Bump version: 0.24.1-beta.0 → 0.24.1-beta.1

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

.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.19.1-beta.5"
+current_version = "0.21.1-beta.0"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

.github/workflows/java.yml

@@ -35,6 +35,9 @@ jobs:
       - uses: Swatinem/rust-cache@v2
         with:
           workspaces: java/core/lancedb-jni
+      - uses: actions-rust-lang/setup-rust-toolchain@v1
+        with:
+          components: rustfmt
       - name: Run cargo fmt
         run: cargo fmt --check
         working-directory: ./java/core/lancedb-jni
@@ -68,6 +71,9 @@ jobs:
       - uses: Swatinem/rust-cache@v2
         with:
           workspaces: java/core/lancedb-jni
+      - uses: actions-rust-lang/setup-rust-toolchain@v1
+        with:
+          components: rustfmt
       - name: Run cargo fmt
         run: cargo fmt --check
         working-directory: ./java/core/lancedb-jni
@@ -110,4 +116,3 @@ jobs:
           -Djdk.reflect.useDirectMethodHandle=false \
           -Dio.netty.tryReflectionSetAccessible=true"
           JAVA_HOME=$JAVA_17 mvn clean test
-  

.github/workflows/make-release-commit.yml

@@ -84,6 +84,7 @@ jobs:
         run: |
           pip install bump-my-version PyGithub packaging
           bash ci/bump_version.sh ${{ inputs.type }} ${{ inputs.bump-minor }} v $COMMIT_BEFORE_BUMP
+          bash ci/update_lockfiles.sh --amend
       - name: Push new version tag
         if: ${{ !inputs.dry_run }}
         uses: ad-m/github-push-action@master
@@ -92,11 +93,3 @@ jobs:
           github_token: ${{ secrets.LANCEDB_RELEASE_TOKEN }}
           branch: ${{ github.ref }}
           tags: true
-      - uses: ./.github/workflows/update_package_lock
-        if: ${{ !inputs.dry_run && inputs.other }}
-        with:
-          github_token: ${{ secrets.GITHUB_TOKEN }}
-      - uses: ./.github/workflows/update_package_lock_nodejs
-        if: ${{ !inputs.dry_run && inputs.other }}
-        with:
-          github_token: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/nodejs.yml

@@ -47,6 +47,9 @@ jobs:
       run: |
         sudo apt update
         sudo apt install -y protobuf-compiler libssl-dev
+    - uses: actions-rust-lang/setup-rust-toolchain@v1
+      with:
+        components: rustfmt, clippy
     - name: Lint
       run: |
         cargo fmt --all -- --check
@@ -113,7 +116,7 @@ jobs:
         set -e
         npm ci
         npm run docs
-        if ! git diff --exit-code; then
+        if ! git diff --exit-code -- . ':(exclude)Cargo.lock'; then
           echo "Docs need to be updated"
           echo "Run 'npm run docs', fix any warnings, and commit the changes."
           exit 1

.github/workflows/npm-publish.yml

@@ -505,6 +505,8 @@ jobs:
     name: vectordb NPM Publish
     needs: [node, node-macos, node-linux-gnu, node-windows]
     runs-on: ubuntu-latest
+    permissions:
+      contents: write
     # Only runs on tags that matches the make-release action
     if: startsWith(github.ref, 'refs/tags/v')
     steps:
@@ -537,6 +539,20 @@ jobs:
         # We need to deprecate the old package to avoid confusion.
         # Each time we publish a new version, it gets undeprecated.
         run: npm deprecate vectordb "Use @lancedb/lancedb instead."
+      - name: Checkout
+        uses: actions/checkout@v4
+        with:
+          ref: main
+      - name: Update package-lock.json
+        run: |
+          git config user.name 'Lance Release'
+          git config user.email 'lance-dev@lancedb.com'
+          bash ci/update_lockfiles.sh
+      - name: Push new commit
+        uses: ad-m/github-push-action@master
+        with:
+          github_token: ${{ secrets.LANCEDB_RELEASE_TOKEN }}
+          branch: main
       - name: Notify Slack Action
         uses: ravsamhq/notify-slack-action@2.3.0
         if: ${{ always() }}
@@ -546,21 +562,3 @@ jobs:
           notification_title: "{workflow} is failing"
         env:
           SLACK_WEBHOOK_URL: ${{ secrets.ACTION_MONITORING_SLACK }}
-
-  update-package-lock:
-    if: startsWith(github.ref, 'refs/tags/v')
-    needs: [release]
-    runs-on: ubuntu-latest
-    permissions:
-      contents: write
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v4
-        with:
-          ref: main
-          token: ${{ secrets.LANCEDB_RELEASE_TOKEN }}
-          fetch-depth: 0
-          lfs: true
-      - uses: ./.github/workflows/update_package_lock
-        with:
-          github_token: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/run_tests/action.yml

@@ -24,8 +24,8 @@ runs:
     - name: pytest (with integration)
       shell: bash
       if: ${{ inputs.integration == 'true' }}
-      run: pytest -m "not slow" -x -v --durations=30 python/python/tests
+      run: pytest -m "not slow" -vv --durations=30 python/python/tests
     - name: pytest (no integration tests)
       shell: bash
       if: ${{ inputs.integration != 'true' }}
-      run: pytest -m "not slow and not s3_test" -x -v --durations=30 python/python/tests
+      run: pytest -m "not slow and not s3_test" -vv --durations=30 python/python/tests

.github/workflows/update_package_lock/action.yml

@@ -1,33 +0,0 @@
-name: update_package_lock
-description: "Update node's package.lock"
-
-inputs:
-  github_token:
-    required: true
-    description: "github token for the repo"
-
-runs:
-  using: "composite"
-  steps:
-    - uses: actions/setup-node@v3
-      with:
-        node-version: 20
-    - name: Set git configs
-      shell: bash
-      run: |
-        git config user.name 'Lance Release'
-        git config user.email 'lance-dev@lancedb.com'
-    - name: Update package-lock.json file
-      working-directory: ./node
-      run: |
-        npm install
-        git add package-lock.json
-        git commit -m "Updating package-lock.json"
-      shell: bash
-    - name: Push changes
-      if: ${{ inputs.dry_run }} == "false"
-      uses: ad-m/github-push-action@master
-      with:
-        github_token: ${{ inputs.github_token }}
-        branch: main
-        tags: true

.github/workflows/update_package_lock_nodejs/action.yml

@@ -1,33 +0,0 @@
-name: update_package_lock_nodejs
-description: "Update nodejs's package.lock"
-
-inputs:
-  github_token:
-    required: true
-    description: "github token for the repo"
-
-runs:
-  using: "composite"
-  steps:
-    - uses: actions/setup-node@v3
-      with:
-        node-version: 20
-    - name: Set git configs
-      shell: bash
-      run: |
-        git config user.name 'Lance Release'
-        git config user.email 'lance-dev@lancedb.com'
-    - name: Update package-lock.json file
-      working-directory: ./nodejs
-      run: |
-        npm install
-        git add package-lock.json
-        git commit -m "Updating package-lock.json"
-      shell: bash
-    - name: Push changes
-      if: ${{ inputs.dry_run }} == "false"
-      uses: ad-m/github-push-action@master
-      with:
-        github_token: ${{ inputs.github_token }}
-        branch: main
-        tags: true

Cargo.toml

@@ -21,49 +21,49 @@ categories = ["database-implementations"]
 rust-version = "1.78.0"
 
 [workspace.dependencies]
-lance = { "version" = "=0.27.2", "features" = ["dynamodb"] }
+lance = { "version" = "=0.31.1", tag="v0.31.1-beta.2", git="https://github.com/lancedb/lance.git", features = ["dynamodb"] }
-lance-io = { version = "=0.27.2" }
+lance-io = { "version" = "=0.31.1", tag="v0.31.1-beta.2", git="https://github.com/lancedb/lance.git" }
-lance-index = { version = "=0.27.2" }
+lance-index = { "version" = "=0.31.1", tag="v0.31.1-beta.2", git="https://github.com/lancedb/lance.git" }
-lance-linalg = { version = "=0.27.2" }
+lance-linalg = { "version" = "=0.31.1", tag="v0.31.1-beta.2", git="https://github.com/lancedb/lance.git" }
-lance-table = { version = "=0.27.2" }
+lance-table = { "version" = "=0.31.1", tag="v0.31.1-beta.2", git="https://github.com/lancedb/lance.git" }
-lance-testing = { version = "=0.27.2" }
+lance-testing = { "version" = "=0.31.1", tag="v0.31.1-beta.2", git="https://github.com/lancedb/lance.git" }
-lance-datafusion = { version = "=0.27.2" }
+lance-datafusion = { "version" = "=0.31.1", tag="v0.31.1-beta.2", git="https://github.com/lancedb/lance.git" }
-lance-encoding = { version = "=0.27.2" }
+lance-encoding = { "version" = "=0.31.1", tag="v0.31.1-beta.2", git="https://github.com/lancedb/lance.git" }
 # Note that this one does not include pyarrow
-arrow = { version = "54.1", optional = false }
+arrow = { version = "55.1", optional = false }
-arrow-array = "54.1"
+arrow-array = "55.1"
-arrow-data = "54.1"
+arrow-data = "55.1"
-arrow-ipc = "54.1"
+arrow-ipc = "55.1"
-arrow-ord = "54.1"
+arrow-ord = "55.1"
-arrow-schema = "54.1"
+arrow-schema = "55.1"
-arrow-arith = "54.1"
+arrow-arith = "55.1"
-arrow-cast = "54.1"
+arrow-cast = "55.1"
 async-trait = "0"
-datafusion = { version = "46.0", default-features = false }
+datafusion = { version = "48.0", default-features = false }
-datafusion-catalog = "46.0"
+datafusion-catalog = "48.0"
-datafusion-common = { version = "46.0", default-features = false }
+datafusion-common = { version = "48.0", default-features = false }
-datafusion-execution = "46.0"
+datafusion-execution = "48.0"
-datafusion-expr = "46.0"
+datafusion-expr = "48.0"
-datafusion-physical-plan = "46.0"
+datafusion-physical-plan = "48.0"
 env_logger = "0.11"
-half = { "version" = "=2.4.1", default-features = false, features = [
+half = { "version" = "2.6.0", default-features = false, features = [
     "num-traits",
 ] }
 futures = "0"
 log = "0.4"
 moka = { version = "0.12", features = ["future"] }
-object_store = "0.11.0"
+object_store = "0.12.0"
 pin-project = "1.0.7"
 snafu = "0.8"
 url = "2"
 num-traits = "0.2"
-rand = "0.8"
+rand = "0.9"
 regex = "1.10"
 lazy_static = "1"
 semver = "1.0.25"
 # Temporary pins to work around downstream issues
 # https://github.com/apache/arrow-rs/commit/2fddf85afcd20110ce783ed5b4cdeb82293da30b
-chrono = "=0.4.39"
+chrono = "=0.4.41"
 # https://github.com/RustCrypto/formats/issues/1684
 base64ct = "=1.6.0"
 # Workaround for: https://github.com/eira-fransham/crunchy/issues/13

README.md

@@ -1,94 +1,97 @@
 <a href="https://cloud.lancedb.com" target="_blank">
   <img src="https://github.com/user-attachments/assets/92dad0a2-2a37-4ce1-b783-0d1b4f30a00c" alt="LanceDB Cloud Public Beta" width="100%" style="max-width: 100%;">
 </a>
-
 <div align="center">
-<p align="center">
 
-<picture>
+[![LanceDB](docs/src/assets/hero-header.png)](https://lancedb.com)
-  <source media="(prefers-color-scheme: dark)" srcset="https://github.com/user-attachments/assets/ac270358-333e-4bea-a132-acefaa94040e">
+[![Website](https://img.shields.io/badge/-Website-100000?style=for-the-badge&labelColor=645cfb&color=645cfb)](https://lancedb.com/)
-  <source media="(prefers-color-scheme: light)" srcset="https://github.com/user-attachments/assets/b864d814-0d29-4784-8fd9-807297c758c0">
+[![Blog](https://img.shields.io/badge/Blog-100000?style=for-the-badge&labelColor=645cfb&color=645cfb)](https://blog.lancedb.com/)
-  <img alt="LanceDB Logo" src="https://github.com/user-attachments/assets/b864d814-0d29-4784-8fd9-807297c758c0" width=300>
+[![Discord](https://img.shields.io/badge/-Discord-100000?style=for-the-badge&logo=discord&logoColor=white&labelColor=645cfb&color=645cfb)](https://discord.gg/zMM32dvNtd)
-</picture>
+[![Twitter](https://img.shields.io/badge/-Twitter-100000?style=for-the-badge&logo=x&logoColor=white&labelColor=645cfb&color=645cfb)](https://twitter.com/lancedb)
+[![LinkedIn](https://img.shields.io/badge/-LinkedIn-100000?style=for-the-badge&logo=linkedin&logoColor=white&labelColor=645cfb&color=645cfb)](https://www.linkedin.com/company/lancedb/)
 
-**Search More, Manage Less**
 
-<a href='https://github.com/lancedb/vectordb-recipes/tree/main' target="_blank"><img alt='LanceDB' src='https://img.shields.io/badge/VectorDB_Recipes-100000?style=for-the-badge&logo=LanceDB&logoColor=white&labelColor=645cfb&color=645cfb'/></a>
+<img src="docs/src/assets/lancedb.png" alt="LanceDB" width="50%">
-<a href='https://lancedb.github.io/lancedb/' target="_blank"><img alt='lancdb' src='https://img.shields.io/badge/DOCS-100000?style=for-the-badge&logo=lancdb&logoColor=white&labelColor=645cfb&color=645cfb'/></a>
-[![Blog](https://img.shields.io/badge/Blog-12100E?style=for-the-badge&logoColor=white)](https://blog.lancedb.com/)
-[![Discord](https://img.shields.io/badge/Discord-%235865F2.svg?style=for-the-badge&logo=discord&logoColor=white)](https://discord.gg/zMM32dvNtd)
-[![Twitter](https://img.shields.io/badge/Twitter-%231DA1F2.svg?style=for-the-badge&logo=Twitter&logoColor=white)](https://twitter.com/lancedb)
-[![Gurubase](https://img.shields.io/badge/Gurubase-Ask%20LanceDB%20Guru-006BFF?style=for-the-badge)](https://gurubase.io/g/lancedb)
 
-</p>
+# **The Multimodal AI Lakehouse**
 
-<img max-width="750px" alt="LanceDB Multimodal Search" src="https://github.com/lancedb/lancedb/assets/917119/09c5afc5-7816-4687-bae4-f2ca194426ec">
+[**How to Install** ](#how-to-install) ✦ [**Detailed Documentation**](https://lancedb.github.io/lancedb/) ✦ [**Tutorials and Recipes**](https://github.com/lancedb/vectordb-recipes/tree/main) ✦  [**Contributors**](#contributors) 
+
+**The ultimate multimodal data platform for AI/ML applications.** 
+
+LanceDB is designed for fast, scalable, and production-ready vector search. It is built on top of the Lance columnar format. You can store, index, and search over petabytes of multimodal data and vectors with ease. 
+LanceDB is a central location where developers can build, train and analyze their AI workloads.
 
-</p>
 </div>
 
-<hr />
+<br>
 
-LanceDB is an open-source database for vector-search built with persistent storage, which greatly simplifies retrieval, filtering and management of embeddings.
+## **Demo: Multimodal Search by Keyword, Vector or with SQL**
+<img max-width="750px" alt="LanceDB Multimodal Search" src="https://github.com/lancedb/lancedb/assets/917119/09c5afc5-7816-4687-bae4-f2ca194426ec">
 
-The key features of LanceDB include:
+## **Star LanceDB to get updates!**
 
-* Production-scale vector search with no servers to manage.
+<details>
+<summary>⭐ Click here ⭐  to see how fast we're growing!</summary>
+<picture>
+  <source media="(prefers-color-scheme: dark)" srcset="https://api.star-history.com/svg?repos=lancedb/lancedb&theme=dark&type=Date">
+  <img width="100%" src="https://api.star-history.com/svg?repos=lancedb/lancedb&theme=dark&type=Date">
+</picture>
+</details>
 
-* Store, query and filter vectors, metadata and multi-modal data (text, images, videos, point clouds, and more).
+## **Key Features**:
 
-* Support for vector similarity search, full-text search and SQL.
+- **Fast Vector Search**: Search billions of vectors in milliseconds with state-of-the-art indexing.
+- **Comprehensive Search**: Support for vector similarity search, full-text search and SQL.
+- **Multimodal Support**: Store, query and filter vectors, metadata and multimodal data (text, images, videos, point clouds, and more).
+- **Advanced Features**: Zero-copy, automatic versioning, manage versions of your data without needing extra infrastructure. GPU support in building vector index.
 
-* Native Python and Javascript/Typescript support.
+### **Products**:
+- **Open Source & Local**: 100% open source, runs locally or in your cloud. No vendor lock-in.
+- **Cloud and Enterprise**: Production-scale vector search with no servers to manage. Complete data sovereignty and security.
 
-* Zero-copy, automatic versioning, manage versions of your data without needing extra infrastructure.
+### **Ecosystem**:
+- **Columnar Storage**: Built on the Lance columnar format for efficient storage and analytics.
+- **Seamless Integration**: Python, Node.js, Rust, and REST APIs for easy integration. Native Python and Javascript/Typescript support.
+- **Rich Ecosystem**: Integrations with [**LangChain** 🦜️🔗](https://python.langchain.com/docs/integrations/vectorstores/lancedb/), [**LlamaIndex** 🦙](https://gpt-index.readthedocs.io/en/latest/examples/vector_stores/LanceDBIndexDemo.html), Apache-Arrow, Pandas, Polars, DuckDB and more on the way.
 
-* GPU support in building vector index(*).
+## **How to Install**:
 
-* Ecosystem integrations with [LangChain 🦜️🔗](https://python.langchain.com/docs/integrations/vectorstores/lancedb/), [LlamaIndex 🦙](https://gpt-index.readthedocs.io/en/latest/examples/vector_stores/LanceDBIndexDemo.html), Apache-Arrow, Pandas, Polars, DuckDB and more on the way.
+Follow the [Quickstart](https://lancedb.github.io/lancedb/basic/) doc to set up LanceDB locally. 
 
-LanceDB's core is written in Rust 🦀 and is built using <a href="https://github.com/lancedb/lance">Lance</a>, an open-source columnar format designed for performant ML workloads.
+**API & SDK:** We also support Python, Typescript and Rust SDKs
 
-## Quick Start
+| Interface | Documentation |
+|-----------|---------------|
+| Python SDK | https://lancedb.github.io/lancedb/python/python/ |
+| Typescript SDK | https://lancedb.github.io/lancedb/js/globals/ |
+| Rust SDK | https://docs.rs/lancedb/latest/lancedb/index.html |
+| REST API | https://docs.lancedb.com/api-reference/introduction |
 
-**Javascript**
+## **Join Us and Contribute**
-```shell
-npm install @lancedb/lancedb
-```
 
-```javascript
+We welcome contributions from everyone! Whether you're a developer, researcher, or just someone who wants to help out. 
-import * as lancedb from "@lancedb/lancedb";
 
-const db = await lancedb.connect("data/sample-lancedb");
+If you have any suggestions or feature requests, please feel free to open an issue on GitHub or discuss it on our [**Discord**](https://discord.gg/G5DcmnZWKB) server.
-const table = await db.createTable("vectors", [
+
-	{ id: 1, vector: [0.1, 0.2], item: "foo", price: 10 },
+[**Check out the GitHub Issues**](https://github.com/lancedb/lancedb/issues) if you would like to work on the features that are planned for the future. If you have any suggestions or feature requests, please feel free to open an issue on GitHub. 
-	{ id: 2, vector: [1.1, 1.2], item: "bar", price: 50 },
+
-], {mode: 'overwrite'});
+## **Contributors**
+
+<a href="https://github.com/lancedb/lancedb/graphs/contributors">
+  <img src="https://contrib.rocks/image?repo=lancedb/lancedb" />
+</a>
 
 
-const query = table.vectorSearch([0.1, 0.3]).limit(2);
+## **Stay in Touch With Us**
-const results = await query.toArray();
+<div align="center">
 
-// You can also search for rows by specific criteria without involving a vector search.
+</br>
-const rowsByCriteria = await table.query().where("price >= 10").toArray();
-```
 
-**Python**
+[![Website](https://img.shields.io/badge/-Website-100000?style=for-the-badge&labelColor=645cfb&color=645cfb)](https://lancedb.com/)
-```shell
+[![Blog](https://img.shields.io/badge/Blog-100000?style=for-the-badge&labelColor=645cfb&color=645cfb)](https://blog.lancedb.com/)
-pip install lancedb
+[![Discord](https://img.shields.io/badge/-Discord-100000?style=for-the-badge&logo=discord&logoColor=white&labelColor=645cfb&color=645cfb)](https://discord.gg/zMM32dvNtd)
-```
+[![Twitter](https://img.shields.io/badge/-Twitter-100000?style=for-the-badge&logo=x&logoColor=white&labelColor=645cfb&color=645cfb)](https://twitter.com/lancedb)
+[![LinkedIn](https://img.shields.io/badge/-LinkedIn-100000?style=for-the-badge&logo=linkedin&logoColor=white&labelColor=645cfb&color=645cfb)](https://www.linkedin.com/company/lancedb/)
 
-```python
+</div>
-import lancedb
-
-uri = "data/sample-lancedb"
-db = lancedb.connect(uri)
-table = db.create_table("my_table",
-                         data=[{"vector": [3.1, 4.1], "item": "foo", "price": 10.0},
-                               {"vector": [5.9, 26.5], "item": "bar", "price": 20.0}])
-result = table.search([100, 100]).limit(2).to_pandas()
-```
-
-## Blogs, Tutorials & Videos
-* 📈 <a href="https://blog.lancedb.com/benchmarking-random-access-in-lance/">2000x better performance with Lance over Parquet</a>
-* 🤖 <a href="https://github.com/lancedb/vectordb-recipes/tree/main/examples/Youtube-Search-QA-Bot">Build a question and answer bot with LanceDB</a>

ci/set_lance_version.py

@@ -0,0 +1,174 @@
+import argparse
+import sys
+import json
+
+
+def run_command(command: str) -> str:
+    """
+    Run a shell command and return stdout as a string.
+    If exit code is not 0, raise an exception with the stderr output.
+    """
+    import subprocess
+
+    result = subprocess.run(command, shell=True, capture_output=True, text=True)
+    if result.returncode != 0:
+        raise Exception(f"Command failed with error: {result.stderr.strip()}")
+    return result.stdout.strip()
+
+
+def get_latest_stable_version() -> str:
+    version_line = run_command("cargo info lance | grep '^version:'")
+    version = version_line.split(" ")[1].strip()
+    return version
+
+
+def get_latest_preview_version() -> str:
+    lance_tags = run_command(
+        "git ls-remote --tags https://github.com/lancedb/lance.git | grep 'refs/tags/v[0-9beta.-]\\+$'"
+    ).splitlines()
+    lance_tags = (
+        tag.split("refs/tags/")[1]
+        for tag in lance_tags
+        if "refs/tags/" in tag and "beta" in tag
+    )
+    from packaging.version import Version
+
+    latest = max(
+        (tag[1:] for tag in lance_tags if tag.startswith("v")), key=lambda t: Version(t)
+    )
+    return str(latest)
+
+
+def extract_features(line: str) -> list:
+    """
+    Extracts the features from a line in Cargo.toml.
+    Example: 'lance = { "version" = "=0.29.0", "features" = ["dynamodb"] }'
+    Returns: ['dynamodb']
+    """
+    import re
+
+    match = re.search(r'"features"\s*=\s*\[(.*?)\]', line)
+    if match:
+        features_str = match.group(1)
+        return [f.strip('"') for f in features_str.split(",")]
+    return []
+
+
+def update_cargo_toml(line_updater):
+    """
+    Updates the Cargo.toml file by applying the line_updater function to each line.
+    The line_updater function should take a line as input and return the updated line.
+    """
+    with open("Cargo.toml", "r") as f:
+        lines = f.readlines()
+
+    new_lines = []
+    for line in lines:
+        if line.startswith("lance"):
+            # Update the line using the provided function
+            new_lines.append(line_updater(line))
+        else:
+            # Keep the line unchanged
+            new_lines.append(line)
+
+    with open("Cargo.toml", "w") as f:
+        f.writelines(new_lines)
+
+
+def set_stable_version(version: str):
+    """
+    Sets lines to
+    lance = { "version" = "=0.29.0", "features" = ["dynamodb"] }
+    lance-io = "=0.29.0"
+    ...
+    """
+
+    def line_updater(line: str) -> str:
+        package_name = line.split("=", maxsplit=1)[0].strip()
+        features = extract_features(line)
+        if features:
+            return f'{package_name} = {{ "version" = "={version}", "features" = {json.dumps(features)} }}\n'
+        else:
+            return f'{package_name} = "={version}"\n'
+
+    update_cargo_toml(line_updater)
+
+
+def set_preview_version(version: str):
+    """
+    Sets lines to
+    lance = { "version" = "=0.29.0", "features" = ["dynamodb"], tag = "v0.29.0-beta.2", git="https://github.com/lancedb/lance.git" }
+    lance-io = { version = "=0.29.0", tag = "v0.29.0-beta.2", git="https://github.com/lancedb/lance.git" }
+    ...
+    """
+
+    def line_updater(line: str) -> str:
+        package_name = line.split("=", maxsplit=1)[0].strip()
+        features = extract_features(line)
+        base_version = version.split("-")[0]  # Get the base version without beta suffix
+        if features:
+            return f'{package_name} = {{ "version" = "={base_version}", "features" = {json.dumps(features)}, "tag" = "v{version}", "git" = "https://github.com/lancedb/lance.git" }}\n'
+        else:
+            return f'{package_name} = {{ "version" = "={base_version}", "tag" = "v{version}", "git" = "https://github.com/lancedb/lance.git" }}\n'
+
+    update_cargo_toml(line_updater)
+
+
+def set_local_version():
+    """
+    Sets lines to
+    lance = { path = "../lance/rust/lance", features = ["dynamodb"] }
+    lance-io = { path = "../lance/rust/lance-io" }
+    ...
+    """
+
+    def line_updater(line: str) -> str:
+        package_name = line.split("=", maxsplit=1)[0].strip()
+        features = extract_features(line)
+        if features:
+            return f'{package_name} = {{ "path" = "../lance/rust/{package_name}", "features" = {json.dumps(features)} }}\n'
+        else:
+            return f'{package_name} = {{ "path" = "../lance/rust/{package_name}" }}\n'
+
+    update_cargo_toml(line_updater)
+
+
+parser = argparse.ArgumentParser(description="Set the version of the Lance package.")
+parser.add_argument(
+    "version",
+    type=str,
+    help="The version to set for the Lance package. Use 'stable' for the latest stable version, 'preview' for latest preview version, or a specific version number (e.g., '0.1.0'). You can also specify 'local' to use a local path.",
+)
+args = parser.parse_args()
+
+if args.version == "stable":
+    latest_stable_version = get_latest_stable_version()
+    print(
+        f"Found latest stable version: \033[1mv{latest_stable_version}\033[0m",
+        file=sys.stderr,
+    )
+    set_stable_version(latest_stable_version)
+elif args.version == "preview":
+    latest_preview_version = get_latest_preview_version()
+    print(
+        f"Found latest preview version: \033[1mv{latest_preview_version}\033[0m",
+        file=sys.stderr,
+    )
+    set_preview_version(latest_preview_version)
+elif args.version == "local":
+    set_local_version()
+else:
+    # Parse the version number.
+    version = args.version
+    # Ignore initial v if present.
+    if version.startswith("v"):
+        version = version[1:]
+
+    if "beta" in version:
+        set_preview_version(version)
+    else:
+        set_stable_version(version)
+
+print("Updating lockfiles...", file=sys.stderr, end="")
+run_command("cargo metadata > /dev/null")
+print(" done.", file=sys.stderr)

ci/update_lockfiles.sh

@@ -0,0 +1,30 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+AMEND=false
+
+for arg in "$@"; do
+  if [[ "$arg" == "--amend" ]]; then
+    AMEND=true
+  fi
+done
+
+# This updates the lockfile without building
+cargo metadata --quiet > /dev/null
+
+pushd nodejs || exit 1
+npm install --package-lock-only --silent
+popd
+pushd node || exit 1
+npm install --package-lock-only --silent
+popd
+
+if git diff --quiet --exit-code; then
+  echo "No lockfile changes to commit; skipping amend."
+elif $AMEND; then
+  git add Cargo.lock nodejs/package-lock.json node/package-lock.json
+  git commit --amend --no-edit
+else
+  git add Cargo.lock nodejs/package-lock.json node/package-lock.json
+  git commit -m "Update lockfiles"
+fi

docs/mkdocs.yml

@@ -193,6 +193,7 @@ nav:
           - Pandas and PyArrow: python/pandas_and_pyarrow.md
           - Polars: python/polars_arrow.md
           - DuckDB: python/duckdb.md
+          - Datafusion: python/datafusion.md
           - LangChain:
               - LangChain 🔗: integrations/langchain.md
               - LangChain demo: notebooks/langchain_demo.ipynb
@@ -248,6 +249,7 @@ nav:
       - Data management: concepts/data_management.md
   - Guides:
       - Working with tables: guides/tables.md
+      - Working with SQL: guides/sql_querying.md
       - Building an ANN index: ann_indexes.md
       - Vector Search: search.md
       - Full-text search (native): fts.md
@@ -324,6 +326,7 @@ nav:
       - Pandas and PyArrow: python/pandas_and_pyarrow.md
       - Polars: python/polars_arrow.md
       - DuckDB: python/duckdb.md
+      - Datafusion: python/datafusion.md
       - LangChain 🦜️🔗↗: integrations/langchain.md
       - LangChain.js 🦜️🔗↗: https://js.langchain.com/docs/integrations/vectorstores/lancedb
       - LlamaIndex 🦙↗: integrations/llamaIndex.md

docs/overrides/partials/main.html

@@ -0,0 +1,5 @@
+{% extends "base.html" %}
+
+{% block announce %}
+  📚 Starting June 1st, 2025, please use <a href="https://lancedb.github.io/documentation" target="_blank" rel="noopener noreferrer">lancedb.github.io/documentation</a> for the latest docs.
+{% endblock %}

docs/src/ann_indexes.md

@@ -291,7 +291,7 @@ Product quantization can lead to approximately `16 * sizeof(float32) / 1 = 64` t
 
 `num_partitions` is used to decide how many partitions the first level `IVF` index uses.
 Higher number of partitions could lead to more efficient I/O during queries and better accuracy, but it takes much more time to train.
-On `SIFT-1M` dataset, our benchmark shows that keeping each partition 1K-4K rows lead to a good latency / recall.
+On `SIFT-1M` dataset, our benchmark shows that keeping each partition 4K-8K rows lead to a good latency / recall.
 
 `num_sub_vectors` specifies how many Product Quantization (PQ) short codes to generate on each vector. The number should be a factor of the vector dimension. Because
 PQ is a lossy compression of the original vector, a higher `num_sub_vectors` usually results in

docs/src/guides/sql_querying.md

@@ -0,0 +1,68 @@
+You can use DuckDB and Apache Datafusion to query your LanceDB tables using SQL.
+This guide will show how to query Lance tables them using both.
+
+We will re-use the dataset [created previously](./pandas_and_pyarrow.md):
+
+```python
+import lancedb
+
+db = lancedb.connect("data/sample-lancedb")
+data = [
+    {"vector": [3.1, 4.1], "item": "foo", "price": 10.0},
+    {"vector": [5.9, 26.5], "item": "bar", "price": 20.0}
+]
+table = db.create_table("pd_table", data=data)
+```
+
+## Querying a LanceDB Table with DuckDb
+
+The `to_lance` method converts the LanceDB table to a `LanceDataset`, which is accessible to DuckDB through the Arrow compatibility layer.
+To query the resulting Lance dataset in DuckDB, all you need to do is reference the dataset by the same name in your SQL query.
+
+```python
+import duckdb
+
+arrow_table = table.to_lance()
+
+duckdb.query("SELECT * FROM arrow_table")
+```
+
+```
+┌─────────────┬─────────┬────────┐
+│   vector    │  item   │ price  │
+│   float[]   │ varchar │ double │
+├─────────────┼─────────┼────────┤
+│ [3.1, 4.1]  │ foo     │   10.0 │
+│ [5.9, 26.5] │ bar     │   20.0 │
+└─────────────┴─────────┴────────┘
+```
+
+## Querying a LanceDB Table with Apache Datafusion
+
+Have the required imports before doing any querying.
+
+=== "Python"
+
+    ```python
+    --8<-- "python/python/tests/docs/test_guide_tables.py:import-lancedb"
+    --8<-- "python/python/tests/docs/test_guide_tables.py:import-session-context"
+    --8<-- "python/python/tests/docs/test_guide_tables.py:import-ffi-dataset"
+    ```
+
+Register the table created with the Datafusion session context.
+
+=== "Python"
+
+    ```python
+    --8<-- "python/python/tests/docs/test_guide_tables.py:lance_sql_basic"
+    ```
+
+```
+┌─────────────┬─────────┬────────┐
+│   vector    │  item   │ price  │
+│   float[]   │ varchar │ double │
+├─────────────┼─────────┼────────┤
+│ [3.1, 4.1]  │ foo     │   10.0 │
+│ [5.9, 26.5] │ bar     │   20.0 │
+└─────────────┴─────────┴────────┘
+```

docs/src/guides/tables.md

@@ -765,7 +765,7 @@ This can be used to update zero to all rows depending on how many rows match the
         ];
         const tbl = await db.createTable("my_table", data)
 
-        await tbl.update({ 
+        await tbl.update({
             values: { vector: [10, 10] },
             where: "x = 2"
         });
@@ -787,9 +787,9 @@ This can be used to update zero to all rows depending on how many rows match the
         ];
         const tbl = await db.createTable("my_table", data)
 
-        await tbl.update({ 
+        await tbl.update({
-            where: "x = 2", 
+            where: "x = 2",
-            values: { vector: [10, 10] } 
+            values: { vector: [10, 10] }
         });
         ```
 

docs/src/js/classes/BooleanQuery.md

@@ -0,0 +1,53 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / BooleanQuery
+
+# Class: BooleanQuery
+
+Represents a full-text query interface.
+This interface defines the structure and behavior for full-text queries,
+including methods to retrieve the query type and convert the query to a dictionary format.
+
+## Implements
+
+- [`FullTextQuery`](../interfaces/FullTextQuery.md)
+
+## Constructors
+
+### new BooleanQuery()
+
+```ts
+new BooleanQuery(queries): BooleanQuery
+```
+
+Creates an instance of BooleanQuery.
+
+#### Parameters
+
+* **queries**: [[`Occur`](../enumerations/Occur.md), [`FullTextQuery`](../interfaces/FullTextQuery.md)][]
+    An array of (Occur, FullTextQuery objects) to combine.
+    Occur specifies whether the query must match, or should match.
+
+#### Returns
+
+[`BooleanQuery`](BooleanQuery.md)
+
+## Methods
+
+### queryType()
+
+```ts
+queryType(): FullTextQueryType
+```
+
+The type of the full-text query.
+
+#### Returns
+
+[`FullTextQueryType`](../enumerations/FullTextQueryType.md)
+
+#### Implementation of
+
+[`FullTextQuery`](../interfaces/FullTextQuery.md).[`queryType`](../interfaces/FullTextQuery.md#querytype)

docs/src/js/classes/MatchQuery.md

@@ -40,6 +40,7 @@ Creates an instance of MatchQuery.
     - `boost`: The boost factor for the query (default is 1.0).
     - `fuzziness`: The fuzziness level for the query (default is 0).
     - `maxExpansions`: The maximum number of terms to consider for fuzzy matching (default is 50).
+    - `operator`: The logical operator to use for combining terms in the query (default is "OR").
 
 * **options.boost?**: `number`
 
@@ -47,6 +48,8 @@ Creates an instance of MatchQuery.
 
 * **options.maxExpansions?**: `number`
 
+* **options.operator?**: [`Operator`](../enumerations/Operator.md)
+
 #### Returns
 
 [`MatchQuery`](MatchQuery.md)

docs/src/js/classes/MultiMatchQuery.md

@@ -38,9 +38,12 @@ Creates an instance of MultiMatchQuery.
 * **options?**
     Optional parameters for the multi-match query.
     - `boosts`: An array of boost factors for each column (default is 1.0 for all).
+    - `operator`: The logical operator to use for combining terms in the query (default is "OR").
 
 * **options.boosts?**: `number`[]
 
+* **options.operator?**: [`Operator`](../enumerations/Operator.md)
+
 #### Returns
 
 [`MultiMatchQuery`](MultiMatchQuery.md)

docs/src/js/classes/PhraseQuery.md

@@ -19,7 +19,10 @@ including methods to retrieve the query type and convert the query to a dictiona
 ### new PhraseQuery()
 
 ```ts
-new PhraseQuery(query, column): PhraseQuery
+new PhraseQuery(
+   query,
+   column,
+   options?): PhraseQuery
 ```
 
 Creates an instance of `PhraseQuery`.
@@ -32,6 +35,12 @@ Creates an instance of `PhraseQuery`.
 * **column**: `string`
     The name of the column to search within.
 
+* **options?**
+    Optional parameters for the phrase query.
+    - `slop`: The maximum number of intervening unmatched positions allowed between words in the phrase (default is 0).
+
+* **options.slop?**: `number`
+
 #### Returns
 
 [`PhraseQuery`](PhraseQuery.md)

docs/src/js/enumerations/FullTextQueryType.md

@@ -15,6 +15,14 @@ Enum representing the types of full-text queries supported.
 
 ## Enumeration Members
 
+### Boolean
+
+```ts
+Boolean: "boolean";
+```
+
+***
+
 ### Boost
 
 ```ts

docs/src/js/enumerations/Occur.md

@@ -0,0 +1,28 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / Occur
+
+# Enumeration: Occur
+
+Enum representing the occurrence of terms in full-text queries.
+
+- `Must`: The term must be present in the document.
+- `Should`: The term should contribute to the document score, but is not required.
+
+## Enumeration Members
+
+### Must
+
+```ts
+Must: "MUST";
+```
+
+***
+
+### Should
+
+```ts
+Should: "SHOULD";
+```

docs/src/js/enumerations/Operator.md

@@ -0,0 +1,28 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / Operator
+
+# Enumeration: Operator
+
+Enum representing the logical operators used in full-text queries.
+
+- `And`: All terms must match.
+- `Or`: At least one term must match.
+
+## Enumeration Members
+
+### And
+
+```ts
+And: "AND";
+```
+
+***
+
+### Or
+
+```ts
+Or: "OR";
+```

docs/src/js/globals.md

@@ -12,9 +12,12 @@
 ## Enumerations
 
 - [FullTextQueryType](enumerations/FullTextQueryType.md)
+- [Occur](enumerations/Occur.md)
+- [Operator](enumerations/Operator.md)
 
 ## Classes
 
+- [BooleanQuery](classes/BooleanQuery.md)
 - [BoostQuery](classes/BoostQuery.md)
 - [Connection](classes/Connection.md)
 - [Index](classes/Index.md)

docs/src/notebooks/Multivector_on_LanceDB.ipynb

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

docs/src/python/datafusion.md

@@ -0,0 +1,53 @@
+# Apache Datafusion
+
+In Python, LanceDB tables can also be queried with [Apache Datafusion](https://datafusion.apache.org/), an extensible query engine written in Rust that uses Apache Arrow as its in-memory format. This means you can write complex SQL queries to analyze your data in LanceDB.
+
+This integration is done via [Datafusion FFI](https://docs.rs/datafusion-ffi/latest/datafusion_ffi/), which provides a native integration between LanceDB and Datafusion.
+The Datafusion FFI allows to pass down column selections and basic filters to LanceDB, reducing the amount of scanned data when executing your query. Additionally, the integration allows streaming data from LanceDB tables which allows to do aggregation larger-than-memory.
+
+We can demonstrate this by first installing `datafusion` and `lancedb`.
+
+```shell
+pip install datafusion lancedb
+```
+
+We will re-use the dataset [created previously](./pandas_and_pyarrow.md):
+
+```python
+import lancedb
+
+from datafusion import SessionContext
+from lance import FFILanceTableProvider
+
+db = lancedb.connect("data/sample-lancedb")
+data = [
+    {"vector": [3.1, 4.1], "item": "foo", "price": 10.0},
+    {"vector": [5.9, 26.5], "item": "bar", "price": 20.0}
+]
+lance_table = db.create_table("lance_table", data)
+
+ctx = SessionContext()
+
+ffi_lance_table = FFILanceTableProvider(
+    lance_table.to_lance(), with_row_id=True, with_row_addr=True
+)
+ctx.register_table_provider("ffi_lance_table", ffi_lance_table)
+```
+
+The `to_lance` method converts the LanceDB table to a `LanceDataset`, which is accessible to Datafusion through the Datafusion FFI integration layer.
+To query the resulting Lance dataset in Datafusion, you first need to register the dataset with Datafusion and then just reference it by the same name in your SQL query.
+
+```python
+ctx.table("ffi_lance_table")
+ctx.sql("SELECT * FROM ffi_lance_table")
+```
+
+```
+┌─────────────┬─────────┬────────┬─────────────────┬─────────────────┐
+│   vector    │  item   │ price  │ _rowid          │ _rowaddr        │
+│   float[]   │ varchar │ double │ bigint unsigned │ bigint unsigned │
+├─────────────┼─────────┼────────┼─────────────────┼─────────────────┤
+│ [3.1, 4.1]  │ foo     │   10.0 │               0 │               0 │
+│ [5.9, 26.5] │ bar     │   20.0 │               1 │               1 │
+└─────────────┴─────────┴────────┴─────────────────┴─────────────────┘
+```

docs/test/requirements.txt

@@ -7,3 +7,4 @@ tantivy==0.20.1
 --extra-index-url https://download.pytorch.org/whl/cpu
 torch
 polars>=0.19, <=1.3.0
+datafusion

java/core/pom.xml

@@ -8,7 +8,7 @@
     <parent>
         <groupId>com.lancedb</groupId>
         <artifactId>lancedb-parent</artifactId>
-        <version>0.19.1-beta.5</version>
+        <version>0.21.1-beta.0</version>
         <relativePath>../pom.xml</relativePath>
     </parent>
 

java/pom.xml

@@ -6,7 +6,7 @@
 
     <groupId>com.lancedb</groupId>
     <artifactId>lancedb-parent</artifactId>
-    <version>0.19.1-beta.5</version>
+    <version>0.21.1-beta.0</version>
     <packaging>pom</packaging>
 
     <name>LanceDB Parent</name>

node/package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "vectordb",
-  "version": "0.19.1-beta.5",
+  "version": "0.21.1-beta.0",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "vectordb",
-      "version": "0.19.1-beta.5",
+      "version": "0.21.1-beta.0",
       "cpu": [
         "x64",
         "arm64"
@@ -52,11 +52,11 @@
         "uuid": "^9.0.0"
       },
       "optionalDependencies": {
-        "@lancedb/vectordb-darwin-arm64": "0.19.1-beta.5",
+        "@lancedb/vectordb-darwin-arm64": "0.21.1-beta.0",
-        "@lancedb/vectordb-darwin-x64": "0.19.1-beta.5",
+        "@lancedb/vectordb-darwin-x64": "0.21.1-beta.0",
-        "@lancedb/vectordb-linux-arm64-gnu": "0.19.1-beta.5",
+        "@lancedb/vectordb-linux-arm64-gnu": "0.21.1-beta.0",
-        "@lancedb/vectordb-linux-x64-gnu": "0.19.1-beta.5",
+        "@lancedb/vectordb-linux-x64-gnu": "0.21.1-beta.0",
-        "@lancedb/vectordb-win32-x64-msvc": "0.19.1-beta.5"
+        "@lancedb/vectordb-win32-x64-msvc": "0.21.1-beta.0"
       },
       "peerDependencies": {
         "@apache-arrow/ts": "^14.0.2",
@@ -327,9 +327,9 @@
       }
     },
     "node_modules/@lancedb/vectordb-darwin-arm64": {
-      "version": "0.19.1-beta.5",
+      "version": "0.21.1-beta.0",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-arm64/-/vectordb-darwin-arm64-0.19.1-beta.5.tgz",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-arm64/-/vectordb-darwin-arm64-0.21.1-beta.0.tgz",
-      "integrity": "sha512-9WcTw67We5HYGayDt5jFquGoyAVzFSt/I65ag8+q7H9q4ZYKxeDhgNyQZJ8BmXEvbJtnYtYBSAtTEdFKYMce6w==",
+      "integrity": "sha512-easypFtN4rFFsSNumFLK/VEhD2BVp+jl6ysICGyutjD/UEiulVdhixBkK5miJOfu/1p67Rjit5C8u3acpX+k2g==",
       "cpu": [
         "arm64"
       ],
@@ -340,9 +340,9 @@
       ]
     },
     "node_modules/@lancedb/vectordb-darwin-x64": {
-      "version": "0.19.1-beta.5",
+      "version": "0.21.1-beta.0",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-x64/-/vectordb-darwin-x64-0.19.1-beta.5.tgz",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-x64/-/vectordb-darwin-x64-0.21.1-beta.0.tgz",
-      "integrity": "sha512-6Pe3PxEMi0VKGsu5R7IhOxTijUM3b5olRAqhxfcu5ti34gXIPNtu7g+T9lS78LKe+0D0v2BjZEY/JQakIFBNRw==",
+      "integrity": "sha512-ez//lKtXu7EWgZlUYgwBM2We4/ty8rOtkDMF3RlveWJAKn+zNX0UM3vTa9W7WbCcBn9Ycs3eQGrBvb0iYFIDgw==",
       "cpu": [
         "x64"
       ],
@@ -353,9 +353,9 @@
       ]
     },
     "node_modules/@lancedb/vectordb-linux-arm64-gnu": {
-      "version": "0.19.1-beta.5",
+      "version": "0.21.1-beta.0",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-arm64-gnu/-/vectordb-linux-arm64-gnu-0.19.1-beta.5.tgz",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-arm64-gnu/-/vectordb-linux-arm64-gnu-0.21.1-beta.0.tgz",
-      "integrity": "sha512-VJbBd+Y+6L2SREaOO1OzuUfTPHXyHE4AcsZuM6VMyoeX8k7lPnaA+vNk96o0w4V2KFEAI6o4QPgrRAXmMAzmbg==",
+      "integrity": "sha512-T+vfr3A/59V8JMB5vonUmFDE8Vcf7Qe+DhQMf6kUlQxx80TujMeTdkaOf9/zBAopN2T8Y2h+GNScjl/WomYOFg==",
       "cpu": [
         "arm64"
       ],
@@ -366,9 +366,9 @@
       ]
     },
     "node_modules/@lancedb/vectordb-linux-x64-gnu": {
-      "version": "0.19.1-beta.5",
+      "version": "0.21.1-beta.0",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-x64-gnu/-/vectordb-linux-x64-gnu-0.19.1-beta.5.tgz",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-x64-gnu/-/vectordb-linux-x64-gnu-0.21.1-beta.0.tgz",
-      "integrity": "sha512-3wS8Zn5NmHoszXfrY4JzMimHoh5LAmVi3pTX4gD+C9kVGoUJcDBP7/CrAbjnAz7VzzAIPmz8kvBuPz8l9X4hjw==",
+      "integrity": "sha512-FpDd4g2+xGrU41gywx4KFPGOlpBZq3VrE+4BBiTrRW6IO5Kbs2Mmq7ufJuDLlLqPs6ZQ5/Wlbcq5PmdRSoeq8A==",
       "cpu": [
         "x64"
       ],
@@ -379,9 +379,9 @@
       ]
     },
     "node_modules/@lancedb/vectordb-win32-x64-msvc": {
-      "version": "0.19.1-beta.5",
+      "version": "0.21.1-beta.0",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-win32-x64-msvc/-/vectordb-win32-x64-msvc-0.19.1-beta.5.tgz",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-win32-x64-msvc/-/vectordb-win32-x64-msvc-0.21.1-beta.0.tgz",
-      "integrity": "sha512-TemM9cvrPa2jFCjvYmKnrL0DTHegi/+LOQ3No9nPDHie2ka2fM9O2q60fAbYsYz+Mo9aV7MvL49ATbNCyl9MLA==",
+      "integrity": "sha512-SEKHecFpgODmrUsAE8pBLu8OMKnAx97Ap0FrH6AGGglJKAVirrrg9BKSPfmHMZCvyPSHzG5TUMxhtNm+Ibg5DQ==",
       "cpu": [
         "x64"
       ],

node/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vectordb",
-  "version": "0.19.1-beta.5",
+  "version": "0.21.1-beta.0",
   "description": " Serverless, low-latency vector database for AI applications",
   "private": false,
   "main": "dist/index.js",
@@ -89,10 +89,10 @@
     }
   },
   "optionalDependencies": {
-    "@lancedb/vectordb-darwin-x64": "0.19.1-beta.5",
+    "@lancedb/vectordb-darwin-x64": "0.21.1-beta.0",
-    "@lancedb/vectordb-darwin-arm64": "0.19.1-beta.5",
+    "@lancedb/vectordb-darwin-arm64": "0.21.1-beta.0",
-    "@lancedb/vectordb-linux-x64-gnu": "0.19.1-beta.5",
+    "@lancedb/vectordb-linux-x64-gnu": "0.21.1-beta.0",
-    "@lancedb/vectordb-linux-arm64-gnu": "0.19.1-beta.5",
+    "@lancedb/vectordb-linux-arm64-gnu": "0.21.1-beta.0",
-    "@lancedb/vectordb-win32-x64-msvc": "0.19.1-beta.5"
+    "@lancedb/vectordb-win32-x64-msvc": "0.21.1-beta.0"
   }
 }

nodejs/Cargo.toml

@@ -1,7 +1,7 @@
 [package]
 name = "lancedb-nodejs"
 edition.workspace = true
-version = "0.19.1-beta.5"
+version = "0.21.1-beta.0"
 license.workspace = true
 description.workspace = true
 repository.workspace = true
@@ -30,6 +30,7 @@ log.workspace = true
 
 # Workaround for build failure until we can fix it.
 aws-lc-sys = "=0.28.0"
+aws-lc-rs = "=1.13.0"
 
 [build-dependencies]
 napi-build = "2.1"

nodejs/__test__/arrow.test.ts

@@ -592,14 +592,14 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
         ).rejects.toThrow("column vector was missing");
       });
 
-      it("will provide a nice error if run twice", async function () {
+      it("will skip embedding application if already applied", async function () {
         const records = sampleRecords();
         const table = await convertToTable(records, dummyEmbeddingConfig);
 
         // fromTableToBuffer will try and apply the embeddings again
-        await expect(
+        // but should skip since the column already has non-null values
-          fromTableToBuffer(table, dummyEmbeddingConfig),
+        const result = await fromTableToBuffer(table, dummyEmbeddingConfig);
-        ).rejects.toThrow("already existed");
+        expect(result.byteLength).toBeGreaterThan(0);
       });
     });
 

nodejs/__test__/table.test.ts

@@ -33,7 +33,12 @@ import {
   register,
 } from "../lancedb/embedding";
 import { Index } from "../lancedb/indices";
-import { instanceOfFullTextQuery } from "../lancedb/query";
+import {
+  BooleanQuery,
+  Occur,
+  Operator,
+  instanceOfFullTextQuery,
+} from "../lancedb/query";
 import exp = require("constants");
 
 describe.each([arrow15, arrow16, arrow17, arrow18])(
@@ -363,9 +368,9 @@ describe("merge insert", () => {
       { a: 4, b: "z" },
     ];
 
-    expect(
+    const result = (await table.toArrow()).toArray().sort((a, b) => a.a - b.a);
-      JSON.parse(JSON.stringify((await table.toArrow()).toArray())),
+
-    ).toEqual(expected);
+    expect(result.map((row) => ({ ...row }))).toEqual(expected);
   });
   test("conditional update", async () => {
     const newData = [
@@ -554,6 +559,32 @@ describe("When creating an index", () => {
     rst = await tbl.query().limit(2).offset(1).nearestTo(queryVec).toArrow();
     expect(rst.numRows).toBe(1);
 
+    // test nprobes
+    rst = await tbl.query().nearestTo(queryVec).limit(2).nprobes(50).toArrow();
+    expect(rst.numRows).toBe(2);
+    rst = await tbl
+      .query()
+      .nearestTo(queryVec)
+      .limit(2)
+      .minimumNprobes(15)
+      .toArrow();
+    expect(rst.numRows).toBe(2);
+    rst = await tbl
+      .query()
+      .nearestTo(queryVec)
+      .limit(2)
+      .minimumNprobes(10)
+      .maximumNprobes(20)
+      .toArrow();
+    expect(rst.numRows).toBe(2);
+
+    expect(() => tbl.query().nearestTo(queryVec).minimumNprobes(0)).toThrow(
+      "Invalid input, minimum_nprobes must be greater than 0",
+    );
+    expect(() => tbl.query().nearestTo(queryVec).maximumNprobes(5)).toThrow(
+      "Invalid input, maximum_nprobes must be greater than minimum_nprobes",
+    );
+
     await tbl.dropIndex("vec_idx");
     const indices2 = await tbl.listIndices();
     expect(indices2.length).toBe(0);
@@ -1506,7 +1537,9 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
       ];
       const table = await db.createTable("test", data);
       await table.createIndex("text", {
-        config: Index.fts(),
+        config: Index.fts({
+          withPosition: true,
+        }),
       });
 
       const results = await table.search("lance").toArray();
@@ -1529,6 +1562,18 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
 
       const results = await table.search("hello").toArray();
       expect(results[0].text).toBe(data[0].text);
+
+      const results2 = await table
+        .search(new MatchQuery("hello world", "text"))
+        .toArray();
+      expect(results2.length).toBe(2);
+
+      const results3 = await table
+        .search(
+          new MatchQuery("hello world", "text", { operator: Operator.And }),
+        )
+        .toArray();
+      expect(results3.length).toBe(1);
     });
 
     test("full text search without lowercase", async () => {
@@ -1559,7 +1604,9 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
       ];
       const table = await db.createTable("test", data);
       await table.createIndex("text", {
-        config: Index.fts(),
+        config: Index.fts({
+          withPosition: true,
+        }),
       });
 
       const results = await table.search("world").toArray();
@@ -1603,6 +1650,60 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
       expect(resultSet.has("fob")).toBe(true);
       expect(resultSet.has("fo")).toBe(true);
       expect(resultSet.has("food")).toBe(true);
+
+      const prefixResults = await table
+        .search(
+          new MatchQuery("foo", "text", { fuzziness: 3, prefixLength: 3 }),
+        )
+        .toArray();
+      expect(prefixResults.length).toBe(2);
+      const resultSet2 = new Set(prefixResults.map((r) => r.text));
+      expect(resultSet2.has("foo")).toBe(true);
+      expect(resultSet2.has("food")).toBe(true);
+    });
+
+    test("full text search boolean query", async () => {
+      const db = await connect(tmpDir.name);
+      const data = [
+        { text: "The cat and dog are playing" },
+        { text: "The cat is sleeping" },
+        { text: "The dog is barking" },
+        { text: "The dog chases the cat" },
+      ];
+      const table = await db.createTable("test", data);
+      await table.createIndex("text", {
+        config: Index.fts({ withPosition: false }),
+      });
+
+      const shouldResults = await table
+        .search(
+          new BooleanQuery([
+            [Occur.Should, new MatchQuery("cat", "text")],
+            [Occur.Should, new MatchQuery("dog", "text")],
+          ]),
+        )
+        .toArray();
+      expect(shouldResults.length).toBe(4);
+
+      const mustResults = await table
+        .search(
+          new BooleanQuery([
+            [Occur.Must, new MatchQuery("cat", "text")],
+            [Occur.Must, new MatchQuery("dog", "text")],
+          ]),
+        )
+        .toArray();
+      expect(mustResults.length).toBe(2);
+
+      const mustNotResults = await table
+        .search(
+          new BooleanQuery([
+            [Occur.Must, new MatchQuery("cat", "text")],
+            [Occur.MustNot, new MatchQuery("dog", "text")],
+          ]),
+        )
+        .toArray();
+      expect(mustNotResults.length).toBe(1);
     });
 
     test.each([

nodejs/lancedb/arrow.ts

@@ -417,7 +417,9 @@ function inferSchema(
         } else {
           const inferredType = inferType(value, path, opts);
           if (inferredType === undefined) {
-            throw new Error(`Failed to infer data type for field ${path.join(".")} at row ${rowI}. \
+            throw new Error(`Failed to infer data type for field ${path.join(
+              ".",
+            )} at row ${rowI}. \
                              Consider providing an explicit schema.`);
           }
           pathTree.set(path, inferredType);
@@ -799,11 +801,17 @@ async function applyEmbeddingsFromMetadata(
         `Cannot apply embedding function because the source column '${functionEntry.sourceColumn}' was not present in the data`,
       );
     }
+
+    // Check if destination column exists and handle accordingly
     if (columns[destColumn] !== undefined) {
-      throw new Error(
+      const existingColumn = columns[destColumn];
-        `Attempt to apply embeddings to table failed because column ${destColumn} already existed`,
+      // If the column exists but is all null, we can fill it with embeddings
-      );
+      if (existingColumn.nullCount !== existingColumn.length) {
+        // Column has non-null values, skip embedding application
+        continue;
+      }
     }
+
     if (table.batches.length > 1) {
       throw new Error(
         "Internal error: `makeArrowTable` unexpectedly created a table with more than one batch",
@@ -903,11 +911,23 @@ async function applyEmbeddings<T>(
       );
     }
   } else {
+    // Check if destination column exists and handle accordingly
     if (Object.prototype.hasOwnProperty.call(newColumns, destColumn)) {
-      throw new Error(
+      const existingColumn = newColumns[destColumn];
-        `Attempt to apply embeddings to table failed because column ${destColumn} already existed`,
+      // If the column exists but is all null, we can fill it with embeddings
-      );
+      if (existingColumn.nullCount !== existingColumn.length) {
+        // Column has non-null values, skip embedding application and return table as-is
+        let newTable = new ArrowTable(newColumns);
+        if (schema != null) {
+          newTable = alignTable(newTable, schema as Schema);
+        }
+        return new ArrowTable(
+          new Schema(newTable.schema.fields, schemaMetadata),
+          newTable.batches,
+        );
+      }
     }
+
     if (table.batches.length > 1) {
       throw new Error(
         "Internal error: `makeArrowTable` unexpectedly created a table with more than one batch",

nodejs/lancedb/index.ts

@@ -64,7 +64,10 @@ export {
   PhraseQuery,
   BoostQuery,
   MultiMatchQuery,
+  BooleanQuery,
   FullTextQueryType,
+  Operator,
+  Occur,
 } from "./query";
 
 export {

nodejs/lancedb/query.ts

@@ -448,6 +448,10 @@ export class VectorQuery extends QueryBase<NativeVectorQuery> {
    * For best results we recommend tuning this parameter with a benchmark against
    * your actual data to find the smallest possible value that will still give
    * you the desired recall.
+   *
+   * For more fine grained control over behavior when you have a very narrow filter
+   * you can use `minimumNprobes` and `maximumNprobes`.  This method sets both
+   * the minimum and maximum to the same value.
    */
   nprobes(nprobes: number): VectorQuery {
     super.doCall((inner) => inner.nprobes(nprobes));
@@ -455,6 +459,33 @@ export class VectorQuery extends QueryBase<NativeVectorQuery> {
     return this;
   }
 
+  /**
+   * Set the minimum number of probes used.
+   *
+   * This controls the minimum number of partitions that will be searched.  This
+   * parameter will impact every query against a vector index, regardless of the
+   * filter.  See `nprobes` for more details.  Higher values will increase recall
+   * but will also increase latency.
+   */
+  minimumNprobes(minimumNprobes: number): VectorQuery {
+    super.doCall((inner) => inner.minimumNprobes(minimumNprobes));
+    return this;
+  }
+
+  /**
+   * Set the maximum number of probes used.
+   *
+   * This controls the maximum number of partitions that will be searched.  If this
+   * number is greater than minimumNprobes then the excess partitions will _only_ be
+   * searched if we have not found enough results.  This can be useful when there is
+   * a narrow filter to allow these queries to spend more time searching and avoid
+   * potential false negatives.
+   */
+  maximumNprobes(maximumNprobes: number): VectorQuery {
+    super.doCall((inner) => inner.maximumNprobes(maximumNprobes));
+    return this;
+  }
+
   /*
    * Set the distance range to use
    *
@@ -762,6 +793,31 @@ export enum FullTextQueryType {
   MatchPhrase = "match_phrase",
   Boost = "boost",
   MultiMatch = "multi_match",
+  Boolean = "boolean",
+}
+
+/**
+ * Enum representing the logical operators used in full-text queries.
+ *
+ * - `And`: All terms must match.
+ * - `Or`: At least one term must match.
+ */
+export enum Operator {
+  And = "AND",
+  Or = "OR",
+}
+
+/**
+ * Enum representing the occurrence of terms in full-text queries.
+ *
+ * - `Must`: The term must be present in the document.
+ * - `Should`: The term should contribute to the document score, but is not required.
+ * - `MustNot`: The term must not be present in the document.
+ */
+export enum Occur {
+  Should = "SHOULD",
+  Must = "MUST",
+  MustNot = "MUST_NOT",
 }
 
 /**
@@ -791,6 +847,7 @@ export function instanceOfFullTextQuery(obj: any): obj is FullTextQuery {
 export class MatchQuery implements FullTextQuery {
   /** @ignore */
   public readonly inner: JsFullTextQuery;
+
   /**
    * Creates an instance of MatchQuery.
    *
@@ -800,6 +857,8 @@ export class MatchQuery implements FullTextQuery {
    *   - `boost`: The boost factor for the query (default is 1.0).
    *   - `fuzziness`: The fuzziness level for the query (default is 0).
    *   - `maxExpansions`: The maximum number of terms to consider for fuzzy matching (default is 50).
+   *   - `operator`: The logical operator to use for combining terms in the query (default is "OR").
+   *   - `prefixLength`: The number of beginning characters being unchanged for fuzzy matching.
    */
   constructor(
     query: string,
@@ -808,6 +867,8 @@ export class MatchQuery implements FullTextQuery {
       boost?: number;
       fuzziness?: number;
       maxExpansions?: number;
+      operator?: Operator;
+      prefixLength?: number;
     },
   ) {
     let fuzziness = options?.fuzziness;
@@ -820,6 +881,8 @@ export class MatchQuery implements FullTextQuery {
       options?.boost ?? 1.0,
       fuzziness,
       options?.maxExpansions ?? 50,
+      options?.operator ?? Operator.Or,
+      options?.prefixLength ?? 0,
     );
   }
 
@@ -836,9 +899,11 @@ export class PhraseQuery implements FullTextQuery {
    *
    * @param query - The phrase to search for in the specified column.
    * @param column - The name of the column to search within.
+   * @param options - Optional parameters for the phrase query.
+   *   - `slop`: The maximum number of intervening unmatched positions allowed between words in the phrase (default is 0).
    */
-  constructor(query: string, column: string) {
+  constructor(query: string, column: string, options?: { slop?: number }) {
-    this.inner = JsFullTextQuery.phraseQuery(query, column);
+    this.inner = JsFullTextQuery.phraseQuery(query, column, options?.slop ?? 0);
   }
 
   queryType(): FullTextQueryType {
@@ -889,18 +954,21 @@ export class MultiMatchQuery implements FullTextQuery {
    * @param columns - An array of column names to search within.
    * @param options - Optional parameters for the multi-match query.
    *  - `boosts`: An array of boost factors for each column (default is 1.0 for all).
+   *  - `operator`: The logical operator to use for combining terms in the query (default is "OR").
    */
   constructor(
     query: string,
     columns: string[],
     options?: {
       boosts?: number[];
+      operator?: Operator;
     },
   ) {
     this.inner = JsFullTextQuery.multiMatchQuery(
       query,
       columns,
       options?.boosts,
+      options?.operator ?? Operator.Or,
     );
   }
 
@@ -908,3 +976,23 @@ export class MultiMatchQuery implements FullTextQuery {
     return FullTextQueryType.MultiMatch;
   }
 }
+
+export class BooleanQuery implements FullTextQuery {
+  /** @ignore */
+  public readonly inner: JsFullTextQuery;
+  /**
+   * Creates an instance of BooleanQuery.
+   *
+   * @param queries - An array of (Occur, FullTextQuery objects) to combine.
+   * Occur specifies whether the query must match, or should match.
+   */
+  constructor(queries: [Occur, FullTextQuery][]) {
+    this.inner = JsFullTextQuery.booleanQuery(
+      queries.map(([occur, query]) => [occur, query.inner]),
+    );
+  }
+
+  queryType(): FullTextQueryType {
+    return FullTextQueryType.Boolean;
+  }
+}

nodejs/npm/darwin-arm64/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-darwin-arm64",
-	"version": "0.19.1-beta.5",
+	"version": "0.21.1-beta.0",
 	"os": ["darwin"],
 	"cpu": ["arm64"],
 	"main": "lancedb.darwin-arm64.node",

nodejs/npm/darwin-x64/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-darwin-x64",
-	"version": "0.19.1-beta.5",
+	"version": "0.21.1-beta.0",
 	"os": ["darwin"],
 	"cpu": ["x64"],
 	"main": "lancedb.darwin-x64.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-gnu",
-	"version": "0.19.1-beta.5",
+	"version": "0.21.1-beta.0",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-gnu.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-musl",
-	"version": "0.19.1-beta.5",
+	"version": "0.21.1-beta.0",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-musl.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-gnu",
-	"version": "0.19.1-beta.5",
+	"version": "0.21.1-beta.0",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-gnu.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-musl",
-	"version": "0.19.1-beta.5",
+	"version": "0.21.1-beta.0",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-musl.node",

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

@@ -1,6 +1,6 @@
 {
   "name": "@lancedb/lancedb-win32-arm64-msvc",
-  "version": "0.19.1-beta.5",
+  "version": "0.21.1-beta.0",
   "os": [
     "win32"
   ],

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-win32-x64-msvc",
-	"version": "0.19.1-beta.5",
+	"version": "0.21.1-beta.0",
 	"os": ["win32"],
 	"cpu": ["x64"],
 	"main": "lancedb.win32-x64-msvc.node",

nodejs/package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "@lancedb/lancedb",
-  "version": "0.19.1-beta.5",
+  "version": "0.21.1-beta.0",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "@lancedb/lancedb",
-      "version": "0.19.1-beta.5",
+      "version": "0.21.1-beta.0",
       "cpu": [
         "x64",
         "arm64"

nodejs/package.json

@@ -11,7 +11,7 @@
     "ann"
   ],
   "private": false,
-  "version": "0.19.1-beta.5",
+  "version": "0.21.1-beta.0",
   "main": "dist/index.js",
   "exports": {
     ".": "./dist/index.js",

nodejs/src/index.rs

@@ -125,32 +125,30 @@ impl Index {
         ascii_folding: Option<bool>,
     ) -> Self {
         let mut opts = FtsIndexBuilder::default();
-        let mut tokenizer_configs = opts.tokenizer_configs.clone();
         if let Some(with_position) = with_position {
             opts = opts.with_position(with_position);
         }
         if let Some(base_tokenizer) = base_tokenizer {
-            tokenizer_configs = tokenizer_configs.base_tokenizer(base_tokenizer);
+            opts = opts.base_tokenizer(base_tokenizer);
         }
         if let Some(language) = language {
-            tokenizer_configs = tokenizer_configs.language(&language).unwrap();
+            opts = opts.language(&language).unwrap();
         }
         if let Some(max_token_length) = max_token_length {
-            tokenizer_configs = tokenizer_configs.max_token_length(Some(max_token_length as usize));
+            opts = opts.max_token_length(Some(max_token_length as usize));
         }
         if let Some(lower_case) = lower_case {
-            tokenizer_configs = tokenizer_configs.lower_case(lower_case);
+            opts = opts.lower_case(lower_case);
         }
         if let Some(stem) = stem {
-            tokenizer_configs = tokenizer_configs.stem(stem);
+            opts = opts.stem(stem);
         }
         if let Some(remove_stop_words) = remove_stop_words {
-            tokenizer_configs = tokenizer_configs.remove_stop_words(remove_stop_words);
+            opts = opts.remove_stop_words(remove_stop_words);
         }
         if let Some(ascii_folding) = ascii_folding {
-            tokenizer_configs = tokenizer_configs.ascii_folding(ascii_folding);
+            opts = opts.ascii_folding(ascii_folding);
         }
-        opts.tokenizer_configs = tokenizer_configs;
 
         Self {
             inner: Mutex::new(Some(LanceDbIndex::FTS(opts))),

nodejs/src/query.rs

@@ -4,7 +4,8 @@
 use std::sync::Arc;
 
 use lancedb::index::scalar::{
-    BoostQuery, FtsQuery, FullTextSearchQuery, MatchQuery, MultiMatchQuery, PhraseQuery,
+    BooleanQuery, BoostQuery, FtsQuery, FullTextSearchQuery, MatchQuery, MultiMatchQuery, Occur,
+    Operator, PhraseQuery,
 };
 use lancedb::query::ExecutableQuery;
 use lancedb::query::Query as LanceDbQuery;
@@ -177,6 +178,31 @@ impl VectorQuery {
         self.inner = self.inner.clone().nprobes(nprobe as usize);
     }
 
+    #[napi]
+    pub fn minimum_nprobes(&mut self, minimum_nprobe: u32) -> napi::Result<()> {
+        self.inner = self
+            .inner
+            .clone()
+            .minimum_nprobes(minimum_nprobe as usize)
+            .default_error()?;
+        Ok(())
+    }
+
+    #[napi]
+    pub fn maximum_nprobes(&mut self, maximum_nprobes: u32) -> napi::Result<()> {
+        let maximum_nprobes = if maximum_nprobes == 0 {
+            None
+        } else {
+            Some(maximum_nprobes as usize)
+        };
+        self.inner = self
+            .inner
+            .clone()
+            .maximum_nprobes(maximum_nprobes)
+            .default_error()?;
+        Ok(())
+    }
+
     #[napi]
     pub fn distance_range(&mut self, lower_bound: Option<f64>, upper_bound: Option<f64>) {
         // napi doesn't support f32, so we have to convert to f32
@@ -308,6 +334,8 @@ impl JsFullTextQuery {
         boost: f64,
         fuzziness: Option<u32>,
         max_expansions: u32,
+        operator: String,
+        prefix_length: u32,
     ) -> napi::Result<Self> {
         Ok(Self {
             inner: MatchQuery::new(query)
@@ -315,14 +343,23 @@ impl JsFullTextQuery {
                 .with_boost(boost as f32)
                 .with_fuzziness(fuzziness)
                 .with_max_expansions(max_expansions as usize)
+                .with_operator(
+                    Operator::try_from(operator.as_str()).map_err(|e| {
+                        napi::Error::from_reason(format!("Invalid operator: {}", e))
+                    })?,
+                )
+                .with_prefix_length(prefix_length)
                 .into(),
         })
     }
 
     #[napi(factory)]
-    pub fn phrase_query(query: String, column: String) -> napi::Result<Self> {
+    pub fn phrase_query(query: String, column: String, slop: u32) -> napi::Result<Self> {
         Ok(Self {
-            inner: PhraseQuery::new(query).with_column(Some(column)).into(),
+            inner: PhraseQuery::new(query)
+                .with_column(Some(column))
+                .with_slop(slop)
+                .into(),
         })
     }
 
@@ -348,6 +385,7 @@ impl JsFullTextQuery {
         query: String,
         columns: Vec<String>,
         boosts: Option<Vec<f64>>,
+        operator: String,
     ) -> napi::Result<Self> {
         let q = match boosts {
             Some(boosts) => MultiMatchQuery::try_new(query, columns)
@@ -358,7 +396,37 @@ impl JsFullTextQuery {
             napi::Error::from_reason(format!("Failed to create multi match query: {}", e))
         })?;
 
-        Ok(Self { inner: q.into() })
+        let operator = Operator::try_from(operator.as_str()).map_err(|e| {
+            napi::Error::from_reason(format!("Invalid operator for multi match query: {}", e))
+        })?;
+
+        Ok(Self {
+            inner: q.with_operator(operator).into(),
+        })
+    }
+
+    #[napi(factory)]
+    pub fn boolean_query(queries: Vec<(String, &JsFullTextQuery)>) -> napi::Result<Self> {
+        let mut sub_queries = Vec::with_capacity(queries.len());
+        for (occur, q) in queries {
+            let occur = Occur::try_from(occur.as_str())
+                .map_err(|e| napi::Error::from_reason(e.to_string()))?;
+            sub_queries.push((occur, q.inner.clone()));
+        }
+        Ok(Self {
+            inner: BooleanQuery::new(sub_queries).into(),
+        })
+    }
+
+    #[napi(getter)]
+    pub fn query_type(&self) -> String {
+        match self.inner {
+            FtsQuery::Match(_) => "match".to_string(),
+            FtsQuery::Phrase(_) => "phrase".to_string(),
+            FtsQuery::Boost(_) => "boost".to_string(),
+            FtsQuery::MultiMatch(_) => "multi_match".to_string(),
+            FtsQuery::Boolean(_) => "boolean".to_string(),
+        }
     }
 }
 

python/.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.22.1"
+current_version = "0.24.1-beta.1"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

python/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb-python"
-version = "0.22.1"
+version = "0.24.1-beta.1"
 edition.workspace = true
 description = "Python bindings for LanceDB"
 license.workspace = true
@@ -14,11 +14,11 @@ name = "_lancedb"
 crate-type = ["cdylib"]
 
 [dependencies]
-arrow = { version = "54.1", features = ["pyarrow"] }
+arrow = { version = "55.1", features = ["pyarrow"] }
 lancedb = { path = "../rust/lancedb", default-features = false }
 env_logger.workspace = true
-pyo3 = { version = "0.23", features = ["extension-module", "abi3-py39"] }
+pyo3 = { version = "0.24", features = ["extension-module", "abi3-py39"] }
-pyo3-async-runtimes = { version = "0.23", features = [
+pyo3-async-runtimes = { version = "0.24", features = [
     "attributes",
     "tokio-runtime",
 ] }
@@ -27,7 +27,7 @@ futures.workspace = true
 tokio = { version = "1.40", features = ["sync"] }
 
 [build-dependencies]
-pyo3-build-config = { version = "0.23", features = [
+pyo3-build-config = { version = "0.24", features = [
     "extension-module",
     "abi3-py39",
 ] }

python/pyproject.toml

@@ -60,6 +60,7 @@ tests = [
     "pyarrow-stubs",
     "pylance>=0.25",
     "requests",
+    "datafusion",
 ]
 dev = [
     "ruff",
@@ -84,7 +85,7 @@ embeddings = [
     "boto3>=1.28.57",
     "awscli>=1.29.57",
     "botocore>=1.31.57",
-    "ollama",
+    "ollama>=0.3.0",
     "ibm-watsonx-ai>=1.1.2",
 ]
 azure = ["adlfs>=2024.2.0"]

python/python/lancedb/_lancedb.pyi

@@ -143,6 +143,8 @@ class VectorQuery:
     def postfilter(self): ...
     def refine_factor(self, refine_factor: int): ...
     def nprobes(self, nprobes: int): ...
+    def minimum_nprobes(self, minimum_nprobes: int): ...
+    def maximum_nprobes(self, maximum_nprobes: int): ...
     def bypass_vector_index(self): ...
     def nearest_to_text(self, query: dict) -> HybridQuery: ...
     def to_query_request(self) -> PyQueryRequest: ...
@@ -158,6 +160,8 @@ class HybridQuery:
     def distance_type(self, distance_type: str): ...
     def refine_factor(self, refine_factor: int): ...
     def nprobes(self, nprobes: int): ...
+    def minimum_nprobes(self, minimum_nprobes: int): ...
+    def maximum_nprobes(self, maximum_nprobes: int): ...
     def bypass_vector_index(self): ...
     def to_vector_query(self) -> VectorQuery: ...
     def to_fts_query(self) -> FTSQuery: ...
@@ -165,23 +169,21 @@ class HybridQuery:
     def get_with_row_id(self) -> bool: ...
     def to_query_request(self) -> PyQueryRequest: ...
 
-class PyFullTextSearchQuery:
+class FullTextQuery:
-    columns: Optional[List[str]]
+    pass
-    query: str
-    limit: Optional[int]
-    wand_factor: Optional[float]
 
 class PyQueryRequest:
     limit: Optional[int]
     offset: Optional[int]
     filter: Optional[Union[str, bytes]]
-    full_text_search: Optional[PyFullTextSearchQuery]
+    full_text_search: Optional[FullTextQuery]
     select: Optional[Union[str, List[str]]]
     fast_search: Optional[bool]
     with_row_id: Optional[bool]
     column: Optional[str]
     query_vector: Optional[List[pa.Array]]
-    nprobes: Optional[int]
+    minimum_nprobes: Optional[int]
+    maximum_nprobes: Optional[int]
     lower_bound: Optional[float]
     upper_bound: Optional[float]
     ef: Optional[int]

python/python/lancedb/embeddings/ollama.py

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

python/python/lancedb/index.py

@@ -102,7 +102,7 @@ class FTS:
 
     Attributes
     ----------
-    with_position : bool, default True
+    with_position : bool, default False
         Whether to store the position of the token in the document. Setting this
         to False can reduce the size of the index and improve indexing speed,
         but it will disable support for phrase queries.
@@ -118,25 +118,25 @@ class FTS:
         ignored.
     lower_case : bool, default True
         Whether to convert the token to lower case. This makes queries case-insensitive.
-    stem : bool, default False
+    stem : bool, default True
         Whether to stem the token. Stemming reduces words to their root form.
         For example, in English "running" and "runs" would both be reduced to "run".
-    remove_stop_words : bool, default False
+    remove_stop_words : bool, default True
         Whether to remove stop words. Stop words are common words that are often
         removed from text before indexing. For example, in English "the" and "and".
-    ascii_folding : bool, default False
+    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".
     """
 
-    with_position: bool = True
+    with_position: bool = False
     base_tokenizer: Literal["simple", "raw", "whitespace"] = "simple"
     language: str = "English"
     max_token_length: Optional[int] = 40
     lower_case: bool = True
-    stem: bool = False
+    stem: bool = True
-    remove_stop_words: bool = False
+    remove_stop_words: bool = True
-    ascii_folding: bool = False
+    ascii_folding: bool = True
 
 
 @dataclass

python/python/lancedb/query.py

@@ -4,7 +4,6 @@
 from __future__ import annotations
 
 from abc import ABC, abstractmethod
-import abc
 from concurrent.futures import ThreadPoolExecutor
 from enum import Enum
 from datetime import timedelta
@@ -88,15 +87,28 @@ def ensure_vector_query(
         return val
 
 
-class FullTextQueryType(Enum):
+class FullTextQueryType(str, Enum):
     MATCH = "match"
     MATCH_PHRASE = "match_phrase"
     BOOST = "boost"
     MULTI_MATCH = "multi_match"
+    BOOLEAN = "boolean"
 
 
-class FullTextQuery(abc.ABC, pydantic.BaseModel):
+class FullTextOperator(str, Enum):
-    @abc.abstractmethod
+    AND = "AND"
+    OR = "OR"
+
+
+class Occur(str, Enum):
+    SHOULD = "SHOULD"
+    MUST = "MUST"
+    MUST_NOT = "MUST_NOT"
+
+
+@pydantic.dataclasses.dataclass
+class FullTextQuery(ABC):
+    @abstractmethod
     def query_type(self) -> FullTextQueryType:
         """
         Get the query type of the query.
@@ -106,193 +118,178 @@ class FullTextQuery(abc.ABC, pydantic.BaseModel):
         str
             The type of the query.
         """
+        pass
 
-    @abc.abstractmethod
+    def __and__(self, other: "FullTextQuery") -> "FullTextQuery":
-    def to_dict(self) -> dict:
         """
-        Convert the query to a dictionary.
+        Combine two queries with a logical AND operation.
-
-        Returns
-        -------
-        dict
-            The query as a dictionary.
-        """
-
-
-class MatchQuery(FullTextQuery):
-    query: str
-    column: str
-    boost: float = 1.0
-    fuzziness: int = 0
-    max_expansions: int = 50
-
-    def __init__(
-        self,
-        query: str,
-        column: str,
-        *,
-        boost: float = 1.0,
-        fuzziness: int = 0,
-        max_expansions: int = 50,
-    ):
-        """
-        Match query for full-text search.
 
         Parameters
         ----------
-        query : str
+        other : FullTextQuery
-            The query string to match against.
+            The other query to combine with.
-        column : str
+
-            The name of the column to match against.
+        Returns
-        boost : float, default 1.0
+        -------
-            The boost factor for the query.
+        FullTextQuery
-            The score of each matching document is multiplied by this value.
+            A new query that combines both queries with AND.
-        fuzziness : int, optional
-            The maximum edit distance for each term in the match query.
-            Defaults to 0 (exact match).
-            If None, fuzziness is applied automatically by the rules:
-                - 0 for terms with length <= 2
-                - 1 for terms with length <= 5
-                - 2 for terms with length > 5
-        max_expansions : int, optional
-            The maximum number of terms to consider for fuzzy matching.
-            Defaults to 50.
         """
-        super().__init__(
+        return BooleanQuery([(Occur.MUST, self), (Occur.MUST, other)])
-            query=query,
+
-            column=column,
+    def __or__(self, other: "FullTextQuery") -> "FullTextQuery":
-            boost=boost,
+        """
-            fuzziness=fuzziness,
+        Combine two queries with a logical OR operation.
-            max_expansions=max_expansions,
+
-        )
+        Parameters
+        ----------
+        other : FullTextQuery
+            The other query to combine with.
+
+        Returns
+        -------
+        FullTextQuery
+            A new query that combines both queries with OR.
+        """
+        return BooleanQuery([(Occur.SHOULD, self), (Occur.SHOULD, other)])
+
+
+@pydantic.dataclasses.dataclass
+class MatchQuery(FullTextQuery):
+    """
+    Match query for full-text search.
+
+    Parameters
+    ----------
+    query : str
+        The query string to match against.
+    column : str
+        The name of the column to match against.
+    boost : float, default 1.0
+        The boost factor for the query.
+        The score of each matching document is multiplied by this value.
+    fuzziness : int, optional
+        The maximum edit distance for each term in the match query.
+        Defaults to 0 (exact match).
+        If None, fuzziness is applied automatically by the rules:
+            - 0 for terms with length <= 2
+            - 1 for terms with length <= 5
+            - 2 for terms with length > 5
+    max_expansions : int, optional
+        The maximum number of terms to consider for fuzzy matching.
+        Defaults to 50.
+    operator : FullTextOperator, default OR
+        The operator to use for combining the query results.
+        Can be either `AND` or `OR`.
+        If `AND`, all terms in the query must match.
+        If `OR`, at least one term in the query must match.
+    prefix_length : int, optional
+        The number of beginning characters being unchanged for fuzzy matching.
+        This is useful to achieve prefix matching.
+    """
+
+    query: str
+    column: str
+    boost: float = pydantic.Field(1.0, kw_only=True)
+    fuzziness: int = pydantic.Field(0, kw_only=True)
+    max_expansions: int = pydantic.Field(50, kw_only=True)
+    operator: FullTextOperator = pydantic.Field(FullTextOperator.OR, kw_only=True)
+    prefix_length: int = pydantic.Field(0, kw_only=True)
 
     def query_type(self) -> FullTextQueryType:
         return FullTextQueryType.MATCH
 
-    def to_dict(self) -> dict:
-        return {
-            "match": {
-                self.column: {
-                    "query": self.query,
-                    "boost": self.boost,
-                    "fuzziness": self.fuzziness,
-                    "max_expansions": self.max_expansions,
-                }
-            }
-        }
-
 
+@pydantic.dataclasses.dataclass
 class PhraseQuery(FullTextQuery):
+    """
+    Phrase query for full-text search.
+
+    Parameters
+    ----------
+    query : str
+        The query string to match against.
+    column : str
+        The name of the column to match against.
+    """
+
     query: str
     column: str
-
+    slop: int = pydantic.Field(0, kw_only=True)
-    def __init__(self, query: str, column: str):
-        """
-        Phrase query for full-text search.
-
-        Parameters
-        ----------
-        query : str
-            The query string to match against.
-        column : str
-            The name of the column to match against.
-        """
-        super().__init__(query=query, column=column)
 
     def query_type(self) -> FullTextQueryType:
         return FullTextQueryType.MATCH_PHRASE
 
-    def to_dict(self) -> dict:
-        return {
-            "match_phrase": {
-                self.column: self.query,
-            }
-        }
-
 
+@pydantic.dataclasses.dataclass
 class BoostQuery(FullTextQuery):
+    """
+    Boost query for full-text search.
+
+    Parameters
+    ----------
+    positive : dict
+        The positive query object.
+    negative : dict
+        The negative query object.
+    negative_boost : float, default 0.5
+        The boost factor for the negative query.
+    """
+
     positive: FullTextQuery
     negative: FullTextQuery
-    negative_boost: float = 0.5
+    negative_boost: float = pydantic.Field(0.5, kw_only=True)
-
-    def __init__(
-        self,
-        positive: FullTextQuery,
-        negative: FullTextQuery,
-        *,
-        negative_boost: float = 0.5,
-    ):
-        """
-        Boost query for full-text search.
-
-        Parameters
-        ----------
-        positive : dict
-            The positive query object.
-        negative : dict
-            The negative query object.
-        negative_boost : float
-            The boost factor for the negative query.
-        """
-        super().__init__(
-            positive=positive, negative=negative, negative_boost=negative_boost
-        )
 
     def query_type(self) -> FullTextQueryType:
         return FullTextQueryType.BOOST
 
-    def to_dict(self) -> dict:
-        return {
-            "boost": {
-                "positive": self.positive.to_dict(),
-                "negative": self.negative.to_dict(),
-                "negative_boost": self.negative_boost,
-            }
-        }
-
 
+@pydantic.dataclasses.dataclass
 class MultiMatchQuery(FullTextQuery):
+    """
+    Multi-match query for full-text search.
+
+    Parameters
+    ----------
+    query : str | list[Query]
+        If a string, the query string to match against.
+    columns : list[str]
+        The list of columns to match against.
+    boosts : list[float], optional
+        The list of boost factors for each column. If not provided,
+        all columns will have the same boost factor.
+    operator : FullTextOperator, default OR
+        The operator to use for combining the query results.
+        Can be either `AND` or `OR`.
+        It would be applied to all columns individually.
+        For example, if the operator is `AND`,
+        then the query "hello world" is equal to
+        `match("hello AND world", column1) OR match("hello AND world", column2)`.
+    """
+
     query: str
     columns: list[str]
-    boosts: list[float]
+    boosts: Optional[list[float]] = pydantic.Field(None, kw_only=True)
-
+    operator: FullTextOperator = pydantic.Field(FullTextOperator.OR, kw_only=True)
-    def __init__(
-        self,
-        query: str,
-        columns: list[str],
-        *,
-        boosts: Optional[list[float]] = None,
-    ):
-        """
-        Multi-match query for full-text search.
-
-        Parameters
-        ----------
-        query : str
-            The query string to match against.
-
-        columns : list[str]
-            The list of columns to match against.
-
-        boosts : list[float], optional
-            The list of boost factors for each column. If not provided,
-            all columns will have the same boost factor.
-        """
-        if boosts is None:
-            boosts = [1.0] * len(columns)
-        super().__init__(query=query, columns=columns, boosts=boosts)
 
     def query_type(self) -> FullTextQueryType:
         return FullTextQueryType.MULTI_MATCH
 
-    def to_dict(self) -> dict:
+
-        return {
+@pydantic.dataclasses.dataclass
-            "multi_match": {
+class BooleanQuery(FullTextQuery):
-                "query": self.query,
+    """
-                "columns": self.columns,
+    Boolean query for full-text search.
-                "boost": self.boosts,
+
-            }
+    Parameters
-        }
+    ----------
+    queries : list[tuple(Occur, FullTextQuery)]
+        The list of queries with their occurrence requirements.
+    """
+
+    queries: list[tuple[Occur, FullTextQuery]]
+
+    def query_type(self) -> FullTextQueryType:
+        return FullTextQueryType.BOOLEAN
 
 
 class FullTextSearchQuery(pydantic.BaseModel):
@@ -445,8 +442,18 @@ class Query(pydantic.BaseModel):
     # which columns to return in the results
     columns: Optional[Union[List[str], Dict[str, str]]] = None
 
-    # number of IVF partitions to search
+    # minimum number of IVF partitions to search
-    nprobes: Optional[int] = None
+    #
+    # If None then a default value (20) will be used.
+    minimum_nprobes: Optional[int] = None
+
+    # maximum number of IVF partitions to search
+    #
+    # If None then a default value (20) will be used.
+    #
+    # If 0 then no limit will be applied and all partitions could be searched
+    # if needed to satisfy the limit.
+    maximum_nprobes: Optional[int] = None
 
     # lower bound for distance search
     lower_bound: Optional[float] = None
@@ -484,7 +491,8 @@ class Query(pydantic.BaseModel):
         query.vector_column = req.column
         query.vector = req.query_vector
         query.distance_type = req.distance_type
-        query.nprobes = req.nprobes
+        query.minimum_nprobes = req.minimum_nprobes
+        query.maximum_nprobes = req.maximum_nprobes
         query.lower_bound = req.lower_bound
         query.upper_bound = req.upper_bound
         query.ef = req.ef
@@ -493,10 +501,8 @@ class Query(pydantic.BaseModel):
         query.postfilter = req.postfilter
         if req.full_text_search is not None:
             query.full_text_query = FullTextSearchQuery(
-                columns=req.full_text_search.columns,
+                columns=None,
-                query=req.full_text_search.query,
+                query=req.full_text_search,
-                limit=req.full_text_search.limit,
-                wand_factor=req.full_text_search.wand_factor,
             )
         return query
 
@@ -1047,7 +1053,8 @@ class LanceVectorQueryBuilder(LanceQueryBuilder):
         super().__init__(table)
         self._query = query
         self._distance_type = None
-        self._nprobes = None
+        self._minimum_nprobes = None
+        self._maximum_nprobes = None
         self._lower_bound = None
         self._upper_bound = None
         self._refine_factor = None
@@ -1110,6 +1117,10 @@ class LanceVectorQueryBuilder(LanceQueryBuilder):
         See discussion in [Querying an ANN Index][querying-an-ann-index] for
         tuning advice.
 
+        This method sets both the minimum and maximum number of probes to the same
+        value. See `minimum_nprobes` and `maximum_nprobes` for more fine-grained
+        control.
+
         Parameters
         ----------
         nprobes: int
@@ -1120,7 +1131,36 @@ class LanceVectorQueryBuilder(LanceQueryBuilder):
         LanceVectorQueryBuilder
             The LanceQueryBuilder object.
         """
-        self._nprobes = nprobes
+        self._minimum_nprobes = nprobes
+        self._maximum_nprobes = nprobes
+        return self
+
+    def minimum_nprobes(self, minimum_nprobes: int) -> LanceVectorQueryBuilder:
+        """Set the minimum number of probes to use.
+
+        See `nprobes` for more details.
+
+        These partitions will be searched on every vector query and will increase recall
+        at the expense of latency.
+        """
+        self._minimum_nprobes = minimum_nprobes
+        return self
+
+    def maximum_nprobes(self, maximum_nprobes: int) -> LanceVectorQueryBuilder:
+        """Set the maximum number of probes to use.
+
+        See `nprobes` for more details.
+
+        If this value is greater than `minimum_nprobes` then the excess partitions
+        will be searched only if we have not found enough results.
+
+        This can be useful when there is a narrow filter to allow these queries to
+        spend more time searching and avoid potential false negatives.
+
+        If this value is 0 then no limit will be applied and all partitions could be
+        searched if needed to satisfy the limit.
+        """
+        self._maximum_nprobes = maximum_nprobes
         return self
 
     def distance_range(
@@ -1224,7 +1264,8 @@ class LanceVectorQueryBuilder(LanceQueryBuilder):
             limit=self._limit,
             distance_type=self._distance_type,
             columns=self._columns,
-            nprobes=self._nprobes,
+            minimum_nprobes=self._minimum_nprobes,
+            maximum_nprobes=self._maximum_nprobes,
             lower_bound=self._lower_bound,
             upper_bound=self._upper_bound,
             refine_factor=self._refine_factor,
@@ -1410,10 +1451,13 @@ class LanceFtsQueryBuilder(LanceQueryBuilder):
 
         query = self._query
         if self._phrase_query:
-            raise NotImplementedError(
+            if isinstance(query, str):
-                "Phrase query is not yet supported in Lance FTS. "
+                if not query.startswith('"') or not query.endswith('"'):
-                "Use tantivy-based index instead for now."
+                    query = f'"{query}"'
-            )
+            elif isinstance(query, FullTextQuery) and not isinstance(
+                query, PhraseQuery
+            ):
+                raise TypeError("Please use PhraseQuery for phrase queries.")
         query = self.to_query_object()
         results = self._table._execute_query(query, timeout=timeout)
         results = results.read_all()
@@ -1588,7 +1632,8 @@ class LanceHybridQueryBuilder(LanceQueryBuilder):
         self._fts_columns = fts_columns
         self._norm = None
         self._reranker = None
-        self._nprobes = None
+        self._minimum_nprobes = None
+        self._maximum_nprobes = None
         self._refine_factor = None
         self._distance_type = None
         self._phrase_query = None
@@ -1820,7 +1865,24 @@ class LanceHybridQueryBuilder(LanceQueryBuilder):
         LanceHybridQueryBuilder
             The LanceHybridQueryBuilder object.
         """
-        self._nprobes = nprobes
+        self._minimum_nprobes = nprobes
+        self._maximum_nprobes = nprobes
+        return self
+
+    def minimum_nprobes(self, minimum_nprobes: int) -> LanceHybridQueryBuilder:
+        """Set the minimum number of probes to use.
+
+        See `nprobes` for more details.
+        """
+        self._minimum_nprobes = minimum_nprobes
+        return self
+
+    def maximum_nprobes(self, maximum_nprobes: int) -> LanceHybridQueryBuilder:
+        """Set the maximum number of probes to use.
+
+        See `nprobes` for more details.
+        """
+        self._maximum_nprobes = maximum_nprobes
         return self
 
     def distance_range(
@@ -2049,8 +2111,10 @@ class LanceHybridQueryBuilder(LanceQueryBuilder):
             self._fts_query.phrase_query(True)
         if self._distance_type:
             self._vector_query.metric(self._distance_type)
-        if self._nprobes:
+        if self._minimum_nprobes:
-            self._vector_query.nprobes(self._nprobes)
+            self._vector_query.minimum_nprobes(self._minimum_nprobes)
+        if self._maximum_nprobes is not None:
+            self._vector_query.maximum_nprobes(self._maximum_nprobes)
         if self._refine_factor:
             self._vector_query.refine_factor(self._refine_factor)
         if self._ef:
@@ -2513,7 +2577,7 @@ class AsyncQuery(AsyncQueryBase):
                 self._inner.nearest_to_text({"query": query, "columns": columns})
             )
         # FullTextQuery object
-        return AsyncFTSQuery(self._inner.nearest_to_text({"query": query.to_dict()}))
+        return AsyncFTSQuery(self._inner.nearest_to_text({"query": query}))
 
 
 class AsyncFTSQuery(AsyncQueryBase):
@@ -2661,6 +2725,34 @@ class AsyncVectorQueryBase:
         self._inner.nprobes(nprobes)
         return self
 
+    def minimum_nprobes(self, minimum_nprobes: int) -> Self:
+        """Set the minimum number of probes to use.
+
+        See `nprobes` for more details.
+
+        These partitions will be searched on every indexed vector query and will
+        increase recall at the expense of latency.
+        """
+        self._inner.minimum_nprobes(minimum_nprobes)
+        return self
+
+    def maximum_nprobes(self, maximum_nprobes: int) -> Self:
+        """Set the maximum number of probes to use.
+
+        See `nprobes` for more details.
+
+        If this value is greater than `minimum_nprobes` then the excess partitions
+        will be searched only if we have not found enough results.
+
+        This can be useful when there is a narrow filter to allow these queries to
+        spend more time searching and avoid potential false negatives.
+
+        If this value is 0 then no limit will be applied and all partitions could be
+        searched if needed to satisfy the limit.
+        """
+        self._inner.maximum_nprobes(maximum_nprobes)
+        return self
+
     def distance_range(
         self, lower_bound: Optional[float] = None, upper_bound: Optional[float] = None
     ) -> Self:
@@ -2835,7 +2927,7 @@ class AsyncVectorQuery(AsyncQueryBase, AsyncVectorQueryBase):
                 self._inner.nearest_to_text({"query": query, "columns": columns})
             )
         # FullTextQuery object
-        return AsyncHybridQuery(self._inner.nearest_to_text({"query": query.to_dict()}))
+        return AsyncHybridQuery(self._inner.nearest_to_text({"query": query}))
 
     async def to_batches(
         self,
@@ -2950,15 +3042,21 @@ class AsyncHybridQuery(AsyncQueryBase, AsyncVectorQueryBase):
         >>> asyncio.run(doctest_example()) # doctest: +ELLIPSIS, +NORMALIZE_WHITESPACE
         Vector Search Plan:
         ProjectionExec: expr=[vector@0 as vector, text@3 as text, _distance@2 as _distance]
-            Take: columns="vector, _rowid, _distance, (text)"
+          Take: columns="vector, _rowid, _distance, (text)"
-                CoalesceBatchesExec: target_batch_size=1024
+            CoalesceBatchesExec: target_batch_size=1024
-                GlobalLimitExec: skip=0, fetch=10
+              GlobalLimitExec: skip=0, fetch=10
-                    FilterExec: _distance@2 IS NOT NULL
+                FilterExec: _distance@2 IS NOT NULL
-                    SortExec: TopK(fetch=10), expr=[_distance@2 ASC NULLS LAST], preserve_partitioning=[false]
+                  SortExec: TopK(fetch=10), expr=[_distance@2 ASC NULLS LAST], preserve_partitioning=[false]
-                        KNNVectorDistance: metric=l2
+                    KNNVectorDistance: metric=l2
-                        LanceScan: uri=..., projection=[vector], row_id=true, row_addr=false, ordered=false
+                      LanceScan: uri=..., projection=[vector], row_id=true, row_addr=false, ordered=false
+        <BLANKLINE>
         FTS Search Plan:
-        LanceScan: uri=..., projection=[vector, text], row_id=false, row_addr=false, ordered=true
+        ProjectionExec: expr=[vector@2 as vector, text@3 as text, _score@1 as _score]
+          Take: columns="_rowid, _score, (vector), (text)"
+            CoalesceBatchesExec: target_batch_size=1024
+              GlobalLimitExec: skip=0, fetch=10
+                MatchQuery: query=hello
+        <BLANKLINE>
 
         Parameters
         ----------

python/python/lancedb/remote/table.py

@@ -18,7 +18,7 @@ from lancedb._lancedb import (
     UpdateResult,
 )
 from lancedb.embeddings.base import EmbeddingFunctionConfig
-from lancedb.index import FTS, BTree, Bitmap, HnswPq, HnswSq, IvfFlat, IvfPq, LabelList
+from lancedb.index import FTS, BTree, Bitmap, HnswSq, IvfFlat, IvfPq, LabelList
 from lancedb.remote.db import LOOP
 import pyarrow as pa
 
@@ -149,15 +149,15 @@ class RemoteTable(Table):
         *,
         replace: bool = False,
         wait_timeout: timedelta = None,
-        with_position: bool = True,
+        with_position: bool = False,
         # tokenizer configs:
         base_tokenizer: str = "simple",
         language: str = "English",
         max_token_length: Optional[int] = 40,
         lower_case: bool = True,
-        stem: bool = False,
+        stem: bool = True,
-        remove_stop_words: bool = False,
+        remove_stop_words: bool = True,
-        ascii_folding: bool = False,
+        ascii_folding: bool = True,
     ):
         config = FTS(
             with_position=with_position,
@@ -186,6 +186,8 @@ class RemoteTable(Table):
         accelerator: Optional[str] = None,
         index_type="vector",
         wait_timeout: Optional[timedelta] = None,
+        *,
+        num_bits: int = 8,
     ):
         """Create an index on the table.
         Currently, the only parameters that matter are
@@ -220,11 +222,6 @@ class RemoteTable(Table):
         >>> table.create_index("l2", "vector") # doctest: +SKIP
         """
 
-        if num_partitions is not None:
-            logging.warning(
-                "num_partitions is not supported on LanceDB cloud."
-                "This parameter will be tuned automatically."
-            )
         if num_sub_vectors is not None:
             logging.warning(
                 "num_sub_vectors is not supported on LanceDB cloud."
@@ -244,13 +241,21 @@ class RemoteTable(Table):
 
         index_type = index_type.upper()
         if index_type == "VECTOR" or index_type == "IVF_PQ":
-            config = IvfPq(distance_type=metric)
+            config = IvfPq(
+                distance_type=metric,
+                num_partitions=num_partitions,
+                num_sub_vectors=num_sub_vectors,
+                num_bits=num_bits,
+            )
         elif index_type == "IVF_HNSW_PQ":
-            config = HnswPq(distance_type=metric)
+            raise ValueError(
+                "IVF_HNSW_PQ is not supported on LanceDB cloud."
+                "Please use IVF_HNSW_SQ instead."
+            )
         elif index_type == "IVF_HNSW_SQ":
-            config = HnswSq(distance_type=metric)
+            config = HnswSq(distance_type=metric, num_partitions=num_partitions)
         elif index_type == "IVF_FLAT":
-            config = IvfFlat(distance_type=metric)
+            config = IvfFlat(distance_type=metric, num_partitions=num_partitions)
         else:
             raise ValueError(
                 f"Unknown vector index type: {index_type}. Valid options are"

python/python/lancedb/table.py

@@ -827,17 +827,17 @@ class Table(ABC):
         ordering_field_names: Optional[Union[str, List[str]]] = None,
         replace: bool = False,
         writer_heap_size: Optional[int] = 1024 * 1024 * 1024,
-        use_tantivy: bool = True,
+        use_tantivy: bool = False,
         tokenizer_name: Optional[str] = None,
-        with_position: bool = True,
+        with_position: bool = False,
         # tokenizer configs:
         base_tokenizer: BaseTokenizerType = "simple",
         language: str = "English",
         max_token_length: Optional[int] = 40,
         lower_case: bool = True,
-        stem: bool = False,
+        stem: bool = True,
-        remove_stop_words: bool = False,
+        remove_stop_words: bool = True,
-        ascii_folding: bool = False,
+        ascii_folding: bool = True,
         wait_timeout: Optional[timedelta] = None,
     ):
         """Create a full-text search index on the table.
@@ -864,10 +864,10 @@ class Table(ABC):
             The tokenizer to use for the index. Can be "raw", "default" or the 2 letter
             language code followed by "_stem". So for english it would be "en_stem".
             For available languages see: https://docs.rs/tantivy/latest/tantivy/tokenizer/enum.Language.html
-        use_tantivy: bool, default True
+        use_tantivy: bool, default False
             If True, use the legacy full-text search implementation based on tantivy.
             If False, use the new full-text search implementation based on lance-index.
-        with_position: bool, default True
+        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.
@@ -885,13 +885,13 @@ class Table(ABC):
         lower_case : bool, default True
             Whether to convert the token to lower case. This makes queries
             case-insensitive.
-        stem : bool, default False
+        stem : bool, default True
             Whether to stem the token. Stemming reduces words to their root form.
             For example, in English "running" and "runs" would both be reduced to "run".
-        remove_stop_words : bool, default False
+        remove_stop_words : bool, default True
             Whether to remove stop words. Stop words are common words that are often
             removed from text before indexing. For example, in English "the" and "and".
-        ascii_folding : bool, default False
+        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".
         wait_timeout: timedelta, optional
@@ -1970,17 +1970,17 @@ class LanceTable(Table):
         ordering_field_names: Optional[Union[str, List[str]]] = None,
         replace: bool = False,
         writer_heap_size: Optional[int] = 1024 * 1024 * 1024,
-        use_tantivy: bool = True,
+        use_tantivy: bool = False,
         tokenizer_name: Optional[str] = None,
-        with_position: bool = True,
+        with_position: bool = False,
         # tokenizer configs:
         base_tokenizer: BaseTokenizerType = "simple",
         language: str = "English",
         max_token_length: Optional[int] = 40,
         lower_case: bool = True,
-        stem: bool = False,
+        stem: bool = True,
-        remove_stop_words: bool = False,
+        remove_stop_words: bool = True,
-        ascii_folding: bool = False,
+        ascii_folding: bool = True,
     ):
         if not use_tantivy:
             if not isinstance(field_names, str):
@@ -1990,6 +1990,7 @@ class LanceTable(Table):
                 tokenizer_configs = {
                     "base_tokenizer": base_tokenizer,
                     "language": language,
+                    "with_position": with_position,
                     "max_token_length": max_token_length,
                     "lower_case": lower_case,
                     "stem": stem,
@@ -2000,7 +2001,6 @@ class LanceTable(Table):
                 tokenizer_configs = self.infer_tokenizer_configs(tokenizer_name)
 
             config = FTS(
-                with_position=with_position,
                 **tokenizer_configs,
             )
 
@@ -3637,8 +3637,10 @@ class AsyncTable:
             )
             if query.distance_type is not None:
                 async_query = async_query.distance_type(query.distance_type)
-            if query.nprobes is not None:
+            if query.minimum_nprobes is not None:
-                async_query = async_query.nprobes(query.nprobes)
+                async_query = async_query.minimum_nprobes(query.minimum_nprobes)
+            if query.maximum_nprobes is not None:
+                async_query = async_query.maximum_nprobes(query.maximum_nprobes)
             if query.refine_factor is not None:
                 async_query = async_query.refine_factor(query.refine_factor)
             if query.vector_column:

python/python/tests/docs/test_guide_tables.py

@@ -25,6 +25,10 @@ import numpy as np
 from lancedb.pydantic import Vector, LanceModel
 
 # --8<-- [end:import-lancedb-pydantic]
+# --8<-- [start:import-session-context]
+from datafusion import SessionContext
+
+# --8<-- [end:import-session-context]
 # --8<-- [start:import-datetime]
 from datetime import timedelta
 
@@ -33,6 +37,10 @@ from datetime import timedelta
 from lancedb.embeddings import get_registry
 
 # --8<-- [end:import-embeddings]
+# --8<-- [start:import-ffi-dataset]
+from lance import FFILanceTableProvider
+
+# --8<-- [end:import-ffi-dataset]
 # --8<-- [start:import-pydantic-basemodel]
 from pydantic import BaseModel
 
@@ -341,6 +349,27 @@ def test_table_with_embedding():
     # --8<-- [end:create_table_with_embedding]
 
 
+def test_sql_query():
+    db = lancedb.connect("data/sample-lancedb")
+    data = [
+        {"vector": [1.1, 1.2], "lat": 45.5, "long": -122.7},
+        {"vector": [0.2, 1.8], "lat": 40.1, "long": -74.1},
+    ]
+    table = db.create_table("lance_table", data)
+
+    # --8<-- [start:lance_sql_basic]
+    ctx = SessionContext()
+    ffi_lance_table = FFILanceTableProvider(
+        table.to_lance(), with_row_id=False, with_row_addr=False
+    )
+
+    ctx.register_table_provider("ffi_lance_table", ffi_lance_table)
+    ctx.table("ffi_lance_table")
+
+    ctx.sql("SELECT vector FROM ffi_lance_table")
+    # --8<-- [end:lance_sql_basic]
+
+
 @pytest.mark.skip
 async def test_table_with_embedding_async():
     async_db = await lancedb.connect_async("data/sample-lancedb")

python/python/tests/docs/test_search.py

@@ -6,7 +6,7 @@ import lancedb
 
 # --8<-- [end:import-lancedb]
 # --8<-- [start:import-numpy]
-from lancedb.query import BoostQuery, MatchQuery
+from lancedb.query import BooleanQuery, BoostQuery, MatchQuery, Occur
 import numpy as np
 import pyarrow as pa
 
@@ -156,6 +156,9 @@ async def test_vector_search_async():
     # --8<-- [end:search_result_async_as_list]
 
 
+@pytest.mark.skipif(
+    os.name == "nt", reason="Need to fix https://github.com/lancedb/lance/issues/3905"
+)
 def test_fts_fuzzy_query():
     uri = "data/fuzzy-example"
     db = lancedb.connect(uri)
@@ -188,7 +191,19 @@ def test_fts_fuzzy_query():
         "food",  # 1 insertion
     }
 
+    results = table.search(
+        MatchQuery("foo", "text", fuzziness=1, prefix_length=3)
+    ).to_pandas()
+    assert len(results) == 2
+    assert set(results["text"].to_list()) == {
+        "foo",
+        "food",
+    }
 
+
+@pytest.mark.skipif(
+    os.name == "nt", reason="Need to fix https://github.com/lancedb/lance/issues/3905"
+)
 def test_fts_boost_query():
     uri = "data/boost-example"
     db = lancedb.connect(uri)
@@ -234,6 +249,63 @@ def test_fts_boost_query():
     )
 
 
+@pytest.mark.skipif(
+    os.name == "nt", reason="Need to fix https://github.com/lancedb/lance/issues/3905"
+)
+def test_fts_boolean_query(tmp_path):
+    uri = tmp_path / "boolean-example"
+    db = lancedb.connect(uri)
+    table = db.create_table(
+        "my_table_fts_boolean",
+        data=[
+            {"text": "The cat and dog are playing"},
+            {"text": "The cat is sleeping"},
+            {"text": "The dog is barking"},
+            {"text": "The dog chases the cat"},
+        ],
+        mode="overwrite",
+    )
+    table.create_fts_index("text", use_tantivy=False, replace=True)
+
+    # SHOULD
+    results = table.search(
+        MatchQuery("cat", "text") | MatchQuery("dog", "text")
+    ).to_pandas()
+    assert len(results) == 4
+    assert set(results["text"].to_list()) == {
+        "The cat and dog are playing",
+        "The cat is sleeping",
+        "The dog is barking",
+        "The dog chases the cat",
+    }
+    # MUST
+    results = table.search(
+        MatchQuery("cat", "text") & MatchQuery("dog", "text")
+    ).to_pandas()
+    assert len(results) == 2
+    assert set(results["text"].to_list()) == {
+        "The cat and dog are playing",
+        "The dog chases the cat",
+    }
+
+    # MUST NOT
+    results = table.search(
+        BooleanQuery(
+            [
+                (Occur.MUST, MatchQuery("cat", "text")),
+                (Occur.MUST_NOT, MatchQuery("dog", "text")),
+            ]
+        )
+    ).to_pandas()
+    assert len(results) == 1
+    assert set(results["text"].to_list()) == {
+        "The cat is sleeping",
+    }
+
+
+@pytest.mark.skipif(
+    os.name == "nt", reason="Need to fix https://github.com/lancedb/lance/issues/3905"
+)
 def test_fts_native():
     # --8<-- [start:basic_fts]
     uri = "data/sample-lancedb"
@@ -282,6 +354,9 @@ def test_fts_native():
     # --8<-- [end:fts_incremental_index]
 
 
+@pytest.mark.skipif(
+    os.name == "nt", reason="Need to fix https://github.com/lancedb/lance/issues/3905"
+)
 @pytest.mark.asyncio
 async def test_fts_native_async():
     # --8<-- [start:basic_fts_async]

python/python/tests/test_fts.py

@@ -215,6 +215,19 @@ def test_search_fts(table, use_tantivy):
         assert len(results) == 5
         assert len(results[0]) == 3  # id, text, _score
 
+        # Test boolean query
+        results = (
+            table.search(MatchQuery("puppy", "text") & MatchQuery("runs", "text"))
+            .select(["id", "text"])
+            .limit(5)
+            .to_list()
+        )
+        assert len(results) == 5
+        assert len(results[0]) == 3  # id, text, _score
+        for r in results:
+            assert "puppy" in r["text"]
+            assert "runs" in r["text"]
+
 
 @pytest.mark.asyncio
 async def test_fts_select_async(async_table):
@@ -287,7 +300,7 @@ def test_search_fts_phrase_query(table):
         assert False
     except Exception:
         pass
-    table.create_fts_index("text", use_tantivy=False, replace=True)
+    table.create_fts_index("text", use_tantivy=False, with_position=True, replace=True)
     results = table.search("puppy").limit(100).to_list()
     phrase_results = table.search('"puppy runs"').limit(100).to_list()
     assert len(results) > len(phrase_results)
@@ -312,7 +325,7 @@ async def test_search_fts_phrase_query_async(async_table):
         assert False
     except Exception:
         pass
-    await async_table.create_index("text", config=FTS())
+    await async_table.create_index("text", config=FTS(with_position=True))
     results = await async_table.query().nearest_to_text("puppy").limit(100).to_list()
     phrase_results = (
         await async_table.query().nearest_to_text('"puppy runs"').limit(100).to_list()
@@ -649,7 +662,7 @@ def test_fts_on_list(mem_db: DBConnection):
         }
     )
     table = mem_db.create_table("test", data=data)
-    table.create_fts_index("text", use_tantivy=False)
+    table.create_fts_index("text", use_tantivy=False, with_position=True)
 
     res = table.search("lance").limit(5).to_list()
     assert len(res) == 3

python/python/tests/test_query.py

@@ -25,6 +25,8 @@ from lancedb.query import (
     AsyncQueryBase,
     AsyncVectorQuery,
     LanceVectorQueryBuilder,
+    MatchQuery,
+    PhraseQuery,
     Query,
     FullTextSearchQuery,
 )
@@ -437,6 +439,33 @@ def test_query_builder_with_filter(table):
     assert all(np.array(rs[0]["vector"]) == [3, 4])
 
 
+def test_invalid_nprobes_sync(table):
+    with pytest.raises(ValueError, match="minimum_nprobes must be greater than 0"):
+        LanceVectorQueryBuilder(table, [0, 0], "vector").minimum_nprobes(0).to_list()
+    with pytest.raises(
+        ValueError, match="maximum_nprobes must be greater than minimum_nprobes"
+    ):
+        LanceVectorQueryBuilder(table, [0, 0], "vector").maximum_nprobes(5).to_list()
+    with pytest.raises(
+        ValueError, match="minimum_nprobes must be less or equal to maximum_nprobes"
+    ):
+        LanceVectorQueryBuilder(table, [0, 0], "vector").minimum_nprobes(100).to_list()
+
+
+@pytest.mark.asyncio
+async def test_invalid_nprobes_async(table_async: AsyncTable):
+    with pytest.raises(ValueError, match="minimum_nprobes must be greater than 0"):
+        await table_async.vector_search([0, 0]).minimum_nprobes(0).to_list()
+    with pytest.raises(
+        ValueError, match="maximum_nprobes must be greater than minimum_nprobes"
+    ):
+        await table_async.vector_search([0, 0]).maximum_nprobes(5).to_list()
+    with pytest.raises(
+        ValueError, match="minimum_nprobes must be less or equal to maximum_nprobes"
+    ):
+        await table_async.vector_search([0, 0]).minimum_nprobes(100).to_list()
+
+
 def test_query_builder_with_prefilter(table):
     df = (
         LanceVectorQueryBuilder(table, [0, 0], "vector")
@@ -583,6 +612,21 @@ async def test_query_async(table_async: AsyncTable):
         table_async.query().nearest_to(pa.array([1, 2])).nprobes(10),
         expected_num_rows=2,
     )
+    await check_query(
+        table_async.query().nearest_to(pa.array([1, 2])).minimum_nprobes(10),
+        expected_num_rows=2,
+    )
+    await check_query(
+        table_async.query().nearest_to(pa.array([1, 2])).maximum_nprobes(30),
+        expected_num_rows=2,
+    )
+    await check_query(
+        table_async.query()
+        .nearest_to(pa.array([1, 2]))
+        .minimum_nprobes(10)
+        .maximum_nprobes(20),
+        expected_num_rows=2,
+    )
     await check_query(
         table_async.query().nearest_to(pa.array([1, 2])).bypass_vector_index(),
         expected_num_rows=2,
@@ -731,6 +775,82 @@ async def test_explain_plan_async(table_async: AsyncTable):
     assert "KNN" in plan
 
 
+@pytest.mark.asyncio
+async def test_explain_plan_fts(table_async: AsyncTable):
+    """Test explain plan for FTS queries"""
+    # Create FTS index
+    from lancedb.index import FTS
+
+    await table_async.create_index("text", config=FTS())
+
+    # Test pure FTS query
+    query = await table_async.search("dog", query_type="fts", fts_columns="text")
+    plan = await query.explain_plan()
+    # Should show FTS details (issue #2465 is now fixed)
+    assert "MatchQuery: query=dog" in plan
+    assert "GlobalLimitExec" in plan  # Default limit
+
+    # Test FTS query with limit
+    query_with_limit = await table_async.search(
+        "dog", query_type="fts", fts_columns="text"
+    )
+    plan_with_limit = await query_with_limit.limit(1).explain_plan()
+    assert "MatchQuery: query=dog" in plan_with_limit
+    assert "GlobalLimitExec: skip=0, fetch=1" in plan_with_limit
+
+    # Test FTS query with offset and limit
+    query_with_offset = await table_async.search(
+        "dog", query_type="fts", fts_columns="text"
+    )
+    plan_with_offset = await query_with_offset.offset(1).limit(1).explain_plan()
+    assert "MatchQuery: query=dog" in plan_with_offset
+    assert "GlobalLimitExec: skip=1, fetch=1" in plan_with_offset
+
+
+@pytest.mark.asyncio
+async def test_explain_plan_vector_with_limit_offset(table_async: AsyncTable):
+    """Test explain plan for vector queries with limit and offset"""
+    # Test vector query with limit
+    plan_with_limit = await (
+        table_async.query().nearest_to(pa.array([1, 2])).limit(1).explain_plan()
+    )
+    assert "KNN" in plan_with_limit
+    assert "GlobalLimitExec: skip=0, fetch=1" in plan_with_limit
+
+    # Test vector query with offset and limit
+    plan_with_offset = await (
+        table_async.query()
+        .nearest_to(pa.array([1, 2]))
+        .offset(1)
+        .limit(1)
+        .explain_plan()
+    )
+    assert "KNN" in plan_with_offset
+    assert "GlobalLimitExec: skip=1, fetch=1" in plan_with_offset
+
+
+@pytest.mark.asyncio
+async def test_explain_plan_with_filters(table_async: AsyncTable):
+    """Test explain plan for queries with filters"""
+    # Test vector query with filter
+    plan_with_filter = await (
+        table_async.query().nearest_to(pa.array([1, 2])).where("id = 1").explain_plan()
+    )
+    assert "KNN" in plan_with_filter
+    assert "FilterExec" in plan_with_filter
+
+    # Test FTS query with filter
+    from lancedb.index import FTS
+
+    await table_async.create_index("text", config=FTS())
+    query_fts_filter = await table_async.search(
+        "dog", query_type="fts", fts_columns="text"
+    )
+    plan_fts_filter = await query_fts_filter.where("id = 1").explain_plan()
+    assert "MatchQuery: query=dog" in plan_fts_filter
+    assert "FilterExec: id@" in plan_fts_filter  # Should show filter details
+
+
 @pytest.mark.asyncio
 async def test_query_camelcase_async(tmp_path):
     db = await lancedb.connect_async(tmp_path)
@@ -909,7 +1029,39 @@ def test_query_serialization_sync(table: lancedb.table.Table):
 
     q = table.search([5.0, 6.0]).nprobes(10).refine_factor(5).to_query_object()
     check_set_props(
-        q, vector_column="vector", vector=[5.0, 6.0], nprobes=10, refine_factor=5
+        q,
+        vector_column="vector",
+        vector=[5.0, 6.0],
+        minimum_nprobes=10,
+        maximum_nprobes=10,
+        refine_factor=5,
+    )
+
+    q = table.search([5.0, 6.0]).minimum_nprobes(10).to_query_object()
+    check_set_props(
+        q,
+        vector_column="vector",
+        vector=[5.0, 6.0],
+        minimum_nprobes=10,
+        maximum_nprobes=None,
+    )
+
+    q = table.search([5.0, 6.0]).nprobes(50).to_query_object()
+    check_set_props(
+        q,
+        vector_column="vector",
+        vector=[5.0, 6.0],
+        minimum_nprobes=50,
+        maximum_nprobes=50,
+    )
+
+    q = table.search([5.0, 6.0]).maximum_nprobes(10).to_query_object()
+    check_set_props(
+        q,
+        vector_column="vector",
+        vector=[5.0, 6.0],
+        maximum_nprobes=10,
+        minimum_nprobes=None,
     )
 
     q = table.search([5.0, 6.0]).distance_range(0.0, 1.0).to_query_object()
@@ -961,7 +1113,8 @@ async def test_query_serialization_async(table_async: AsyncTable):
         limit=10,
         vector=sample_vector,
         postfilter=False,
-        nprobes=20,
+        minimum_nprobes=20,
+        maximum_nprobes=20,
         with_row_id=False,
         bypass_vector_index=False,
     )
@@ -971,7 +1124,20 @@ async def test_query_serialization_async(table_async: AsyncTable):
         q,
         vector=sample_vector,
         postfilter=False,
-        nprobes=20,
+        minimum_nprobes=20,
+        maximum_nprobes=20,
+        with_row_id=False,
+        bypass_vector_index=False,
+        limit=10,
+    )
+
+    q = (await table_async.search([5.0, 6.0])).nprobes(50).to_query_object()
+    check_set_props(
+        q,
+        vector=sample_vector,
+        postfilter=False,
+        minimum_nprobes=50,
+        maximum_nprobes=50,
         with_row_id=False,
         bypass_vector_index=False,
         limit=10,
@@ -990,7 +1156,8 @@ async def test_query_serialization_async(table_async: AsyncTable):
         filter="id = 1",
         postfilter=True,
         vector=sample_vector,
-        nprobes=20,
+        minimum_nprobes=20,
+        maximum_nprobes=20,
         with_row_id=False,
         bypass_vector_index=False,
     )
@@ -1004,7 +1171,8 @@ async def test_query_serialization_async(table_async: AsyncTable):
     check_set_props(
         q,
         vector=sample_vector,
-        nprobes=10,
+        minimum_nprobes=10,
+        maximum_nprobes=10,
         refine_factor=5,
         postfilter=False,
         with_row_id=False,
@@ -1012,6 +1180,18 @@ async def test_query_serialization_async(table_async: AsyncTable):
         limit=10,
     )
 
+    q = (await table_async.search([5.0, 6.0])).minimum_nprobes(5).to_query_object()
+    check_set_props(
+        q,
+        vector=sample_vector,
+        minimum_nprobes=5,
+        maximum_nprobes=20,
+        postfilter=False,
+        with_row_id=False,
+        bypass_vector_index=False,
+        limit=10,
+    )
+
     q = (
         (await table_async.search([5.0, 6.0]))
         .distance_range(0.0, 1.0)
@@ -1023,7 +1203,8 @@ async def test_query_serialization_async(table_async: AsyncTable):
         lower_bound=0.0,
         upper_bound=1.0,
         postfilter=False,
-        nprobes=20,
+        minimum_nprobes=20,
+        maximum_nprobes=20,
         with_row_id=False,
         bypass_vector_index=False,
         limit=10,
@@ -1035,7 +1216,8 @@ async def test_query_serialization_async(table_async: AsyncTable):
         distance_type="cosine",
         vector=sample_vector,
         postfilter=False,
-        nprobes=20,
+        minimum_nprobes=20,
+        maximum_nprobes=20,
         with_row_id=False,
         bypass_vector_index=False,
         limit=10,
@@ -1047,7 +1229,8 @@ async def test_query_serialization_async(table_async: AsyncTable):
         ef=7,
         vector=sample_vector,
         postfilter=False,
-        nprobes=20,
+        minimum_nprobes=20,
+        maximum_nprobes=20,
         with_row_id=False,
         bypass_vector_index=False,
         limit=10,
@@ -1059,24 +1242,34 @@ async def test_query_serialization_async(table_async: AsyncTable):
         bypass_vector_index=True,
         vector=sample_vector,
         postfilter=False,
-        nprobes=20,
+        minimum_nprobes=20,
+        maximum_nprobes=20,
         with_row_id=False,
         limit=10,
     )
 
     # FTS queries
-    q = (await table_async.search("foo")).limit(10).to_query_object()
+    match_query = MatchQuery("foo", "text")
+    q = (await table_async.search(match_query)).limit(10).to_query_object()
     check_set_props(
         q,
         limit=10,
-        full_text_query=FullTextSearchQuery(columns=[], query="foo"),
+        full_text_query=FullTextSearchQuery(columns=None, query=match_query),
         with_row_id=False,
     )
 
-    q = (await table_async.search("foo", query_type="fts")).to_query_object()
+    q = (await table_async.search(match_query)).to_query_object()
     check_set_props(
         q,
-        full_text_query=FullTextSearchQuery(columns=[], query="foo"),
+        full_text_query=FullTextSearchQuery(columns=None, query=match_query),
+        with_row_id=False,
+    )
+
+    phrase_query = PhraseQuery("foo", "text", slop=1)
+    q = (await table_async.search(phrase_query)).to_query_object()
+    check_set_props(
+        q,
+        full_text_query=FullTextSearchQuery(columns=None, query=phrase_query),
         with_row_id=False,
     )
 

python/python/tests/test_remote_db.py

@@ -496,6 +496,8 @@ def test_query_sync_minimal():
             "ef": None,
             "vector": [1.0, 2.0, 3.0],
             "nprobes": 20,
+            "minimum_nprobes": 20,
+            "maximum_nprobes": 20,
             "version": None,
         }
 
@@ -536,6 +538,8 @@ def test_query_sync_maximal():
             "refine_factor": 10,
             "vector": [1.0, 2.0, 3.0],
             "nprobes": 5,
+            "minimum_nprobes": 5,
+            "maximum_nprobes": 5,
             "lower_bound": None,
             "upper_bound": None,
             "ef": None,
@@ -564,6 +568,66 @@ def test_query_sync_maximal():
         )
 
 
+def test_query_sync_nprobes():
+    def handler(body):
+        assert body == {
+            "distance_type": "l2",
+            "k": 10,
+            "prefilter": True,
+            "fast_search": True,
+            "vector_column": "vector2",
+            "refine_factor": None,
+            "lower_bound": None,
+            "upper_bound": None,
+            "ef": None,
+            "vector": [1.0, 2.0, 3.0],
+            "nprobes": 5,
+            "minimum_nprobes": 5,
+            "maximum_nprobes": 15,
+            "version": None,
+        }
+
+        return pa.table({"id": [1, 2, 3], "name": ["a", "b", "c"]})
+
+    with query_test_table(handler) as table:
+        (
+            table.search([1, 2, 3], vector_column_name="vector2", fast_search=True)
+            .minimum_nprobes(5)
+            .maximum_nprobes(15)
+            .to_list()
+        )
+
+
+def test_query_sync_no_max_nprobes():
+    def handler(body):
+        assert body == {
+            "distance_type": "l2",
+            "k": 10,
+            "prefilter": True,
+            "fast_search": True,
+            "vector_column": "vector2",
+            "refine_factor": None,
+            "lower_bound": None,
+            "upper_bound": None,
+            "ef": None,
+            "vector": [1.0, 2.0, 3.0],
+            "nprobes": 5,
+            "minimum_nprobes": 5,
+            "maximum_nprobes": 0,
+            "version": None,
+        }
+
+        return pa.table({"id": [1, 2, 3], "name": ["a", "b", "c"]})
+
+    with query_test_table(handler) as table:
+        (
+            table.search([1, 2, 3], vector_column_name="vector2", fast_search=True)
+            .minimum_nprobes(5)
+            .maximum_nprobes(0)
+            .to_list()
+        )
+
+
 @pytest.mark.parametrize("server_version", [Version("0.1.0"), Version("0.2.0")])
 def test_query_sync_batch_queries(server_version):
     def handler(body):
@@ -666,6 +730,8 @@ def test_query_sync_hybrid():
                 "refine_factor": None,
                 "vector": [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0],
                 "nprobes": 20,
+                "minimum_nprobes": 20,
+                "maximum_nprobes": 20,
                 "lower_bound": None,
                 "upper_bound": None,
                 "ef": None,

python/python/tests/test_s3.py

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

python/src/index.rs

@@ -3,7 +3,7 @@
 
 use lancedb::index::vector::IvfFlatIndexBuilder;
 use lancedb::index::{
-    scalar::{BTreeIndexBuilder, FtsIndexBuilder, TokenizerConfig},
+    scalar::{BTreeIndexBuilder, FtsIndexBuilder},
     vector::{IvfHnswPqIndexBuilder, IvfHnswSqIndexBuilder, IvfPqIndexBuilder},
     Index as LanceDbIndex,
 };
@@ -38,19 +38,17 @@ pub fn extract_index_params(source: &Option<Bound<'_, PyAny>>) -> PyResult<Lance
             "LabelList" => Ok(LanceDbIndex::LabelList(Default::default())),
             "FTS" => {
                 let params = source.extract::<FtsParams>()?;
-                let inner_opts = TokenizerConfig::default()
+                let inner_opts = FtsIndexBuilder::default()
                     .base_tokenizer(params.base_tokenizer)
                     .language(&params.language)
                     .map_err(|_| PyValueError::new_err(format!("LanceDB does not support the requested language: '{}'", params.language)))?
+                    .with_position(params.with_position)
                     .lower_case(params.lower_case)
                     .max_token_length(params.max_token_length)
                     .remove_stop_words(params.remove_stop_words)
                     .stem(params.stem)
                     .ascii_folding(params.ascii_folding);
-                let mut opts = FtsIndexBuilder::default()
+                Ok(LanceDbIndex::FTS(inner_opts))
-                    .with_position(params.with_position);
-                opts.tokenizer_configs = inner_opts;
-                Ok(LanceDbIndex::FTS(opts))
             },
             "IvfFlat" => {
                 let params = source.extract::<IvfFlatParams>()?;

python/src/query.rs

@@ -9,15 +9,16 @@ use arrow::array::Array;
 use arrow::array::ArrayData;
 use arrow::pyarrow::FromPyArrow;
 use arrow::pyarrow::IntoPyArrow;
-use lancedb::index::scalar::{FtsQuery, FullTextSearchQuery, MatchQuery, PhraseQuery};
+use lancedb::index::scalar::{
+    BooleanQuery, BoostQuery, FtsQuery, FullTextSearchQuery, MatchQuery, MultiMatchQuery, Occur,
+    Operator, PhraseQuery,
+};
 use lancedb::query::QueryExecutionOptions;
 use lancedb::query::QueryFilter;
 use lancedb::query::{
     ExecutableQuery, Query as LanceDbQuery, QueryBase, Select, VectorQuery as LanceDbVectorQuery,
 };
 use lancedb::table::AnyQuery;
-use pyo3::exceptions::PyRuntimeError;
-use pyo3::exceptions::{PyNotImplementedError, PyValueError};
 use pyo3::prelude::{PyAnyMethods, PyDictMethods};
 use pyo3::pymethods;
 use pyo3::types::PyList;
@@ -27,30 +28,172 @@ use pyo3::IntoPyObject;
 use pyo3::PyAny;
 use pyo3::PyRef;
 use pyo3::PyResult;
+use pyo3::{exceptions::PyRuntimeError, FromPyObject};
+use pyo3::{
+    exceptions::{PyNotImplementedError, PyValueError},
+    intern,
+};
 use pyo3::{pyclass, PyErr};
 use pyo3_async_runtimes::tokio::future_into_py;
 
-use crate::arrow::RecordBatchStream;
+use crate::util::parse_distance_type;
-use crate::error::PythonErrorExt;
+use crate::{arrow::RecordBatchStream, util::PyLanceDB};
-use crate::util::{parse_distance_type, parse_fts_query};
+use crate::{error::PythonErrorExt, index::class_name};
 
-// Python representation of full text search parameters
+impl FromPyObject<'_> for PyLanceDB<FtsQuery> {
-#[derive(Clone)]
+    fn extract_bound(ob: &Bound<'_, PyAny>) -> PyResult<Self> {
-#[pyclass(get_all)]
+        match class_name(ob)?.as_str() {
-pub struct PyFullTextSearchQuery {
+            "MatchQuery" => {
-    pub columns: Vec<String>,
+                let query = ob.getattr("query")?.extract()?;
-    pub query: String,
+                let column = ob.getattr("column")?.extract()?;
-    pub limit: Option<i64>,
+                let boost = ob.getattr("boost")?.extract()?;
-    pub wand_factor: Option<f32>,
+                let fuzziness = ob.getattr("fuzziness")?.extract()?;
+                let max_expansions = ob.getattr("max_expansions")?.extract()?;
+                let operator = ob.getattr("operator")?.extract::<String>()?;
+                let prefix_length = ob.getattr("prefix_length")?.extract()?;
+
+                Ok(Self(
+                    MatchQuery::new(query)
+                        .with_column(Some(column))
+                        .with_boost(boost)
+                        .with_fuzziness(fuzziness)
+                        .with_max_expansions(max_expansions)
+                        .with_operator(Operator::try_from(operator.as_str()).map_err(|e| {
+                            PyValueError::new_err(format!("Invalid operator: {}", e))
+                        })?)
+                        .with_prefix_length(prefix_length)
+                        .into(),
+                ))
+            }
+            "PhraseQuery" => {
+                let query = ob.getattr("query")?.extract()?;
+                let column = ob.getattr("column")?.extract()?;
+                let slop = ob.getattr("slop")?.extract()?;
+
+                Ok(Self(
+                    PhraseQuery::new(query)
+                        .with_column(Some(column))
+                        .with_slop(slop)
+                        .into(),
+                ))
+            }
+            "BoostQuery" => {
+                let positive: Self = ob.getattr("positive")?.extract()?;
+                let negative: Self = ob.getattr("negative")?.extract()?;
+                let negative_boost = ob.getattr("negative_boost")?.extract()?;
+                Ok(Self(
+                    BoostQuery::new(positive.0, negative.0, negative_boost).into(),
+                ))
+            }
+            "MultiMatchQuery" => {
+                let query = ob.getattr("query")?.extract()?;
+                let columns = ob.getattr("columns")?.extract()?;
+                let boosts: Option<Vec<f32>> = ob.getattr("boosts")?.extract()?;
+                let operator: String = ob.getattr("operator")?.extract()?;
+
+                let q = MultiMatchQuery::try_new(query, columns)
+                    .map_err(|e| PyValueError::new_err(format!("Invalid query: {}", e)))?;
+                let q = if let Some(boosts) = boosts {
+                    q.try_with_boosts(boosts)
+                        .map_err(|e| PyValueError::new_err(format!("Invalid boosts: {}", e)))?
+                } else {
+                    q
+                };
+
+                let op = Operator::try_from(operator.as_str())
+                    .map_err(|e| PyValueError::new_err(format!("Invalid operator: {}", e)))?;
+
+                Ok(Self(q.with_operator(op).into()))
+            }
+            "BooleanQuery" => {
+                let queries: Vec<(String, Self)> = ob.getattr("queries")?.extract()?;
+                let mut sub_queries = Vec::with_capacity(queries.len());
+                for (occur, q) in queries {
+                    let occur = Occur::try_from(occur.as_str())
+                        .map_err(|e| PyValueError::new_err(e.to_string()))?;
+                    sub_queries.push((occur, q.0));
+                }
+                Ok(Self(BooleanQuery::new(sub_queries).into()))
+            }
+            name => Err(PyValueError::new_err(format!(
+                "Unsupported FTS query type: {}",
+                name
+            ))),
+        }
+    }
 }
 
-impl From<FullTextSearchQuery> for PyFullTextSearchQuery {
+impl<'py> IntoPyObject<'py> for PyLanceDB<FtsQuery> {
-    fn from(query: FullTextSearchQuery) -> Self {
+    type Target = PyAny;
-        Self {
+    type Output = Bound<'py, Self::Target>;
-            columns: query.columns().into_iter().collect(),
+    type Error = PyErr;
-            query: query.query.query().to_owned(),
+
-            limit: query.limit,
+    fn into_pyobject(self, py: pyo3::Python<'py>) -> PyResult<Self::Output> {
-            wand_factor: query.wand_factor,
+        let namespace = py
+            .import(intern!(py, "lancedb"))
+            .and_then(|m| m.getattr(intern!(py, "query")))
+            .expect("Failed to import namespace");
+
+        match self.0 {
+            FtsQuery::Match(query) => {
+                let kwargs = PyDict::new(py);
+                kwargs.set_item("boost", query.boost)?;
+                kwargs.set_item("fuzziness", query.fuzziness)?;
+                kwargs.set_item("max_expansions", query.max_expansions)?;
+                kwargs.set_item::<_, &str>("operator", query.operator.into())?;
+                kwargs.set_item("prefix_length", query.prefix_length)?;
+                namespace
+                    .getattr(intern!(py, "MatchQuery"))?
+                    .call((query.terms, query.column.unwrap()), Some(&kwargs))
+            }
+            FtsQuery::Phrase(query) => {
+                let kwargs = PyDict::new(py);
+                kwargs.set_item("slop", query.slop)?;
+                namespace
+                    .getattr(intern!(py, "PhraseQuery"))?
+                    .call((query.terms, query.column.unwrap()), Some(&kwargs))
+            }
+            FtsQuery::Boost(query) => {
+                let positive = Self(query.positive.as_ref().clone()).into_pyobject(py)?;
+                let negative = Self(query.negative.as_ref().clone()).into_pyobject(py)?;
+                let kwargs = PyDict::new(py);
+                kwargs.set_item("negative_boost", query.negative_boost)?;
+                namespace
+                    .getattr(intern!(py, "BoostQuery"))?
+                    .call((positive, negative), Some(&kwargs))
+            }
+            FtsQuery::MultiMatch(query) => {
+                let first = &query.match_queries[0];
+                let (columns, boosts): (Vec<_>, Vec<_>) = query
+                    .match_queries
+                    .iter()
+                    .map(|q| (q.column.as_ref().unwrap().clone(), q.boost))
+                    .unzip();
+                let kwargs = PyDict::new(py);
+                kwargs.set_item("boosts", boosts)?;
+                kwargs.set_item::<_, &str>("operator", first.operator.into())?;
+                namespace
+                    .getattr(intern!(py, "MultiMatchQuery"))?
+                    .call((first.terms.clone(), columns), Some(&kwargs))
+            }
+            FtsQuery::Boolean(query) => {
+                let mut queries: Vec<(&str, Bound<'py, PyAny>)> = Vec::with_capacity(
+                    query.should.len() + query.must.len() + query.must_not.len(),
+                );
+                for q in query.should {
+                    queries.push((Occur::Should.into(), Self(q).into_pyobject(py)?));
+                }
+                for q in query.must {
+                    queries.push((Occur::Must.into(), Self(q).into_pyobject(py)?));
+                }
+                for q in query.must_not {
+                    queries.push((Occur::MustNot.into(), Self(q).into_pyobject(py)?));
+                }
+
+                namespace
+                    .getattr(intern!(py, "BooleanQuery"))?
+                    .call1((queries,))
+            }
         }
     }
 }
@@ -80,13 +223,16 @@ pub struct PyQueryRequest {
     pub limit: Option<usize>,
     pub offset: Option<usize>,
     pub filter: Option<PyQueryFilter>,
-    pub full_text_search: Option<PyFullTextSearchQuery>,
+    pub full_text_search: Option<PyLanceDB<FtsQuery>>,
     pub select: PySelect,
     pub fast_search: Option<bool>,
     pub with_row_id: Option<bool>,
     pub column: Option<String>,
     pub query_vector: Option<PyQueryVectors>,
-    pub nprobes: Option<usize>,
+    pub minimum_nprobes: Option<usize>,
+    // None means user did not set it and default shoud be used (currenty 20)
+    // Some(0) means user set it to None and there is no limit
+    pub maximum_nprobes: Option<usize>,
     pub lower_bound: Option<f32>,
     pub upper_bound: Option<f32>,
     pub ef: Option<usize>,
@@ -106,13 +252,14 @@ impl From<AnyQuery> for PyQueryRequest {
                 filter: query_request.filter.map(PyQueryFilter),
                 full_text_search: query_request
                     .full_text_search
-                    .map(PyFullTextSearchQuery::from),
+                    .map(|fts| PyLanceDB(fts.query)),
                 select: PySelect(query_request.select),
                 fast_search: Some(query_request.fast_search),
                 with_row_id: Some(query_request.with_row_id),
                 column: None,
                 query_vector: None,
-                nprobes: None,
+                minimum_nprobes: None,
+                maximum_nprobes: None,
                 lower_bound: None,
                 upper_bound: None,
                 ef: None,
@@ -132,7 +279,11 @@ impl From<AnyQuery> for PyQueryRequest {
                 with_row_id: Some(vector_query.base.with_row_id),
                 column: vector_query.column,
                 query_vector: Some(PyQueryVectors(vector_query.query_vector)),
-                nprobes: Some(vector_query.nprobes),
+                minimum_nprobes: Some(vector_query.minimum_nprobes),
+                maximum_nprobes: match vector_query.maximum_nprobes {
+                    None => Some(0),
+                    Some(value) => Some(value),
+                },
                 lower_bound: vector_query.lower_bound,
                 upper_bound: vector_query.upper_bound,
                 ef: vector_query.ef,
@@ -269,8 +420,8 @@ impl Query {
                 }
             };
             let mut query = FullTextSearchQuery::new_query(query);
-            if let Some(cols) = columns {
+            match columns {
-                if !cols.is_empty() {
+                Some(cols) if !cols.is_empty() => {
                     query = query.with_columns(&cols).map_err(|e| {
                         PyValueError::new_err(format!(
                             "Failed to set full text search columns: {}",
@@ -278,15 +429,12 @@ impl Query {
                         ))
                     })?;
                 }
+                _ => {}
             }
             query
-        } else if let Ok(query) = fts_query.downcast::<PyDict>() {
-            let query = parse_fts_query(query)?;
-            FullTextSearchQuery::new_query(query)
         } else {
-            return Err(PyValueError::new_err(
+            let query = fts_query.extract::<PyLanceDB<FtsQuery>>()?;
-                "query must be a string or a Query object",
+            FullTextSearchQuery::new_query(query.0)
-            ));
         };
 
         Ok(FTSQuery {
@@ -414,7 +562,10 @@ impl FTSQuery {
     }
 
     pub fn explain_plan(self_: PyRef<'_, Self>, verbose: bool) -> PyResult<Bound<'_, PyAny>> {
-        let inner = self_.inner.clone();
+        let inner = self_
+            .inner
+            .clone()
+            .full_text_search(self_.fts_query.clone());
         future_into_py(self_.py(), async move {
             inner
                 .explain_plan(verbose)
@@ -424,7 +575,10 @@ impl FTSQuery {
     }
 
     pub fn analyze_plan(self_: PyRef<'_, Self>) -> PyResult<Bound<'_, PyAny>> {
-        let inner = self_.inner.clone();
+        let inner = self_
+            .inner
+            .clone()
+            .full_text_search(self_.fts_query.clone());
         future_into_py(self_.py(), async move {
             inner
                 .analyze_plan()
@@ -509,6 +663,29 @@ impl VectorQuery {
         self.inner = self.inner.clone().nprobes(nprobe as usize);
     }
 
+    pub fn minimum_nprobes(&mut self, minimum_nprobes: u32) -> PyResult<()> {
+        self.inner = self
+            .inner
+            .clone()
+            .minimum_nprobes(minimum_nprobes as usize)
+            .infer_error()?;
+        Ok(())
+    }
+
+    pub fn maximum_nprobes(&mut self, maximum_nprobes: u32) -> PyResult<()> {
+        let maximum_nprobes = if maximum_nprobes == 0 {
+            None
+        } else {
+            Some(maximum_nprobes as usize)
+        };
+        self.inner = self
+            .inner
+            .clone()
+            .maximum_nprobes(maximum_nprobes)
+            .infer_error()?;
+        Ok(())
+    }
+
     #[pyo3(signature = (lower_bound=None, upper_bound=None))]
     pub fn distance_range(&mut self, lower_bound: Option<f32>, upper_bound: Option<f32>) {
         self.inner = self.inner.clone().distance_range(lower_bound, upper_bound);

python/src/util.rs

@@ -3,15 +3,11 @@
 
 use std::sync::Mutex;
 
-use lancedb::index::scalar::{BoostQuery, FtsQuery, MatchQuery, MultiMatchQuery, PhraseQuery};
 use lancedb::DistanceType;
-use pyo3::prelude::{PyAnyMethods, PyDictMethods, PyListMethods};
-use pyo3::types::PyDict;
 use pyo3::{
     exceptions::{PyRuntimeError, PyValueError},
     pyfunction, PyResult,
 };
-use pyo3::{Bound, PyAny};
 
 /// A wrapper around a rust builder
 ///
@@ -64,116 +60,6 @@ pub fn validate_table_name(table_name: &str) -> PyResult<()> {
         .map_err(|e| PyValueError::new_err(e.to_string()))
 }
 
-pub fn parse_fts_query(query: &Bound<'_, PyDict>) -> PyResult<FtsQuery> {
+/// A wrapper around a LanceDB type to allow it to be used in Python
-    let query_type = query.keys().get_item(0)?.extract::<String>()?;
+#[derive(Debug, Clone)]
-    let query_value = query
+pub struct PyLanceDB<T>(pub T);
-        .get_item(&query_type)?
-        .ok_or(PyValueError::new_err(format!(
-            "Query type {} not found",
-            query_type
-        )))?;
-    let query_value = query_value.downcast::<PyDict>()?;
-
-    match query_type.as_str() {
-        "match" => {
-            let column = query_value.keys().get_item(0)?.extract::<String>()?;
-            let params = query_value
-                .get_item(&column)?
-                .ok_or(PyValueError::new_err(format!(
-                    "column {} not found",
-                    column
-                )))?;
-            let params = params.downcast::<PyDict>()?;
-
-            let query = params
-                .get_item("query")?
-                .ok_or(PyValueError::new_err("query not found"))?
-                .extract::<String>()?;
-            let boost = params
-                .get_item("boost")?
-                .ok_or(PyValueError::new_err("boost not found"))?
-                .extract::<f32>()?;
-            let fuzziness = params
-                .get_item("fuzziness")?
-                .ok_or(PyValueError::new_err("fuzziness not found"))?
-                .extract::<Option<u32>>()?;
-            let max_expansions = params
-                .get_item("max_expansions")?
-                .ok_or(PyValueError::new_err("max_expansions not found"))?
-                .extract::<usize>()?;
-
-            let query = MatchQuery::new(query)
-                .with_column(Some(column))
-                .with_boost(boost)
-                .with_fuzziness(fuzziness)
-                .with_max_expansions(max_expansions);
-            Ok(query.into())
-        }
-
-        "match_phrase" => {
-            let column = query_value.keys().get_item(0)?.extract::<String>()?;
-            let query = query_value
-                .get_item(&column)?
-                .ok_or(PyValueError::new_err(format!(
-                    "column {} not found",
-                    column
-                )))?
-                .extract::<String>()?;
-
-            let query = PhraseQuery::new(query).with_column(Some(column));
-            Ok(query.into())
-        }
-
-        "boost" => {
-            let positive: Bound<'_, PyAny> = query_value
-                .get_item("positive")?
-                .ok_or(PyValueError::new_err("positive not found"))?;
-            let positive = positive.downcast::<PyDict>()?;
-
-            let negative = query_value
-                .get_item("negative")?
-                .ok_or(PyValueError::new_err("negative not found"))?;
-            let negative = negative.downcast::<PyDict>()?;
-
-            let negative_boost = query_value
-                .get_item("negative_boost")?
-                .ok_or(PyValueError::new_err("negative_boost not found"))?
-                .extract::<f32>()?;
-
-            let positive_query = parse_fts_query(positive)?;
-            let negative_query = parse_fts_query(negative)?;
-            let query = BoostQuery::new(positive_query, negative_query, Some(negative_boost));
-
-            Ok(query.into())
-        }
-
-        "multi_match" => {
-            let query = query_value
-                .get_item("query")?
-                .ok_or(PyValueError::new_err("query not found"))?
-                .extract::<String>()?;
-
-            let columns = query_value
-                .get_item("columns")?
-                .ok_or(PyValueError::new_err("columns not found"))?
-                .extract::<Vec<String>>()?;
-
-            let boost = query_value
-                .get_item("boost")?
-                .ok_or(PyValueError::new_err("boost not found"))?
-                .extract::<Vec<f32>>()?;
-
-            let query = MultiMatchQuery::try_new(query, columns)
-                .and_then(|q| q.try_with_boosts(boost))
-                .map_err(|e| {
-                    PyValueError::new_err(format!("Error creating MultiMatchQuery: {}", e))
-                })?;
-            Ok(query.into())
-        }
-
-        _ => Err(PyValueError::new_err(format!(
-            "Unsupported query type: {}",
-            query_type
-        ))),
-    }
-}

rust/ffi/node/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb-node"
-version = "0.19.1-beta.5"
+version = "0.21.1-beta.0"
 description = "Serverless, low-latency vector database for AI applications"
 license.workspace = true
 edition.workspace = true

rust/lancedb/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb"
-version = "0.19.1-beta.5"
+version = "0.21.1-beta.0"
 edition.workspace = true
 description = "LanceDB: A serverless, low-latency vector database for AI applications"
 license.workspace = true
@@ -60,15 +60,15 @@ reqwest = { version = "0.12.0", default-features = false, features = [
     "macos-system-configuration",
     "stream",
 ], optional = true }
-rand = { version = "0.8.3", features = ["small_rng"], optional = true }
+rand = { version = "0.9", features = ["small_rng"], optional = true }
 http = { version = "1", optional = true } # Matching what is in reqwest
 uuid = { version = "1.7.0", features = ["v4"], optional = true }
 polars-arrow = { version = ">=0.37,<0.40.0", optional = true }
 polars = { version = ">=0.37,<0.40.0", optional = true }
 hf-hub = { version = "0.4.1", optional = true, default-features = false, features = ["rustls-tls", "tokio", "ureq"]}
-candle-core = { version = "0.6.0", optional = true }
+candle-core = { version = "0.9.1", optional = true }
-candle-transformers = { version = "0.6.0", optional = true }
+candle-transformers = { version = "0.9.1", optional = true }
-candle-nn = { version = "0.6.0", optional = true }
+candle-nn = { version = "0.9.1", optional = true }
 tokenizers = { version = "0.19.1", optional = true }
 semver = { workspace = true }
 
@@ -78,7 +78,7 @@ bytemuck_derive.workspace = true
 
 [dev-dependencies]
 tempfile = "3.5.0"
-rand = { version = "0.8.3", features = ["small_rng"] }
+rand = { version = "0.9", features = ["small_rng"] }
 random_word = { version = "0.4.3", features = ["en"] }
 uuid = { version = "1.7.0", features = ["v4"] }
 walkdir = "2"

rust/lancedb/examples/full_text_search.rs

@@ -51,7 +51,7 @@ fn create_some_records() -> Result<Box<dyn RecordBatchReader + Send>> {
                 Arc::new(Int32Array::from_iter_values(0..TOTAL as i32)),
                 Arc::new(StringArray::from_iter_values((0..TOTAL).map(|_| {
                     (0..n_terms)
-                        .map(|_| words[random::<usize>() % words.len()])
+                        .map(|_| words[random::<u32>() as usize % words.len()])
                         .collect::<Vec<_>>()
                         .join(" ")
                 }))),

rust/lancedb/src/catalog/listing.rs

@@ -105,7 +105,7 @@ impl ListingCatalog {
     }
 
     async fn open_path(path: &str) -> Result<Self> {
-        let (object_store, base_path) = ObjectStore::from_uri(path).await.unwrap();
+        let (object_store, base_path) = ObjectStore::from_uri(path).await?;
         if object_store.is_local() {
             Self::try_create_dir(path).context(CreateDirSnafu { path })?;
         }

rust/lancedb/src/embeddings/sentence_transformers.rs

@@ -214,7 +214,7 @@ impl SentenceTransformersEmbeddings {
 
         let embeddings = self
             .model
-            .forward(&input_ids, &token_type_ids)
+            .forward(&input_ids, &token_type_ids, None)
             // TODO: it'd be nice to support other devices
             .and_then(|output| output.to_device(&Device::Cpu))?;
 
@@ -310,7 +310,7 @@ impl SentenceTransformersEmbeddings {
         let embeddings = Tensor::stack(&tokens, 0)
             .and_then(|tokens| {
                 let token_type_ids = tokens.zeros_like()?;
-                self.model.forward(&tokens, &token_type_ids)
+                self.model.forward(&tokens, &token_type_ids, None)
             })
             // TODO: it'd be nice to support other devices
             .and_then(|tokens| tokens.to_device(&Device::Cpu))

rust/lancedb/src/index/scalar.rs

@@ -51,35 +51,7 @@ pub struct BitmapIndexBuilder {}
 #[derive(Debug, Clone, Default)]
 pub struct LabelListIndexBuilder {}
 
-/// Builder for a full text search index
-///
-/// A full text search index is an index on a string column that allows for full text search
-#[derive(Debug, Clone)]
-pub struct FtsIndexBuilder {
-    /// Whether to store the position of the tokens
-    /// This is used for phrase queries
-    pub with_position: bool,
-
-    pub tokenizer_configs: TokenizerConfig,
-}
-
-impl Default for FtsIndexBuilder {
-    fn default() -> Self {
-        Self {
-            with_position: true,
-            tokenizer_configs: TokenizerConfig::default(),
-        }
-    }
-}
-
-impl FtsIndexBuilder {
-    /// Set the with_position flag
-    pub fn with_position(mut self, with_position: bool) -> Self {
-        self.with_position = with_position;
-        self
-    }
-}
-
 pub use lance_index::scalar::inverted::query::*;
-pub use lance_index::scalar::inverted::TokenizerConfig;
 pub use lance_index::scalar::FullTextSearchQuery;
+pub use lance_index::scalar::InvertedIndexParams as FtsIndexBuilder;
+pub use lance_index::scalar::InvertedIndexParams;

rust/lancedb/src/io/object_store.rs

@@ -107,7 +107,7 @@ impl ObjectStore for MirroringObjectStore {
         self.primary.delete(location).await
     }
 
-    fn list(&self, prefix: Option<&Path>) -> BoxStream<'_, Result<ObjectMeta>> {
+    fn list(&self, prefix: Option<&Path>) -> BoxStream<'static, Result<ObjectMeta>> {
         self.primary.list(prefix)
     }
 
@@ -197,16 +197,8 @@ mod test {
 
     #[tokio::test]
     async fn test_e2e() {
-        let dir1 = tempfile::tempdir()
+        let dir1 = tempfile::tempdir().unwrap().keep().canonicalize().unwrap();
-            .unwrap()
+        let dir2 = tempfile::tempdir().unwrap().keep().canonicalize().unwrap();
-            .into_path()
-            .canonicalize()
-            .unwrap();
-        let dir2 = tempfile::tempdir()
-            .unwrap()
-            .into_path()
-            .canonicalize()
-            .unwrap();
 
         let secondary_store = LocalFileSystem::new_with_prefix(dir2.to_str().unwrap()).unwrap();
         let object_store_wrapper = Arc::new(MirroringObjectStoreWrapper {

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

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

rust/lancedb/src/query.rs

@@ -796,8 +796,10 @@ pub struct VectorQueryRequest {
     pub column: Option<String>,
     /// The vector(s) to search for
     pub query_vector: Vec<Arc<dyn Array>>,
-    /// The number of partitions to search
+    /// The minimum number of partitions to search
-    pub nprobes: usize,
+    pub minimum_nprobes: usize,
+    /// The maximum number of partitions to search
+    pub maximum_nprobes: Option<usize>,
     /// The lower bound (inclusive) of the distance to search for.
     pub lower_bound: Option<f32>,
     /// The upper bound (exclusive) of the distance to search for.
@@ -819,7 +821,8 @@ impl Default for VectorQueryRequest {
             base: QueryRequest::default(),
             column: None,
             query_vector: Vec::new(),
-            nprobes: 20,
+            minimum_nprobes: 20,
+            maximum_nprobes: Some(20),
             lower_bound: None,
             upper_bound: None,
             ef: None,
@@ -925,11 +928,75 @@ impl VectorQuery {
     /// For best results we recommend tuning this parameter with a benchmark against
     /// your actual data to find the smallest possible value that will still give
     /// you the desired recall.
+    ///
+    /// This method sets both the minimum and maximum number of partitions to search.
+    /// For more fine-grained control see [`VectorQuery::minimum_nprobes`] and
+    /// [`VectorQuery::maximum_nprobes`].
     pub fn nprobes(mut self, nprobes: usize) -> Self {
-        self.request.nprobes = nprobes;
+        self.request.minimum_nprobes = nprobes;
+        self.request.maximum_nprobes = Some(nprobes);
         self
     }
 
+    /// Set the minimum number of partitions to search
+    ///
+    /// This argument is only used when the vector column has an IVF PQ index.
+    /// If there is no index then this value is ignored.
+    ///
+    /// See [`VectorQuery::nprobes`] for more details.
+    ///
+    /// These partitions will be searched on every indexed vector query.
+    ///
+    /// Will return an error if the value is not greater than 0 or if maximum_nprobes
+    /// has been set and is less than the minimum_nprobes.
+    pub fn minimum_nprobes(mut self, minimum_nprobes: usize) -> Result<Self> {
+        if minimum_nprobes == 0 {
+            return Err(Error::InvalidInput {
+                message: "minimum_nprobes must be greater than 0".to_string(),
+            });
+        }
+        if let Some(maximum_nprobes) = self.request.maximum_nprobes {
+            if minimum_nprobes > maximum_nprobes {
+                return Err(Error::InvalidInput {
+                    message: "minimum_nprobes must be less or equal to maximum_nprobes".to_string(),
+                });
+            }
+        }
+        self.request.minimum_nprobes = minimum_nprobes;
+        Ok(self)
+    }
+
+    /// Set the maximum number of partitions to search
+    ///
+    /// This argument is only used when the vector column has an IVF PQ index.
+    /// If there is no index then this value is ignored.
+    ///
+    /// See [`VectorQuery::nprobes`] for more details.
+    ///
+    /// If this value is greater than minimum_nprobes then the excess partitions will
+    /// only be searched if the initial search does not return enough results.
+    ///
+    /// This can be useful when there is a narrow filter to allow these queries to
+    /// spend more time searching and avoid potential false negatives.
+    ///
+    /// Set to None to search all partitions, if needed, to satsify the limit
+    pub fn maximum_nprobes(mut self, maximum_nprobes: Option<usize>) -> Result<Self> {
+        if let Some(maximum_nprobes) = maximum_nprobes {
+            if maximum_nprobes == 0 {
+                return Err(Error::InvalidInput {
+                    message: "maximum_nprobes must be greater than 0".to_string(),
+                });
+            }
+            if maximum_nprobes < self.request.minimum_nprobes {
+                return Err(Error::InvalidInput {
+                    message: "maximum_nprobes must be greater than minimum_nprobes".to_string(),
+                });
+            }
+        }
+        self.request.maximum_nprobes = maximum_nprobes;
+        Ok(self)
+    }
+
     /// Set the distance range for vector search,
     /// only rows with distances in the range [lower_bound, upper_bound) will be returned
     pub fn distance_range(mut self, lower_bound: Option<f32>, upper_bound: Option<f32>) -> Self {
@@ -1208,7 +1275,8 @@ mod tests {
         );
         assert_eq!(query.request.base.limit.unwrap(), 100);
         assert_eq!(query.request.base.offset.unwrap(), 1);
-        assert_eq!(query.request.nprobes, 1000);
+        assert_eq!(query.request.minimum_nprobes, 1000);
+        assert_eq!(query.request.maximum_nprobes, Some(1000));
         assert!(query.request.use_index);
         assert_eq!(query.request.distance_type, Some(DistanceType::Cosine));
         assert_eq!(query.request.refine_factor, Some(999));

rust/lancedb/src/remote/table.rs

@@ -32,6 +32,7 @@ use lance::dataset::{ColumnAlteration, NewColumnTransform, Version};
 use lance_datafusion::exec::{execute_plan, OneShotExec};
 use reqwest::{RequestBuilder, Response};
 use serde::{Deserialize, Serialize};
+use serde_json::Number;
 use std::collections::HashMap;
 use std::io::Cursor;
 use std::pin::Pin;
@@ -56,6 +57,8 @@ use crate::{
 };
 
 const REQUEST_TIMEOUT_HEADER: HeaderName = HeaderName::from_static("x-request-timeout-ms");
+const METRIC_TYPE_KEY: &str = "metric_type";
+const INDEX_TYPE_KEY: &str = "index_type";
 
 pub struct RemoteTags<'a, S: HttpSend = Sender> {
     inner: &'a RemoteTable<S>,
@@ -438,7 +441,18 @@ impl<S: HttpSend> RemoteTable<S> {
 
         // Apply general parameters, before we dispatch based on number of query vectors.
         body["distance_type"] = serde_json::json!(query.distance_type.unwrap_or_default());
-        body["nprobes"] = query.nprobes.into();
+        // In 0.23.1 we migrated from `nprobes` to `minimum_nprobes` and `maximum_nprobes`.
+        // Old client / new server: since minimum_nprobes is missing, fallback to nprobes
+        // New client / old server: old server will only see nprobes, make sure to set both
+        //                          nprobes and minimum_nprobes
+        // New client / new server: since minimum_nprobes is present, server can ignore nprobes
+        body["nprobes"] = query.minimum_nprobes.into();
+        body["minimum_nprobes"] = query.minimum_nprobes.into();
+        if let Some(maximum_nprobes) = query.maximum_nprobes {
+            body["maximum_nprobes"] = maximum_nprobes.into();
+        } else {
+            body["maximum_nprobes"] = serde_json::Value::Number(Number::from_u128(0).unwrap())
+        }
         body["lower_bound"] = query.lower_bound.into();
         body["upper_bound"] = query.upper_bound.into();
         body["ef"] = query.ef.into();
@@ -985,27 +999,53 @@ impl<S: HttpSend> BaseTable for RemoteTable<S> {
             "column": column
         });
 
-        let (index_type, distance_type) = match index.index {
+        match index.index {
             // TODO: Should we pass the actual index parameters? SaaS does not
             // yet support them.
-            Index::IvfFlat(index) => ("IVF_FLAT", Some(index.distance_type)),
+            Index::IvfFlat(index) => {
-            Index::IvfPq(index) => ("IVF_PQ", Some(index.distance_type)),
+                body[INDEX_TYPE_KEY] = serde_json::Value::String("IVF_FLAT".to_string());
-            Index::IvfHnswSq(index) => ("IVF_HNSW_SQ", Some(index.distance_type)),
+                body[METRIC_TYPE_KEY] =
-            Index::BTree(_) => ("BTREE", None),
+                    serde_json::Value::String(index.distance_type.to_string().to_lowercase());
-            Index::Bitmap(_) => ("BITMAP", None),
+                if let Some(num_partitions) = index.num_partitions {
-            Index::LabelList(_) => ("LABEL_LIST", None),
+                    body["num_partitions"] = serde_json::Value::Number(num_partitions.into());
+                }
+            }
+            Index::IvfPq(index) => {
+                body[INDEX_TYPE_KEY] = serde_json::Value::String("IVF_PQ".to_string());
+                body[METRIC_TYPE_KEY] =
+                    serde_json::Value::String(index.distance_type.to_string().to_lowercase());
+                if let Some(num_partitions) = index.num_partitions {
+                    body["num_partitions"] = serde_json::Value::Number(num_partitions.into());
+                }
+                if let Some(num_bits) = index.num_bits {
+                    body["num_bits"] = serde_json::Value::Number(num_bits.into());
+                }
+            }
+            Index::IvfHnswSq(index) => {
+                body[INDEX_TYPE_KEY] = serde_json::Value::String("IVF_HNSW_SQ".to_string());
+                body[METRIC_TYPE_KEY] =
+                    serde_json::Value::String(index.distance_type.to_string().to_lowercase());
+                if let Some(num_partitions) = index.num_partitions {
+                    body["num_partitions"] = serde_json::Value::Number(num_partitions.into());
+                }
+            }
+            Index::BTree(_) => {
+                body[INDEX_TYPE_KEY] = serde_json::Value::String("BTREE".to_string());
+            }
+            Index::Bitmap(_) => {
+                body[INDEX_TYPE_KEY] = serde_json::Value::String("BITMAP".to_string());
+            }
+            Index::LabelList(_) => {
+                body[INDEX_TYPE_KEY] = serde_json::Value::String("LABEL_LIST".to_string());
+            }
             Index::FTS(fts) => {
-                let with_position = fts.with_position;
+                body[INDEX_TYPE_KEY] = serde_json::Value::String("FTS".to_string());
-                let configs = serde_json::to_value(fts.tokenizer_configs).map_err(|e| {
+                let params = serde_json::to_value(&fts).map_err(|e| Error::InvalidInput {
-                    Error::InvalidInput {
+                    message: format!("failed to serialize FTS index params {:?}", e),
-                        message: format!("failed to serialize FTS index params {:?}", e),
-                    }
                 })?;
-                for (key, value) in configs.as_object().unwrap() {
+                for (key, value) in params.as_object().unwrap() {
                     body[key] = value.clone();
                 }
-                body["with_position"] = serde_json::Value::Bool(with_position);
-                ("FTS", None)
             }
             Index::Auto => {
                 let schema = self.schema().await?;
@@ -1015,9 +1055,11 @@ impl<S: HttpSend> BaseTable for RemoteTable<S> {
                         message: format!("Column {} not found in schema", column),
                     })?;
                 if supported_vector_data_type(field.data_type()) {
-                    ("IVF_PQ", Some(DistanceType::L2))
+                    body[INDEX_TYPE_KEY] = serde_json::Value::String("IVF_PQ".to_string());
+                    body[METRIC_TYPE_KEY] =
+                        serde_json::Value::String(DistanceType::L2.to_string().to_lowercase());
                 } else if supported_btree_data_type(field.data_type()) {
-                    ("BTREE", None)
+                    body[INDEX_TYPE_KEY] = serde_json::Value::String("BTREE".to_string());
                 } else {
                     return Err(Error::NotSupported {
                         message: format!(
@@ -1034,12 +1076,6 @@ impl<S: HttpSend> BaseTable for RemoteTable<S> {
                 })
             }
         };
-        body["index_type"] = serde_json::Value::String(index_type.into());
-        if let Some(distance_type) = distance_type {
-            // Phalanx expects this to be lowercase right now.
-            body["metric_type"] =
-                serde_json::Value::String(distance_type.to_string().to_lowercase());
-        }
 
         let request = request.json(&body);
 
@@ -1421,11 +1457,12 @@ mod tests {
     use chrono::{DateTime, Utc};
     use futures::{future::BoxFuture, StreamExt, TryFutureExt};
     use lance_index::scalar::inverted::query::MatchQuery;
-    use lance_index::scalar::FullTextSearchQuery;
+    use lance_index::scalar::{FullTextSearchQuery, InvertedIndexParams};
     use reqwest::Body;
     use rstest::rstest;
+    use serde_json::json;
 
-    use crate::index::vector::IvfFlatIndexBuilder;
+    use crate::index::vector::{IvfFlatIndexBuilder, IvfHnswSqIndexBuilder};
     use crate::remote::db::DEFAULT_SERVER_VERSION;
     use crate::remote::JSON_CONTENT_TYPE;
     use crate::{
@@ -2079,6 +2116,8 @@ mod tests {
                 "prefilter": true,
                 "distance_type": "l2",
                 "nprobes": 20,
+                "minimum_nprobes": 20,
+                "maximum_nprobes": 20,
                 "lower_bound": Option::<f32>::None,
                 "upper_bound": Option::<f32>::None,
                 "k": 10,
@@ -2179,6 +2218,8 @@ mod tests {
                 "bypass_vector_index": true,
                 "columns": ["a", "b"],
                 "nprobes": 12,
+                "minimum_nprobes": 12,
+                "maximum_nprobes": 12,
                 "lower_bound": Option::<f32>::None,
                 "upper_bound": Option::<f32>::None,
                 "ef": Option::<usize>::None,
@@ -2306,6 +2347,7 @@ mod tests {
                                 "fuzziness": 0,
                                 "max_expansions": 50,
                                 "operator": "Or",
+                                "prefix_length": 0,
                             },
                         }
                     },
@@ -2420,29 +2462,79 @@ mod tests {
         let cases = [
             (
                 "IVF_FLAT",
-                Some("hamming"),
+                json!({
+                    "metric_type": "hamming",
+                }),
                 Index::IvfFlat(IvfFlatIndexBuilder::default().distance_type(DistanceType::Hamming)),
             ),
-            ("IVF_PQ", Some("l2"), Index::IvfPq(Default::default())),
+            (
+                "IVF_FLAT",
+                json!({
+                    "metric_type": "hamming",
+                    "num_partitions": 128,
+                }),
+                Index::IvfFlat(
+                    IvfFlatIndexBuilder::default()
+                        .distance_type(DistanceType::Hamming)
+                        .num_partitions(128),
+                ),
+            ),
             (
                 "IVF_PQ",
-                Some("cosine"),
+                json!({
-                Index::IvfPq(IvfPqIndexBuilder::default().distance_type(DistanceType::Cosine)),
+                    "metric_type": "l2",
+                }),
+                Index::IvfPq(Default::default()),
+            ),
+            (
+                "IVF_PQ",
+                json!({
+                    "metric_type": "cosine",
+                    "num_partitions": 128,
+                    "num_bits": 4,
+                }),
+                Index::IvfPq(
+                    IvfPqIndexBuilder::default()
+                        .distance_type(DistanceType::Cosine)
+                        .num_partitions(128)
+                        .num_bits(4),
+                ),
             ),
             (
                 "IVF_HNSW_SQ",
-                Some("l2"),
+                json!({
+                    "metric_type": "l2",
+                }),
                 Index::IvfHnswSq(Default::default()),
             ),
+            (
+                "IVF_HNSW_SQ",
+                json!({
+                    "metric_type": "l2",
+                    "num_partitions": 128,
+                }),
+                Index::IvfHnswSq(
+                    IvfHnswSqIndexBuilder::default()
+                        .distance_type(DistanceType::L2)
+                        .num_partitions(128),
+                ),
+            ),
             // HNSW_PQ isn't yet supported on SaaS
-            ("BTREE", None, Index::BTree(Default::default())),
+            ("BTREE", json!({}), Index::BTree(Default::default())),
-            ("BITMAP", None, Index::Bitmap(Default::default())),
+            ("BITMAP", json!({}), Index::Bitmap(Default::default())),
-            ("LABEL_LIST", None, Index::LabelList(Default::default())),
+            (
-            ("FTS", None, Index::FTS(Default::default())),
+                "LABEL_LIST",
+                json!({}),
+                Index::LabelList(Default::default()),
+            ),
+            (
+                "FTS",
+                serde_json::to_value(InvertedIndexParams::default()).unwrap(),
+                Index::FTS(Default::default()),
+            ),
         ];
 
-        for (index_type, distance_type, index) in cases {
+        for (index_type, expected_body, index) in cases {
-            let params = index.clone();
             let table = Table::new_with_handler("my_table", move |request| {
                 assert_eq!(request.method(), "POST");
                 assert_eq!(request.url().path(), "/v1/table/my_table/create_index/");
@@ -2452,23 +2544,9 @@ mod tests {
                 );
                 let body = request.body().unwrap().as_bytes().unwrap();
                 let body: serde_json::Value = serde_json::from_slice(body).unwrap();
-                let mut expected_body = serde_json::json!({
+                let mut expected_body = expected_body.clone();
-                    "column": "a",
+                expected_body["column"] = "a".into();
-                    "index_type": index_type,
+                expected_body[INDEX_TYPE_KEY] = index_type.into();
-                });
-                if let Some(distance_type) = distance_type {
-                    expected_body["metric_type"] = distance_type.to_lowercase().into();
-                }
-                if let Index::FTS(fts) = &params {
-                    expected_body["with_position"] = fts.with_position.into();
-                    expected_body["base_tokenizer"] = "simple".into();
-                    expected_body["language"] = "English".into();
-                    expected_body["max_token_length"] = 40.into();
-                    expected_body["lower_case"] = true.into();
-                    expected_body["stem"] = false.into();
-                    expected_body["remove_stop_words"] = false.into();
-                    expected_body["ascii_folding"] = false.into();
-                }
 
                 assert_eq!(body, expected_body);
 

rust/lancedb/src/table.rs

@@ -14,7 +14,7 @@ use datafusion_physical_plan::projection::ProjectionExec;
 use datafusion_physical_plan::repartition::RepartitionExec;
 use datafusion_physical_plan::union::UnionExec;
 use datafusion_physical_plan::ExecutionPlan;
-use futures::{FutureExt, StreamExt, TryFutureExt, TryStreamExt};
+use futures::{FutureExt, StreamExt, TryFutureExt};
 use lance::dataset::builder::DatasetBuilder;
 use lance::dataset::cleanup::RemovalStats;
 use lance::dataset::optimize::{compact_files, CompactionMetrics, IndexRemapperOptions};
@@ -85,6 +85,7 @@ pub use lance::dataset::optimize::CompactionOptions;
 pub use lance::dataset::refs::{TagContents, Tags as LanceTags};
 pub use lance::dataset::scanner::DatasetRecordBatchStream;
 use lance::dataset::statistics::DatasetStatisticsExt;
+use lance_index::frag_reuse::FRAG_REUSE_INDEX_NAME;
 pub use lance_index::optimize::OptimizeOptions;
 use serde_with::skip_serializing_none;
 
@@ -1977,16 +1978,12 @@ impl NativeTable {
         }
 
         let mut dataset = self.dataset.get_mut().await?;
-        let fts_params = lance_index::scalar::InvertedIndexParams {
-            with_position: fts_opts.with_position,
-            tokenizer_config: fts_opts.tokenizer_configs,
-        };
         dataset
             .create_index(
                 &[field.name()],
                 IndexType::Inverted,
                 None,
-                &fts_params,
+                &fts_opts,
                 replace,
             )
             .await?;
@@ -2357,12 +2354,15 @@ impl BaseTable for NativeTable {
                     query.base.limit.unwrap_or(DEFAULT_TOP_K),
                 )?;
             }
+            scanner.minimum_nprobes(query.minimum_nprobes);
+            if let Some(maximum_nprobes) = query.maximum_nprobes {
+                scanner.maximum_nprobes(maximum_nprobes);
+            }
         }
         scanner.limit(
             query.base.limit.map(|limit| limit as i64),
             query.base.offset.map(|offset| offset as i64),
         )?;
-        scanner.nprobs(query.nprobes);
         if let Some(ef) = query.ef {
             scanner.ef(ef);
         }
@@ -2605,28 +2605,56 @@ impl BaseTable for NativeTable {
     async fn list_indices(&self) -> Result<Vec<IndexConfig>> {
         let dataset = self.dataset.get().await?;
         let indices = dataset.load_indices().await?;
-        futures::stream::iter(indices.as_slice()).then(|idx| async {
+        let results = futures::stream::iter(indices.as_slice()).then(|idx| async {
-            let stats = dataset.index_statistics(idx.name.as_str()).await?;
+
-            let stats: serde_json::Value = serde_json::from_str(&stats).map_err(|e| Error::Runtime {
+            // skip Lance internal indexes
-                message: format!("error deserializing index statistics: {}", e),
+            if idx.name == FRAG_REUSE_INDEX_NAME {
-            })?;
+                return None;
-            let index_type = stats.get("index_type").and_then(|v| v.as_str())
+            }
-            .ok_or_else(|| Error::Runtime {
+
-                message: "index statistics was missing index type".to_string(),
+            let stats = match dataset.index_statistics(idx.name.as_str()).await {
-            })?;
+                Ok(stats) => stats,
-            let index_type: crate::index::IndexType = index_type.parse().map_err(|e| Error::Runtime {
+                Err(e) => {
-                message: format!("error parsing index type: {}", e),
+                    log::warn!("Failed to get statistics for index {} ({}): {}", idx.name, idx.uuid, e);
-            })?;
+                    return None;
+                }
+            };
+
+            let stats: serde_json::Value = match serde_json::from_str(&stats) {
+                Ok(stats) => stats,
+                Err(e) => {
+                    log::warn!("Failed to deserialize index statistics for index {} ({}): {}", idx.name, idx.uuid, e);
+                    return None;
+                }
+            };
+
+            let Some(index_type) = stats.get("index_type").and_then(|v| v.as_str()) else {
+                log::warn!("Index statistics was missing 'index_type' field for index {} ({})", idx.name, idx.uuid);
+                return None;
+            };
+
+            let index_type: crate::index::IndexType = match index_type.parse() {
+                Ok(index_type) => index_type,
+                Err(e) => {
+                    log::warn!("Failed to parse index type for index {} ({}): {}", idx.name, idx.uuid, e);
+                    return None;
+                }
+            };
 
             let mut columns = Vec::with_capacity(idx.fields.len());
             for field_id in &idx.fields {
-                let field = dataset.schema().field_by_id(*field_id).ok_or_else(|| Error::Runtime { message: format!("The index with name {} and uuid {} referenced a field with id {} which does not exist in the schema", idx.name, idx.uuid, field_id) })?;
+                let Some(field) = dataset.schema().field_by_id(*field_id) else {
+                    log::warn!("The index {} ({}) referenced a field with id {} which does not exist in the schema", idx.name, idx.uuid, field_id);
+                    return None;
+                };
                 columns.push(field.name.clone());
             }
 
             let name = idx.name.clone();
-            Ok(IndexConfig { index_type, columns, name })
+            Some(IndexConfig { index_type, columns, name })
-        }).try_collect::<Vec<_>>().await
+        }).collect::<Vec<_>>().await;
+
+        Ok(results.into_iter().flatten().collect())
     }
 
     fn dataset_uri(&self) -> &str {
@@ -2819,7 +2847,7 @@ mod tests {
     use super::*;
     use crate::connect;
     use crate::connection::ConnectBuilder;
-    use crate::index::scalar::BTreeIndexBuilder;
+    use crate::index::scalar::{BTreeIndexBuilder, BitmapIndexBuilder};
     use crate::query::{ExecutableQuery, QueryBase};
 
     #[tokio::test]
@@ -4271,4 +4299,65 @@ mod tests {
             }
         )
     }
+
+    #[tokio::test]
+    pub async fn test_list_indices_skip_frag_reuse() {
+        let tmp_dir = tempdir().unwrap();
+        let uri = tmp_dir.path().to_str().unwrap();
+
+        let conn = ConnectBuilder::new(uri).execute().await.unwrap();
+
+        let schema = Arc::new(Schema::new(vec![
+            Field::new("id", DataType::Int32, false),
+            Field::new("foo", DataType::Int32, true),
+        ]));
+        let batch = RecordBatch::try_new(
+            schema.clone(),
+            vec![
+                Arc::new(Int32Array::from_iter_values(0..100)),
+                Arc::new(Int32Array::from_iter_values(0..100)),
+            ],
+        )
+        .unwrap();
+
+        let table = conn
+            .create_table(
+                "test_list_indices_skip_frag_reuse",
+                RecordBatchIterator::new(vec![Ok(batch.clone())], batch.schema()),
+            )
+            .execute()
+            .await
+            .unwrap();
+
+        table
+            .add(RecordBatchIterator::new(
+                vec![Ok(batch.clone())],
+                batch.schema(),
+            ))
+            .execute()
+            .await
+            .unwrap();
+
+        table
+            .create_index(&["id"], Index::Bitmap(BitmapIndexBuilder {}))
+            .execute()
+            .await
+            .unwrap();
+
+        table
+            .optimize(OptimizeAction::Compact {
+                options: CompactionOptions {
+                    target_rows_per_fragment: 2_000,
+                    defer_index_remap: true,
+                    ..Default::default()
+                },
+                remap_options: None,
+            })
+            .await
+            .unwrap();
+
+        let result = table.list_indices().await.unwrap();
+        assert_eq!(result.len(), 1);
+        assert_eq!(result[0].index_type, crate::index::IndexType::Bitmap);
+    }
 }

rust/lancedb/src/table/datafusion.rs

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

rust/lancedb/src/table/dataset.rs

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