feat(nodejs): surface skip_auto_cleanup on add and merge insert

### Summary 

Adds a `connectNamespace(implName, properties, options?)` to the NodeJS
SDK`. Closes #3380.

### Testing
- pnpm test
- Ran smoke test

```
import { connectNamespace } from "lancedb"
import { tmpdir } from "os";
import { mkdtempSync } from "fs";
import { join } from "path";

const dir = mkdtempSync(join(tmpdir(), "lancedb-connect-namespace-smoke-"));
console.log(`Using temp dir: ${dir}\n`);

// 1. Happy path: connect via the "dir" namespace impl, create + list a table.
console.log('Connecting via connectNamespace("dir", { root })...');
const db = await connectNamespace("dir", { root: dir });
console.log("  ✓ connected:", db.display());

console.log("Creating a table and listing it...");
await db.createTable("users", [
  { id: 1, name: "alice" },
  { id: 2, name: "bob" },
]);
console.log("  ✓ tableNames ->", await db.tableNames());

const table = await db.openTable("users");
console.log("  ✓ users.countRows ->", await table.countRows());

// 2. Storage options pass-through.
console.log("\nReconnecting with storageOptions (plumbing check)...");
const dbWithOpts = await connectNamespace(
  "dir",
  { root: dir },
  { storageOptions: { newTableDataStorageVersion: "stable" } },
);
console.log("  ✓ connected with storageOptions:", dbWithOpts.display());
await dbWithOpts.close();

// 3. Empty implName -> clear error.
console.log("\nCalling connectNamespace('', {}) (expect error)...");
try {
  await connectNamespace("", {});
  console.error("  UNEXPECTED: empty implName did not throw");
} catch (err) {
  console.log(`  ✓ Got expected error: ${err.message.split("\n")[0]}`);
}

// 4. Unknown impl -> error.
console.log("\nCalling connectNamespace('not-a-real-impl', {}) (expect error)...");
try {
  await connectNamespace("not-a-real-impl", {});
  console.error("  UNEXPECTED: unknown impl did not throw");
} catch (err) {
  console.log(`  ✓ Got expected error: ${err.message.split("\n")[0]}`);
}

// 5. Create a table inside a child namespace, then reconnect with a fresh
//    connectNamespace call and confirm the table is reachable via that
//    namespace path. (The dir+manifest impl keeps the namespace hierarchy in
//    a root manifest, so "scoping" happens via namespacePath args, not by
//    pointing root at a subdir.)
console.log("\nCreating a table inside a child namespace...");
const dir2 = mkdtempSync(join(tmpdir(), "lancedb-connect-namespace-smoke-"));
const writer = await connectNamespace("dir", {
  root: dir2,
  manifest_enabled: "true",
});
await writer.createNamespace(["analytics"]);
await writer.createTable(
  "orders",
  [
    { id: 1, total: 10 },
    { id: 2, total: 20 },
  ],
  ["analytics"],
);
console.log(
  "  ✓ writer sees tables under [analytics] ->",
  await writer.tableNames(["analytics"]),
);
await writer.close();

console.log("Reconnecting and reading the table via its namespace path...");
const reader = await connectNamespace("dir", {
  root: dir2,
  manifest_enabled: "true",
});
console.log(
  "  ✓ reader tableNames(['analytics']) ->",
  await reader.tableNames(["analytics"]),
);
const orders = await reader.openTable("orders", ["analytics"]);
console.log("  ✓ orders.countRows via reader ->", await orders.countRows());
await reader.close();

await db.close();
console.log("\nAll checks passed.");
```

```
Using temp dir: /var/folders/bj/hn6jv9c50y301d1nx0y8xmn00000gn/T/lancedb-connect-namespace-smoke-WByF1P

Connecting via connectNamespace("dir", { root })...
  ✓ connected: LanceNamespaceDatabase
Creating a table and listing it...
  ✓ tableNames -> [ 'users' ]
  ✓ users.countRows -> 2

Reconnecting with storageOptions (plumbing check)...
  ✓ connected with storageOptions: LanceNamespaceDatabase

Calling connectNamespace('', {}) (expect error)...
  ✓ Got expected error: implName must be a non-empty string

Calling connectNamespace('not-a-real-impl', {}) (expect error)...
  ✓ Got expected error: Invalid input, Failed to connect to namespace: Namespace { source: Unsupported { message: "Implementation 'not-a-real-impl' is not available. Supported: dir, rest" }, location: Location { file: "/Users/brendan/.cargo/git/checkouts/lance-8ddea23c38163eda/f693245/rust/lance-namespace-impls/src/connect.rs", line: 216, column: 14 } }

Creating a table inside a child namespace...
  ✓ writer sees tables under [analytics] -> [ 'orders' ]
Reconnecting and reading the table via its namespace path...
  ✓ reader tableNames(['analytics']) -> [ 'orders' ]
  ✓ orders.countRows via reader -> 2

All checks passed.
```

### Docs
- regenerated docs

.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.27.2-beta.1"
+current_version = "0.28.0-beta.11"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

.github/ISSUE_TEMPLATE/documentation.yml

@@ -18,6 +18,6 @@ body:
       label: Link
       description: >
         Provide a link to the existing documentation, if applicable.
-      placeholder: ex. https://lancedb.com/docs/tables/...
+      placeholder: ex. https://docs.lancedb.com/tables/...
     validations:
       required: false

.github/dependabot.yml

@@ -0,0 +1,18 @@
+version: 2
+
+# Scope: the root Cargo workspace, which produces the Rust binaries we
+# ship to users (the Node.js and Python native extensions). The
+# `rust/lancedb` library crate shares the same lockfile; its consumers
+# pick their own dependency versions, but bumping transitive deps here
+# keeps the binaries we ship current.
+updates:
+  - package-ecosystem: cargo
+    directory: /
+    schedule:
+      interval: weekly
+    open-pull-requests-limit: 10
+    groups:
+      rust-minor-patch:
+        update-types:
+          - minor
+          - patch

.github/workflows/codex-fix-ci.yml

@@ -45,7 +45,9 @@ jobs:
       - name: Set up Node.js
         uses: actions/setup-node@v4
         with:
-          node-version: 20
+          # pnpm 11 (used by the nodejs install step below) requires
+          # Node >= 22.13; use 24 since 22 hits EOL in October.
+          node-version: 24
 
       - name: Install Codex CLI
         run: npm install -g @openai/codex
@@ -79,10 +81,14 @@ jobs:
           java-version: '11'
           cache: maven
 
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 11.1.1
       - name: Install Node.js dependencies for TypeScript bindings
         run: |
           cd nodejs
-          npm ci
+          pnpm install --frozen-lockfile
 
       - name: Configure git user
         run: |
@@ -137,7 +143,7 @@ jobs:
                - For Rust test failures: Run the specific test with "cargo test -p <crate> <test_name>"
                - For Python test failures: Build with "cd python && maturin develop" then run "pytest <specific_test_file>::<test_name>"
                - For Java test failures: Run "cd java && mvn test -Dtest=<TestClass>#<testMethod>"
-               - For TypeScript test failures: Run "cd nodejs && npm run build && npm test -- --testNamePattern='<test_name>'"
+               - For TypeScript test failures: Run "cd nodejs && pnpm build && pnpm test -- --testNamePattern='<test_name>'"
                - Do NOT run the full test suite - only run the tests that were failing
 
           7. If the additional guidelines are provided, follow them as well.

.github/workflows/dev.yml

@@ -8,6 +8,9 @@ concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
   cancel-in-progress: true
 
+permissions:
+  contents: read
+
 jobs:
   labeler:
     permissions:

.github/workflows/java-publish.yml

@@ -19,6 +19,9 @@ on:
     paths:
       - .github/workflows/java-publish.yml
 
+permissions:
+  contents: read
+
 jobs:
   publish:
     name: Build and Publish
@@ -40,7 +43,7 @@ jobs:
           server-username: SONATYPE_USER
           server-password: SONATYPE_TOKEN
           gpg-private-key: ${{ secrets.GPG_PRIVATE_KEY }}
-          gpg-passphrase: ${{ secrets.GPG_PASSPHRASE }}
+          gpg-passphrase: MAVEN_GPG_PASSPHRASE
       - name: Set git config
         run: |
           git config --global user.email "dev+gha@lancedb.com"
@@ -55,10 +58,11 @@ jobs:
           echo "use-agent" >> ~/.gnupg/gpg.conf
           echo "pinentry-mode loopback" >> ~/.gnupg/gpg.conf
           export GPG_TTY=$(tty)
-          ./mvnw --batch-mode -DskipTests -DpushChanges=false -Dgpg.passphrase=${{ secrets.GPG_PASSPHRASE }} deploy -pl lancedb-core -am -P deploy-to-ossrh
+          ./mvnw --batch-mode -DskipTests -DpushChanges=false deploy -pl lancedb-core -am -P deploy-to-ossrh
         env:
           SONATYPE_USER: ${{ secrets.SONATYPE_USER }}
           SONATYPE_TOKEN: ${{ secrets.SONATYPE_TOKEN }}
+          MAVEN_GPG_PASSPHRASE: ${{ secrets.GPG_PASSPHRASE }}
 
   report-failure:
     name: Report Workflow Failure

.github/workflows/java.yml

@@ -24,6 +24,9 @@ on:
       - java/**
       - .github/workflows/java.yml
 
+permissions:
+  contents: read
+
 jobs:
   build-java:
     runs-on: ubuntu-24.04

.github/workflows/license-header-check.yml

@@ -10,6 +10,10 @@ on:
       - nodejs/**
       - java/**
       - .github/workflows/license-header-check.yml
+
+permissions:
+  contents: read
+
 jobs:
   check-licenses:
     runs-on: ubuntu-latest

.github/workflows/nodejs.yml

@@ -8,12 +8,16 @@ on:
     paths:
       - Cargo.toml
       - Cargo.lock
+      - rust-toolchain.toml
       - nodejs/**
       - rust/**
       - docs/src/js/**
       - .github/workflows/nodejs.yml
       - docker-compose.yml
 
+permissions:
+  contents: read
+
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
   cancel-in-progress: true
@@ -38,11 +42,17 @@ jobs:
       with:
         fetch-depth: 0
         lfs: true
+    - uses: pnpm/action-setup@v4
+      with:
+        version: 11.1.1
     - uses: actions/setup-node@v4
       with:
-        node-version: 20
+        # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
-        cache: 'npm'
+        # in October. The library itself still supports Node >= 18
-        cache-dependency-path: nodejs/package-lock.json
+        # (see test matrix below).
+        node-version: 24
+        cache: 'pnpm'
+        cache-dependency-path: nodejs/pnpm-lock.yaml
     - uses: actions-rust-lang/setup-rust-toolchain@v1
       with:
         components: rustfmt, clippy
@@ -57,11 +67,13 @@ jobs:
       run: cargo clippy --profile ci --all --all-features -- -D warnings
     - name: Lint Typescript
       run: |
-        npm ci
+        pnpm install --frozen-lockfile
-        npm run lint-ci
+        pnpm lint-ci
     - name: Lint examples
       working-directory: nodejs/examples
-      run: npm ci && npm run lint-ci
+      # The `@lancedb/lancedb` dep points at file:../dist; pnpm errors if
+      # that dir is missing, so create an empty one for lint-only runs.
+      run: mkdir -p ../dist && pnpm install --frozen-lockfile && pnpm lint-ci
   linux:
     name: Linux (NodeJS ${{ matrix.node-version }})
     timeout-minutes: 30
@@ -78,14 +90,18 @@ jobs:
       with:
         fetch-depth: 0
         lfs: true
-    - uses: actions/setup-node@v4
+    - uses: pnpm/action-setup@v4
-      name: Setup Node.js 20 for build
       with:
-        # @napi-rs/cli v3 requires Node >= 20.12 (via @inquirer/prompts@8).
+        version: 11.1.1
-        # Build always on Node 20; tests run on the matrix version below.
+    - uses: actions/setup-node@v4
-        node-version: 20
+      name: Setup Node.js 24 for build
-        cache: 'npm'
+      with:
-        cache-dependency-path: nodejs/package-lock.json
+        # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
+        # in October. Build/install runs on Node 24; tests run on the
+        # matrix version below using direct jest invocation.
+        node-version: 24
+        cache: 'pnpm'
+        cache-dependency-path: nodejs/pnpm-lock.yaml
     - uses: Swatinem/rust-cache@v2
     - name: Install dependencies
       run: |
@@ -93,45 +109,52 @@ jobs:
         sudo apt install -y protobuf-compiler libssl-dev
     - name: Build
       run: |
-        npm ci --include=optional
+        pnpm install --frozen-lockfile
-        npm run build:debug -- --profile ci
+        # No `--` separator: pnpm forwards it literally, which would
+        # make napi-rs treat `--profile ci` as a cargo passthrough arg.
+        pnpm build:debug --profile ci
+        pnpm tsc
+    - name: Setup examples
+      working-directory: nodejs/examples
+      run: pnpm install --frozen-lockfile
+    - name: Check docs
+      run: |
+        # We run this as part of the job because the binary needs to be built
+        # first to export the types of the native code.
+        set -e
+        # `pnpm docs` would invoke pnpm's built-in `docs` command, not
+        # the script — use `pnpm run docs`.
+        pnpm run docs
+        if ! git diff --exit-code -- ../ ':(exclude)Cargo.lock'; then
+          echo "Docs need to be updated"
+          echo "Run 'pnpm run docs', fix any warnings, and commit the changes."
+          exit 1
+        fi
     - uses: actions/setup-node@v4
       name: Setup Node.js ${{ matrix.node-version }} for test
       with:
         node-version: ${{ matrix.node-version }}
-    - name: Compile TypeScript
-      run: npm run tsc
     - name: Setup localstack
       working-directory: .
       run: docker compose up --detach --wait
     - name: Test
       env:
         S3_TEST: "1"
-      run: npm run test
+        # Newer @smithy/core uses dynamic ESM imports.
-    - name: Setup examples
+        NODE_OPTIONS: "--experimental-vm-modules"
-      working-directory: nodejs/examples
+      # Invoke jest directly because pnpm 11 itself requires Node 22+
-      run: npm ci
+      # while the matrix tests on older Node versions.
+      run: npx jest --verbose
     - name: Test examples
       working-directory: ./
       env:
         OPENAI_API_KEY: test
         OPENAI_BASE_URL: http://0.0.0.0:8000
+        NODE_OPTIONS: "--experimental-vm-modules"
       run: |
         python ci/mock_openai.py &
         cd nodejs/examples
-        npm test
+        npx jest --testEnvironment jest-environment-node-single-context --verbose
-    - name: Check docs
-      run: |
-        # We run this as part of the job because the binary needs to be built
-        # first to export the types of the native code.
-        set -e
-        npm ci
-        npm run docs
-        if ! git diff --exit-code -- ../ ':(exclude)Cargo.lock'; then
-          echo "Docs need to be updated"
-          echo "Run 'npm run docs', fix any warnings, and commit the changes."
-          exit 1
-        fi
   macos:
     timeout-minutes: 30
     runs-on: "macos-14"
@@ -144,20 +167,28 @@ jobs:
       with:
         fetch-depth: 0
         lfs: true
+    - uses: pnpm/action-setup@v4
+      with:
+        version: 11.1.1
     - uses: actions/setup-node@v4
       with:
-        node-version: 20
+        # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
-        cache: 'npm'
+        # in October.
-        cache-dependency-path: nodejs/package-lock.json
+        node-version: 24
+        cache: 'pnpm'
+        cache-dependency-path: nodejs/pnpm-lock.yaml
+    - uses: dtolnay/rust-toolchain@stable
     - uses: Swatinem/rust-cache@v2
     - name: Install dependencies
       run: |
         brew install protobuf
     - name: Build
       run: |
-        npm ci --include=optional
+        pnpm install --frozen-lockfile
-        npm run build:debug -- --profile ci
+        # No `--` separator: pnpm forwards it literally, which would
-        npm run tsc
+        # make napi-rs treat `--profile ci` as a cargo passthrough arg.
+        pnpm build:debug --profile ci
+        pnpm tsc
     - name: Test
       run: |
-        npm run test
+        pnpm test

.github/workflows/npm-publish.yml

@@ -171,13 +171,18 @@ jobs:
         working-directory: nodejs
     steps:
       - uses: actions/checkout@v4
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 11.1.1
       - name: Setup node
         uses: actions/setup-node@v4
-        if: ${{ !matrix.settings.docker }}
         with:
-          node-version: 20
+          # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
-          cache: npm
+          # in October.
-          cache-dependency-path: nodejs/package-lock.json
+          node-version: 24
+          cache: pnpm
+          cache-dependency-path: nodejs/pnpm-lock.yaml
       - name: Install
         uses: dtolnay/rust-toolchain@stable
         if: ${{ !matrix.settings.docker }}
@@ -195,7 +200,7 @@ jobs:
             target/
           key: nodejs-${{ matrix.settings.target }}-cargo-${{ matrix.settings.host }}
       - name: Install dependencies
-        run: npm ci
+        run: pnpm install --frozen-lockfile
       - name: Install Zig
         uses: mlugg/setup-zig@v2
         if: ${{ contains(matrix.settings.target, 'musl') }}
@@ -248,7 +253,7 @@ jobs:
       # one to do the upload.
       - name: Make generic artifacts
         if: ${{ matrix.settings.target == 'aarch64-apple-darwin' }}
-        run: npm run tsc
+        run: pnpm tsc
       - name: Upload Generic Artifacts
         if: ${{ matrix.settings.target == 'aarch64-apple-darwin' }}
         uses: actions/upload-artifact@v4
@@ -283,14 +288,24 @@ jobs:
         working-directory: nodejs
     steps:
       - uses: actions/checkout@v4
-      - name: Setup node
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 11.1.1
+      - name: Setup Node.js 24 for install
+        uses: actions/setup-node@v4
+        with:
+          # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
+          # in October.
+          node-version: 24
+          cache: pnpm
+          cache-dependency-path: nodejs/pnpm-lock.yaml
+      - name: Install dependencies
+        run: pnpm install --frozen-lockfile
+      - name: Setup Node.js ${{ matrix.node }} for test
         uses: actions/setup-node@v4
         with:
           node-version: ${{ matrix.node }}
-          cache: npm
-          cache-dependency-path: nodejs/package-lock.json
-      - name: Install dependencies
-        run: npm ci
       - name: Download artifacts
         uses: actions/download-artifact@v4
         with:
@@ -311,7 +326,9 @@ jobs:
       - name: Move built files
         run: cp dist/native.d.ts dist/native.js dist/*.node lancedb/
       - name: Test bindings
-        run: npm test
+        # Invoke jest directly because pnpm 11 itself requires Node 22+
+        # while the matrix tests on older Node versions.
+        run: npx jest --verbose
   publish:
     name: Publish
     runs-on: ubuntu-latest
@@ -323,15 +340,19 @@ jobs:
       - test-lancedb
     steps:
       - uses: actions/checkout@v4
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 11.1.1
       - name: Setup node
         uses: actions/setup-node@v4
         with:
           node-version: 24
-          cache: npm
+          cache: pnpm
-          cache-dependency-path: nodejs/package-lock.json
+          cache-dependency-path: nodejs/pnpm-lock.yaml
           registry-url: "https://registry.npmjs.org"
       - name: Install dependencies
-        run: npm ci
+        run: pnpm install --frozen-lockfile
       - uses: actions/download-artifact@v4
         with:
           name: nodejs-dist
@@ -351,7 +372,7 @@ jobs:
       - name: Display structure of downloaded files
         run: find dist && find nodejs-artifacts
       - name: Move artifacts
-        run: npx napi artifacts -d nodejs-artifacts
+        run: pnpm exec napi artifacts -d nodejs-artifacts
       - name: List packages
         run: find npm
       - name: Publish

.github/workflows/pypi-publish.yml

@@ -14,10 +14,16 @@ on:
 env:
   PIP_EXTRA_INDEX_URL: "https://pypi.fury.io/lance-format/ https://pypi.fury.io/lancedb/"
 
+permissions:
+  contents: read
+
 jobs:
   linux:
     name: Python ${{ matrix.config.platform }} manylinux${{ matrix.config.manylinux }}
     timeout-minutes: 60
+    permissions:
+      id-token: write
+      contents: read
     strategy:
       matrix:
         config:
@@ -57,10 +63,12 @@ jobs:
       - uses: ./.github/workflows/upload_wheel
         if: startsWith(github.ref, 'refs/tags/python-v')
         with:
-          pypi_token: ${{ secrets.LANCEDB_PYPI_API_TOKEN }}
           fury_token: ${{ secrets.FURY_TOKEN }}
   mac:
     timeout-minutes: 90
+    permissions:
+      id-token: write
+      contents: read
     runs-on: ${{ matrix.config.runner }}
     strategy:
       matrix:
@@ -85,10 +93,12 @@ jobs:
       - uses: ./.github/workflows/upload_wheel
         if: startsWith(github.ref, 'refs/tags/python-v')
         with:
-          pypi_token: ${{ secrets.LANCEDB_PYPI_API_TOKEN }}
           fury_token: ${{ secrets.FURY_TOKEN }}
   windows:
     timeout-minutes: 60
+    permissions:
+      id-token: write
+      contents: read
     runs-on: windows-latest
     steps:
       - uses: actions/checkout@v4
@@ -107,7 +117,6 @@ jobs:
       - uses: ./.github/workflows/upload_wheel
         if: startsWith(github.ref, 'refs/tags/python-v')
         with:
-          pypi_token: ${{ secrets.LANCEDB_PYPI_API_TOKEN }}
           fury_token: ${{ secrets.FURY_TOKEN }}
   gh-release:
     if: startsWith(github.ref, 'refs/tags/python-v')

.github/workflows/python.yml

@@ -8,6 +8,7 @@ on:
     paths:
       - Cargo.toml
       - Cargo.lock
+      - rust-toolchain.toml
       - python/**
       - rust/**
       - .github/workflows/python.yml
@@ -16,6 +17,9 @@ on:
       - .github/workflows/build_windows_wheel/**
       - .github/workflows/run_tests/**
 
+permissions:
+  contents: read
+
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
   cancel-in-progress: true
@@ -107,7 +111,6 @@ jobs:
       - name: Install
         run: |
           pip install --extra-index-url https://pypi.fury.io/lance-format/ --extra-index-url https://pypi.fury.io/lancedb/ -e .[tests,dev,embeddings]
-          pip install tantivy
           pip install mlx
       - name: Doctest
         run: pytest --doctest-modules python/lancedb
@@ -226,6 +229,5 @@ jobs:
           pip install "pydantic<2"
           pip install pyarrow==16
           pip install --extra-index-url https://pypi.fury.io/lance-format/ --extra-index-url https://pypi.fury.io/lancedb/ -e .[tests]
-          pip install tantivy
       - name: Run tests
         run: pytest -m "not slow and not s3_test" -x -v --durations=30 python/tests

.github/workflows/rust.yml

@@ -8,9 +8,16 @@ on:
     paths:
       - Cargo.toml
       - Cargo.lock
+      - rust-toolchain.toml
+      - deny.toml
       - rust/**
+      - nodejs/Cargo.toml
+      - python/Cargo.toml
       - .github/workflows/rust.yml
 
+permissions:
+  contents: read
+
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
   cancel-in-progress: true
@@ -52,6 +59,17 @@ jobs:
       - name: Run clippy (without remote feature)
         run: cargo clippy --profile ci --workspace --tests -- -D warnings
 
+  deny:
+    # Supply-chain checks: advisories, licenses, banned crates, and source
+    # restrictions. Configuration lives in `deny.toml` at the workspace root.
+    timeout-minutes: 10
+    runs-on: ubuntu-24.04
+    steps:
+      - uses: actions/checkout@v4
+      - uses: EmbarkStudios/cargo-deny-action@v2
+        with:
+          command: check advisories bans licenses sources
+
   build-no-lock:
     runs-on: ubuntu-24.04
     timeout-minutes: 30

.github/workflows/update_package_lock_run.yml

@@ -3,6 +3,9 @@ name: Update package-lock.json
 on:
   workflow_dispatch:
 
+permissions:
+  contents: read
+
 jobs:
   publish:
     runs-on: ubuntu-latest

.github/workflows/update_package_lock_run_nodejs.yml

@@ -3,6 +3,9 @@ name: Update NodeJs package-lock.json
 on:
   workflow_dispatch:
 
+permissions:
+  contents: read
+
 jobs:
   publish:
     runs-on: ubuntu-latest

.github/workflows/upload_wheel/action.yml

@@ -2,9 +2,6 @@ name: upload-wheel
 
 description: "Upload wheels to Pypi"
 inputs:
-  pypi_token:
-    required: true
-    description: "release token for the repo"
   fury_token:
     required: true
     description: "release token for the fury repo"
@@ -12,12 +9,6 @@ inputs:
 runs:
   using: "composite"
   steps:
-  - name: Install dependencies
-    shell: bash
-    run: |
-      python -m pip install --upgrade pip
-      pip install twine
-      python3 -m pip install --upgrade pkginfo
   - name: Choose repo
     shell: bash
     id: choose_repo
@@ -27,19 +18,17 @@ runs:
       else
         echo "repo=pypi" >> $GITHUB_OUTPUT
       fi
-  - name: Publish to PyPI
+  - name: Publish to Fury
+    if: steps.choose_repo.outputs.repo == 'fury'
     shell: bash
     env:
       FURY_TOKEN: ${{ inputs.fury_token }}
-      PYPI_TOKEN: ${{ inputs.pypi_token }}
     run: |
-      if [[ ${{ steps.choose_repo.outputs.repo }} == fury ]]; then
+      WHEEL=$(ls target/wheels/lancedb-*.whl 2> /dev/null | head -n 1)
-        WHEEL=$(ls target/wheels/lancedb-*.whl 2> /dev/null | head -n 1)
+      echo "Uploading $WHEEL to Fury"
-        echo "Uploading $WHEEL to Fury"
+      curl -f -F package=@$WHEEL https://$FURY_TOKEN@push.fury.io/lancedb/
-        curl -f -F package=@$WHEEL https://$FURY_TOKEN@push.fury.io/lancedb/
+  - name: Publish to PyPI
-      else
+    if: steps.choose_repo.outputs.repo == 'pypi'
-        twine upload --repository ${{ steps.choose_repo.outputs.repo }} \
+    uses: pypa/gh-action-pypi-publish@release/v1
-          --username __token__ \
+    with:
-          --password $PYPI_TOKEN \
+      packages-dir: target/wheels/
-          target/wheels/lancedb-*.whl
-      fi

Cargo.toml

@@ -1,7 +1,5 @@
 [workspace]
 members = ["rust/lancedb", "nodejs", "python"]
-# Python package needs to be built by maturin.
-exclude = ["python"]
 resolver = "2"
 
 [workspace.package]
@@ -15,40 +13,40 @@ categories = ["database-implementations"]
 rust-version = "1.91.0"
 
 [workspace.dependencies]
-lance = { version = "=4.0.0", default-features = false }
+lance = { "version" = "=7.0.0-beta.7", default-features = false, "tag" = "v7.0.0-beta.7", "git" = "https://github.com/lance-format/lance.git" }
-lance-core = { version = "=4.0.0" }
+lance-core = { "version" = "=7.0.0-beta.7", "tag" = "v7.0.0-beta.7", "git" = "https://github.com/lance-format/lance.git" }
-lance-datagen = { version = "=4.0.0" }
+lance-datagen = { "version" = "=7.0.0-beta.7", "tag" = "v7.0.0-beta.7", "git" = "https://github.com/lance-format/lance.git" }
-lance-file = { version = "=4.0.0" }
+lance-file = { "version" = "=7.0.0-beta.7", "tag" = "v7.0.0-beta.7", "git" = "https://github.com/lance-format/lance.git" }
-lance-io = { version = "=4.0.0", default-features = false }
+lance-io = { "version" = "=7.0.0-beta.7", default-features = false, "tag" = "v7.0.0-beta.7", "git" = "https://github.com/lance-format/lance.git" }
-lance-index = { version = "=4.0.0" }
+lance-index = { "version" = "=7.0.0-beta.7", "tag" = "v7.0.0-beta.7", "git" = "https://github.com/lance-format/lance.git" }
-lance-linalg = { version = "=4.0.0" }
+lance-linalg = { "version" = "=7.0.0-beta.7", "tag" = "v7.0.0-beta.7", "git" = "https://github.com/lance-format/lance.git" }
-lance-namespace = { version = "=4.0.0" }
+lance-namespace = { "version" = "=7.0.0-beta.7", "tag" = "v7.0.0-beta.7", "git" = "https://github.com/lance-format/lance.git" }
-lance-namespace-impls = { version = "=4.0.0", default-features = false }
+lance-namespace-impls = { "version" = "=7.0.0-beta.7", default-features = false, "tag" = "v7.0.0-beta.7", "git" = "https://github.com/lance-format/lance.git" }
-lance-table = { version = "=4.0.0" }
+lance-table = { "version" = "=7.0.0-beta.7", "tag" = "v7.0.0-beta.7", "git" = "https://github.com/lance-format/lance.git" }
-lance-testing = { version = "=4.0.0" }
+lance-testing = { "version" = "=7.0.0-beta.7", "tag" = "v7.0.0-beta.7", "git" = "https://github.com/lance-format/lance.git" }
-lance-datafusion = { version = "=4.0.0" }
+lance-datafusion = { "version" = "=7.0.0-beta.7", "tag" = "v7.0.0-beta.7", "git" = "https://github.com/lance-format/lance.git" }
-lance-encoding = { version = "=4.0.0" }
+lance-encoding = { "version" = "=7.0.0-beta.7", "tag" = "v7.0.0-beta.7", "git" = "https://github.com/lance-format/lance.git" }
-lance-arrow = { version = "=4.0.0" }
+lance-arrow = { "version" = "=7.0.0-beta.7", "tag" = "v7.0.0-beta.7", "git" = "https://github.com/lance-format/lance.git" }
 ahash = "0.8"
 # Note that this one does not include pyarrow
-arrow = { version = "57.2", optional = false }
+arrow = { version = "58.0.0", optional = false }
-arrow-array = "57.2"
+arrow-array = "58.0.0"
-arrow-data = "57.2"
+arrow-data = "58.0.0"
-arrow-ipc = "57.2"
+arrow-ipc = "58.0.0"
-arrow-ord = "57.2"
+arrow-ord = "58.0.0"
-arrow-schema = "57.2"
+arrow-schema = "58.0.0"
-arrow-select = "57.2"
+arrow-select = "58.0.0"
-arrow-cast = "57.2"
+arrow-cast = "58.0.0"
 async-trait = "0"
-datafusion = { version = "52.1", default-features = false }
+datafusion = { version = "53.0.0", default-features = false }
-datafusion-catalog = "52.1"
+datafusion-catalog = "53.0.0"
-datafusion-common = { version = "52.1", default-features = false }
+datafusion-common = { version = "53.0.0", default-features = false }
-datafusion-execution = "52.1"
+datafusion-execution = "53.0.0"
-datafusion-expr = "52.1"
+datafusion-expr = "53.0.0"
-datafusion-functions = "52.1"
+datafusion-functions = "53.0.0"
-datafusion-physical-plan = "52.1"
+datafusion-physical-plan = "53.0.0"
-datafusion-physical-expr = "52.1"
+datafusion-physical-expr = "53.0.0"
-datafusion-sql = "52.1"
+datafusion-sql = "53.0.0"
 env_logger = "0.11"
 half = { "version" = "2.7.1", default-features = false, features = [
     "num-traits",
@@ -56,7 +54,7 @@ half = { "version" = "2.7.1", default-features = false, features = [
 futures = "0"
 log = "0.4"
 moka = { version = "0.12", features = ["future"] }
-object_store = "0.12.0"
+object_store = "0.13.2"
 pin-project = "1.0.7"
 rand = "0.9"
 snafu = "0.8"

README.md

@@ -15,7 +15,7 @@
 
 # **The Multimodal AI Lakehouse**
 
-[**How to Install** ](#how-to-install) ✦ [**Detailed Documentation**](https://lancedb.com/docs) ✦ [**Tutorials and Recipes**](https://github.com/lancedb/vectordb-recipes/tree/main) ✦  [**Contributors**](#contributors) 
+[**How to Install** ](#how-to-install) ✦ [**Detailed Documentation**](https://docs.lancedb.com) ✦ [**Tutorials and Recipes**](https://github.com/lancedb/vectordb-recipes/tree/main) ✦  [**Contributors**](#contributors) 
 
 **The ultimate multimodal data platform for AI/ML applications.** 
 
@@ -57,7 +57,7 @@ LanceDB is a central location where developers can build, train and analyze thei
 
 ## **How to Install**:
 
-Follow the [Quickstart](https://lancedb.com/docs/quickstart/) doc to set up LanceDB locally. 
+Follow the [Quickstart](https://docs.lancedb.com/quickstart) doc to set up LanceDB locally. 
 
 **API & SDK:** We also support Python, Typescript and Rust SDKs
 

deny.toml

@@ -0,0 +1,196 @@
+# cargo-deny configuration for LanceDB.
+#
+# Run locally with `cargo deny check`. See
+# https://embarkstudios.github.io/cargo-deny/ for the full reference.
+
+# The set of target triples we care about. cargo-deny will only consider
+# dependencies that are used on at least one of these targets. Keeping this
+# explicit avoids noise from platform-specific crates (e.g. wasm, android,
+# ios) that we never actually ship.
+[graph]
+targets = [
+    "x86_64-unknown-linux-gnu",
+    "aarch64-unknown-linux-gnu",
+    "x86_64-apple-darwin",
+    "aarch64-apple-darwin",
+    "x86_64-pc-windows-msvc",
+    "aarch64-pc-windows-msvc",
+]
+all-features = true
+
+[output]
+feature-depth = 1
+
+# ---------------------------------------------------------------------------
+# Advisories: security vulnerabilities and yanked crates.
+# ---------------------------------------------------------------------------
+[advisories]
+version = 2
+# Fail the check if any crate in the lockfile has been yanked from crates.io.
+# Yanked crates are a signal the author retracted the release (often due to
+# bugs or security issues) and should not be depended on.
+yanked = "deny"
+# Advisory IDs we have explicitly reviewed and chosen to accept. Every
+# entry must include a rationale and, where possible, an upstream issue
+# pointing to a fix. Revisit this list whenever dependencies are updated.
+ignore = [
+    # rsa: Marvin Attack timing side-channel in PKCS#1 v1.5 decryption.
+    # Reached only through opendal → reqsign → rsa. We do not use RSA
+    # decryption in LanceDB ourselves; this is dormant in the signing path.
+    # No fixed release exists upstream as of this writing.
+    # https://rustsec.org/advisories/RUSTSEC-2023-0071
+    { id = "RUSTSEC-2023-0071", reason = "rsa crate via opendal/reqsign; no fixed upstream release" },
+
+    # instant: unmaintained. Pulled in via backoff → instant. Upstream
+    # recommends switching to `web-time`; fix has to come from backoff.
+    # https://rustsec.org/advisories/RUSTSEC-2024-0384
+    { id = "RUSTSEC-2024-0384", reason = "transitive via backoff; waiting on backoff replacement" },
+
+    # paste: unmaintained (author archived the repo). Used transitively by
+    # datafusion and the arrow ecosystem; widespread, no drop-in replacement.
+    # https://rustsec.org/advisories/RUSTSEC-2024-0436
+    { id = "RUSTSEC-2024-0436", reason = "transitive via datafusion; awaiting ecosystem migration" },
+
+    # encoding: unmaintained. Reached through lindera-dictionary, which is
+    # required by the native Lindera tokenizer path. Lindera has not migrated
+    # off this crate yet.
+    # https://rustsec.org/advisories/RUSTSEC-2021-0153
+    { id = "RUSTSEC-2021-0153", reason = "transitive via lindera-dictionary for native Lindera tokenizer" },
+
+    # fast-float: unsound and unmaintained. Reached only through polars-arrow
+    # from the optional Polars integration; replacement requires a Polars
+    # dependency upgrade.
+    # https://rustsec.org/advisories/RUSTSEC-2024-0379
+    { id = "RUSTSEC-2024-0379", reason = "transitive via polars-arrow; waiting on Polars migration" },
+
+    # tantivy: segfault on malformed input due to missing bounds check.
+    # Pulled in via lance for full-text search. We only feed tantivy
+    # documents we construct ourselves, not attacker-controlled bytes.
+    # Tracked for a lance dependency bump.
+    # https://rustsec.org/advisories/RUSTSEC-2025-0003
+    { id = "RUSTSEC-2025-0003", reason = "tantivy via lance; inputs are internally produced, not user-supplied bytes" },
+
+    # backoff: unmaintained. Reached only via async-openai. Replacement
+    # requires async-openai to migrate (or us to drop async-openai).
+    # https://rustsec.org/advisories/RUSTSEC-2025-0012
+    { id = "RUSTSEC-2025-0012", reason = "transitive via async-openai; waiting on upstream migration" },
+
+    # number_prefix: unmaintained. Transitive via indicatif → hf-hub.
+    # No security impact, just maintenance status.
+    # https://rustsec.org/advisories/RUSTSEC-2025-0119
+    { id = "RUSTSEC-2025-0119", reason = "transitive via hf-hub/indicatif; cosmetic formatting crate" },
+
+    # bincode: unmaintained. Reached through lindera and lindera-dictionary,
+    # which are required by the native Lindera tokenizer path. Lindera has not
+    # migrated to another serialization format yet.
+    # https://rustsec.org/advisories/RUSTSEC-2025-0141
+    { id = "RUSTSEC-2025-0141", reason = "transitive via lindera/lindera-dictionary for native Lindera tokenizer" },
+
+    # lru: soundness issue in IterMut. Reached only through aws-sdk-s3 in
+    # LanceDB's dev-dependency graph; LanceDB does not use that iterator
+    # directly. Clearing this requires the AWS SDK chain to update lru.
+    # https://rustsec.org/advisories/RUSTSEC-2026-0002
+    { id = "RUSTSEC-2026-0002", reason = "transitive via aws-sdk-s3 dev-dependency; waiting on AWS SDK lru upgrade" },
+
+    # rustls-webpki 0.101.7 (old major line): name-constraint checks for
+    # URI / wildcard names. Pulled in only via the legacy rustls 0.21 chain
+    # from aws-smithy-http-client. The 0.103 line we actively use is patched.
+    # Clearing the 0.101 copy requires the aws-sdk chain to migrate off
+    # rustls 0.21.
+    # https://rustsec.org/advisories/RUSTSEC-2026-0098
+    # https://rustsec.org/advisories/RUSTSEC-2026-0099
+    { id = "RUSTSEC-2026-0098", reason = "only affects rustls-webpki 0.101 from legacy aws-smithy/rustls 0.21 chain" },
+    { id = "RUSTSEC-2026-0099", reason = "only affects rustls-webpki 0.101 from legacy aws-smithy/rustls 0.21 chain" },
+
+    # rustls-webpki 0.101.7: reachable panic in CRL parsing. Same legacy
+    # rustls 0.21 chain from aws-smithy-http-client as above. The 0.103 line
+    # we actively use is upgraded to 0.103.13 which contains the fix.
+    # https://rustsec.org/advisories/RUSTSEC-2026-0104
+    { id = "RUSTSEC-2026-0104", reason = "only affects rustls-webpki 0.101 from legacy aws-smithy/rustls 0.21 chain" },
+
+    # rand 0.8.5: soundness issue only when ThreadRng reseeds inside a custom
+    # logger. Reached through several transitive chains. LanceDB does not use
+    # rand from a custom logger; upgrade once all pinned chains accept 0.8.6+.
+    # https://rustsec.org/advisories/RUSTSEC-2026-0097
+    { id = "RUSTSEC-2026-0097", reason = "transitive rand 0.8.5; LanceDB does not call ThreadRng from custom logging" },
+]
+
+# ---------------------------------------------------------------------------
+# Licenses: only allow licenses we've reviewed as compatible with Apache-2.0.
+# ---------------------------------------------------------------------------
+[licenses]
+version = 2
+# SPDX identifiers for licenses that are compatible with our Apache-2.0
+# distribution. Additions require legal review.
+allow = [
+    "Apache-2.0",
+    "Apache-2.0 WITH LLVM-exception",
+    "MIT",
+    "BSD-2-Clause",
+    "BSD-3-Clause",
+    "ISC",
+    "Unicode-3.0",
+    "Unicode-DFS-2016",
+    "Zlib",
+    "CC0-1.0",
+    "MPL-2.0",
+    "BSL-1.0",
+    "OpenSSL",
+    # 0BSD ("BSD Zero Clause") is effectively public domain — no attribution
+    # required. Pulled in by `mock_instant`.
+    "0BSD",
+    # bzip2-1.0.6 is the permissive upstream bzip2 license (BSD-like). Pulled
+    # in by `libbz2-rs-sys`, the pure-Rust bzip2 implementation.
+    "bzip2-1.0.6",
+    # CDLA-Permissive-2.0 is a permissive data license used by `webpki-roots`
+    # for the Mozilla CA root bundle. Data-only, distribution-compatible.
+    "CDLA-Permissive-2.0",
+]
+confidence-threshold = 0.8
+# Crates whose license cannot be determined from Cargo metadata but whose
+# license we've manually confirmed from upstream. Keep this list minimal.
+[[licenses.clarify]]
+# polars-arrow-format omits the `license` field in its Cargo.toml, but the
+# upstream repo (pola-rs/polars-arrow-format) is dual-licensed Apache-2.0 OR
+# MIT. See https://github.com/pola-rs/polars-arrow-format/blob/main/LICENSE
+crate = "polars-arrow-format"
+expression = "Apache-2.0 OR MIT"
+license-files = []
+
+# ---------------------------------------------------------------------------
+# Bans: disallow specific crates and flag dependency hygiene issues.
+# ---------------------------------------------------------------------------
+[bans]
+# Warn (not deny) on duplicate versions of the same crate. In a large
+# workspace like this one, duplicates are common and often unavoidable
+# transitively. We surface them to discourage growth, but don't fail CI.
+multiple-versions = "warn"
+# Wildcard version requirements (`foo = "*"`) are a footgun — they let any
+# future release in without review. Ban them outright.
+wildcards = "deny"
+# Internal workspace crates reference each other via `path = "..."`, which
+# cargo-deny sees as a wildcard version. That's fine for private workspace
+# members (not published to crates.io), so allow it specifically for paths.
+allow-wildcard-paths = true
+# Features that, if enabled, should cause the check to fail.
+deny = []
+# Crates to skip when checking for duplicate versions.
+skip = []
+# Similar to `skip`, but also skips the entire transitive subtree.
+skip-tree = []
+
+# ---------------------------------------------------------------------------
+# Sources: restrict where crates can come from.
+# ---------------------------------------------------------------------------
+[sources]
+# Deny any registry other than the ones explicitly listed below.
+unknown-registry = "deny"
+# Deny any git dependency whose host isn't in the allow-list below. This
+# prevents accidental pulls from arbitrary forks.
+unknown-git = "deny"
+allow-registry = ["https://github.com/rust-lang/crates.io-index"]
+# Lance is developed in a sibling repo and pulled as a git dependency until
+# releases are cut to crates.io. Allow that specific host.
+allow-git = [
+    "https://github.com/lance-format/lance",
+]

dockerfiles/Dockerfile

@@ -24,4 +24,4 @@ RUN python --version && \
   rustc --version && \
   protoc --version
 
-RUN pip install --no-cache-dir tantivy lancedb
+RUN pip install --no-cache-dir lancedb

docs/README.md

@@ -1,6 +1,6 @@
 # LanceDB Documentation
 
-LanceDB docs are available at [lancedb.com/docs](https://lancedb.com/docs).
+LanceDB docs are available at [docs.lancedb.com](https://docs.lancedb.com).
 
 The SDK docs are built and deployed automatically by [Github Actions](../.github/workflows/docs.yml)
 whenever a commit is pushed to the `main` branch. So it is possible for the docs to show

docs/src/java/java.md

@@ -14,7 +14,7 @@ Add the following dependency to your `pom.xml`:
 <dependency>
     <groupId>com.lancedb</groupId>
     <artifactId>lancedb-core</artifactId>
-    <version>0.27.2-beta.1</version>
+    <version>0.28.0-beta.11</version>
 </dependency>
 ```
 
@@ -57,32 +57,32 @@ LanceNamespace namespaceClient = LanceDbNamespaceClientBuilder.newBuilder()
 
 ## Metadata Operations
 
-### Creating a Namespace
+### Creating a Namespace Path
 
-Namespaces organize tables hierarchically. Create a namespace before creating tables within it:
+Namespace paths organize tables hierarchically. Create the desired namespace path before creating tables within it:
 
 ```java
 import org.lance.namespace.model.CreateNamespaceRequest;
 import org.lance.namespace.model.CreateNamespaceResponse;
 
-// Create a child namespace
+// Create a child namespace path
 CreateNamespaceRequest request = new CreateNamespaceRequest();
 request.setId(Arrays.asList("my_namespace"));
 
 CreateNamespaceResponse response = namespaceClient.createNamespace(request);
 ```
 
-You can also create nested namespaces:
+You can also create nested namespace paths:
 
 ```java
-// Create a nested namespace: parent/child
+// Create a nested namespace path: parent/child
 CreateNamespaceRequest request = new CreateNamespaceRequest();
 request.setId(Arrays.asList("parent_namespace", "child_namespace"));
 
 CreateNamespaceResponse response = namespaceClient.createNamespace(request);
 ```
 
-### Describing a Namespace
+### Describing a Namespace Path
 
 ```java
 import org.lance.namespace.model.DescribeNamespaceRequest;
@@ -95,22 +95,22 @@ DescribeNamespaceResponse response = namespaceClient.describeNamespace(request);
 System.out.println("Namespace properties: " + response.getProperties());
 ```
 
-### Listing Namespaces
+### Listing Namespace Paths
 
 ```java
 import org.lance.namespace.model.ListNamespacesRequest;
 import org.lance.namespace.model.ListNamespacesResponse;
 
-// List all namespaces at root level
+// List all namespace paths at the root level
 ListNamespacesRequest request = new ListNamespacesRequest();
 request.setId(Arrays.asList());  // Empty for root
 
 ListNamespacesResponse response = namespaceClient.listNamespaces(request);
 for (String ns : response.getNamespaces()) {
-    System.out.println("Namespace: " + ns);
+    System.out.println("Namespace path: " + ns);
 }
 
-// List child namespaces under a parent
+// List child namespace paths under a parent path
 ListNamespacesRequest childRequest = new ListNamespacesRequest();
 childRequest.setId(Arrays.asList("parent_namespace"));
 
@@ -123,7 +123,7 @@ ListNamespacesResponse childResponse = namespaceClient.listNamespaces(childReque
 import org.lance.namespace.model.ListTablesRequest;
 import org.lance.namespace.model.ListTablesResponse;
 
-// List tables in a namespace
+// List tables in a namespace path
 ListTablesRequest request = new ListTablesRequest();
 request.setId(Arrays.asList("my_namespace"));
 
@@ -133,7 +133,7 @@ for (String table : response.getTables()) {
 }
 ```
 
-### Dropping a Namespace
+### Dropping a Namespace Path
 
 ```java
 import org.lance.namespace.model.DropNamespaceRequest;
@@ -175,7 +175,7 @@ DropTableResponse response = namespaceClient.dropTable(request);
 
 ### Creating a Table
 
-Tables are created within a namespace by providing data in Apache Arrow IPC format:
+Tables are created within a namespace path by providing data in Apache Arrow IPC format:
 
 ```java
 import org.lance.namespace.LanceNamespace;
@@ -242,7 +242,7 @@ try (BufferAllocator allocator = new RootAllocator();
     }
     byte[] tableData = out.toByteArray();
 
-    // Create table in a namespace
+    // Create a table in a namespace path
     CreateTableRequest request = new CreateTableRequest();
     request.setId(Arrays.asList("my_namespace", "my_table"));
     CreateTableResponse response = namespaceClient.createTable(request, tableData);

docs/src/js/README.md

@@ -34,7 +34,7 @@ const results = await table.vectorSearch([0.1, 0.3]).limit(20).toArray();
 console.log(results);
 ```
 
-The [quickstart](https://lancedb.com/docs/quickstart/basic-usage/) contains more complete examples.
+The [quickstart](https://docs.lancedb.com/quickstart/) contains more complete examples.
 
 ## Development
 

docs/src/js/_media/CONTRIBUTING.md

@@ -12,20 +12,22 @@ Typescript.
 * `src/`: Rust bindings source code
 * `lancedb/`: Typescript package source code
 * `__test__/`: Unit tests
-* `examples/`: An npm package with the examples shown in the documentation
+* `examples/`: A pnpm package with the examples shown in the documentation
 
 ## Development environment
 
 To set up your development environment, you will need to install the following:
 
-1. Node.js 14 or later
+1. Node.js 22 or later (required by pnpm 11)
-2. Rust's package manager, Cargo. Use [rustup](https://rustup.rs/) to install.
+2. [pnpm](https://pnpm.io/installation) 11 or later (or run via `corepack enable`,
-3. [protoc](https://grpc.io/docs/protoc-installation/) (Protocol Buffers compiler)
+   which uses the `packageManager` field in `package.json`)
+3. Rust's package manager, Cargo. Use [rustup](https://rustup.rs/) to install.
+4. [protoc](https://grpc.io/docs/protoc-installation/) (Protocol Buffers compiler)
 
 Initial setup:
 
 ```shell
-npm install
+pnpm install
 ```
 
 ### Commit Hooks
@@ -39,38 +41,38 @@ pre-commit install
 
 ## Development
 
-Most common development commands can be run using the npm scripts.
+Most common development commands can be run using the pnpm scripts.
 
 Build the package
 
 ```shell
-npm install
+pnpm install
-npm run build
+pnpm build
 ```
 
 Lint:
 
 ```shell
-npm run lint
+pnpm lint
 ```
 
 Format and fix lints:
 
 ```shell
-npm run lint-fix
+pnpm lint-fix
 ```
 
 Run tests:
 
 ```shell
-npm test
+pnpm test
 ```
 
 To run a single test:
 
 ```shell
 # Single file: table.test.ts
-npm test -- table.test.ts
+pnpm test -- table.test.ts
 # Single test: 'merge insert' in table.test.ts
-npm test -- table.test.ts --testNamePattern=merge\ insert
+pnpm test -- table.test.ts --testNamePattern=merge\ insert
 ```

docs/src/js/classes/Connection.md

@@ -61,8 +61,8 @@ sharing the same data, deletion, and index files.
 * **options.sourceVersion?**: `number`
     The version of the source table to clone.
 
-* **options.targetNamespace?**: `string`[]
+* **options.targetNamespacePath?**: `string`[]
-    The namespace for the target table (defaults to root namespace).
+    The namespace path for the target table (defaults to root namespace).
 
 #### Returns
 
@@ -116,13 +116,13 @@ Creates a new empty Table
 
 `Promise`<[`Table`](Table.md)>
 
-#### createEmptyTable(name, schema, namespace, options)
+#### createEmptyTable(name, schema, namespacePath, options)
 
 ```ts
 abstract createEmptyTable(
    name,
    schema,
-   namespace?,
+   namespacePath?,
    options?): Promise<Table>
 ```
 
@@ -136,8 +136,8 @@ Creates a new empty Table
 * **schema**: [`SchemaLike`](../type-aliases/SchemaLike.md)
     The schema of the table
 
-* **namespace?**: `string`[]
+* **namespacePath?**: `string`[]
-    The namespace to create the table in (defaults to root namespace)
+    The namespace path to create the table in (defaults to root namespace)
 
 * **options?**: `Partial`&lt;[`CreateTableOptions`](../interfaces/CreateTableOptions.md)&gt;
     Additional options
@@ -148,12 +148,39 @@ Creates a new empty Table
 
 ***
 
-### createTable()
+### createNamespace()
-
-#### createTable(options, namespace)
 
 ```ts
-abstract createTable(options, namespace?): Promise<Table>
+abstract createNamespace(namespacePath, options?): Promise<CreateNamespaceResponse>
+```
+
+Create a new namespace at the given path.
+
+#### Parameters
+
+* **namespacePath**: `string`[]
+    The namespace path to create.
+
+* **options?**: `Partial`&lt;[`CreateNamespaceOptions`](../interfaces/CreateNamespaceOptions.md)&gt;
+    Creation `mode`
+    ("create" | "exist_ok" | "overwrite") and optional `properties`
+    to attach to the namespace.
+
+#### Returns
+
+`Promise`&lt;[`CreateNamespaceResponse`](../interfaces/CreateNamespaceResponse.md)&gt;
+
+The properties of the
+  created namespace and an optional transaction id.
+
+***
+
+### createTable()
+
+#### createTable(options, namespacePath)
+
+```ts
+abstract createTable(options, namespacePath?): Promise<Table>
 ```
 
 Creates a new Table and initialize it with new data.
@@ -163,8 +190,8 @@ Creates a new Table and initialize it with new data.
 * **options**: `object` & `Partial`&lt;[`CreateTableOptions`](../interfaces/CreateTableOptions.md)&gt;
     The options object.
 
-* **namespace?**: `string`[]
+* **namespacePath?**: `string`[]
-    The namespace to create the table in (defaults to root namespace)
+    The namespace path to create the table in (defaults to root namespace)
 
 ##### Returns
 
@@ -197,13 +224,13 @@ Creates a new Table and initialize it with new data.
 
 `Promise`&lt;[`Table`](Table.md)&gt;
 
-#### createTable(name, data, namespace, options)
+#### createTable(name, data, namespacePath, options)
 
 ```ts
 abstract createTable(
    name,
    data,
-   namespace?,
+   namespacePath?,
    options?): Promise<Table>
 ```
 
@@ -218,8 +245,8 @@ Creates a new Table and initialize it with new data.
     Non-empty Array of Records
     to be inserted into the table
 
-* **namespace?**: `string`[]
+* **namespacePath?**: `string`[]
-    The namespace to create the table in (defaults to root namespace)
+    The namespace path to create the table in (defaults to root namespace)
 
 * **options?**: `Partial`&lt;[`CreateTableOptions`](../interfaces/CreateTableOptions.md)&gt;
     Additional options
@@ -230,6 +257,29 @@ Creates a new Table and initialize it with new data.
 
 ***
 
+### describeNamespace()
+
+```ts
+abstract describeNamespace(namespacePath): Promise<DescribeNamespaceResponse>
+```
+
+Describe a namespace, returning its properties.
+
+#### Parameters
+
+* **namespacePath**: `string`[]
+    The namespace path to describe, in
+    parent → child order, e.g. `["analytics", "sales"]`.
+
+#### Returns
+
+`Promise`&lt;[`DescribeNamespaceResponse`](../interfaces/DescribeNamespaceResponse.md)&gt;
+
+The namespace's properties
+  (may be undefined if the namespace has none).
+
+***
+
 ### display()
 
 ```ts
@@ -247,15 +297,15 @@ Return a brief description of the connection
 ### dropAllTables()
 
 ```ts
-abstract dropAllTables(namespace?): Promise<void>
+abstract dropAllTables(namespacePath?): Promise<void>
 ```
 
 Drop all tables in the database.
 
 #### Parameters
 
-* **namespace?**: `string`[]
+* **namespacePath?**: `string`[]
-    The namespace to drop tables from (defaults to root namespace).
+    The namespace path to drop tables from (defaults to root namespace).
 
 #### Returns
 
@@ -263,10 +313,40 @@ Drop all tables in the database.
 
 ***
 
+### dropNamespace()
+
+```ts
+abstract dropNamespace(namespacePath, options?): Promise<DropNamespaceResponse>
+```
+
+Drop a namespace.
+
+Use `behavior: "cascade"` to also drop everything contained in the
+namespace (sub-namespaces and tables). The default `"restrict"`
+behavior refuses to drop a non-empty namespace.
+
+#### Parameters
+
+* **namespacePath**: `string`[]
+    The namespace path to drop.
+
+* **options?**: `Partial`&lt;[`DropNamespaceOptions`](../interfaces/DropNamespaceOptions.md)&gt;
+    `mode` ("skip" | "fail"
+    for missing-namespace handling) and `behavior` ("restrict" | "cascade").
+
+#### Returns
+
+`Promise`&lt;[`DropNamespaceResponse`](../interfaces/DropNamespaceResponse.md)&gt;
+
+Any properties returned by
+  the server and an optional transaction id.
+
+***
+
 ### dropTable()
 
 ```ts
-abstract dropTable(name, namespace?): Promise<void>
+abstract dropTable(name, namespacePath?): Promise<void>
 ```
 
 Drop an existing table.
@@ -276,8 +356,8 @@ Drop an existing table.
 * **name**: `string`
     The name of the table to drop.
 
-* **namespace?**: `string`[]
+* **namespacePath?**: `string`[]
-    The namespace of the table (defaults to root namespace).
+    The namespace path of the table (defaults to root namespace).
 
 #### Returns
 
@@ -299,12 +379,42 @@ Return true if the connection has not been closed
 
 ***
 
+### listNamespaces()
+
+```ts
+abstract listNamespaces(namespacePath?, options?): Promise<ListNamespacesResponse>
+```
+
+List the immediate child namespaces under the given parent.
+
+Results may be paginated. To retrieve subsequent pages, pass the
+`pageToken` returned by a previous call.
+
+#### Parameters
+
+* **namespacePath?**: `string`[]
+    The parent namespace path. Defaults
+    to the root namespace if omitted.
+
+* **options?**: `Partial`&lt;[`ListNamespacesOptions`](../interfaces/ListNamespacesOptions.md)&gt;
+    Pagination options
+    (`pageToken`, `limit`).
+
+#### Returns
+
+`Promise`&lt;[`ListNamespacesResponse`](../interfaces/ListNamespacesResponse.md)&gt;
+
+Child namespace names and
+  an optional token for fetching the next page.
+
+***
+
 ### openTable()
 
 ```ts
 abstract openTable(
    name,
-   namespace?,
+   namespacePath?,
    options?): Promise<Table>
 ```
 
@@ -315,8 +425,8 @@ Open a table in the database.
 * **name**: `string`
     The name of the table
 
-* **namespace?**: `string`[]
+* **namespacePath?**: `string`[]
-    The namespace of the table (defaults to root namespace)
+    The namespace path of the table (defaults to root namespace)
 
 * **options?**: `Partial`&lt;[`OpenTableOptions`](../interfaces/OpenTableOptions.md)&gt;
     Additional options
@@ -349,10 +459,10 @@ Tables will be returned in lexicographical order.
 
 `Promise`&lt;`string`[]&gt;
 
-#### tableNames(namespace, options)
+#### tableNames(namespacePath, options)
 
 ```ts
-abstract tableNames(namespace?, options?): Promise<string[]>
+abstract tableNames(namespacePath?, options?): Promise<string[]>
 ```
 
 List all the table names in this database.
@@ -361,8 +471,8 @@ Tables will be returned in lexicographical order.
 
 ##### Parameters
 
-* **namespace?**: `string`[]
+* **namespacePath?**: `string`[]
-    The namespace to list tables from (defaults to root namespace)
+    The namespace path to list tables from (defaults to root namespace)
 
 * **options?**: `Partial`&lt;[`TableNamesOptions`](../interfaces/TableNamesOptions.md)&gt;
     options to control the

docs/src/js/classes/Table.md

@@ -501,6 +501,34 @@ Modeled after ``VACUUM`` in PostgreSQL.
 
 ***
 
+### prewarmData()
+
+```ts
+abstract prewarmData(columns?): Promise<void>
+```
+
+Prewarm one or more columns of data in the table.
+
+#### Parameters
+
+* **columns?**: `string`[]
+    The columns to prewarm. If undefined, all columns are prewarmed.
+    This will load the column data into the page cache so that future queries that
+    read those columns avoid the initial cold-start latency.  This call initiates
+    prewarming and returns once the request is accepted; the warming itself may
+    continue in the background.  Calling it on already-prewarmed columns is a
+    no-op on the server.
+    Prewarming is generally useful for columns used in filters or projections.
+    Large columns (e.g. high-dimensional vectors or binary data) may not be
+    practical to prewarm.
+    This feature is currently only supported on remote tables.
+
+#### Returns
+
+`Promise`&lt;`void`&gt;
+
+***
+
 ### prewarmIndex()
 
 ```ts

docs/src/js/functions/connectNamespace.md

@@ -0,0 +1,131 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / connectNamespace
+
+# Function: connectNamespace()
+
+## connectNamespace(implName, config, options)
+
+```ts
+function connectNamespace(
+   implName,
+   config,
+   options?): Promise<Connection>
+```
+
+Connect to a LanceDB database through a namespace.
+
+Unlike [connect](connect.md), which routes by URI scheme (local path vs.
+`db://` cloud), `connectNamespace` always returns a namespace-backed
+connection. The `implName` selects the namespace implementation:
+
+- `"dir"` — directory namespace, configured with [DirNamespaceConfig](../interfaces/DirNamespaceConfig.md).
+- `"rest"` — remote REST catalog, configured with [RestNamespaceConfig](../interfaces/RestNamespaceConfig.md).
+- Any other string — full module path for a custom implementation,
+  configured with a free-form string-keyed `properties` map.
+
+### Parameters
+
+* **implName**: `"dir"`
+
+* **config**: [`DirNamespaceConfig`](../interfaces/DirNamespaceConfig.md)
+
+* **options?**: `Partial`&lt;[`ConnectNamespaceOptions`](../interfaces/ConnectNamespaceOptions.md)&gt;
+
+### Returns
+
+`Promise`&lt;[`Connection`](../classes/Connection.md)&gt;
+
+### Examples
+
+```ts
+const db = await connectNamespace("dir", { root: "/path/to/db" });
+await db.createTable("users", [{ id: 1 }]);
+```
+
+```ts
+const db = await connectNamespace("rest", {
+  uri: "https://catalog.example.com",
+  headers: { "x-api-key": process.env.CATALOG_KEY ?? "" },
+});
+```
+
+```ts
+const db = await connectNamespace("my.custom.Namespace", {
+  endpoint: "...",
+});
+```
+
+## connectNamespace(implName, config, options)
+
+```ts
+function connectNamespace(
+   implName,
+   config,
+   options?): Promise<Connection>
+```
+
+Connect through the built-in REST namespace.
+
+Configured with [RestNamespaceConfig](../interfaces/RestNamespaceConfig.md). See the function-level
+documentation above for the full surface, examples, and how this
+relates to [connect](connect.md).
+
+### Parameters
+
+* **implName**: `"rest"`
+
+* **config**: [`RestNamespaceConfig`](../interfaces/RestNamespaceConfig.md)
+
+* **options?**: `Partial`&lt;[`ConnectNamespaceOptions`](../interfaces/ConnectNamespaceOptions.md)&gt;
+
+### Returns
+
+`Promise`&lt;[`Connection`](../classes/Connection.md)&gt;
+
+### Example
+
+```ts
+const db = await connectNamespace("rest", {
+  uri: "https://catalog.example.com",
+  headers: { "x-api-key": process.env.CATALOG_KEY ?? "" },
+});
+```
+
+## connectNamespace(implName, properties, options)
+
+```ts
+function connectNamespace(
+   implName,
+   properties,
+   options?): Promise<Connection>
+```
+
+Connect through a custom namespace implementation by full module path,
+configured with a free-form string-keyed `properties` map. Use the
+typed overloads above for the built-in `"dir"` and `"rest"` impls.
+
+See the function-level documentation above for examples and how this
+relates to [connect](connect.md).
+
+### Parameters
+
+* **implName**: `string`
+
+* **properties**: `Record`&lt;`string`, `string`&gt;
+
+* **options?**: `Partial`&lt;[`ConnectNamespaceOptions`](../interfaces/ConnectNamespaceOptions.md)&gt;
+
+### Returns
+
+`Promise`&lt;[`Connection`](../classes/Connection.md)&gt;
+
+### Example
+
+```ts
+const db = await connectNamespace("my.custom.Namespace", {
+  endpoint: "...",
+});
+```

docs/src/js/globals.md

@@ -51,10 +51,17 @@
 - [ClientConfig](interfaces/ClientConfig.md)
 - [ColumnAlteration](interfaces/ColumnAlteration.md)
 - [CompactionStats](interfaces/CompactionStats.md)
+- [ConnectNamespaceOptions](interfaces/ConnectNamespaceOptions.md)
 - [ConnectionOptions](interfaces/ConnectionOptions.md)
+- [CreateNamespaceOptions](interfaces/CreateNamespaceOptions.md)
+- [CreateNamespaceResponse](interfaces/CreateNamespaceResponse.md)
 - [CreateTableOptions](interfaces/CreateTableOptions.md)
 - [DeleteResult](interfaces/DeleteResult.md)
+- [DescribeNamespaceResponse](interfaces/DescribeNamespaceResponse.md)
+- [DirNamespaceConfig](interfaces/DirNamespaceConfig.md)
 - [DropColumnsResult](interfaces/DropColumnsResult.md)
+- [DropNamespaceOptions](interfaces/DropNamespaceOptions.md)
+- [DropNamespaceResponse](interfaces/DropNamespaceResponse.md)
 - [ExecutableQuery](interfaces/ExecutableQuery.md)
 - [FragmentStatistics](interfaces/FragmentStatistics.md)
 - [FragmentSummaryStats](interfaces/FragmentSummaryStats.md)
@@ -69,12 +76,15 @@
 - [IvfFlatOptions](interfaces/IvfFlatOptions.md)
 - [IvfPqOptions](interfaces/IvfPqOptions.md)
 - [IvfRqOptions](interfaces/IvfRqOptions.md)
+- [ListNamespacesOptions](interfaces/ListNamespacesOptions.md)
+- [ListNamespacesResponse](interfaces/ListNamespacesResponse.md)
 - [MergeResult](interfaces/MergeResult.md)
 - [OpenTableOptions](interfaces/OpenTableOptions.md)
 - [OptimizeOptions](interfaces/OptimizeOptions.md)
 - [OptimizeStats](interfaces/OptimizeStats.md)
 - [QueryExecutionOptions](interfaces/QueryExecutionOptions.md)
 - [RemovalStats](interfaces/RemovalStats.md)
+- [RestNamespaceConfig](interfaces/RestNamespaceConfig.md)
 - [RetryConfig](interfaces/RetryConfig.md)
 - [ShuffleOptions](interfaces/ShuffleOptions.md)
 - [SplitCalculatedOptions](interfaces/SplitCalculatedOptions.md)
@@ -107,6 +117,7 @@
 
 - [RecordBatchIterator](functions/RecordBatchIterator.md)
 - [connect](functions/connect.md)
+- [connectNamespace](functions/connectNamespace.md)
 - [makeArrowTable](functions/makeArrowTable.md)
 - [packBits](functions/packBits.md)
 - [permutationBuilder](functions/permutationBuilder.md)

docs/src/js/interfaces/ClientConfig.md

@@ -53,3 +53,18 @@ optional tlsConfig: TlsConfig;
 ```ts
 optional userAgent: string;
 ```
+
+***
+
+### userId?
+
+```ts
+optional userId: string;
+```
+
+User identifier for tracking purposes.
+
+This is sent as the `x-lancedb-user-id` header in requests to LanceDB Cloud/Enterprise.
+It can be set directly, or via the `LANCEDB_USER_ID` environment variable.
+Alternatively, set `LANCEDB_USER_ID_ENV_KEY` to specify another environment
+variable that contains the user ID value.

docs/src/js/interfaces/ConnectNamespaceOptions.md

@@ -0,0 +1,54 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / ConnectNamespaceOptions
+
+# Interface: ConnectNamespaceOptions
+
+## Properties
+
+### namespaceClientProperties?
+
+```ts
+optional namespaceClientProperties: Record<string, string>;
+```
+
+Extra properties for the backing namespace client.
+
+***
+
+### readConsistencyInterval?
+
+```ts
+optional readConsistencyInterval: number;
+```
+
+The interval, in seconds, at which to check for updates to the table
+from other processes. If None, then consistency is not checked. For
+performance reasons, this is the default. For strong consistency, set
+this to zero seconds. Then every read will check for updates from other
+processes. As a compromise, you can set this to a non-zero value for
+eventual consistency.
+
+***
+
+### session?
+
+```ts
+optional session: Session;
+```
+
+The session to use for this connection. Holds shared caches and other
+session-specific state.
+
+***
+
+### storageOptions?
+
+```ts
+optional storageOptions: Record<string, string>;
+```
+
+Configuration for object storage. The available options are described
+at https://docs.lancedb.com/storage/

docs/src/js/interfaces/ConnectionOptions.md

@@ -41,6 +41,29 @@ for testing purposes.
 
 ***
 
+### manifestEnabled?
+
+```ts
+optional manifestEnabled: boolean;
+```
+
+(For LanceDB OSS only): use directory namespace manifests as the source
+of truth for table metadata. Existing directory-listed root tables are
+migrated into the manifest on access.
+
+***
+
+### namespaceClientProperties?
+
+```ts
+optional namespaceClientProperties: Record<string, string>;
+```
+
+(For LanceDB OSS only): extra properties for the backing namespace
+client used by manifest-enabled native connections.
+
+***
+
 ### readConsistencyInterval?
 
 ```ts
@@ -89,4 +112,4 @@ optional storageOptions: Record<string, string>;
 
 (For LanceDB OSS only): configuration for object storage.
 
-The available options are described at https://lancedb.com/docs/storage/
+The available options are described at https://docs.lancedb.com/storage/

docs/src/js/interfaces/CreateNamespaceOptions.md

@@ -0,0 +1,27 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / CreateNamespaceOptions
+
+# Interface: CreateNamespaceOptions
+
+## Properties
+
+### mode?
+
+```ts
+optional mode: "overwrite" | "create" | "exist_ok";
+```
+
+Creation mode.
+
+***
+
+### properties?
+
+```ts
+optional properties: Record<string, string>;
+```
+
+Properties to set on the new namespace.

docs/src/js/interfaces/CreateNamespaceResponse.md

@@ -0,0 +1,23 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / CreateNamespaceResponse
+
+# Interface: CreateNamespaceResponse
+
+## Properties
+
+### properties?
+
+```ts
+optional properties: Record<string, string>;
+```
+
+***
+
+### transactionId?
+
+```ts
+optional transactionId: string;
+```

docs/src/js/interfaces/CreateTableOptions.md

@@ -97,4 +97,4 @@ Configuration for object storage.
 Options already set on the connection will be inherited by the table,
 but can be overridden here.
 
-The available options are described at https://lancedb.com/docs/storage/
+The available options are described at https://docs.lancedb.com/storage/

docs/src/js/interfaces/DescribeNamespaceResponse.md

@@ -0,0 +1,15 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / DescribeNamespaceResponse
+
+# Interface: DescribeNamespaceResponse
+
+## Properties
+
+### properties?
+
+```ts
+optional properties: Record<string, string>;
+```

docs/src/js/interfaces/DirNamespaceConfig.md

@@ -0,0 +1,47 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / DirNamespaceConfig
+
+# Interface: DirNamespaceConfig
+
+Configuration for the built-in directory namespace (`"dir"`).
+
+The directory namespace stores tables under a single root path (local
+filesystem or object storage URI). See
+[https://docs.lancedb.com/namespaces](https://docs.lancedb.com/namespaces) for the documented surface;
+less-common knobs live under [DirNamespaceConfig.extraProperties](DirNamespaceConfig.md#extraproperties).
+
+## Properties
+
+### extraProperties?
+
+```ts
+optional extraProperties: Record<string, string>;
+```
+
+Additional raw properties passed verbatim to the namespace
+implementation (e.g. `storage.*`, `credential_vendor.*`). Typed
+fields above take precedence on key collision.
+
+***
+
+### manifestEnabled?
+
+```ts
+optional manifestEnabled: boolean;
+```
+
+Whether to maintain a namespace manifest at the root. Required for
+child namespaces. Defaults to true on the impl side.
+
+***
+
+### root
+
+```ts
+root: string;
+```
+
+Root path or URI containing the LanceDB tables.

docs/src/js/interfaces/DropNamespaceOptions.md

@@ -0,0 +1,27 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / DropNamespaceOptions
+
+# Interface: DropNamespaceOptions
+
+## Properties
+
+### behavior?
+
+```ts
+optional behavior: "restrict" | "cascade";
+```
+
+Refuse to drop if non-empty (restrict) or drop recursively (cascade).
+
+***
+
+### mode?
+
+```ts
+optional mode: "fail" | "skip";
+```
+
+Whether to skip if the namespace doesn't exist, or fail.

docs/src/js/interfaces/DropNamespaceResponse.md

@@ -0,0 +1,23 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / DropNamespaceResponse
+
+# Interface: DropNamespaceResponse
+
+## Properties
+
+### properties?
+
+```ts
+optional properties: Record<string, string>;
+```
+
+***
+
+### transactionId?
+
+```ts
+optional transactionId: string[];
+```

docs/src/js/interfaces/ListNamespacesOptions.md

@@ -0,0 +1,27 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / ListNamespacesOptions
+
+# Interface: ListNamespacesOptions
+
+## Properties
+
+### limit?
+
+```ts
+optional limit: number;
+```
+
+An optional limit to the number of results to return.
+
+***
+
+### pageToken?
+
+```ts
+optional pageToken: string;
+```
+
+Token from a previous response for pagination.

docs/src/js/interfaces/ListNamespacesResponse.md

@@ -0,0 +1,23 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / ListNamespacesResponse
+
+# Interface: ListNamespacesResponse
+
+## Properties
+
+### namespaces
+
+```ts
+namespaces: string[];
+```
+
+***
+
+### pageToken?
+
+```ts
+optional pageToken: string;
+```

docs/src/js/interfaces/OpenTableOptions.md

@@ -42,4 +42,4 @@ Configuration for object storage.
 Options already set on the connection will be inherited by the table,
 but can be overridden here.
 
-The available options are described at https://lancedb.com/docs/storage/
+The available options are described at https://docs.lancedb.com/storage/

docs/src/js/interfaces/RestNamespaceConfig.md

@@ -0,0 +1,47 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / RestNamespaceConfig
+
+# Interface: RestNamespaceConfig
+
+Configuration for the built-in REST namespace (`"rest"`).
+
+The REST namespace talks to a remote catalog server over HTTP. See
+[https://docs.lancedb.com/namespaces](https://docs.lancedb.com/namespaces) for the documented surface;
+less-common knobs (TLS, metrics) live under
+[RestNamespaceConfig.extraProperties](RestNamespaceConfig.md#extraproperties).
+
+## Properties
+
+### extraProperties?
+
+```ts
+optional extraProperties: Record<string, string>;
+```
+
+Additional raw properties passed verbatim to the namespace
+implementation (e.g. `tls.*`, `ops_metrics_enabled`, `delimiter`).
+Typed fields above take precedence on key collision.
+
+***
+
+### headers?
+
+```ts
+optional headers: Record<string, string>;
+```
+
+HTTP headers forwarded with each request. Keys are passed through
+as-is (e.g. `"x-api-key"`, `"Authorization"`).
+
+***
+
+### uri
+
+```ts
+uri: string;
+```
+
+Catalog endpoint URL.

docs/src/python/python.md

@@ -94,11 +94,11 @@ of raw SQL strings with [where][lancedb.query.LanceQueryBuilder.where] and
 
 ## Full text search
 
-::: lancedb.fts.create_index
+Use [lancedb.table.Table.create_fts_index][] for the synchronous API or
+[lancedb.table.AsyncTable.create_index][] with [lancedb.index.FTS][] for the
+asynchronous API.
 
-::: lancedb.fts.populate_index
+::: lancedb.index.FTS
-
-::: lancedb.fts.search_index
 
 ## Utilities
 

java/lancedb-core/pom.xml

@@ -8,7 +8,7 @@
     <parent>
       <groupId>com.lancedb</groupId>
       <artifactId>lancedb-parent</artifactId>
-      <version>0.27.2-beta.1</version>
+      <version>0.28.0-beta.11</version>
       <relativePath>../pom.xml</relativePath>
     </parent>
 

java/pom.xml

@@ -6,7 +6,7 @@
 
     <groupId>com.lancedb</groupId>
     <artifactId>lancedb-parent</artifactId>
-    <version>0.27.2-beta.1</version>
+    <version>0.28.0-beta.11</version>
     <packaging>pom</packaging>
     <name>${project.artifactId}</name>
     <description>LanceDB Java SDK Parent POM</description>
@@ -28,7 +28,7 @@
     <properties>
         <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
         <arrow.version>15.0.0</arrow.version>
-        <lance-core.version>3.0.1</lance-core.version>
+        <lance-core.version>7.0.0-beta.7</lance-core.version>
         <spotless.skip>false</spotless.skip>
         <spotless.version>2.30.0</spotless.version>
         <spotless.java.googlejavaformat.version>1.7</spotless.java.googlejavaformat.version>

nodejs/AGENTS.md

@@ -3,11 +3,11 @@ The core Rust library is in the `../rust/lancedb` directory, the rust binding
 code is in the `src/` directory and the typescript bindings are in
 the `lancedb/` directory.
 
-Whenever you change the Rust code, you will need to recompile: `npm run build`.
+Whenever you change the Rust code, you will need to recompile: `pnpm build`.
 
 Common commands:
-* Build: `npm run build`
+* Build: `pnpm build`
-* Lint: `npm run lint`
+* Lint: `pnpm lint`
-* Fix lints: `npm run lint-fix`
+* Fix lints: `pnpm lint-fix`
-* Test: `npm test`
+* Test: `pnpm test`
-* Run single test file: `npm test __test__/arrow.test.ts`
+* Run single test file: `pnpm test __test__/arrow.test.ts`

nodejs/CONTRIBUTING.md

@@ -12,20 +12,22 @@ Typescript.
 * `src/`: Rust bindings source code
 * `lancedb/`: Typescript package source code
 * `__test__/`: Unit tests
-* `examples/`: An npm package with the examples shown in the documentation
+* `examples/`: A pnpm package with the examples shown in the documentation
 
 ## Development environment
 
 To set up your development environment, you will need to install the following:
 
-1. Node.js 14 or later
+1. Node.js 22 or later (required by pnpm 11)
-2. Rust's package manager, Cargo. Use [rustup](https://rustup.rs/) to install.
+2. [pnpm](https://pnpm.io/installation) 11 or later (or run via `corepack enable`,
-3. [protoc](https://grpc.io/docs/protoc-installation/) (Protocol Buffers compiler)
+   which uses the `packageManager` field in `package.json`)
+3. Rust's package manager, Cargo. Use [rustup](https://rustup.rs/) to install.
+4. [protoc](https://grpc.io/docs/protoc-installation/) (Protocol Buffers compiler)
 
 Initial setup:
 
 ```shell
-npm install
+pnpm install
 ```
 
 ### Commit Hooks
@@ -39,38 +41,38 @@ pre-commit install
 
 ## Development
 
-Most common development commands can be run using the npm scripts.
+Most common development commands can be run using the pnpm scripts.
 
 Build the package
 
 ```shell
-npm install
+pnpm install
-npm run build
+pnpm build
 ```
 
 Lint:
 
 ```shell
-npm run lint
+pnpm lint
 ```
 
 Format and fix lints:
 
 ```shell
-npm run lint-fix
+pnpm lint-fix
 ```
 
 Run tests:
 
 ```shell
-npm test
+pnpm test
 ```
 
 To run a single test:
 
 ```shell
 # Single file: table.test.ts
-npm test -- table.test.ts
+pnpm test -- table.test.ts
 # Single test: 'merge insert' in table.test.ts
-npm test -- table.test.ts --testNamePattern=merge\ insert
+pnpm test -- table.test.ts --testNamePattern=merge\ insert
 ```

nodejs/Cargo.toml

@@ -1,7 +1,8 @@
 [package]
 name = "lancedb-nodejs"
 edition.workspace = true
-version = "0.27.2-beta.1"
+version = "0.28.0-beta.11"
+publish = false
 license.workspace = true
 description.workspace = true
 repository.workspace = true
@@ -15,12 +16,13 @@ crate-type = ["cdylib"]
 async-trait.workspace = true
 arrow-ipc.workspace = true
 arrow-array.workspace = true
-arrow-buffer = "57.2"
+arrow-buffer = "58.0.0"
 half.workspace = true
 arrow-schema.workspace = true
 env_logger.workspace = true
 futures.workspace = true
 lancedb = { path = "../rust/lancedb", default-features = false }
+lance-namespace.workspace = true
 napi = { version = "3.8.3", default-features = false, features = [
     "napi9",
     "async"
@@ -31,8 +33,8 @@ lzma-sys = { version = "0.1", features = ["static"] }
 log.workspace = true
 
 # Pin to resolve build failures; update periodically for security patches.
-aws-lc-sys = "=0.38.0"
+aws-lc-sys = "=0.40.0"
-aws-lc-rs = "=1.16.1"
+aws-lc-rs = "=1.16.3"
 
 [build-dependencies]
 napi-build = "2.3.1"

nodejs/README.md

@@ -30,7 +30,7 @@ const results = await table.vectorSearch([0.1, 0.3]).limit(20).toArray();
 console.log(results);
 ```
 
-The [quickstart](https://lancedb.com/docs/quickstart/basic-usage/) contains more complete examples.
+The [quickstart](https://docs.lancedb.com/quickstart/) contains more complete examples.
 
 ## Development
 

nodejs/__test__/connection.test.ts

@@ -4,7 +4,7 @@
 import { readdirSync } from "fs";
 import { Field, Float64, Schema } from "apache-arrow";
 import * as tmp from "tmp";
-import { Connection, Table, connect } from "../lancedb";
+import { Connection, Table, connect, connectNamespace } from "../lancedb";
 import { LocalTable } from "../lancedb/table";
 
 describe("when connecting", () => {
@@ -306,3 +306,186 @@ describe("clone table functionality", () => {
     ).rejects.toThrow("Deep clone is not yet implemented");
   });
 });
+
+describe("namespaces", () => {
+  let tmpDir: tmp.DirResult;
+  let db: Connection;
+
+  beforeEach(async () => {
+    tmpDir = tmp.dirSync({ unsafeCleanup: true });
+    // The local DirectoryNamespace backend only supports child namespaces
+    // when manifest mode is enabled (see lance-namespace-impls/src/dir.rs).
+    db = await connect(tmpDir.name, {
+      // biome-ignore lint/style/useNamingConvention: opaque backend property key, must match Rust
+      namespaceClientProperties: { manifest_enabled: "true" },
+    });
+  });
+  afterEach(() => tmpDir.removeCallback());
+
+  it("should create and describe a namespace", async () => {
+    await db.createNamespace(["myns"]);
+    const desc = await db.describeNamespace(["myns"]);
+    expect(desc).toBeDefined();
+  });
+
+  it("should list namespaces created at the root", async () => {
+    await db.createNamespace(["alpha"]);
+    await db.createNamespace(["beta"]);
+    const list = await db.listNamespaces();
+    expect(list.namespaces).toEqual(expect.arrayContaining(["alpha", "beta"]));
+  });
+
+  it("should list child namespaces under a parent", async () => {
+    await db.createNamespace(["parent"]);
+    await db.createNamespace(["parent", "child"]);
+    const list = await db.listNamespaces(["parent"]);
+    expect(list.namespaces).toContain("child");
+  });
+
+  it("should drop a namespace", async () => {
+    await db.createNamespace(["ephemeral"]);
+    await db.dropNamespace(["ephemeral"]);
+    const list = await db.listNamespaces();
+    expect(list.namespaces).not.toContain("ephemeral");
+  });
+
+  it("should raise an error on any namespace op after close", async () => {
+    await db.close();
+    await expect(db.describeNamespace(["foo"])).rejects.toThrow(
+      "Connection is closed",
+    );
+    await expect(db.listNamespaces()).rejects.toThrow("Connection is closed");
+    await expect(db.createNamespace(["foo"])).rejects.toThrow(
+      "Connection is closed",
+    );
+    await expect(db.dropNamespace(["foo"])).rejects.toThrow(
+      "Connection is closed",
+    );
+  });
+
+  it("should raise an understandable error when describing a non-existent namespace", async () => {
+    await expect(db.describeNamespace(["does-not-exist"])).rejects.toThrow(
+      /not found/i,
+    );
+  });
+
+  it("should raise an error when creating a namespace that already exists", async () => {
+    await db.createNamespace(["dup"]);
+    await expect(db.createNamespace(["dup"])).rejects.toThrow();
+  });
+
+  it("should reject an unrecognized createNamespace mode with a clear error", async () => {
+    await expect(
+      // biome-ignore lint/suspicious/noExplicitAny: deliberately bypass TS to test runtime validation
+      db.createNamespace(["x"], { mode: "frobnicate" as any }),
+    ).rejects.toThrow(/Invalid mode 'frobnicate'/);
+  });
+
+  it("should reject an unrecognized dropNamespace mode with a clear error", async () => {
+    await db.createNamespace(["x"]);
+    await expect(
+      // biome-ignore lint/suspicious/noExplicitAny: deliberately bypass TS to test runtime validation
+      db.dropNamespace(["x"], { mode: "frobnicate" as any }),
+    ).rejects.toThrow(/Invalid mode 'frobnicate'/);
+  });
+
+  it("should reject an unrecognized dropNamespace behavior with a clear error", async () => {
+    await db.createNamespace(["x"]);
+    await expect(
+      // biome-ignore lint/suspicious/noExplicitAny: deliberately bypass TS to test runtime validation
+      db.dropNamespace(["x"], { behavior: "frobnicate" as any }),
+    ).rejects.toThrow(/Invalid behavior 'frobnicate'/);
+  });
+});
+
+describe("connectNamespace", () => {
+  let tmpDir: tmp.DirResult;
+  beforeEach(() => {
+    tmpDir = tmp.dirSync({ unsafeCleanup: true });
+  });
+  afterEach(() => tmpDir.removeCallback());
+
+  it("connects via the dir implementation and supports table ops", async () => {
+    const db = await connectNamespace("dir", { root: tmpDir.name });
+    await db.createTable("users", [{ id: 1 }, { id: 2 }]);
+    await expect(db.tableNames()).resolves.toContain("users");
+  });
+
+  it("throws a clear error when implName is empty", async () => {
+    await expect(connectNamespace("", {})).rejects.toThrow(
+      "implName must be a non-empty string",
+    );
+  });
+
+  it("throws when the namespace implementation is unknown", async () => {
+    await expect(connectNamespace("not-a-real-impl", {})).rejects.toThrow();
+  });
+
+  it("passes storage options through to the namespace", async () => {
+    const db = await connectNamespace(
+      "dir",
+      { root: tmpDir.name },
+      { storageOptions: { newTableDataStorageVersion: "stable" } },
+    );
+    await db.createTable("plumbing", [{ id: 1 }]);
+    await expect(db.tableNames()).resolves.toContain("plumbing");
+  });
+
+  it("supports child namespaces when manifestEnabled is true on the dir config", async () => {
+    const writer = await connectNamespace("dir", {
+      root: tmpDir.name,
+      manifestEnabled: true,
+    });
+    await writer.createNamespace(["analytics"]);
+    await writer.createTable("orders", [{ id: 1 }, { id: 2 }], ["analytics"]);
+    await writer.close();
+
+    const reader = await connectNamespace("dir", {
+      root: tmpDir.name,
+      manifestEnabled: true,
+    });
+    await expect(reader.tableNames(["analytics"])).resolves.toContain("orders");
+    const orders = await reader.openTable("orders", ["analytics"]);
+    await expect(orders.countRows()).resolves.toBe(2);
+  });
+
+  it("merges extraProperties into the dir config and is overridden by typed fields", async () => {
+    // Two observable assertions:
+    // - Typed `root` overrides extraProperties.root: createTable would fail
+    //   under the bogus path if the override didn't happen.
+    // - extraProperties.manifest_enabled="false" is honored end-to-end. Child
+    //   namespaces require manifest mode (default true), so explicitly
+    //   disabling it via extraProperties must make createNamespace reject. If
+    //   extraProperties pass-through were silently broken, the default would
+    //   let createNamespace succeed.
+    const db = await connectNamespace("dir", {
+      root: tmpDir.name,
+      extraProperties: {
+        root: "/should/be/overridden",
+        // biome-ignore lint/style/useNamingConvention: backend property key
+        manifest_enabled: "false",
+      },
+    });
+    await db.createTable("base", [{ id: 1 }]);
+    await expect(db.tableNames()).resolves.toContain("base");
+    await expect(db.createNamespace(["analytics"])).rejects.toThrow();
+  });
+
+  it("flows unknown top-level keys through when implName is dynamic (no silent drop)", async () => {
+    // Routes via the third overload because `impl` is `string`, not the
+    // literal `"dir"`. The dispatcher still notices the runtime value is
+    // "dir", but unknown keys like `manifest_enabled` must not be silently
+    // dropped during the conversion.
+    //
+    // Asserting a *negative* outcome (manifest disabled -> createNamespace
+    // rejects) is required for observability, since the backend default for
+    // `manifest_enabled` is true.
+    const impl: string = "dir";
+    const db = await connectNamespace(impl, {
+      root: tmpDir.name,
+      // biome-ignore lint/style/useNamingConvention: backend property key
+      manifest_enabled: "false",
+    });
+    await expect(db.createNamespace(["mixed"])).rejects.toThrow();
+  });
+});

nodejs/__test__/rerankers.test.ts

@@ -1,6 +1,8 @@
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
+import { spawn } from "node:child_process";
+import * as path from "node:path";
 import { RecordBatch } from "apache-arrow";
 import * as tmp from "tmp";
 import { Connection, Index, Table, connect, makeArrowTable } from "../lancedb";
@@ -76,4 +78,91 @@ describe("rerankers", function () {
 
     expect(result).toHaveLength(2);
   });
+
+  it("does not keep process alive after rerank query", async function () {
+    const script = `
+import * as lancedb from "./dist/index.js";
+import * as os from "node:os";
+import * as path from "node:path";
+import * as fs from "node:fs/promises";
+
+const dir = await fs.mkdtemp(path.join(os.tmpdir(), "lancedb-rerank-exit-"));
+const db = await lancedb.connect(dir);
+const table = await db.createTable("test", [{ text: "hello", vector: [1, 2, 3] }], {
+  mode: "overwrite",
+});
+await table.createIndex("text", { config: lancedb.Index.fts() });
+await table.waitForIndex(["text_idx"], 30);
+
+const reranker = await lancedb.rerankers.RRFReranker.create();
+await table
+  .query()
+  .nearestTo([1, 2, 3])
+  .fullTextSearch("hello")
+  .rerank(reranker)
+  .toArray();
+
+table.close();
+db.close();
+`;
+
+    await new Promise<void>((resolve, reject) => {
+      const child = spawn(
+        process.execPath,
+        ["--input-type=module", "-e", script],
+        {
+          cwd: path.resolve(__dirname, ".."),
+          stdio: ["ignore", "pipe", "pipe"],
+        },
+      );
+
+      let stdout = "";
+      let stderr = "";
+
+      child.stdout.on("data", (chunk) => {
+        stdout += chunk.toString();
+      });
+
+      child.stderr.on("data", (chunk) => {
+        stderr += chunk.toString();
+      });
+
+      const timeout = setTimeout(() => {
+        child.kill();
+        reject(
+          new Error(
+            `child process did not exit in time\nstdout:\n${stdout}\nstderr:\n${stderr}`,
+          ),
+        );
+      }, 20_000);
+
+      child.on("error", (err) => {
+        clearTimeout(timeout);
+        reject(err);
+      });
+
+      child.on("exit", (code, signal) => {
+        clearTimeout(timeout);
+        if (signal !== null) {
+          reject(
+            new Error(
+              `child process exited with signal ${signal}\nstdout:\n${stdout}\nstderr:\n${stderr}`,
+            ),
+          );
+          return;
+        }
+
+        if (code !== 0) {
+          reject(
+            new Error(
+              `child process exited with code ${code}\nstdout:\n${stdout}\nstderr:\n${stderr}`,
+            ),
+          );
+          return;
+        }
+
+        resolve();
+      });
+    });
+  });
 });

nodejs/__test__/table.test.ts

@@ -103,7 +103,7 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
         },
         numIndices: 0,
         numRows: 3,
-        totalBytes: 24,
+        totalBytes: 44,
       });
     });
 
@@ -115,6 +115,12 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
       await expect(table.countRows()).resolves.toBe(1);
     });
 
+    it("should accept skipAutoCleanup on add()", async () => {
+      await table.add([{ id: 1 }], { skipAutoCleanup: true });
+      await table.add([{ id: 2 }], { skipAutoCleanup: true });
+      await expect(table.countRows()).resolves.toBe(2);
+    });
+
     it("should let me close the table", async () => {
       expect(table.isOpen()).toBe(true);
       table.close();
@@ -1870,6 +1876,25 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
       expect(results.length).toBe(3);
     });
 
+    test("prewarmData errors on local tables", async () => {
+      const db = await connect(tmpDir.name);
+      const data = [
+        { text: "alpha", vector: [0.1, 0.2, 0.3] },
+        { text: "beta", vector: [0.4, 0.5, 0.6] },
+      ];
+      const table = await db.createTable("prewarm_data_test", data);
+
+      // prewarmData is only supported on remote tables. We verify the call
+      // is wired through napi and surfaces the expected error for both
+      // arg shapes (undefined and string[]).
+      await expect(table.prewarmData()).rejects.toThrow(
+        "prewarm_data is currently only supported on remote tables",
+      );
+      await expect(table.prewarmData(["text"])).rejects.toThrow(
+        "prewarm_data is currently only supported on remote tables",
+      );
+    });
+
     test("full text index on list", async () => {
       const db = await connect(tmpDir.name);
       const data = [

nodejs/examples/package.json

@@ -11,16 +11,17 @@
     "test": "node --experimental-vm-modules node_modules/.bin/jest --testEnvironment jest-environment-node-single-context --verbose",
     "lint": "biome check *.ts && biome format *.ts",
     "lint-ci": "biome ci .",
-    "lint-fix": "biome check --write *.ts && npm run format",
+    "lint-fix": "biome check --write *.ts && pnpm format",
     "format": "biome format --write *.ts"
   },
   "author": "Lance Devs",
   "license": "Apache-2.0",
+  "packageManager": "pnpm@11.1.1",
   "dependencies": {
-    "@huggingface/transformers": "^3.0.2",
+    "@huggingface/transformers": "3.0.2",
     "@lancedb/lancedb": "file:../dist",
-    "openai": "^4.29.2",
+    "openai": "4.29.2",
-    "sharp": "^0.33.5"
+    "sharp": "0.33.5"
   },
   "devDependencies": {
     "@biomejs/biome": "^1.7.3",

nodejs/examples/pnpm-workspace.yaml

@@ -0,0 +1,13 @@
+# Block resolution of versions less than 24h old (Shai-Hulud window).
+# This is the pnpm 11 default but pinned here so it's visible to
+# reviewers and survives a future pnpm major flipping the default.
+minimumReleaseAge: 1440
+
+# Fail install if a transitive dep tries to run an unapproved script.
+strictDepBuilds: true
+
+allowBuilds:
+  '@biomejs/biome': true
+  onnxruntime-node: true
+  protobufjs: true
+  sharp: true

nodejs/lancedb/connection.ts

@@ -16,6 +16,18 @@ import {
 } from "./arrow";
 import { EmbeddingFunctionConfig, getRegistry } from "./embedding/registry";
 import { Connection as LanceDbConnection } from "./native";
+import type {
+  CreateNamespaceResponse,
+  DescribeNamespaceResponse,
+  DropNamespaceResponse,
+  ListNamespacesResponse,
+} from "./native";
+export type {
+  CreateNamespaceResponse,
+  DescribeNamespaceResponse,
+  DropNamespaceResponse,
+  ListNamespacesResponse,
+};
 import { sanitizeTable } from "./sanitize";
 import { LocalTable, Table } from "./table";
 
@@ -42,7 +54,7 @@ export interface CreateTableOptions {
    * Options already set on the connection will be inherited by the table,
    * but can be overridden here.
    *
-   * The available options are described at https://lancedb.com/docs/storage/
+   * The available options are described at https://docs.lancedb.com/storage/
    */
   storageOptions?: Record<string, string>;
 
@@ -78,7 +90,7 @@ export interface OpenTableOptions {
    * Options already set on the connection will be inherited by the table,
    * but can be overridden here.
    *
-   * The available options are described at https://lancedb.com/docs/storage/
+   * The available options are described at https://docs.lancedb.com/storage/
    */
   storageOptions?: Record<string, string>;
   /**
@@ -110,6 +122,28 @@ export interface TableNamesOptions {
   /** An optional limit to the number of results to return. */
   limit?: number;
 }
+
+export interface ListNamespacesOptions {
+  /** Token from a previous response for pagination. */
+  pageToken?: string;
+  /** An optional limit to the number of results to return. */
+  limit?: number;
+}
+
+export interface CreateNamespaceOptions {
+  /** Creation mode. */
+  mode?: "create" | "exist_ok" | "overwrite";
+  /** Properties to set on the new namespace. */
+  properties?: Record<string, string>;
+}
+
+export interface DropNamespaceOptions {
+  /** Whether to skip if the namespace doesn't exist, or fail. */
+  mode?: "skip" | "fail";
+  /** Refuse to drop if non-empty (restrict) or drop recursively (cascade). */
+  behavior?: "restrict" | "cascade";
+}
+
 /**
  * A LanceDB Connection that allows you to open tables and create new ones.
  *
@@ -166,25 +200,25 @@ export abstract class Connection {
    * List all the table names in this database.
    *
    * Tables will be returned in lexicographical order.
-   * @param {string[]} namespace - The namespace to list tables from (defaults to root namespace)
+   * @param {string[]} namespacePath - The namespace path to list tables from (defaults to root namespace)
    * @param {Partial<TableNamesOptions>} options - options to control the
    * paging / start point
    *
    */
   abstract tableNames(
-    namespace?: string[],
+    namespacePath?: string[],
     options?: Partial<TableNamesOptions>,
   ): Promise<string[]>;
 
   /**
    * Open a table in the database.
    * @param {string} name - The name of the table
-   * @param {string[]} namespace - The namespace of the table (defaults to root namespace)
+   * @param {string[]} namespacePath - The namespace path of the table (defaults to root namespace)
    * @param {Partial<OpenTableOptions>} options - Additional options
    */
   abstract openTable(
     name: string,
-    namespace?: string[],
+    namespacePath?: string[],
     options?: Partial<OpenTableOptions>,
   ): Promise<Table>;
 
@@ -193,7 +227,7 @@ export abstract class Connection {
    * @param {object} options - The options object.
    * @param {string} options.name - The name of the table.
    * @param {Data} options.data - Non-empty Array of Records to be inserted into the table
-   * @param {string[]} namespace - The namespace to create the table in (defaults to root namespace)
+   * @param {string[]} namespacePath - The namespace path to create the table in (defaults to root namespace)
    *
    */
   abstract createTable(
@@ -201,7 +235,7 @@ export abstract class Connection {
       name: string;
       data: Data;
     } & Partial<CreateTableOptions>,
-    namespace?: string[],
+    namespacePath?: string[],
   ): Promise<Table>;
   /**
    * Creates a new Table and initialize it with new data.
@@ -220,13 +254,13 @@ export abstract class Connection {
    * @param {string} name - The name of the table.
    * @param {Record<string, unknown>[] | TableLike} data - Non-empty Array of Records
    * to be inserted into the table
-   * @param {string[]} namespace - The namespace to create the table in (defaults to root namespace)
+   * @param {string[]} namespacePath - The namespace path to create the table in (defaults to root namespace)
    * @param {Partial<CreateTableOptions>} options - Additional options
    */
   abstract createTable(
     name: string,
     data: Record<string, unknown>[] | TableLike,
-    namespace?: string[],
+    namespacePath?: string[],
     options?: Partial<CreateTableOptions>,
   ): Promise<Table>;
 
@@ -245,28 +279,91 @@ export abstract class Connection {
    * Creates a new empty Table
    * @param {string} name - The name of the table.
    * @param {Schema} schema - The schema of the table
-   * @param {string[]} namespace - The namespace to create the table in (defaults to root namespace)
+   * @param {string[]} namespacePath - The namespace path to create the table in (defaults to root namespace)
    * @param {Partial<CreateTableOptions>} options - Additional options
    */
   abstract createEmptyTable(
     name: string,
     schema: import("./arrow").SchemaLike,
-    namespace?: string[],
+    namespacePath?: string[],
     options?: Partial<CreateTableOptions>,
   ): Promise<Table>;
 
   /**
    * Drop an existing table.
    * @param {string} name The name of the table to drop.
-   * @param {string[]} namespace The namespace of the table (defaults to root namespace).
+   * @param {string[]} namespacePath The namespace path of the table (defaults to root namespace).
    */
-  abstract dropTable(name: string, namespace?: string[]): Promise<void>;
+  abstract dropTable(name: string, namespacePath?: string[]): Promise<void>;
 
   /**
    * Drop all tables in the database.
-   * @param {string[]} namespace The namespace to drop tables from (defaults to root namespace).
+   * @param {string[]} namespacePath The namespace path to drop tables from (defaults to root namespace).
    */
-  abstract dropAllTables(namespace?: string[]): Promise<void>;
+  abstract dropAllTables(namespacePath?: string[]): Promise<void>;
+
+  /**
+   * Describe a namespace, returning its properties.
+   *
+   * @param {string[]} namespacePath - The namespace path to describe, in
+   *   parent → child order, e.g. `["analytics", "sales"]`.
+   * @returns {Promise<DescribeNamespaceResponse>} The namespace's properties
+   *   (may be undefined if the namespace has none).
+   */
+  abstract describeNamespace(
+    namespacePath: string[],
+  ): Promise<DescribeNamespaceResponse>;
+
+  /**
+   * List the immediate child namespaces under the given parent.
+   *
+   * Results may be paginated. To retrieve subsequent pages, pass the
+   * `pageToken` returned by a previous call.
+   *
+   * @param {string[]} namespacePath - The parent namespace path. Defaults
+   *   to the root namespace if omitted.
+   * @param {Partial<ListNamespacesOptions>} options - Pagination options
+   *   (`pageToken`, `limit`).
+   * @returns {Promise<ListNamespacesResponse>} Child namespace names and
+   *   an optional token for fetching the next page.
+   */
+  abstract listNamespaces(
+    namespacePath?: string[],
+    options?: Partial<ListNamespacesOptions>,
+  ): Promise<ListNamespacesResponse>;
+
+  /**
+   * Create a new namespace at the given path.
+   *
+   * @param {string[]} namespacePath - The namespace path to create.
+   * @param {Partial<CreateNamespaceOptions>} options - Creation `mode`
+   *   ("create" | "exist_ok" | "overwrite") and optional `properties`
+   *   to attach to the namespace.
+   * @returns {Promise<CreateNamespaceResponse>} The properties of the
+   *   created namespace and an optional transaction id.
+   */
+  abstract createNamespace(
+    namespacePath: string[],
+    options?: Partial<CreateNamespaceOptions>,
+  ): Promise<CreateNamespaceResponse>;
+
+  /**
+   * Drop a namespace.
+   *
+   * Use `behavior: "cascade"` to also drop everything contained in the
+   * namespace (sub-namespaces and tables). The default `"restrict"`
+   * behavior refuses to drop a non-empty namespace.
+   *
+   * @param {string[]} namespacePath - The namespace path to drop.
+   * @param {Partial<DropNamespaceOptions>} options - `mode` ("skip" | "fail"
+   *   for missing-namespace handling) and `behavior` ("restrict" | "cascade").
+   * @returns {Promise<DropNamespaceResponse>} Any properties returned by
+   *   the server and an optional transaction id.
+   */
+  abstract dropNamespace(
+    namespacePath: string[],
+    options?: Partial<DropNamespaceOptions>,
+  ): Promise<DropNamespaceResponse>;
 
   /**
    * Clone a table from a source table.
@@ -279,7 +376,7 @@ export abstract class Connection {
    * @param {string} targetTableName - The name of the target table to create.
    * @param {string} sourceUri - The URI of the source table to clone from.
    * @param {object} options - Clone options.
-   * @param {string[]} options.targetNamespace - The namespace for the target table (defaults to root namespace).
+   * @param {string[]} options.targetNamespacePath - The namespace path for the target table (defaults to root namespace).
    * @param {number} options.sourceVersion - The version of the source table to clone.
    * @param {string} options.sourceTag - The tag of the source table to clone.
    * @param {boolean} options.isShallow - Whether to perform a shallow clone (defaults to true).
@@ -288,7 +385,7 @@ export abstract class Connection {
     targetTableName: string,
     sourceUri: string,
     options?: {
-      targetNamespace?: string[];
+      targetNamespacePath?: string[];
       sourceVersion?: number;
       sourceTag?: string;
       isShallow?: boolean;
@@ -319,25 +416,25 @@ export class LocalConnection extends Connection {
   }
 
   async tableNames(
-    namespaceOrOptions?: string[] | Partial<TableNamesOptions>,
+    namespacePathOrOptions?: string[] | Partial<TableNamesOptions>,
     options?: Partial<TableNamesOptions>,
   ): Promise<string[]> {
-    // Detect if first argument is namespace array or options object
+    // Detect if first argument is namespacePath array or options object
-    let namespace: string[] | undefined;
+    let namespacePath: string[] | undefined;
     let tableNamesOptions: Partial<TableNamesOptions> | undefined;
 
-    if (Array.isArray(namespaceOrOptions)) {
+    if (Array.isArray(namespacePathOrOptions)) {
-      // First argument is namespace array
+      // First argument is namespacePath array
-      namespace = namespaceOrOptions;
+      namespacePath = namespacePathOrOptions;
       tableNamesOptions = options;
     } else {
       // First argument is options object (backwards compatibility)
-      namespace = undefined;
+      namespacePath = undefined;
-      tableNamesOptions = namespaceOrOptions;
+      tableNamesOptions = namespacePathOrOptions;
     }
 
     return this.inner.tableNames(
-      namespace ?? [],
+      namespacePath ?? [],
       tableNamesOptions?.startAfter,
       tableNamesOptions?.limit,
     );
@@ -345,12 +442,12 @@ export class LocalConnection extends Connection {
 
   async openTable(
     name: string,
-    namespace?: string[],
+    namespacePath?: string[],
     options?: Partial<OpenTableOptions>,
   ): Promise<Table> {
     const innerTable = await this.inner.openTable(
       name,
-      namespace ?? [],
+      namespacePath ?? [],
       cleanseStorageOptions(options?.storageOptions),
       options?.indexCacheSize,
     );
@@ -362,7 +459,7 @@ export class LocalConnection extends Connection {
     targetTableName: string,
     sourceUri: string,
     options?: {
-      targetNamespace?: string[];
+      targetNamespacePath?: string[];
       sourceVersion?: number;
       sourceTag?: string;
       isShallow?: boolean;
@@ -371,7 +468,7 @@ export class LocalConnection extends Connection {
     const innerTable = await this.inner.cloneTable(
       targetTableName,
       sourceUri,
-      options?.targetNamespace ?? [],
+      options?.targetNamespacePath ?? [],
       options?.sourceVersion ?? null,
       options?.sourceTag ?? null,
       options?.isShallow ?? true,
@@ -406,42 +503,42 @@ export class LocalConnection extends Connection {
     nameOrOptions:
       | string
       | ({ name: string; data: Data } & Partial<CreateTableOptions>),
-    dataOrNamespace?: Record<string, unknown>[] | TableLike | string[],
+    dataOrNamespacePath?: Record<string, unknown>[] | TableLike | string[],
-    namespaceOrOptions?: string[] | Partial<CreateTableOptions>,
+    namespacePathOrOptions?: string[] | Partial<CreateTableOptions>,
     options?: Partial<CreateTableOptions>,
   ): Promise<Table> {
     if (typeof nameOrOptions !== "string" && "name" in nameOrOptions) {
-      // First overload: createTable(options, namespace?)
+      // First overload: createTable(options, namespacePath?)
       const { name, data, ...createOptions } = nameOrOptions;
-      const namespace = dataOrNamespace as string[] | undefined;
+      const namespacePath = dataOrNamespacePath as string[] | undefined;
-      return this._createTableImpl(name, data, namespace, createOptions);
+      return this._createTableImpl(name, data, namespacePath, createOptions);
     }
 
-    // Second overload: createTable(name, data, namespace?, options?)
+    // Second overload: createTable(name, data, namespacePath?, options?)
     const name = nameOrOptions;
-    const data = dataOrNamespace as Record<string, unknown>[] | TableLike;
+    const data = dataOrNamespacePath as Record<string, unknown>[] | TableLike;
 
-    // Detect if third argument is namespace array or options object
+    // Detect if third argument is namespacePath array or options object
-    let namespace: string[] | undefined;
+    let namespacePath: string[] | undefined;
     let createOptions: Partial<CreateTableOptions> | undefined;
 
-    if (Array.isArray(namespaceOrOptions)) {
+    if (Array.isArray(namespacePathOrOptions)) {
-      // Third argument is namespace array
+      // Third argument is namespacePath array
-      namespace = namespaceOrOptions;
+      namespacePath = namespacePathOrOptions;
       createOptions = options;
     } else {
       // Third argument is options object (backwards compatibility)
-      namespace = undefined;
+      namespacePath = undefined;
-      createOptions = namespaceOrOptions;
+      createOptions = namespacePathOrOptions;
     }
 
-    return this._createTableImpl(name, data, namespace, createOptions);
+    return this._createTableImpl(name, data, namespacePath, createOptions);
   }
 
   private async _createTableImpl(
     name: string,
     data: Data,
-    namespace?: string[],
+    namespacePath?: string[],
     options?: Partial<CreateTableOptions>,
   ): Promise<Table> {
     if (data === undefined) {
@@ -455,7 +552,7 @@ export class LocalConnection extends Connection {
       name,
       buf,
       mode,
-      namespace ?? [],
+      namespacePath ?? [],
       storageOptions,
     );
 
@@ -465,21 +562,21 @@ export class LocalConnection extends Connection {
   async createEmptyTable(
     name: string,
     schema: import("./arrow").SchemaLike,
-    namespaceOrOptions?: string[] | Partial<CreateTableOptions>,
+    namespacePathOrOptions?: string[] | Partial<CreateTableOptions>,
     options?: Partial<CreateTableOptions>,
   ): Promise<Table> {
-    // Detect if third argument is namespace array or options object
+    // Detect if third argument is namespacePath array or options object
-    let namespace: string[] | undefined;
+    let namespacePath: string[] | undefined;
     let createOptions: Partial<CreateTableOptions> | undefined;
 
-    if (Array.isArray(namespaceOrOptions)) {
+    if (Array.isArray(namespacePathOrOptions)) {
-      // Third argument is namespace array
+      // Third argument is namespacePath array
-      namespace = namespaceOrOptions;
+      namespacePath = namespacePathOrOptions;
       createOptions = options;
     } else {
       // Third argument is options object (backwards compatibility)
-      namespace = undefined;
+      namespacePath = undefined;
-      createOptions = namespaceOrOptions;
+      createOptions = namespacePathOrOptions;
     }
 
     let mode: string = createOptions?.mode ?? "create";
@@ -502,18 +599,57 @@ export class LocalConnection extends Connection {
       name,
       buf,
       mode,
-      namespace ?? [],
+      namespacePath ?? [],
       storageOptions,
     );
     return new LocalTable(innerTable);
   }
 
-  async dropTable(name: string, namespace?: string[]): Promise<void> {
+  async dropTable(name: string, namespacePath?: string[]): Promise<void> {
-    return this.inner.dropTable(name, namespace ?? []);
+    return this.inner.dropTable(name, namespacePath ?? []);
   }
 
-  async dropAllTables(namespace?: string[]): Promise<void> {
+  async dropAllTables(namespacePath?: string[]): Promise<void> {
-    return this.inner.dropAllTables(namespace ?? []);
+    return this.inner.dropAllTables(namespacePath ?? []);
+  }
+
+  describeNamespace(
+    namespacePath: string[],
+  ): Promise<DescribeNamespaceResponse> {
+    return this.inner.describeNamespace(namespacePath);
+  }
+
+  listNamespaces(
+    namespacePath?: string[],
+    options?: Partial<ListNamespacesOptions>,
+  ): Promise<ListNamespacesResponse> {
+    return this.inner.listNamespaces(
+      namespacePath ?? [],
+      options?.pageToken,
+      options?.limit,
+    );
+  }
+
+  createNamespace(
+    namespacePath: string[],
+    options?: Partial<CreateNamespaceOptions>,
+  ): Promise<CreateNamespaceResponse> {
+    return this.inner.createNamespace(
+      namespacePath,
+      options?.mode,
+      options?.properties,
+    );
+  }
+
+  dropNamespace(
+    namespacePath: string[],
+    options?: Partial<DropNamespaceOptions>,
+  ): Promise<DropNamespaceResponse> {
+    return this.inner.dropNamespace(
+      namespacePath,
+      options?.mode,
+      options?.behavior,
+    );
   }
 }
 

nodejs/lancedb/index.ts

@@ -8,6 +8,7 @@ import {
 } from "./connection";
 
 import {
+  ConnectNamespaceOptions,
   ConnectionOptions,
   Connection as LanceDbConnection,
   JsHeaderProvider as NativeJsHeaderProvider,
@@ -22,6 +23,7 @@ export { JsHeaderProvider as NativeJsHeaderProvider } from "./native.js";
 export {
   AddColumnsSql,
   ConnectionOptions,
+  ConnectNamespaceOptions,
   IndexStatistics,
   IndexConfig,
   ClientConfig,
@@ -62,6 +64,13 @@ export {
   CreateTableOptions,
   TableNamesOptions,
   OpenTableOptions,
+  ListNamespacesOptions,
+  CreateNamespaceOptions,
+  DropNamespaceOptions,
+  ListNamespacesResponse,
+  CreateNamespaceResponse,
+  DropNamespaceResponse,
+  DescribeNamespaceResponse,
 } from "./connection";
 
 export { Session } from "./native.js";
@@ -293,3 +302,197 @@ export async function connect(
   );
   return new LocalConnection(nativeConn);
 }
+
+/**
+ * Configuration for the built-in directory namespace (`"dir"`).
+ *
+ * The directory namespace stores tables under a single root path (local
+ * filesystem or object storage URI). See
+ * {@link https://docs.lancedb.com/namespaces} for the documented surface;
+ * less-common knobs live under {@link DirNamespaceConfig.extraProperties}.
+ */
+export interface DirNamespaceConfig {
+  /** Root path or URI containing the LanceDB tables. */
+  root: string;
+  /**
+   * Whether to maintain a namespace manifest at the root. Required for
+   * child namespaces. Defaults to true on the impl side.
+   */
+  manifestEnabled?: boolean;
+  /**
+   * Additional raw properties passed verbatim to the namespace
+   * implementation (e.g. `storage.*`, `credential_vendor.*`). Typed
+   * fields above take precedence on key collision.
+   */
+  extraProperties?: Record<string, string>;
+}
+
+/**
+ * Configuration for the built-in REST namespace (`"rest"`).
+ *
+ * The REST namespace talks to a remote catalog server over HTTP. See
+ * {@link https://docs.lancedb.com/namespaces} for the documented surface;
+ * less-common knobs (TLS, metrics) live under
+ * {@link RestNamespaceConfig.extraProperties}.
+ */
+export interface RestNamespaceConfig {
+  /** Catalog endpoint URL. */
+  uri: string;
+  /**
+   * HTTP headers forwarded with each request. Keys are passed through
+   * as-is (e.g. `"x-api-key"`, `"Authorization"`).
+   */
+  headers?: Record<string, string>;
+  /**
+   * Additional raw properties passed verbatim to the namespace
+   * implementation (e.g. `tls.*`, `ops_metrics_enabled`, `delimiter`).
+   * Typed fields above take precedence on key collision.
+   */
+  extraProperties?: Record<string, string>;
+}
+
+function dirConfigToProperties(
+  config: DirNamespaceConfig,
+): Record<string, string> {
+  // Spread the whole input so that unknown keys (e.g. a raw `manifest_enabled`
+  // passed via the dynamic-impl path) flow through instead of being dropped.
+  // Typed transformations layer on top.
+  const { manifestEnabled, extraProperties, ...rest } = config;
+  const properties: Record<string, string> = {
+    ...(extraProperties ?? {}),
+    ...(rest as Record<string, string>),
+  };
+  if (manifestEnabled !== undefined) {
+    properties.manifest_enabled = String(manifestEnabled);
+  }
+  return properties;
+}
+
+function restConfigToProperties(
+  config: RestNamespaceConfig,
+): Record<string, string> {
+  const { headers, extraProperties, ...rest } = config;
+  const properties: Record<string, string> = {
+    ...(extraProperties ?? {}),
+    ...(rest as Record<string, string>),
+  };
+  if (headers) {
+    for (const [name, value] of Object.entries(headers)) {
+      properties[`headers.${name}`] = value;
+    }
+  }
+  return properties;
+}
+
+/**
+ * Connect to a LanceDB database through a namespace.
+ *
+ * Unlike {@link connect}, which routes by URI scheme (local path vs.
+ * `db://` cloud), `connectNamespace` always returns a namespace-backed
+ * connection. The `implName` selects the namespace implementation:
+ *
+ * - `"dir"` — directory namespace, configured with {@link DirNamespaceConfig}.
+ * - `"rest"` — remote REST catalog, configured with {@link RestNamespaceConfig}.
+ * - Any other string — full module path for a custom implementation,
+ *   configured with a free-form string-keyed `properties` map.
+ *
+ * @example Typed dir namespace
+ * ```ts
+ * const db = await connectNamespace("dir", { root: "/path/to/db" });
+ * await db.createTable("users", [{ id: 1 }]);
+ * ```
+ *
+ * @example Typed REST namespace with auth headers
+ * ```ts
+ * const db = await connectNamespace("rest", {
+ *   uri: "https://catalog.example.com",
+ *   headers: { "x-api-key": process.env.CATALOG_KEY ?? "" },
+ * });
+ * ```
+ *
+ * @example Custom implementation with raw properties
+ * ```ts
+ * const db = await connectNamespace("my.custom.Namespace", {
+ *   endpoint: "...",
+ * });
+ * ```
+ */
+export function connectNamespace(
+  implName: "dir",
+  config: DirNamespaceConfig,
+  options?: Partial<ConnectNamespaceOptions>,
+): Promise<Connection>;
+/**
+ * Connect through the built-in REST namespace.
+ *
+ * Configured with {@link RestNamespaceConfig}. See the function-level
+ * documentation above for the full surface, examples, and how this
+ * relates to {@link connect}.
+ *
+ * @example
+ * ```ts
+ * const db = await connectNamespace("rest", {
+ *   uri: "https://catalog.example.com",
+ *   headers: { "x-api-key": process.env.CATALOG_KEY ?? "" },
+ * });
+ * ```
+ */
+export function connectNamespace(
+  implName: "rest",
+  config: RestNamespaceConfig,
+  options?: Partial<ConnectNamespaceOptions>,
+): Promise<Connection>;
+/**
+ * Connect through a custom namespace implementation by full module path,
+ * configured with a free-form string-keyed `properties` map. Use the
+ * typed overloads above for the built-in `"dir"` and `"rest"` impls.
+ *
+ * See the function-level documentation above for examples and how this
+ * relates to {@link connect}.
+ *
+ * @example
+ * ```ts
+ * const db = await connectNamespace("my.custom.Namespace", {
+ *   endpoint: "...",
+ * });
+ * ```
+ */
+export function connectNamespace(
+  implName: string,
+  properties: Record<string, string>,
+  options?: Partial<ConnectNamespaceOptions>,
+): Promise<Connection>;
+export async function connectNamespace(
+  implName: string,
+  configOrProperties:
+    | DirNamespaceConfig
+    | RestNamespaceConfig
+    | Record<string, string>,
+  options?: Partial<ConnectNamespaceOptions>,
+): Promise<Connection> {
+  let properties: Record<string, string>;
+  if (implName === "dir") {
+    properties = dirConfigToProperties(
+      configOrProperties as DirNamespaceConfig,
+    );
+  } else if (implName === "rest") {
+    properties = restConfigToProperties(
+      configOrProperties as RestNamespaceConfig,
+    );
+  } else {
+    properties = configOrProperties as Record<string, string>;
+  }
+
+  const finalOptions: ConnectNamespaceOptions = (options ??
+    {}) as ConnectNamespaceOptions;
+  finalOptions.storageOptions = cleanseStorageOptions(
+    finalOptions.storageOptions,
+  );
+
+  const nativeConn = await LanceDbConnection.newWithNamespace(
+    implName,
+    properties,
+    finalOptions,
+  );
+  return new LocalConnection(nativeConn);
+}

nodejs/lancedb/merge.ts

@@ -87,6 +87,23 @@ export class MergeInsertBuilder {
       this.#schema,
     );
   }
+
+  /**
+   * Skip the automatic cleanup of old dataset versions that would otherwise
+   * run as part of this merge insert's commit. Forwards to
+   * `MergeInsertBuilder::skip_auto_cleanup` in lance-core.
+   *
+   * Useful for high-frequency writers that prefer to manage version cleanup
+   * themselves, or writers without delete permissions on the underlying storage.
+   *
+   * @param skip - If true, the auto-cleanup step is skipped at commit time.
+   */
+  skipAutoCleanup(skip: boolean): MergeInsertBuilder {
+    return new MergeInsertBuilder(
+      this.#native.skipAutoCleanup(skip),
+      this.#schema,
+    );
+  }
   /**
    * Executes the merge insert operation
    *

nodejs/lancedb/table.ts

@@ -56,6 +56,18 @@ export interface AddDataOptions {
    * If "overwrite" then the new data will replace the existing data in the table.
    */
   mode: "append" | "overwrite";
+  /**
+   * If true, skip the automatic cleanup of old dataset versions that would
+   * otherwise run as part of this write's commit. Forwards to
+   * `WriteParams.skip_auto_cleanup` in lance-core.
+   *
+   * Useful for high-frequency writers that prefer to manage version cleanup
+   * themselves (for example, via a separate periodic optimize job), or for
+   * writers that don't have delete permissions on the underlying storage.
+   *
+   * Defaults to false.
+   */
+  skipAutoCleanup?: boolean;
 }
 
 export interface UpdateOptions {
@@ -285,6 +297,25 @@ export abstract class Table {
    */
   abstract prewarmIndex(name: string): Promise<void>;
 
+  /**
+   * Prewarm one or more columns of data in the table.
+   *
+   * @param columns The columns to prewarm. If undefined, all columns are prewarmed.
+   *
+   * This will load the column data into the page cache so that future queries that
+   * read those columns avoid the initial cold-start latency.  This call initiates
+   * prewarming and returns once the request is accepted; the warming itself may
+   * continue in the background.  Calling it on already-prewarmed columns is a
+   * no-op on the server.
+   *
+   * Prewarming is generally useful for columns used in filters or projections.
+   * Large columns (e.g. high-dimensional vectors or binary data) may not be
+   * practical to prewarm.
+   *
+   * This feature is currently only supported on remote tables.
+   */
+  abstract prewarmData(columns?: string[]): Promise<void>;
+
   /**
    * Waits for asynchronous indexing to complete on the table.
    *
@@ -617,7 +648,7 @@ export class LocalTable extends Table {
     const schema = await this.schema();
 
     const buffer = await fromDataToBuffer(data, undefined, schema);
-    return await this.inner.add(buffer, mode);
+    return await this.inner.add(buffer, mode, options?.skipAutoCleanup);
   }
 
   async update(
@@ -710,6 +741,10 @@ export class LocalTable extends Table {
     await this.inner.prewarmIndex(name);
   }
 
+  async prewarmData(columns?: string[]): Promise<void> {
+    await this.inner.prewarmData(columns);
+  }
+
   async waitForIndex(
     indexNames: string[],
     timeoutSeconds: number,

nodejs/npm/darwin-arm64/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-darwin-arm64",
-	"version": "0.27.2-beta.1",
+	"version": "0.28.0-beta.11",
 	"os": ["darwin"],
 	"cpu": ["arm64"],
 	"main": "lancedb.darwin-arm64.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-gnu",
-	"version": "0.27.2-beta.1",
+	"version": "0.28.0-beta.11",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-gnu.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-musl",
-	"version": "0.27.2-beta.1",
+	"version": "0.28.0-beta.11",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-musl.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-gnu",
-	"version": "0.27.2-beta.1",
+	"version": "0.28.0-beta.11",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-gnu.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-musl",
-	"version": "0.27.2-beta.1",
+	"version": "0.28.0-beta.11",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-musl.node",

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

@@ -1,6 +1,6 @@
 {
   "name": "@lancedb/lancedb-win32-arm64-msvc",
-  "version": "0.27.2-beta.1",
+  "version": "0.28.0-beta.11",
   "os": [
     "win32"
   ],

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-win32-x64-msvc",
-	"version": "0.27.2-beta.1",
+	"version": "0.28.0-beta.11",
 	"os": ["win32"],
 	"cpu": ["x64"],
 	"main": "lancedb.win32-x64-msvc.node",

nodejs/package.json

@@ -11,7 +11,7 @@
     "ann"
   ],
   "private": false,
-  "version": "0.27.2-beta.1",
+  "version": "0.28.0-beta.11",
   "main": "dist/index.js",
   "exports": {
     ".": "./dist/index.js",
@@ -38,15 +38,15 @@
     "url": "https://github.com/lancedb/lancedb"
   },
   "devDependencies": {
-    "@aws-sdk/client-dynamodb": "^3.33.0",
+    "@aws-sdk/client-dynamodb": "3.1003.0",
-    "@aws-sdk/client-kms": "^3.33.0",
+    "@aws-sdk/client-kms": "3.1003.0",
-    "@aws-sdk/client-s3": "^3.33.0",
+    "@aws-sdk/client-s3": "3.1003.0",
     "@biomejs/biome": "^1.7.3",
     "@jest/globals": "^29.7.0",
-    "@napi-rs/cli": "^3.5.1",
+    "@napi-rs/cli": "3.5.1",
     "@types/axios": "^0.14.0",
     "@types/jest": "^29.1.2",
-    "@types/node": "^22.7.4",
+    "@types/node": "22.7.4",
     "@types/tmp": "^0.2.6",
     "apache-arrow-15": "npm:apache-arrow@15.0.0",
     "apache-arrow-16": "npm:apache-arrow@16.0.0",
@@ -57,9 +57,9 @@
     "shx": "^0.3.4",
     "tmp": "^0.2.3",
     "ts-jest": "^29.1.2",
-    "typedoc": "^0.26.4",
+    "typedoc": "0.26.4",
-    "typedoc-plugin-markdown": "^4.2.1",
+    "typedoc-plugin-markdown": "4.2.1",
-    "typescript": "^5.5.4",
+    "typescript": "5.5.4",
     "typescript-eslint": "^7.1.0"
   },
   "ava": {
@@ -68,16 +68,16 @@
   "engines": {
     "node": ">= 18"
   },
+  "packageManager": "pnpm@11.1.1",
   "cpu": ["x64", "arm64"],
   "os": ["darwin", "linux", "win32"],
   "scripts": {
     "artifacts": "napi artifacts",
     "build:debug": "napi build --platform --dts ../lancedb/native.d.ts --js ../lancedb/native.js --output-dir lancedb",
-    "postbuild:debug": "shx mkdir -p dist && shx cp lancedb/*.node dist/",
+    "postbuild:debug": "shx mkdir -p dist && shx cp lancedb/*.node dist/ && node -e \"require('fs').writeFileSync('dist/package.json', JSON.stringify({name:'@lancedb/lancedb',type:'commonjs'}))\"",
     "build:release": "napi build --platform --release --dts ../lancedb/native.d.ts --js ../lancedb/native.js --output-dir dist",
-    "postbuild:release": "shx mkdir -p dist && shx cp lancedb/*.node dist/",
+    "build": "pnpm build:debug && pnpm tsc",
-    "build": "npm run build:debug && npm run tsc",
+    "build-release": "pnpm build:release && pnpm tsc",
-    "build-release": "npm run build:release && npm run tsc",
     "tsc": "tsc -b",
     "posttsc": "shx cp lancedb/native.d.ts dist/native.d.ts",
     "lint-ci": "biome ci .",
@@ -87,7 +87,7 @@
     "lint-fix": "biome check --write . && biome format --write .",
     "prepublishOnly": "napi prepublish -t npm",
     "test": "jest --verbose",
-    "integration": "S3_TEST=1 npm run test",
+    "integration": "S3_TEST=1 pnpm test",
     "universal": "napi universalize",
     "version": "napi version"
   },
@@ -95,8 +95,8 @@
     "reflect-metadata": "^0.2.2"
   },
   "optionalDependencies": {
-    "@huggingface/transformers": "^3.0.2",
+    "@huggingface/transformers": "3.0.2",
-    "openai": "^4.29.2"
+    "openai": "4.29.2"
   },
   "peerDependencies": {
     "apache-arrow": ">=15.0.0 <=18.1.0"

nodejs/pnpm-workspace.yaml

@@ -0,0 +1,18 @@
+# Flat node_modules layout. The @napi-rs/cli build step fails to locate
+# the cdylib artifact under pnpm's isolated layout; the hoisted linker
+# mirrors npm's structure and unblocks the native build.
+nodeLinker: hoisted
+
+# Block resolution of versions less than 24h old (Shai-Hulud window).
+# This is the pnpm 11 default but pinned here so it's visible to
+# reviewers and survives a future pnpm major flipping the default.
+minimumReleaseAge: 1440
+
+# Fail install if a transitive dep tries to run an unapproved script.
+strictDepBuilds: true
+
+allowBuilds:
+  '@biomejs/biome': true
+  onnxruntime-node: true
+  protobufjs: true
+  sharp: true

nodejs/src/connection.rs

@@ -8,12 +8,16 @@ use lancedb::database::{CreateTableMode, Database};
 use napi::bindgen_prelude::*;
 use napi_derive::*;
 
+use crate::ConnectNamespaceOptions;
 use crate::ConnectionOptions;
 use crate::error::NapiErrorExt;
 use crate::header::JsHeaderProvider;
 use crate::table::Table;
-use lancedb::connection::{ConnectBuilder, Connection as LanceDBConnection};
+use lancedb::connection::{ConnectBuilder, Connection as LanceDBConnection, connect_namespace};
 
+use lance_namespace::models::{
+    CreateNamespaceRequest, DescribeNamespaceRequest, DropNamespaceRequest, ListNamespacesRequest,
+};
 use lancedb::ipc::{ipc_file_to_batches, ipc_file_to_schema};
 
 #[napi]
@@ -21,6 +25,29 @@ pub struct Connection {
     inner: Option<LanceDBConnection>,
 }
 
+#[napi(object)]
+pub struct DescribeNamespaceResponse {
+    pub properties: Option<HashMap<String, String>>,
+}
+
+#[napi(object)]
+pub struct ListNamespacesResponse {
+    pub namespaces: Vec<String>,
+    pub page_token: Option<String>,
+}
+
+#[napi(object)]
+pub struct CreateNamespaceResponse {
+    pub properties: Option<HashMap<String, String>>,
+    pub transaction_id: Option<String>,
+}
+
+#[napi(object)]
+pub struct DropNamespaceResponse {
+    pub properties: Option<HashMap<String, String>>,
+    pub transaction_id: Option<Vec<String>>,
+}
+
 impl Connection {
     pub(crate) fn inner_new(inner: LanceDBConnection) -> Self {
         Self { inner: Some(inner) }
@@ -67,6 +94,12 @@ impl Connection {
                 builder = builder.storage_option(key, value);
             }
         }
+        if let Some(manifest_enabled) = options.manifest_enabled {
+            builder = builder.manifest_enabled(manifest_enabled);
+        }
+        if let Some(namespace_client_properties) = options.namespace_client_properties {
+            builder = builder.namespace_client_properties(namespace_client_properties);
+        }
 
         // Create client config, optionally with header provider
         let client_config = options.client_config.unwrap_or_default();
@@ -100,6 +133,39 @@ impl Connection {
         Ok(Self::inner_new(builder.execute().await.default_error()?))
     }
 
+    /// Create a new Connection instance backed by a namespace implementation.
+    #[napi(factory)]
+    pub async fn new_with_namespace(
+        impl_name: String,
+        properties: HashMap<String, String>,
+        options: ConnectNamespaceOptions,
+    ) -> napi::Result<Self> {
+        if impl_name.is_empty() {
+            return Err(napi::Error::from_reason(
+                "implName must be a non-empty string",
+            ));
+        }
+
+        let mut builder = connect_namespace(&impl_name, properties);
+        if let Some(interval) = options.read_consistency_interval {
+            builder =
+                builder.read_consistency_interval(std::time::Duration::from_secs_f64(interval));
+        }
+        if let Some(storage_options) = options.storage_options {
+            for (key, value) in storage_options {
+                builder = builder.storage_option(key, value);
+            }
+        }
+        if let Some(namespace_client_properties) = options.namespace_client_properties {
+            builder = builder.namespace_client_properties(namespace_client_properties);
+        }
+        if let Some(session) = options.session {
+            builder = builder.session(session.inner.clone());
+        }
+
+        Ok(Self::inner_new(builder.execute().await.default_error()?))
+    }
+
     #[napi]
     pub fn display(&self) -> napi::Result<String> {
         Ok(self.get_inner()?.to_string())
@@ -119,12 +185,12 @@ impl Connection {
     #[napi(catch_unwind)]
     pub async fn table_names(
         &self,
-        namespace: Vec<String>,
+        namespace_path: Option<Vec<String>>,
         start_after: Option<String>,
         limit: Option<u32>,
     ) -> napi::Result<Vec<String>> {
         let mut op = self.get_inner()?.table_names();
-        op = op.namespace(namespace);
+        op = op.namespace(namespace_path.unwrap_or_default());
         if let Some(start_after) = start_after {
             op = op.start_after(start_after);
         }
@@ -146,7 +212,7 @@ impl Connection {
         name: String,
         buf: Buffer,
         mode: String,
-        namespace: Vec<String>,
+        namespace_path: Option<Vec<String>>,
         storage_options: Option<HashMap<String, String>>,
     ) -> napi::Result<Table> {
         let batches = ipc_file_to_batches(buf.to_vec())
@@ -154,7 +220,7 @@ impl Connection {
         let mode = Self::parse_create_mode_str(&mode)?;
         let mut builder = self.get_inner()?.create_table(&name, batches).mode(mode);
 
-        builder = builder.namespace(namespace);
+        builder = builder.namespace(namespace_path.unwrap_or_default());
 
         if let Some(storage_options) = storage_options {
             for (key, value) in storage_options {
@@ -171,7 +237,7 @@ impl Connection {
         name: String,
         schema_buf: Buffer,
         mode: String,
-        namespace: Vec<String>,
+        namespace_path: Option<Vec<String>>,
         storage_options: Option<HashMap<String, String>>,
     ) -> napi::Result<Table> {
         let schema = ipc_file_to_schema(schema_buf.to_vec()).map_err(|e| {
@@ -183,7 +249,7 @@ impl Connection {
             .create_empty_table(&name, schema)
             .mode(mode);
 
-        builder = builder.namespace(namespace);
+        builder = builder.namespace(namespace_path.unwrap_or_default());
 
         if let Some(storage_options) = storage_options {
             for (key, value) in storage_options {
@@ -198,13 +264,13 @@ impl Connection {
     pub async fn open_table(
         &self,
         name: String,
-        namespace: Vec<String>,
+        namespace_path: Option<Vec<String>>,
         storage_options: Option<HashMap<String, String>>,
         index_cache_size: Option<u32>,
     ) -> napi::Result<Table> {
         let mut builder = self.get_inner()?.open_table(&name);
 
-        builder = builder.namespace(namespace);
+        builder = builder.namespace(namespace_path.unwrap_or_default());
 
         if let Some(storage_options) = storage_options {
             for (key, value) in storage_options {
@@ -223,7 +289,7 @@ impl Connection {
         &self,
         target_table_name: String,
         source_uri: String,
-        target_namespace: Vec<String>,
+        target_namespace_path: Option<Vec<String>>,
         source_version: Option<i64>,
         source_tag: Option<String>,
         is_shallow: bool,
@@ -232,7 +298,7 @@ impl Connection {
             .get_inner()?
             .clone_table(&target_table_name, &source_uri);
 
-        builder = builder.target_namespace(target_namespace);
+        builder = builder.target_namespace(target_namespace_path.unwrap_or_default());
 
         if let Some(version) = source_version {
             builder = builder.source_version(version as u64);
@@ -250,18 +316,147 @@ impl Connection {
 
     /// Drop table with the name. Or raise an error if the table does not exist.
     #[napi(catch_unwind)]
-    pub async fn drop_table(&self, name: String, namespace: Vec<String>) -> napi::Result<()> {
+    pub async fn drop_table(
+        &self,
+        name: String,
+        namespace_path: Option<Vec<String>>,
+    ) -> napi::Result<()> {
+        let ns = namespace_path.unwrap_or_default();
         self.get_inner()?
-            .drop_table(&name, &namespace)
+            .drop_table(&name, &ns)
             .await
             .default_error()
     }
 
     #[napi(catch_unwind)]
-    pub async fn drop_all_tables(&self, namespace: Vec<String>) -> napi::Result<()> {
+    pub async fn drop_all_tables(&self, namespace_path: Option<Vec<String>>) -> napi::Result<()> {
-        self.get_inner()?
+        let ns = namespace_path.unwrap_or_default();
-            .drop_all_tables(&namespace)
+        self.get_inner()?.drop_all_tables(&ns).await.default_error()
+    }
+
+    #[napi(catch_unwind)]
+    /// Describe a namespace and return its properties.
+    pub async fn describe_namespace(
+        &self,
+        namespace_path: Vec<String>,
+    ) -> napi::Result<DescribeNamespaceResponse> {
+        let req = DescribeNamespaceRequest {
+            id: Some(namespace_path),
+            ..Default::default()
+        };
+        let resp = self
+            .get_inner()?
+            .describe_namespace(req)
             .await
-            .default_error()
+            .default_error()?;
+        Ok(DescribeNamespaceResponse {
+            properties: resp.properties,
+        })
+    }
+
+    #[napi(catch_unwind)]
+    /// List child namespaces under the given namespace path
+    pub async fn list_namespaces(
+        &self,
+        namespace_path: Option<Vec<String>>,
+        page_token: Option<String>,
+        limit: Option<u32>,
+    ) -> napi::Result<ListNamespacesResponse> {
+        let req = ListNamespacesRequest {
+            id: namespace_path,
+            page_token,
+            limit: limit.map(|l| l as i32),
+            ..Default::default()
+        };
+        let resp = self
+            .get_inner()?
+            .list_namespaces(req)
+            .await
+            .default_error()?;
+        Ok(ListNamespacesResponse {
+            namespaces: resp.namespaces,
+            page_token: resp.page_token,
+        })
+    }
+
+    #[napi(catch_unwind)]
+    /// Create a new namespace with optional properties.
+    pub async fn create_namespace(
+        &self,
+        namespace_path: Vec<String>,
+        mode: Option<String>,
+        properties: Option<HashMap<String, String>>,
+    ) -> napi::Result<CreateNamespaceResponse> {
+        let mode_str = mode
+            .map(|m| match m.to_lowercase().as_str() {
+                "create" => Ok("Create".to_string()),
+                "exist_ok" => Ok("ExistOk".to_string()),
+                "overwrite" => Ok("Overwrite".to_string()),
+                _ => Err(napi::Error::from_reason(format!(
+                    "Invalid mode '{}': expected one of 'create', 'exist_ok', 'overwrite'",
+                    m
+                ))),
+            })
+            .transpose()?;
+        let req = CreateNamespaceRequest {
+            id: Some(namespace_path),
+            mode: mode_str,
+            properties,
+            ..Default::default()
+        };
+        let resp = self
+            .get_inner()?
+            .create_namespace(req)
+            .await
+            .default_error()?;
+        Ok(CreateNamespaceResponse {
+            properties: resp.properties,
+            transaction_id: resp.transaction_id,
+        })
+    }
+
+    #[napi(catch_unwind)]
+    /// Drop a namespace.
+    pub async fn drop_namespace(
+        &self,
+        namespace_path: Vec<String>,
+        mode: Option<String>,
+        behavior: Option<String>,
+    ) -> napi::Result<DropNamespaceResponse> {
+        let mode_str = mode
+            .map(|m| match m.to_lowercase().as_str() {
+                "skip" => Ok("Skip".to_string()),
+                "fail" => Ok("Fail".to_string()),
+                _ => Err(napi::Error::from_reason(format!(
+                    "Invalid mode '{}': expected one of 'skip', 'fail'",
+                    m
+                ))),
+            })
+            .transpose()?;
+        let behavior_str = behavior
+            .map(|b| match b.to_lowercase().as_str() {
+                "restrict" => Ok("Restrict".to_string()),
+                "cascade" => Ok("Cascade".to_string()),
+                _ => Err(napi::Error::from_reason(format!(
+                    "Invalid behavior '{}': expected one of 'restrict', 'cascade'",
+                    b
+                ))),
+            })
+            .transpose()?;
+        let req = DropNamespaceRequest {
+            id: Some(namespace_path),
+            mode: mode_str,
+            behavior: behavior_str,
+            ..Default::default()
+        };
+        let resp = self
+            .get_inner()?
+            .drop_namespace(req)
+            .await
+            .default_error()?;
+        Ok(DropNamespaceResponse {
+            properties: resp.properties,
+            transaction_id: resp.transaction_id,
+        })
     }
 }

nodejs/src/lib.rs

@@ -35,8 +35,15 @@ pub struct ConnectionOptions {
     pub read_consistency_interval: Option<f64>,
     /// (For LanceDB OSS only): configuration for object storage.
     ///
-    /// The available options are described at https://lancedb.com/docs/storage/
+    /// The available options are described at https://docs.lancedb.com/storage/
     pub storage_options: Option<HashMap<String, String>>,
+    /// (For LanceDB OSS only): use directory namespace manifests as the source
+    /// of truth for table metadata. Existing directory-listed root tables are
+    /// migrated into the manifest on access.
+    pub manifest_enabled: Option<bool>,
+    /// (For LanceDB OSS only): extra properties for the backing namespace
+    /// client used by manifest-enabled native connections.
+    pub namespace_client_properties: Option<HashMap<String, String>>,
     /// (For LanceDB OSS only): the session to use for this connection. Holds
     /// shared caches and other session-specific state.
     pub session: Option<session::Session>,
@@ -60,6 +67,26 @@ pub struct OpenTableOptions {
     pub storage_options: Option<HashMap<String, String>>,
 }
 
+#[napi(object)]
+#[derive(Debug)]
+pub struct ConnectNamespaceOptions {
+    /// The interval, in seconds, at which to check for updates to the table
+    /// from other processes. If None, then consistency is not checked. For
+    /// performance reasons, this is the default. For strong consistency, set
+    /// this to zero seconds. Then every read will check for updates from other
+    /// processes. As a compromise, you can set this to a non-zero value for
+    /// eventual consistency.
+    pub read_consistency_interval: Option<f64>,
+    /// Configuration for object storage. The available options are described
+    /// at https://docs.lancedb.com/storage/
+    pub storage_options: Option<HashMap<String, String>>,
+    /// Extra properties for the backing namespace client.
+    pub namespace_client_properties: Option<HashMap<String, String>>,
+    /// The session to use for this connection. Holds shared caches and other
+    /// session-specific state.
+    pub session: Option<session::Session>,
+}
+
 #[napi_derive::module_init]
 fn init() {
     let env = Env::new()

nodejs/src/merge.rs

@@ -50,6 +50,13 @@ impl NativeMergeInsertBuilder {
         this
     }
 
+    #[napi]
+    pub fn skip_auto_cleanup(&self, skip: bool) -> Self {
+        let mut this = self.clone();
+        this.inner.skip_auto_cleanup(skip);
+        this
+    }
+
     #[napi(catch_unwind)]
     pub async fn execute(&self, buf: Buffer) -> napi::Result<MergeResult> {
         let data = ipc_file_to_batches(buf.to_vec())

nodejs/src/remote.rs

@@ -92,6 +92,13 @@ pub struct ClientConfig {
     pub extra_headers: Option<HashMap<String, String>>,
     pub id_delimiter: Option<String>,
     pub tls_config: Option<TlsConfig>,
+    /// User identifier for tracking purposes.
+    ///
+    /// This is sent as the `x-lancedb-user-id` header in requests to LanceDB Cloud/Enterprise.
+    /// It can be set directly, or via the `LANCEDB_USER_ID` environment variable.
+    /// Alternatively, set `LANCEDB_USER_ID_ENV_KEY` to specify another environment
+    /// variable that contains the user ID value.
+    pub user_id: Option<String>,
 }
 
 impl From<TimeoutConfig> for lancedb::remote::TimeoutConfig {
@@ -145,6 +152,7 @@ impl From<ClientConfig> for lancedb::remote::ClientConfig {
             id_delimiter: config.id_delimiter,
             tls_config: config.tls_config.map(Into::into),
             header_provider: None, // the header provider is set separately later
+            user_id: config.user_id,
         }
     }
 }

nodejs/src/rerankers.rs

@@ -18,6 +18,7 @@ type RerankHybridFn = ThreadsafeFunction<
     RerankHybridCallbackArgs,
     Status,
     false,
+    true,
 >;
 
 /// Reranker implementation that "wraps" a NodeJS Reranker implementation.
@@ -32,7 +33,10 @@ impl Reranker {
     pub fn new(
         rerank_hybrid: Function<RerankHybridCallbackArgs, Promise<Buffer>>,
     ) -> napi::Result<Self> {
-        let rerank_hybrid = rerank_hybrid.build_threadsafe_function().build()?;
+        let rerank_hybrid = rerank_hybrid
+            .build_threadsafe_function()
+            .weak::<true>()
+            .build()?;
         Ok(Self { rerank_hybrid })
     }
 }

nodejs/src/table.rs

@@ -6,7 +6,7 @@ use std::collections::HashMap;
 use lancedb::ipc::{ipc_file_to_batches, ipc_file_to_schema};
 use lancedb::table::{
     AddDataMode, ColumnAlteration as LanceColumnAlteration, Duration, NewColumnTransform,
-    OptimizeAction, OptimizeOptions, Table as LanceDbTable,
+    OptimizeAction, OptimizeOptions, Table as LanceDbTable, WriteOptions,
 };
 use napi::bindgen_prelude::*;
 use napi_derive::napi;
@@ -68,7 +68,12 @@ impl Table {
     }
 
     #[napi(catch_unwind)]
-    pub async fn add(&self, buf: Buffer, mode: String) -> napi::Result<AddResult> {
+    pub async fn add(
+        &self,
+        buf: Buffer,
+        mode: String,
+        skip_auto_cleanup: Option<bool>,
+    ) -> napi::Result<AddResult> {
         let batches = ipc_file_to_batches(buf.to_vec())
             .map_err(|e| napi::Error::from_reason(format!("Failed to read IPC file: {}", e)))?;
         let batches = batches
@@ -92,6 +97,13 @@ impl Table {
             return Err(napi::Error::from_reason(format!("Invalid mode: {}", mode)));
         };
 
+        if skip_auto_cleanup.unwrap_or(false) {
+            op = op.write_options(WriteOptions {
+                skip_auto_cleanup: true,
+                ..Default::default()
+            });
+        }
+
         let res = op.execute().await.default_error()?;
         Ok(res.into())
     }
@@ -159,6 +171,14 @@ impl Table {
             .default_error()
     }
 
+    #[napi(catch_unwind)]
+    pub async fn prewarm_data(&self, columns: Option<Vec<String>>) -> napi::Result<()> {
+        self.inner_ref()?
+            .prewarm_data(columns)
+            .await
+            .default_error()
+    }
+
     #[napi(catch_unwind)]
     pub async fn wait_for_index(&self, index_names: Vec<String>, timeout_s: i64) -> Result<()> {
         let timeout = std::time::Duration::from_secs(timeout_s.try_into().unwrap());

python/.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.30.2"
+current_version = "0.31.0-beta.11"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

python/Cargo.toml

@@ -1,6 +1,7 @@
 [package]
 name = "lancedb-python"
-version = "0.30.2"
+version = "0.31.0-beta.11"
+publish = false
 edition.workspace = true
 description = "Python bindings for LanceDB"
 license.workspace = true
@@ -14,7 +15,7 @@ name = "_lancedb"
 crate-type = ["cdylib"]
 
 [dependencies]
-arrow = { version = "57.2", features = ["pyarrow"] }
+arrow = { version = "58.0.0", features = ["pyarrow"] }
 async-trait = "0.1"
 bytes = "1"
 lancedb = { path = "../rust/lancedb", default-features = false }
@@ -24,8 +25,8 @@ lance-namespace-impls.workspace = true
 lance-io.workspace = true
 env_logger.workspace = true
 log.workspace = true
-pyo3 = { version = "0.26", features = ["extension-module", "abi3-py39"] }
+pyo3 = { version = "0.28", features = ["extension-module", "abi3-py39"] }
-pyo3-async-runtimes = { version = "0.26", features = [
+pyo3-async-runtimes = { version = "0.28", features = [
     "attributes",
     "tokio-runtime",
 ] }
@@ -34,10 +35,11 @@ futures.workspace = true
 serde = "1"
 serde_json = "1"
 snafu.workspace = true
-tokio = { version = "1.40", features = ["sync"] }
+tokio = { version = "1.40", features = ["sync", "rt-multi-thread"] }
+libc = "0.2"
 
 [build-dependencies]
-pyo3-build-config = { version = "0.26", features = [
+pyo3-build-config = { version = "0.28", features = [
     "extension-module",
     "abi3-py39",
 ] }

python/PYTHON_THIRD_PARTY_LICENSES.md

@@ -183,7 +183,6 @@
 | stack-data                     | 0.6.3           | MIT License                                                                                      | http://github.com/alexmojaki/stack_data                                                           |
 | sympy                          | 1.14.0          | BSD License                                                                                      | https://sympy.org                                                                                 |
 | tabulate                       | 0.9.0           | MIT License                                                                                      | https://github.com/astanin/python-tabulate                                                        |
-| tantivy                        | 0.25.1          | UNKNOWN                                                                                          | UNKNOWN                                                                                           |
 | threadpoolctl                  | 3.6.0           | BSD License                                                                                      | https://github.com/joblib/threadpoolctl                                                           |
 | timm                           | 1.0.24          | Apache Software License                                                                          | https://github.com/huggingface/pytorch-image-models                                               |
 | tinycss2                       | 1.4.0           | BSD License                                                                                      | https://www.courtbouillon.org/tinycss2                                                            |

python/pyproject.toml

@@ -45,7 +45,7 @@ repository = "https://github.com/lancedb/lancedb"
 
 [project.optional-dependencies]
 pylance = [
-    "pylance>=4.0.0b7",
+    "pylance>=5.0.0b5",
 ]
 tests = [
     "aiohttp>=3.9.0",
@@ -57,9 +57,8 @@ tests = [
     "duckdb>=0.9.0",
     "pytz>=2023.3",
     "polars>=0.19, <=1.3.0",
-    "tantivy>=0.20.0",
     "pyarrow-stubs>=16.0",
-    "pylance>=4.0.0b7",
+    "pylance>=5.0.0b5",
     "requests>=2.31.0",
     "datafusion>=52,<53",
 ]
@@ -83,7 +82,7 @@ embeddings = [
     "colpali-engine>=0.3.10",
     "huggingface_hub>=0.19.0",
     "InstructorEmbedding>=1.0.1",
-    "google.generativeai>=0.3.0",
+    "google-genai>=1.0.0",
     "boto3>=1.28.57",
     "awscli>=1.44.38",
     "botocore>=1.31.57",

python/python/lancedb/__init__.py

@@ -6,8 +6,7 @@ import importlib.metadata
 import os
 from concurrent.futures import ThreadPoolExecutor
 from datetime import timedelta
-from typing import Dict, Optional, Union, Any
+from typing import Dict, Optional, Union, Any, List
-import warnings
 
 __version__ = importlib.metadata.version("lancedb")
 
@@ -15,7 +14,6 @@ from ._lancedb import connect as lancedb_connect
 from .common import URI, sanitize_uri
 from urllib.parse import urlparse
 from .db import AsyncConnection, DBConnection, LanceDBConnection
-from .io import StorageOptionsProvider
 from .remote import ClientConfig
 from .remote.db import RemoteDBConnection
 from .expr import Expr, col, lit, func
@@ -64,7 +62,7 @@ def _check_s3_bucket_with_dots(
 
 
 def connect(
-    uri: URI,
+    uri: Optional[URI] = None,
     *,
     api_key: Optional[str] = None,
     region: str = "us-east-1",
@@ -74,14 +72,19 @@ def connect(
     client_config: Union[ClientConfig, Dict[str, Any], None] = None,
     storage_options: Optional[Dict[str, str]] = None,
     session: Optional[Session] = None,
+    manifest_enabled: bool = False,
+    namespace_client_impl: Optional[str] = None,
+    namespace_client_properties: Optional[Dict[str, str]] = None,
+    namespace_client_pushdown_operations: Optional[List[str]] = None,
     **kwargs: Any,
 ) -> DBConnection:
     """Connect to a LanceDB database.
 
     Parameters
     ----------
-    uri: str or Path
+    uri: str or Path, optional
-        The uri of the database.
+        The uri of the database. When ``namespace_client_impl`` is provided you may
+        omit ``uri`` and connect through a namespace client instead.
     api_key: str, optional
         If presented, connect to LanceDB cloud.
         Otherwise, connect to a database on file system or cloud storage.
@@ -107,13 +110,29 @@ def connect(
         default configuration is used.
     storage_options: dict, optional
         Additional options for the storage backend. See available options at
-        <https://lancedb.com/docs/storage/>
+        <https://docs.lancedb.com/storage/>
+    manifest_enabled : bool, default False
+        When true for local/native connections, use directory namespace
+        manifests as the source of truth for table metadata. Existing
+        directory-listed root tables are migrated into the manifest on access.
     session: Session, optional
         (For LanceDB OSS only)
         A session to use for this connection. Sessions allow you to configure
         cache sizes for index and metadata caches, which can significantly
         impact memory use and performance. They can also be re-used across
         multiple connections to share the same cache state.
+    namespace_client_impl : str, optional
+        When provided along with ``namespace_client_properties``, ``connect``
+        returns a namespace-backed connection by delegating to
+        :func:`connect_namespace`. The value identifies which namespace
+        implementation to load (e.g., ``"dir"`` or ``"rest"``).
+    namespace_client_properties : dict, optional
+        Configuration to pass to the namespace client implementation. Required
+        when ``namespace_client_impl`` is set.
+    namespace_client_pushdown_operations : list[str], optional
+        Only used when ``namespace_client_properties`` is provided. Forwards to
+        :func:`connect_namespace` to control which operations are executed on the
+        namespace service (e.g., ``["QueryTable", "CreateTable"]``).
 
     Examples
     --------
@@ -133,11 +152,48 @@ def connect(
     >>> db = lancedb.connect("db://my_database", api_key="ldb_...",
     ...                      client_config={"retry_config": {"retries": 5}})
 
+    Connect to a namespace-backed database:
+
+    >>> db = lancedb.connect(namespace_client_impl="dir",
+    ...                      namespace_client_properties={"root": "/tmp/ns"})
+
     Returns
     -------
     conn : DBConnection
         A connection to a LanceDB database.
     """
+    if namespace_client_impl is not None:
+        if namespace_client_properties is None:
+            raise ValueError(
+                "namespace_client_properties must be provided when "
+                "namespace_client_impl is set"
+            )
+        if kwargs:
+            raise ValueError(f"Unknown keyword arguments: {kwargs}")
+        return connect_namespace(
+            namespace_client_impl,
+            namespace_client_properties,
+            read_consistency_interval=read_consistency_interval,
+            storage_options=storage_options,
+            session=session,
+            namespace_client_pushdown_operations=namespace_client_pushdown_operations,
+        )
+
+    if namespace_client_properties is not None and not manifest_enabled:
+        raise ValueError(
+            "namespace_client_impl must be provided when using "
+            "namespace_client_properties unless manifest_enabled=True"
+        )
+
+    if namespace_client_pushdown_operations is not None:
+        raise ValueError(
+            "namespace_client_pushdown_operations is only valid when "
+            "connecting through a namespace"
+        )
+    if uri is None:
+        raise ValueError(
+            "uri is required when not connecting through a namespace client"
+        )
     if isinstance(uri, str) and uri.startswith("db://"):
         if api_key is None:
             api_key = os.environ.get("LANCEDB_API_KEY")
@@ -166,9 +222,92 @@ def connect(
         read_consistency_interval=read_consistency_interval,
         storage_options=storage_options,
         session=session,
+        manifest_enabled=manifest_enabled,
+        namespace_client_properties=namespace_client_properties,
     )
 
 
+WORKER_PROPERTY_PREFIX = "_lancedb_worker_"
+
+
+def _apply_worker_overrides(props: dict[str, str]) -> dict[str, str]:
+    """Apply worker property overrides.
+
+    Any key starting with ``_lancedb_worker_`` is extracted, the prefix
+    is stripped, and the resulting key-value pair is put back into the
+    map (overriding the existing value if present).  The original
+    prefixed key is removed.
+    """
+    worker_keys = [k for k in props if k.startswith(WORKER_PROPERTY_PREFIX)]
+    if not worker_keys:
+        return props
+    result = dict(props)
+    for key in worker_keys:
+        value = result.pop(key)
+        real_key = key[len(WORKER_PROPERTY_PREFIX) :]
+        result[real_key] = value
+    return result
+
+
+def deserialize_conn(
+    data: str,
+    *,
+    for_worker: bool = False,
+) -> DBConnection:
+    """Reconstruct a DBConnection from a serialized string.
+
+    The string must have been produced by
+    :meth:`DBConnection.serialize`.
+
+    Parameters
+    ----------
+    data : str
+        String produced by ``serialize()``.
+    for_worker : bool, default False
+        When ``True``, any namespace client property whose key starts
+        with ``_lancedb_worker_`` has that prefix stripped and the
+        value overrides the corresponding property.  For example,
+        ``_lancedb_worker_uri`` replaces ``uri``.
+
+    Returns
+    -------
+    DBConnection
+        A new connection matching the serialized state.
+    """
+    import json
+
+    parsed = json.loads(data)
+    connection_type = parsed.get("connection_type")
+
+    rci_secs = parsed.get("read_consistency_interval_seconds")
+    rci = timedelta(seconds=rci_secs) if rci_secs is not None else None
+    storage_options = parsed.get("storage_options")
+
+    if connection_type == "namespace":
+        props = dict(parsed.get("namespace_client_properties") or {})
+        if for_worker:
+            props = _apply_worker_overrides(props)
+        return connect_namespace(
+            namespace_client_impl=parsed["namespace_client_impl"],
+            namespace_client_properties=props,
+            read_consistency_interval=rci,
+            storage_options=storage_options,
+            namespace_client_pushdown_operations=parsed.get(
+                "namespace_client_pushdown_operations"
+            ),
+        )
+    elif connection_type == "local":
+        return LanceDBConnection(
+            parsed["uri"],
+            read_consistency_interval=rci,
+            storage_options=storage_options,
+            manifest_enabled=parsed.get("manifest_enabled", False),
+            namespace_client_properties=parsed.get("namespace_client_properties"),
+        )
+    else:
+        raise ValueError(f"Unknown connection_type: {connection_type}")
+
+
 async def connect_async(
     uri: URI,
     *,
@@ -179,6 +318,8 @@ async def connect_async(
     client_config: Optional[Union[ClientConfig, Dict[str, Any]]] = None,
     storage_options: Optional[Dict[str, str]] = None,
     session: Optional[Session] = None,
+    manifest_enabled: bool = False,
+    namespace_client_properties: Optional[Dict[str, str]] = None,
 ) -> AsyncConnection:
     """Connect to a LanceDB database.
 
@@ -211,13 +352,20 @@ async def connect_async(
         default configuration is used.
     storage_options: dict, optional
         Additional options for the storage backend. See available options at
-        <https://lancedb.com/docs/storage/>
+        <https://docs.lancedb.com/storage/>
     session: Session, optional
         (For LanceDB OSS only)
         A session to use for this connection. Sessions allow you to configure
         cache sizes for index and metadata caches, which can significantly
         impact memory use and performance. They can also be re-used across
         multiple connections to share the same cache state.
+    manifest_enabled : bool, default False
+        When true for local/native connections, use directory namespace
+        manifests as the source of truth for table metadata. Existing
+        directory-listed root tables are migrated into the manifest on access.
+    namespace_client_properties : dict, optional
+        Additional directory namespace client properties to use with
+        ``manifest_enabled=True``.
 
     Examples
     --------
@@ -260,6 +408,8 @@ async def connect_async(
             client_config,
             storage_options,
             session,
+            manifest_enabled,
+            namespace_client_properties,
         )
     )
 
@@ -284,17 +434,6 @@ __all__ = [
     "LanceNamespaceDBConnection",
     "RemoteDBConnection",
     "Session",
-    "StorageOptionsProvider",
     "Table",
     "__version__",
 ]
-
-
-def __warn_on_fork():
-    warnings.warn(
-        "lance is not fork-safe. If you are using multiprocessing, use spawn instead.",
-    )
-
-
-if hasattr(os, "register_at_fork"):
-    os.register_at_fork(before=__warn_on_fork)  # type: ignore[attr-defined]

python/python/lancedb/_lancedb.pyi

@@ -12,9 +12,9 @@ from .index import (
     LabelList,
     HnswPq,
     HnswSq,
+    HnswFlat,
     FTS,
 )
-from .io import StorageOptionsProvider
 from lance_namespace import (
     ListNamespacesResponse,
     CreateNamespaceResponse,
@@ -26,6 +26,7 @@ from .remote import ClientConfig
 
 IvfHnswPq: type[HnswPq] = HnswPq
 IvfHnswSq: type[HnswSq] = HnswSq
+IvfHnswFlat: type[HnswFlat] = HnswFlat
 
 class PyExpr:
     """A type-safe DataFusion expression node (Rust-side handle)."""
@@ -72,35 +73,35 @@ class Connection(object):
     async def close(self): ...
     async def list_namespaces(
         self,
-        namespace: Optional[List[str]] = None,
+        namespace_path: Optional[List[str]] = None,
         page_token: Optional[str] = None,
         limit: Optional[int] = None,
     ) -> ListNamespacesResponse: ...
     async def create_namespace(
         self,
-        namespace: List[str],
+        namespace_path: List[str],
         mode: Optional[str] = None,
         properties: Optional[Dict[str, str]] = None,
     ) -> CreateNamespaceResponse: ...
     async def drop_namespace(
         self,
-        namespace: List[str],
+        namespace_path: List[str],
         mode: Optional[str] = None,
         behavior: Optional[str] = None,
     ) -> DropNamespaceResponse: ...
     async def describe_namespace(
         self,
-        namespace: List[str],
+        namespace_path: List[str],
     ) -> DescribeNamespaceResponse: ...
     async def list_tables(
         self,
-        namespace: Optional[List[str]] = None,
+        namespace_path: Optional[List[str]] = None,
         page_token: Optional[str] = None,
         limit: Optional[int] = None,
     ) -> ListTablesResponse: ...
     async def table_names(
         self,
-        namespace: Optional[List[str]],
+        namespace_path: Optional[List[str]],
         start_after: Optional[str],
         limit: Optional[int],
     ) -> list[str]: ...  # Deprecated: Use list_tables instead
@@ -109,9 +110,8 @@ class Connection(object):
         name: str,
         mode: str,
         data: pa.RecordBatchReader,
-        namespace: Optional[List[str]] = None,
+        namespace_path: Optional[List[str]] = None,
         storage_options: Optional[Dict[str, str]] = None,
-        storage_options_provider: Optional[StorageOptionsProvider] = None,
         location: Optional[str] = None,
     ) -> Table: ...
     async def create_empty_table(
@@ -119,17 +119,15 @@ class Connection(object):
         name: str,
         mode: str,
         schema: pa.Schema,
-        namespace: Optional[List[str]] = None,
+        namespace_path: Optional[List[str]] = None,
         storage_options: Optional[Dict[str, str]] = None,
-        storage_options_provider: Optional[StorageOptionsProvider] = None,
         location: Optional[str] = None,
     ) -> Table: ...
     async def open_table(
         self,
         name: str,
-        namespace: Optional[List[str]] = None,
+        namespace_path: Optional[List[str]] = None,
         storage_options: Optional[Dict[str, str]] = None,
-        storage_options_provider: Optional[StorageOptionsProvider] = None,
         index_cache_size: Optional[int] = None,
         location: Optional[str] = None,
     ) -> Table: ...
@@ -137,7 +135,7 @@ class Connection(object):
         self,
         target_table_name: str,
         source_uri: str,
-        target_namespace: Optional[List[str]] = None,
+        target_namespace_path: Optional[List[str]] = None,
         source_version: Optional[int] = None,
         source_tag: Optional[str] = None,
         is_shallow: bool = True,
@@ -146,13 +144,18 @@ class Connection(object):
         self,
         cur_name: str,
         new_name: str,
-        cur_namespace: Optional[List[str]] = None,
+        cur_namespace_path: Optional[List[str]] = None,
-        new_namespace: Optional[List[str]] = None,
+        new_namespace_path: Optional[List[str]] = None,
     ) -> None: ...
     async def drop_table(
-        self, name: str, namespace: Optional[List[str]] = None
+        self, name: str, namespace_path: Optional[List[str]] = None
     ) -> None: ...
-    async def drop_all_tables(self, namespace: Optional[List[str]] = None) -> None: ...
+    async def drop_all_tables(
+        self, namespace_path: Optional[List[str]] = None
+    ) -> None: ...
+    async def namespace_client_config(
+        self,
+    ) -> Dict[str, Any]: ...
 
 class Table:
     def name(self) -> str: ...
@@ -179,6 +182,7 @@ class Table:
             IvfPq,
             HnswPq,
             HnswSq,
+            HnswFlat,
             BTree,
             Bitmap,
             LabelList,
@@ -241,6 +245,8 @@ async def connect(
     client_config: Optional[Union[ClientConfig, Dict[str, Any]]],
     storage_options: Optional[Dict[str, str]],
     session: Optional[Session],
+    manifest_enabled: bool = False,
+    namespace_client_properties: Optional[Dict[str, str]] = None,
 ) -> Connection: ...
 
 class RecordBatchStream:
@@ -439,7 +445,7 @@ class AsyncPermutationBuilder:
     async def execute(self) -> Table: ...
 
 def async_permutation_builder(
-    table: Table, dest_table_name: str
+    table: Table,
 ) -> AsyncPermutationBuilder: ...
 def fts_query_to_json(query: Any) -> str: ...
 

python/python/lancedb/background_loop.py

@@ -2,7 +2,9 @@
 # SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
 import asyncio
+import os
 import threading
+import warnings
 
 
 class BackgroundEventLoop:
@@ -13,6 +15,9 @@ class BackgroundEventLoop:
     """
 
     def __init__(self):
+        self._start()
+
+    def _start(self):
         self.loop = asyncio.new_event_loop()
         self.thread = threading.Thread(
             target=self.loop.run_forever,
@@ -31,3 +36,30 @@ class BackgroundEventLoop:
 
 
 LOOP = BackgroundEventLoop()
+
+_FORK_WARNED = False
+
+
+def _reset_after_fork():
+    # Threads do not survive fork(), so the asyncio loop in LOOP.thread is
+    # dead in the child. Re-initialize the singleton in place so existing
+    # `from .background_loop import LOOP` references in other modules see
+    # the new state. The Rust-side tokio runtime is reset analogously by a
+    # pthread_atfork hook installed in the _lancedb extension.
+    LOOP._start()
+    global _FORK_WARNED
+    if not _FORK_WARNED:
+        _FORK_WARNED = True
+        warnings.warn(
+            "lancedb fork support is experimental: the internal async "
+            "runtime has been reset in the forked child, but a small chance "
+            "of deadlock remains if other state was mid-operation at fork "
+            "time. The 'forkserver' or 'spawn' multiprocessing start method "
+            "is likely a safer alternative.",
+            RuntimeWarning,
+            stacklevel=2,
+        )
+
+
+if hasattr(os, "register_at_fork"):
+    os.register_at_fork(after_in_child=_reset_after_fork)

python/python/lancedb/common.py

@@ -96,7 +96,7 @@ def data_to_reader(
             f"Unknown data type {type(data)}. "
             "Supported types: list of dicts, pandas DataFrame, polars DataFrame, "
             "pyarrow Table/RecordBatch, or Pydantic models. "
-            "See https://lancedb.com/docs/tables/ for examples."
+            "See https://docs.lancedb.com/tables/ for examples."
         )
 
 

python/python/lancedb/embeddings/gemini_text.py

@@ -19,10 +19,10 @@ from .utils import TEXT, api_key_not_found_help
 @register("gemini-text")
 class GeminiText(TextEmbeddingFunction):
     """
-    An embedding function that uses the Google's Gemini API. Requires GOOGLE_API_KEY to
+    An embedding function that uses Google's Gemini API. Requires GOOGLE_API_KEY to
     be set.
 
-    https://ai.google.dev/docs/embeddings_guide
+    https://ai.google.dev/gemini-api/docs/embeddings
 
     Supports various tasks types:
     | Task Type               | Description                                            |
@@ -46,9 +46,12 @@ class GeminiText(TextEmbeddingFunction):
 
     Parameters
     ----------
-    name: str, default "models/embedding-001"
+    name: str, default "gemini-embedding-001"
-        The name of the model to use. See the Gemini documentation for a list of
+        The name of the model to use. Supported models include:
-        available models.
+        - "gemini-embedding-001" (768 dimensions)
+
+        Note: The legacy "models/embedding-001" format is also supported but
+        "gemini-embedding-001" is recommended.
 
     query_task_type: str, default "retrieval_query"
         Sets the task type for the queries.
@@ -77,7 +80,7 @@ class GeminiText(TextEmbeddingFunction):
 
     """
 
-    name: str = "models/embedding-001"
+    name: str = "gemini-embedding-001"
     query_task_type: str = "retrieval_query"
     source_task_type: str = "retrieval_document"
 
@@ -114,23 +117,48 @@ class GeminiText(TextEmbeddingFunction):
         texts: list[str] or np.ndarray (of str)
             The texts to embed
         """
-        if (
+        from google.genai import types
-            kwargs.get("task_type") == "retrieval_document"
-        ):  # Provide a title to use existing API design
-            title = "Embedding of a document"
-            kwargs["title"] = title
 
-        return [
+        task_type = kwargs.get("task_type")
-            self.client.embed_content(model=self.name, content=text, **kwargs)[
+
-                "embedding"
+        # Build content objects for embed_content
-            ]
+        contents = []
-            for text in texts
+        for text in texts:
-        ]
+            if task_type == "retrieval_document":
+                # Provide a title for retrieval_document task
+                contents.append(
+                    {"parts": [{"text": "Embedding of a document"}, {"text": text}]}
+                )
+            else:
+                contents.append({"parts": [{"text": text}]})
+
+        # Build config
+        config_kwargs = {}
+        if task_type:
+            config_kwargs["task_type"] = task_type.upper()  # API expects uppercase
+
+        # Call embed_content for each content
+        embeddings = []
+        for content in contents:
+            config = (
+                types.EmbedContentConfig(**config_kwargs) if config_kwargs else None
+            )
+            response = self.client.models.embed_content(
+                model=self.name,
+                contents=content,
+                config=config,
+            )
+            embeddings.append(response.embeddings[0].values)
+
+        return embeddings
 
     @cached_property
     def client(self):
-        genai = attempt_import_or_raise("google.generativeai", "google.generativeai")
+        attempt_import_or_raise("google.genai", "google-genai")
 
         if not os.environ.get("GOOGLE_API_KEY"):
             api_key_not_found_help("google")
-        return genai
+
+        from google import genai as genai_module
+
+        return genai_module.Client(api_key=os.environ.get("GOOGLE_API_KEY"))

python/python/lancedb/fts.py

@@ -1,201 +0,0 @@
-# SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright The LanceDB Authors
-
-"""Full text search index using tantivy-py"""
-
-import os
-from typing import List, Tuple, Optional
-
-import pyarrow as pa
-
-try:
-    import tantivy
-except ImportError:
-    raise ImportError(
-        "Please install tantivy-py `pip install tantivy` to use the full text search feature."  # noqa: E501
-    )
-
-from .table import LanceTable
-
-
-def create_index(
-    index_path: str,
-    text_fields: List[str],
-    ordering_fields: Optional[List[str]] = None,
-    tokenizer_name: str = "default",
-) -> tantivy.Index:
-    """
-    Create a new Index (not populated)
-
-    Parameters
-    ----------
-    index_path : str
-        Path to the index directory
-    text_fields : List[str]
-        List of text fields to index
-    ordering_fields: List[str]
-        List of unsigned type fields to order by at search time
-    tokenizer_name : str, default "default"
-        The tokenizer to use
-
-    Returns
-    -------
-    index : tantivy.Index
-        The index object (not yet populated)
-    """
-    if ordering_fields is None:
-        ordering_fields = []
-    # Declaring our schema.
-    schema_builder = tantivy.SchemaBuilder()
-    # special field that we'll populate with row_id
-    schema_builder.add_integer_field("doc_id", stored=True)
-    # data fields
-    for name in text_fields:
-        schema_builder.add_text_field(name, stored=True, tokenizer_name=tokenizer_name)
-    if ordering_fields:
-        for name in ordering_fields:
-            schema_builder.add_unsigned_field(name, fast=True)
-    schema = schema_builder.build()
-    os.makedirs(index_path, exist_ok=True)
-    index = tantivy.Index(schema, path=index_path)
-    return index
-
-
-def populate_index(
-    index: tantivy.Index,
-    table: LanceTable,
-    fields: List[str],
-    writer_heap_size: Optional[int] = None,
-    ordering_fields: Optional[List[str]] = None,
-) -> int:
-    """
-    Populate an index with data from a LanceTable
-
-    Parameters
-    ----------
-    index : tantivy.Index
-        The index object
-    table : LanceTable
-        The table to index
-    fields : List[str]
-        List of fields to index
-    writer_heap_size : int
-        The writer heap size in bytes, defaults to 1GB
-
-    Returns
-    -------
-    int
-        The number of rows indexed
-    """
-    if ordering_fields is None:
-        ordering_fields = []
-    writer_heap_size = writer_heap_size or 1024 * 1024 * 1024
-    # first check the fields exist and are string or large string type
-    nested = []
-
-    for name in fields:
-        try:
-            f = table.schema.field(name)  # raises KeyError if not found
-        except KeyError:
-            f = resolve_path(table.schema, name)
-            nested.append(name)
-
-        if not pa.types.is_string(f.type) and not pa.types.is_large_string(f.type):
-            raise TypeError(f"Field {name} is not a string type")
-
-    # create a tantivy writer
-    writer = index.writer(heap_size=writer_heap_size)
-    # write data into index
-    dataset = table.to_lance()
-    row_id = 0
-
-    max_nested_level = 0
-    if len(nested) > 0:
-        max_nested_level = max([len(name.split(".")) for name in nested])
-
-    for b in dataset.to_batches(columns=fields + ordering_fields):
-        if max_nested_level > 0:
-            b = pa.Table.from_batches([b])
-            for _ in range(max_nested_level - 1):
-                b = b.flatten()
-        for i in range(b.num_rows):
-            doc = tantivy.Document()
-            for name in fields:
-                value = b[name][i].as_py()
-                if value is not None:
-                    doc.add_text(name, value)
-            for name in ordering_fields:
-                value = b[name][i].as_py()
-                if value is not None:
-                    doc.add_unsigned(name, value)
-            if not doc.is_empty:
-                doc.add_integer("doc_id", row_id)
-                writer.add_document(doc)
-            row_id += 1
-    # commit changes
-    writer.commit()
-    return row_id
-
-
-def resolve_path(schema, field_name: str) -> pa.Field:
-    """
-    Resolve a nested field path to a list of field names
-
-    Parameters
-    ----------
-    field_name : str
-        The field name to resolve
-
-    Returns
-    -------
-    List[str]
-        The resolved path
-    """
-    path = field_name.split(".")
-    field = schema.field(path.pop(0))
-    for segment in path:
-        if pa.types.is_struct(field.type):
-            field = field.type.field(segment)
-        else:
-            raise KeyError(f"field {field_name} not found in schema {schema}")
-    return field
-
-
-def search_index(
-    index: tantivy.Index, query: str, limit: int = 10, ordering_field=None
-) -> Tuple[Tuple[int], Tuple[float]]:
-    """
-    Search an index for a query
-
-    Parameters
-    ----------
-    index : tantivy.Index
-        The index object
-    query : str
-        The query string
-    limit : int
-        The maximum number of results to return
-
-    Returns
-    -------
-    ids_and_score: list[tuple[int], tuple[float]]
-        A tuple of two tuples, the first containing the document ids
-        and the second containing the scores
-    """
-    searcher = index.searcher()
-    query = index.parse_query(query)
-    # get top results
-    if ordering_field:
-        results = searcher.search(query, limit, order_by_field=ordering_field)
-    else:
-        results = searcher.search(query, limit)
-    if results.count == 0:
-        return tuple(), tuple()
-    return tuple(
-        zip(
-            *[
-                (searcher.doc(doc_address)["doc_id"][0], score)
-                for score, doc_address in results.hits
-            ]
-        )
-    )

python/python/lancedb/index.py

@@ -7,6 +7,7 @@ from typing import Literal, Optional
 from ._lancedb import (
     IndexConfig,
 )
+from .types import BaseTokenizerType
 
 lang_mapping = {
     "ar": "Arabic",
@@ -111,8 +112,12 @@ class FTS:
         - "simple": Splits text by whitespace and punctuation.
         - "whitespace": Split text by whitespace, but not punctuation.
         - "raw": No tokenization. The entire text is treated as a single token.
+        - "ngram": N-gram tokenizer for substring-style matching.
+        - "jieba/*": Jieba tokenizer loaded from Lance's language model home.
+        - "lindera/*": Lindera tokenizer loaded from Lance's language model home.
     language : str, default "English"
-        The language to use for tokenization.
+        The language to use for stemming and stop-word removal. This is not the
+        primary way to enable CJK tokenization.
     max_token_length : int, default 40
         The maximum token length to index. Tokens longer than this length will be
         ignored.
@@ -127,10 +132,17 @@ class FTS:
     ascii_folding : bool, default True
         Whether to fold ASCII characters. This converts accented characters to
         their ASCII equivalent. For example, "café" would be converted to "cafe".
+
+    Notes
+    -----
+    Model-backed tokenizers such as ``jieba/default`` and ``lindera/ipadic``
+    require tokenizer models in Lance's language model home. Set
+    ``LANCE_LANGUAGE_MODEL_HOME`` to override the default platform data
+    directory under ``lance/language_models``.
     """
 
     with_position: bool = False
-    base_tokenizer: Literal["simple", "raw", "whitespace"] = "simple"
+    base_tokenizer: BaseTokenizerType = "simple"
     language: str = "English"
     max_token_length: Optional[int] = 40
     lower_case: bool = True
@@ -376,9 +388,98 @@ class HnswSq:
     target_partition_size: Optional[int] = None
 
 
+@dataclass
+class HnswFlat:
+    """Describe a HNSW-FLAT index configuration.
+
+    HNSW-FLAT stands for Hierarchical Navigable Small World without quantization.
+    It stores raw vectors in the HNSW graph, providing the highest recall among
+    the IVF_HNSW family at the cost of more memory and disk space compared to
+    :class:`HnswSq` or :class:`HnswPq`.
+
+    Parameters
+    ----------
+
+    distance_type: str, default "l2"
+
+        The distance metric used to train the index.
+
+        The following distance types are available:
+
+        "l2" - Euclidean distance. This is a very common distance metric that
+        accounts for both magnitude and direction when determining the distance
+        between vectors. l2 distance has a range of [0, ∞).
+
+        "cosine" - Cosine distance.  Cosine distance is a distance metric
+        calculated from the cosine similarity between two vectors. Cosine
+        similarity is a measure of similarity between two non-zero vectors of an
+        inner product space. It is defined to equal the cosine of the angle
+        between them.  Unlike l2, the cosine distance is not affected by the
+        magnitude of the vectors.  Cosine distance has a range of [0, 2].
+
+        "dot" - Dot product. Dot distance is the dot product of two vectors. Dot
+        distance has a range of (-∞, ∞). If the vectors are normalized (i.e. their
+        l2 norm is 1), then dot distance is equivalent to the cosine distance.
+
+    num_partitions, default sqrt(num_rows)
+
+        The number of IVF partitions to create.
+
+        For HNSW, we recommend a small number of partitions. Setting this to 1
+        works well for most tables. For very large tables, training just one HNSW
+        graph will require too much memory. Each partition becomes its own HNSW
+        graph, so setting this value higher reduces the peak memory use of
+        training.
+
+    max_iterations, default 50
+
+        Max iterations to train kmeans.
+
+        When training an IVF index we use kmeans to calculate the partitions.
+        This parameter controls how many iterations of kmeans to run.
+
+    sample_rate, default 256
+
+        The rate used to calculate the number of training vectors for kmeans.
+
+    m, default 20
+
+        The number of neighbors to select for each vector in the HNSW graph.
+
+        This value controls the tradeoff between search speed and accuracy.
+        The higher the value the more accurate the search but the slower it
+        will be.
+
+    ef_construction, default 300
+
+        The number of candidates to evaluate during the construction of the HNSW
+        graph.
+
+        This value controls the tradeoff between build speed and accuracy.
+        The higher the value the more accurate the build but the slower it will
+        be.  150 to 300 is the typical range. 100 is a minimum for good quality
+        search results. In most cases, there is no benefit to setting this higher
+        than 500.  This value should be set to a value that is not less than `ef`
+        in the search phase.
+
+    target_partition_size, default is 1,048,576
+
+        The target size of each partition.
+    """
+
+    distance_type: Literal["l2", "cosine", "dot"] = "l2"
+    num_partitions: Optional[int] = None
+    max_iterations: int = 50
+    sample_rate: int = 256
+    m: int = 20
+    ef_construction: int = 300
+    target_partition_size: Optional[int] = None
+
+
 # Backwards-compatible aliases
 IvfHnswPq = HnswPq
 IvfHnswSq = HnswSq
+IvfHnswFlat = HnswFlat
 
 
 @dataclass
@@ -698,11 +799,13 @@ __all__ = [
     "IvfPq",
     "IvfHnswPq",
     "IvfHnswSq",
+    "IvfHnswFlat",
     "IvfSq",
     "IvfRq",
     "IvfFlat",
     "HnswPq",
     "HnswSq",
+    "HnswFlat",
     "IndexConfig",
     "FTS",
     "Bitmap",

python/python/lancedb/io.py

@@ -2,70 +2,3 @@
 # SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
 """I/O utilities and interfaces for LanceDB."""
-
-from abc import ABC, abstractmethod
-from typing import Dict
-
-
-class StorageOptionsProvider(ABC):
-    """Abstract base class for providing storage options to LanceDB tables.
-
-    Storage options providers enable automatic credential refresh for cloud
-    storage backends (e.g., AWS S3, Azure Blob Storage, GCS). When credentials
-    have an expiration time, the provider's fetch_storage_options() method will
-    be called periodically to get fresh credentials before they expire.
-
-    Example
-    -------
-    >>> class MyProvider(StorageOptionsProvider):
-    ...     def fetch_storage_options(self) -> Dict[str, str]:
-    ...         # Fetch fresh credentials from your credential manager
-    ...         return {
-    ...             "aws_access_key_id": "...",
-    ...             "aws_secret_access_key": "...",
-    ...             "expires_at_millis": "1234567890000"  # Optional
-    ...         }
-    """
-
-    @abstractmethod
-    def fetch_storage_options(self) -> Dict[str, str]:
-        """Fetch fresh storage credentials.
-
-        This method is called by LanceDB when credentials need to be refreshed.
-        If the returned dictionary contains an "expires_at_millis" key with a
-        Unix timestamp in milliseconds, LanceDB will automatically refresh the
-        credentials before that time. If the key is not present, credentials
-        are assumed to not expire.
-
-        Returns
-        -------
-        Dict[str, str]
-            Dictionary containing cloud storage credentials and optionally an
-            expiration time:
-            - "expires_at_millis" (optional): Unix timestamp in milliseconds when
-              credentials expire
-            - Provider-specific credential keys (e.g., aws_access_key_id,
-              aws_secret_access_key, etc.)
-
-        Raises
-        ------
-        RuntimeError
-            If credentials cannot be fetched or are invalid
-        """
-        pass
-
-    def provider_id(self) -> str:
-        """Return a human-readable unique identifier for this provider instance.
-
-        This identifier is used for caching and equality comparison. Two providers
-        with the same ID will share the same cached object store connection.
-
-        The default implementation uses the class name and string representation.
-        Override this method if you need custom identification logic.
-
-        Returns
-        -------
-        str
-            A unique identifier for this provider instance
-        """
-        return f"{self.__class__.__name__} {{ repr: {str(self)!r} }}"

python/python/lancedb/permutation.py

@@ -1,11 +1,12 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
-from deprecation import deprecated
+import copy
-from lancedb import AsyncConnection, DBConnection
-import pyarrow as pa
 import json
 
+from deprecation import deprecated
+import pyarrow as pa
+
 from ._lancedb import async_permutation_builder, PermutationReader
 from .table import LanceTable
 from .background_loop import LOOP
@@ -36,10 +37,7 @@ class PermutationBuilder:
     be referenced by name in the future.  If names are not provided then they can only
     be referenced by their ordinal index.  There is no requirement to name every split.
 
-    By default, the permutation will be stored in memory and will be lost when the
+    The permutation is stored in memory and will be lost when the program exits.
-    program exits.  To persist the permutation (for very large datasets or to share
-    the permutation across multiple workers) use the [persist](#persist) method to
-    create a permanent table.
     """
 
     def __init__(self, table: LanceTable):
@@ -51,15 +49,6 @@ class PermutationBuilder:
         """
         self._async = async_permutation_builder(table)
 
-    def persist(
-        self, database: Union[DBConnection, AsyncConnection], table_name: str
-    ) -> "PermutationBuilder":
-        """
-        Persist the permutation to the given database.
-        """
-        self._async.persist(database, table_name)
-        return self
-
     def split_random(
         self,
         *,
@@ -284,9 +273,8 @@ class Permutations:
         self.permutation_table = permutation_table
 
         if permutation_table.schema.metadata is not None:
-            split_names = permutation_table.schema.metadata.get(
+            raw = permutation_table.schema.metadata.get(b"split_names")
-                b"split_names", None
+            split_names = raw.decode("utf-8") if raw is not None else None
-            ).decode("utf-8")
             if split_names is not None:
                 self.split_names = json.loads(split_names)
                 self.split_dict = {
@@ -381,20 +369,44 @@ class Permutation:
 
     def __init__(
         self,
-        reader: PermutationReader,
+        base_table: LanceTable,
+        permutation_table: Optional[LanceTable],
+        split: int,
         selection: dict[str, str],
         batch_size: int,
         transform_fn: Callable[pa.RecordBatch, Any],
+        offset: Optional[int] = None,
+        limit: Optional[int] = None,
+        connection_factory: Optional[Callable[[str], LanceTable]] = None,
+        _reader: Optional[PermutationReader] = None,
     ):
         """
         Internal constructor.  Use [from_tables](#from_tables) instead.
         """
-        assert reader is not None, "reader is required"
+        assert base_table is not None, "base_table is required"
         assert selection is not None, "selection is required"
-        self.reader = reader
+        self.base_table = base_table
+        self.permutation_table = permutation_table
+        self.split = split
         self.selection = selection
         self.transform_fn = transform_fn
         self.batch_size = batch_size
+        self.offset = offset
+        self.limit = limit
+        self.connection_factory = connection_factory
+        if _reader is None:
+            _reader = LOOP.run(self._build_reader())
+        self.reader: PermutationReader = _reader
+
+    async def _build_reader(self) -> PermutationReader:
+        reader = await PermutationReader.from_tables(
+            self.base_table, self.permutation_table, self.split
+        )
+        if self.offset is not None:
+            reader = await reader.with_offset(self.offset)
+        if self.limit is not None:
+            reader = await reader.with_limit(self.limit)
+        return reader
 
     def _with_selection(self, selection: dict[str, str]) -> "Permutation":
         """
@@ -403,21 +415,97 @@ class Permutation:
         Does not validation of the selection and it replaces it entirely.  This is not
         intended for public use.
         """
-        return Permutation(self.reader, selection, self.batch_size, self.transform_fn)
+        new = copy.copy(self)
-
+        new.selection = selection
-    def _with_reader(self, reader: PermutationReader) -> "Permutation":
+        return new
-        """
-        Creates a new permutation with the given reader
-
-        This is an internal method and should not be used directly.
-        """
-        return Permutation(reader, self.selection, self.batch_size, self.transform_fn)
 
     def with_batch_size(self, batch_size: int) -> "Permutation":
         """
         Creates a new permutation with the given batch size
         """
-        return Permutation(self.reader, self.selection, batch_size, self.transform_fn)
+        new = copy.copy(self)
+        new.batch_size = batch_size
+        return new
+
+    def with_connection_factory(
+        self, connection_factory: Callable[[str], LanceTable]
+    ) -> "Permutation":
+        """
+        Creates a new permutation that will use ``connection_factory`` to reopen
+        the base table when this permutation is unpickled in a worker process.
+
+        The factory is a callable that takes a single argument — the base table
+        name — and returns a [LanceTable]. It must be picklable; the worker
+        will pickle it via standard ``pickle`` and call it to recover the base
+        table. Picklable callables in practice means top-level (module-level)
+        functions, ``functools.partial`` of such functions, or instances of
+        picklable classes implementing ``__call__``. Lambdas and closures over
+        local variables don't pickle with the default protocol.
+
+        Setting a factory is necessary when the URI alone is not enough to
+        re-open the connection — most importantly for LanceDB Cloud (``db://``)
+        connections, where ``api_key`` and ``region`` aren't recoverable from
+        the connection object after construction.
+
+        For local file or cloud-storage paths the factory is optional: if not
+        set, ``__getstate__`` falls back to capturing
+        ``(uri, storage_options, namespace_path)`` and re-opening via
+        ``lancedb.connect(uri, storage_options=...)``.
+
+        Examples
+        --------
+        Basic native (file-system path), parameterized via ``functools.partial``::
+
+            import functools, lancedb
+            from lancedb.permutation import Permutation
+
+            def open_native_table(uri: str, table_name: str):
+                return lancedb.connect(uri).open_table(table_name)
+
+            factory = functools.partial(open_native_table, "/data/lance_db")
+            permutation = Permutation.identity(
+                factory("training")
+            ).with_connection_factory(factory)
+
+        Native via :func:`lancedb.connect_namespace` (e.g. a directory- or
+        REST-backed namespace client). The factory takes the
+        implementation name and properties dict as partial-bound args so
+        the worker can rebuild the same namespace connection::
+
+            def open_via_namespace(
+                impl: str, properties: dict[str, str], table_name: str,
+            ):
+                return lancedb.connect_namespace(impl, properties).open_table(
+                    table_name,
+                )
+
+            factory = functools.partial(
+                open_via_namespace,
+                "dir",
+                {"root": "/data/lance_db"},
+            )
+
+        LanceDB Cloud, reading credentials from env vars at worker startup
+        so secrets aren't pickled into the dataset::
+
+            import os, lancedb
+
+            def open_remote_table(table_name: str):
+                db = lancedb.connect(
+                    "db://my-database",
+                    api_key=os.environ["LANCEDB_API_KEY"],
+                    region=os.environ.get("LANCEDB_REGION", "us-east-1"),
+                )
+                return db.open_table(table_name)
+
+            permutation = Permutation.identity(
+                open_remote_table("training")
+            ).with_connection_factory(open_remote_table)
+        """
+        assert connection_factory is not None, "connection_factory is required"
+        new = copy.copy(self)
+        new.connection_factory = connection_factory
+        return new
 
     @classmethod
     def identity(cls, table: LanceTable) -> "Permutation":
@@ -460,9 +548,8 @@ class Permutation:
                         f"Cannot create a permutation on split `{split}`"
                         " because no split names are defined in the permutation table"
                     )
-                split_names = permutation_table.schema.metadata.get(
+                raw = permutation_table.schema.metadata.get(b"split_names")
-                    b"split_names", None
+                split_names = raw.decode("utf-8") if raw is not None else None
-                ).decode("utf-8")
                 if split_names is None:
                     raise ValueError(
                         f"Cannot create a permutation on split `{split}`"
@@ -491,11 +578,126 @@ class Permutation:
             schema = await reader.output_schema(None)
             initial_selection = {name: name for name in schema.names}
             return cls(
-                reader, initial_selection, DEFAULT_BATCH_SIZE, Transforms.arrow2python
+                base_table,
+                permutation_table,
+                split,
+                initial_selection,
+                DEFAULT_BATCH_SIZE,
+                Transforms.arrow2python,
+                _reader=reader,
             )
 
         return LOOP.run(do_from_tables())
 
+    def __getstate__(self) -> dict[str, Any]:
+        """Build a picklable state dict for this permutation.
+
+        The base table is captured either via a user-supplied
+        ``connection_factory`` (see [with_connection_factory]) or, as a
+        fallback, by introspecting ``(uri, storage_options, namespace_path)``
+        on the connection. The permutation table — always an in-memory
+        LanceDB table — is captured as a pyarrow Table (which pickles via
+        Arrow IPC natively). The reader is dropped from the wire format;
+        ``__setstate__`` rebuilds it from the restored tables.
+        """
+        permutation_data: Optional[pa.Table] = None
+        if self.permutation_table is not None:
+            permutation_data = self.permutation_table.to_arrow()
+
+        common = {
+            "base_table_name": self.base_table.name,
+            "permutation_data": permutation_data,
+            "split": self.split,
+            "selection": self.selection,
+            "batch_size": self.batch_size,
+            "transform_fn": self.transform_fn,
+            "offset": self.offset,
+            "limit": self.limit,
+            "connection_factory": self.connection_factory,
+        }
+
+        if self.connection_factory is not None:
+            # The factory carries enough state to recover the base table on
+            # its own; we don't need to capture the URI / storage options /
+            # namespace from the existing connection.
+            return common
+
+        # URI-introspection fallback: only viable for native (OSS) connections
+        # where (uri, storage_options) is enough to reopen. Remote / cloud
+        # connections don't expose recoverable api_key / region — those users
+        # must call with_connection_factory().
+        try:
+            base_uri = self.base_table._conn.uri
+            storage_options = self.base_table._conn.storage_options
+        except AttributeError as e:
+            raise ValueError(
+                "Cannot pickle this Permutation: the base table's connection "
+                "does not expose a uri/storage_options, which usually means it "
+                "is a remote (LanceDB Cloud) connection. Call "
+                "Permutation.with_connection_factory(...) first to provide a "
+                "picklable callable that re-opens the base table from a worker "
+                "process."
+            ) from e
+
+        if base_uri.startswith("memory://"):
+            # In-memory base tables don't exist in any worker process by
+            # default, so dump the entire base table into the pickle. This
+            # can be expensive for large datasets — users with large
+            # in-memory base tables should either persist them or set a
+            # connection_factory.
+            return {
+                **common,
+                "base_table_data": self.base_table.to_arrow(),
+            }
+
+        return {
+            **common,
+            "base_table_uri": base_uri,
+            "base_table_namespace": self.base_table._namespace_path,
+            "base_table_storage_options": storage_options,
+        }
+
+    def __setstate__(self, state: dict[str, Any]) -> None:
+        from . import connect
+
+        connection_factory = state["connection_factory"]
+        if connection_factory is not None:
+            base_table = connection_factory(state["base_table_name"])
+        elif "base_table_data" in state:
+            # In-memory base table inlined into the pickle; rebuild the same
+            # way we rebuild the in-memory permutation table.
+            mem_db = connect("memory://")
+            base_table = mem_db.create_table(
+                state["base_table_name"], state["base_table_data"]
+            )
+        else:
+            base_db = connect(
+                state["base_table_uri"],
+                storage_options=state["base_table_storage_options"],
+            )
+            base_table = base_db.open_table(
+                state["base_table_name"],
+                namespace_path=state["base_table_namespace"] or None,
+            )
+
+        permutation_table: Optional[LanceTable] = None
+        if state["permutation_data"] is not None:
+            mem_db = connect("memory://")
+            permutation_table = mem_db.create_table(
+                "permutation", state["permutation_data"]
+            )
+
+        self.base_table = base_table
+        self.permutation_table = permutation_table
+        self.split = state["split"]
+        self.selection = state["selection"]
+        self.batch_size = state["batch_size"]
+        self.transform_fn = state["transform_fn"]
+        self.offset = state["offset"]
+        self.limit = state["limit"]
+        self.connection_factory = connection_factory
+        self.reader = LOOP.run(self._build_reader())
+
     @property
     def schema(self) -> pa.Schema:
         async def do_output_schema():
@@ -762,7 +964,9 @@ class Permutation:
         for expensive operations such as image decoding.
         """
         assert transform is not None, "transform is required"
-        return Permutation(self.reader, self.selection, self.batch_size, transform)
+        new = copy.copy(self)
+        new.transform_fn = transform
+        return new
 
     def __getitem__(self, index: int) -> Any:
         """
@@ -797,12 +1001,10 @@ class Permutation:
         """
         Skip the first `skip` rows of the permutation
         """
-
+        new = copy.copy(self)
-        async def do_with_skip():
+        new.offset = skip
-            reader = await self.reader.with_offset(skip)
+        new.reader = LOOP.run(new._build_reader())
-            return self._with_reader(reader)
+        return new
-
-        return LOOP.run(do_with_skip())
 
     @deprecated(details="Use with_take instead")
     def take(self, limit: int) -> "Permutation":
@@ -820,12 +1022,10 @@ class Permutation:
         """
         Limit the permutation to `limit` rows (following any `skip`)
         """
-
+        new = copy.copy(self)
-        async def do_with_take():
+        new.limit = limit
-            reader = await self.reader.with_limit(limit)
+        new.reader = LOOP.run(new._build_reader())
-            return self._with_reader(reader)
+        return new
-
-        return LOOP.run(do_with_take())
 
     @deprecated(details="Use with_repeat instead")
     def repeat(self, times: int) -> "Permutation":

python/python/lancedb/pydantic.py

@@ -10,6 +10,7 @@ import sys
 import types
 from abc import ABC, abstractmethod
 from datetime import date, datetime
+from enum import Enum
 from typing import (
     TYPE_CHECKING,
     Any,
@@ -314,6 +315,19 @@ def _pydantic_type_to_arrow_type(tp: Any, field: FieldInfo) -> pa.DataType:
                 return pa.list_(pa.list_(tp.value_arrow_type(), tp.dim()))
             # For regular Vector
             return pa.list_(tp.value_arrow_type(), tp.dim())
+        if _safe_issubclass(tp, Enum):
+            # Map Enum to the Arrow type of its value.
+            # For string-valued enums, use dictionary encoding for efficiency.
+            # For integer enums, use the native type.
+            # Fall back to utf8 for mixed-type or empty enums.
+            value_types = {type(m.value) for m in tp}
+            if len(value_types) == 1:
+                value_type = value_types.pop()
+                if value_type is str:
+                    # Use dictionary encoding for string enums
+                    return pa.dictionary(pa.int32(), pa.utf8())
+                return _py_type_to_arrow_type(value_type, field)
+            return pa.utf8()
     return _py_type_to_arrow_type(tp, field)
 
 

python/python/lancedb/query.py

@@ -25,7 +25,6 @@ import deprecation
 import numpy as np
 import pyarrow as pa
 import pyarrow.compute as pc
-import pyarrow.fs as pa_fs
 import pydantic
 
 from lancedb.pydantic import PYDANTIC_VERSION
@@ -1526,9 +1525,7 @@ class LanceFtsQueryBuilder(LanceQueryBuilder):
         return self._table._output_schema(self.to_query_object())
 
     def to_arrow(self, *, timeout: Optional[timedelta] = None) -> pa.Table:
-        path, fs, exist = self._table._get_fts_index_path()
+        self._table._ensure_no_legacy_fts_index()
-        if exist:
-            return self.tantivy_to_arrow()
 
         query = self._query
         if self._phrase_query:
@@ -1552,90 +1549,6 @@ class LanceFtsQueryBuilder(LanceQueryBuilder):
     ):
         raise NotImplementedError("to_batches on an FTS query")
 
-    def tantivy_to_arrow(self) -> pa.Table:
-        try:
-            import tantivy
-        except ImportError:
-            raise ImportError(
-                "Please install tantivy-py `pip install tantivy` to use the full text search feature."  # noqa: E501
-            )
-
-        from .fts import search_index
-
-        # get the index path
-        path, fs, exist = self._table._get_fts_index_path()
-
-        # check if the index exist
-        if not exist:
-            raise FileNotFoundError(
-                "Fts index does not exist. "
-                "Please first call table.create_fts_index(['<field_names>']) to "
-                "create the fts index."
-            )
-
-        # Check that we are on local filesystem
-        if not isinstance(fs, pa_fs.LocalFileSystem):
-            raise NotImplementedError(
-                "Tantivy-based full text search "
-                "is only supported on the local filesystem"
-            )
-        # open the index
-        index = tantivy.Index.open(path)
-        # get the scores and doc ids
-        query = self._query
-        if self._phrase_query:
-            query = query.replace('"', "'")
-            query = f'"{query}"'
-        limit = self._limit if self._limit is not None else 10
-        row_ids, scores = search_index(
-            index, query, limit, ordering_field=self.ordering_field_name
-        )
-        if len(row_ids) == 0:
-            empty_schema = pa.schema([pa.field("_score", pa.float32())])
-            return pa.Table.from_batches([], schema=empty_schema)
-        scores = pa.array(scores)
-        output_tbl = self._table.to_lance().take(row_ids, columns=self._columns)
-        output_tbl = output_tbl.append_column("_score", scores)
-        # this needs to match vector search results which are uint64
-        row_ids = pa.array(row_ids, type=pa.uint64())
-
-        if self._where is not None:
-            tmp_name = "__lancedb__duckdb__indexer__"
-            output_tbl = output_tbl.append_column(
-                tmp_name, pa.array(range(len(output_tbl)))
-            )
-            try:
-                # TODO would be great to have Substrait generate pyarrow compute
-                # expressions or conversely have pyarrow support SQL expressions
-                # using Substrait
-                import duckdb
-
-                indexer = duckdb.sql(
-                    f"SELECT {tmp_name} FROM output_tbl WHERE {self._where}"
-                ).to_arrow_table()[tmp_name]
-                output_tbl = output_tbl.take(indexer).drop([tmp_name])
-                row_ids = row_ids.take(indexer)
-
-            except ImportError:
-                import tempfile
-
-                import lance
-
-                # TODO Use "memory://" instead once that's supported
-                with tempfile.TemporaryDirectory() as tmp:
-                    ds = lance.write_dataset(output_tbl, tmp)
-                    output_tbl = ds.to_table(filter=self._where)
-                    indexer = output_tbl[tmp_name]
-                    row_ids = row_ids.take(indexer)
-                    output_tbl = output_tbl.drop([tmp_name])
-
-        if self._with_row_id:
-            output_tbl = output_tbl.append_column("_rowid", row_ids)
-
-        if self._reranker is not None:
-            output_tbl = self._reranker.rerank_fts(self._query, output_tbl)
-        return output_tbl
-
     def rerank(self, reranker: Reranker) -> LanceFtsQueryBuilder:
         """Rerank the results using the specified reranker.
 
@@ -1730,7 +1643,7 @@ class LanceHybridQueryBuilder(LanceQueryBuilder):
     def _validate_query(self, query, vector=None, text=None):
         if query is not None and (vector is not None or text is not None):
             raise ValueError(
-                "You can either provide a string query in search() method"
+                "You can either provide a string query in search() method "
                 "or set `vector()` and `text()` explicitly for hybrid search."
                 "But not both."
             )

python/python/lancedb/remote/__init__.py

@@ -145,6 +145,33 @@ class TlsConfig:
 
 @dataclass
 class ClientConfig:
+    """Configuration for the LanceDB Cloud HTTP client.
+
+    Attributes
+    ----------
+    user_agent: str
+        User agent string sent with requests.
+    retry_config: RetryConfig
+        Configuration for retrying failed requests.
+    timeout_config: Optional[TimeoutConfig]
+        Configuration for request timeouts.
+    extra_headers: Optional[dict]
+        Additional headers to include in requests.
+    id_delimiter: Optional[str]
+        The delimiter to use when constructing object identifiers.
+    tls_config: Optional[TlsConfig]
+        TLS/mTLS configuration for secure connections.
+    header_provider: Optional[HeaderProvider]
+        Provider for dynamic headers to be added to each request.
+    user_id: Optional[str]
+        User identifier for tracking purposes. This is sent as the
+        `x-lancedb-user-id` header in requests to LanceDB Cloud/Enterprise.
+
+        This can also be set via the `LANCEDB_USER_ID` environment variable.
+        Alternatively, set `LANCEDB_USER_ID_ENV_KEY` to specify another
+        environment variable that contains the user ID value.
+    """
+
     user_agent: str = f"LanceDB-Python-Client/{__version__}"
     retry_config: RetryConfig = field(default_factory=RetryConfig)
     timeout_config: Optional[TimeoutConfig] = field(default_factory=TimeoutConfig)
@@ -152,6 +179,7 @@ class ClientConfig:
     id_delimiter: Optional[str] = None
     tls_config: Optional[TlsConfig] = None
     header_provider: Optional["HeaderProvider"] = None
+    user_id: Optional[str] = None
 
     def __post_init__(self):
         if isinstance(self.retry_config, dict):

python/python/lancedb/remote/db.py

@@ -24,6 +24,7 @@ from ..common import DATA
 from ..db import DBConnection, LOOP
 from ..embeddings import EmbeddingFunctionConfig
 from lance_namespace import (
+    LanceNamespace,
     CreateNamespaceResponse,
     DescribeNamespaceResponse,
     DropNamespaceResponse,
@@ -111,7 +112,7 @@ class RemoteDBConnection(DBConnection):
     @override
     def list_namespaces(
         self,
-        namespace: Optional[List[str]] = None,
+        namespace_path: Optional[List[str]] = None,
         page_token: Optional[str] = None,
         limit: Optional[int] = None,
     ) -> ListNamespacesResponse:
@@ -119,7 +120,7 @@ class RemoteDBConnection(DBConnection):
 
         Parameters
         ----------
-        namespace: List[str], optional
+        namespace_path: List[str], optional
             The parent namespace to list namespaces in.
             None or empty list represents root namespace.
         page_token: str, optional
@@ -133,18 +134,18 @@ class RemoteDBConnection(DBConnection):
         ListNamespacesResponse
             Response containing namespace names and optional page_token for pagination.
         """
-        if namespace is None:
+        if namespace_path is None:
-            namespace = []
+            namespace_path = []
         return LOOP.run(
             self._conn.list_namespaces(
-                namespace=namespace, page_token=page_token, limit=limit
+                namespace_path=namespace_path, page_token=page_token, limit=limit
             )
         )
 
     @override
     def create_namespace(
         self,
-        namespace: List[str],
+        namespace_path: List[str],
         mode: Optional[str] = None,
         properties: Optional[Dict[str, str]] = None,
     ) -> CreateNamespaceResponse:
@@ -152,7 +153,7 @@ class RemoteDBConnection(DBConnection):
 
         Parameters
         ----------
-        namespace: List[str]
+        namespace_path: List[str]
             The namespace identifier to create.
         mode: str, optional
             Creation mode - "create" (fail if exists), "exist_ok" (skip if exists),
@@ -167,14 +168,14 @@ class RemoteDBConnection(DBConnection):
         """
         return LOOP.run(
             self._conn.create_namespace(
-                namespace=namespace, mode=mode, properties=properties
+                namespace_path=namespace_path, mode=mode, properties=properties
             )
         )
 
     @override
     def drop_namespace(
         self,
-        namespace: List[str],
+        namespace_path: List[str],
         mode: Optional[str] = None,
         behavior: Optional[str] = None,
     ) -> DropNamespaceResponse:
@@ -182,7 +183,7 @@ class RemoteDBConnection(DBConnection):
 
         Parameters
         ----------
-        namespace: List[str]
+        namespace_path: List[str]
             The namespace identifier to drop.
         mode: str, optional
             Whether to skip if not exists ("SKIP") or fail ("FAIL"). Case insensitive.
@@ -196,16 +197,20 @@ class RemoteDBConnection(DBConnection):
             Response containing properties and transaction_id if applicable.
         """
         return LOOP.run(
-            self._conn.drop_namespace(namespace=namespace, mode=mode, behavior=behavior)
+            self._conn.drop_namespace(
+                namespace_path=namespace_path, mode=mode, behavior=behavior
+            )
         )
 
     @override
-    def describe_namespace(self, namespace: List[str]) -> DescribeNamespaceResponse:
+    def describe_namespace(
+        self, namespace_path: List[str]
+    ) -> DescribeNamespaceResponse:
         """Describe a namespace.
 
         Parameters
         ----------
-        namespace: List[str]
+        namespace_path: List[str]
             The namespace identifier to describe.
 
         Returns
@@ -213,12 +218,12 @@ class RemoteDBConnection(DBConnection):
         DescribeNamespaceResponse
             Response containing the namespace properties.
         """
-        return LOOP.run(self._conn.describe_namespace(namespace=namespace))
+        return LOOP.run(self._conn.describe_namespace(namespace_path=namespace_path))
 
     @override
     def list_tables(
         self,
-        namespace: Optional[List[str]] = None,
+        namespace_path: Optional[List[str]] = None,
         page_token: Optional[str] = None,
         limit: Optional[int] = None,
     ) -> ListTablesResponse:
@@ -226,7 +231,7 @@ class RemoteDBConnection(DBConnection):
 
         Parameters
         ----------
-        namespace: List[str], optional
+        namespace_path: List[str], optional
             The namespace to list tables in.
             None or empty list represents root namespace.
         page_token: str, optional
@@ -240,11 +245,11 @@ class RemoteDBConnection(DBConnection):
         ListTablesResponse
             Response containing table names and optional page_token for pagination.
         """
-        if namespace is None:
+        if namespace_path is None:
-            namespace = []
+            namespace_path = []
         return LOOP.run(
             self._conn.list_tables(
-                namespace=namespace, page_token=page_token, limit=limit
+                namespace_path=namespace_path, page_token=page_token, limit=limit
             )
         )
 
@@ -254,7 +259,7 @@ class RemoteDBConnection(DBConnection):
         page_token: Optional[str] = None,
         limit: int = 10,
         *,
-        namespace: Optional[List[str]] = None,
+        namespace_path: Optional[List[str]] = None,
     ) -> Iterable[str]:
         """List the names of all tables in the database.
 
@@ -263,7 +268,7 @@ class RemoteDBConnection(DBConnection):
 
         Parameters
         ----------
-        namespace: List[str], default []
+        namespace_path: List[str], default []
             The namespace to list tables in.
             Empty list represents root namespace.
         page_token: str
@@ -282,11 +287,11 @@ class RemoteDBConnection(DBConnection):
             DeprecationWarning,
             stacklevel=2,
         )
-        if namespace is None:
+        if namespace_path is None:
-            namespace = []
+            namespace_path = []
         return LOOP.run(
             self._conn.table_names(
-                namespace=namespace, start_after=page_token, limit=limit
+                namespace_path=namespace_path, start_after=page_token, limit=limit
             )
         )
 
@@ -295,7 +300,7 @@ class RemoteDBConnection(DBConnection):
         self,
         name: str,
         *,
-        namespace: Optional[List[str]] = None,
+        namespace_path: Optional[List[str]] = None,
         storage_options: Optional[Dict[str, str]] = None,
         index_cache_size: Optional[int] = None,
     ) -> Table:
@@ -305,7 +310,7 @@ class RemoteDBConnection(DBConnection):
         ----------
         name: str
             The name of the table.
-        namespace: List[str], optional
+        namespace_path: List[str], optional
             The namespace to open the table from.
             None or empty list represents root namespace.
 
@@ -315,15 +320,15 @@ class RemoteDBConnection(DBConnection):
         """
         from .table import RemoteTable
 
-        if namespace is None:
+        if namespace_path is None:
-            namespace = []
+            namespace_path = []
         if index_cache_size is not None:
             logging.info(
                 "index_cache_size is ignored in LanceDb Cloud"
                 " (there is no local cache to configure)"
             )
 
-        table = LOOP.run(self._conn.open_table(name, namespace=namespace))
+        table = LOOP.run(self._conn.open_table(name, namespace_path=namespace_path))
         return RemoteTable(table, self.db_name)
 
     def clone_table(
@@ -331,7 +336,7 @@ class RemoteDBConnection(DBConnection):
         target_table_name: str,
         source_uri: str,
         *,
-        target_namespace: Optional[List[str]] = None,
+        target_namespace_path: Optional[List[str]] = None,
         source_version: Optional[int] = None,
         source_tag: Optional[str] = None,
         is_shallow: bool = True,
@@ -344,7 +349,7 @@ class RemoteDBConnection(DBConnection):
             The name of the target table to create.
         source_uri: str
             The URI of the source table to clone from.
-        target_namespace: List[str], optional
+        target_namespace_path: List[str], optional
             The namespace for the target table.
             None or empty list represents root namespace.
         source_version: int, optional
@@ -361,13 +366,13 @@ class RemoteDBConnection(DBConnection):
         """
         from .table import RemoteTable
 
-        if target_namespace is None:
+        if target_namespace_path is None:
-            target_namespace = []
+            target_namespace_path = []
         table = LOOP.run(
             self._conn.clone_table(
                 target_table_name,
                 source_uri,
-                target_namespace=target_namespace,
+                target_namespace_path=target_namespace_path,
                 source_version=source_version,
                 source_tag=source_tag,
                 is_shallow=is_shallow,
@@ -387,7 +392,7 @@ class RemoteDBConnection(DBConnection):
         exist_ok: bool = False,
         embedding_functions: Optional[List[EmbeddingFunctionConfig]] = None,
         *,
-        namespace: Optional[List[str]] = None,
+        namespace_path: Optional[List[str]] = None,
     ) -> Table:
         """Create a [Table][lancedb.table.Table] in the database.
 
@@ -395,7 +400,7 @@ class RemoteDBConnection(DBConnection):
         ----------
         name: str
             The name of the table.
-        namespace: List[str], optional
+        namespace_path: List[str], optional
             The namespace to create the table in.
             None or empty list represents root namespace.
         data: The data to initialize the table, *optional*
@@ -495,8 +500,8 @@ class RemoteDBConnection(DBConnection):
                 mode = "exist_ok"
             elif not mode:
                 mode = "exist_ok"
-        if namespace is None:
+        if namespace_path is None:
-            namespace = []
+            namespace_path = []
         validate_table_name(name)
         if embedding_functions is not None:
             logging.warning(
@@ -511,7 +516,7 @@ class RemoteDBConnection(DBConnection):
             self._conn.create_table(
                 name,
                 data,
-                namespace=namespace,
+                namespace_path=namespace_path,
                 mode=mode,
                 schema=schema,
                 on_bad_vectors=on_bad_vectors,
@@ -521,28 +526,28 @@ class RemoteDBConnection(DBConnection):
         return RemoteTable(table, self.db_name)
 
     @override
-    def drop_table(self, name: str, namespace: Optional[List[str]] = None):
+    def drop_table(self, name: str, namespace_path: Optional[List[str]] = None):
         """Drop a table from the database.
 
         Parameters
         ----------
         name: str
             The name of the table.
-        namespace: List[str], optional
+        namespace_path: List[str], optional
             The namespace to drop the table from.
             None or empty list represents root namespace.
         """
-        if namespace is None:
+        if namespace_path is None:
-            namespace = []
+            namespace_path = []
-        LOOP.run(self._conn.drop_table(name, namespace=namespace))
+        LOOP.run(self._conn.drop_table(name, namespace_path=namespace_path))
 
     @override
     def rename_table(
         self,
         cur_name: str,
         new_name: str,
-        cur_namespace: Optional[List[str]] = None,
+        cur_namespace_path: Optional[List[str]] = None,
-        new_namespace: Optional[List[str]] = None,
+        new_namespace_path: Optional[List[str]] = None,
     ):
         """Rename a table in the database.
 
@@ -553,19 +558,32 @@ class RemoteDBConnection(DBConnection):
         new_name: str
             The new name of the table.
         """
-        if cur_namespace is None:
+        if cur_namespace_path is None:
-            cur_namespace = []
+            cur_namespace_path = []
-        if new_namespace is None:
+        if new_namespace_path is None:
-            new_namespace = []
+            new_namespace_path = []
         LOOP.run(
             self._conn.rename_table(
                 cur_name,
                 new_name,
-                cur_namespace=cur_namespace,
+                cur_namespace_path=cur_namespace_path,
-                new_namespace=new_namespace,
+                new_namespace_path=new_namespace_path,
             )
         )
 
+    @override
+    def namespace_client(self) -> LanceNamespace:
+        """Get the equivalent namespace client for this connection.
+
+        Returns a RestNamespace with the same URI and authentication headers.
+
+        Returns
+        -------
+        LanceNamespace
+            The namespace client for this connection.
+        """
+        return LOOP.run(self._conn.namespace_client())
+
     async def close(self):
         """Close the connection to the database."""
         self._conn.close()

python/python/lancedb/remote/table.py

@@ -22,6 +22,7 @@ from lancedb.index import (
     FTS,
     BTree,
     Bitmap,
+    HnswFlat,
     HnswSq,
     IvfFlat,
     IvfPq,
@@ -39,6 +40,7 @@ from lancedb.table import _normalize_progress
 
 from ..query import LanceVectorQueryBuilder, LanceQueryBuilder, LanceTakeQueryBuilder
 from ..table import AsyncTable, IndexStatistics, Query, Table, Tags
+from ..types import BaseTokenizerType
 
 
 class RemoteTable(Table):
@@ -167,7 +169,7 @@ class RemoteTable(Table):
         wait_timeout: Optional[timedelta] = None,
         with_position: bool = False,
         # tokenizer configs:
-        base_tokenizer: str = "simple",
+        base_tokenizer: BaseTokenizerType = "simple",
         language: str = "English",
         max_token_length: Optional[int] = 40,
         lower_case: bool = True,
@@ -284,13 +286,15 @@ class RemoteTable(Table):
             )
         elif index_type == "IVF_HNSW_SQ":
             config = HnswSq(distance_type=metric, num_partitions=num_partitions)
+        elif index_type == "IVF_HNSW_FLAT":
+            config = HnswFlat(distance_type=metric, num_partitions=num_partitions)
         elif index_type == "IVF_FLAT":
             config = IvfFlat(distance_type=metric, num_partitions=num_partitions)
         else:
             raise ValueError(
                 f"Unknown vector index type: {index_type}. Valid options are"
                 " 'IVF_FLAT', 'IVF_PQ', 'IVF_RQ', 'IVF_SQ',"
-                " 'IVF_HNSW_PQ', 'IVF_HNSW_SQ'"
+                " 'IVF_HNSW_PQ', 'IVF_HNSW_SQ', 'IVF_HNSW_FLAT'"
             )
 
         LOOP.run(

python/python/lancedb/table.py

@@ -57,6 +57,7 @@ from .index import (
     LabelList,
     HnswPq,
     HnswSq,
+    HnswFlat,
     FTS,
 )
 from .merge import LanceMergeInsertBuilder
@@ -86,10 +87,62 @@ from .util import (
 )
 from .index import lang_mapping
 
+_MODEL_BACKED_TOKENIZER_PREFIXES = ("jieba", "lindera")
+_MODEL_BACKED_TOKENIZER_ERRORS = (
+    "unknown base tokenizer",
+    "Invalid directory path:",
+    "Failed to load Jieba",
+    "Failed to load tokenizer config",
+    "Failed to initialize default tokenizer",
+)
+
+
+def _add_unique_note(exception: BaseException, note: str) -> None:
+    existing_notes = getattr(exception, "__notes__", ()) or ()
+    message = (
+        exception.args[0]
+        if exception.args and isinstance(exception.args[0], str)
+        else ""
+    )
+    if note not in existing_notes and note not in message:
+        add_note(exception, note)
+
+
+def _is_model_backed_tokenizer(base_tokenizer: str) -> bool:
+    return any(
+        base_tokenizer == prefix or base_tokenizer.startswith(f"{prefix}/")
+        for prefix in _MODEL_BACKED_TOKENIZER_PREFIXES
+    )
+
+
+def _maybe_add_fts_error_note(
+    exception: BaseException, *, base_tokenizer: str, language: Optional[str] = None
+) -> None:
+    message = str(exception)
+    if language is not None and "not support the requested language" in message:
+        supported_langs = ", ".join(lang_mapping.values())
+        _add_unique_note(exception, f"Supported languages: {supported_langs}")
+        return
+
+    if not _is_model_backed_tokenizer(base_tokenizer):
+        return
+
+    if not any(marker in message for marker in _MODEL_BACKED_TOKENIZER_ERRORS):
+        return
+
+    _add_unique_note(
+        exception,
+        "Model-backed tokenizers such as 'jieba/default' and 'lindera/ipadic' "
+        "require tokenizer models in Lance's language model home. Set "
+        "LANCE_LANGUAGE_MODEL_HOME to override the default platform data "
+        "directory under 'lance/language_models'. Expected layouts include "
+        "'<model-home>/jieba/default/...' and "
+        "'<model-home>/lindera/ipadic/...'.",
+    )
+
 
 if TYPE_CHECKING:
     from .db import LanceDBConnection
-    from .io import StorageOptionsProvider
     from ._lancedb import (
         Table as LanceDBTable,
         OptimizeStats,
@@ -192,7 +245,7 @@ def _into_pyarrow_reader(
             f"Unknown data type {type(data)}. "
             "Supported types: list of dicts, pandas DataFrame, polars DataFrame, "
             "pyarrow Table/RecordBatch, or Pydantic models. "
-            "See https://lancedb.com/docs/tables/ for examples."
+            "See https://docs.lancedb.com/tables/ for examples."
         )
 
 
@@ -271,15 +324,17 @@ def _sanitize_data(
         reader,
         on_bad_vectors=on_bad_vectors,
         fill_value=fill_value,
+        target_schema=target_schema,
+        metadata=metadata,
     )
 
     if target_schema is None:
         target_schema, reader = _infer_target_schema(reader)
 
     if metadata:
-        new_metadata = target_schema.metadata or {}
+        target_schema = target_schema.with_metadata(
-        new_metadata.update(metadata)
+            _merge_metadata(target_schema.metadata, metadata)
-        target_schema = target_schema.with_metadata(new_metadata)
+        )
 
     _validate_schema(target_schema)
     reader = _cast_to_target_schema(reader, target_schema, allow_subschema)
@@ -295,7 +350,7 @@ def _cast_to_target_schema(
     # pa.Table.cast expects field order not to be changed.
     # Lance doesn't care about field order, so we don't need to rearrange fields
     # to match the target schema. We just need to correctly cast the fields.
-    if reader.schema == target_schema:
+    if reader.schema.equals(target_schema, check_metadata=True):
         # Fast path when the schemas are already the same
         return reader
 
@@ -315,7 +370,13 @@ def _cast_to_target_schema(
     def gen():
         for batch in reader:
             # Table but not RecordBatch has cast.
-            yield pa.Table.from_batches([batch]).cast(reordered_schema).to_batches()[0]
+            cast_batches = (
+                pa.Table.from_batches([batch]).cast(reordered_schema).to_batches()
+            )
+            if cast_batches:
+                yield pa.RecordBatch.from_arrays(
+                    cast_batches[0].columns, schema=reordered_schema
+                )
 
     return pa.RecordBatchReader.from_batches(reordered_schema, gen())
 
@@ -333,37 +394,51 @@ def _align_field_types(
         if target_field is None:
             raise ValueError(f"Field '{field.name}' not found in target schema")
         if pa.types.is_struct(target_field.type):
-            new_type = pa.struct(
+            if pa.types.is_struct(field.type):
-                _align_field_types(
+                new_type = pa.struct(
-                    field.type.fields,
+                    _align_field_types(
-                    target_field.type.fields,
+                        field.type.fields,
+                        target_field.type.fields,
+                    )
                 )
-            )
+            else:
+                new_type = target_field.type
         elif pa.types.is_list(target_field.type):
-            new_type = pa.list_(
+            if _is_list_like(field.type):
-                _align_field_types(
+                new_type = pa.list_(
-                    [field.type.value_field],
+                    _align_field_types(
-                    [target_field.type.value_field],
+                        [field.type.value_field],
-                )[0]
+                        [target_field.type.value_field],
-            )
+                    )[0]
+                )
+            else:
+                new_type = target_field.type
         elif pa.types.is_large_list(target_field.type):
-            new_type = pa.large_list(
+            if _is_list_like(field.type):
-                _align_field_types(
+                new_type = pa.large_list(
-                    [field.type.value_field],
+                    _align_field_types(
-                    [target_field.type.value_field],
+                        [field.type.value_field],
-                )[0]
+                        [target_field.type.value_field],
-            )
+                    )[0]
+                )
+            else:
+                new_type = target_field.type
         elif pa.types.is_fixed_size_list(target_field.type):
-            new_type = pa.list_(
+            if _is_list_like(field.type):
-                _align_field_types(
+                new_type = pa.list_(
-                    [field.type.value_field],
+                    _align_field_types(
-                    [target_field.type.value_field],
+                        [field.type.value_field],
-                )[0],
+                        [target_field.type.value_field],
-                target_field.type.list_size,
+                    )[0],
-            )
+                    target_field.type.list_size,
+                )
+            else:
+                new_type = target_field.type
         else:
             new_type = target_field.type
-        new_fields.append(pa.field(field.name, new_type, field.nullable))
+        new_fields.append(
+            pa.field(field.name, new_type, field.nullable, target_field.metadata)
+        )
     return new_fields
 
 
@@ -441,6 +516,7 @@ def sanitize_create_table(
             schema = data.schema
 
     if metadata:
+        metadata = _merge_metadata(schema.metadata, metadata)
         schema = schema.with_metadata(metadata)
         # Need to apply metadata to the data as well
         if isinstance(data, pa.Table):
@@ -493,9 +569,9 @@ def _append_vector_columns(
     vector columns to the table.
     """
     if schema is None:
-        metadata = metadata or {}
+        metadata = _merge_metadata(metadata)
     else:
-        metadata = schema.metadata or metadata or {}
+        metadata = _merge_metadata(schema.metadata, metadata)
     functions = EmbeddingFunctionRegistry.get_instance().parse_functions(metadata)
 
     if not functions:
@@ -921,29 +997,29 @@ class Table(ABC):
         Parameters
         ----------
         field_names: str or list of str
-            The name(s) of the field to index.
+            The name of the field to index. Native FTS indexes can only be
-            If ``use_tantivy`` is False (default), only a single field name
+            created on a single field at a time. To search over multiple text
-            (str) is supported. To index multiple fields, create a separate
+            fields, create a separate FTS index for each field.
-            FTS index for each field.
         replace: bool, default False
             If True, replace the existing index if it exists. Note that this is
             not yet an atomic operation; the index will be temporarily
             unavailable while the new index is being created.
         writer_heap_size: int, default 1GB
-            Only available with use_tantivy=True
+            Deprecated legacy Tantivy parameter. Any value other than the
+            default raises an error.
         ordering_field_names:
-            A list of unsigned type fields to index to optionally order
+            Deprecated legacy Tantivy parameter. Setting this raises an error.
-            results on at search time.
-            only available with use_tantivy=True
         tokenizer_name: str, default "default"
-            The tokenizer to use for the index. Can be "raw", "default" or the 2 letter
+            A compatibility alias for native tokenizer configs. Can be "raw",
-            language code followed by "_stem". So for english it would be "en_stem".
+            "default" or the 2 letter language code followed by "_stem". So
-            For available languages see: https://docs.rs/tantivy/latest/tantivy/tokenizer/enum.Language.html
+            for english it would be "en_stem". For new native FTS indexes, use
+            ``base_tokenizer`` directly; ``tokenizer_name`` is a legacy
+            compatibility alias and does not expose model-backed tokenizer names
+            such as ``jieba/default`` or ``lindera/ipadic``.
         use_tantivy: bool, default False
-            If True, use the legacy full-text search implementation based on tantivy.
+            Deprecated legacy Tantivy parameter. Setting this to True raises an
-            If False, use the new full-text search implementation based on lance-index.
+            error.
         with_position: bool, default False
-            Only available with use_tantivy=False
             If False, do not store the positions of the terms in the text.
             This can reduce the size of the index and improve indexing speed.
             But it will raise an exception for phrase queries.
@@ -953,8 +1029,11 @@ class Table(ABC):
             - "whitespace": Split text by whitespace, but not punctuation.
             - "raw": No tokenization. The entire text is treated as a single token.
             - "ngram": N-Gram tokenizer.
+            - "jieba/*": Jieba tokenizer loaded from Lance's language model home.
+            - "lindera/*": Lindera tokenizer loaded from Lance's language model home.
         language : str, default "English"
-            The language to use for tokenization.
+            The language to use for stemming and stop-word removal. This is not
+            the primary way to enable CJK tokenization.
         max_token_length : int, default 40
             The maximum token length to index. Tokens longer than this length will be
             ignored.
@@ -980,6 +1059,13 @@ class Table(ABC):
             The timeout to wait if indexing is asynchronous.
         name: str, optional
             The name of the index. If not provided, a default name will be generated.
+
+        Notes
+        -----
+        Model-backed tokenizers such as ``jieba/default`` and ``lindera/ipadic``
+        require tokenizer models in Lance's language model home. Set
+        ``LANCE_LANGUAGE_MODEL_HOME`` to override the default platform data
+        directory under ``lance/language_models``.
         """
         raise NotImplementedError
 
@@ -1724,6 +1810,16 @@ class Table(ABC):
         index_exists = fs.get_file_info(path).type != pa_fs.FileType.NotFound
         return (path, fs, index_exists)
 
+    def _ensure_no_legacy_fts_index(self):
+        path, _, exists = self._get_fts_index_path()
+        if exists:
+            raise ValueError(
+                "Legacy Tantivy FTS index detected at "
+                f"{path}. Tantivy-based FTS has been removed. "
+                "Delete the legacy index and recreate it with "
+                "table.create_fts_index(...)."
+            )
+
     @abstractmethod
     def uses_v2_manifest_paths(self) -> bool:
         """
@@ -1776,30 +1872,30 @@ class LanceTable(Table):
         connection: "LanceDBConnection",
         name: str,
         *,
-        namespace: Optional[List[str]] = None,
+        namespace_path: Optional[List[str]] = None,
         storage_options: Optional[Dict[str, str]] = None,
-        storage_options_provider: Optional["StorageOptionsProvider"] = None,
         index_cache_size: Optional[int] = None,
         location: Optional[str] = None,
         namespace_client: Optional[Any] = None,
         managed_versioning: Optional[bool] = None,
+        pushdown_operations: Optional[set] = None,
         _async: AsyncTable = None,
     ):
-        if namespace is None:
+        if namespace_path is None:
-            namespace = []
+            namespace_path = []
         self._conn = connection
-        self._namespace = namespace
+        self._namespace_path = namespace_path
         self._location = location  # Store location for use in _dataset_path
         self._namespace_client = namespace_client
+        self._pushdown_operations = pushdown_operations or set()
         if _async is not None:
             self._table = _async
         else:
             self._table = LOOP.run(
                 connection._conn.open_table(
                     name,
-                    namespace=namespace,
+                    namespace_path=namespace_path,
                     storage_options=storage_options,
-                    storage_options_provider=storage_options_provider,
                     index_cache_size=index_cache_size,
                     location=location,
                     namespace_client=namespace_client,
@@ -1814,13 +1910,13 @@ class LanceTable(Table):
     @property
     def namespace(self) -> List[str]:
         """Return the namespace path of the table."""
-        return self._namespace
+        return self._namespace_path
 
     @property
     def id(self) -> str:
         """Return the full identifier of the table (namespace$name)."""
-        if self._namespace:
+        if self._namespace_path:
-            return "$".join(self._namespace + [self.name])
+            return "$".join(self._namespace_path + [self.name])
         return self.name
 
     @classmethod
@@ -1841,26 +1937,26 @@ class LanceTable(Table):
         db,
         name,
         *,
-        namespace: Optional[List[str]] = None,
+        namespace_path: Optional[List[str]] = None,
         storage_options: Optional[Dict[str, str]] = None,
-        storage_options_provider: Optional["StorageOptionsProvider"] = None,
         index_cache_size: Optional[int] = None,
         location: Optional[str] = None,
         namespace_client: Optional[Any] = None,
         managed_versioning: Optional[bool] = None,
+        pushdown_operations: Optional[set] = None,
     ):
-        if namespace is None:
+        if namespace_path is None:
-            namespace = []
+            namespace_path = []
         tbl = cls(
             db,
             name,
-            namespace=namespace,
+            namespace_path=namespace_path,
             storage_options=storage_options,
-            storage_options_provider=storage_options_provider,
             index_cache_size=index_cache_size,
             location=location,
             namespace_client=namespace_client,
             managed_versioning=managed_versioning,
+            pushdown_operations=pushdown_operations,
         )
 
         # check the dataset exists
@@ -1893,11 +1989,11 @@ class LanceTable(Table):
             )
 
         if self._namespace_client is not None:
-            table_id = self._namespace + [self.name]
+            table_id = self._namespace_path + [self.name]
             return lance.dataset(
                 version=self.version,
                 storage_options=self._conn.storage_options,
-                namespace=self._namespace_client,
+                namespace_client=self._namespace_client,
                 table_id=table_id,
                 **kwargs,
             )
@@ -2141,7 +2237,13 @@ class LanceTable(Table):
         index_cache_size: Optional[int] = None,
         num_bits: int = 8,
         index_type: Literal[
-            "IVF_FLAT", "IVF_SQ", "IVF_PQ", "IVF_RQ", "IVF_HNSW_SQ", "IVF_HNSW_PQ"
+            "IVF_FLAT",
+            "IVF_SQ",
+            "IVF_PQ",
+            "IVF_RQ",
+            "IVF_HNSW_SQ",
+            "IVF_HNSW_PQ",
+            "IVF_HNSW_FLAT",
         ] = "IVF_PQ",
         max_iterations: int = 50,
         sample_rate: int = 256,
@@ -2228,6 +2330,16 @@ class LanceTable(Table):
                 ef_construction=ef_construction,
                 target_partition_size=target_partition_size,
             )
+        elif index_type == "IVF_HNSW_FLAT":
+            config = HnswFlat(
+                distance_type=metric,
+                num_partitions=num_partitions,
+                max_iterations=max_iterations,
+                sample_rate=sample_rate,
+                m=m,
+                ef_construction=ef_construction,
+                target_partition_size=target_partition_size,
+            )
         else:
             raise ValueError(f"Unknown index type {index_type}")
 
@@ -2383,41 +2495,57 @@ class LanceTable(Table):
         prefix_only: bool = False,
         name: Optional[str] = None,
     ):
-        if not use_tantivy:
+        self._ensure_no_legacy_fts_index()
-            if not isinstance(field_names, str):
-                raise ValueError(
-                    "Native FTS indexes can only be created on a single field "
-                    "at a time. To search over multiple text fields, create a "
-                    "separate FTS index for each field."
-                )
 
-            if tokenizer_name is None:
+        if use_tantivy:
-                tokenizer_configs = {
+            raise ValueError(
-                    "base_tokenizer": base_tokenizer,
+                "Tantivy-based FTS has been removed. "
-                    "language": language,
+                "Remove use_tantivy and recreate the index with native FTS."
-                    "with_position": with_position,
+            )
-                    "max_token_length": max_token_length,
+        if ordering_field_names is not None:
-                    "lower_case": lower_case,
+            raise ValueError(
-                    "stem": stem,
+                "ordering_field_names was only supported by the removed "
-                    "remove_stop_words": remove_stop_words,
+                "Tantivy-based FTS implementation."
-                    "ascii_folding": ascii_folding,
+            )
-                    "ngram_min_length": ngram_min_length,
+        if writer_heap_size != 1024 * 1024 * 1024:
-                    "ngram_max_length": ngram_max_length,
+            raise ValueError(
-                    "prefix_only": prefix_only,
+                "writer_heap_size was only supported by the removed "
-                }
+                "Tantivy-based FTS implementation."
-            else:
+            )
-                tokenizer_configs = self.infer_tokenizer_configs(tokenizer_name)
+        if not isinstance(field_names, str):
-
+            raise ValueError(
-            config = FTS(
+                "Native FTS indexes can only be created on a single field "
-                **tokenizer_configs,
+                "at a time. To search over multiple text fields, create a "
+                "separate FTS index for each field."
+            )
+        if "." in field_names:
+            raise ValueError(
+                "Native FTS indexes can only be created on top-level fields. "
+                f"Received nested field path: {field_names!r}."
             )
 
-            # delete the existing legacy index if it exists
+        if tokenizer_name is None:
-            if replace:
+            tokenizer_configs = {
-                path, fs, exist = self._get_fts_index_path()
+                "base_tokenizer": base_tokenizer,
-                if exist:
+                "language": language,
-                    fs.delete_dir(path)
+                "with_position": with_position,
+                "max_token_length": max_token_length,
+                "lower_case": lower_case,
+                "stem": stem,
+                "remove_stop_words": remove_stop_words,
+                "ascii_folding": ascii_folding,
+                "ngram_min_length": ngram_min_length,
+                "ngram_max_length": ngram_max_length,
+                "prefix_only": prefix_only,
+            }
+        else:
+            tokenizer_configs = self.infer_tokenizer_configs(tokenizer_name)
 
+        config = FTS(
+            **tokenizer_configs,
+        )
+
+        try:
             LOOP.run(
                 self._table.create_index(
                     field_names,
@@ -2426,42 +2554,13 @@ class LanceTable(Table):
                     name=name,
                 )
             )
-            return
+        except (ValueError, RuntimeError) as e:
-
+            _maybe_add_fts_error_note(
-        from .fts import create_index, populate_index
+                e,
-
+                base_tokenizer=config.base_tokenizer,
-        if isinstance(field_names, str):
+                language=config.language,
-            field_names = [field_names]
-
-        if isinstance(ordering_field_names, str):
-            ordering_field_names = [ordering_field_names]
-
-        path, fs, exist = self._get_fts_index_path()
-        if exist:
-            if not replace:
-                raise ValueError("Index already exists. Use replace=True to overwrite.")
-            fs.delete_dir(path)
-
-        if not isinstance(fs, pa_fs.LocalFileSystem):
-            raise NotImplementedError(
-                "Full-text search is only supported on the local filesystem"
             )
-
+            raise e
-        if tokenizer_name is None:
-            tokenizer_name = "default"
-        index = create_index(
-            path,
-            field_names,
-            ordering_fields=ordering_field_names,
-            tokenizer_name=tokenizer_name,
-        )
-        populate_index(
-            index,
-            self,
-            field_names,
-            ordering_fields=ordering_field_names,
-            writer_heap_size=writer_heap_size,
-        )
 
     @staticmethod
     def infer_tokenizer_configs(tokenizer_name: str) -> dict:
@@ -2803,13 +2902,13 @@ class LanceTable(Table):
         fill_value: float = 0.0,
         embedding_functions: Optional[List[EmbeddingFunctionConfig]] = None,
         *,
-        namespace: Optional[List[str]] = None,
+        namespace_path: Optional[List[str]] = None,
         storage_options: Optional[Dict[str, str | bool]] = None,
-        storage_options_provider: Optional["StorageOptionsProvider"] = None,
         data_storage_version: Optional[str] = None,
         enable_v2_manifest_paths: Optional[bool] = None,
         location: Optional[str] = None,
         namespace_client: Optional[Any] = None,
+        pushdown_operations: Optional[set] = None,
     ):
         """
         Create a new table.
@@ -2864,13 +2963,14 @@ class LanceTable(Table):
             Deprecated.  Set `storage_options` when connecting to the database and set
             `new_table_enable_v2_manifest_paths` in the options.
         """
-        if namespace is None:
+        if namespace_path is None:
-            namespace = []
+            namespace_path = []
         self = cls.__new__(cls)
         self._conn = db
-        self._namespace = namespace
+        self._namespace_path = namespace_path
         self._location = location
         self._namespace_client = namespace_client
+        self._pushdown_operations = pushdown_operations or set()
 
         if data_storage_version is not None:
             warnings.warn(
@@ -2903,10 +3003,10 @@ class LanceTable(Table):
                 on_bad_vectors=on_bad_vectors,
                 fill_value=fill_value,
                 embedding_functions=embedding_functions,
-                namespace=namespace,
+                namespace_path=namespace_path,
                 storage_options=storage_options,
-                storage_options_provider=storage_options_provider,
                 location=location,
+                namespace_client=namespace_client,
             )
         )
         return self
@@ -2974,6 +3074,15 @@ class LanceTable(Table):
         batch_size: Optional[int] = None,
         timeout: Optional[timedelta] = None,
     ) -> pa.RecordBatchReader:
+        if (
+            "QueryTable" in self._pushdown_operations
+            and self._namespace_client is not None
+        ):
+            from lancedb.namespace import _execute_server_side_query
+
+            table_id = self._namespace_path + [self.name]
+            return _execute_server_side_query(self._namespace_client, table_id, query)
+
         async_iter = LOOP.run(
             self._table._execute_query(query, batch_size=batch_size, timeout=timeout)
         )
@@ -3203,43 +3312,157 @@ def _handle_bad_vectors(
     reader: pa.RecordBatchReader,
     on_bad_vectors: Literal["error", "drop", "fill", "null"] = "error",
     fill_value: float = 0.0,
+    target_schema: Optional[pa.Schema] = None,
+    metadata: Optional[dict] = None,
 ) -> pa.RecordBatchReader:
-    vector_columns = []
+    vector_columns = _find_vector_columns(reader.schema, target_schema, metadata)
+    if not vector_columns:
+        return reader
 
-    for field in reader.schema:
+    output_schema = _vector_output_schema(reader.schema, vector_columns)
-        # They can provide a 'vector' column that isn't yet a FSL
-        named_vector_col = (
-            (
-                pa.types.is_list(field.type)
-                or pa.types.is_large_list(field.type)
-                or pa.types.is_fixed_size_list(field.type)
-            )
-            and pa.types.is_floating(field.type.value_type)
-            and field.name == VECTOR_COLUMN_NAME
-        )
-        # TODO: we're making an assumption that fixed size list of 10 or more
-        # is a vector column. This is definitely a bit hacky.
-        likely_vector_col = (
-            pa.types.is_fixed_size_list(field.type)
-            and pa.types.is_floating(field.type.value_type)
-            and (field.type.list_size >= 10)
-        )
-
-        if named_vector_col or likely_vector_col:
-            vector_columns.append(field.name)
 
     def gen():
         for batch in reader:
-            for name in vector_columns:
+            pending_dims = []
+            for vector_column in vector_columns:
+                dim = vector_column["expected_dim"]
+                if target_schema is not None and dim is None:
+                    dim = _infer_vector_dim(batch[vector_column["name"]])
+                    pending_dims.append(vector_column)
                 batch = _handle_bad_vector_column(
                     batch,
-                    vector_column_name=name,
+                    vector_column_name=vector_column["name"],
                     on_bad_vectors=on_bad_vectors,
                     fill_value=fill_value,
+                    expected_dim=dim,
+                    expected_value_type=vector_column["expected_value_type"],
                 )
-            yield batch
+            for vector_column in pending_dims:
+                if vector_column["expected_dim"] is None:
+                    vector_column["expected_dim"] = _infer_vector_dim(
+                        batch[vector_column["name"]]
+                    )
+            if batch.schema.equals(output_schema, check_metadata=True):
+                yield batch
+                continue
 
-    return pa.RecordBatchReader.from_batches(reader.schema, gen())
+            cast_batches = (
+                pa.Table.from_batches([batch]).cast(output_schema).to_batches()
+            )
+            if cast_batches:
+                yield pa.RecordBatch.from_arrays(
+                    cast_batches[0].columns,
+                    schema=output_schema,
+                )
+
+    return pa.RecordBatchReader.from_batches(output_schema, gen())
+
+
+def _find_vector_columns(
+    reader_schema: pa.Schema,
+    target_schema: Optional[pa.Schema],
+    metadata: Optional[dict],
+) -> List[dict]:
+    if target_schema is None:
+        vector_columns = []
+        for field in reader_schema:
+            named_vector_col = (
+                _is_list_like(field.type)
+                and pa.types.is_floating(field.type.value_type)
+                and field.name == VECTOR_COLUMN_NAME
+            )
+            likely_vector_col = (
+                pa.types.is_fixed_size_list(field.type)
+                and pa.types.is_floating(field.type.value_type)
+                and (field.type.list_size >= 10)
+            )
+            if named_vector_col or likely_vector_col:
+                vector_columns.append(
+                    {
+                        "name": field.name,
+                        "expected_dim": None,
+                        "expected_value_type": None,
+                    }
+                )
+        return vector_columns
+
+    reader_column_names = set(reader_schema.names)
+    active_metadata = _merge_metadata(target_schema.metadata, metadata)
+    embedding_function_columns = set(
+        EmbeddingFunctionRegistry.get_instance().parse_functions(active_metadata).keys()
+    )
+    vector_columns = []
+    for field in target_schema:
+        if field.name not in reader_column_names:
+            continue
+        if not _is_list_like(field.type) or not pa.types.is_floating(
+            field.type.value_type
+        ):
+            continue
+
+        reader_field = reader_schema.field(field.name)
+        named_vector_col = (
+            field.name in embedding_function_columns
+            or field.name == VECTOR_COLUMN_NAME
+            or (field.name == "embedding" and pa.types.is_fixed_size_list(field.type))
+        )
+        typed_fixed_vector_col = (
+            pa.types.is_fixed_size_list(reader_field.type)
+            and pa.types.is_floating(reader_field.type.value_type)
+            and reader_field.type.list_size >= 10
+        )
+
+        if named_vector_col or typed_fixed_vector_col:
+            vector_columns.append(
+                {
+                    "name": field.name,
+                    "expected_dim": (
+                        field.type.list_size
+                        if pa.types.is_fixed_size_list(field.type)
+                        else None
+                    ),
+                    "expected_value_type": field.type.value_type,
+                }
+            )
+
+    return vector_columns
+
+
+def _vector_output_schema(
+    reader_schema: pa.Schema,
+    vector_columns: List[dict],
+) -> pa.Schema:
+    columns_by_name = {column["name"]: column for column in vector_columns}
+    fields = []
+    for field in reader_schema:
+        column = columns_by_name.get(field.name)
+        if column is None:
+            output_type = field.type
+        else:
+            output_type = _vector_output_type(field, column)
+        fields.append(pa.field(field.name, output_type, field.nullable, field.metadata))
+    return pa.schema(fields, metadata=reader_schema.metadata)
+
+
+def _vector_output_type(field: pa.Field, vector_column: dict) -> pa.DataType:
+    if not _is_list_like(field.type):
+        return field.type
+
+    if vector_column["expected_value_type"] is not None and (
+        pa.types.is_null(field.type.value_type)
+        or pa.types.is_integer(field.type.value_type)
+        or pa.types.is_unsigned_integer(field.type.value_type)
+    ):
+        return pa.list_(vector_column["expected_value_type"])
+
+    if (
+        vector_column["expected_dim"] is not None
+        and pa.types.is_fixed_size_list(field.type)
+        and field.type.list_size != vector_column["expected_dim"]
+    ):
+        return pa.list_(field.type.value_type)
+
+    return field.type
 
 
 def _handle_bad_vector_column(
@@ -3247,6 +3470,8 @@ def _handle_bad_vector_column(
     vector_column_name: str,
     on_bad_vectors: str = "error",
     fill_value: float = 0.0,
+    expected_dim: Optional[int] = None,
+    expected_value_type: Optional[pa.DataType] = None,
 ) -> pa.RecordBatch:
     """
     Ensure that the vector column exists and has type fixed_size_list(float)
@@ -3263,14 +3488,39 @@ def _handle_bad_vector_column(
     fill_value: float, default 0.0
         The value to use when filling vectors. Only used if on_bad_vectors="fill".
     """
+    position = data.column_names.index(vector_column_name)
     vec_arr = data[vector_column_name]
+    if not _is_list_like(vec_arr.type):
+        return data
 
-    has_nan = has_nan_values(vec_arr)
+    if (
+        expected_dim is not None
+        and pa.types.is_fixed_size_list(vec_arr.type)
+        and vec_arr.type.list_size != expected_dim
+    ):
+        vec_arr = pa.array(vec_arr.to_pylist(), type=pa.list_(vec_arr.type.value_type))
+        data = data.set_column(position, vector_column_name, vec_arr)
 
-    if pa.types.is_fixed_size_list(vec_arr.type):
+    if expected_value_type is not None and (
+        pa.types.is_integer(vec_arr.type.value_type)
+        or pa.types.is_unsigned_integer(vec_arr.type.value_type)
+    ):
+        vec_arr = pa.array(vec_arr.to_pylist(), type=pa.list_(expected_value_type))
+        data = data.set_column(position, vector_column_name, vec_arr)
+
+    if pa.types.is_floating(vec_arr.type.value_type):
+        has_nan = has_nan_values(vec_arr)
+    else:
+        has_nan = pa.array([False] * len(vec_arr))
+
+    if expected_dim is not None:
+        dim = expected_dim
+    elif pa.types.is_fixed_size_list(vec_arr.type):
         dim = vec_arr.type.list_size
     else:
-        dim = _modal_list_size(vec_arr)
+        dim = _infer_vector_dim(vec_arr)
+        if dim is None:
+            return data
     has_wrong_dim = pc.not_equal(pc.list_value_length(vec_arr), dim)
 
     has_bad_vectors = pc.any(has_nan).as_py() or pc.any(has_wrong_dim).as_py()
@@ -3308,13 +3558,12 @@ def _handle_bad_vector_column(
                 )
             vec_arr = pc.if_else(
                 is_bad,
-                pa.scalar([fill_value] * dim),
+                pa.scalar([fill_value] * dim, type=vec_arr.type),
                 vec_arr,
             )
         else:
             raise ValueError(f"Invalid value for on_bad_vectors: {on_bad_vectors}")
 
-    position = data.column_names.index(vector_column_name)
     return data.set_column(position, vector_column_name, vec_arr)
 
 
@@ -3335,6 +3584,28 @@ def has_nan_values(arr: Union[pa.ListArray, pa.ChunkedArray]) -> pa.BooleanArray
     return pc.is_in(indices, has_nan_indices)
 
 
+def _is_list_like(data_type: pa.DataType) -> bool:
+    return (
+        pa.types.is_list(data_type)
+        or pa.types.is_large_list(data_type)
+        or pa.types.is_fixed_size_list(data_type)
+    )
+
+
+def _merge_metadata(*metadata_dicts: Optional[dict]) -> dict:
+    merged = {}
+    for metadata in metadata_dicts:
+        if metadata is None:
+            continue
+        for key, value in metadata.items():
+            if isinstance(key, str):
+                key = key.encode("utf-8")
+            if isinstance(value, str):
+                value = value.encode("utf-8")
+            merged[key] = value
+    return merged
+
+
 def _name_suggests_vector_column(field_name: str) -> bool:
     """Check if a field name indicates a vector column."""
     name_lower = field_name.lower()
@@ -3402,6 +3673,16 @@ def _modal_list_size(arr: Union[pa.ListArray, pa.ChunkedArray]) -> int:
     return pc.mode(pc.list_value_length(arr))[0].as_py()["mode"]
 
 
+def _infer_vector_dim(arr: Union[pa.Array, pa.ChunkedArray]) -> Optional[int]:
+    if not _is_list_like(arr.type):
+        return None
+    lengths = pc.list_value_length(arr)
+    lengths = pc.filter(lengths, pc.greater(lengths, 0))
+    if len(lengths) == 0:
+        return None
+    return pc.mode(lengths)[0].as_py()["mode"]
+
+
 def _validate_schema(schema: pa.Schema):
     """
     Make sure the metadata is valid utf8
@@ -3609,7 +3890,18 @@ class AsyncTable:
         *,
         replace: Optional[bool] = None,
         config: Optional[
-            Union[IvfFlat, IvfPq, IvfRq, HnswPq, HnswSq, BTree, Bitmap, LabelList, FTS]
+            Union[
+                IvfFlat,
+                IvfPq,
+                IvfRq,
+                HnswPq,
+                HnswSq,
+                HnswFlat,
+                BTree,
+                Bitmap,
+                LabelList,
+                FTS,
+            ]
         ] = None,
         wait_timeout: Optional[timedelta] = None,
         name: Optional[str] = None,
@@ -3656,6 +3948,7 @@ class AsyncTable:
                     IvfRq,
                     HnswPq,
                     HnswSq,
+                    HnswFlat,
                     BTree,
                     Bitmap,
                     LabelList,
@@ -3675,11 +3968,13 @@ class AsyncTable:
                 name=name,
                 train=train,
             )
-        except ValueError as e:
+        except (ValueError, RuntimeError) as e:
-            if "not support the requested language" in str(e):
+            if isinstance(config, FTS):
-                supported_langs = ", ".join(lang_mapping.values())
+                _maybe_add_fts_error_note(
-                help_msg = f"Supported languages: {supported_langs}"
+                    e,
-                add_note(e, help_msg)
+                    base_tokenizer=config.base_tokenizer,
+                    language=config.language,
+                )
             raise e
 
     async def drop_index(self, name: str) -> None:
@@ -4824,6 +5119,7 @@ class IndexStatistics:
         "IVF_RQ",
         "IVF_HNSW_SQ",
         "IVF_HNSW_PQ",
+        "IVF_HNSW_FLAT",
         "FTS",
         "BTREE",
         "BITMAP",

python/python/lancedb/types.py

@@ -24,6 +24,7 @@ VectorIndexType = Literal[
     "IVF_PQ",
     "IVF_HNSW_SQ",
     "IVF_HNSW_PQ",
+    "IVF_HNSW_FLAT",
     "IVF_RQ",
 ]
 ScalarIndexType = Literal["BTREE", "BITMAP", "LABEL_LIST"]
@@ -31,6 +32,7 @@ IndexType = Literal[
     "IVF_PQ",
     "IVF_HNSW_PQ",
     "IVF_HNSW_SQ",
+    "IVF_HNSW_FLAT",
     "IVF_SQ",
     "FTS",
     "BTREE",
@@ -40,4 +42,5 @@ IndexType = Literal[
 ]
 
 # Tokenizer literals
-BaseTokenizerType = Literal["simple", "raw", "whitespace", "ngram"]
+BuiltinTokenizerType = Literal["simple", "raw", "whitespace", "ngram"]
+BaseTokenizerType = BuiltinTokenizerType | str