert", () => {
         .execute(newData, { timeoutMs: 0 }),
     ).rejects.toThrow("merge insert timed out");
   });
+
+  test("useIndex", async () => {
+    const newData = [
+      { a: 2, b: "x" },
+      { a: 4, b: "z" },
+    ];
+
+    // Test with useIndex(true) - should work fine
+    const result1 = await table
+      .mergeInsert("a")
+      .whenNotMatchedInsertAll()
+      .useIndex(true)
+      .execute(newData);
+
+    expect(result1.numInsertedRows).toBe(1); // Only a=4 should be inserted
+
+    // Test with useIndex(false) - should also work fine
+    const newData2 = [{ a: 5, b: "w" }];
+    const result2 = await table
+      .mergeInsert("a")
+      .whenNotMatchedInsertAll()
+      .useIndex(false)
+      .execute(newData2);
+
+    expect(result2.numInsertedRows).toBe(1); // a=5 should be inserted
+  });
 });
 
 describe("When creating an index", () => {
@@ -779,6 +861,15 @@ describe("When creating an index", () => {
     });
   });
 
+  it("should be able to create IVF_RQ", async () => {
+    await tbl.createIndex("vec", {
+      config: Index.ivfRq({
+        numPartitions: 10,
+        numBits: 1,
+      }),
+    });
+  });
+
   it("should allow me to replace (or not) an existing index", async () => {
     await tbl.createIndex("id");
     // Default is replace=true
@@ -1429,7 +1520,9 @@ describe("when optimizing a dataset", () => {
 
   it("delete unverified", async () => {
     const version = await table.version();
-    const versionFile = `${tmpDir.name}/${table.name}.lance/_versions/${version - 1}.manifest`;
+    const versionFile = `${tmpDir.name}/${table.name}.lance/_versions/${
+      version - 1
+    }.manifest`;
     fs.rmSync(versionFile);
 
     let stats = await table.optimize({ deleteUnverified: false });
@@ -1943,3 +2036,52 @@ describe("column name options", () => {
     expect(results2.length).toBe(10);
   });
 });
+
+describe("when creating an empty table", () => {
+  let con: Connection;
+  beforeEach(async () => {
+    const tmpDir = tmp.dirSync({ unsafeCleanup: true });
+    con = await connect(tmpDir.name);
+  });
+  afterEach(() => {
+    con.close();
+  });
+
+  it("can create an empty table from an arrow Schema", async () => {
+    const schema = new Schema([
+      new Field("id", new Int64()),
+      new Field("vector", new Float64()),
+    ]);
+    const table = await con.createEmptyTable("test", schema);
+    const actualSchema = await table.schema();
+    expect(actualSchema.fields[0].type.typeId).toBe(Type.Int);
+    expect((actualSchema.fields[0].type as Int64).bitWidth).toBe(64);
+    expect(actualSchema.fields[1].type.typeId).toBe(Type.Float);
+    expect((actualSchema.fields[1].type as Float64).precision).toBe(2);
+  });
+
+  it("can create an empty table from schema that specifies field types by name", async () => {
+    const schemaLike = {
+      fields: [
+        {
+          name: "id",
+          type: "int64",
+          nullable: true,
+        },
+        {
+          name: "vector",
+          type: "float64",
+          nullable: true,
+        },
+      ],
+      metadata: new Map(),
+      names: ["id", "vector"],
+    } satisfies SchemaLike;
+    const table = await con.createEmptyTable("test", schemaLike);
+    const actualSchema = await table.schema();
+    expect(actualSchema.fields[0].type.typeId).toBe(Type.Int);
+    expect((actualSchema.fields[0].type as Int64).bitWidth).toBe(64);
+    expect(actualSchema.fields[1].type.typeId).toBe(Type.Float);
+    expect((actualSchema.fields[1].type as Float64).precision).toBe(2);
+  });
+});

nodejs/biome.json

@@ -48,6 +48,7 @@
         "noUnreachableSuper": "error",
         "noUnsafeFinally": "error",
         "noUnsafeOptionalChaining": "error",
+        "noUnusedImports": "error",
         "noUnusedLabels": "error",
         "noUnusedVariables": "warn",
         "useIsNan": "error",

nodejs/lancedb/arrow.ts

@@ -41,7 +41,6 @@ import {
   vectorFromArray as badVectorFromArray,
   makeBuilder,
   makeData,
-  makeTable,
 } from "apache-arrow";
 import { Buffers } from "apache-arrow/data";
 import { type EmbeddingFunction } from "./embedding/embedding_function";
@@ -74,7 +73,7 @@ export type FieldLike =
   | {
       type: string;
       name: string;
-      nullable?: boolean;
+      nullable: boolean;
       metadata?: Map<string, string>;
     };
 
@@ -279,7 +278,7 @@ export class MakeArrowTableOptions {
 }
 
 /**
- * An enhanced version of the {@link makeTable} function from Apache Arrow
+ * An enhanced version of the apache-arrow makeTable function from Apache Arrow
  * that supports nested fields and embeddings columns.
  *
  * (typically you do not need to call this function.  It will be called automatically
@@ -512,7 +511,11 @@ function* rowPathsAndValues(
     if (isObject(value)) {
       yield* rowPathsAndValues(value, [...basePath, key]);
     } else {
-      yield [[...basePath, key], value];
+      // Skip undefined values - they should be treated the same as missing fields
+      // for embedding function purposes
+      if (value !== undefined) {
+        yield [[...basePath, key], value];
+      }
     }
   }
 }
@@ -701,7 +704,7 @@ function transposeData(
       }
       return current;
     });
-    return makeVector(values, field.type);
+    return makeVector(values, field.type, undefined, field.nullable);
   }
 }
 
@@ -748,9 +751,30 @@ function makeVector(
   values: unknown[],
   type?: DataType,
   stringAsDictionary?: boolean,
+  nullable?: boolean,
   // biome-ignore lint/suspicious/noExplicitAny: skip
 ): Vector<any> {
   if (type !== undefined) {
+    // Convert undefined values to null for nullable fields
+    if (nullable) {
+      values = values.map((v) => (v === undefined ? null : v));
+    }
+
+    // workaround for: https://github.com/apache/arrow-js/issues/68
+    if (DataType.isBool(type)) {
+      const hasNonNullValue = values.some((v) => v !== null && v !== undefined);
+      if (!hasNonNullValue) {
+        const nullBitmap = new Uint8Array(Math.ceil(values.length / 8));
+        const data = makeData({
+          type: type,
+          length: values.length,
+          nullCount: values.length,
+          nullBitmap,
+        });
+        return arrowMakeVector(data);
+      }
+    }
+
     // No need for inference, let Arrow create it
     if (type instanceof Int) {
       if (DataType.isInt(type) && type.bitWidth === 64) {
@@ -875,7 +899,12 @@ async function applyEmbeddingsFromMetadata(
   for (const field of schema.fields) {
     if (!(field.name in columns)) {
       const nullValues = new Array(table.numRows).fill(null);
-      columns[field.name] = makeVector(nullValues, field.type);
+      columns[field.name] = makeVector(
+        nullValues,
+        field.type,
+        undefined,
+        field.nullable,
+      );
     }
   }
 
@@ -939,7 +968,12 @@ async function applyEmbeddings<T>(
     } else if (schema != null) {
       const destField = schema.fields.find((f) => f.name === destColumn);
       if (destField != null) {
-        newColumns[destColumn] = makeVector([], destField.type);
+        newColumns[destColumn] = makeVector(
+          [],
+          destField.type,
+          undefined,
+          destField.nullable,
+        );
       } else {
         throw new Error(
           `Attempt to apply embeddings to an empty table failed because schema was missing embedding column '${destColumn}'`,
@@ -1251,19 +1285,36 @@ function validateSchemaEmbeddings(
     if (isFixedSizeList(field.type)) {
       field = sanitizeField(field);
       if (data.length !== 0 && data?.[0]?.[field.name] === undefined) {
+        // Check if there's an embedding function registered for this field
+        let hasEmbeddingFunction = false;
+
+        // Check schema metadata for embedding functions
         if (schema.metadata.has("embedding_functions")) {
           const embeddings = JSON.parse(
             schema.metadata.get("embedding_functions")!,
           );
-          if (
+          // biome-ignore lint/suspicious/noExplicitAny: we don't know the type of `f`
-            // biome-ignore lint/suspicious/noExplicitAny: we don't know the type of `f`
+          if (embeddings.find((f: any) => f["vectorColumn"] === field.name)) {
-            embeddings.find((f: any) => f["vectorColumn"] === field.name) ===
+            hasEmbeddingFunction = true;
-            undefined
+          }
-          ) {
+        }
+
+        // Check passed embedding function parameter
+        if (embeddings && embeddings.vectorColumn === field.name) {
+          hasEmbeddingFunction = true;
+        }
+
+        // If the field is nullable AND there's no embedding function, allow undefined/omitted values
+        if (field.nullable && !hasEmbeddingFunction) {
+          fields.push(field);
+        } else {
+          // Either not nullable OR has embedding function - require explicit values
+          if (hasEmbeddingFunction) {
+            // Don't add to missingEmbeddingFields since this is expected to be filled by embedding function
+            fields.push(field);
+          } else {
             missingEmbeddingFields.push(field);
           }
-        } else {
-          missingEmbeddingFields.push(field);
         }
       } else {
         fields.push(field);

nodejs/lancedb/connection.ts

@@ -3,7 +3,6 @@
 
 import {
   Data,
-  Schema,
   SchemaLike,
   TableLike,
   fromTableToStreamBuffer,
@@ -268,6 +267,33 @@ export abstract class Connection {
    * @param {string[]} namespace The namespace to drop tables from (defaults to root namespace).
    */
   abstract dropAllTables(namespace?: string[]): Promise<void>;
+
+  /**
+   * Clone a table from a source table.
+   *
+   * A shallow clone creates a new table that shares the underlying data files
+   * with the source table but has its own independent manifest. This allows
+   * both the source and cloned tables to evolve independently while initially
+   * sharing the same data, deletion, and index files.
+   *
+   * @param {string} targetTableName - The name of the target table to create.
+   * @param {string} sourceUri - The URI of the source table to clone from.
+   * @param {object} options - Clone options.
+   * @param {string[]} options.targetNamespace - The namespace for the target table (defaults to root namespace).
+   * @param {number} options.sourceVersion - The version of the source table to clone.
+   * @param {string} options.sourceTag - The tag of the source table to clone.
+   * @param {boolean} options.isShallow - Whether to perform a shallow clone (defaults to true).
+   */
+  abstract cloneTable(
+    targetTableName: string,
+    sourceUri: string,
+    options?: {
+      targetNamespace?: string[];
+      sourceVersion?: number;
+      sourceTag?: string;
+      isShallow?: boolean;
+    },
+  ): Promise<Table>;
 }
 
 /** @hideconstructor */
@@ -332,6 +358,28 @@ export class LocalConnection extends Connection {
     return new LocalTable(innerTable);
   }
 
+  async cloneTable(
+    targetTableName: string,
+    sourceUri: string,
+    options?: {
+      targetNamespace?: string[];
+      sourceVersion?: number;
+      sourceTag?: string;
+      isShallow?: boolean;
+    },
+  ): Promise<Table> {
+    const innerTable = await this.inner.cloneTable(
+      targetTableName,
+      sourceUri,
+      options?.targetNamespace ?? [],
+      options?.sourceVersion ?? null,
+      options?.sourceTag ?? null,
+      options?.isShallow ?? true,
+    );
+
+    return new LocalTable(innerTable);
+  }
+
   private getStorageOptions(
     options?: Partial<CreateTableOptions>,
   ): Record<string, string> | undefined {

nodejs/lancedb/index.ts

@@ -43,6 +43,10 @@ export {
   DeleteResult,
   DropColumnsResult,
   UpdateResult,
+  SplitRandomOptions,
+  SplitHashOptions,
+  SplitSequentialOptions,
+  ShuffleOptions,
 } from "./native.js";
 
 export {
@@ -85,6 +89,7 @@ export {
   Index,
   IndexOptions,
   IvfPqOptions,
+  IvfRqOptions,
   IvfFlatOptions,
   HnswPqOptions,
   HnswSqOptions,
@@ -110,6 +115,7 @@ export {
 export { MergeInsertBuilder, WriteExecutionOptions } from "./merge";
 
 export * as embedding from "./embedding";
+export { permutationBuilder, PermutationBuilder } from "./permutation";
 export * as rerankers from "./rerankers";
 export {
   SchemaLike,

nodejs/lancedb/indices.ts

@@ -112,6 +112,77 @@ export interface IvfPqOptions {
   sampleRate?: number;
 }
 
+export interface IvfRqOptions {
+  /**
+   * The number of IVF partitions to create.
+   *
+   * This value should generally scale with the number of rows in the dataset.
+   * By default the number of partitions is the square root of the number of
+   * rows.
+   *
+   * If this value is too large then the first part of the search (picking the
+   * right partition) will be slow. If this value is too small then the second
+   * part of the search (searching within a partition) will be slow.
+   */
+  numPartitions?: number;
+
+  /**
+   * Number of bits per dimension for residual quantization.
+   *
+   * This value controls how much each residual component is compressed. The more
+   * bits, the more accurate the index will be but the slower search. Typical values
+   * are small integers; the default is 1 bit per dimension.
+   */
+  numBits?: number;
+
+  /**
+   * Distance type to use to build the index.
+   *
+   * Default value is "l2".
+   *
+   * This is used when training the index to calculate the IVF partitions
+   * (vectors are grouped in partitions with similar vectors according to this
+   * distance type) and during quantization.
+   *
+   * The distance type used to train an index MUST match the distance type used
+   * to search the index. Failure to do so will yield inaccurate results.
+   *
+   * The following distance types are available:
+   *
+   * "l2" - Euclidean distance.
+   * "cosine" - Cosine distance.
+   * "dot" - Dot product.
+   */
+  distanceType?: "l2" | "cosine" | "dot";
+
+  /**
+   * Max iterations to train IVF kmeans.
+   *
+   * When training an IVF index we use kmeans to calculate the partitions. This parameter
+   * controls how many iterations of kmeans to run.
+   *
+   * The default value is 50.
+   */
+  maxIterations?: number;
+
+  /**
+   * The number of vectors, per partition, to sample when training IVF kmeans.
+   *
+   * When an IVF index is trained, we need to calculate partitions. These are groups
+   * of vectors that are similar to each other. To do this we use an algorithm called kmeans.
+   *
+   * Running kmeans on a large dataset can be slow. To speed this up we run kmeans on a
+   * random sample of the data. This parameter controls the size of the sample. The total
+   * number of vectors used to train the index is `sample_rate * num_partitions`.
+   *
+   * Increasing this value might improve the quality of the index but in most cases the
+   * default should be sufficient.
+   *
+   * The default value is 256.
+   */
+  sampleRate?: number;
+}
+
 /**
  * Options to create an `HNSW_PQ` index
  */
@@ -523,6 +594,35 @@ export class Index {
         options?.distanceType,
         options?.numPartitions,
         options?.numSubVectors,
+        options?.numBits,
+        options?.maxIterations,
+        options?.sampleRate,
+      ),
+    );
+  }
+
+  /**
+   * Create an IvfRq index
+   *
+   * IVF-RQ (RabitQ Quantization) compresses vectors using RabitQ quantization
+   * and organizes them into IVF partitions.
+   *
+   * The compression scheme is called RabitQ quantization. Each dimension is quantized into a small number of bits.
+   * The parameters `num_bits` and `num_partitions` control this process, providing a tradeoff
+   * between index size (and thus search speed) and index accuracy.
+   *
+   * The partitioning process is called IVF and the `num_partitions` parameter controls how
+   * many groups to create.
+   *
+   * Note that training an IVF RQ index on a large dataset is a slow operation and
+   * currently is also a memory intensive operation.
+   */
+  static ivfRq(options?: Partial<IvfRqOptions>) {
+    return new Index(
+      LanceDbIndex.ivfRq(
+        options?.distanceType,
+        options?.numPartitions,
+        options?.numBits,
         options?.maxIterations,
         options?.sampleRate,
       ),

nodejs/lancedb/merge.ts

@@ -70,6 +70,23 @@ export class MergeInsertBuilder {
       this.#schema,
     );
   }
+
+  /**
+   * Controls whether to use indexes for the merge operation.
+   *
+   * When set to `true` (the default), the operation will use an index if available
+   * on the join key for improved performance. When set to `false`, it forces a full
+   * table scan even if an index exists. This can be useful for benchmarking or when
+   * the query optimizer chooses a suboptimal path.
+   *
+   * @param useIndex - Whether to use indices for the merge operation. Defaults to `true`.
+   */
+  useIndex(useIndex: boolean): MergeInsertBuilder {
+    return new MergeInsertBuilder(
+      this.#native.useIndex(useIndex),
+      this.#schema,
+    );
+  }
   /**
    * Executes the merge insert operation
    *

nodejs/lancedb/permutation.ts

@@ -0,0 +1,188 @@
+// SPDX-License-Identifier: Apache-2.0
+// SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+import {
+  PermutationBuilder as NativePermutationBuilder,
+  Table as NativeTable,
+  ShuffleOptions,
+  SplitHashOptions,
+  SplitRandomOptions,
+  SplitSequentialOptions,
+  permutationBuilder as nativePermutationBuilder,
+} from "./native.js";
+import { LocalTable, Table } from "./table";
+
+/**
+ * A PermutationBuilder for creating data permutations with splits, shuffling, and filtering.
+ *
+ * This class provides a TypeScript wrapper around the native Rust PermutationBuilder,
+ * offering methods to configure data splits, shuffling, and filtering before executing
+ * the permutation to create a new table.
+ */
+export class PermutationBuilder {
+  private inner: NativePermutationBuilder;
+
+  /**
+   * @hidden
+   */
+  constructor(inner: NativePermutationBuilder) {
+    this.inner = inner;
+  }
+
+  /**
+   * Configure random splits for the permutation.
+   *
+   * @param options - Configuration for random splitting
+   * @returns A new PermutationBuilder instance
+   * @example
+   * ```ts
+   * // Split by ratios
+   * builder.splitRandom({ ratios: [0.7, 0.3], seed: 42 });
+   *
+   * // Split by counts
+   * builder.splitRandom({ counts: [1000, 500], seed: 42 });
+   *
+   * // Split with fixed size
+   * builder.splitRandom({ fixed: 100, seed: 42 });
+   * ```
+   */
+  splitRandom(options: SplitRandomOptions): PermutationBuilder {
+    const newInner = this.inner.splitRandom(options);
+    return new PermutationBuilder(newInner);
+  }
+
+  /**
+   * Configure hash-based splits for the permutation.
+   *
+   * @param options - Configuration for hash-based splitting
+   * @returns A new PermutationBuilder instance
+   * @example
+   * ```ts
+   * builder.splitHash({
+   *   columns: ["user_id"],
+   *   splitWeights: [70, 30],
+   *   discardWeight: 0
+   * });
+   * ```
+   */
+  splitHash(options: SplitHashOptions): PermutationBuilder {
+    const newInner = this.inner.splitHash(options);
+    return new PermutationBuilder(newInner);
+  }
+
+  /**
+   * Configure sequential splits for the permutation.
+   *
+   * @param options - Configuration for sequential splitting
+   * @returns A new PermutationBuilder instance
+   * @example
+   * ```ts
+   * // Split by ratios
+   * builder.splitSequential({ ratios: [0.8, 0.2] });
+   *
+   * // Split by counts
+   * builder.splitSequential({ counts: [800, 200] });
+   *
+   * // Split with fixed size
+   * builder.splitSequential({ fixed: 1000 });
+   * ```
+   */
+  splitSequential(options: SplitSequentialOptions): PermutationBuilder {
+    const newInner = this.inner.splitSequential(options);
+    return new PermutationBuilder(newInner);
+  }
+
+  /**
+   * Configure calculated splits for the permutation.
+   *
+   * @param calculation - SQL expression for calculating splits
+   * @returns A new PermutationBuilder instance
+   * @example
+   * ```ts
+   * builder.splitCalculated("user_id % 3");
+   * ```
+   */
+  splitCalculated(calculation: string): PermutationBuilder {
+    const newInner = this.inner.splitCalculated(calculation);
+    return new PermutationBuilder(newInner);
+  }
+
+  /**
+   * Configure shuffling for the permutation.
+   *
+   * @param options - Configuration for shuffling
+   * @returns A new PermutationBuilder instance
+   * @example
+   * ```ts
+   * // Basic shuffle
+   * builder.shuffle({ seed: 42 });
+   *
+   * // Shuffle with clump size
+   * builder.shuffle({ seed: 42, clumpSize: 10 });
+   * ```
+   */
+  shuffle(options: ShuffleOptions): PermutationBuilder {
+    const newInner = this.inner.shuffle(options);
+    return new PermutationBuilder(newInner);
+  }
+
+  /**
+   * Configure filtering for the permutation.
+   *
+   * @param filter - SQL filter expression
+   * @returns A new PermutationBuilder instance
+   * @example
+   * ```ts
+   * builder.filter("age > 18 AND status = 'active'");
+   * ```
+   */
+  filter(filter: string): PermutationBuilder {
+    const newInner = this.inner.filter(filter);
+    return new PermutationBuilder(newInner);
+  }
+
+  /**
+   * Execute the permutation and create the destination table.
+   *
+   * @returns A Promise that resolves to the new Table instance
+   * @example
+   * ```ts
+   * const permutationTable = await builder.execute();
+   * console.log(`Created table: ${permutationTable.name}`);
+   * ```
+   */
+  async execute(): Promise<Table> {
+    const nativeTable: NativeTable = await this.inner.execute();
+    return new LocalTable(nativeTable);
+  }
+}
+
+/**
+ * Create a permutation builder for the given table.
+ *
+ * @param table - The source table to create a permutation from
+ * @param destTableName - The name for the destination permutation table
+ * @returns A PermutationBuilder instance
+ * @example
+ * ```ts
+ * const builder = permutationBuilder(sourceTable, "training_data")
+ *   .splitRandom({ ratios: [0.8, 0.2], seed: 42 })
+ *   .shuffle({ seed: 123 });
+ *
+ * const trainingTable = await builder.execute();
+ * ```
+ */
+export function permutationBuilder(
+  table: Table,
+  destTableName: string,
+): PermutationBuilder {
+  // Extract the inner native table from the TypeScript wrapper
+  const localTable = table as LocalTable;
+  // Access inner through type assertion since it's private
+  const nativeBuilder = nativePermutationBuilder(
+    // biome-ignore lint/suspicious/noExplicitAny: need access to private variable
+    (localTable as any).inner,
+    destTableName,
+  );
+  return new PermutationBuilder(nativeBuilder);
+}

nodejs/lancedb/sanitize.ts

@@ -326,6 +326,9 @@ export function sanitizeDictionary(typeLike: object) {
 
 // biome-ignore lint/suspicious/noExplicitAny: skip
 export function sanitizeType(typeLike: unknown): DataType<any> {
+  if (typeof typeLike === "string") {
+    return dataTypeFromName(typeLike);
+  }
   if (typeof typeLike !== "object" || typeLike === null) {
     throw Error("Expected a Type but object was null/undefined");
   }
@@ -447,7 +450,7 @@ export function sanitizeType(typeLike: unknown): DataType<any> {
     case Type.DurationSecond:
       return new DurationSecond();
     default:
-      throw new Error("Unrecoginized type id in schema: " + typeId);
+      throw new Error("Unrecognized type id in schema: " + typeId);
   }
 }
 
@@ -467,7 +470,15 @@ export function sanitizeField(fieldLike: unknown): Field {
       "The field passed in is missing a `type`/`name`/`nullable` property",
     );
   }
-  const type = sanitizeType(fieldLike.type);
+  let type: DataType;
+  try {
+    type = sanitizeType(fieldLike.type);
+  } catch (error: unknown) {
+    throw Error(
+      `Unable to sanitize type for field: ${fieldLike.name} due to error: ${error}`,
+      { cause: error },
+    );
+  }
   const name = fieldLike.name;
   if (!(typeof name === "string")) {
     throw Error("The field passed in had a non-string `name` property");
@@ -581,3 +592,46 @@ function sanitizeData(
     },
   );
 }
+
+const constructorsByTypeName = {
+  null: () => new Null(),
+  binary: () => new Binary(),
+  utf8: () => new Utf8(),
+  bool: () => new Bool(),
+  int8: () => new Int8(),
+  int16: () => new Int16(),
+  int32: () => new Int32(),
+  int64: () => new Int64(),
+  uint8: () => new Uint8(),
+  uint16: () => new Uint16(),
+  uint32: () => new Uint32(),
+  uint64: () => new Uint64(),
+  float16: () => new Float16(),
+  float32: () => new Float32(),
+  float64: () => new Float64(),
+  datemillisecond: () => new DateMillisecond(),
+  dateday: () => new DateDay(),
+  timenanosecond: () => new TimeNanosecond(),
+  timemicrosecond: () => new TimeMicrosecond(),
+  timemillisecond: () => new TimeMillisecond(),
+  timesecond: () => new TimeSecond(),
+  intervaldaytime: () => new IntervalDayTime(),
+  intervalyearmonth: () => new IntervalYearMonth(),
+  durationnanosecond: () => new DurationNanosecond(),
+  durationmicrosecond: () => new DurationMicrosecond(),
+  durationmillisecond: () => new DurationMillisecond(),
+  durationsecond: () => new DurationSecond(),
+} as const;
+
+type MappableTypeName = keyof typeof constructorsByTypeName;
+
+export function dataTypeFromName(typeName: string): DataType {
+  const normalizedTypeName = typeName.toLowerCase() as MappableTypeName;
+  const _constructor = constructorsByTypeName[normalizedTypeName];
+
+  if (!_constructor) {
+    throw new Error("Unrecognized type name in schema: " + typeName);
+  }
+
+  return _constructor();
+}

nodejs/npm/darwin-arm64/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-darwin-arm64",
-	"version": "0.22.1-beta.0",
+	"version": "0.22.2",
 	"os": ["darwin"],
 	"cpu": ["arm64"],
 	"main": "lancedb.darwin-arm64.node",

nodejs/npm/darwin-x64/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-darwin-x64",
-	"version": "0.22.1-beta.0",
+	"version": "0.22.2",
 	"os": ["darwin"],
 	"cpu": ["x64"],
 	"main": "lancedb.darwin-x64.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-gnu",
-	"version": "0.22.1-beta.0",
+	"version": "0.22.2",
 	"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": "0.22.1-beta.0",
+	"version": "0.22.2",
 	"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": "0.22.1-beta.0",
+	"version": "0.22.2",
 	"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": "0.22.1-beta.0",
+	"version": "0.22.2",
 	"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": "0.22.1-beta.0",
+  "version": "0.22.2",
   "os": [
     "win32"
   ],

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-win32-x64-msvc",
-	"version": "0.22.1-beta.0",
+	"version": "0.22.2",
 	"os": ["win32"],
 	"cpu": ["x64"],
 	"main": "lancedb.win32-x64-msvc.node",

nodejs/package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "@lancedb/lancedb",
-  "version": "0.22.1-beta.0",
+  "version": "0.22.2",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "@lancedb/lancedb",
-      "version": "0.22.1-beta.0",
+      "version": "0.22.2",
       "cpu": [
         "x64",
         "arm64"

nodejs/package.json

@@ -11,7 +11,7 @@
     "ann"
   ],
   "private": false,
-  "version": "0.22.1-beta.0",
+  "version": "0.22.2",
   "main": "dist/index.js",
   "exports": {
     ".": "./dist/index.js",

nodejs/src/connection.rs

@@ -213,6 +213,36 @@ impl Connection {
         Ok(Table::new(tbl))
     }
 
+    #[napi(catch_unwind)]
+    pub async fn clone_table(
+        &self,
+        target_table_name: String,
+        source_uri: String,
+        target_namespace: Vec<String>,
+        source_version: Option<i64>,
+        source_tag: Option<String>,
+        is_shallow: bool,
+    ) -> napi::Result<Table> {
+        let mut builder = self
+            .get_inner()?
+            .clone_table(&target_table_name, &source_uri);
+
+        builder = builder.target_namespace(target_namespace);
+
+        if let Some(version) = source_version {
+            builder = builder.source_version(version as u64);
+        }
+
+        if let Some(tag) = source_tag {
+            builder = builder.source_tag(tag);
+        }
+
+        builder = builder.is_shallow(is_shallow);
+
+        let tbl = builder.execute().await.default_error()?;
+        Ok(Table::new(tbl))
+    }
+
     /// Drop table with the name. Or raise an error if the table does not exist.
     #[napi(catch_unwind)]
     pub async fn drop_table(&self, name: String, namespace: Vec<String>) -> napi::Result<()> {

nodejs/src/index.rs

@@ -6,6 +6,7 @@ use std::sync::Mutex;
 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;
@@ -65,6 +66,36 @@ impl Index {
         })
     }
 
+    #[napi(factory)]
+    pub fn ivf_rq(
+        distance_type: Option<String>,
+        num_partitions: Option<u32>,
+        num_bits: Option<u32>,
+        max_iterations: Option<u32>,
+        sample_rate: Option<u32>,
+    ) -> napi::Result<Self> {
+        let mut ivf_rq_builder = IvfRqIndexBuilder::default();
+        if let Some(distance_type) = distance_type {
+            let distance_type = parse_distance_type(distance_type)?;
+            ivf_rq_builder = ivf_rq_builder.distance_type(distance_type);
+        }
+        if let Some(num_partitions) = num_partitions {
+            ivf_rq_builder = ivf_rq_builder.num_partitions(num_partitions);
+        }
+        if let Some(num_bits) = num_bits {
+            ivf_rq_builder = ivf_rq_builder.num_bits(num_bits);
+        }
+        if let Some(max_iterations) = max_iterations {
+            ivf_rq_builder = ivf_rq_builder.max_iterations(max_iterations);
+        }
+        if let Some(sample_rate) = sample_rate {
+            ivf_rq_builder = ivf_rq_builder.sample_rate(sample_rate);
+        }
+        Ok(Self {
+            inner: Mutex::new(Some(LanceDbIndex::IvfRq(ivf_rq_builder))),
+        })
+    }
+
     #[napi(factory)]
     pub fn ivf_flat(
         distance_type: Option<String>,

nodejs/src/lib.rs

@@ -12,6 +12,7 @@ mod header;
 mod index;
 mod iterator;
 pub mod merge;
+pub mod permutation;
 mod query;
 pub mod remote;
 mod rerankers;

nodejs/src/merge.rs

@@ -43,6 +43,13 @@ impl NativeMergeInsertBuilder {
         self.inner.timeout(Duration::from_millis(timeout as u64));
     }
 
+    #[napi]
+    pub fn use_index(&self, use_index: bool) -> Self {
+        let mut this = self.clone();
+        this.inner.use_index(use_index);
+        this
+    }
+
     #[napi(catch_unwind)]
     pub async fn execute(&self, buf: Buffer) -> napi::Result<MergeResult> {
         let data = ipc_file_to_batches(buf.to_vec())

nodejs/src/permutation.rs

@@ -0,0 +1,222 @@
+// SPDX-License-Identifier: Apache-2.0
+// SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+use std::sync::{Arc, Mutex};
+
+use crate::{error::NapiErrorExt, table::Table};
+use lancedb::dataloader::{
+    permutation::{PermutationBuilder as LancePermutationBuilder, ShuffleStrategy},
+    split::{SplitSizes, SplitStrategy},
+};
+use napi_derive::napi;
+
+#[napi(object)]
+pub struct SplitRandomOptions {
+    pub ratios: Option<Vec<f64>>,
+    pub counts: Option<Vec<i64>>,
+    pub fixed: Option<i64>,
+    pub seed: Option<i64>,
+}
+
+#[napi(object)]
+pub struct SplitHashOptions {
+    pub columns: Vec<String>,
+    pub split_weights: Vec<i64>,
+    pub discard_weight: Option<i64>,
+}
+
+#[napi(object)]
+pub struct SplitSequentialOptions {
+    pub ratios: Option<Vec<f64>>,
+    pub counts: Option<Vec<i64>>,
+    pub fixed: Option<i64>,
+}
+
+#[napi(object)]
+pub struct ShuffleOptions {
+    pub seed: Option<i64>,
+    pub clump_size: Option<i64>,
+}
+
+pub struct PermutationBuilderState {
+    pub builder: Option<LancePermutationBuilder>,
+    pub dest_table_name: String,
+}
+
+#[napi]
+pub struct PermutationBuilder {
+    state: Arc<Mutex<PermutationBuilderState>>,
+}
+
+impl PermutationBuilder {
+    pub fn new(builder: LancePermutationBuilder, dest_table_name: String) -> Self {
+        Self {
+            state: Arc::new(Mutex::new(PermutationBuilderState {
+                builder: Some(builder),
+                dest_table_name,
+            })),
+        }
+    }
+}
+
+impl PermutationBuilder {
+    fn modify(
+        &self,
+        func: impl FnOnce(LancePermutationBuilder) -> LancePermutationBuilder,
+    ) -> napi::Result<Self> {
+        let mut state = self.state.lock().unwrap();
+        let builder = state
+            .builder
+            .take()
+            .ok_or_else(|| napi::Error::from_reason("Builder already consumed"))?;
+        state.builder = Some(func(builder));
+        Ok(Self {
+            state: self.state.clone(),
+        })
+    }
+}
+
+#[napi]
+impl PermutationBuilder {
+    /// Configure random splits
+    #[napi]
+    pub fn split_random(&self, options: SplitRandomOptions) -> napi::Result<Self> {
+        // Check that exactly one split type is provided
+        let split_args_count = [
+            options.ratios.is_some(),
+            options.counts.is_some(),
+            options.fixed.is_some(),
+        ]
+        .iter()
+        .filter(|&&x| x)
+        .count();
+
+        if split_args_count != 1 {
+            return Err(napi::Error::from_reason(
+                "Exactly one of 'ratios', 'counts', or 'fixed' must be provided",
+            ));
+        }
+
+        let sizes = if let Some(ratios) = options.ratios {
+            SplitSizes::Percentages(ratios)
+        } else if let Some(counts) = options.counts {
+            SplitSizes::Counts(counts.into_iter().map(|c| c as u64).collect())
+        } else if let Some(fixed) = options.fixed {
+            SplitSizes::Fixed(fixed as u64)
+        } else {
+            unreachable!("One of the split arguments must be provided");
+        };
+
+        let seed = options.seed.map(|s| s as u64);
+
+        self.modify(|builder| builder.with_split_strategy(SplitStrategy::Random { seed, sizes }))
+    }
+
+    /// Configure hash-based splits
+    #[napi]
+    pub fn split_hash(&self, options: SplitHashOptions) -> napi::Result<Self> {
+        let split_weights = options
+            .split_weights
+            .into_iter()
+            .map(|w| w as u64)
+            .collect();
+        let discard_weight = options.discard_weight.unwrap_or(0) as u64;
+
+        self.modify(|builder| {
+            builder.with_split_strategy(SplitStrategy::Hash {
+                columns: options.columns,
+                split_weights,
+                discard_weight,
+            })
+        })
+    }
+
+    /// Configure sequential splits
+    #[napi]
+    pub fn split_sequential(&self, options: SplitSequentialOptions) -> napi::Result<Self> {
+        // Check that exactly one split type is provided
+        let split_args_count = [
+            options.ratios.is_some(),
+            options.counts.is_some(),
+            options.fixed.is_some(),
+        ]
+        .iter()
+        .filter(|&&x| x)
+        .count();
+
+        if split_args_count != 1 {
+            return Err(napi::Error::from_reason(
+                "Exactly one of 'ratios', 'counts', or 'fixed' must be provided",
+            ));
+        }
+
+        let sizes = if let Some(ratios) = options.ratios {
+            SplitSizes::Percentages(ratios)
+        } else if let Some(counts) = options.counts {
+            SplitSizes::Counts(counts.into_iter().map(|c| c as u64).collect())
+        } else if let Some(fixed) = options.fixed {
+            SplitSizes::Fixed(fixed as u64)
+        } else {
+            unreachable!("One of the split arguments must be provided");
+        };
+
+        self.modify(|builder| builder.with_split_strategy(SplitStrategy::Sequential { sizes }))
+    }
+
+    /// Configure calculated splits
+    #[napi]
+    pub fn split_calculated(&self, calculation: String) -> napi::Result<Self> {
+        self.modify(|builder| {
+            builder.with_split_strategy(SplitStrategy::Calculated { calculation })
+        })
+    }
+
+    /// Configure shuffling
+    #[napi]
+    pub fn shuffle(&self, options: ShuffleOptions) -> napi::Result<Self> {
+        let seed = options.seed.map(|s| s as u64);
+        let clump_size = options.clump_size.map(|c| c as u64);
+
+        self.modify(|builder| {
+            builder.with_shuffle_strategy(ShuffleStrategy::Random { seed, clump_size })
+        })
+    }
+
+    /// Configure filtering
+    #[napi]
+    pub fn filter(&self, filter: String) -> napi::Result<Self> {
+        self.modify(|builder| builder.with_filter(filter))
+    }
+
+    /// Execute the permutation builder and create the table
+    #[napi]
+    pub async fn execute(&self) -> napi::Result<Table> {
+        let (builder, dest_table_name) = {
+            let mut state = self.state.lock().unwrap();
+            let builder = state
+                .builder
+                .take()
+                .ok_or_else(|| napi::Error::from_reason("Builder already consumed"))?;
+
+            let dest_table_name = std::mem::take(&mut state.dest_table_name);
+            (builder, dest_table_name)
+        };
+
+        let table = builder.build(&dest_table_name).await.default_error()?;
+        Ok(Table::new(table))
+    }
+}
+
+/// Create a permutation builder for the given table
+#[napi]
+pub fn permutation_builder(
+    table: &crate::table::Table,
+    dest_table_name: String,
+) -> napi::Result<PermutationBuilder> {
+    use lancedb::dataloader::permutation::PermutationBuilder as LancePermutationBuilder;
+
+    let inner_table = table.inner_ref()?.clone();
+    let inner_builder = LancePermutationBuilder::new(inner_table);
+
+    Ok(PermutationBuilder::new(inner_builder, dest_table_name))
+}

nodejs/src/table.rs

@@ -26,7 +26,7 @@ pub struct Table {
 }
 
 impl Table {
-    fn inner_ref(&self) -> napi::Result<&LanceDbTable> {
+    pub(crate) fn inner_ref(&self) -> napi::Result<&LanceDbTable> {
         self.inner
             .as_ref()
             .ok_or_else(|| napi::Error::from_reason(format!("Table {} is closed", self.name)))

python/.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.25.1-beta.1"
+current_version = "0.25.3-beta.0"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.
@@ -24,6 +24,19 @@ commit = true
 message = "Bump version: {current_version} → {new_version}"
 commit_args = ""
 
+# Update Cargo.lock after version bump
+pre_commit_hooks = [
+  """
+    cd python && cargo update -p lancedb-python
+    if git diff --quiet ../Cargo.lock; then
+        echo "Cargo.lock unchanged"
+    else
+        git add ../Cargo.lock
+        echo "Updated and staged Cargo.lock"
+    fi
+    """,
+]
+
 [tool.bumpversion.parts.pre_l]
 values = ["beta", "final"]
 optional_value = "final"

python/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb-python"
-version = "0.25.1-beta.1"
+version = "0.25.3-beta.0"
 edition.workspace = true
 description = "Python bindings for LanceDB"
 license.workspace = true
@@ -14,12 +14,12 @@ name = "_lancedb"
 crate-type = ["cdylib"]
 
 [dependencies]
-arrow = { version = "55.1", features = ["pyarrow"] }
+arrow = { version = "56.2", features = ["pyarrow"] }
 async-trait = "0.1"
 lancedb = { path = "../rust/lancedb", default-features = false }
 env_logger.workspace = true
-pyo3 = { version = "0.24", features = ["extension-module", "abi3-py39"] }
+pyo3 = { version = "0.25", features = ["extension-module", "abi3-py39"] }
-pyo3-async-runtimes = { version = "0.24", features = [
+pyo3-async-runtimes = { version = "0.25", features = [
     "attributes",
     "tokio-runtime",
 ] }
@@ -28,7 +28,7 @@ futures.workspace = true
 tokio = { version = "1.40", features = ["sync"] }
 
 [build-dependencies]
-pyo3-build-config = { version = "0.24", features = [
+pyo3-build-config = { version = "0.25", features = [
     "extension-module",
     "abi3-py39",
 ] }

python/pyproject.toml

@@ -5,12 +5,12 @@ dynamic = ["version"]
 dependencies = [
     "deprecation",
     "numpy",
-    "overrides>=0.7",
+    "overrides>=0.7; python_version<'3.12'",
     "packaging",
     "pyarrow>=16",
     "pydantic>=1.10",
     "tqdm>=4.27.0",
-    "lance-namespace==0.0.6"
+    "lance-namespace>=0.0.16"
 ]
 description = "lancedb"
 authors = [{ name = "LanceDB Devs", email = "dev@lancedb.com" }]

python/python/lancedb/_lancedb.pyi

@@ -60,6 +60,15 @@ class Connection(object):
         storage_options: Optional[Dict[str, str]] = None,
         index_cache_size: Optional[int] = None,
     ) -> Table: ...
+    async def clone_table(
+        self,
+        target_table_name: str,
+        source_uri: str,
+        target_namespace: List[str] = [],
+        source_version: Optional[int] = None,
+        source_tag: Optional[str] = None,
+        is_shallow: bool = True,
+    ) -> Table: ...
     async def rename_table(
         self,
         cur_name: str,
@@ -124,6 +133,7 @@ class Tags:
     async def update(self, tag: str, version: int): ...
 
 class IndexConfig:
+    name: str
     index_type: str
     columns: List[str]
 
@@ -286,3 +296,34 @@ class AlterColumnsResult:
 
 class DropColumnsResult:
     version: int
+
+class AsyncPermutationBuilder:
+    def select(self, projections: Dict[str, str]) -> "AsyncPermutationBuilder": ...
+    def split_random(
+        self,
+        *,
+        ratios: Optional[List[float]] = None,
+        counts: Optional[List[int]] = None,
+        fixed: Optional[int] = None,
+        seed: Optional[int] = None,
+    ) -> "AsyncPermutationBuilder": ...
+    def split_hash(
+        self, columns: List[str], split_weights: List[int], *, discard_weight: int = 0
+    ) -> "AsyncPermutationBuilder": ...
+    def split_sequential(
+        self,
+        *,
+        ratios: Optional[List[float]] = None,
+        counts: Optional[List[int]] = None,
+        fixed: Optional[int] = None,
+    ) -> "AsyncPermutationBuilder": ...
+    def split_calculated(self, calculation: str) -> "AsyncPermutationBuilder": ...
+    def shuffle(
+        self, seed: Optional[int], clump_size: Optional[int]
+    ) -> "AsyncPermutationBuilder": ...
+    def filter(self, filter: str) -> "AsyncPermutationBuilder": ...
+    async def execute(self) -> Table: ...
+
+def async_permutation_builder(
+    table: Table, dest_table_name: str
+) -> AsyncPermutationBuilder: ...

python/python/lancedb/db.py

@@ -5,11 +5,20 @@
 from __future__ import annotations
 
 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
 
+if sys.version_info >= (3, 12):
+    from typing import override
+
+    class EnforceOverrides:
+        pass
+else:
+    from overrides import EnforceOverrides, override  # type: ignore
+
 from lancedb.embeddings.registry import EmbeddingFunctionRegistry
-from overrides import EnforceOverrides, override  # type: ignore
 
 from lancedb.common import data_to_reader, sanitize_uri, validate_schema
 from lancedb.background_loop import LOOP
@@ -32,7 +41,6 @@ import deprecation
 if TYPE_CHECKING:
     import pyarrow as pa
     from .pydantic import LanceModel
-    from datetime import timedelta
 
     from ._lancedb import Connection as LanceDbConnection
     from .common import DATA, URI
@@ -444,7 +452,12 @@ class LanceDBConnection(DBConnection):
         read_consistency_interval: Optional[timedelta] = None,
         storage_options: Optional[Dict[str, str]] = None,
         session: Optional[Session] = None,
+        _inner: Optional[LanceDbConnection] = None,
     ):
+        if _inner is not None:
+            self._conn = _inner
+            return
+
         if not isinstance(uri, Path):
             scheme = get_uri_scheme(uri)
         is_local = isinstance(uri, Path) or scheme == "file"
@@ -453,11 +466,6 @@ class LanceDBConnection(DBConnection):
                 uri = Path(uri)
             uri = uri.expanduser().absolute()
             Path(uri).mkdir(parents=True, exist_ok=True)
-        self._uri = str(uri)
-        self._entered = False
-        self.read_consistency_interval = read_consistency_interval
-        self.storage_options = storage_options
-        self.session = session
 
         if read_consistency_interval is not None:
             read_consistency_interval_secs = read_consistency_interval.total_seconds()
@@ -476,10 +484,32 @@ class LanceDBConnection(DBConnection):
                 session,
             )
 
+        # TODO: It would be nice if we didn't store self.storage_options but it is
+        # currently used by the LanceTable.to_lance method.  This doesn't _really_
+        # work because some paths like LanceDBConnection.from_inner will lose the
+        # storage_options.  Also, this class really shouldn't be holding any state
+        # beyond _conn.
+        self.storage_options = storage_options
         self._conn = AsyncConnection(LOOP.run(do_connect()))
 
+    @property
+    def read_consistency_interval(self) -> Optional[timedelta]:
+        return LOOP.run(self._conn.get_read_consistency_interval())
+
+    @property
+    def session(self) -> Optional[Session]:
+        return self._conn.session
+
+    @property
+    def uri(self) -> str:
+        return self._conn.uri
+
+    @classmethod
+    def from_inner(cls, inner: LanceDbConnection):
+        return cls(None, _inner=inner)
+
     def __repr__(self) -> str:
-        val = f"{self.__class__.__name__}(uri={self._uri!r}"
+        val = f"{self.__class__.__name__}(uri={self._conn.uri!r}"
         if self.read_consistency_interval is not None:
             val += f", read_consistency_interval={repr(self.read_consistency_interval)}"
         val += ")"
@@ -489,6 +519,10 @@ class LanceDBConnection(DBConnection):
         conn = AsyncConnection(await lancedb_connect(self.uri))
         return await conn.table_names(start_after=start_after, limit=limit)
 
+    @property
+    def _inner(self) -> LanceDbConnection:
+        return self._conn._inner
+
     @override
     def list_namespaces(
         self,
@@ -665,6 +699,60 @@ class LanceDBConnection(DBConnection):
             index_cache_size=index_cache_size,
         )
 
+    def clone_table(
+        self,
+        target_table_name: str,
+        source_uri: str,
+        *,
+        target_namespace: List[str] = [],
+        source_version: Optional[int] = None,
+        source_tag: Optional[str] = None,
+        is_shallow: bool = True,
+    ) -> LanceTable:
+        """Clone a table from a source table.
+
+        A shallow clone creates a new table that shares the underlying data files
+        with the source table but has its own independent manifest. This allows
+        both the source and cloned tables to evolve independently while initially
+        sharing the same data, deletion, and index files.
+
+        Parameters
+        ----------
+        target_table_name: str
+            The name of the target table to create.
+        source_uri: str
+            The URI of the source table to clone from.
+        target_namespace: List[str], optional
+            The namespace for the target table.
+            None or empty list represents root namespace.
+        source_version: int, optional
+            The version of the source table to clone.
+        source_tag: str, optional
+            The tag of the source table to clone.
+        is_shallow: bool, default True
+            Whether to perform a shallow clone (True) or deep clone (False).
+            Currently only shallow clone is supported.
+
+        Returns
+        -------
+        A LanceTable object representing the cloned table.
+        """
+        LOOP.run(
+            self._conn.clone_table(
+                target_table_name,
+                source_uri,
+                target_namespace=target_namespace,
+                source_version=source_version,
+                source_tag=source_tag,
+                is_shallow=is_shallow,
+            )
+        )
+        return LanceTable.open(
+            self,
+            target_table_name,
+            namespace=target_namespace,
+        )
+
     @override
     def drop_table(
         self,
@@ -794,6 +882,13 @@ class AsyncConnection(object):
     def uri(self) -> str:
         return self._inner.uri
 
+    async def get_read_consistency_interval(self) -> Optional[timedelta]:
+        interval_secs = await self._inner.get_read_consistency_interval()
+        if interval_secs is not None:
+            return timedelta(seconds=interval_secs)
+        else:
+            return None
+
     async def list_namespaces(
         self,
         namespace: List[str] = [],
@@ -1136,6 +1231,54 @@ class AsyncConnection(object):
         )
         return AsyncTable(table)
 
+    async def clone_table(
+        self,
+        target_table_name: str,
+        source_uri: str,
+        *,
+        target_namespace: List[str] = [],
+        source_version: Optional[int] = None,
+        source_tag: Optional[str] = None,
+        is_shallow: bool = True,
+    ) -> AsyncTable:
+        """Clone a table from a source table.
+
+        A shallow clone creates a new table that shares the underlying data files
+        with the source table but has its own independent manifest. This allows
+        both the source and cloned tables to evolve independently while initially
+        sharing the same data, deletion, and index files.
+
+        Parameters
+        ----------
+        target_table_name: str
+            The name of the target table to create.
+        source_uri: str
+            The URI of the source table to clone from.
+        target_namespace: List[str], optional
+            The namespace for the target table.
+            None or empty list represents root namespace.
+        source_version: int, optional
+            The version of the source table to clone.
+        source_tag: str, optional
+            The tag of the source table to clone.
+        is_shallow: bool, default True
+            Whether to perform a shallow clone (True) or deep clone (False).
+            Currently only shallow clone is supported.
+
+        Returns
+        -------
+        An AsyncTable object representing the cloned table.
+        """
+        table = await self._inner.clone_table(
+            target_table_name,
+            source_uri,
+            target_namespace=target_namespace,
+            source_version=source_version,
+            source_tag=source_tag,
+            is_shallow=is_shallow,
+        )
+        return AsyncTable(table)
+
     async def rename_table(
         self,
         cur_name: str,

python/python/lancedb/embeddings/registry.py

@@ -122,7 +122,7 @@ class EmbeddingFunctionRegistry:
             obj["vector_column"]: EmbeddingFunctionConfig(
                 vector_column=obj["vector_column"],
                 source_column=obj["source_column"],
-                function=self.get(obj["name"])(**obj["model"]),
+                function=self.get(obj["name"]).create(**obj["model"]),
             )
             for obj in raw_list
         }

python/python/lancedb/index.py

@@ -251,6 +251,13 @@ class HnswPq:
         results. In most cases, there is no benefit to setting this higher than 500.
         This value should be set to a value that is not less than `ef` in the
         search phase.
+
+    target_partition_size, default is 1,048,576
+
+        The target size of each partition.
+
+        This value controls the tradeoff between search performance and accuracy.
+        faster search but less accurate results as higher value.
     """
 
     distance_type: Literal["l2", "cosine", "dot"] = "l2"
@@ -261,6 +268,7 @@ class HnswPq:
     sample_rate: int = 256
     m: int = 20
     ef_construction: int = 300
+    target_partition_size: Optional[int] = None
 
 
 @dataclass
@@ -351,6 +359,12 @@ class HnswSq:
         This value should be set to a value that is not less than `ef` in the search
         phase.
 
+    target_partition_size, default is 1,048,576
+
+        The target size of each partition.
+
+        This value controls the tradeoff between search performance and accuracy.
+        faster search but less accurate results as higher value.
     """
 
     distance_type: Literal["l2", "cosine", "dot"] = "l2"
@@ -359,6 +373,7 @@ class HnswSq:
     sample_rate: int = 256
     m: int = 20
     ef_construction: int = 300
+    target_partition_size: Optional[int] = None
 
 
 @dataclass
@@ -444,12 +459,20 @@ class IvfFlat:
         cases the default should be sufficient.
 
         The default value is 256.
+
+    target_partition_size, default is 8192
+
+        The target size of each partition.
+
+        This value controls the tradeoff between search performance and accuracy.
+        faster search but less accurate results as higher value.
     """
 
     distance_type: Literal["l2", "cosine", "dot", "hamming"] = "l2"
     num_partitions: Optional[int] = None
     max_iterations: int = 50
     sample_rate: int = 256
+    target_partition_size: Optional[int] = None
 
 
 @dataclass
@@ -564,6 +587,13 @@ class IvfPq:
         cases the default should be sufficient.
 
         The default value is 256.
+
+    target_partition_size, default is 8192
+
+        The target size of each partition.
+
+        This value controls the tradeoff between search performance and accuracy.
+        faster search but less accurate results as higher value.
     """
 
     distance_type: Literal["l2", "cosine", "dot"] = "l2"
@@ -572,11 +602,56 @@ class IvfPq:
     num_bits: int = 8
     max_iterations: int = 50
     sample_rate: int = 256
+    target_partition_size: Optional[int] = None
+
+
+@dataclass
+class IvfRq:
+    """Describes an IVF RQ Index
+
+    IVF-RQ (Residual Quantization) stores a compressed copy of each vector using
+    residual quantization and organizes them into IVF partitions. Parameters
+    largely mirror IVF-PQ for consistency.
+
+    Attributes
+    ----------
+    distance_type: str, default "l2"
+        Distance metric used to train the index and for quantization.
+
+        The following distance types are available:
+
+        "l2" - Euclidean distance.
+        "cosine" - Cosine distance.
+        "dot" - Dot product.
+
+    num_partitions: int, default sqrt(num_rows)
+        Number of IVF partitions to create.
+
+    num_bits: int, default 1
+        Number of bits to encode each dimension.
+
+    max_iterations: int, default 50
+        Max iterations to train kmeans when computing IVF partitions.
+
+    sample_rate: int, default 256
+        Controls the number of training vectors: sample_rate * num_partitions.
+
+    target_partition_size, default is 8192
+        Target size of each partition.
+    """
+
+    distance_type: Literal["l2", "cosine", "dot"] = "l2"
+    num_partitions: Optional[int] = None
+    num_bits: int = 1
+    max_iterations: int = 50
+    sample_rate: int = 256
+    target_partition_size: Optional[int] = None
 
 
 __all__ = [
     "BTree",
     "IvfPq",
+    "IvfRq",
     "IvfFlat",
     "HnswPq",
     "HnswSq",

python/python/lancedb/merge.py

@@ -33,6 +33,7 @@ class LanceMergeInsertBuilder(object):
         self._when_not_matched_by_source_delete = False
         self._when_not_matched_by_source_condition = None
         self._timeout = None
+        self._use_index = True
 
     def when_matched_update_all(
         self, *, where: Optional[str] = None
@@ -78,6 +79,23 @@ class LanceMergeInsertBuilder(object):
             self._when_not_matched_by_source_condition = condition
         return self
 
+    def use_index(self, use_index: bool) -> LanceMergeInsertBuilder:
+        """
+        Controls whether to use indexes for the merge operation.
+
+        When set to `True` (the default), the operation will use an index if available
+        on the join key for improved performance. When set to `False`, it forces a full
+        table scan even if an index exists. This can be useful for benchmarking or when
+        the query optimizer chooses a suboptimal path.
+
+        Parameters
+        ----------
+        use_index: bool
+            Whether to use indices for the merge operation. Defaults to `True`.
+        """
+        self._use_index = use_index
+        return self
+
     def execute(
         self,
         new_data: DATA,

python/python/lancedb/namespace.py

@@ -12,13 +12,18 @@ from __future__ import annotations
 
 from typing import Dict, Iterable, List, Optional, Union
 import os
+import sys
+
+if sys.version_info >= (3, 12):
+    from typing import override
+else:
+    from overrides import override
 
 from lancedb.db import DBConnection
 from lancedb.table import LanceTable, Table
 from lancedb.util import validate_table_name
 from lancedb.common import validate_schema
 from lancedb.table import sanitize_create_table
-from overrides import override
 
 from lance_namespace import LanceNamespace, connect as namespace_connect
 from lance_namespace_urllib3_client.models import (

python/python/lancedb/permutation.py

@@ -0,0 +1,72 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+from ._lancedb import async_permutation_builder
+from .table import LanceTable
+from .background_loop import LOOP
+from typing import Optional
+
+
+class PermutationBuilder:
+    def __init__(self, table: LanceTable, dest_table_name: str):
+        self._async = async_permutation_builder(table, dest_table_name)
+
+    def select(self, projections: dict[str, str]) -> "PermutationBuilder":
+        self._async.select(projections)
+        return self
+
+    def split_random(
+        self,
+        *,
+        ratios: Optional[list[float]] = None,
+        counts: Optional[list[int]] = None,
+        fixed: Optional[int] = None,
+        seed: Optional[int] = None,
+    ) -> "PermutationBuilder":
+        self._async.split_random(ratios=ratios, counts=counts, fixed=fixed, seed=seed)
+        return self
+
+    def split_hash(
+        self,
+        columns: list[str],
+        split_weights: list[int],
+        *,
+        discard_weight: Optional[int] = None,
+    ) -> "PermutationBuilder":
+        self._async.split_hash(columns, split_weights, discard_weight=discard_weight)
+        return self
+
+    def split_sequential(
+        self,
+        *,
+        ratios: Optional[list[float]] = None,
+        counts: Optional[list[int]] = None,
+        fixed: Optional[int] = None,
+    ) -> "PermutationBuilder":
+        self._async.split_sequential(ratios=ratios, counts=counts, fixed=fixed)
+        return self
+
+    def split_calculated(self, calculation: str) -> "PermutationBuilder":
+        self._async.split_calculated(calculation)
+        return self
+
+    def shuffle(
+        self, *, seed: Optional[int] = None, clump_size: Optional[int] = None
+    ) -> "PermutationBuilder":
+        self._async.shuffle(seed=seed, clump_size=clump_size)
+        return self
+
+    def filter(self, filter: str) -> "PermutationBuilder":
+        self._async.filter(filter)
+        return self
+
+    def execute(self) -> LanceTable:
+        async def do_execute():
+            inner_tbl = await self._async.execute()
+            return LanceTable.from_inner(inner_tbl)
+
+        return LOOP.run(do_execute())
+
+
+def permutation_builder(table: LanceTable, dest_table_name: str) -> PermutationBuilder:
+    return PermutationBuilder(table, dest_table_name)

python/python/lancedb/remote/db.py

@@ -5,15 +5,20 @@
 from datetime import timedelta
 import logging
 from concurrent.futures import ThreadPoolExecutor
+import sys
 from typing import Any, Dict, Iterable, List, Optional, Union
 from urllib.parse import urlparse
 import warnings
 
+if sys.version_info >= (3, 12):
+    from typing import override
+else:
+    from overrides import override
+
 # Remove this import to fix circular dependency
 # from lancedb import connect_async
 from lancedb.remote import ClientConfig
 import pyarrow as pa
-from overrides import override
 
 from ..common import DATA
 from ..db import DBConnection, LOOP
@@ -212,6 +217,53 @@ class RemoteDBConnection(DBConnection):
         table = LOOP.run(self._conn.open_table(name, namespace=namespace))
         return RemoteTable(table, self.db_name)
 
+    def clone_table(
+        self,
+        target_table_name: str,
+        source_uri: str,
+        *,
+        target_namespace: List[str] = [],
+        source_version: Optional[int] = None,
+        source_tag: Optional[str] = None,
+        is_shallow: bool = True,
+    ) -> Table:
+        """Clone a table from a source table.
+
+        Parameters
+        ----------
+        target_table_name: str
+            The name of the target table to create.
+        source_uri: str
+            The URI of the source table to clone from.
+        target_namespace: List[str], optional
+            The namespace for the target table.
+            None or empty list represents root namespace.
+        source_version: int, optional
+            The version of the source table to clone.
+        source_tag: str, optional
+            The tag of the source table to clone.
+        is_shallow: bool, default True
+            Whether to perform a shallow clone (True) or deep clone (False).
+            Currently only shallow clone is supported.
+
+        Returns
+        -------
+        A RemoteTable object representing the cloned table.
+        """
+        from .table import RemoteTable
+
+        table = LOOP.run(
+            self._conn.clone_table(
+                target_table_name,
+                source_uri,
+                target_namespace=target_namespace,
+                source_version=source_version,
+                source_tag=source_tag,
+                is_shallow=is_shallow,
+            )
+        )
+        return RemoteTable(table, self.db_name)
+
     @override
     def create_table(
         self,

python/python/lancedb/remote/table.py

@@ -114,7 +114,7 @@ class RemoteTable(Table):
         index_type: Literal["BTREE", "BITMAP", "LABEL_LIST", "scalar"] = "scalar",
         *,
         replace: bool = False,
-        wait_timeout: timedelta = None,
+        wait_timeout: Optional[timedelta] = None,
         name: Optional[str] = None,
     ):
         """Creates a scalar index
@@ -153,7 +153,7 @@ class RemoteTable(Table):
         column: str,
         *,
         replace: bool = False,
-        wait_timeout: timedelta = None,
+        wait_timeout: Optional[timedelta] = None,
         with_position: bool = False,
         # tokenizer configs:
         base_tokenizer: str = "simple",

python/python/lancedb/rerankers/__init__.py

@@ -9,6 +9,7 @@ from .linear_combination import LinearCombinationReranker
 from .openai import OpenaiReranker
 from .jinaai import JinaReranker
 from .rrf import RRFReranker
+from .mrr import MRRReranker
 from .answerdotai import AnswerdotaiRerankers
 from .voyageai import VoyageAIReranker
 
@@ -23,4 +24,5 @@ __all__ = [
     "RRFReranker",
     "AnswerdotaiRerankers",
     "VoyageAIReranker",
+    "MRRReranker",
 ]

python/python/lancedb/rerankers/mrr.py

@@ -0,0 +1,169 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+
+from typing import Union, List, TYPE_CHECKING
+import pyarrow as pa
+import numpy as np
+
+from collections import defaultdict
+from .base import Reranker
+
+if TYPE_CHECKING:
+    from ..table import LanceVectorQueryBuilder
+
+
+class MRRReranker(Reranker):
+    """
+    Reranks the results using Mean Reciprocal Rank (MRR) algorithm based
+    on the scores of vector and FTS search.
+    Algorithm reference - https://en.wikipedia.org/wiki/Mean_reciprocal_rank
+
+    MRR calculates the average of reciprocal ranks across different search results.
+    For each document, it computes the reciprocal of its rank in each system,
+    then takes the mean of these reciprocal ranks as the final score.
+
+    Parameters
+    ----------
+    weight_vector : float, default 0.5
+        Weight for vector search results (0.0 to 1.0)
+    weight_fts : float, default 0.5
+        Weight for FTS search results (0.0 to 1.0)
+        Note: weight_vector + weight_fts should equal 1.0
+    return_score : str, default "relevance"
+        Options are "relevance" or "all"
+        The type of score to return. If "relevance", will return only the relevance
+        score. If "all", will return all scores from the vector and FTS search along
+        with the relevance score.
+    """
+
+    def __init__(
+        self,
+        weight_vector: float = 0.5,
+        weight_fts: float = 0.5,
+        return_score="relevance",
+    ):
+        if not (0.0 <= weight_vector <= 1.0):
+            raise ValueError("weight_vector must be between 0.0 and 1.0")
+        if not (0.0 <= weight_fts <= 1.0):
+            raise ValueError("weight_fts must be between 0.0 and 1.0")
+        if abs(weight_vector + weight_fts - 1.0) > 1e-6:
+            raise ValueError("weight_vector + weight_fts must equal 1.0")
+
+        super().__init__(return_score)
+        self.weight_vector = weight_vector
+        self.weight_fts = weight_fts
+
+    def rerank_hybrid(
+        self,
+        query: str,  # noqa: F821
+        vector_results: pa.Table,
+        fts_results: pa.Table,
+    ):
+        vector_ids = vector_results["_rowid"].to_pylist() if vector_results else []
+        fts_ids = fts_results["_rowid"].to_pylist() if fts_results else []
+
+        # Maps result_id to list of (type, reciprocal_rank)
+        mrr_score_map = defaultdict(list)
+
+        if vector_ids:
+            for rank, result_id in enumerate(vector_ids, 1):
+                reciprocal_rank = 1.0 / rank
+                mrr_score_map[result_id].append(("vector", reciprocal_rank))
+
+        if fts_ids:
+            for rank, result_id in enumerate(fts_ids, 1):
+                reciprocal_rank = 1.0 / rank
+                mrr_score_map[result_id].append(("fts", reciprocal_rank))
+
+        final_mrr_scores = {}
+        for result_id, scores in mrr_score_map.items():
+            vector_rr = 0.0
+            fts_rr = 0.0
+
+            for score_type, reciprocal_rank in scores:
+                if score_type == "vector":
+                    vector_rr = reciprocal_rank
+                elif score_type == "fts":
+                    fts_rr = reciprocal_rank
+
+            # If a document doesn't appear, its reciprocal rank is 0
+            weighted_mrr = self.weight_vector * vector_rr + self.weight_fts * fts_rr
+            final_mrr_scores[result_id] = weighted_mrr
+
+        combined_results = self.merge_results(vector_results, fts_results)
+        combined_row_ids = combined_results["_rowid"].to_pylist()
+        relevance_scores = [final_mrr_scores[row_id] for row_id in combined_row_ids]
+        combined_results = combined_results.append_column(
+            "_relevance_score", pa.array(relevance_scores, type=pa.float32())
+        )
+        combined_results = combined_results.sort_by(
+            [("_relevance_score", "descending")]
+        )
+
+        if self.score == "relevance":
+            combined_results = self._keep_relevance_score(combined_results)
+
+        return combined_results
+
+    def rerank_multivector(
+        self,
+        vector_results: Union[List[pa.Table], List["LanceVectorQueryBuilder"]],
+        query: str = None,
+        deduplicate: bool = True,  # noqa: F821
+    ):
+        """
+        Reranks the results from multiple vector searches using MRR algorithm.
+        Each vector search result is treated as a separate ranking system,
+        and MRR calculates the mean of reciprocal ranks across all systems.
+        This cannot reuse rerank_hybrid because MRR semantics require treating
+        each vector result as a separate ranking system.
+        """
+        if not all(isinstance(v, type(vector_results[0])) for v in vector_results):
+            raise ValueError(
+                "All elements in vector_results should be of the same type"
+            )
+
+        # avoid circular import
+        if type(vector_results[0]).__name__ == "LanceVectorQueryBuilder":
+            vector_results = [result.to_arrow() for result in vector_results]
+        elif not isinstance(vector_results[0], pa.Table):
+            raise ValueError(
+                "vector_results should be a list of pa.Table or LanceVectorQueryBuilder"
+            )
+
+        if not all("_rowid" in result.column_names for result in vector_results):
+            raise ValueError(
+                "'_rowid' is required for deduplication. \
+                    add _rowid to search results like this: \
+                    `search().with_row_id(True)`"
+            )
+
+        mrr_score_map = defaultdict(list)
+
+        for result_table in vector_results:
+            result_ids = result_table["_rowid"].to_pylist()
+            for rank, result_id in enumerate(result_ids, 1):
+                reciprocal_rank = 1.0 / rank
+                mrr_score_map[result_id].append(reciprocal_rank)
+
+        final_mrr_scores = {}
+        for result_id, reciprocal_ranks in mrr_score_map.items():
+            mean_rr = np.mean(reciprocal_ranks)
+            final_mrr_scores[result_id] = mean_rr
+
+        combined = pa.concat_tables(vector_results, **self._concat_tables_args)
+        combined = self._deduplicate(combined)
+
+        combined_row_ids = combined["_rowid"].to_pylist()
+
+        relevance_scores = [final_mrr_scores[row_id] for row_id in combined_row_ids]
+        combined = combined.append_column(
+            "_relevance_score", pa.array(relevance_scores, type=pa.float32())
+        )
+        combined = combined.sort_by([("_relevance_score", "descending")])
+
+        if self.score == "relevance":
+            combined = self._keep_relevance_score(combined)
+
+        return combined

python/python/lancedb/table.py

@@ -44,7 +44,7 @@ import numpy as np
 
 from .common import DATA, VEC, VECTOR_COLUMN_NAME
 from .embeddings import EmbeddingFunctionConfig, EmbeddingFunctionRegistry
-from .index import BTree, IvfFlat, IvfPq, Bitmap, LabelList, HnswPq, HnswSq, FTS
+from .index import BTree, IvfFlat, IvfPq, Bitmap, IvfRq, LabelList, HnswPq, HnswSq, FTS
 from .merge import LanceMergeInsertBuilder
 from .pydantic import LanceModel, model_to_dict
 from .query import (
@@ -74,6 +74,7 @@ from .index import lang_mapping
 
 
 if TYPE_CHECKING:
+    from .db import LanceDBConnection
     from ._lancedb import (
         Table as LanceDBTable,
         OptimizeStats,
@@ -88,7 +89,6 @@ if TYPE_CHECKING:
         MergeResult,
         UpdateResult,
     )
-    from .db import LanceDBConnection
     from .index import IndexConfig
     import pandas
     import PIL
@@ -691,6 +691,7 @@ class Table(ABC):
         ef_construction: int = 300,
         name: Optional[str] = None,
         train: bool = True,
+        target_partition_size: Optional[int] = None,
     ):
         """Create an index on the table.
 
@@ -1469,10 +1470,7 @@ class Table(ABC):
             be deleted unless they are at least 7 days old. If delete_unverified is True
             then these files will be deleted regardless of their age.
         retrain: bool, default False
-            If True, retrain the vector indices, this would refine the IVF clustering
+            This parameter is no longer used and is deprecated.
-            and quantization, which may improve the search accuracy. It's faster than
-            re-creating the index from scratch, so it's recommended to try this first,
-            when the data distribution has changed significantly.
 
         Experimental API
         ----------------
@@ -1709,22 +1707,38 @@ class LanceTable(Table):
         namespace: List[str] = [],
         storage_options: Optional[Dict[str, str]] = None,
         index_cache_size: Optional[int] = None,
+        _async: AsyncTable = None,
     ):
         self._conn = connection
         self._namespace = namespace
-        self._table = LOOP.run(
+        if _async is not None:
-            connection._conn.open_table(
+            self._table = _async
-                name,
+        else:
-                namespace=namespace,
+            self._table = LOOP.run(
-                storage_options=storage_options,
+                connection._conn.open_table(
-                index_cache_size=index_cache_size,
+                    name,
+                    namespace=namespace,
+                    storage_options=storage_options,
+                    index_cache_size=index_cache_size,
+                )
             )
-        )
 
     @property
     def name(self) -> str:
         return self._table.name
 
+    @classmethod
+    def from_inner(cls, tbl: LanceDBTable):
+        from .db import LanceDBConnection
+
+        async_tbl = AsyncTable(tbl)
+        conn = LanceDBConnection.from_inner(tbl.database())
+        return cls(
+            conn,
+            async_tbl.name,
+            _async=async_tbl,
+        )
+
     @classmethod
     def open(cls, db, name, *, namespace: List[str] = [], **kwargs):
         tbl = cls(db, name, namespace=namespace, **kwargs)
@@ -1993,7 +2007,7 @@ class LanceTable(Table):
         index_cache_size: Optional[int] = None,
         num_bits: int = 8,
         index_type: Literal[
-            "IVF_FLAT", "IVF_PQ", "IVF_HNSW_SQ", "IVF_HNSW_PQ"
+            "IVF_FLAT", "IVF_PQ", "IVF_RQ", "IVF_HNSW_SQ", "IVF_HNSW_PQ"
         ] = "IVF_PQ",
         max_iterations: int = 50,
         sample_rate: int = 256,
@@ -2002,6 +2016,7 @@ class LanceTable(Table):
         *,
         name: Optional[str] = None,
         train: bool = True,
+        target_partition_size: Optional[int] = None,
     ):
         """Create an index on the table."""
         if accelerator is not None:
@@ -2018,6 +2033,7 @@ class LanceTable(Table):
                 num_bits=num_bits,
                 m=m,
                 ef_construction=ef_construction,
+                target_partition_size=target_partition_size,
             )
             self.checkout_latest()
             return
@@ -2027,6 +2043,7 @@ class LanceTable(Table):
                 num_partitions=num_partitions,
                 max_iterations=max_iterations,
                 sample_rate=sample_rate,
+                target_partition_size=target_partition_size,
             )
         elif index_type == "IVF_PQ":
             config = IvfPq(
@@ -2036,6 +2053,16 @@ class LanceTable(Table):
                 num_bits=num_bits,
                 max_iterations=max_iterations,
                 sample_rate=sample_rate,
+                target_partition_size=target_partition_size,
+            )
+        elif index_type == "IVF_RQ":
+            config = IvfRq(
+                distance_type=metric,
+                num_partitions=num_partitions,
+                num_bits=num_bits,
+                max_iterations=max_iterations,
+                sample_rate=sample_rate,
+                target_partition_size=target_partition_size,
             )
         elif index_type == "IVF_HNSW_PQ":
             config = HnswPq(
@@ -2047,6 +2074,7 @@ class LanceTable(Table):
                 sample_rate=sample_rate,
                 m=m,
                 ef_construction=ef_construction,
+                target_partition_size=target_partition_size,
             )
         elif index_type == "IVF_HNSW_SQ":
             config = HnswSq(
@@ -2056,6 +2084,7 @@ class LanceTable(Table):
                 sample_rate=sample_rate,
                 m=m,
                 ef_construction=ef_construction,
+                target_partition_size=target_partition_size,
             )
         else:
             raise ValueError(f"Unknown index type {index_type}")
@@ -2743,6 +2772,10 @@ class LanceTable(Table):
             self._table._do_merge(merge, new_data, on_bad_vectors, fill_value)
         )
 
+    @property
+    def _inner(self) -> LanceDBTable:
+        return self._table._inner
+
     @deprecation.deprecated(
         deprecated_in="0.21.0",
         current_version=__version__,
@@ -2828,10 +2861,7 @@ class LanceTable(Table):
             be deleted unless they are at least 7 days old. If delete_unverified is True
             then these files will be deleted regardless of their age.
         retrain: bool, default False
-            If True, retrain the vector indices, this would refine the IVF clustering
+            This parameter is no longer used and is deprecated.
-            and quantization, which may improve the search accuracy. It's faster than
-            re-creating the index from scratch, so it's recommended to try this first,
-            when the data distribution has changed significantly.
 
         Experimental API
         ----------------
@@ -3329,7 +3359,7 @@ class AsyncTable:
         *,
         replace: Optional[bool] = None,
         config: Optional[
-            Union[IvfFlat, IvfPq, HnswPq, HnswSq, BTree, Bitmap, LabelList, FTS]
+            Union[IvfFlat, IvfPq, IvfRq, HnswPq, HnswSq, BTree, Bitmap, LabelList, FTS]
         ] = None,
         wait_timeout: Optional[timedelta] = None,
         name: Optional[str] = None,
@@ -3368,11 +3398,12 @@ class AsyncTable:
         """
         if config is not None:
             if not isinstance(
-                config, (IvfFlat, IvfPq, HnswPq, HnswSq, BTree, Bitmap, LabelList, FTS)
+                config,
+                (IvfFlat, IvfPq, IvfRq, HnswPq, HnswSq, BTree, Bitmap, LabelList, FTS),
             ):
                 raise TypeError(
-                    "config must be an instance of IvfPq, HnswPq, HnswSq, BTree,"
+                    "config must be an instance of IvfPq, IvfRq, HnswPq, HnswSq, BTree,"
-                    " Bitmap, LabelList, or FTS"
+                    " Bitmap, LabelList, or FTS, but got " + str(type(config))
                 )
         try:
             await self._inner.create_index(
@@ -3919,6 +3950,7 @@ class AsyncTable:
                 when_not_matched_by_source_delete=merge._when_not_matched_by_source_delete,
                 when_not_matched_by_source_condition=merge._when_not_matched_by_source_condition,
                 timeout=merge._timeout,
+                use_index=merge._use_index,
             ),
         )
 
@@ -4291,10 +4323,7 @@ class AsyncTable:
             be deleted unless they are at least 7 days old. If delete_unverified is True
             then these files will be deleted regardless of their age.
         retrain: bool, default False
-            If True, retrain the vector indices, this would refine the IVF clustering
+            This parameter is no longer used and is deprecated.
-            and quantization, which may improve the search accuracy. It's faster than
-            re-creating the index from scratch, so it's recommended to try this first,
-            when the data distribution has changed significantly.
 
         Experimental API
         ----------------
@@ -4317,10 +4346,19 @@ class AsyncTable:
         cleanup_since_ms: Optional[int] = None
         if cleanup_older_than is not None:
             cleanup_since_ms = round(cleanup_older_than.total_seconds() * 1000)
+
+        if retrain:
+            import warnings
+
+            warnings.warn(
+                "The 'retrain' parameter is deprecated and will be removed in a "
+                "future version.",
+                DeprecationWarning,
+            )
+
         return await self._inner.optimize(
             cleanup_since_ms=cleanup_since_ms,
             delete_unverified=delete_unverified,
-            retrain=retrain,
         )
 
     async def list_indices(self) -> Iterable[IndexConfig]:

python/python/lancedb/types.py

@@ -18,10 +18,17 @@ AddMode = Literal["append", "overwrite"]
 CreateMode = Literal["create", "overwrite"]
 
 # Index type literals
-VectorIndexType = Literal["IVF_FLAT", "IVF_PQ", "IVF_HNSW_SQ", "IVF_HNSW_PQ"]
+VectorIndexType = Literal["IVF_FLAT", "IVF_PQ", "IVF_HNSW_SQ", "IVF_HNSW_PQ", "IVF_RQ"]
 ScalarIndexType = Literal["BTREE", "BITMAP", "LABEL_LIST"]
 IndexType = Literal[
-    "IVF_PQ", "IVF_HNSW_PQ", "IVF_HNSW_SQ", "FTS", "BTREE", "BITMAP", "LABEL_LIST"
+    "IVF_PQ",
+    "IVF_HNSW_PQ",
+    "IVF_HNSW_SQ",
+    "FTS",
+    "BTREE",
+    "BITMAP",
+    "LABEL_LIST",
+    "IVF_RQ",
 ]
 
 # Tokenizer literals

python/python/tests/test_db.py

@@ -747,15 +747,16 @@ def test_local_namespace_operations(tmp_path):
     # Create a local database connection
     db = lancedb.connect(tmp_path)
 
-    # Test list_namespaces returns empty list
+    # Test list_namespaces returns empty list for root namespace
     namespaces = list(db.list_namespaces())
     assert namespaces == []
 
-    # Test list_namespaces with parameters still returns empty list
+    # Test list_namespaces with non-empty namespace raises NotImplementedError
-    namespaces_with_params = list(
+    with pytest.raises(
-        db.list_namespaces(namespace=["test"], page_token="token", limit=5)
+        NotImplementedError,
-    )
+        match="Namespace operations are not supported for listing database",
-    assert namespaces_with_params == []
+    ):
+        list(db.list_namespaces(namespace=["test"]))
 
 
 def test_local_create_namespace_not_supported(tmp_path):
@@ -830,3 +831,119 @@ def test_local_table_operations_with_namespace_raise_error(tmp_path):
     # Test table_names without namespace - should work normally
     tables_root = list(db.table_names())
     assert "test_table" in tables_root
+
+
+def test_clone_table_latest_version(tmp_path):
+    """Test cloning a table with the latest version (default behavior)"""
+    import os
+
+    db = lancedb.connect(tmp_path)
+
+    # Create source table with some data
+    data = [
+        {"id": 1, "text": "hello", "vector": [1.0, 2.0]},
+        {"id": 2, "text": "world", "vector": [3.0, 4.0]},
+    ]
+    source_table = db.create_table("source", data=data)
+
+    # Add more data to create a new version
+    more_data = [{"id": 3, "text": "test", "vector": [5.0, 6.0]}]
+    source_table.add(more_data)
+
+    # Clone the table (should get latest version with 3 rows)
+    source_uri = os.path.join(tmp_path, "source.lance")
+    cloned_table = db.clone_table("cloned", source_uri)
+
+    # Verify cloned table has all 3 rows
+    assert cloned_table.count_rows() == 3
+    assert "cloned" in db.table_names()
+
+    # Verify data matches
+    cloned_data = cloned_table.to_pandas()
+    assert len(cloned_data) == 3
+    assert set(cloned_data["id"].tolist()) == {1, 2, 3}
+
+
+def test_clone_table_specific_version(tmp_path):
+    """Test cloning a table from a specific version"""
+    import os
+
+    db = lancedb.connect(tmp_path)
+
+    # Create source table with initial data
+    data = [
+        {"id": 1, "text": "hello", "vector": [1.0, 2.0]},
+        {"id": 2, "text": "world", "vector": [3.0, 4.0]},
+    ]
+    source_table = db.create_table("source", data=data)
+
+    # Get the initial version
+    initial_version = source_table.version
+
+    # Add more data to create a new version
+    more_data = [{"id": 3, "text": "test", "vector": [5.0, 6.0]}]
+    source_table.add(more_data)
+
+    # Verify source now has 3 rows
+    assert source_table.count_rows() == 3
+
+    # Clone from the initial version (should have only 2 rows)
+    source_uri = os.path.join(tmp_path, "source.lance")
+    cloned_table = db.clone_table("cloned", source_uri, source_version=initial_version)
+
+    # Verify cloned table has only the initial 2 rows
+    assert cloned_table.count_rows() == 2
+    cloned_data = cloned_table.to_pandas()
+    assert set(cloned_data["id"].tolist()) == {1, 2}
+
+
+def test_clone_table_with_tag(tmp_path):
+    """Test cloning a table from a tagged version"""
+    import os
+
+    db = lancedb.connect(tmp_path)
+
+    # Create source table with initial data
+    data = [
+        {"id": 1, "text": "hello", "vector": [1.0, 2.0]},
+        {"id": 2, "text": "world", "vector": [3.0, 4.0]},
+    ]
+    source_table = db.create_table("source", data=data)
+
+    # Create a tag for the current version
+    source_table.tags.create("v1.0", source_table.version)
+
+    # Add more data after the tag
+    more_data = [{"id": 3, "text": "test", "vector": [5.0, 6.0]}]
+    source_table.add(more_data)
+
+    # Verify source now has 3 rows
+    assert source_table.count_rows() == 3
+
+    # Clone from the tagged version (should have only 2 rows)
+    source_uri = os.path.join(tmp_path, "source.lance")
+    cloned_table = db.clone_table("cloned", source_uri, source_tag="v1.0")
+
+    # Verify cloned table has only the tagged version's 2 rows
+    assert cloned_table.count_rows() == 2
+    cloned_data = cloned_table.to_pandas()
+    assert set(cloned_data["id"].tolist()) == {1, 2}
+
+
+def test_clone_table_deep_clone_fails(tmp_path):
+    """Test that deep clone raises an unsupported error"""
+    import os
+
+    db = lancedb.connect(tmp_path)
+
+    # Create source table with some data
+    data = [
+        {"id": 1, "text": "hello", "vector": [1.0, 2.0]},
+        {"id": 2, "text": "world", "vector": [3.0, 4.0]},
+    ]
+    db.create_table("source", data=data)
+
+    # Try to create a deep clone (should fail)
+    source_uri = os.path.join(tmp_path, "source.lance")
+    with pytest.raises(Exception, match="Deep clone is not yet implemented"):
+        db.clone_table("cloned", source_uri, is_shallow=False)

python/python/tests/test_embeddings.py

@@ -114,6 +114,63 @@ def test_embedding_function_variables():
     assert func.safe_model_dump()["secret_key"] == "$var:secret"
 
 
+def test_parse_functions_with_variables():
+    @register("variable-parsing-test")
+    class VariableParsingFunction(TextEmbeddingFunction):
+        api_key: str
+        base_url: Optional[str] = None
+
+        @staticmethod
+        def sensitive_keys():
+            return ["api_key"]
+
+        def ndims(self):
+            return 10
+
+        def generate_embeddings(self, texts):
+            # Mock implementation that just returns random embeddings
+            # In real usage, this would use the api_key to call an API
+            return [np.random.rand(self.ndims()).tolist() for _ in texts]
+
+    registry = EmbeddingFunctionRegistry.get_instance()
+
+    registry.set_var("test_api_key", "sk-test-key-12345")
+    registry.set_var("test_base_url", "https://api.example.com")
+
+    conf = EmbeddingFunctionConfig(
+        source_column="text",
+        vector_column="vector",
+        function=registry.get("variable-parsing-test").create(
+            api_key="$var:test_api_key", base_url="$var:test_base_url"
+        ),
+    )
+
+    metadata = registry.get_table_metadata([conf])
+
+    # Create a mock arrow table with the metadata
+    schema = pa.schema(
+        [pa.field("text", pa.string()), pa.field("vector", pa.list_(pa.float32(), 10))]
+    )
+    table = pa.table({"text": [], "vector": []}, schema=schema)
+    table = table.replace_schema_metadata(metadata)
+
+    ds = lance.write_dataset(table, "memory://")
+
+    configs = registry.parse_functions(ds.schema.metadata)
+
+    assert "vector" in configs
+    parsed_func = configs["vector"].function
+
+    assert parsed_func.api_key == "sk-test-key-12345"
+    assert parsed_func.base_url == "https://api.example.com"
+
+    embeddings = parsed_func.generate_embeddings(["test text"])
+    assert len(embeddings) == 1
+    assert len(embeddings[0]) == 10
+
+    assert parsed_func.safe_model_dump()["api_key"] == "$var:test_api_key"
+
+
 def test_embedding_with_bad_results(tmp_path):
     @register("null-embedding")
     class NullEmbeddingFunction(TextEmbeddingFunction):

python/python/tests/test_index.py

@@ -8,7 +8,17 @@ import pyarrow as pa
 import pytest
 import pytest_asyncio
 from lancedb import AsyncConnection, AsyncTable, connect_async
-from lancedb.index import BTree, IvfFlat, IvfPq, Bitmap, LabelList, HnswPq, HnswSq, FTS
+from lancedb.index import (
+    BTree,
+    IvfFlat,
+    IvfPq,
+    IvfRq,
+    Bitmap,
+    LabelList,
+    HnswPq,
+    HnswSq,
+    FTS,
+)
 
 
 @pytest_asyncio.fixture
@@ -35,6 +45,8 @@ async def some_table(db_async):
             "tags": [
                 [f"tag{random.randint(0, 8)}" for _ in range(2)] for _ in range(NROWS)
             ],
+            "is_active": [random.choice([True, False]) for _ in range(NROWS)],
+            "data": [random.randbytes(random.randint(0, 128)) for _ in range(NROWS)],
         }
     )
     return await db_async.create_table(
@@ -99,10 +111,17 @@ async def test_create_fixed_size_binary_index(some_table: AsyncTable):
 @pytest.mark.asyncio
 async def test_create_bitmap_index(some_table: AsyncTable):
     await some_table.create_index("id", config=Bitmap())
+    await some_table.create_index("is_active", config=Bitmap())
+    await some_table.create_index("data", config=Bitmap())
     indices = await some_table.list_indices()
-    assert str(indices) == '[Index(Bitmap, columns=["id"], name="id_idx")]'
+    assert len(indices) == 3
-    indices = await some_table.list_indices()
+    assert indices[0].index_type == "Bitmap"
-    assert len(indices) == 1
+    assert indices[0].columns == ["id"]
+    assert indices[1].index_type == "Bitmap"
+    assert indices[1].columns == ["is_active"]
+    assert indices[2].index_type == "Bitmap"
+    assert indices[2].columns == ["data"]
+
     index_name = indices[0].name
     stats = await some_table.index_stats(index_name)
     assert stats.index_type == "BITMAP"
@@ -111,6 +130,11 @@ async def test_create_bitmap_index(some_table: AsyncTable):
     assert stats.num_unindexed_rows == 0
     assert stats.num_indices == 1
 
+    assert (
+        "ScalarIndexQuery"
+        in await some_table.query().where("is_active = TRUE").explain_plan()
+    )
+
 
 @pytest.mark.asyncio
 async def test_create_label_list_index(some_table: AsyncTable):
@@ -181,6 +205,16 @@ async def test_create_4bit_ivfpq_index(some_table: AsyncTable):
     assert stats.loss >= 0.0
 
 
+@pytest.mark.asyncio
+async def test_create_ivfrq_index(some_table: AsyncTable):
+    await some_table.create_index("vector", config=IvfRq(num_bits=1))
+    indices = await some_table.list_indices()
+    assert len(indices) == 1
+    assert indices[0].index_type == "IvfRq"
+    assert indices[0].columns == ["vector"]
+    assert indices[0].name == "vector_idx"
+
+
 @pytest.mark.asyncio
 async def test_create_hnswpq_index(some_table: AsyncTable):
     await some_table.create_index("vector", config=HnswPq(num_partitions=10))

python/python/tests/test_permutation.py

@@ -0,0 +1,496 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+import pyarrow as pa
+import pytest
+
+from lancedb.permutation import permutation_builder
+
+
+def test_split_random_ratios(mem_db):
+    """Test random splitting with ratios."""
+    tbl = mem_db.create_table(
+        "test_table", pa.table({"x": range(100), "y": range(100)})
+    )
+    permutation_tbl = (
+        permutation_builder(tbl, "test_permutation")
+        .split_random(ratios=[0.3, 0.7])
+        .execute()
+    )
+
+    # Check that the table was created and has data
+    assert permutation_tbl.count_rows() == 100
+
+    # Check that split_id column exists and has correct values
+    data = permutation_tbl.search(None).to_arrow().to_pydict()
+    split_ids = data["split_id"]
+    assert set(split_ids) == {0, 1}
+
+    # Check approximate split sizes (allowing for rounding)
+    split_0_count = split_ids.count(0)
+    split_1_count = split_ids.count(1)
+    assert 25 <= split_0_count <= 35  # ~30% ± tolerance
+    assert 65 <= split_1_count <= 75  # ~70% ± tolerance
+
+
+def test_split_random_counts(mem_db):
+    """Test random splitting with absolute counts."""
+    tbl = mem_db.create_table(
+        "test_table", pa.table({"x": range(100), "y": range(100)})
+    )
+    permutation_tbl = (
+        permutation_builder(tbl, "test_permutation")
+        .split_random(counts=[20, 30])
+        .execute()
+    )
+
+    # Check that we have exactly the requested counts
+    assert permutation_tbl.count_rows() == 50
+
+    data = permutation_tbl.search(None).to_arrow().to_pydict()
+    split_ids = data["split_id"]
+    assert split_ids.count(0) == 20
+    assert split_ids.count(1) == 30
+
+
+def test_split_random_fixed(mem_db):
+    """Test random splitting with fixed number of splits."""
+    tbl = mem_db.create_table(
+        "test_table", pa.table({"x": range(100), "y": range(100)})
+    )
+    permutation_tbl = (
+        permutation_builder(tbl, "test_permutation").split_random(fixed=4).execute()
+    )
+
+    # Check that we have 4 splits with 25 rows each
+    assert permutation_tbl.count_rows() == 100
+
+    data = permutation_tbl.search(None).to_arrow().to_pydict()
+    split_ids = data["split_id"]
+    assert set(split_ids) == {0, 1, 2, 3}
+
+    for split_id in range(4):
+        assert split_ids.count(split_id) == 25
+
+
+def test_split_random_with_seed(mem_db):
+    """Test that seeded random splits are reproducible."""
+    tbl = mem_db.create_table("test_table", pa.table({"x": range(50), "y": range(50)}))
+
+    # Create two identical permutations with same seed
+    perm1 = (
+        permutation_builder(tbl, "perm1")
+        .split_random(ratios=[0.6, 0.4], seed=42)
+        .execute()
+    )
+
+    perm2 = (
+        permutation_builder(tbl, "perm2")
+        .split_random(ratios=[0.6, 0.4], seed=42)
+        .execute()
+    )
+
+    # Results should be identical
+    data1 = perm1.search(None).to_arrow().to_pydict()
+    data2 = perm2.search(None).to_arrow().to_pydict()
+
+    assert data1["row_id"] == data2["row_id"]
+    assert data1["split_id"] == data2["split_id"]
+
+
+def test_split_hash(mem_db):
+    """Test hash-based splitting."""
+    tbl = mem_db.create_table(
+        "test_table",
+        pa.table(
+            {
+                "id": range(100),
+                "category": (["A", "B", "C"] * 34)[:100],  # Repeating pattern
+                "value": range(100),
+            }
+        ),
+    )
+
+    permutation_tbl = (
+        permutation_builder(tbl, "test_permutation")
+        .split_hash(["category"], [1, 1], discard_weight=0)
+        .execute()
+    )
+
+    # Should have all 100 rows (no discard)
+    assert permutation_tbl.count_rows() == 100
+
+    data = permutation_tbl.search(None).to_arrow().to_pydict()
+    split_ids = data["split_id"]
+    assert set(split_ids) == {0, 1}
+
+    # Verify that each split has roughly 50 rows (allowing for hash variance)
+    split_0_count = split_ids.count(0)
+    split_1_count = split_ids.count(1)
+    assert 30 <= split_0_count <= 70  # ~50 ± 20 tolerance for hash distribution
+    assert 30 <= split_1_count <= 70  # ~50 ± 20 tolerance for hash distribution
+
+    # Hash splits should be deterministic - same category should go to same split
+    # Let's verify by creating another permutation and checking consistency
+    perm2 = (
+        permutation_builder(tbl, "test_permutation2")
+        .split_hash(["category"], [1, 1], discard_weight=0)
+        .execute()
+    )
+
+    data2 = perm2.search(None).to_arrow().to_pydict()
+    assert data["split_id"] == data2["split_id"]  # Should be identical
+
+
+def test_split_hash_with_discard(mem_db):
+    """Test hash-based splitting with discard weight."""
+    tbl = mem_db.create_table(
+        "test_table",
+        pa.table({"id": range(100), "category": ["A", "B"] * 50, "value": range(100)}),
+    )
+
+    permutation_tbl = (
+        permutation_builder(tbl, "test_permutation")
+        .split_hash(["category"], [1, 1], discard_weight=2)  # Should discard ~50%
+        .execute()
+    )
+
+    # Should have fewer than 100 rows due to discard
+    row_count = permutation_tbl.count_rows()
+    assert row_count < 100
+    assert row_count > 0  # But not empty
+
+
+def test_split_sequential(mem_db):
+    """Test sequential splitting."""
+    tbl = mem_db.create_table(
+        "test_table", pa.table({"x": range(100), "y": range(100)})
+    )
+
+    permutation_tbl = (
+        permutation_builder(tbl, "test_permutation")
+        .split_sequential(counts=[30, 40])
+        .execute()
+    )
+
+    assert permutation_tbl.count_rows() == 70
+
+    data = permutation_tbl.search(None).to_arrow().to_pydict()
+    row_ids = data["row_id"]
+    split_ids = data["split_id"]
+
+    # Sequential should maintain order
+    assert row_ids == sorted(row_ids)
+
+    # First 30 should be split 0, next 40 should be split 1
+    assert split_ids[:30] == [0] * 30
+    assert split_ids[30:] == [1] * 40
+
+
+def test_split_calculated(mem_db):
+    """Test calculated splitting."""
+    tbl = mem_db.create_table(
+        "test_table", pa.table({"id": range(100), "value": range(100)})
+    )
+
+    permutation_tbl = (
+        permutation_builder(tbl, "test_permutation")
+        .split_calculated("id % 3")  # Split based on id modulo 3
+        .execute()
+    )
+
+    assert permutation_tbl.count_rows() == 100
+
+    data = permutation_tbl.search(None).to_arrow().to_pydict()
+    row_ids = data["row_id"]
+    split_ids = data["split_id"]
+
+    # Verify the calculation: each row's split_id should equal row_id % 3
+    for i, (row_id, split_id) in enumerate(zip(row_ids, split_ids)):
+        assert split_id == row_id % 3
+
+
+def test_split_error_cases(mem_db):
+    """Test error handling for invalid split parameters."""
+    tbl = mem_db.create_table("test_table", pa.table({"x": range(10), "y": range(10)}))
+
+    # Test split_random with no parameters
+    with pytest.raises(Exception):
+        permutation_builder(tbl, "error1").split_random().execute()
+
+    # Test split_random with multiple parameters
+    with pytest.raises(Exception):
+        permutation_builder(tbl, "error2").split_random(
+            ratios=[0.5, 0.5], counts=[5, 5]
+        ).execute()
+
+    # Test split_sequential with no parameters
+    with pytest.raises(Exception):
+        permutation_builder(tbl, "error3").split_sequential().execute()
+
+    # Test split_sequential with multiple parameters
+    with pytest.raises(Exception):
+        permutation_builder(tbl, "error4").split_sequential(
+            ratios=[0.5, 0.5], fixed=2
+        ).execute()
+
+
+def test_shuffle_no_seed(mem_db):
+    """Test shuffling without a seed."""
+    tbl = mem_db.create_table(
+        "test_table", pa.table({"id": range(100), "value": range(100)})
+    )
+
+    # Create a permutation with shuffling (no seed)
+    permutation_tbl = permutation_builder(tbl, "test_permutation").shuffle().execute()
+
+    assert permutation_tbl.count_rows() == 100
+
+    data = permutation_tbl.search(None).to_arrow().to_pydict()
+    row_ids = data["row_id"]
+
+    # Row IDs should not be in sequential order due to shuffling
+    # This is probabilistic but with 100 rows, it's extremely unlikely they'd stay
+    # in order
+    assert row_ids != list(range(100))
+
+
+def test_shuffle_with_seed(mem_db):
+    """Test that shuffling with a seed is reproducible."""
+    tbl = mem_db.create_table(
+        "test_table", pa.table({"id": range(50), "value": range(50)})
+    )
+
+    # Create two identical permutations with same shuffle seed
+    perm1 = permutation_builder(tbl, "perm1").shuffle(seed=42).execute()
+
+    perm2 = permutation_builder(tbl, "perm2").shuffle(seed=42).execute()
+
+    # Results should be identical due to same seed
+    data1 = perm1.search(None).to_arrow().to_pydict()
+    data2 = perm2.search(None).to_arrow().to_pydict()
+
+    assert data1["row_id"] == data2["row_id"]
+    assert data1["split_id"] == data2["split_id"]
+
+
+def test_shuffle_with_clump_size(mem_db):
+    """Test shuffling with clump size."""
+    tbl = mem_db.create_table(
+        "test_table", pa.table({"id": range(100), "value": range(100)})
+    )
+
+    # Create a permutation with shuffling using clumps
+    permutation_tbl = (
+        permutation_builder(tbl, "test_permutation")
+        .shuffle(clump_size=10)  # 10-row clumps
+        .execute()
+    )
+
+    assert permutation_tbl.count_rows() == 100
+
+    data = permutation_tbl.search(None).to_arrow().to_pydict()
+    row_ids = data["row_id"]
+
+    for i in range(10):
+        start = row_ids[i * 10]
+        assert row_ids[i * 10 : (i + 1) * 10] == list(range(start, start + 10))
+
+
+def test_shuffle_different_seeds(mem_db):
+    """Test that different seeds produce different shuffle orders."""
+    tbl = mem_db.create_table(
+        "test_table", pa.table({"id": range(50), "value": range(50)})
+    )
+
+    # Create two permutations with different shuffle seeds
+    perm1 = (
+        permutation_builder(tbl, "perm1")
+        .split_random(fixed=2)
+        .shuffle(seed=42)
+        .execute()
+    )
+
+    perm2 = (
+        permutation_builder(tbl, "perm2")
+        .split_random(fixed=2)
+        .shuffle(seed=123)
+        .execute()
+    )
+
+    # Results should be different due to different seeds
+    data1 = perm1.search(None).to_arrow().to_pydict()
+    data2 = perm2.search(None).to_arrow().to_pydict()
+
+    # Row order should be different
+    assert data1["row_id"] != data2["row_id"]
+
+
+def test_shuffle_combined_with_splits(mem_db):
+    """Test shuffling combined with different split strategies."""
+    tbl = mem_db.create_table(
+        "test_table",
+        pa.table(
+            {
+                "id": range(100),
+                "category": (["A", "B", "C"] * 34)[:100],
+                "value": range(100),
+            }
+        ),
+    )
+
+    # Test shuffle with random splits
+    perm_random = (
+        permutation_builder(tbl, "perm_random")
+        .split_random(ratios=[0.6, 0.4], seed=42)
+        .shuffle(seed=123, clump_size=None)
+        .execute()
+    )
+
+    # Test shuffle with hash splits
+    perm_hash = (
+        permutation_builder(tbl, "perm_hash")
+        .split_hash(["category"], [1, 1], discard_weight=0)
+        .shuffle(seed=456, clump_size=5)
+        .execute()
+    )
+
+    # Test shuffle with sequential splits
+    perm_sequential = (
+        permutation_builder(tbl, "perm_sequential")
+        .split_sequential(counts=[40, 35])
+        .shuffle(seed=789, clump_size=None)
+        .execute()
+    )
+
+    # Verify all permutations work and have expected properties
+    assert perm_random.count_rows() == 100
+    assert perm_hash.count_rows() == 100
+    assert perm_sequential.count_rows() == 75
+
+    # Verify shuffle affected the order
+    data_random = perm_random.search(None).to_arrow().to_pydict()
+    data_sequential = perm_sequential.search(None).to_arrow().to_pydict()
+
+    assert data_random["row_id"] != list(range(100))
+    assert data_sequential["row_id"] != list(range(75))
+
+
+def test_no_shuffle_maintains_order(mem_db):
+    """Test that not calling shuffle maintains the original order."""
+    tbl = mem_db.create_table(
+        "test_table", pa.table({"id": range(50), "value": range(50)})
+    )
+
+    # Create permutation without shuffle (should maintain some order)
+    permutation_tbl = (
+        permutation_builder(tbl, "test_permutation")
+        .split_sequential(counts=[25, 25])  # Sequential maintains order
+        .execute()
+    )
+
+    assert permutation_tbl.count_rows() == 50
+
+    data = permutation_tbl.search(None).to_arrow().to_pydict()
+    row_ids = data["row_id"]
+
+    # With sequential splits and no shuffle, should maintain order
+    assert row_ids == list(range(50))
+
+
+def test_filter_basic(mem_db):
+    """Test basic filtering functionality."""
+    tbl = mem_db.create_table(
+        "test_table", pa.table({"id": range(100), "value": range(100, 200)})
+    )
+
+    # Filter to only include rows where id < 50
+    permutation_tbl = (
+        permutation_builder(tbl, "test_permutation").filter("id < 50").execute()
+    )
+
+    assert permutation_tbl.count_rows() == 50
+
+    data = permutation_tbl.search(None).to_arrow().to_pydict()
+    row_ids = data["row_id"]
+
+    # All row_ids should be less than 50
+    assert all(row_id < 50 for row_id in row_ids)
+
+
+def test_filter_with_splits(mem_db):
+    """Test filtering combined with split strategies."""
+    tbl = mem_db.create_table(
+        "test_table",
+        pa.table(
+            {
+                "id": range(100),
+                "category": (["A", "B", "C"] * 34)[:100],
+                "value": range(100),
+            }
+        ),
+    )
+
+    # Filter to only category A and B, then split
+    permutation_tbl = (
+        permutation_builder(tbl, "test_permutation")
+        .filter("category IN ('A', 'B')")
+        .split_random(ratios=[0.5, 0.5])
+        .execute()
+    )
+
+    # Should have fewer than 100 rows due to filtering
+    row_count = permutation_tbl.count_rows()
+    assert row_count == 67
+
+    data = permutation_tbl.search(None).to_arrow().to_pydict()
+    categories = data["category"]
+
+    # All categories should be A or B
+    assert all(cat in ["A", "B"] for cat in categories)
+
+
+def test_filter_with_shuffle(mem_db):
+    """Test filtering combined with shuffling."""
+    tbl = mem_db.create_table(
+        "test_table",
+        pa.table(
+            {
+                "id": range(100),
+                "category": (["A", "B", "C", "D"] * 25)[:100],
+                "value": range(100),
+            }
+        ),
+    )
+
+    # Filter and shuffle
+    permutation_tbl = (
+        permutation_builder(tbl, "test_permutation")
+        .filter("category IN ('A', 'C')")
+        .shuffle(seed=42)
+        .execute()
+    )
+
+    row_count = permutation_tbl.count_rows()
+    assert row_count == 50  # Should have 50 rows (A and C categories)
+
+    data = permutation_tbl.search(None).to_arrow().to_pydict()
+    row_ids = data["row_id"]
+
+    assert row_ids != sorted(row_ids)
+
+
+def test_filter_empty_result(mem_db):
+    """Test filtering that results in empty set."""
+    tbl = mem_db.create_table(
+        "test_table", pa.table({"id": range(10), "value": range(10)})
+    )
+
+    # Filter that matches nothing
+    permutation_tbl = (
+        permutation_builder(tbl, "test_permutation")
+        .filter("value > 100")  # No values > 100 in our data
+        .execute()
+    )
+
+    assert permutation_tbl.count_rows() == 0

python/python/tests/test_rerankers.py

@@ -22,6 +22,7 @@ from lancedb.rerankers import (
     JinaReranker,
     AnswerdotaiRerankers,
     VoyageAIReranker,
+    MRRReranker,
 )
 from lancedb.table import LanceTable
 
@@ -46,6 +47,7 @@ def get_test_table(tmp_path, use_tantivy):
         db,
         "my_table",
         schema=MyTable,
+        mode="overwrite",
     )
 
     # Need to test with a bunch of phrases to make sure sorting is consistent
@@ -96,7 +98,7 @@ def get_test_table(tmp_path, use_tantivy):
     )
 
     # Create a fts index
-    table.create_fts_index("text", use_tantivy=use_tantivy)
+    table.create_fts_index("text", use_tantivy=use_tantivy, replace=True)
 
     return table, MyTable
 
@@ -320,6 +322,34 @@ def test_rrf_reranker(tmp_path, use_tantivy):
     _run_test_hybrid_reranker(reranker, tmp_path, use_tantivy)
 
 
+@pytest.mark.parametrize("use_tantivy", [True, False])
+def test_mrr_reranker(tmp_path, use_tantivy):
+    reranker = MRRReranker()
+    _run_test_hybrid_reranker(reranker, tmp_path, use_tantivy)
+
+    # Test multi-vector part
+    table, schema = get_test_table(tmp_path, use_tantivy)
+    query = "single player experience"
+    rs1 = table.search(query, vector_column_name="vector").limit(10).with_row_id(True)
+    rs2 = (
+        table.search(query, vector_column_name="meta_vector")
+        .limit(10)
+        .with_row_id(True)
+    )
+    result = reranker.rerank_multivector([rs1, rs2])
+    assert "_relevance_score" in result.column_names
+    assert len(result) <= 20
+
+    if len(result) > 1:
+        assert np.all(np.diff(result.column("_relevance_score").to_numpy()) <= 0), (
+            "The _relevance_score should be descending."
+        )
+
+    # Test with duplicate results
+    result_deduped = reranker.rerank_multivector([rs1, rs2, rs1])
+    assert len(result_deduped) == len(result)
+
+
 def test_rrf_reranker_distance():
     data = pa.table(
         {

python/python/tests/test_table.py

@@ -674,6 +674,45 @@ def test_create_index_method(mock_create_index, mem_db: DBConnection):
         "vector", replace=True, config=expected_config, name=None, train=True
     )
 
+    # Test with target_partition_size
+    table.create_index(
+        metric="l2",
+        num_sub_vectors=96,
+        vector_column_name="vector",
+        replace=True,
+        index_cache_size=256,
+        num_bits=4,
+        target_partition_size=8192,
+    )
+    expected_config = IvfPq(
+        distance_type="l2",
+        num_sub_vectors=96,
+        num_bits=4,
+        target_partition_size=8192,
+    )
+    mock_create_index.assert_called_with(
+        "vector", replace=True, config=expected_config, name=None, train=True
+    )
+
+    # target_partition_size has a default value,
+    # so `num_partitions` and `target_partition_size` are not required
+    table.create_index(
+        metric="l2",
+        num_sub_vectors=96,
+        vector_column_name="vector",
+        replace=True,
+        index_cache_size=256,
+        num_bits=4,
+    )
+    expected_config = IvfPq(
+        distance_type="l2",
+        num_sub_vectors=96,
+        num_bits=4,
+    )
+    mock_create_index.assert_called_with(
+        "vector", replace=True, config=expected_config, name=None, train=True
+    )
+
     table.create_index(
         vector_column_name="my_vector",
         metric="dot",

python/src/connection.rs

@@ -4,7 +4,10 @@
 use std::{collections::HashMap, sync::Arc, time::Duration};
 
 use arrow::{datatypes::Schema, ffi_stream::ArrowArrayStreamReader, pyarrow::FromPyArrow};
-use lancedb::{connection::Connection as LanceConnection, database::CreateTableMode};
+use lancedb::{
+    connection::Connection as LanceConnection,
+    database::{CreateTableMode, ReadConsistency},
+};
 use pyo3::{
     exceptions::{PyRuntimeError, PyValueError},
     pyclass, pyfunction, pymethods, Bound, FromPyObject, Py, PyAny, PyRef, PyResult, Python,
@@ -23,7 +26,7 @@ impl Connection {
         Self { inner: Some(inner) }
     }
 
-    fn get_inner(&self) -> PyResult<&LanceConnection> {
+    pub(crate) fn get_inner(&self) -> PyResult<&LanceConnection> {
         self.inner
             .as_ref()
             .ok_or_else(|| PyRuntimeError::new_err("Connection is closed"))
@@ -63,6 +66,18 @@ impl Connection {
         self.get_inner().map(|inner| inner.uri().to_string())
     }
 
+    #[pyo3(signature = ())]
+    pub fn get_read_consistency_interval(self_: PyRef<'_, Self>) -> PyResult<Bound<'_, PyAny>> {
+        let inner = self_.get_inner()?.clone();
+        future_into_py(self_.py(), async move {
+            Ok(match inner.read_consistency().await.infer_error()? {
+                ReadConsistency::Manual => None,
+                ReadConsistency::Eventual(duration) => Some(duration.as_secs_f64()),
+                ReadConsistency::Strong => Some(0.0_f64),
+            })
+        })
+    }
+
     #[pyo3(signature = (namespace=vec![], start_after=None, limit=None))]
     pub fn table_names(
         self_: PyRef<'_, Self>,
@@ -163,6 +178,34 @@ impl Connection {
         })
     }
 
+    #[pyo3(signature = (target_table_name, source_uri, target_namespace=vec![], source_version=None, source_tag=None, is_shallow=true))]
+    pub fn clone_table(
+        self_: PyRef<'_, Self>,
+        target_table_name: String,
+        source_uri: String,
+        target_namespace: Vec<String>,
+        source_version: Option<u64>,
+        source_tag: Option<String>,
+        is_shallow: bool,
+    ) -> PyResult<Bound<'_, PyAny>> {
+        let inner = self_.get_inner()?.clone();
+
+        let mut builder = inner.clone_table(target_table_name, source_uri);
+        builder = builder.target_namespace(target_namespace);
+        if let Some(version) = source_version {
+            builder = builder.source_version(version);
+        }
+        if let Some(tag) = source_tag {
+            builder = builder.source_tag(tag);
+        }
+        builder = builder.is_shallow(is_shallow);
+
+        future_into_py(self_.py(), async move {
+            let table = builder.execute().await.infer_error()?;
+            Ok(Table::new(table))
+        })
+    }
+
     #[pyo3(signature = (cur_name, new_name, cur_namespace=vec![], new_namespace=vec![]))]
     pub fn rename_table(
         self_: PyRef<'_, Self>,
@@ -255,7 +298,7 @@ impl Connection {
 #[pyo3(signature = (uri, api_key=None, region=None, host_override=None, read_consistency_interval=None, client_config=None, storage_options=None, session=None))]
 #[allow(clippy::too_many_arguments)]
 pub fn connect(
-    py: Python,
+    py: Python<'_>,
     uri: String,
     api_key: Option<String>,
     region: Option<String>,

python/src/index.rs

@@ -1,7 +1,7 @@
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
-use lancedb::index::vector::IvfFlatIndexBuilder;
+use lancedb::index::vector::{IvfFlatIndexBuilder, IvfRqIndexBuilder};
 use lancedb::index::{
     scalar::{BTreeIndexBuilder, FtsIndexBuilder},
     vector::{IvfHnswPqIndexBuilder, IvfHnswSqIndexBuilder, IvfPqIndexBuilder},
@@ -63,6 +63,9 @@ pub fn extract_index_params(source: &Option<Bound<'_, PyAny>>) -> PyResult<Lance
                 if let Some(num_partitions) = params.num_partitions {
                     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);
+                }
                 Ok(LanceDbIndex::IvfFlat(ivf_flat_builder))
             },
             "IvfPq" => {
@@ -76,11 +79,30 @@ pub fn extract_index_params(source: &Option<Bound<'_, PyAny>>) -> PyResult<Lance
                 if let Some(num_partitions) = params.num_partitions {
                     ivf_pq_builder = ivf_pq_builder.num_partitions(num_partitions);
                 }
+                if let Some(target_partition_size) = params.target_partition_size {
+                    ivf_pq_builder = ivf_pq_builder.target_partition_size(target_partition_size);
+                }
                 if let Some(num_sub_vectors) = params.num_sub_vectors {
                     ivf_pq_builder = ivf_pq_builder.num_sub_vectors(num_sub_vectors);
                 }
                 Ok(LanceDbIndex::IvfPq(ivf_pq_builder))
             },
+            "IvfRq" => {
+                let params = source.extract::<IvfRqParams>()?;
+                let distance_type = parse_distance_type(params.distance_type)?;
+                let mut ivf_rq_builder = IvfRqIndexBuilder::default()
+                    .distance_type(distance_type)
+                    .max_iterations(params.max_iterations)
+                    .sample_rate(params.sample_rate)
+                    .num_bits(params.num_bits);
+                if let Some(num_partitions) = params.num_partitions {
+                    ivf_rq_builder = ivf_rq_builder.num_partitions(num_partitions);
+                }
+                if let Some(target_partition_size) = params.target_partition_size {
+                    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)?;
@@ -94,6 +116,9 @@ pub fn extract_index_params(source: &Option<Bound<'_, PyAny>>) -> PyResult<Lance
                 if let Some(num_partitions) = params.num_partitions {
                     hnsw_pq_builder = hnsw_pq_builder.num_partitions(num_partitions);
                 }
+                if let Some(target_partition_size) = params.target_partition_size {
+                    hnsw_pq_builder = hnsw_pq_builder.target_partition_size(target_partition_size);
+                }
                 if let Some(num_sub_vectors) = params.num_sub_vectors {
                     hnsw_pq_builder = hnsw_pq_builder.num_sub_vectors(num_sub_vectors);
                 }
@@ -111,6 +136,9 @@ pub fn extract_index_params(source: &Option<Bound<'_, PyAny>>) -> PyResult<Lance
                 if let Some(num_partitions) = params.num_partitions {
                     hnsw_sq_builder = hnsw_sq_builder.num_partitions(num_partitions);
                 }
+                if let Some(target_partition_size) = params.target_partition_size {
+                    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!(
@@ -144,6 +172,7 @@ struct IvfFlatParams {
     num_partitions: Option<u32>,
     max_iterations: u32,
     sample_rate: u32,
+    target_partition_size: Option<u32>,
 }
 
 #[derive(FromPyObject)]
@@ -154,6 +183,17 @@ struct IvfPqParams {
     num_bits: u32,
     max_iterations: u32,
     sample_rate: u32,
+    target_partition_size: Option<u32>,
+}
+
+#[derive(FromPyObject)]
+struct IvfRqParams {
+    distance_type: String,
+    num_partitions: Option<u32>,
+    num_bits: u32,
+    max_iterations: u32,
+    sample_rate: u32,
+    target_partition_size: Option<u32>,
 }
 
 #[derive(FromPyObject)]
@@ -166,6 +206,7 @@ struct IvfHnswPqParams {
     sample_rate: u32,
     m: u32,
     ef_construction: u32,
+    target_partition_size: Option<u32>,
 }
 
 #[derive(FromPyObject)]
@@ -176,6 +217,7 @@ struct IvfHnswSqParams {
     sample_rate: u32,
     m: u32,
     ef_construction: u32,
+    target_partition_size: Option<u32>,
 }
 
 #[pyclass(get_all)]

python/src/lib.rs

@@ -5,6 +5,7 @@ use arrow::RecordBatchStream;
 use connection::{connect, Connection};
 use env_logger::Env;
 use index::IndexConfig;
+use permutation::PyAsyncPermutationBuilder;
 use pyo3::{
     pymodule,
     types::{PyModule, PyModuleMethods},
@@ -22,6 +23,7 @@ pub mod connection;
 pub mod error;
 pub mod header;
 pub mod index;
+pub mod permutation;
 pub mod query;
 pub mod session;
 pub mod table;
@@ -49,7 +51,9 @@ pub fn _lancedb(_py: Python, m: &Bound<'_, PyModule>) -> PyResult<()> {
     m.add_class::<DeleteResult>()?;
     m.add_class::<DropColumnsResult>()?;
     m.add_class::<UpdateResult>()?;
+    m.add_class::<PyAsyncPermutationBuilder>()?;
     m.add_function(wrap_pyfunction!(connect, m)?)?;
+    m.add_function(wrap_pyfunction!(permutation::async_permutation_builder, m)?)?;
     m.add_function(wrap_pyfunction!(util::validate_table_name, m)?)?;
     m.add("__version__", env!("CARGO_PKG_VERSION"))?;
     Ok(())

python/src/permutation.rs

@@ -0,0 +1,177 @@
+// SPDX-License-Identifier: Apache-2.0
+// SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+use std::sync::{Arc, Mutex};
+
+use crate::{error::PythonErrorExt, table::Table};
+use lancedb::dataloader::{
+    permutation::{PermutationBuilder as LancePermutationBuilder, ShuffleStrategy},
+    split::{SplitSizes, SplitStrategy},
+};
+use pyo3::{
+    exceptions::PyRuntimeError, pyclass, pymethods, types::PyAnyMethods, Bound, PyAny, PyRefMut,
+    PyResult,
+};
+use pyo3_async_runtimes::tokio::future_into_py;
+
+/// Create a permutation builder for the given table
+#[pyo3::pyfunction]
+pub fn async_permutation_builder(
+    table: Bound<'_, PyAny>,
+    dest_table_name: String,
+) -> PyResult<PyAsyncPermutationBuilder> {
+    let table = table.getattr("_inner")?.downcast_into::<Table>()?;
+    let inner_table = table.borrow().inner_ref()?.clone();
+    let inner_builder = LancePermutationBuilder::new(inner_table);
+
+    Ok(PyAsyncPermutationBuilder {
+        state: Arc::new(Mutex::new(PyAsyncPermutationBuilderState {
+            builder: Some(inner_builder),
+            dest_table_name,
+        })),
+    })
+}
+
+struct PyAsyncPermutationBuilderState {
+    builder: Option<LancePermutationBuilder>,
+    dest_table_name: String,
+}
+
+#[pyclass(name = "AsyncPermutationBuilder")]
+pub struct PyAsyncPermutationBuilder {
+    state: Arc<Mutex<PyAsyncPermutationBuilderState>>,
+}
+
+impl PyAsyncPermutationBuilder {
+    fn modify(
+        &self,
+        func: impl FnOnce(LancePermutationBuilder) -> LancePermutationBuilder,
+    ) -> PyResult<Self> {
+        let mut state = self.state.lock().unwrap();
+        let builder = state
+            .builder
+            .take()
+            .ok_or_else(|| PyRuntimeError::new_err("Builder already consumed"))?;
+        state.builder = Some(func(builder));
+        Ok(Self {
+            state: self.state.clone(),
+        })
+    }
+}
+
+#[pymethods]
+impl PyAsyncPermutationBuilder {
+    #[pyo3(signature = (*, ratios=None, counts=None, fixed=None, seed=None))]
+    pub fn split_random(
+        slf: PyRefMut<'_, Self>,
+        ratios: Option<Vec<f64>>,
+        counts: Option<Vec<u64>>,
+        fixed: Option<u64>,
+        seed: Option<u64>,
+    ) -> PyResult<Self> {
+        // Check that exactly one split type is provided
+        let split_args_count = [ratios.is_some(), counts.is_some(), fixed.is_some()]
+            .iter()
+            .filter(|&&x| x)
+            .count();
+
+        if split_args_count != 1 {
+            return Err(pyo3::exceptions::PyValueError::new_err(
+                "Exactly one of 'ratios', 'counts', or 'fixed' must be provided",
+            ));
+        }
+
+        let sizes = if let Some(ratios) = ratios {
+            SplitSizes::Percentages(ratios)
+        } else if let Some(counts) = counts {
+            SplitSizes::Counts(counts)
+        } else if let Some(fixed) = fixed {
+            SplitSizes::Fixed(fixed)
+        } else {
+            unreachable!("One of the split arguments must be provided");
+        };
+
+        slf.modify(|builder| builder.with_split_strategy(SplitStrategy::Random { seed, sizes }))
+    }
+
+    #[pyo3(signature = (columns, split_weights, *, discard_weight=0))]
+    pub fn split_hash(
+        slf: PyRefMut<'_, Self>,
+        columns: Vec<String>,
+        split_weights: Vec<u64>,
+        discard_weight: u64,
+    ) -> PyResult<Self> {
+        slf.modify(|builder| {
+            builder.with_split_strategy(SplitStrategy::Hash {
+                columns,
+                split_weights,
+                discard_weight,
+            })
+        })
+    }
+
+    #[pyo3(signature = (*, ratios=None, counts=None, fixed=None))]
+    pub fn split_sequential(
+        slf: PyRefMut<'_, Self>,
+        ratios: Option<Vec<f64>>,
+        counts: Option<Vec<u64>>,
+        fixed: Option<u64>,
+    ) -> PyResult<Self> {
+        // Check that exactly one split type is provided
+        let split_args_count = [ratios.is_some(), counts.is_some(), fixed.is_some()]
+            .iter()
+            .filter(|&&x| x)
+            .count();
+
+        if split_args_count != 1 {
+            return Err(pyo3::exceptions::PyValueError::new_err(
+                "Exactly one of 'ratios', 'counts', or 'fixed' must be provided",
+            ));
+        }
+
+        let sizes = if let Some(ratios) = ratios {
+            SplitSizes::Percentages(ratios)
+        } else if let Some(counts) = counts {
+            SplitSizes::Counts(counts)
+        } else if let Some(fixed) = fixed {
+            SplitSizes::Fixed(fixed)
+        } else {
+            unreachable!("One of the split arguments must be provided");
+        };
+
+        slf.modify(|builder| builder.with_split_strategy(SplitStrategy::Sequential { sizes }))
+    }
+
+    pub fn split_calculated(slf: PyRefMut<'_, Self>, calculation: String) -> PyResult<Self> {
+        slf.modify(|builder| builder.with_split_strategy(SplitStrategy::Calculated { calculation }))
+    }
+
+    pub fn shuffle(
+        slf: PyRefMut<'_, Self>,
+        seed: Option<u64>,
+        clump_size: Option<u64>,
+    ) -> PyResult<Self> {
+        slf.modify(|builder| {
+            builder.with_shuffle_strategy(ShuffleStrategy::Random { seed, clump_size })
+        })
+    }
+
+    pub fn filter(slf: PyRefMut<'_, Self>, filter: String) -> PyResult<Self> {
+        slf.modify(|builder| builder.with_filter(filter))
+    }
+
+    pub fn execute(slf: PyRefMut<'_, Self>) -> PyResult<Bound<'_, PyAny>> {
+        let mut state = slf.state.lock().unwrap();
+        let builder = state
+            .builder
+            .take()
+            .ok_or_else(|| PyRuntimeError::new_err("Builder already consumed"))?;
+
+        let dest_table_name = std::mem::take(&mut state.dest_table_name);
+
+        future_into_py(slf.py(), async move {
+            let table = builder.build(&dest_table_name).await.infer_error()?;
+            Ok(Table::new(table))
+        })
+    }
+}

python/src/table.rs

@@ -3,6 +3,7 @@
 use std::{collections::HashMap, sync::Arc};
 
 use crate::{
+    connection::Connection,
     error::PythonErrorExt,
     index::{extract_index_params, IndexConfig},
     query::{Query, TakeQuery},
@@ -249,7 +250,7 @@ impl Table {
 }
 
 impl Table {
-    fn inner_ref(&self) -> PyResult<&LanceDbTable> {
+    pub(crate) fn inner_ref(&self) -> PyResult<&LanceDbTable> {
         self.inner
             .as_ref()
             .ok_or_else(|| PyRuntimeError::new_err(format!("Table {} is closed", self.name)))
@@ -272,6 +273,13 @@ impl Table {
         self.inner.take();
     }
 
+    pub fn database(&self) -> PyResult<Connection> {
+        let inner = self.inner_ref()?.clone();
+        let inner_connection =
+            lancedb::Connection::new(inner.database().clone(), inner.embedding_registry().clone());
+        Ok(Connection::new(inner_connection))
+    }
+
     pub fn schema(self_: PyRef<'_, Self>) -> PyResult<Bound<'_, PyAny>> {
         let inner = self_.inner_ref()?.clone();
         future_into_py(self_.py(), async move {
@@ -591,12 +599,11 @@ impl Table {
     }
 
     /// Optimize the on-disk data by compacting and pruning old data, for better performance.
-    #[pyo3(signature = (cleanup_since_ms=None, delete_unverified=None, retrain=None))]
+    #[pyo3(signature = (cleanup_since_ms=None, delete_unverified=None))]
     pub fn optimize(
         self_: PyRef<'_, Self>,
         cleanup_since_ms: Option<u64>,
         delete_unverified: Option<bool>,
-        retrain: Option<bool>,
     ) -> PyResult<Bound<'_, PyAny>> {
         let inner = self_.inner_ref()?.clone();
         let older_than = if let Some(ms) = cleanup_since_ms {
@@ -632,10 +639,9 @@ impl Table {
                 .prune
                 .unwrap();
             inner
-                .optimize(lancedb::table::OptimizeAction::Index(match retrain {
+                .optimize(lancedb::table::OptimizeAction::Index(
-                    Some(true) => OptimizeOptions::retrain(),
+                    OptimizeOptions::default(),
-                    _ => OptimizeOptions::default(),
+                ))
-                }))
                 .await
                 .infer_error()?;
             Ok(OptimizeStats {
@@ -674,6 +680,9 @@ impl Table {
         if let Some(timeout) = parameters.timeout {
             builder.timeout(timeout);
         }
+        if let Some(use_index) = parameters.use_index {
+            builder.use_index(use_index);
+        }
 
         future_into_py(self_.py(), async move {
             let res = builder.execute(Box::new(batches)).await.infer_error()?;
@@ -833,6 +842,7 @@ pub struct MergeInsertParams {
     when_not_matched_by_source_delete: bool,
     when_not_matched_by_source_condition: Option<String>,
     timeout: Option<std::time::Duration>,
+    use_index: Option<bool>,
 }
 
 #[pyclass]

rust-toolchain.toml

@@ -1,2 +1,2 @@
 [toolchain]
-channel = "1.86.0"
+channel = "1.90.0"

rust/lancedb/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb"
-version = "0.22.1-beta.0"
+version = "0.22.2"
 edition.workspace = true
 description = "LanceDB: A serverless, low-latency vector database for AI applications"
 license.workspace = true
@@ -11,6 +11,7 @@ rust-version.workspace = true
 
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 [dependencies]
+ahash = { workspace = true }
 arrow = { workspace = true }
 arrow-array = { workspace = true }
 arrow-data = { workspace = true }
@@ -24,18 +25,23 @@ datafusion-common.workspace = true
 datafusion-execution.workspace = true
 datafusion-expr.workspace = true
 datafusion-physical-plan.workspace = true
+datafusion.workspace = true
 object_store = { workspace = true }
 snafu = { workspace = true }
 half = { workspace = true }
 lazy_static.workspace = true
 lance = { workspace = true }
+lance-core = { workspace = true }
 lance-datafusion.workspace = true
+lance-datagen = { workspace = true }
+lance-file = { workspace = true }
 lance-io = { workspace = true }
 lance-index = { workspace = true }
 lance-table = { workspace = true }
 lance-linalg = { workspace = true }
 lance-testing = { workspace = true }
 lance-encoding = { workspace = true }
+lance-namespace = { workspace = true }
 moka = { workspace = true }
 pin-project = { workspace = true }
 tokio = { version = "1.23", features = ["rt-multi-thread"] }
@@ -45,11 +51,13 @@ bytes = "1"
 futures.workspace = true
 num-traits.workspace = true
 url.workspace = true
+rand.workspace = true
 regex.workspace = true
 serde = { version = "^1" }
 serde_json = { version = "1" }
 async-openai = { version = "0.20.0", optional = true }
 serde_with = { version = "3.8.1" }
+tempfile = "3.5.0"
 aws-sdk-bedrockruntime = { version = "1.27.0", optional = true }
 # For remote feature
 reqwest = { version = "0.12.0", default-features = false, features = [
@@ -60,9 +68,8 @@ reqwest = { version = "0.12.0", default-features = false, features = [
     "macos-system-configuration",
     "stream",
 ], optional = true }
-rand = { version = "0.9", features = ["small_rng"], optional = true }
 http = { version = "1", optional = true } # Matching what is in reqwest
-uuid = { version = "1.7.0", features = ["v4"], optional = true }
+uuid = { version = "1.7.0", features = ["v4"] }
 polars-arrow = { version = ">=0.37,<0.40.0", optional = true }
 polars = { version = ">=0.37,<0.40.0", optional = true }
 hf-hub = { version = "0.4.1", optional = true, default-features = false, features = [
@@ -81,19 +88,20 @@ crunchy.workspace = true
 bytemuck_derive.workspace = true
 
 [dev-dependencies]
+anyhow = "1"
 tempfile = "3.5.0"
-rand = { version = "0.9", features = ["small_rng"] }
 random_word = { version = "0.4.3", features = ["en"] }
 uuid = { version = "1.7.0", features = ["v4"] }
 walkdir = "2"
-aws-sdk-dynamodb = { version = "1.38.0" }
+aws-sdk-dynamodb = { version = "1.55.0" }
-aws-sdk-s3 = { version = "1.38.0" }
+aws-sdk-s3 = { version = "1.55.0" }
-aws-sdk-kms = { version = "1.37" }
+aws-sdk-kms = { version = "1.48.0" }
-aws-config = { version = "1.0" }
+aws-config = { version = "1.5.10" }
-aws-smithy-runtime = { version = "1.3" }
+aws-smithy-runtime = { version = "1.9.1" }
 datafusion.workspace = true
 http-body = "1"                                        # Matching reqwest
 rstest = "0.23.0"
+test-log = "0.2"
 
 
 [features]
@@ -103,7 +111,7 @@ oss = ["lance/oss", "lance-io/oss"]
 gcs = ["lance/gcp", "lance-io/gcp"]
 azure = ["lance/azure", "lance-io/azure"]
 dynamodb = ["lance/dynamodb", "aws"]
-remote = ["dep:reqwest", "dep:http", "dep:rand", "dep:uuid"]
+remote = ["dep:reqwest", "dep:http"]
 fp16kernels = ["lance-linalg/fp16kernels"]
 s3-test = []
 bedrock = ["dep:aws-sdk-bedrockruntime"]