Bump version: 0.17.0-beta.2 → 0.17.0-beta.3

Add remote database folder prefix
support db://bucket/path/to/folder/

.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.13.0"
+current_version = "0.14.0-beta.1"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

.github/workflows/npm-publish.yml

@@ -133,7 +133,7 @@ jobs:
           free -h
       - name: Build Linux Artifacts
         run: |
-          bash ci/build_linux_artifacts.sh ${{ matrix.config.arch }}
+          bash ci/build_linux_artifacts.sh ${{ matrix.config.arch }} ${{ matrix.config.arch }}-unknown-linux-gnu
       - name: Upload Linux Artifacts
         uses: actions/upload-artifact@v4
         with:
@@ -143,7 +143,7 @@ jobs:
 
   node-linux-musl:
     name: vectordb (${{ matrix.config.arch}}-unknown-linux-musl)
-    runs-on: ubuntu-latest
+    runs-on: ${{ matrix.config.runner }}
     container: alpine:edge
     # Only runs on tags that matches the make-release action
     if: startsWith(github.ref, 'refs/tags/v')
@@ -152,7 +152,10 @@ jobs:
       matrix:
         config:
           - arch: x86_64
+            runner: ubuntu-latest
           - arch: aarch64
+            # For successful fat LTO builds, we need a large runner to avoid OOM errors.
+            runner: buildjet-16vcpu-ubuntu-2204-arm
     steps:
       - name: Checkout
         uses: actions/checkout@v4
@@ -185,7 +188,7 @@ jobs:
       - name: Build Linux Artifacts
         run: |
           source ./saved_env
-          bash ci/manylinux_node/build_vectordb.sh ${{ matrix.config.arch }}
+          bash ci/manylinux_node/build_vectordb.sh ${{ matrix.config.arch }} ${{ matrix.config.arch }}-unknown-linux-musl
       - name: Upload Linux Artifacts
         uses: actions/upload-artifact@v4
         with:
@@ -246,7 +249,7 @@ jobs:
 
   nodejs-linux-musl:
     name: lancedb (${{ matrix.config.arch}}-unknown-linux-musl
-    runs-on: ubuntu-latest
+    runs-on: ${{ matrix.config.runner }}
     container: alpine:edge
     # Only runs on tags that matches the make-release action
     if: startsWith(github.ref, 'refs/tags/v')
@@ -255,7 +258,10 @@ jobs:
       matrix:
         config:
           - arch: x86_64
+            runner: ubuntu-latest
           - arch: aarch64
+            # For successful fat LTO builds, we need a large runner to avoid OOM errors.
+            runner: buildjet-16vcpu-ubuntu-2204-arm
     steps:
       - name: Checkout
         uses: actions/checkout@v4
@@ -334,6 +340,50 @@ jobs:
           path: |
             node/dist/lancedb-vectordb-win32*.tgz
 
+  node-windows-arm64:
+    name: vectordb ${{ matrix.config.arch }}-pc-windows-msvc
+    runs-on: ubuntu-latest
+    container: alpine:edge
+    strategy:
+      fail-fast: false
+      matrix:
+        config:
+          # - arch: x86_64
+          - arch: aarch64
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Install dependencies
+        run: |
+          apk add protobuf-dev curl clang lld llvm19 grep npm bash msitools sed
+          curl --proto '=https' --tlsv1.3 -sSf https://raw.githubusercontent.com/rust-lang/rustup/refs/heads/master/rustup-init.sh | sh -s -- -y --default-toolchain 1.80.0
+          echo "source $HOME/.cargo/env" >> saved_env
+          echo "export CC=clang" >> saved_env
+          echo "export AR=llvm-ar" >> saved_env
+          source "$HOME/.cargo/env"
+          rustup target add ${{ matrix.config.arch }}-pc-windows-msvc --toolchain 1.80.0
+          (mkdir -p sysroot && cd sysroot && sh ../ci/sysroot-${{ matrix.config.arch }}-pc-windows-msvc.sh)
+          echo "export C_INCLUDE_PATH=/usr/${{ matrix.config.arch }}-pc-windows-msvc/usr/include" >> saved_env
+          echo "export CARGO_BUILD_TARGET=${{ matrix.config.arch }}-pc-windows-msvc" >> saved_env
+      - name: Configure x86_64 build
+        if: ${{ matrix.config.arch == 'x86_64' }}
+        run: |
+          echo "export RUSTFLAGS='-Ctarget-cpu=haswell -Ctarget-feature=+crt-static,+avx2,+fma,+f16c -Clinker=lld -Clink-arg=/LIBPATH:/usr/x86_64-pc-windows-msvc/usr/lib'" >> saved_env
+      - name: Configure aarch64 build
+        if: ${{ matrix.config.arch == 'aarch64' }}
+        run: |
+          echo "export RUSTFLAGS='-Ctarget-feature=+crt-static,+neon,+fp16,+fhm,+dotprod -Clinker=lld -Clink-arg=/LIBPATH:/usr/aarch64-pc-windows-msvc/usr/lib -Clink-arg=arm64rt.lib'" >> saved_env
+      - name: Build Windows Artifacts
+        run: |
+          source ./saved_env
+          bash ci/manylinux_node/build_vectordb.sh ${{ matrix.config.arch }} ${{ matrix.config.arch }}-pc-windows-msvc
+      - name: Upload Windows Artifacts
+        uses: actions/upload-artifact@v4
+        with:
+          name: node-native-windows-${{ matrix.config.arch }}
+          path: |
+            node/dist/lancedb-vectordb-win32*.tgz
+
   # TODO: re-enable once working https://github.com/lancedb/lancedb/pull/1831
   # node-windows-arm64:
   #   name: vectordb win32-arm64-msvc
@@ -472,6 +522,52 @@ jobs:
           path: |
             nodejs/dist/*.node
 
+  nodejs-windows-arm64:
+    name: lancedb ${{ matrix.config.arch }}-pc-windows-msvc
+    runs-on: ubuntu-latest
+    container: alpine:edge
+    strategy:
+      fail-fast: false
+      matrix:
+        config:
+          # - arch: x86_64
+          - arch: aarch64
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Install dependencies
+        run: |
+          apk add protobuf-dev curl clang lld llvm19 grep npm bash msitools sed
+          curl --proto '=https' --tlsv1.3 -sSf https://raw.githubusercontent.com/rust-lang/rustup/refs/heads/master/rustup-init.sh | sh -s -- -y --default-toolchain 1.80.0
+          echo "source $HOME/.cargo/env" >> saved_env
+          echo "export CC=clang" >> saved_env
+          echo "export AR=llvm-ar" >> saved_env
+          source "$HOME/.cargo/env"
+          rustup target add ${{ matrix.config.arch }}-pc-windows-msvc --toolchain 1.80.0
+          (mkdir -p sysroot && cd sysroot && sh ../ci/sysroot-${{ matrix.config.arch }}-pc-windows-msvc.sh)
+          echo "export C_INCLUDE_PATH=/usr/${{ matrix.config.arch }}-pc-windows-msvc/usr/include" >> saved_env
+          echo "export CARGO_BUILD_TARGET=${{ matrix.config.arch }}-pc-windows-msvc" >> saved_env
+          printf '#!/bin/sh\ncargo "$@"' > $HOME/.cargo/bin/cargo-xwin
+          chmod u+x $HOME/.cargo/bin/cargo-xwin
+      - name: Configure x86_64 build
+        if: ${{ matrix.config.arch == 'x86_64' }}
+        run: |
+          echo "export RUSTFLAGS='-Ctarget-cpu=haswell -Ctarget-feature=+crt-static,+avx2,+fma,+f16c -Clinker=lld -Clink-arg=/LIBPATH:/usr/x86_64-pc-windows-msvc/usr/lib'" >> saved_env
+      - name: Configure aarch64 build
+        if: ${{ matrix.config.arch == 'aarch64' }}
+        run: |
+          echo "export RUSTFLAGS='-Ctarget-feature=+crt-static,+neon,+fp16,+fhm,+dotprod -Clinker=lld -Clink-arg=/LIBPATH:/usr/aarch64-pc-windows-msvc/usr/lib -Clink-arg=arm64rt.lib'" >> saved_env
+      - name: Build Windows Artifacts
+        run: |
+          source ./saved_env
+          bash ci/manylinux_node/build_lancedb.sh ${{ matrix.config.arch }}
+      - name: Upload Windows Artifacts
+        uses: actions/upload-artifact@v4
+        with:
+          name: nodejs-native-windows-${{ matrix.config.arch }}
+          path: |
+            nodejs/dist/*.node
+
   # TODO: re-enable once working https://github.com/lancedb/lancedb/pull/1831
   # nodejs-windows-arm64:
   #   name: lancedb win32-arm64-msvc
@@ -568,7 +664,7 @@ jobs:
 
   release:
     name: vectordb NPM Publish
-    needs: [node, node-macos, node-linux-gnu, node-linux-musl, node-windows]
+    needs: [node, node-macos, node-linux-gnu, node-linux-musl, node-windows, node-windows-arm64]
     runs-on: ubuntu-latest
     # Only runs on tags that matches the make-release action
     if: startsWith(github.ref, 'refs/tags/v')
@@ -608,7 +704,7 @@ jobs:
 
   release-nodejs:
     name: lancedb NPM Publish
-    needs: [nodejs-macos, nodejs-linux-gnu, nodejs-linux-musl, nodejs-windows]
+    needs: [nodejs-macos, nodejs-linux-gnu, nodejs-linux-musl, nodejs-windows, nodejs-windows-arm64]
     runs-on: ubuntu-latest
     # Only runs on tags that matches the make-release action
     if: startsWith(github.ref, 'refs/tags/v')

.github/workflows/pypi-publish.yml

@@ -83,7 +83,7 @@ jobs:
       - name: Set up Python
         uses: actions/setup-python@v4
         with:
-          python-version: 3.8
+          python-version: 3.12
       - uses: ./.github/workflows/build_windows_wheel
         with:
           python-minor-version: 8

.github/workflows/upload_wheel/action.yml

@@ -17,6 +17,7 @@ runs:
     run: |
       python -m pip install --upgrade pip
       pip install twine
+      python3 -m pip install --upgrade pkginfo
   - name: Choose repo
     shell: bash
     id: choose_repo

Cargo.toml

@@ -23,13 +23,14 @@ rust-version = "1.80.0"                                                     # TO
 [workspace.dependencies]
 lance = { "version" = "=0.20.0", "features" = [
     "dynamodb",
-], git = "https://github.com/lancedb/lance.git", tag = "v0.20.0-beta.2" }
+], git = "https://github.com/lancedb/lance.git", tag = "v0.20.0-beta.3" }
-lance-index = { version = "=0.20.0", git = "https://github.com/lancedb/lance.git", tag = "v0.20.0-beta.2" }
+lance-io = { version = "=0.20.0", git = "https://github.com/lancedb/lance.git", tag = "v0.20.0-beta.3" }
-lance-linalg = { version = "=0.20.0", git = "https://github.com/lancedb/lance.git", tag = "v0.20.0-beta.2" }
+lance-index = { version = "=0.20.0", git = "https://github.com/lancedb/lance.git", tag = "v0.20.0-beta.3" }
-lance-table = { version = "=0.20.0", git = "https://github.com/lancedb/lance.git", tag = "v0.20.0-beta.2" }
+lance-linalg = { version = "=0.20.0", git = "https://github.com/lancedb/lance.git", tag = "v0.20.0-beta.3" }
-lance-testing = { version = "=0.20.0", git = "https://github.com/lancedb/lance.git", tag = "v0.20.0-beta.2" }
+lance-table = { version = "=0.20.0", git = "https://github.com/lancedb/lance.git", tag = "v0.20.0-beta.3" }
-lance-datafusion = { version = "=0.20.0", git = "https://github.com/lancedb/lance.git", tag = "v0.20.0-beta.2" }
+lance-testing = { version = "=0.20.0", git = "https://github.com/lancedb/lance.git", tag = "v0.20.0-beta.3" }
-lance-encoding = { version = "=0.20.0", git = "https://github.com/lancedb/lance.git", tag = "v0.20.0-beta.2" }
+lance-datafusion = { version = "=0.20.0", git = "https://github.com/lancedb/lance.git", tag = "v0.20.0-beta.3" }
+lance-encoding = { version = "=0.20.0", git = "https://github.com/lancedb/lance.git", tag = "v0.20.0-beta.3" }
 # Note that this one does not include pyarrow
 arrow = { version = "52.2", optional = false }
 arrow-array = "52.2"

ci/build_linux_artifacts.sh

@@ -1,8 +1,9 @@
 #!/bin/bash
 set -e
 ARCH=${1:-x86_64}
+TARGET_TRIPLE=${2:-x86_64-unknown-linux-gnu}
 
-# We pass down the current user so that when we later mount the local files 
+# We pass down the current user so that when we later mount the local files
 # into the container, the files are accessible by the current user.
 pushd ci/manylinux_node
 docker build \
@@ -18,4 +19,4 @@ docker run \
     -v $(pwd):/io -w /io \
     --memory-swap=-1 \
     lancedb-node-manylinux \
-    bash ci/manylinux_node/build_vectordb.sh $ARCH
+    bash ci/manylinux_node/build_vectordb.sh $ARCH $TARGET_TRIPLE

ci/manylinux_node/build_vectordb.sh

@@ -2,6 +2,7 @@
 # Builds the node module for manylinux. Invoked by ci/build_linux_artifacts.sh.
 set -e
 ARCH=${1:-x86_64}
+TARGET_TRIPLE=${2:-x86_64-unknown-linux-gnu}
 
 if [ "$ARCH" = "x86_64" ]; then
     export OPENSSL_LIB_DIR=/usr/local/lib64/
@@ -17,4 +18,4 @@ FILE=$HOME/.bashrc && test -f $FILE && source $FILE
 cd node
 npm ci
 npm run build-release
-npm run pack-build
+npm run pack-build -- -t $TARGET_TRIPLE

ci/sysroot-aarch64-pc-windows-msvc.sh

@@ -0,0 +1,105 @@
+#!/bin/sh
+
+# https://github.com/mstorsjo/msvc-wine/blob/master/vsdownload.py
+# https://github.com/mozilla/gecko-dev/blob/6027d1d91f2d3204a3992633b3ef730ff005fc64/build/vs/vs2022-car.yaml
+
+# function dl() {
+# 	curl -O https://download.visualstudio.microsoft.com/download/pr/$1
+# }
+
+# [[.h]]
+
+# "id": "Win11SDK_10.0.26100"
+# "version": "10.0.26100.7"
+
+# libucrt.lib
+
+# example: <assert.h>
+# dir: ucrt/
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/2ee3a5fc6e9fc832af7295b138e93839/universal%20crt%20headers%20libraries%20and%20sources-x86_en-us.msi
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/b1aa09b90fe314aceb090f6ec7626624/16ab2ea2187acffa6435e334796c8c89.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/400609bb0ff5804e36dbe6dcd42a7f01/6ee7bbee8435130a869cf971694fd9e2.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/2ac327317abb865a0e3f56b2faefa918/78fa3c824c2c48bd4a49ab5969adaaf7.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/f034bc0b2680f67dccd4bfeea3d0f932/7afc7b670accd8e3cc94cfffd516f5cb.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/7ed5e12f9d50f80825a8b27838cf4c7f/96076045170fe5db6d5dcf14b6f6688e.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/764edc185a696bda9e07df8891dddbbb/a1e2a83aa8a71c48c742eeaff6e71928.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/66854bedc6dbd5ccb5dd82c8e2412231/b2f03f34ff83ec013b9e45c7cd8e8a73.cab
+
+# example: <windows.h>
+# dir: um/
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/b286efac4d83a54fc49190bddef1edc9/windows%20sdk%20for%20windows%20store%20apps%20headers-x86_en-us.msi
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/e0dc3811d92ab96fcb72bf63d6c08d71/766c0ffd568bbb31bf7fb6793383e24a.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/613503da4b5628768497822826aed39f/8125ee239710f33ea485965f76fae646.cab
+
+# example: <winapifamily.h>
+# dir: /shared
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/122979f0348d3a2a36b6aa1a111d5d0c/windows%20sdk%20for%20windows%20store%20apps%20headers%20onecoreuap-x86_en-us.msi
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/766e04beecdfccff39e91dd9eb32834a/e89e3dcbb016928c7e426238337d69eb.cab
+
+
+# "id": "Microsoft.VisualC.14.16.CRT.Headers"
+# "version": "14.16.27045"
+
+# example: <vcruntime.h>
+# dir: MSVC/
+curl -O https://download.visualstudio.microsoft.com/download/pr/bac0afd7-cc9e-4182-8a83-9898fa20e092/87bbe41e09a2f83711e72696f49681429327eb7a4b90618c35667a6ba2e2880e/Microsoft.VisualC.14.16.CRT.Headers.vsix
+
+# [[.lib]]
+
+# advapi32.lib bcrypt.lib kernel32.lib ntdll.lib user32.lib uuid.lib ws2_32.lib userenv.lib cfgmgr32.lib runtimeobject.lib
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/944c4153b849a1f7d0c0404a4f1c05ea/windows%20sdk%20for%20windows%20store%20apps%20libs-x86_en-us.msi
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/5306aed3e1a38d1e8bef5934edeb2a9b/05047a45609f311645eebcac2739fc4c.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/13c8a73a0f5a6474040b26d016a26fab/13d68b8a7b6678a368e2d13ff4027521.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/149578fb3b621cdb61ee1813b9b3e791/463ad1b0783ebda908fd6c16a4abfe93.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/5c986c4f393c6b09d5aec3b539e9fb4a/5a22e5cde814b041749fb271547f4dd5.cab
+
+# fwpuclnt.lib arm64rt.lib
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/7a332420d812f7c1d41da865ae5a7c52/windows%20sdk%20desktop%20libs%20arm64-x86_en-us.msi
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/19de98ed4a79938d0045d19c047936b3/3e2f7be479e3679d700ce0782e4cc318.cab
+
+# libcmt.lib libvcruntime.lib
+curl -O https://download.visualstudio.microsoft.com/download/pr/bac0afd7-cc9e-4182-8a83-9898fa20e092/227f40682a88dc5fa0ccb9cadc9ad30af99ad1f1a75db63407587d079f60d035/Microsoft.VisualC.14.16.CRT.ARM64.Desktop.vsix
+
+
+msiextract universal%20crt%20headers%20libraries%20and%20sources-x86_en-us.msi
+msiextract windows%20sdk%20for%20windows%20store%20apps%20headers-x86_en-us.msi
+msiextract windows%20sdk%20for%20windows%20store%20apps%20headers%20onecoreuap-x86_en-us.msi
+msiextract windows%20sdk%20for%20windows%20store%20apps%20libs-x86_en-us.msi
+msiextract windows%20sdk%20desktop%20libs%20arm64-x86_en-us.msi
+unzip -o Microsoft.VisualC.14.16.CRT.Headers.vsix
+unzip -o Microsoft.VisualC.14.16.CRT.ARM64.Desktop.vsix
+
+mkdir -p /usr/aarch64-pc-windows-msvc/usr/include
+mkdir -p /usr/aarch64-pc-windows-msvc/usr/lib
+
+# lowercase folder/file names
+echo "$(find . -regex ".*/[^/]*[A-Z][^/]*")" | xargs -I{} sh -c 'mv "$(echo "{}" | sed -E '"'"'s/(.*\/)/\L\1/'"'"')" "$(echo "{}" | tr [A-Z] [a-z])"'
+
+# .h
+(cd 'program files/windows kits/10/include/10.0.26100.0' && cp -r ucrt/* um/* shared/* -t /usr/aarch64-pc-windows-msvc/usr/include)
+
+cp -r contents/vc/tools/msvc/14.16.27023/include/* /usr/aarch64-pc-windows-msvc/usr/include
+
+# lowercase #include "" and #include <>
+find /usr/aarch64-pc-windows-msvc/usr/include -type f -exec sed -i -E 's/(#include <[^<>]*?[A-Z][^<>]*?>)|(#include "[^"]*?[A-Z][^"]*?")/\L\1\2/' "{}" ';'
+
+# ARM intrinsics
+# original dir: MSVC/
+
+# '__n128x4' redefined in arm_neon.h
+# "arm64_neon.h" included from intrin.h
+
+(cd /usr/lib/llvm19/lib/clang/19/include && cp arm_neon.h intrin.h -t /usr/aarch64-pc-windows-msvc/usr/include)
+
+# .lib
+
+# _Interlocked intrinsics
+# must always link with arm64rt.lib
+# reason: https://developercommunity.visualstudio.com/t/libucrtlibstreamobj-error-lnk2001-unresolved-exter/1544787#T-ND1599818
+# I don't understand the 'correct' fix for this, arm64rt.lib is supposed to be the workaround
+
+(cd 'program files/windows kits/10/lib/10.0.26100.0/um/arm64' && cp advapi32.lib bcrypt.lib kernel32.lib ntdll.lib user32.lib uuid.lib ws2_32.lib userenv.lib cfgmgr32.lib runtimeobject.lib fwpuclnt.lib arm64rt.lib -t /usr/aarch64-pc-windows-msvc/usr/lib)
+
+(cd 'contents/vc/tools/msvc/14.16.27023/lib/arm64' && cp libcmt.lib libvcruntime.lib -t /usr/aarch64-pc-windows-msvc/usr/lib)
+
+cp 'program files/windows kits/10/lib/10.0.26100.0/ucrt/arm64/libucrt.lib' /usr/aarch64-pc-windows-msvc/usr/lib

ci/sysroot-x86_64-pc-windows-msvc.sh

@@ -0,0 +1,105 @@
+#!/bin/sh
+
+# https://github.com/mstorsjo/msvc-wine/blob/master/vsdownload.py
+# https://github.com/mozilla/gecko-dev/blob/6027d1d91f2d3204a3992633b3ef730ff005fc64/build/vs/vs2022-car.yaml
+
+# function dl() {
+# 	curl -O https://download.visualstudio.microsoft.com/download/pr/$1
+# }
+
+# [[.h]]
+
+# "id": "Win11SDK_10.0.26100"
+# "version": "10.0.26100.7"
+
+# libucrt.lib
+
+# example: <assert.h>
+# dir: ucrt/
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/2ee3a5fc6e9fc832af7295b138e93839/universal%20crt%20headers%20libraries%20and%20sources-x86_en-us.msi
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/b1aa09b90fe314aceb090f6ec7626624/16ab2ea2187acffa6435e334796c8c89.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/400609bb0ff5804e36dbe6dcd42a7f01/6ee7bbee8435130a869cf971694fd9e2.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/2ac327317abb865a0e3f56b2faefa918/78fa3c824c2c48bd4a49ab5969adaaf7.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/f034bc0b2680f67dccd4bfeea3d0f932/7afc7b670accd8e3cc94cfffd516f5cb.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/7ed5e12f9d50f80825a8b27838cf4c7f/96076045170fe5db6d5dcf14b6f6688e.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/764edc185a696bda9e07df8891dddbbb/a1e2a83aa8a71c48c742eeaff6e71928.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/66854bedc6dbd5ccb5dd82c8e2412231/b2f03f34ff83ec013b9e45c7cd8e8a73.cab
+
+# example: <windows.h>
+# dir: um/
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/b286efac4d83a54fc49190bddef1edc9/windows%20sdk%20for%20windows%20store%20apps%20headers-x86_en-us.msi
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/e0dc3811d92ab96fcb72bf63d6c08d71/766c0ffd568bbb31bf7fb6793383e24a.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/613503da4b5628768497822826aed39f/8125ee239710f33ea485965f76fae646.cab
+
+# example: <winapifamily.h>
+# dir: /shared
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/122979f0348d3a2a36b6aa1a111d5d0c/windows%20sdk%20for%20windows%20store%20apps%20headers%20onecoreuap-x86_en-us.msi
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/766e04beecdfccff39e91dd9eb32834a/e89e3dcbb016928c7e426238337d69eb.cab
+
+
+# "id": "Microsoft.VisualC.14.16.CRT.Headers"
+# "version": "14.16.27045"
+
+# example: <vcruntime.h>
+# dir: MSVC/
+curl -O https://download.visualstudio.microsoft.com/download/pr/bac0afd7-cc9e-4182-8a83-9898fa20e092/87bbe41e09a2f83711e72696f49681429327eb7a4b90618c35667a6ba2e2880e/Microsoft.VisualC.14.16.CRT.Headers.vsix
+
+# [[.lib]]
+
+# advapi32.lib bcrypt.lib kernel32.lib ntdll.lib user32.lib uuid.lib ws2_32.lib userenv.lib cfgmgr32.lib
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/944c4153b849a1f7d0c0404a4f1c05ea/windows%20sdk%20for%20windows%20store%20apps%20libs-x86_en-us.msi
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/5306aed3e1a38d1e8bef5934edeb2a9b/05047a45609f311645eebcac2739fc4c.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/13c8a73a0f5a6474040b26d016a26fab/13d68b8a7b6678a368e2d13ff4027521.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/149578fb3b621cdb61ee1813b9b3e791/463ad1b0783ebda908fd6c16a4abfe93.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/5c986c4f393c6b09d5aec3b539e9fb4a/5a22e5cde814b041749fb271547f4dd5.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/bfc3904a0195453419ae4dfea7abd6fb/e10768bb6e9d0ea730280336b697da66.cab
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/637f9f3be880c71f9e3ca07b4d67345c/f9b24c8280986c0683fbceca5326d806.cab
+
+# dbghelp.lib fwpuclnt.lib
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/9f51690d5aa804b1340ce12d1ec80f89/windows%20sdk%20desktop%20libs%20x64-x86_en-us.msi
+curl -O https://download.visualstudio.microsoft.com/download/pr/32863b8d-a46d-4231-8e84-0888519d20a9/d3a7df4ca3303a698640a29e558a5e5b/58314d0646d7e1a25e97c902166c3155.cab
+
+# libcmt.lib libvcruntime.lib
+curl -O https://download.visualstudio.microsoft.com/download/pr/bac0afd7-cc9e-4182-8a83-9898fa20e092/8728f21ae09940f1f4b4ee47b4a596be2509e2a47d2f0c83bbec0ea37d69644b/Microsoft.VisualC.14.16.CRT.x64.Desktop.vsix
+
+
+msiextract universal%20crt%20headers%20libraries%20and%20sources-x86_en-us.msi
+msiextract windows%20sdk%20for%20windows%20store%20apps%20headers-x86_en-us.msi
+msiextract windows%20sdk%20for%20windows%20store%20apps%20headers%20onecoreuap-x86_en-us.msi
+msiextract windows%20sdk%20for%20windows%20store%20apps%20libs-x86_en-us.msi
+msiextract windows%20sdk%20desktop%20libs%20x64-x86_en-us.msi
+unzip -o Microsoft.VisualC.14.16.CRT.Headers.vsix
+unzip -o Microsoft.VisualC.14.16.CRT.x64.Desktop.vsix
+
+mkdir -p /usr/x86_64-pc-windows-msvc/usr/include
+mkdir -p /usr/x86_64-pc-windows-msvc/usr/lib
+
+# lowercase folder/file names
+echo "$(find . -regex ".*/[^/]*[A-Z][^/]*")" | xargs -I{} sh -c 'mv "$(echo "{}" | sed -E '"'"'s/(.*\/)/\L\1/'"'"')" "$(echo "{}" | tr [A-Z] [a-z])"'
+
+# .h
+(cd 'program files/windows kits/10/include/10.0.26100.0' && cp -r ucrt/* um/* shared/* -t /usr/x86_64-pc-windows-msvc/usr/include)
+
+cp -r contents/vc/tools/msvc/14.16.27023/include/* /usr/x86_64-pc-windows-msvc/usr/include
+
+# lowercase #include "" and #include <>
+find /usr/x86_64-pc-windows-msvc/usr/include -type f -exec sed -i -E 's/(#include <[^<>]*?[A-Z][^<>]*?>)|(#include "[^"]*?[A-Z][^"]*?")/\L\1\2/' "{}" ';'
+
+# x86 intrinsics
+# original dir: MSVC/
+
+# '_mm_movemask_epi8' defined in emmintrin.h
+# '__v4sf' defined in xmmintrin.h
+# '__v2si' defined in mmintrin.h
+# '__m128d' redefined in immintrin.h
+# '__m128i' redefined in intrin.h
+# '_mm_comlt_epu8' defined in ammintrin.h
+
+(cd /usr/lib/llvm19/lib/clang/19/include && cp emmintrin.h xmmintrin.h mmintrin.h immintrin.h intrin.h ammintrin.h -t /usr/x86_64-pc-windows-msvc/usr/include)
+
+# .lib
+(cd 'program files/windows kits/10/lib/10.0.26100.0/um/x64' && cp advapi32.lib bcrypt.lib kernel32.lib ntdll.lib user32.lib uuid.lib ws2_32.lib userenv.lib cfgmgr32.lib dbghelp.lib fwpuclnt.lib -t /usr/x86_64-pc-windows-msvc/usr/lib)
+
+(cd 'contents/vc/tools/msvc/14.16.27023/lib/x64' && cp libcmt.lib libvcruntime.lib -t /usr/x86_64-pc-windows-msvc/usr/lib)
+
+cp 'program files/windows kits/10/lib/10.0.26100.0/ucrt/x64/libucrt.lib' /usr/x86_64-pc-windows-msvc/usr/lib

docs/mkdocs.yml

@@ -55,6 +55,9 @@ plugins:
             show_signature_annotations: true
             show_root_heading: true
             members_order: source
+            docstring_section_style: list
+            signature_crossrefs: true
+            separate_signature: true
           import:
             # for cross references
             - https://arrow.apache.org/docs/objects.inv

docs/src/guides/scalar_index.md

@@ -1,23 +1,35 @@
-# Building Scalar Index
+# Building a Scalar Index
 
-Similar to many SQL databases, LanceDB supports several types of Scalar indices to accelerate search
+Scalar indices organize data by scalar attributes (e.g. numbers, categorical values), enabling fast filtering of vector data. In vector databases, scalar indices accelerate the retrieval of scalar data associated with vectors, thus enhancing the query performance when searching for vectors that meet certain scalar criteria. 
+
+Similar to many SQL databases, LanceDB supports several types of scalar indices to accelerate search
 over scalar columns.
 
-- `BTREE`: The most common type is BTREE. This index is inspired by the btree data structure
+- `BTREE`: The most common type is BTREE. The index stores a copy of the
-  although only the first few layers of the btree are cached in memory.
+  column in sorted order. This sorted copy allows a binary search to be used to
-  It will perform well on columns with a large number of unique values and few rows per value.
+  satisfy queries.
-- `BITMAP`: this index stores a bitmap for each unique value in the column.
+- `BITMAP`: this index stores a bitmap for each unique value in the column. It 
-  This index is useful for columns with a finite number of unique values and many rows per value.
+  uses a series of bits to indicate whether a value is present in a row of a table
-  For example, columns that represent "categories", "labels", or "tags"
+- `LABEL_LIST`: a special index that can be used on `List<T>` columns to
-- `LABEL_LIST`: a special index that is used to index list columns whose values have a finite set of possibilities.
+  support queries with `array_contains_all` and `array_contains_any`
+  using an underlying bitmap index.
   For example, a column that contains lists of tags (e.g. `["tag1", "tag2", "tag3"]`) can be indexed with a `LABEL_LIST` index.
 
+!!! tips "How to choose the right scalar index type"
+
+    `BTREE`: This index is good for scalar columns with mostly distinct values and does best when the query is highly selective.
+    
+    `BITMAP`: This index works best for low-cardinality numeric or string columns, where the number of unique values is small (i.e., less than a few thousands).
+    
+    `LABEL_LIST`: This index should be used for columns containing list-type data.
+
 | Data Type                                                       | Filter                                    | Index Type   |
 | --------------------------------------------------------------- | ----------------------------------------- | ------------ |
 | Numeric, String, Temporal                                       | `<`, `=`, `>`, `in`, `between`, `is null` | `BTREE`      |
 | Boolean, numbers or strings with fewer than 1,000 unique values | `<`, `=`, `>`, `in`, `between`, `is null` | `BITMAP`     |
 | List of low cardinality of numbers or strings                   | `array_has_any`, `array_has_all`          | `LABEL_LIST` |
 
+### Create a scalar index
 === "Python"
 
     ```python
@@ -46,7 +58,7 @@ over scalar columns.
         await tlb.create_index("publisher", { config: lancedb.Index.bitmap() })
         ```
 
-For example, the following scan will be faster if the column `my_col` has a scalar index:
+The following scan will be faster if the column `book_id` has a scalar index:
 
 === "Python"
 
@@ -106,3 +118,30 @@ Scalar indices can also speed up scans containing a vector search or full text s
           .limit(10)
           .toArray();
         ```
+### Update a scalar index
+Updating the table data (adding, deleting, or modifying records) requires that you also update the scalar index. This can be done by calling `optimize`, which will trigger an update to the existing scalar index.
+=== "Python"
+
+    ```python
+    table.add([{"vector": [7, 8], "book_id": 4}])
+    table.optimize()
+    ```
+
+=== "TypeScript"
+
+    ```typescript
+    await tbl.add([{ vector: [7, 8], book_id: 4 }]);
+    await tbl.optimize();
+    ```
+
+=== "Rust"
+
+    ```rust
+    let more_data: Box<dyn RecordBatchReader + Send> = create_some_records()?;
+    tbl.add(more_data).execute().await?;
+    tbl.optimize(OptimizeAction::All).execute().await?;
+    ```
+
+!!! note
+
+    New data added after creating the scalar index will still appear in search results if optimize is not used, but with increased latency due to a flat search on the unindexed portion. LanceDB Cloud automates the optimize process, minimizing the impact on search speed.

docs/src/python/python.md

@@ -1,6 +1,16 @@
 # Python API Reference
 
-This section contains the API reference for the OSS Python API.
+This section contains the API reference for the Python API. There is a
+synchronous and an asynchronous API client.
+
+The general flow of using the API is:
+
+1. Use [lancedb.connect][] or [lancedb.connect_async][] to connect to a database.
+2. Use the returned [lancedb.DBConnection][] or [lancedb.AsyncConnection][] to
+   create or open tables.
+3. Use the returned [lancedb.table.Table][] or [lancedb.AsyncTable][] to query
+   or modify tables.
+
 
 ## Installation
 

docs/src/sql.md

@@ -7,6 +7,10 @@ performed on the top-k results returned by the vector search. However, pre-filte
 option that performs the filter prior to vector search. This can be useful to narrow down on
 the search space on a very large dataset to reduce query latency.
 
+Note that both pre-filtering and post-filtering can yield false positives. For pre-filtering, if the filter is too selective, it might eliminate relevant items that the vector search would have otherwise identified as a good match. In this case, increasing `nprobes` parameter will help reduce such false positives. It is recommended to set `use_index=false` if you know that the filter is highly selective.
+
+Similarly, a highly selective post-filter can lead to false positives. Increasing both `nprobes` and `refine_factor` can mitigate this issue. When deciding between pre-filtering and post-filtering, pre-filtering is generally the safer choice if you're uncertain.
+
 <!-- Setup Code
 ```python
 import lancedb
@@ -57,6 +61,9 @@ const tbl = await db.createTable('myVectors', data)
         ```ts
         --8<-- "docs/src/sql_legacy.ts:search"
         ```
+!!! note
+
+    Creating a [scalar index](guides/scalar_index.md) accelerates filtering
 
 ## SQL filters
 

java/core/pom.xml

@@ -8,7 +8,7 @@
     <parent>
         <groupId>com.lancedb</groupId>
         <artifactId>lancedb-parent</artifactId>
-        <version>0.13.0-final.0</version>
+        <version>0.14.0-beta.1</version>
         <relativePath>../pom.xml</relativePath>
     </parent>
 

java/pom.xml

@@ -6,7 +6,7 @@
 
     <groupId>com.lancedb</groupId>
     <artifactId>lancedb-parent</artifactId>
-    <version>0.13.0-final.0</version>
+    <version>0.14.0-beta.1</version>
     <packaging>pom</packaging>
 
     <name>LanceDB Parent</name>

node/package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "vectordb",
-  "version": "0.13.0",
+  "version": "0.14.0-beta.1",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "vectordb",
-      "version": "0.13.0",
+      "version": "0.14.0-beta.1",
       "cpu": [
         "x64",
         "arm64"
@@ -52,12 +52,14 @@
         "uuid": "^9.0.0"
       },
       "optionalDependencies": {
-        "@lancedb/vectordb-darwin-arm64": "0.13.0",
+        "@lancedb/vectordb-darwin-arm64": "0.14.0-beta.1",
-        "@lancedb/vectordb-darwin-x64": "0.13.0",
+        "@lancedb/vectordb-darwin-x64": "0.14.0-beta.1",
-        "@lancedb/vectordb-linux-arm64-gnu": "0.13.0",
+        "@lancedb/vectordb-linux-arm64-gnu": "0.14.0-beta.1",
-        "@lancedb/vectordb-linux-x64-gnu": "0.13.0",
+        "@lancedb/vectordb-linux-arm64-musl": "0.14.0-beta.1",
-        "@lancedb/vectordb-win32-arm64-msvc": "0.13.0",
+        "@lancedb/vectordb-linux-x64-gnu": "0.14.0-beta.1",
-        "@lancedb/vectordb-win32-x64-msvc": "0.13.0"
+        "@lancedb/vectordb-linux-x64-musl": "0.14.0-beta.1",
+        "@lancedb/vectordb-win32-arm64-msvc": "0.14.0-beta.1",
+        "@lancedb/vectordb-win32-x64-msvc": "0.14.0-beta.1"
       },
       "peerDependencies": {
         "@apache-arrow/ts": "^14.0.2",
@@ -327,66 +329,6 @@
         "@jridgewell/sourcemap-codec": "^1.4.10"
       }
     },
-    "node_modules/@lancedb/vectordb-darwin-arm64": {
-      "version": "0.13.0",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-arm64/-/vectordb-darwin-arm64-0.13.0.tgz",
-      "integrity": "sha512-8hdcjkRmgrdQYf1jN+DyZae40LIv8UUfnWy70Uid5qy63sSvRW/+MvIdqIPFr9QlLUXmpyyQuX0y3bZhUR99cQ==",
-      "cpu": [
-        "arm64"
-      ],
-      "optional": true,
-      "os": [
-        "darwin"
-      ]
-    },
-    "node_modules/@lancedb/vectordb-darwin-x64": {
-      "version": "0.13.0",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-x64/-/vectordb-darwin-x64-0.13.0.tgz",
-      "integrity": "sha512-fWzAY4l5SQtNfMYh80v+M66ugZHhdxbkpk5mNEv6Zsug3DL6kRj3Uv31/i0wgzY6F5G3LUlbjZerN+eTnDLwOw==",
-      "cpu": [
-        "x64"
-      ],
-      "optional": true,
-      "os": [
-        "darwin"
-      ]
-    },
-    "node_modules/@lancedb/vectordb-linux-arm64-gnu": {
-      "version": "0.13.0",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-arm64-gnu/-/vectordb-linux-arm64-gnu-0.13.0.tgz",
-      "integrity": "sha512-ltwAT9baOSuR5YiGykQXPC8/HGYF13vpI47qxhP9yfgiz9pA8EUn8p8YrBRzq7J4DIZ4b8JSVDXQnMIqEtB4Kg==",
-      "cpu": [
-        "arm64"
-      ],
-      "optional": true,
-      "os": [
-        "linux"
-      ]
-    },
-    "node_modules/@lancedb/vectordb-linux-x64-gnu": {
-      "version": "0.13.0",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-x64-gnu/-/vectordb-linux-x64-gnu-0.13.0.tgz",
-      "integrity": "sha512-MiT/RBlMPGGRh7BX+MXwRuNiiUnKmuDcHH8nm88IH28T7TQxXIbA9w6UpSg5m9f3DgKQI2K8oLi29oKIB8ZwDQ==",
-      "cpu": [
-        "x64"
-      ],
-      "optional": true,
-      "os": [
-        "linux"
-      ]
-    },
-    "node_modules/@lancedb/vectordb-win32-x64-msvc": {
-      "version": "0.13.0",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-win32-x64-msvc/-/vectordb-win32-x64-msvc-0.13.0.tgz",
-      "integrity": "sha512-SovP/hwWYLJIy65DKbVuXlBPTb/nwvVpTO6dh9zRch+L5ek6JmVAkwsfeTS2p5bMa8VPujsCXYUAVuCDEJU8wg==",
-      "cpu": [
-        "x64"
-      ],
-      "optional": true,
-      "os": [
-        "win32"
-      ]
-    },
     "node_modules/@neon-rs/cli": {
       "version": "0.0.160",
       "resolved": "https://registry.npmjs.org/@neon-rs/cli/-/cli-0.0.160.tgz",

node/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vectordb",
-  "version": "0.13.0",
+  "version": "0.14.0-beta.1",
   "description": " Serverless, low-latency vector database for AI applications",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",
@@ -84,18 +84,20 @@
       "aarch64-apple-darwin": "@lancedb/vectordb-darwin-arm64",
       "x86_64-unknown-linux-gnu": "@lancedb/vectordb-linux-x64-gnu",
       "aarch64-unknown-linux-gnu": "@lancedb/vectordb-linux-arm64-gnu",
+      "x86_64-unknown-linux-musl": "@lancedb/vectordb-linux-x64-musl",
+      "aarch64-unknown-linux-musl": "@lancedb/vectordb-linux-arm64-musl",
       "x86_64-pc-windows-msvc": "@lancedb/vectordb-win32-x64-msvc",
       "aarch64-pc-windows-msvc": "@lancedb/vectordb-win32-arm64-msvc"
     }
   },
   "optionalDependencies": {
-    "@lancedb/vectordb-darwin-x64": "0.13.0",
+    "@lancedb/vectordb-darwin-x64": "0.14.0-beta.1",
-    "@lancedb/vectordb-darwin-arm64": "0.13.0",
+    "@lancedb/vectordb-darwin-arm64": "0.14.0-beta.1",
-    "@lancedb/vectordb-linux-x64-gnu": "0.13.0",
+    "@lancedb/vectordb-linux-x64-gnu": "0.14.0-beta.1",
-    "@lancedb/vectordb-linux-arm64-gnu": "0.13.0",
+    "@lancedb/vectordb-linux-arm64-gnu": "0.14.0-beta.1",
-    "@lancedb/vectordb-linux-x64-musl": "0.13.0",
+    "@lancedb/vectordb-linux-x64-musl": "0.14.0-beta.1",
-    "@lancedb/vectordb-linux-arm64-musl": "0.13.0",
+    "@lancedb/vectordb-linux-arm64-musl": "0.14.0-beta.1",
-    "@lancedb/vectordb-win32-x64-msvc": "0.13.0",
+    "@lancedb/vectordb-win32-x64-msvc": "0.14.0-beta.1",
-    "@lancedb/vectordb-win32-arm64-msvc": "0.13.0"
+    "@lancedb/vectordb-win32-arm64-msvc": "0.14.0-beta.1"
   }
 }

nodejs/Cargo.toml

@@ -1,7 +1,7 @@
 [package]
 name = "lancedb-nodejs"
 edition.workspace = true
-version = "0.13.0"
+version = "0.14.0-beta.1"
 license.workspace = true
 description.workspace = true
 repository.workspace = true

nodejs/__test__/connection.test.ts

@@ -110,7 +110,10 @@ describe("given a connection", () => {
     let table = await db.createTable("test", data, { useLegacyFormat: true });
 
     const isV2 = async (table: Table) => {
-      const data = await table.query().toArrow({ maxBatchLength: 100000 });
+      const data = await table
+        .query()
+        .limit(10000)
+        .toArrow({ maxBatchLength: 100000 });
       console.log(data.batches.length);
       return data.batches.length < 5;
     };

nodejs/__test__/table.test.ts

@@ -585,11 +585,11 @@ describe("When creating an index", () => {
     expect(fs.readdirSync(indexDir)).toHaveLength(1);
 
     for await (const r of tbl.query().where("id > 1").select(["id"])) {
-      expect(r.numRows).toBe(298);
+      expect(r.numRows).toBe(10);
     }
     // should also work with 'filter' alias
     for await (const r of tbl.query().filter("id > 1").select(["id"])) {
-      expect(r.numRows).toBe(298);
+      expect(r.numRows).toBe(10);
     }
   });
 

nodejs/npm/darwin-arm64/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-darwin-arm64",
-	"version": "0.13.0",
+	"version": "0.14.0-beta.1",
 	"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": "0.13.0",
+	"version": "0.14.0-beta.1",
 	"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": "0.13.0",
+	"version": "0.14.0-beta.1",
 	"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": "0.13.0",
+	"version": "0.14.0-beta.1",
 	"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": "0.13.0",
+	"version": "0.14.0-beta.1",
 	"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": "0.13.0",
+	"version": "0.14.0-beta.1",
 	"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": "0.13.0",
+  "version": "0.14.0-beta.1",
   "os": [
     "win32"
   ],

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-win32-x64-msvc",
-	"version": "0.13.0",
+	"version": "0.14.0-beta.1",
 	"os": ["win32"],
 	"cpu": ["x64"],
 	"main": "lancedb.win32-x64-msvc.node",

nodejs/package.json

@@ -10,7 +10,7 @@
     "vector database",
     "ann"
   ],
-  "version": "0.13.0",
+  "version": "0.14.0-beta.1",
   "main": "dist/index.js",
   "exports": {
     ".": "./dist/index.js",

python/.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.16.1-beta.0"
+current_version = "0.17.0-beta.3"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

python/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb-python"
-version = "0.16.1-beta.0"
+version = "0.17.0-beta.3"
 edition.workspace = true
 description = "Python bindings for LanceDB"
 license.workspace = true
@@ -17,11 +17,17 @@ crate-type = ["cdylib"]
 arrow = { version = "52.1", features = ["pyarrow"] }
 lancedb = { path = "../rust/lancedb", default-features = false }
 env_logger.workspace = true
-pyo3 = { version = "0.21", features = ["extension-module", "abi3-py38", "gil-refs"] }
+pyo3 = { version = "0.21", features = [
+    "extension-module",
+    "abi3-py39",
+    "gil-refs"
+] }
 # Using this fork for now: https://github.com/awestlake87/pyo3-asyncio/issues/119
 # pyo3-asyncio = { version = "0.20", features = ["attributes", "tokio-runtime"] }
-pyo3-asyncio-0-21 = { version = "0.21.0", features = ["attributes", "tokio-runtime"] }
+pyo3-asyncio-0-21 = { version = "0.21.0", features = [
-
+    "attributes",
+    "tokio-runtime"
+] }
 pin-project = "1.1.5"
 futures.workspace = true
 tokio = { version = "1.36.0", features = ["sync"] }
@@ -29,14 +35,13 @@ tokio = { version = "1.36.0", features = ["sync"] }
 [build-dependencies]
 pyo3-build-config = { version = "0.20.3", features = [
     "extension-module",
-    "abi3-py38",
+    "abi3-py39",
 ] }
 
 [features]
 default = ["default-tls", "remote"]
 fp16kernels = ["lancedb/fp16kernels"]
 remote = ["lancedb/remote"]
-
 # TLS
 default-tls = ["lancedb/default-tls"]
 native-tls = ["lancedb/native-tls"]

python/pyproject.toml

@@ -3,8 +3,7 @@ name = "lancedb"
 # version in Cargo.toml
 dependencies = [
     "deprecation",
-    "nest-asyncio~=1.0",
+    "pylance==0.20.0b3",
-    "pylance==0.20.0b2",
     "tqdm>=4.27.0",
     "pydantic>=1.10",
     "packaging",
@@ -31,7 +30,6 @@ classifiers = [
     "Programming Language :: Python",
     "Programming Language :: Python :: 3",
     "Programming Language :: Python :: 3 :: Only",
-    "Programming Language :: Python :: 3.8",
     "Programming Language :: Python :: 3.9",
     "Programming Language :: Python :: 3.10",
     "Programming Language :: Python :: 3.11",

python/python/lancedb/embeddings/openai.py

@@ -83,25 +83,33 @@ class OpenAIEmbeddings(TextEmbeddingFunction):
         """
         openai = attempt_import_or_raise("openai")
 
+        valid_texts = []
+        valid_indices = []
+        for idx, text in enumerate(texts):
+            if text:
+                valid_texts.append(text)
+                valid_indices.append(idx)
+
         # TODO retry, rate limit, token limit
         try:
-            if self.name == "text-embedding-ada-002":
+            kwargs = {
-                rs = self._openai_client.embeddings.create(input=texts, model=self.name)
+                "input": valid_texts,
-            else:
+                "model": self.name,
-                kwargs = {
+            }
-                    "input": texts,
+            if self.name != "text-embedding-ada-002":
-                    "model": self.name,
+                kwargs["dimensions"] = self.dim
-                }
+
-                if self.dim:
+            rs = self._openai_client.embeddings.create(**kwargs)
-                    kwargs["dimensions"] = self.dim
+            valid_embeddings = {
-                rs = self._openai_client.embeddings.create(**kwargs)
+                idx: v.embedding for v, idx in zip(rs.data, valid_indices)
+            }
         except openai.BadRequestError:
             logging.exception("Bad request: %s", texts)
             return [None] * len(texts)
         except Exception:
             logging.exception("OpenAI embeddings error")
             raise
-        return [v.embedding for v in rs.data]
+        return [valid_embeddings.get(idx, None) for idx in range(len(texts))]
 
     @cached_property
     def _openai_client(self):

python/python/lancedb/embeddings/voyageai.py

@@ -12,18 +12,22 @@
 #  limitations under the License.
 
 import os
-from typing import ClassVar, List, Union
+from typing import ClassVar, TYPE_CHECKING, List, Union
 
 import numpy as np
+import pyarrow as pa
 
 from ..util import attempt_import_or_raise
-from .base import TextEmbeddingFunction
+from .base import EmbeddingFunction
 from .registry import register
-from .utils import api_key_not_found_help, TEXT
+from .utils import api_key_not_found_help, IMAGES
+
+if TYPE_CHECKING:
+    import PIL
 
 
 @register("voyageai")
-class VoyageAIEmbeddingFunction(TextEmbeddingFunction):
+class VoyageAIEmbeddingFunction(EmbeddingFunction):
     """
     An embedding function that uses the VoyageAI API
 
@@ -36,6 +40,7 @@ class VoyageAIEmbeddingFunction(TextEmbeddingFunction):
 
             * voyage-3
             * voyage-3-lite
+            * voyage-multimodal-3
             * voyage-finance-2
             * voyage-multilingual-2
             * voyage-law-2
@@ -54,7 +59,7 @@ class VoyageAIEmbeddingFunction(TextEmbeddingFunction):
         .create(name="voyage-3")
 
     class TextModel(LanceModel):
-        text: str = voyageai.SourceField()
+        data: str = voyageai.SourceField()
         vector: Vector(voyageai.ndims()) =  voyageai.VectorField()
 
     data = [ { "text": "hello world" },
@@ -77,6 +82,7 @@ class VoyageAIEmbeddingFunction(TextEmbeddingFunction):
             return 1536
         elif self.name in [
             "voyage-3",
+            "voyage-multimodal-3",
             "voyage-finance-2",
             "voyage-multilingual-2",
             "voyage-law-2",
@@ -85,19 +91,19 @@ class VoyageAIEmbeddingFunction(TextEmbeddingFunction):
         else:
             raise ValueError(f"Model {self.name} not supported")
 
-    def compute_query_embeddings(self, query: str, *args, **kwargs) -> List[np.array]:
+    def sanitize_input(self, images: IMAGES) -> Union[List[bytes], np.ndarray]:
-        return self.compute_source_embeddings(query, input_type="query")
+        """
+        Sanitize the input to the embedding function.
+        """
+        if isinstance(images, (str, bytes)):
+            images = [images]
+        elif isinstance(images, pa.Array):
+            images = images.to_pylist()
+        elif isinstance(images, pa.ChunkedArray):
+            images = images.combine_chunks().to_pylist()
+        return images
 
-    def compute_source_embeddings(self, texts: TEXT, *args, **kwargs) -> List[np.array]:
+    def generate_text_embeddings(self, text: str, **kwargs) -> np.ndarray:
-        texts = self.sanitize_input(texts)
-        input_type = (
-            kwargs.get("input_type") or "document"
-        )  # assume source input type if not passed by `compute_query_embeddings`
-        return self.generate_embeddings(texts, input_type=input_type)
-
-    def generate_embeddings(
-        self, texts: Union[List[str], np.ndarray], *args, **kwargs
-    ) -> List[np.array]:
         """
         Get the embeddings for the given texts
 
@@ -109,15 +115,55 @@ class VoyageAIEmbeddingFunction(TextEmbeddingFunction):
 
         truncation: Optional[bool]
         """
-        VoyageAIEmbeddingFunction._init_client()
+        if self.name in ["voyage-multimodal-3"]:
-        rs = VoyageAIEmbeddingFunction.client.embed(
+            rs = VoyageAIEmbeddingFunction._get_client().multimodal_embed(
-            texts=texts, model=self.name, **kwargs
+                inputs=[[text]], model=self.name, **kwargs
-        )
+            )
+        else:
+            rs = VoyageAIEmbeddingFunction._get_client().embed(
+                texts=[text], model=self.name, **kwargs
+            )
 
-        return [emb for emb in rs.embeddings]
+        return rs.embeddings[0]
+
+    def generate_image_embedding(
+        self, image: "PIL.Image.Image", **kwargs
+    ) -> np.ndarray:
+        rs = VoyageAIEmbeddingFunction._get_client().multimodal_embed(
+            inputs=[[image]], model=self.name, **kwargs
+        )
+        return rs.embeddings[0]
+
+    def compute_query_embeddings(
+        self, query: Union[str, "PIL.Image.Image"], *args, **kwargs
+    ) -> List[np.ndarray]:
+        """
+        Compute the embeddings for a given user query
+
+        Parameters
+        ----------
+        query : Union[str, PIL.Image.Image]
+            The query to embed. A query can be either text or an image.
+        """
+        if isinstance(query, str):
+            return [self.generate_text_embeddings(query, input_type="query")]
+        else:
+            PIL = attempt_import_or_raise("PIL", "pillow")
+            if isinstance(query, PIL.Image.Image):
+                return [self.generate_image_embedding(query, input_type="query")]
+            else:
+                raise TypeError("Only text PIL images supported as query")
+
+    def compute_source_embeddings(
+        self, images: IMAGES, *args, **kwargs
+    ) -> List[np.array]:
+        images = self.sanitize_input(images)
+        return [
+            self.generate_image_embedding(img, input_type="document") for img in images
+        ]
 
     @staticmethod
-    def _init_client():
+    def _get_client():
         if VoyageAIEmbeddingFunction.client is None:
             voyageai = attempt_import_or_raise("voyageai")
             if os.environ.get("VOYAGE_API_KEY") is None:
@@ -125,3 +171,4 @@ class VoyageAIEmbeddingFunction(TextEmbeddingFunction):
             VoyageAIEmbeddingFunction.client = voyageai.Client(
                 os.environ["VOYAGE_API_KEY"]
             )
+        return VoyageAIEmbeddingFunction.client

python/python/lancedb/integrations/pyarrow.py

@@ -0,0 +1,248 @@
+import logging
+from typing import Any, List, Optional, Tuple, Union, Literal
+
+import pyarrow as pa
+
+from ..table import Table
+
+Filter = Union[str, pa.compute.Expression]
+Keys = Union[str, List[str]]
+JoinType = Literal[
+    "left semi",
+    "right semi",
+    "left anti",
+    "right anti",
+    "inner",
+    "left outer",
+    "right outer",
+    "full outer",
+]
+
+
+class PyarrowScannerAdapter(pa.dataset.Scanner):
+    def __init__(
+        self,
+        table: Table,
+        columns: Optional[List[str]] = None,
+        filter: Optional[Filter] = None,
+        batch_size: Optional[int] = None,
+        batch_readahead: Optional[int] = None,
+        fragment_readahead: Optional[int] = None,
+        fragment_scan_options: Optional[Any] = None,
+        use_threads: bool = True,
+        memory_pool: Optional[Any] = None,
+    ):
+        self.table = table
+        self.columns = columns
+        self.filter = filter
+        self.batch_size = batch_size
+        if batch_readahead is not None:
+            logging.debug("ignoring batch_readahead which has no lance equivalent")
+        if fragment_readahead is not None:
+            logging.debug("ignoring fragment_readahead which has no lance equivalent")
+        if fragment_scan_options is not None:
+            raise NotImplementedError("fragment_scan_options not supported")
+        if use_threads is False:
+            raise NotImplementedError("use_threads=False not supported")
+        if memory_pool is not None:
+            raise NotImplementedError("memory_pool not supported")
+
+    def count_rows(self):
+        return self.table.count_rows(self.filter)
+
+    def from_batches(self, **kwargs):
+        raise NotImplementedError
+
+    def from_dataset(self, **kwargs):
+        raise NotImplementedError
+
+    def from_fragment(self, **kwargs):
+        raise NotImplementedError
+
+    def head(self, num_rows: int):
+        return self.to_reader(limit=num_rows).read_all()
+
+    @property
+    def projected_schema(self):
+        return self.head(1).schema
+
+    def scan_batches(self):
+        return self.to_reader()
+
+    def take(self, indices: List[int]):
+        raise NotImplementedError
+
+    def to_batches(self):
+        return self.to_reader()
+
+    def to_table(self):
+        return self.to_reader().read_all()
+
+    def to_reader(self, *, limit: Optional[int] = None):
+        query = self.table.search()
+        # Disable the builtin limit
+        if limit is None:
+            num_rows = self.count_rows()
+            query.limit(num_rows)
+        elif limit <= 0:
+            raise ValueError("limit must be positive")
+        else:
+            query.limit(limit)
+        if self.columns is not None:
+            query = query.select(self.columns)
+        if self.filter is not None:
+            query = query.where(self.filter, prefilter=True)
+        return query.to_batches(batch_size=self.batch_size)
+
+
+class PyarrowDatasetAdapter(pa.dataset.Dataset):
+    def __init__(self, table: Table):
+        self.table = table
+
+    def count_rows(self, filter: Optional[Filter] = None):
+        return self.table.count_rows(filter)
+
+    def get_fragments(self, filter: Optional[Filter] = None):
+        raise NotImplementedError
+
+    def head(
+        self,
+        num_rows: int,
+        columns: Optional[List[str]] = None,
+        filter: Optional[Filter] = None,
+        batch_size: Optional[int] = None,
+        batch_readahead: Optional[int] = None,
+        fragment_readahead: Optional[int] = None,
+        fragment_scan_options: Optional[Any] = None,
+        use_threads: bool = True,
+        memory_pool: Optional[Any] = None,
+    ):
+        return self.scanner(
+            columns,
+            filter,
+            batch_size,
+            batch_readahead,
+            fragment_readahead,
+            fragment_scan_options,
+            use_threads,
+            memory_pool,
+        ).head(num_rows)
+
+    def join(
+        self,
+        right_dataset: Any,
+        keys: Keys,
+        right_keys: Optional[Keys] = None,
+        join_type: Optional[JoinType] = None,
+        left_suffix: Optional[str] = None,
+        right_suffix: Optional[str] = None,
+        coalesce_keys: bool = True,
+        use_threads: bool = True,
+    ):
+        raise NotImplementedError
+
+    def join_asof(
+        self,
+        right_dataset: Any,
+        on: str,
+        by: Keys,
+        tolerance: int,
+        right_on: Optional[str] = None,
+        right_by: Optional[Keys] = None,
+    ):
+        raise NotImplementedError
+
+    @property
+    def partition_expression(self):
+        raise NotImplementedError
+
+    def replace_schema(self, schema: pa.Schema):
+        raise NotImplementedError
+
+    def scanner(
+        self,
+        columns: Optional[List[str]] = None,
+        filter: Optional[Filter] = None,
+        batch_size: Optional[int] = None,
+        batch_readahead: Optional[int] = None,
+        fragment_readahead: Optional[int] = None,
+        fragment_scan_options: Optional[Any] = None,
+        use_threads: bool = True,
+        memory_pool: Optional[Any] = None,
+    ):
+        return PyarrowScannerAdapter(
+            self.table,
+            columns,
+            filter,
+            batch_size,
+            batch_readahead,
+            fragment_readahead,
+            fragment_scan_options,
+            use_threads,
+            memory_pool,
+        )
+
+    @property
+    def schema(self):
+        return self.table.schema
+
+    def sort_by(self, sorting: Union[str, List[Tuple[str, bool]]]):
+        raise NotImplementedError
+
+    def take(
+        self,
+        indices: List[int],
+        columns: Optional[List[str]] = None,
+        filter: Optional[Filter] = None,
+        batch_size: Optional[int] = None,
+        batch_readahead: Optional[int] = None,
+        fragment_readahead: Optional[int] = None,
+        fragment_scan_options: Optional[Any] = None,
+        use_threads: bool = True,
+        memory_pool: Optional[Any] = None,
+    ):
+        raise NotImplementedError
+
+    def to_batches(
+        self,
+        columns: Optional[List[str]] = None,
+        filter: Optional[Filter] = None,
+        batch_size: Optional[int] = None,
+        batch_readahead: Optional[int] = None,
+        fragment_readahead: Optional[int] = None,
+        fragment_scan_options: Optional[Any] = None,
+        use_threads: bool = True,
+        memory_pool: Optional[Any] = None,
+    ):
+        return self.scanner(
+            columns,
+            filter,
+            batch_size,
+            batch_readahead,
+            fragment_readahead,
+            fragment_scan_options,
+            use_threads,
+            memory_pool,
+        ).to_batches()
+
+    def to_table(
+        self,
+        columns: Optional[List[str]] = None,
+        filter: Optional[Filter] = None,
+        batch_size: Optional[int] = None,
+        batch_readahead: Optional[int] = None,
+        fragment_readahead: Optional[int] = None,
+        fragment_scan_options: Optional[Any] = None,
+        use_threads: bool = True,
+        memory_pool: Optional[Any] = None,
+    ):
+        return self.scanner(
+            columns,
+            filter,
+            batch_size,
+            batch_readahead,
+            fragment_readahead,
+            fragment_scan_options,
+            use_threads,
+            memory_pool,
+        ).to_table()

python/python/lancedb/pydantic.py

@@ -1,15 +1,5 @@
-#  Copyright 2023 LanceDB Developers
+# SPDX-License-Identifier: Apache-2.0
-#
+# SPDX-FileCopyrightText: Copyright The LanceDB Authors
-#  Licensed under the Apache License, Version 2.0 (the "License");
-#  you may not use this file except in compliance with the License.
-#  You may obtain a copy of the License at
-#      http://www.apache.org/licenses/LICENSE-2.0
-#
-#  Unless required by applicable law or agreed to in writing, software
-#  distributed under the License is distributed on an "AS IS" BASIS,
-#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-#  See the License for the specific language governing permissions and
-#  limitations under the License.
 
 """Pydantic (v1 / v2) adapter for LanceDB"""
 
@@ -30,6 +20,7 @@ from typing import (
     Type,
     Union,
     _GenericAlias,
+    GenericAlias,
 )
 
 import numpy as np
@@ -75,7 +66,7 @@ def vector(dim: int, value_type: pa.DataType = pa.float32()):
 
 
 def Vector(
-    dim: int, value_type: pa.DataType = pa.float32()
+    dim: int, value_type: pa.DataType = pa.float32(), nullable: bool = True
 ) -> Type[FixedSizeListMixin]:
     """Pydantic Vector Type.
 
@@ -88,6 +79,8 @@ def Vector(
         The dimension of the vector.
     value_type : pyarrow.DataType, optional
         The value type of the vector, by default pa.float32()
+    nullable : bool, optional
+        Whether the vector is nullable, by default it is True.
 
     Examples
     --------
@@ -103,7 +96,7 @@ def Vector(
     >>> assert schema == pa.schema([
     ...     pa.field("id", pa.int64(), False),
     ...     pa.field("url", pa.utf8(), False),
-    ...     pa.field("embeddings", pa.list_(pa.float32(), 768), False)
+    ...     pa.field("embeddings", pa.list_(pa.float32(), 768))
     ... ])
     """
 
@@ -112,6 +105,10 @@ def Vector(
         def __repr__(self):
             return f"FixedSizeList(dim={dim})"
 
+        @staticmethod
+        def nullable() -> bool:
+            return nullable
+
         @staticmethod
         def dim() -> int:
             return dim
@@ -205,9 +202,7 @@ else:
 def _pydantic_to_arrow_type(field: FieldInfo) -> pa.DataType:
     """Convert a Pydantic FieldInfo to Arrow DataType"""
 
-    if isinstance(field.annotation, _GenericAlias) or (
+    if isinstance(field.annotation, (_GenericAlias, GenericAlias)):
-        sys.version_info > (3, 9) and isinstance(field.annotation, types.GenericAlias)
-    ):
         origin = field.annotation.__origin__
         args = field.annotation.__args__
         if origin is list:
@@ -235,7 +230,7 @@ def _pydantic_to_arrow_type(field: FieldInfo) -> pa.DataType:
 
 def is_nullable(field: FieldInfo) -> bool:
     """Check if a Pydantic FieldInfo is nullable."""
-    if isinstance(field.annotation, _GenericAlias):
+    if isinstance(field.annotation, (_GenericAlias, GenericAlias)):
         origin = field.annotation.__origin__
         args = field.annotation.__args__
         if origin == Union:
@@ -246,6 +241,10 @@ def is_nullable(field: FieldInfo) -> bool:
         for typ in args:
             if typ is type(None):
                 return True
+    elif inspect.isclass(field.annotation) and issubclass(
+        field.annotation, FixedSizeListMixin
+    ):
+        return field.annotation.nullable()
     return False
 
 

python/python/lancedb/query.py

@@ -325,6 +325,14 @@ class LanceQueryBuilder(ABC):
         """
         raise NotImplementedError
 
+    @abstractmethod
+    def to_batches(self, /, batch_size: Optional[int] = None) -> pa.Table:
+        """
+        Execute the query and return the results as a pyarrow
+        [RecordBatchReader](https://arrow.apache.org/docs/python/generated/pyarrow.RecordBatchReader.html)
+        """
+        raise NotImplementedError
+
     def to_list(self) -> List[dict]:
         """
         Execute the query and return the results as a list of dictionaries.
@@ -869,6 +877,9 @@ class LanceFtsQueryBuilder(LanceQueryBuilder):
             check_reranker_result(results)
         return results
 
+    def to_batches(self, /, batch_size: Optional[int] = None):
+        raise NotImplementedError("to_batches on an FTS query")
+
     def tantivy_to_arrow(self) -> pa.Table:
         try:
             import tantivy
@@ -971,6 +982,9 @@ class LanceFtsQueryBuilder(LanceQueryBuilder):
 
 class LanceEmptyQueryBuilder(LanceQueryBuilder):
     def to_arrow(self) -> pa.Table:
+        return self.to_batches().read_all()
+
+    def to_batches(self, /, batch_size: Optional[int] = None) -> pa.RecordBatchReader:
         query = Query(
             columns=self._columns,
             filter=self._where,
@@ -980,7 +994,7 @@ class LanceEmptyQueryBuilder(LanceQueryBuilder):
             # not actually respected in remote query
             offset=self._offset or 0,
         )
-        return self._table._execute_query(query).read_all()
+        return self._table._execute_query(query)
 
     def rerank(self, reranker: Reranker) -> LanceEmptyQueryBuilder:
         """Rerank the results using the specified reranker.
@@ -1135,6 +1149,9 @@ class LanceHybridQueryBuilder(LanceQueryBuilder):
             results = results.drop(["_rowid"])
         return results
 
+    def to_batches(self):
+        raise NotImplementedError("to_batches not yet supported on a hybrid query")
+
     def _rank(self, results: pa.Table, column: str, ascending: bool = True):
         if len(results) == 0:
             return results
@@ -1502,10 +1519,11 @@ class AsyncQueryBase(object):
         ...     print(plan)
         >>> asyncio.run(doctest_example()) # doctest: +ELLIPSIS, +NORMALIZE_WHITESPACE
         ProjectionExec: expr=[vector@0 as vector, _distance@2 as _distance]
-          FilterExec: _distance@2 IS NOT NULL
+          GlobalLimitExec: skip=0, fetch=10
-            SortExec: TopK(fetch=10), expr=[_distance@2 ASC NULLS LAST], preserve_partitioning=[false]
+            FilterExec: _distance@2 IS NOT NULL
-              KNNVectorDistance: metric=l2
+              SortExec: TopK(fetch=10), expr=[_distance@2 ASC NULLS LAST], preserve_partitioning=[false]
-                LanceScan: uri=..., projection=[vector], row_id=true, row_addr=false, ordered=false
+                KNNVectorDistance: metric=l2
+                  LanceScan: uri=..., projection=[vector], row_id=true, row_addr=false, ordered=false
 
         Parameters
         ----------

python/python/lancedb/remote/background_loop.py

@@ -0,0 +1,25 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+import asyncio
+import threading
+
+
+class BackgroundEventLoop:
+    """
+    A background event loop that can run futures.
+
+    Used to bridge sync and async code, without messing with users event loops.
+    """
+
+    def __init__(self):
+        self.loop = asyncio.new_event_loop()
+        self.thread = threading.Thread(
+            target=self.loop.run_forever,
+            name="LanceDBBackgroundEventLoop",
+            daemon=True,
+        )
+        self.thread.start()
+
+    def run(self, future):
+        return asyncio.run_coroutine_threadsafe(future, self.loop).result()

python/python/lancedb/remote/db.py

@@ -11,7 +11,6 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 
-import asyncio
 from datetime import timedelta
 import logging
 from concurrent.futures import ThreadPoolExecutor
@@ -21,6 +20,7 @@ import warnings
 
 from lancedb import connect_async
 from lancedb.remote import ClientConfig
+from lancedb.remote.background_loop import BackgroundEventLoop
 import pyarrow as pa
 from overrides import override
 
@@ -31,6 +31,8 @@ from ..pydantic import LanceModel
 from ..table import Table
 from ..util import validate_table_name
 
+LOOP = BackgroundEventLoop()
+
 
 class RemoteDBConnection(DBConnection):
     """A connection to a remote LanceDB database."""
@@ -86,18 +88,9 @@ class RemoteDBConnection(DBConnection):
             raise ValueError(f"Invalid scheme: {parsed.scheme}, only accepts db://")
         self.db_name = parsed.netloc
 
-        import nest_asyncio
-
-        nest_asyncio.apply()
-        try:
-            self._loop = asyncio.get_running_loop()
-        except RuntimeError:
-            self._loop = asyncio.new_event_loop()
-            asyncio.set_event_loop(self._loop)
-
         self.client_config = client_config
 
-        self._conn = self._loop.run_until_complete(
+        self._conn = LOOP.run(
             connect_async(
                 db_url,
                 api_key=api_key,
@@ -127,9 +120,7 @@ class RemoteDBConnection(DBConnection):
         -------
         An iterator of table names.
         """
-        return self._loop.run_until_complete(
+        return LOOP.run(self._conn.table_names(start_after=page_token, limit=limit))
-            self._conn.table_names(start_after=page_token, limit=limit)
-        )
 
     @override
     def open_table(self, name: str, *, index_cache_size: Optional[int] = None) -> Table:
@@ -152,8 +143,8 @@ class RemoteDBConnection(DBConnection):
                 " (there is no local cache to configure)"
             )
 
-        table = self._loop.run_until_complete(self._conn.open_table(name))
+        table = LOOP.run(self._conn.open_table(name))
-        return RemoteTable(table, self.db_name, self._loop)
+        return RemoteTable(table, self.db_name)
 
     @override
     def create_table(
@@ -268,7 +259,7 @@ class RemoteDBConnection(DBConnection):
 
         from .table import RemoteTable
 
-        table = self._loop.run_until_complete(
+        table = LOOP.run(
             self._conn.create_table(
                 name,
                 data,
@@ -278,7 +269,7 @@ class RemoteDBConnection(DBConnection):
                 fill_value=fill_value,
             )
         )
-        return RemoteTable(table, self.db_name, self._loop)
+        return RemoteTable(table, self.db_name)
 
     @override
     def drop_table(self, name: str):
@@ -289,7 +280,7 @@ class RemoteDBConnection(DBConnection):
         name: str
             The name of the table.
         """
-        self._loop.run_until_complete(self._conn.drop_table(name))
+        LOOP.run(self._conn.drop_table(name))
 
     @override
     def rename_table(self, cur_name: str, new_name: str):
@@ -302,7 +293,7 @@ class RemoteDBConnection(DBConnection):
         new_name: str
             The new name of the table.
         """
-        self._loop.run_until_complete(self._conn.rename_table(cur_name, new_name))
+        LOOP.run(self._conn.rename_table(cur_name, new_name))
 
     async def close(self):
         """Close the connection to the database."""

python/python/lancedb/remote/table.py

@@ -12,12 +12,12 @@
 #  limitations under the License.
 
 from datetime import timedelta
-import asyncio
 import logging
 from functools import cached_property
 from typing import Dict, Iterable, List, Optional, Union, Literal
 
 from lancedb.index import FTS, BTree, Bitmap, HnswPq, HnswSq, IvfPq, LabelList
+from lancedb.remote.db import LOOP
 import pyarrow as pa
 
 from lancedb.common import DATA, VEC, VECTOR_COLUMN_NAME
@@ -33,9 +33,7 @@ class RemoteTable(Table):
         self,
         table: AsyncTable,
         db_name: str,
-        loop: Optional[asyncio.AbstractEventLoop] = None,
     ):
-        self._loop = loop
         self._table = table
         self.db_name = db_name
 
@@ -56,12 +54,12 @@ class RemoteTable(Table):
         of this Table
 
         """
-        return self._loop.run_until_complete(self._table.schema())
+        return LOOP.run(self._table.schema())
 
     @property
     def version(self) -> int:
         """Get the current version of the table"""
-        return self._loop.run_until_complete(self._table.version())
+        return LOOP.run(self._table.version())
 
     @cached_property
     def embedding_functions(self) -> dict:
@@ -98,11 +96,11 @@ class RemoteTable(Table):
 
     def list_indices(self):
         """List all the indices on the table"""
-        return self._loop.run_until_complete(self._table.list_indices())
+        return LOOP.run(self._table.list_indices())
 
     def index_stats(self, index_uuid: str):
         """List all the stats of a specified index"""
-        return self._loop.run_until_complete(self._table.index_stats(index_uuid))
+        return LOOP.run(self._table.index_stats(index_uuid))
 
     def create_scalar_index(
         self,
@@ -132,9 +130,7 @@ class RemoteTable(Table):
         else:
             raise ValueError(f"Unknown index type: {index_type}")
 
-        self._loop.run_until_complete(
+        LOOP.run(self._table.create_index(column, config=config, replace=replace))
-            self._table.create_index(column, config=config, replace=replace)
-        )
 
     def create_fts_index(
         self,
@@ -144,9 +140,7 @@ class RemoteTable(Table):
         with_position: bool = True,
     ):
         config = FTS(with_position=with_position)
-        self._loop.run_until_complete(
+        LOOP.run(self._table.create_index(column, config=config, replace=replace))
-            self._table.create_index(column, config=config, replace=replace)
-        )
 
     def create_index(
         self,
@@ -227,9 +221,7 @@ class RemoteTable(Table):
                 " 'IVF_PQ', 'IVF_HNSW_PQ', 'IVF_HNSW_SQ'"
             )
 
-        self._loop.run_until_complete(
+        LOOP.run(self._table.create_index(vector_column_name, config=config))
-            self._table.create_index(vector_column_name, config=config)
-        )
 
     def add(
         self,
@@ -261,7 +253,7 @@ class RemoteTable(Table):
             The value to use when filling vectors. Only used if on_bad_vectors="fill".
 
         """
-        self._loop.run_until_complete(
+        LOOP.run(
             self._table.add(
                 data, mode=mode, on_bad_vectors=on_bad_vectors, fill_value=fill_value
             )
@@ -349,9 +341,7 @@ class RemoteTable(Table):
     def _execute_query(
         self, query: Query, batch_size: Optional[int] = None
     ) -> pa.RecordBatchReader:
-        return self._loop.run_until_complete(
+        return LOOP.run(self._table._execute_query(query, batch_size=batch_size))
-            self._table._execute_query(query, batch_size=batch_size)
-        )
 
     def merge_insert(self, on: Union[str, Iterable[str]]) -> LanceMergeInsertBuilder:
         """Returns a [`LanceMergeInsertBuilder`][lancedb.merge.LanceMergeInsertBuilder]
@@ -368,9 +358,7 @@ class RemoteTable(Table):
         on_bad_vectors: str,
         fill_value: float,
     ):
-        self._loop.run_until_complete(
+        LOOP.run(self._table._do_merge(merge, new_data, on_bad_vectors, fill_value))
-            self._table._do_merge(merge, new_data, on_bad_vectors, fill_value)
-        )
 
     def delete(self, predicate: str):
         """Delete rows from the table.
@@ -419,7 +407,7 @@ class RemoteTable(Table):
            x      vector  _distance # doctest: +SKIP
         0  2  [3.0, 4.0]       85.0 # doctest: +SKIP
         """
-        self._loop.run_until_complete(self._table.delete(predicate))
+        LOOP.run(self._table.delete(predicate))
 
     def update(
         self,
@@ -469,7 +457,7 @@ class RemoteTable(Table):
         2  2  [10.0, 10.0] # doctest: +SKIP
 
         """
-        self._loop.run_until_complete(
+        LOOP.run(
             self._table.update(where=where, updates=values, updates_sql=values_sql)
         )
 
@@ -499,7 +487,7 @@ class RemoteTable(Table):
         )
 
     def count_rows(self, filter: Optional[str] = None) -> int:
-        return self._loop.run_until_complete(self._table.count_rows(filter))
+        return LOOP.run(self._table.count_rows(filter))
 
     def add_columns(self, transforms: Dict[str, str]):
         raise NotImplementedError(

python/python/tests/test_db.py

@@ -599,7 +599,9 @@ async def test_create_in_v2_mode(tmp_path):
     )
 
     async def is_in_v2_mode(tbl):
-        batches = await tbl.query().to_batches(max_batch_length=1024 * 10)
+        batches = (
+            await tbl.query().limit(10 * 1024).to_batches(max_batch_length=1024 * 10)
+        )
         num_batches = 0
         async for batch in batches:
             num_batches += 1

python/python/tests/test_duckdb.py

@@ -0,0 +1,21 @@
+import duckdb
+import pyarrow as pa
+
+import lancedb
+from lancedb.integrations.pyarrow import PyarrowDatasetAdapter
+
+
+def test_basic_query(tmp_path):
+    data = pa.table({"x": [1, 2, 3, 4], "y": [5, 6, 7, 8]})
+    conn = lancedb.connect(tmp_path)
+    tbl = conn.create_table("test", data)
+
+    adapter = PyarrowDatasetAdapter(tbl)  # noqa: F841
+
+    duck_conn = duckdb.connect()
+
+    results = duck_conn.sql("SELECT SUM(x) FROM adapter").fetchall()
+    assert results[0][0] == 10
+
+    results = duck_conn.sql("SELECT SUM(y) FROM adapter").fetchall()
+    assert results[0][0] == 26

python/python/tests/test_embeddings.py

@@ -90,10 +90,13 @@ def test_embedding_with_bad_results(tmp_path):
             self, texts: Union[List[str], np.ndarray]
         ) -> list[Union[np.array, None]]:
             # Return None, which is bad if field is non-nullable
-            return [
+            a = [
-                None if i % 2 == 0 else np.random.randn(self.ndims())
+                np.full(self.ndims(), np.nan)
+                if i % 2 == 0
+                else np.random.randn(self.ndims())
                 for i in range(len(texts))
             ]
+            return a
 
     db = lancedb.connect(tmp_path)
     registry = EmbeddingFunctionRegistry.get_instance()

python/python/tests/test_embeddings_slow.py

@@ -1,15 +1,6 @@
-#  Copyright (c) 2023. LanceDB Developers
+# SPDX-License-Identifier: Apache-2.0
-#
+# SPDX-FileCopyrightText: Copyright The LanceDB Authors
-#  Licensed under the Apache License, Version 2.0 (the "License");
+
-#  you may not use this file except in compliance with the License.
-#  You may obtain a copy of the License at
-#      http://www.apache.org/licenses/LICENSE-2.0
-#
-#  Unless required by applicable law or agreed to in writing, software
-#  distributed under the License is distributed on an "AS IS" BASIS,
-#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-#  See the License for the specific language governing permissions and
-#  limitations under the License.
 import importlib
 import io
 import os
@@ -17,6 +8,7 @@ import os
 import lancedb
 import numpy as np
 import pandas as pd
+import pyarrow as pa
 import pytest
 from lancedb.embeddings import get_registry
 from lancedb.pydantic import LanceModel, Vector
@@ -444,6 +436,30 @@ def test_watsonx_embedding(tmp_path):
         assert tbl.search("hello").limit(1).to_pandas()["text"][0] == "hello world"
 
 
+@pytest.mark.slow
+@pytest.mark.skipif(
+    os.environ.get("OPENAI_API_KEY") is None, reason="OPENAI_API_KEY not set"
+)
+def test_openai_with_empty_strs(tmp_path):
+    model = get_registry().get("openai").create(max_retries=0)
+
+    class TextModel(LanceModel):
+        text: str = model.SourceField()
+        vector: Vector(model.ndims()) = model.VectorField()
+
+    df = pd.DataFrame({"text": ["hello world", ""]})
+    db = lancedb.connect(tmp_path)
+    tbl = db.create_table("test", schema=TextModel, mode="overwrite")
+
+    tbl.add(df, on_bad_vectors="skip")
+    tb = tbl.to_arrow()
+    assert tb.schema.field_by_name("vector").type == pa.list_(
+        pa.float32(), model.ndims()
+    )
+    assert len(tb) == 2
+    assert tb["vector"].is_null().to_pylist() == [False, True]
+
+
 @pytest.mark.slow
 @pytest.mark.skipif(
     importlib.util.find_spec("ollama") is None, reason="Ollama not installed"

python/python/tests/test_pyarrow.py

@@ -0,0 +1,47 @@
+import pyarrow as pa
+
+import lancedb
+from lancedb.integrations.pyarrow import PyarrowDatasetAdapter
+
+
+def test_dataset_adapter(tmp_path):
+    data = pa.table({"x": [1, 2, 3, 4], "y": [5, 6, 7, 8]})
+    conn = lancedb.connect(tmp_path)
+    tbl = conn.create_table("test", data)
+
+    adapter = PyarrowDatasetAdapter(tbl)
+
+    assert adapter.count_rows() == 4
+    assert adapter.count_rows("x > 2") == 2
+    assert adapter.schema == data.schema
+    assert adapter.head(2) == data.slice(0, 2)
+    assert adapter.to_table() == data
+    assert adapter.to_batches().read_all() == data
+    assert adapter.scanner().to_table() == data
+    assert adapter.scanner().to_batches().read_all() == data
+
+    assert adapter.scanner().projected_schema == data.schema
+    assert adapter.scanner(columns=["x"]).projected_schema == pa.schema(
+        [data.schema.field("x")]
+    )
+    assert adapter.scanner(columns=["x"]).to_table() == pa.table({"x": [1, 2, 3, 4]})
+
+    # Make sure we bypass the limit
+    data = pa.table({"x": range(100)})
+    tbl = conn.create_table("test2", data)
+
+    adapter = PyarrowDatasetAdapter(tbl)
+
+    assert adapter.count_rows() == 100
+    assert adapter.to_table().num_rows == 100
+    assert adapter.head(10).num_rows == 10
+
+    # Empty table
+    tbl = conn.create_table("test3", None, schema=pa.schema({"x": pa.int64()}))
+    adapter = PyarrowDatasetAdapter(tbl)
+
+    assert adapter.count_rows() == 0
+    assert adapter.to_table().num_rows == 0
+    assert adapter.head(10).num_rows == 0
+
+    assert adapter.scanner().projected_schema == pa.schema({"x": pa.int64()})

python/python/tests/test_pydantic.py

@@ -1,16 +1,5 @@
-#  Copyright 2023 LanceDB Developers
+# SPDX-License-Identifier: Apache-2.0
-#
+# SPDX-FileCopyrightText: Copyright The LanceDB Authors
-#  Licensed under the Apache License, Version 2.0 (the "License");
-#  you may not use this file except in compliance with the License.
-#  You may obtain a copy of the License at
-#      http://www.apache.org/licenses/LICENSE-2.0
-#
-#  Unless required by applicable law or agreed to in writing, software
-#  distributed under the License is distributed on an "AS IS" BASIS,
-#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-#  See the License for the specific language governing permissions and
-#  limitations under the License.
-
 
 import json
 import sys
@@ -172,6 +161,26 @@ def test_pydantic_to_arrow_py38():
     assert schema == expect_schema
 
 
+def test_nullable_vector():
+    class NullableModel(pydantic.BaseModel):
+        vec: Vector(16, nullable=False)
+
+    schema = pydantic_to_schema(NullableModel)
+    assert schema == pa.schema([pa.field("vec", pa.list_(pa.float32(), 16), False)])
+
+    class DefaultModel(pydantic.BaseModel):
+        vec: Vector(16)
+
+    schema = pydantic_to_schema(DefaultModel)
+    assert schema == pa.schema([pa.field("vec", pa.list_(pa.float32(), 16), True)])
+
+    class NotNullableModel(pydantic.BaseModel):
+        vec: Vector(16)
+
+    schema = pydantic_to_schema(NotNullableModel)
+    assert schema == pa.schema([pa.field("vec", pa.list_(pa.float32(), 16), True)])
+
+
 def test_fixed_size_list_field():
     class TestModel(pydantic.BaseModel):
         vec: Vector(16)
@@ -192,7 +201,7 @@ def test_fixed_size_list_field():
     schema = pydantic_to_schema(TestModel)
     assert schema == pa.schema(
         [
-            pa.field("vec", pa.list_(pa.float32(), 16), False),
+            pa.field("vec", pa.list_(pa.float32(), 16)),
             pa.field("li", pa.list_(pa.int64()), False),
         ]
     )

python/python/tests/test_remote_db.py

@@ -1,6 +1,7 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
+from concurrent.futures import ThreadPoolExecutor
 import contextlib
 from datetime import timedelta
 import http.server
@@ -187,6 +188,47 @@ async def test_retry_error():
         assert cause.status_code == 429
 
 
+def test_table_add_in_threadpool():
+    def handler(request):
+        if request.path == "/v1/table/test/insert/":
+            request.send_response(200)
+            request.end_headers()
+        elif request.path == "/v1/table/test/create/?mode=create":
+            request.send_response(200)
+            request.send_header("Content-Type", "application/json")
+            request.end_headers()
+            request.wfile.write(b"{}")
+        elif request.path == "/v1/table/test/describe/":
+            request.send_response(200)
+            request.send_header("Content-Type", "application/json")
+            request.end_headers()
+            payload = json.dumps(
+                dict(
+                    version=1,
+                    schema=dict(
+                        fields=[
+                            dict(name="id", type={"type": "int64"}, nullable=False),
+                        ]
+                    ),
+                )
+            )
+            request.wfile.write(payload.encode())
+        else:
+            request.send_response(404)
+            request.end_headers()
+
+    with mock_lancedb_connection(handler) as db:
+        table = db.create_table("test", [{"id": 1}])
+        with ThreadPoolExecutor(3) as executor:
+            futures = []
+            for _ in range(10):
+                future = executor.submit(table.add, [{"id": 1}])
+                futures.append(future)
+
+            for future in futures:
+                future.result()
+
+
 @contextlib.contextmanager
 def query_test_table(query_handler):
     def handler(request):

rust/ffi/node/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb-node"
-version = "0.13.0"
+version = "0.14.0-beta.1"
 description = "Serverless, low-latency vector database for AI applications"
 license.workspace = true
 edition.workspace = true

rust/lancedb/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb"
-version = "0.13.0"
+version = "0.14.0-beta.1"
 edition.workspace = true
 description = "LanceDB: A serverless, low-latency vector database for AI applications"
 license.workspace = true
@@ -27,6 +27,7 @@ half = { workspace = true }
 lazy_static.workspace = true
 lance = { workspace = true }
 lance-datafusion.workspace = true
+lance-io = { workspace = true }
 lance-index = { workspace = true }
 lance-table = { workspace = true }
 lance-linalg = { workspace = true }

rust/lancedb/src/connection.rs

@@ -38,6 +38,9 @@ use crate::table::{NativeTable, TableDefinition, WriteOptions};
 use crate::utils::validate_table_name;
 use crate::Table;
 pub use lance_encoding::version::LanceFileVersion;
+#[cfg(feature = "remote")]
+use lance_io::object_store::StorageOptions;
+use lance_table::io::commit::commit_handler_from_url;
 
 pub const LANCE_FILE_EXTENSION: &str = "lance";
 
@@ -133,7 +136,7 @@ impl IntoArrow for NoData {
 
 /// A builder for configuring a [`Connection::create_table`] operation
 pub struct CreateTableBuilder<const HAS_DATA: bool, T: IntoArrow> {
-    parent: Arc<dyn ConnectionInternal>,
+    pub(crate) parent: Arc<dyn ConnectionInternal>,
     pub(crate) name: String,
     pub(crate) data: Option<T>,
     pub(crate) mode: CreateTableMode,
@@ -341,7 +344,7 @@ pub struct OpenTableBuilder {
 }
 
 impl OpenTableBuilder {
-    fn new(parent: Arc<dyn ConnectionInternal>, name: String) -> Self {
+    pub(crate) fn new(parent: Arc<dyn ConnectionInternal>, name: String) -> Self {
         Self {
             parent,
             name,
@@ -717,12 +720,14 @@ impl ConnectBuilder {
             message: "An api_key is required when connecting to LanceDb Cloud".to_string(),
         })?;
 
+        let storage_options = StorageOptions(self.storage_options.clone());
         let internal = Arc::new(crate::remote::db::RemoteDatabase::try_new(
             &self.uri,
             &api_key,
             &region,
             self.host_override,
             self.client_config,
+            storage_options.into(),
         )?);
         Ok(Connection {
             internal,
@@ -855,7 +860,7 @@ impl Database {
                 let table_base_uri = if let Some(store) = engine {
                     static WARN_ONCE: std::sync::Once = std::sync::Once::new();
                     WARN_ONCE.call_once(|| {
-                        log::warn!("Specifing engine is not a publicly supported feature in lancedb yet. THE API WILL CHANGE");
+                        log::warn!("Specifying engine is not a publicly supported feature in lancedb yet. THE API WILL CHANGE");
                     });
                     let old_scheme = url.scheme().to_string();
                     let new_scheme = format!("{}+{}", old_scheme, store);
@@ -1036,6 +1041,7 @@ impl ConnectionInternal for Database {
         };
 
         let mut write_params = options.write_options.lance_write_params.unwrap_or_default();
+
         if matches!(&options.mode, CreateTableMode::Overwrite) {
             write_params.mode = WriteMode::Overwrite;
         }
@@ -1122,7 +1128,7 @@ impl ConnectionInternal for Database {
         let dir_name = format!("{}.{}", name, LANCE_EXTENSION);
         let full_path = self.base_path.child(dir_name.clone());
         self.object_store
-            .remove_dir_all(full_path)
+            .remove_dir_all(full_path.clone())
             .await
             .map_err(|err| match err {
                 // this error is not lance::Error::DatasetNotFound,
@@ -1132,6 +1138,19 @@ impl ConnectionInternal for Database {
                 },
                 _ => Error::from(err),
             })?;
+
+        let object_store_params = ObjectStoreParams {
+            storage_options: Some(self.storage_options.clone()),
+            ..Default::default()
+        };
+        let mut uri = self.uri.clone();
+        if let Some(query_string) = &self.query_string {
+            uri.push_str(&format!("?{}", query_string));
+        }
+        let commit_handler = commit_handler_from_url(&uri, &Some(object_store_params))
+            .await
+            .unwrap();
+        commit_handler.delete(&full_path).await.unwrap();
         Ok(())
     }
 
@@ -1169,6 +1188,7 @@ mod tests {
     use lance_testing::datagen::{BatchGenerator, IncrementingInt32};
     use tempfile::tempdir;
 
+    use crate::query::QueryBase;
     use crate::query::{ExecutableQuery, QueryExecutionOptions};
 
     use super::*;
@@ -1296,6 +1316,7 @@ mod tests {
         // In v1 the row group size will trump max_batch_length
         let batches = tbl
             .query()
+            .limit(20000)
             .execute_with_options(QueryExecutionOptions {
                 max_batch_length: 50000,
                 ..Default::default()

rust/lancedb/src/query.rs

@@ -596,7 +596,7 @@ impl Query {
     pub(crate) fn new(parent: Arc<dyn TableInternal>) -> Self {
         Self {
             parent,
-            limit: None,
+            limit: Some(DEFAULT_TOP_K),
             offset: None,
             filter: None,
             full_text_search: None,

rust/lancedb/src/remote/client.rs

@@ -21,6 +21,7 @@ use reqwest::{
 };
 
 use crate::error::{Error, Result};
+use crate::remote::db::RemoteOptions;
 
 const REQUEST_ID_HEADER: &str = "x-request-id";
 
@@ -215,6 +216,7 @@ impl RestfulLanceDbClient<Sender> {
         region: &str,
         host_override: Option<String>,
         client_config: ClientConfig,
+        options: &RemoteOptions,
     ) -> Result<Self> {
         let parsed_url = url::Url::parse(db_url).map_err(|err| Error::InvalidInput {
             message: format!("db_url is not a valid URL. '{db_url}'. Error: {err}"),
@@ -226,6 +228,14 @@ impl RestfulLanceDbClient<Sender> {
             });
         }
         let db_name = parsed_url.host_str().unwrap();
+        let db_prefix = {
+            let prefix = parsed_url.path().trim_start_matches('/');
+            if prefix.is_empty() {
+                None
+            } else {
+                Some(prefix)
+            }
+        };
 
         // Get the timeouts
         let connect_timeout = Self::get_timeout(
@@ -255,6 +265,8 @@ impl RestfulLanceDbClient<Sender> {
                 region,
                 db_name,
                 host_override.is_some(),
+                options,
+                db_prefix,
             )?)
             .user_agent(client_config.user_agent)
             .build()
@@ -262,6 +274,7 @@ impl RestfulLanceDbClient<Sender> {
                 message: "Failed to build HTTP client".into(),
                 source: Some(Box::new(err)),
             })?;
+
         let host = match host_override {
             Some(host_override) => host_override,
             None => format!("https://{}.{}.api.lancedb.com", db_name, region),
@@ -287,6 +300,8 @@ impl<S: HttpSend> RestfulLanceDbClient<S> {
         region: &str,
         db_name: &str,
         has_host_override: bool,
+        options: &RemoteOptions,
+        db_prefix: Option<&str>,
     ) -> Result<HeaderMap> {
         let mut headers = HeaderMap::new();
         headers.insert(
@@ -312,6 +327,34 @@ impl<S: HttpSend> RestfulLanceDbClient<S> {
                 })?,
             );
         }
+        if db_prefix.is_some() {
+            headers.insert(
+                "x-lancedb-database-prefix",
+                HeaderValue::from_str(db_prefix.unwrap()).map_err(|_| Error::InvalidInput {
+                    message: format!(
+                        "non-ascii database prefix '{}' provided",
+                        db_prefix.unwrap()
+                    ),
+                })?,
+            );
+        }
+
+        if let Some(v) = options.0.get("account_name") {
+            headers.insert(
+                "x-azure-storage-account-name",
+                HeaderValue::from_str(v).map_err(|_| Error::InvalidInput {
+                    message: format!("non-ascii storage account name '{}' provided", db_name),
+                })?,
+            );
+        }
+        if let Some(v) = options.0.get("azure_storage_account_name") {
+            headers.insert(
+                "x-azure-storage-account-name",
+                HeaderValue::from_str(v).map_err(|_| Error::InvalidInput {
+                    message: format!("non-ascii storage account name '{}' provided", db_name),
+                })?,
+            );
+        }
 
         Ok(headers)
     }

rust/lancedb/src/remote/db.rs

@@ -12,18 +12,21 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
+use std::collections::HashMap;
 use std::sync::Arc;
 
 use arrow_array::RecordBatchReader;
 use async_trait::async_trait;
 use http::StatusCode;
+use lance_io::object_store::StorageOptions;
 use moka::future::Cache;
 use reqwest::header::CONTENT_TYPE;
 use serde::Deserialize;
 use tokio::task::spawn_blocking;
 
 use crate::connection::{
-    ConnectionInternal, CreateTableBuilder, NoData, OpenTableBuilder, TableNamesBuilder,
+    ConnectionInternal, CreateTableBuilder, CreateTableMode, NoData, OpenTableBuilder,
+    TableNamesBuilder,
 };
 use crate::embeddings::EmbeddingRegistry;
 use crate::error::Result;
@@ -52,9 +55,16 @@ impl RemoteDatabase {
         region: &str,
         host_override: Option<String>,
         client_config: ClientConfig,
+        options: RemoteOptions,
     ) -> Result<Self> {
-        let client =
+        let client = RestfulLanceDbClient::try_new(
-            RestfulLanceDbClient::try_new(uri, api_key, region, host_override, client_config)?;
+            uri,
+            api_key,
+            region,
+            host_override,
+            client_config,
+            &options,
+        )?;
 
         let table_cache = Cache::builder()
             .time_to_live(std::time::Duration::from_secs(300))
@@ -95,6 +105,16 @@ impl<S: HttpSend> std::fmt::Display for RemoteDatabase<S> {
     }
 }
 
+impl From<&CreateTableMode> for &'static str {
+    fn from(val: &CreateTableMode) -> Self {
+        match val {
+            CreateTableMode::Create => "create",
+            CreateTableMode::Overwrite => "overwrite",
+            CreateTableMode::ExistOk(_) => "exist_ok",
+        }
+    }
+}
+
 #[async_trait]
 impl<S: HttpSend> ConnectionInternal for RemoteDatabase<S> {
     async fn table_names(&self, options: TableNamesBuilder) -> Result<Vec<String>> {
@@ -133,14 +153,40 @@ impl<S: HttpSend> ConnectionInternal for RemoteDatabase<S> {
         let req = self
             .client
             .post(&format!("/v1/table/{}/create/", options.name))
+            .query(&[("mode", Into::<&str>::into(&options.mode))])
             .body(data_buffer)
             .header(CONTENT_TYPE, ARROW_STREAM_CONTENT_TYPE);
+
         let (request_id, rsp) = self.client.send(req, false).await?;
 
         if rsp.status() == StatusCode::BAD_REQUEST {
             let body = rsp.text().await.err_to_http(request_id.clone())?;
             if body.contains("already exists") {
-                return Err(crate::Error::TableAlreadyExists { name: options.name });
+                return match options.mode {
+                    CreateTableMode::Create => {
+                        Err(crate::Error::TableAlreadyExists { name: options.name })
+                    }
+                    CreateTableMode::ExistOk(callback) => {
+                        let builder = OpenTableBuilder::new(options.parent, options.name);
+                        let builder = (callback)(builder);
+                        builder.execute().await
+                    }
+
+                    // This should not happen, as we explicitly set the mode to overwrite and the server
+                    // shouldn't return an error if the table already exists.
+                    //
+                    // However if the server is an older version that doesn't support the mode parameter,
+                    // then we'll get the 400 response.
+                    CreateTableMode::Overwrite => Err(crate::Error::Http {
+                        source: format!(
+                            "unexpected response from server for create mode overwrite: {}",
+                            body
+                        )
+                        .into(),
+                        request_id,
+                        status_code: Some(StatusCode::BAD_REQUEST),
+                    }),
+                };
             } else {
                 return Err(crate::Error::InvalidInput { message: body });
             }
@@ -206,6 +252,29 @@ impl<S: HttpSend> ConnectionInternal for RemoteDatabase<S> {
     }
 }
 
+/// RemoteOptions contains a subset of StorageOptions that are compatible with Remote LanceDB connections
+#[derive(Clone, Debug, Default)]
+pub struct RemoteOptions(pub HashMap<String, String>);
+
+impl RemoteOptions {
+    pub fn new(options: HashMap<String, String>) -> Self {
+        Self(options)
+    }
+}
+
+impl From<StorageOptions> for RemoteOptions {
+    fn from(options: StorageOptions) -> Self {
+        let supported_opts = vec!["account_name", "azure_storage_account_name"];
+        let mut filtered = HashMap::new();
+        for opt in supported_opts {
+            if let Some(v) = options.0.get(opt) {
+                filtered.insert(opt.to_string(), v.to_string());
+            }
+        }
+        RemoteOptions::new(filtered)
+    }
+}
+
 #[cfg(test)]
 mod tests {
     use std::sync::{Arc, OnceLock};
@@ -213,7 +282,9 @@ mod tests {
     use arrow_array::{Int32Array, RecordBatch, RecordBatchIterator};
     use arrow_schema::{DataType, Field, Schema};
 
+    use crate::connection::ConnectBuilder;
     use crate::{
+        connection::CreateTableMode,
         remote::{ARROW_STREAM_CONTENT_TYPE, JSON_CONTENT_TYPE},
         Connection, Error,
     };
@@ -382,6 +453,73 @@ mod tests {
         );
     }
 
+    #[tokio::test]
+    async fn test_create_table_modes() {
+        let test_cases = [
+            (None, "mode=create"),
+            (Some(CreateTableMode::Create), "mode=create"),
+            (Some(CreateTableMode::Overwrite), "mode=overwrite"),
+            (
+                Some(CreateTableMode::ExistOk(Box::new(|b| b))),
+                "mode=exist_ok",
+            ),
+        ];
+
+        for (mode, expected_query_string) in test_cases {
+            let conn = Connection::new_with_handler(move |request| {
+                assert_eq!(request.method(), &reqwest::Method::POST);
+                assert_eq!(request.url().path(), "/v1/table/table1/create/");
+                assert_eq!(request.url().query(), Some(expected_query_string));
+
+                http::Response::builder().status(200).body("").unwrap()
+            });
+
+            let data = RecordBatch::try_new(
+                Arc::new(Schema::new(vec![Field::new("a", DataType::Int32, false)])),
+                vec![Arc::new(Int32Array::from(vec![1, 2, 3]))],
+            )
+            .unwrap();
+            let reader = RecordBatchIterator::new([Ok(data.clone())], data.schema());
+            let mut builder = conn.create_table("table1", reader);
+            if let Some(mode) = mode {
+                builder = builder.mode(mode);
+            }
+            builder.execute().await.unwrap();
+        }
+
+        // check that the open table callback is called with exist_ok
+        let conn = Connection::new_with_handler(|request| match request.url().path() {
+            "/v1/table/table1/create/" => http::Response::builder()
+                .status(400)
+                .body("Table table1 already exists")
+                .unwrap(),
+            "/v1/table/table1/describe/" => http::Response::builder().status(200).body("").unwrap(),
+            _ => {
+                panic!("unexpected path: {:?}", request.url().path());
+            }
+        });
+        let data = RecordBatch::try_new(
+            Arc::new(Schema::new(vec![Field::new("a", DataType::Int32, false)])),
+            vec![Arc::new(Int32Array::from(vec![1, 2, 3]))],
+        )
+        .unwrap();
+
+        let called: Arc<OnceLock<bool>> = Arc::new(OnceLock::new());
+        let reader = RecordBatchIterator::new([Ok(data.clone())], data.schema());
+        let called_in_cb = called.clone();
+        conn.create_table("table1", reader)
+            .mode(CreateTableMode::ExistOk(Box::new(move |b| {
+                called_in_cb.clone().set(true).unwrap();
+                b
+            })))
+            .execute()
+            .await
+            .unwrap();
+
+        let called = *called.get().unwrap_or(&false);
+        assert!(called);
+    }
+
     #[tokio::test]
     async fn test_create_table_empty() {
         let conn = Connection::new_with_handler(|request| {
@@ -436,4 +574,16 @@ mod tests {
         });
         conn.rename_table("table1", "table2").await.unwrap();
     }
+
+    #[tokio::test]
+    async fn test_connect_remote_options() {
+        let db_uri = "db://my-container/my-prefix";
+        let _ = ConnectBuilder::new(db_uri)
+            .region("us-east-1")
+            .api_key("my-api-key")
+            .storage_options(vec![("azure_storage_account_name", "my-storage-account")])
+            .execute()
+            .await
+            .unwrap();
+    }
 }

rust/lancedb/src/remote/table.rs

@@ -1227,6 +1227,7 @@ mod tests {
                 "prefilter": true,
                 "distance_type": "l2",
                 "nprobes": 20,
+                "k": 10,
                 "ef": Option::<usize>::None,
                 "refine_factor": null,
                 "version": null,