[python] Bump version: 0.3.0 → 0.3.1

#488

.bumpversion.cfg

@@ -1,5 +1,5 @@
 [bumpversion]
-current_version = 0.1.19
+current_version = 0.3.0
 commit = True
 message = Bump version: {current_version} → {new_version}
 tag = True

.github/workflows/node.yml

@@ -9,6 +9,7 @@ on:
       - node/**
       - rust/ffi/node/**
       - .github/workflows/node.yml
+      - docker-compose.yml
 
 env:
   # Disable full debug symbol generation to speed up CI build and keep memory down
@@ -107,3 +108,56 @@ jobs:
     - name: Test
       run: |
         npm run test
+  aws-integtest:
+    timeout-minutes: 45
+    runs-on: "ubuntu-22.04"
+    defaults:
+      run:
+        shell: bash
+        working-directory: node
+    env:
+      AWS_ACCESS_KEY_ID: ACCESSKEY
+      AWS_SECRET_ACCESS_KEY: SECRETKEY
+      AWS_DEFAULT_REGION: us-west-2
+      # this one is for s3
+      AWS_ENDPOINT: http://localhost:4566
+      # this one is for dynamodb
+      DYNAMODB_ENDPOINT: http://localhost:4566
+    steps:
+    - uses: actions/checkout@v3
+      with:
+        fetch-depth: 0
+        lfs: true
+    - uses: actions/setup-node@v3
+      with:
+        node-version: 18
+        cache: 'npm'
+        cache-dependency-path: node/package-lock.json
+    - name: start local stack
+      run: docker compose -f ../docker-compose.yml up -d --wait
+    - name: create s3
+      run: aws s3 mb s3://lancedb-integtest --endpoint $AWS_ENDPOINT
+    - name: create ddb
+      run: |
+        aws dynamodb create-table \
+          --table-name lancedb-integtest \
+          --attribute-definitions '[{"AttributeName": "base_uri", "AttributeType": "S"}, {"AttributeName": "version", "AttributeType": "N"}]' \
+          --key-schema '[{"AttributeName": "base_uri", "KeyType": "HASH"}, {"AttributeName": "version", "KeyType": "RANGE"}]' \
+          --provisioned-throughput '{"ReadCapacityUnits": 10, "WriteCapacityUnits": 10}' \
+          --endpoint-url $DYNAMODB_ENDPOINT
+    - uses: Swatinem/rust-cache@v2
+    - name: Install dependencies
+      run: |
+        sudo apt update
+        sudo apt install -y protobuf-compiler libssl-dev
+    - name: Build
+      run: |
+        npm ci
+        npm run tsc
+        npm run build
+        npm run pack-build
+        npm install --no-save ./dist/lancedb-vectordb-*.tgz
+        # Remove index.node to test with dependency installed
+        rm index.node
+    - name: Test
+      run: npm run integration-test

.github/workflows/python.yml

@@ -38,7 +38,7 @@ jobs:
     - name: isort
       run: isort --check --diff --quiet .
     - name: Run tests
-      run: pytest -x -v --durations=30 tests
+      run: pytest -m "not slow" -x -v --durations=30 tests
     - name: doctest
       run: pytest --doctest-modules lancedb
   mac:
@@ -65,4 +65,34 @@ jobs:
     - name: Black
       run: black --check --diff --no-color --quiet .
     - name: Run tests
-      run: pytest -x -v --durations=30 tests
+      run: pytest -m "not slow" -x -v --durations=30 tests
+  pydantic1x:
+    timeout-minutes: 30
+    runs-on: "ubuntu-22.04"
+    defaults:
+      run:
+        shell: bash
+        working-directory: python
+    steps:
+    - uses: actions/checkout@v3
+      with:
+        fetch-depth: 0
+        lfs: true
+    - name: Set up Python
+      uses: actions/setup-python@v4
+      with:
+        python-version: 3.9
+    - name: Install lancedb
+      run: |
+        pip install "pydantic<2"
+        pip install -e .[tests]
+        pip install tantivy@git+https://github.com/quickwit-oss/tantivy-py#164adc87e1a033117001cf70e38c82a53014d985
+        pip install pytest pytest-mock black isort
+    - name: Black
+      run: black --check --diff --no-color --quiet .
+    - name: isort
+      run: isort --check --diff --quiet .
+    - name: Run tests
+      run: pytest -m "not slow" -x -v --durations=30 tests
+    - name: doctest
+      run: pytest --doctest-modules lancedb

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

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

Cargo.toml

@@ -1,17 +1,27 @@
 [workspace]
-members = [
-    "rust/vectordb",
-    "rust/ffi/node"
-]
+members = ["rust/ffi/node", "rust/vectordb"]
+# Python package needs to be built by maturin.
+exclude = ["python"]
 resolver = "2"
 
 [workspace.dependencies]
-lance = "=0.5.9"
-arrow-array = "42.0"
-arrow-data = "42.0"
-arrow-schema = "42.0"
-arrow-ipc = "42.0"
-half = { "version" = "=2.2.1", default-features = false }
+lance = { "version" = "=0.8.3", "features" = ["dynamodb"] }
+lance-linalg = { "version" = "=0.8.3" }
+lance-testing = { "version" = "=0.8.3" }
+# Note that this one does not include pyarrow
+arrow = { version = "43.0.0", optional = false }
+arrow-array = "43.0"
+arrow-data = "43.0"
+arrow-ipc = "43.0"
+arrow-ord = "43.0"
+arrow-schema = "43.0"
+arrow-arith = "43.0"
+arrow-cast = "43.0"
+chrono = "0.4.23"
+half = { "version" = "=2.2.1", default-features = false, features = [
+    "num-traits"
+] }
+log = "0.4"
 object_store = "0.6.1"
 snafu = "0.7.4"
-
+url = "2"

README.md

@@ -33,6 +33,8 @@ The key features of LanceDB include:
 
 * Zero-copy, automatic versioning, manage versions of your data without needing extra infrastructure.
 
+* GPU support in building vector index(*).
+
 * Ecosystem integrations with [LangChain 🦜️🔗](https://python.langchain.com/en/latest/modules/indexes/vectorstores/examples/lanecdb.html), [LlamaIndex 🦙](https://gpt-index.readthedocs.io/en/latest/examples/vector_stores/LanceDBIndexDemo.html), Apache-Arrow, Pandas, Polars, DuckDB and more on the way.
 
 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.
@@ -52,8 +54,7 @@ const table = await db.createTable('vectors',
       [{ id: 1, vector: [0.1, 0.2], item: "foo", price: 10 },
        { id: 2, vector: [1.1, 1.2], item: "bar", price: 50 }])
 
-const query = table.search([0.1, 0.3]);
-query.limit = 20;
+const query = table.search([0.1, 0.3]).limit(2);
 const results = await query.execute();
 ```
 
@@ -70,7 +71,7 @@ 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_df()
+result = table.search([100, 100]).limit(2).to_pandas()
 ```
 
 ## Blogs, Tutorials & Videos

docker-compose.yml

@@ -0,0 +1,18 @@
+version: "3.9"
+services:
+  localstack:
+    image: localstack/localstack:0.14
+    ports:
+      - 4566:4566
+    environment:
+      - SERVICES=s3,dynamodb
+      - DEBUG=1
+      - LS_LOG=trace
+      - DOCKER_HOST=unix:///var/run/docker.sock
+      - AWS_ACCESS_KEY_ID=ACCESSKEY
+      - AWS_SECRET_ACCESS_KEY=SECRETKEY
+    healthcheck:
+      test: [ "CMD", "curl", "-f", "http://localhost:4566/health" ]
+      interval: 5s
+      retries: 3
+      start_period: 10s

docs/mkdocs.yml

@@ -1,5 +1,6 @@
 site_name: LanceDB Docs
 repo_url: https://github.com/lancedb/lancedb
+edit_uri: https://github.com/lancedb/lancedb/tree/main/docs/src
 repo_name: lancedb/lancedb
 docs_dir: src
 
@@ -10,6 +11,17 @@ theme:
   features:
     - content.code.copy
     - content.tabs.link
+    - content.action.edit
+    - toc.follow
+    - toc.integrate
+    - navigation.top
+    - navigation.tabs
+    - navigation.tabs.sticky
+    - navigation.footer
+    - navigation.tracking
+    - navigation.instant
+    - navigation.indexes
+    - navigation.expand
   icon:
     repo: fontawesome/brands/github
   custom_dir: overrides
@@ -53,11 +65,50 @@ markdown_extensions:
 - md_in_html
 
 nav:
-- Home: index.md
+- Home:
+  - 🏢 Home: index.md
+  - 💡 Basics: basic.md
+  - 📚 Guides:
+    - Create Ingest Update Delete: guides/tables.md
+    - Vector Search: search.md
+    - SQL filters: sql.md
+    - Indexing: ann_indexes.md
+  - 🧬 Embeddings: embedding.md
+  - 🔍 Python full-text search: fts.md
+  - 🔌 Integrations:
+    - integrations/index.md
+    - Pandas and PyArrow: python/arrow.md
+    - DuckDB: python/duckdb.md
+    - LangChain 🔗: https://python.langchain.com/en/latest/modules/indexes/vectorstores/examples/lancedb.html
+    - LangChain JS/TS 🔗: https://js.langchain.com/docs/modules/data_connection/vectorstores/integrations/lancedb
+    - LlamaIndex 🦙: https://gpt-index.readthedocs.io/en/latest/examples/vector_stores/LanceDBIndexDemo.html
+    - Pydantic: python/pydantic.md
+    - Voxel51: integrations/voxel51.md
+    - PromptTools: integrations/prompttools.md
+  - 🐍 Python examples:
+    - examples/index.md
+    - YouTube Transcript Search: notebooks/youtube_transcript_search.ipynb
+    - Documentation QA Bot using LangChain: notebooks/code_qa_bot.ipynb
+    - Multimodal search using CLIP: notebooks/multimodal_search.ipynb
+    - Serverless QA Bot with S3 and Lambda: examples/serverless_lancedb_with_s3_and_lambda.md
+    - Serverless QA Bot with Modal: examples/serverless_qa_bot_with_modal_and_langchain.md
+  - 🌐 Javascript examples:
+    - Examples: examples/index_js.md
+    - Serverless Website Chatbot: examples/serverless_website_chatbot.md
+    - YouTube Transcript Search: examples/youtube_transcript_bot_with_nodejs.md
+    - TransformersJS Embedding Search: examples/transformerjs_embedding_search_nodejs.md
+  - ⚙️ CLI & Config: cli_config.md
+
 - Basics: basic.md
+- Guides:
+  - Create Ingest Update Delete: guides/tables.md
+  - Vector Search: search.md
+  - SQL filters: sql.md
+  - Indexing: ann_indexes.md
 - Embeddings: embedding.md
 - Python full-text search: fts.md
 - Integrations:
+  - integrations/index.md
   - Pandas and PyArrow: python/arrow.md
   - DuckDB: python/duckdb.md
   - LangChain 🦜️🔗: https://python.langchain.com/en/latest/modules/indexes/vectorstores/examples/lancedb.html
@@ -65,23 +116,28 @@ nav:
   - LlamaIndex 🦙: https://gpt-index.readthedocs.io/en/latest/examples/vector_stores/LanceDBIndexDemo.html
   - Pydantic: python/pydantic.md
   - Voxel51: integrations/voxel51.md
+  - PromptTools: integrations/prompttools.md
 - Python examples:
+  - examples/index.md
   - YouTube Transcript Search: notebooks/youtube_transcript_search.ipynb
   - Documentation QA Bot using LangChain: notebooks/code_qa_bot.ipynb
   - Multimodal search using CLIP: notebooks/multimodal_search.ipynb
   - Serverless QA Bot with S3 and Lambda: examples/serverless_lancedb_with_s3_and_lambda.md
   - Serverless QA Bot with Modal: examples/serverless_qa_bot_with_modal_and_langchain.md
 - Javascript examples:
+  - examples/index_js.md
   - YouTube Transcript Search: examples/youtube_transcript_bot_with_nodejs.md
+  - Serverless Chatbot from any website: examples/serverless_website_chatbot.md
   - TransformersJS Embedding Search: examples/transformerjs_embedding_search_nodejs.md
-
-- References:
-  - Vector Search: search.md
-  - SQL filters: sql.md
-  - Indexing: ann_indexes.md
 - API references:
   - Python API: python/python.md
   - Javascript API: javascript/modules.md
+- LanceDB Cloud↗: https://noteforms.com/forms/lancedb-mailing-list-cloud-kty1o5?notionforms=1&utm_source=notionforms
 
 extra_css:
   - styles/global.css
+
+extra:
+  analytics:
+    provider: google
+    property: G-B7NFM40W74

docs/src/ann_indexes.md

@@ -6,7 +6,7 @@ LanceDB provides many parameters to fine-tune the index's size, the speed of que
 
 Currently, LanceDB does *not* automatically create the ANN index.
 LanceDB has optimized code for KNN as well. For many use-cases, datasets under 100K vectors won't require index creation at all.
-If you can live with <100ms latency, skipping index creation is a simpler workflow while guaranteeing 100% recall.
+If you can live with < 100ms latency, skipping index creation is a simpler workflow while guaranteeing 100% recall.
 
 In the future we will look to automatically create and configure the ANN index.
 
@@ -68,6 +68,12 @@ a single PQ code.
   <figcaption>IVF_PQ index with <code>num_partitions=2, num_sub_vectors=4</code></figcaption>
 </figure>
 
+### Use GPU to build vector index
+
+Lance Python SDK has experimental GPU support for creating IVF index.
+You can specify the GPU device to train IVF partitions via
+
+- **accelerator**: Specify to `"cuda"`` to enable GPU training.
 
 ## Querying an ANN Index
 
@@ -91,10 +97,10 @@ There are a couple of parameters that can be used to fine-tune the search:
          .limit(2) \
          .nprobes(20) \
          .refine_factor(10) \
-         .to_df()
+         .to_pandas()
      ```
      ```
-                                              vector       item       score
+                                              vector       item       _distance
      0  [0.44949695, 0.8444449, 0.06281311, 0.23338133...  item 1141  103.575333
      1  [0.48587373, 0.269207, 0.15095535, 0.65531915,...  item 3953  108.393867
      ```
@@ -109,9 +115,8 @@ There are a couple of parameters that can be used to fine-tune the search:
          .execute()
      ```
 
-The search will return the data requested in addition to the score of each item.
+The search will return the data requested in addition to the distance of each item.
 
-**Note:** The score is the distance between the query vector and the element. A lower number means that the result is more relevant.
 
 ### Filtering (where clause)
 
@@ -119,7 +124,7 @@ You can further filter the elements returned by a search using a where clause.
 
 === "Python"
      ```python
-     tbl.search(np.random.random((1536))).where("item != 'item 1141'").to_df()
+     tbl.search(np.random.random((1536))).where("item != 'item 1141'").to_pandas()
      ```
 
 === "Javascript"
@@ -136,10 +141,10 @@ You can select the columns returned by the query using a select clause.
 
 === "Python"
      ```python
-     tbl.search(np.random.random((1536))).select(["vector"]).to_df()
+     tbl.search(np.random.random((1536))).select(["vector"]).to_pandas()
      ```
      ```
-     vector      score
+        vector                                             _distance
      0  [0.30928212, 0.022668175, 0.1756372, 0.4911822...  93.971092
      1  [0.2525465, 0.01723831, 0.261568, 0.002007689,...  95.173485
      ...
@@ -155,28 +160,28 @@ You can select the columns returned by the query using a select clause.
 
 ## FAQ
 
-### When is it necessary to create an ANN vector index.
+### When is it necessary to create an ANN vector index?
 
-`LanceDB` has manually tuned SIMD code for computing vector distances.
-In our benchmarks, computing 100K pairs of 1K dimension vectors only take less than 20ms.
-For small dataset (<100K rows) or the applications which can accept 100ms latency, vector indices are usually not necessary.
+`LanceDB` has manually-tuned SIMD code for computing vector distances.
+In our benchmarks, computing 100K pairs of 1K dimension vectors takes **less than 20ms**.
+For small datasets (< 100K rows) or applications that can accept 100ms latency, vector indices are usually not necessary.
 
 For large-scale or higher dimension vectors, it is beneficial to create vector index.
 
-### How big is my index, and how many memory will it take.
+### How big is my index, and how many memory will it take?
 
-In LanceDB, all vector indices are disk-based, meaning that when responding to a vector query, only the relevant pages from the index file are loaded from disk and cached in memory. Additionally, each sub-vector is usually encoded into 1 byte PQ code.
+In LanceDB, all vector indices are **disk-based**, meaning that when responding to a vector query, only the relevant pages from the index file are loaded from disk and cached in memory. Additionally, each sub-vector is usually encoded into 1 byte PQ code.
 
 For example, with a 1024-dimension dataset, if we choose `num_sub_vectors=64`, each sub-vector has `1024 / 64 = 16` float32 numbers.
 Product quantization can lead to approximately `16 * sizeof(float32) / 1 = 64` times of space reduction.
 
-### How to choose `num_partitions` and `num_sub_vectors` for `IVF_PQ` index.
+### How to choose `num_partitions` and `num_sub_vectors` for `IVF_PQ` index?
 
 `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.
 
-`num_sub_vectors` decides how many Product Quantization code to generate on each vector. Because
-Product Quantization is a lossy compression of the original vector, the more `num_sub_vectors` usually results to
-less space distortion, and thus yield better accuracy. However, similarly, more `num_sub_vectors` causes heavier I/O and
-more PQ computation, thus, higher latency. `dimension / num_sub_vectors` should be aligned with 8 for better SIMD efficiency.
+`num_sub_vectors` specifies how many Product Quantization (PQ) short codes to generate on each vector. Because
+PQ is a lossy compression of the original vector, a higher `num_sub_vectors` usually results in
+less space distortion, and thus yields better accuracy. However, a higher `num_sub_vectors` also causes heavier I/O and
+more PQ computation, and thus, higher latency. `dimension / num_sub_vectors` should be a multiple of 8 for optimum SIMD efficiency.

docs/src/basic.md

@@ -79,6 +79,18 @@ We'll cover the basics of using LanceDB on your local machine in this section.
 
 ??? info "Under the hood, LanceDB is converting the input data into an Apache Arrow table and persisting it to disk in [Lance format](https://www.github.com/lancedb/lance)."
 
+### Creating an empty table
+
+Sometimes you may not have the data to insert into the table at creation time.
+In this case, you can create an empty table and specify the schema.
+
+=== "Python"
+      ```python
+      import pyarrow as pa
+      schema = pa.schema([pa.field("vector", pa.list_(pa.float32(), list_size=2))])
+      tbl = db.create_table("empty_table", schema=schema)
+      ```
+
 ## How to open an existing table
 
 Once created, you can open a table using the following code:
@@ -111,9 +123,15 @@ After a table has been created, you can always add more data to it using
 
 === "Python"
       ```python
-      df = pd.DataFrame([{"vector": [1.3, 1.4], "item": "fizz", "price": 100.0},
-                        {"vector": [9.5, 56.2], "item": "buzz", "price": 200.0}])
-      tbl.add(df)
+
+      # Option 1: Add a list of dicts to a table
+      data = [{"vector": [1.3, 1.4], "item": "fizz", "price": 100.0},
+            {"vector": [9.5, 56.2], "item": "buzz", "price": 200.0}]
+      tbl.add(data)
+
+      # Option 2: Add a pandas DataFrame to a table
+      df = pd.DataFrame(data)
+      tbl.add(data)
       ```
 
 === "Javascript"
@@ -122,6 +140,22 @@ After a table has been created, you can always add more data to it using
               {vector: [9.5, 56.2], item: "buzz", price: 200.0}])
       ```
 
+## How to search for (approximate) nearest neighbors
+
+Once you've embedded the query, you can find its nearest neighbors using the following code:
+
+=== "Python"
+      ```python
+      tbl.search([100, 100]).limit(2).to_pandas()
+      ```
+
+      This returns a pandas DataFrame with the results.
+
+=== "Javascript"
+      ```javascript
+      const query = await tbl.search([100, 100]).limit(2).execute();
+      ```
+
 ## How to delete rows from a table
 
 Use the `delete()` method on tables to delete rows from a table. To choose
@@ -151,24 +185,34 @@ To see what expressions are supported, see the [SQL filters](sql.md) section.
 
       Read more: [vectordb.Table.delete](javascript/interfaces/Table.md#delete)
 
-## How to search for (approximate) nearest neighbors
+## How to remove a table
 
-Once you've embedded the query, you can find its nearest neighbors using the following code:
+Use the `drop_table()` method on the database to remove a table.
 
 === "Python"
       ```python
-      tbl.search([100, 100]).limit(2).to_df()
+      db.drop_table("my_table")
       ```
 
-      This returns a pandas DataFrame with the results.
+This permanently removes the table and is not recoverable, unlike deleting rows.
+By default, if the table does not exist an exception is raised. To suppress this,
+you can pass in `ignore_missing=True`.
 
-=== "Javascript"
-      ```javascript
-      const query = await tbl.search([100, 100]).limit(2).execute();
-      ```
 
 ## What's next
 
 This section covered the very basics of the LanceDB API.
 LanceDB supports many additional features when creating indices to speed up search and options for search.
 These are contained in the next section of the documentation.
+
+## Note: Bundling vectorDB apps with webpack
+Since LanceDB contains a prebuilt Node binary, you must configure `next.config.js` to exclude it from webpack. This is required for both using Next.js and deploying on Vercel.
+```javascript
+/** @type {import('next').NextConfig} */
+module.exports = ({
+  webpack(config) {
+    config.externals.push({ vectordb: 'vectordb' })
+    return config;
+  }
+})
+```

docs/src/cli_config.md

@@ -0,0 +1,37 @@
+
+## LanceDB CLI
+Once lanceDB is installed, you can access the CLI using `lancedb` command on the console
+```
+lancedb
+```
+This lists out all the various command-line options available. You can get the usage or help for a particular command
+```
+lancedb {command} --help
+```
+
+## LanceDB config
+LanceDB uses a global config file to store certain settings. These settings are configurable using the lanceDB cli.
+To view your config settings, you can use:
+```
+lancedb config
+```
+These config parameters can be tuned using the cli.
+```
+lancedb {config_name} --{argument}
+```
+
+## LanceDB Opt-in Diagnostics
+When enabled, LanceDB will send anonymous events to help us improve LanceDB. These diagnostics are used only for error reporting and no data is collected. Error & stats allow us to automate certain aspects of bug reporting, prioritization of fixes and feature requests.
+These diagnostics are opt-in and can be enabled or disabled using the `lancedb diagnostics` command. These are enabled by default.
+Get usage help.
+```
+lancedb diagnostics --help
+```
+Disable diagnostics
+```
+lancedb diagnostics --disabled
+```
+Enable diagnostics
+```
+lancedb diagnostics --enabled
+```

docs/src/embedding.md

@@ -66,7 +66,7 @@ You can also use an external API like OpenAI to generate embeddings
     to generate embeddings for each row.
 
     Say if you have a pandas DataFrame with a `text` column that you want to be embedded,
-    you can use the [with_embeddings](https://lancedb.github.io/lancedb/python/#lancedb.embeddings.with_embeddings)
+    you can use the [with_embeddings](https://lancedb.github.io/lancedb/python/python/#lancedb.embeddings.with_embeddings)
     function to generate embeddings and add create a combined pyarrow table:
 
 
@@ -118,7 +118,7 @@ belong in the same latent space and your results will be nonsensical.
      ```python
      query = "What's the best pizza topping?"
      query_vector = embed_func([query])[0]
-     tbl.search(query_vector).limit(10).to_df()
+     tbl.search(query_vector).limit(10).to_pandas()
      ```
 
      The above snippet returns a pandas DataFrame with the 10 closest vectors to the query.

docs/src/examples/index.md

@@ -0,0 +1,23 @@
+# Examples
+
+Here are some of the examples, projects and applications using LanceDB python library. Some examples are covered in detail in the next sections. You can find more on [VectorDB Recipes](https://github.com/lancedb/vectordb-recipes)
+
+| Example | Interactive Envs | Scripts  |
+|-------- | ---------------- | ------   |
+| | | |
+| [Youtube transcript search bot](https://github.com/lancedb/vectordb-recipes/tree/main/examples/youtube_bot/) | <a href="https://colab.research.google.com/github/lancedb/vectordb-recipes/blob/main/examples/youtube_bot/main.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a>| [![Python](https://img.shields.io/badge/python-3670A0?style=for-the-badge&logo=python&logoColor=ffdd54)](https://github.com/lancedb/vectordb-recipes/tree/main/examples/youtube_bot/main.py)|
+| [Langchain: Code Docs QA bot](https://github.com/lancedb/vectordb-recipes/tree/main/examples/Code-Documentation-QA-Bot/) | <a href="https://colab.research.google.com/github/lancedb/vectordb-recipes/blob/main/examples/Code-Documentation-QA-Bot/main.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a>| [![Python](https://img.shields.io/badge/python-3670A0?style=for-the-badge&logo=python&logoColor=ffdd54)](https://github.com/lancedb/vectordb-recipes/tree/main/examples/Code-Documentation-QA-Bot/main.py) |
+| [AI Agents: Reducing Hallucination](https://github.com/lancedb/vectordb-recipes/tree/main/examples/reducing_hallucinations_ai_agents/) | <a href="https://colab.research.google.com/github/lancedb/vectordb-recipes/blob/main/examples/reducing_hallucinations_ai_agents/main.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a>| [![Python](https://img.shields.io/badge/python-3670A0?style=for-the-badge&logo=python&logoColor=ffdd54)](https://github.com/lancedb/vectordb-recipes/tree/main/examples/reducing_hallucinations_ai_agents/main.py)|
+| [Multimodal CLIP: DiffusionDB](https://github.com/lancedb/vectordb-recipes/tree/main/examples/multimodal_clip/) | <a href="https://colab.research.google.com/github/lancedb/vectordb-recipes/blob/main/examples/multimodal_clip/main.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a>| [![Python](https://img.shields.io/badge/python-3670A0?style=for-the-badge&logo=python&logoColor=ffdd54)](https://github.com/lancedb/vectordb-recipes/tree/main/examples/multimodal_clip/main.py)  |
+| [Multimodal CLIP: Youtube videos](https://github.com/lancedb/vectordb-recipes/tree/main/examples/multimodal_video_search/) | <a href="https://colab.research.google.com/github/lancedb/vectordb-recipes/blob/main/examples/multimodal_video_search/main.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a>| [![Python](https://img.shields.io/badge/python-3670A0?style=for-the-badge&logo=python&logoColor=ffdd54)](https://github.com/lancedb/vectordb-recipes/tree/main/examples/multimodal_video_search/main.py)  |
+| [Movie Recommender](https://github.com/lancedb/vectordb-recipes/tree/main/examples/movie-recommender/) | <a href="https://colab.research.google.com/github/lancedb/vectordb-recipes/blob/main/examples/movie-recommender/main.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> | [![Python](https://img.shields.io/badge/python-3670A0?style=for-the-badge&logo=python&logoColor=ffdd54)](https://github.com/lancedb/vectordb-recipes/tree/main/examples/movie-recommender/main.py)  |
+| [Audio Search](https://github.com/lancedb/vectordb-recipes/tree/main/examples/audio_search/) | <a href="https://colab.research.google.com/github/lancedb/vectordb-recipes/blob/main/examples/audio_search/main.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> | [![Python](https://img.shields.io/badge/python-3670A0?style=for-the-badge&logo=python&logoColor=ffdd54)](https://github.com/lancedb/vectordb-recipes/tree/main/examples/audio_search/main.py)  |
+| [Multimodal Image + Text Search](https://github.com/lancedb/vectordb-recipes/tree/main/examples/multimodal_search/) | <a href="https://colab.research.google.com/github/lancedb/vectordb-recipes/blob/main/examples/multimodal_search/main.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> | [![Python](https://img.shields.io/badge/python-3670A0?style=for-the-badge&logo=python&logoColor=ffdd54)](https://github.com/lancedb/vectordb-recipes/tree/main/examples/multimodal_search/main.py)  |
+| [Evaluating Prompts with Prompttools](https://github.com/lancedb/vectordb-recipes/tree/main/examples/prompttools-eval-prompts/) | <a href="https://colab.research.google.com/github/lancedb/vectordb-recipes/blob/main/examples/prompttools-eval-prompts/main.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a> |  |
+
+## Projects & Applications powered by LanceDB
+
+| Project Name                                        | Description                                                                                                          | Screenshot                                |
+|-----------------------------------------------------|----------------------------------------------------------------------------------------------------------------------|-------------------------------------------|
+| [YOLOExplorer](https://github.com/lancedb/yoloexplorer) | Iterate on your YOLO / CV datasets using SQL, Vector semantic search, and more within seconds                  | ![YOLOExplorer](https://github.com/lancedb/vectordb-recipes/assets/15766192/ae513a29-8f15-4e0b-99a1-ccd8272b6131) |
+| [Website Chatbot (Deployable Vercel Template)](https://github.com/lancedb/lancedb-vercel-chatbot) | Create a chatbot from the sitemap of any website/docs of your choice. Built using vectorDB serverless native javascript package. | ![Chatbot](../assets/vercel-template.gif)    |

docs/src/examples/index_js.md

@@ -0,0 +1,19 @@
+# Examples
+
+Here are some of the examples, projects and applications using vectordb native javascript library.
+Some examples are covered in detail in the next sections. You can find more on [VectorDB Recipes](https://github.com/lancedb/vectordb-recipes)
+
+| Example | Scripts  |
+|-------- | ------   |
+| | |
+| [Youtube transcript search bot](https://github.com/lancedb/vectordb-recipes/tree/main/examples/youtube_bot/) | [![JavaScript](https://img.shields.io/badge/javascript-%23323330.svg?style=for-the-badge&logo=javascript&logoColor=%23F7DF1E)](https://github.com/lancedb/vectordb-recipes/tree/main/examples/youtube_bot/index.js)|
+| [Langchain: Code Docs QA bot](https://github.com/lancedb/vectordb-recipes/tree/main/examples/Code-Documentation-QA-Bot/) | [![JavaScript](https://img.shields.io/badge/javascript-%23323330.svg?style=for-the-badge&logo=javascript&logoColor=%23F7DF1E)](https://github.com/lancedb/vectordb-recipes/tree/main/examples/Code-Documentation-QA-Bot/index.js)|
+| [AI Agents: Reducing Hallucination](https://github.com/lancedb/vectordb-recipes/tree/main/examples/reducing_hallucinations_ai_agents/) | [![JavaScript](https://img.shields.io/badge/javascript-%23323330.svg?style=for-the-badge&logo=javascript&logoColor=%23F7DF1E)](https://github.com/lancedb/vectordb-recipes/tree/main/examples/reducing_hallucinations_ai_agents/index.js)|
+| [TransformersJS Embedding example](https://github.com/lancedb/vectordb-recipes/tree/main/examples/js-transformers/) | [![JavaScript](https://img.shields.io/badge/javascript-%23323330.svg?style=for-the-badge&logo=javascript&logoColor=%23F7DF1E)](https://github.com/lancedb/vectordb-recipes/tree/main/examples/js-transformers/index.js)  |
+
+## Projects & Applications
+
+| Project Name                                        | Description                                                                                                          | Screenshot                                |
+|-----------------------------------------------------|----------------------------------------------------------------------------------------------------------------------|-------------------------------------------|
+| [YOLOExplorer](https://github.com/lancedb/yoloexplorer) | Iterate on your YOLO / CV datasets using SQL, Vector semantic search, and more within seconds                  | ![YOLOExplorer](https://github.com/lancedb/vectordb-recipes/assets/15766192/ae513a29-8f15-4e0b-99a1-ccd8272b6131) |
+| [Website Chatbot (Deployable Vercel Template)](https://github.com/lancedb/lancedb-vercel-chatbot) | Create a chatbot from the sitemap of any website/docs of your choice. Built using vectorDB serverless native javascript package. | ![Chatbot](../assets/vercel-template.gif)    |

docs/src/examples/serverless_lancedb_with_s3_and_lambda.md

@@ -80,14 +80,14 @@ def handler(event, context):
     # Shape of SIFT is (128,1M), d=float32
     query_vector = np.array(event['query_vector'], dtype=np.float32)
 
-    rs = table.search(query_vector).limit(2).to_df()
+    rs = table.search(query_vector).limit(2).to_list()
 
     return {
         "statusCode": status_code,
         "headers": {
             "Content-Type": "application/json"
         },
-        "body": rs.to_json()
+        "body": json.dumps(rs)
     }
 ``` 
 

docs/src/examples/serverless_website_chatbot.md

@@ -0,0 +1,61 @@
+# LanceDB Chatbot - Vercel Next.js Template
+Use an AI chatbot with website context retrieved from a vector store like LanceDB. LanceDB is lightweight and can be embedded directly into Next.js, with data stored on-prem.
+
+## One click deploy on Vercel
+[![Deploy with Vercel](https://vercel.com/button)](https://vercel.com/new/clone?repository-url=https%3A%2F%2Fgithub.com%2Flancedb%2Flancedb-vercel-chatbot&env=OPENAI_API_KEY&envDescription=OpenAI%20API%20Key%20for%20chat%20completion.&project-name=lancedb-vercel-chatbot&repository-name=lancedb-vercel-chatbot&demo-title=LanceDB%20Chatbot%20Demo&demo-description=Demo%20website%20chatbot%20with%20LanceDB.&demo-url=https%3A%2F%2Flancedb.vercel.app&demo-image=https%3A%2F%2Fi.imgur.com%2FazVJtvr.png)
+
+![Demo website landing page](../assets/vercel-template.gif)
+
+## Development
+
+First, rename `.env.example` to `.env.local`, and fill out `OPENAI_API_KEY` with your OpenAI API key. You can get one [here](https://openai.com/blog/openai-api).
+
+Run the development server:
+
+```bash
+npm run dev
+# or
+yarn dev
+# or
+pnpm dev
+```
+
+Open [http://localhost:3000](http://localhost:3000) with your browser to see the result.
+
+This project uses [`next/font`](https://nextjs.org/docs/basic-features/font-optimization) to automatically optimize and load Inter, a custom Google Font.
+
+## Learn More
+
+To learn more about LanceDB or Next.js, take a look at the following resources:
+
+- [LanceDB Documentation](https://lancedb.github.io/lancedb/) - learn about LanceDB, the developer-friendly serverless vector database.
+- [Next.js Documentation](https://nextjs.org/docs) - learn about Next.js features and API.
+- [Learn Next.js](https://nextjs.org/learn) - an interactive Next.js tutorial.
+
+## LanceDB on Next.js and Vercel
+
+FYI: these configurations have been pre-implemented in this template.
+
+Since LanceDB contains a prebuilt Node binary, you must configure `next.config.js` to exclude it from webpack. This is required for both using Next.js and deploying on Vercel.
+```js
+/** @type {import('next').NextConfig} */
+module.exports = ({
+  webpack(config) {
+    config.externals.push({ vectordb: 'vectordb' })
+    return config;
+  }
+})
+```
+
+To deploy on Vercel, we need to make sure that the NodeJS runtime static file analysis for Vercel can find the binary, since LanceDB uses dynamic imports by default. We can do this by modifying `package.json` in the `scripts` section.
+```json
+{
+  ...
+  "scripts": {
+    ...
+    "vercel-build": "sed -i 's/nativeLib = require(`@lancedb\\/vectordb-\\${currentTarget()}`);/nativeLib = require(`@lancedb\\/vectordb-linux-x64-gnu`);/' node_modules/vectordb/native.js && next build",
+    ...
+  },
+  ...
+}
+```

docs/src/examples/transformerjs_embedding_search_nodejs.md

@@ -99,7 +99,7 @@ Output of `results`:
     id: 5,
     text: 'Banana',
     type: 'fruit',
-    score: 0.4919965863227844
+    _distance: 0.4919965863227844
   },
   {
     vector: Float32Array(384) [
@@ -111,7 +111,7 @@ Output of `results`:
     id: 1,
     text: 'Cherry',
     type: 'fruit',
-    score: 0.5540297031402588
+    _distance: 0.5540297031402588
   }
 ]
 ```

docs/src/examples/youtube_transcript_bot.md

@@ -7,7 +7,7 @@
 
 <a href="https://colab.research.google.com/github/lancedb/vectordb-recipes/blob/main/examples/youtube_bot/main.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab">
 
-Scripts - [![Python](https://img.shields.io/badge/python-3670A0?style=for-the-badge&logo=python&logoColor=ffdd54)](./examples/youtube_bot/main.py)  [![JavaScript](https://img.shields.io/badge/javascript-%23323330.svg?style=for-the-badge&logo=javascript&logoColor=%23F7DF1E)](./examples/youtube_bot/index.js)
+Scripts - [![Python](https://img.shields.io/badge/python-3670A0?style=for-the-badge&logo=python&logoColor=ffdd54)](https://github.com/lancedb/vectordb-recipesexamples/youtube_bot/main.py)  [![JavaScript](https://img.shields.io/badge/javascript-%23323330.svg?style=for-the-badge&logo=javascript&logoColor=%23F7DF1E)](https://github.com/lancedb/vectordb-recipes/examples/youtube_bot/index.js)
 
 
 This example is in a [notebook](https://github.com/lancedb/lancedb/blob/main/docs/src/notebooks/youtube_transcript_search.ipynb)

docs/src/fts.md

@@ -6,17 +6,19 @@ to make this available for JS as well.
 
 ## Installation
 
-To use full text search, you must install optional dependency tantivy-py:
+To use full text search, you must install the dependency `tantivy-py`:
 
-# tantivy 0.19.2
-pip install tantivy@git+https://github.com/quickwit-oss/tantivy-py#164adc87e1a033117001cf70e38c82a53014d985
+# tantivy 0.20.1
+```sh
+pip install tantivy==0.20.1
+```
 
 
 ## Quickstart
 
 Assume:
 1. `table` is a LanceDB Table
-2. `text` is the name of the Table column that we want to index
+2. `text` is the name of the `Table` column that we want to index
 
 For example,
 
@@ -41,7 +43,13 @@ table.create_fts_index("text")
 To search:
 
 ```python
-df = table.search("puppy").limit(10).select(["text"]).to_df()
+table.search("puppy").limit(10).select(["text"]).to_list()
+```
+
+Which returns a list of dictionaries:
+
+```python
+[{'text': 'Frodo was a happy puppy', 'score': 0.6931471824645996}]
 ```
 
 LanceDB automatically looks for an FTS index if the input is str.

docs/src/guides/tables.md

@@ -0,0 +1,411 @@
+ <a href="https://colab.research.google.com/github/lancedb/lancedb/blob/main/docs/src/notebooks/tables_guide.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a><br/>
+A Table is a collection of Records in a LanceDB Database. You can follow along on colab!
+
+## Creating a LanceDB Table
+
+=== "Python"
+    ### LanceDB Connection
+
+    ```python
+    import lancedb
+    db = lancedb.connect("./.lancedb")
+    ```
+
+    LanceDB allows ingesting data from various sources - `dict`, `list[dict]`, `pd.DataFrame`, `pa.Table` or a `Iterator[pa.RecordBatch]`. Let's take a look at some of the these.
+
+    ### From list of tuples or dictionaries
+
+    ```python
+    import lancedb
+
+    db = lancedb.connect("./.lancedb")
+
+    data = [{"vector": [1.1, 1.2], "lat": 45.5, "long": -122.7},
+            {"vector": [0.2, 1.8], "lat": 40.1, "long": -74.1}]
+
+    db.create_table("my_table", data)
+
+    db["my_table"].head()
+    ```
+
+    !!! info "Note"
+        If the table already exists, LanceDB will raise an error by default. If you want to overwrite the table, you can pass in mode="overwrite" to the createTable function.
+
+        ```python
+        db.create_table("name", data, mode="overwrite")
+        ```
+
+
+    ### From pandas DataFrame
+
+    ```python
+    import pandas as pd
+
+    data = pd.DataFrame({
+        "vector": [[1.1, 1.2, 1.3, 1.4], [0.2, 1.8, 0.4, 3.6]],
+        "lat": [45.5, 40.1],
+        "long": [-122.7, -74.1]
+    })
+
+    db.create_table("table2", data)
+
+    db["table2"].head()
+    ```
+    !!! info "Note"
+        Data is converted to Arrow before being written to disk. For maximum control over how data is saved, either provide the PyArrow schema to convert to or else provide a PyArrow Table directly.
+
+    ```python
+    custom_schema = pa.schema([
+    pa.field("vector", pa.list_(pa.float32(), 4)),
+    pa.field("lat", pa.float32()),
+    pa.field("long", pa.float32())
+    ])
+
+    table = db.create_table("table3", data, schema=custom_schema)
+    ```
+
+    ### From PyArrow Tables
+    You can also create LanceDB tables directly from pyarrow tables
+
+    ```python
+    table = pa.Table.from_arrays(
+            [
+                pa.array([[3.1, 4.1, 5.1, 6.1], [5.9, 26.5, 4.7, 32.8]],
+                        pa.list_(pa.float32(), 4)),
+                pa.array(["foo", "bar"]),
+                pa.array([10.0, 20.0]),
+            ],
+            ["vector", "item", "price"],
+        )
+
+    db = lancedb.connect("db")
+
+    tbl = db.create_table("test1", table)
+    ```
+
+    ### From Pydantic Models
+    When you create an empty table without data, you must specify the table schema.
+    LanceDB supports creating tables by specifying a pyarrow schema or a specialized
+    pydantic model called `LanceModel`.
+
+    For example, the following Content model specifies a table with 5 columns:
+    movie_id, vector, genres, title, and imdb_id. When you create a table, you can
+    pass the class as the value of the `schema` parameter to `create_table`.
+    The `vector` column is a `Vector` type, which is a specialized pydantic type that
+    can be configured with the vector dimensions. It is also important to note that
+    LanceDB only understands subclasses of `lancedb.pydantic.LanceModel`
+    (which itself derives from `pydantic.BaseModel`).
+
+    ```python
+    from lancedb.pydantic import Vector, LanceModel
+
+    class Content(LanceModel):
+        movie_id: int
+        vector: Vector(128)
+        genres: str
+        title: str
+        imdb_id: int
+
+        @property
+        def imdb_url(self) -> str:
+            return f"https://www.imdb.com/title/tt{self.imdb_id}"
+
+    import pyarrow as pa
+    db = lancedb.connect("~/.lancedb")
+    table_name = "movielens_small"
+    table = db.create_table(table_name, schema=Content)
+    ```
+
+    ### Using Iterators / Writing Large Datasets
+
+    It is recommended to use itertators to add large datasets in batches when creating your table in one go. This does not create multiple versions of your dataset unlike manually adding batches using `table.add()`
+
+    LanceDB additionally supports pyarrow's `RecordBatch` Iterators or other generators producing supported data types.
+
+    Here's an example using using `RecordBatch` iterator for creating tables.
+
+    ```python
+    import pyarrow as pa
+
+    def make_batches():
+        for i in range(5):
+            yield pa.RecordBatch.from_arrays(
+                [
+                    pa.array([[3.1, 4.1, 5.1, 6.1], [5.9, 26.5, 4.7, 32.8]],
+                            pa.list_(pa.float32(), 4)),
+                    pa.array(["foo", "bar"]),
+                    pa.array([10.0, 20.0]),
+                ],
+                ["vector", "item", "price"],
+            )
+
+    schema = pa.schema([
+        pa.field("vector", pa.list_(pa.float32(), 4)),
+        pa.field("item", pa.utf8()),
+        pa.field("price", pa.float32()),
+    ])
+
+    db.create_table("table4", make_batches(), schema=schema)
+    ```
+
+    You can also use iterators of other types like Pandas dataframe or Pylists directly in the above example.
+
+    ## Creating Empty Table
+    You can also create empty tables in python. Initialize it with schema and later ingest data into it.
+
+    ```python
+    import lancedb
+    import pyarrow as pa
+
+    schema = pa.schema(
+      [
+          pa.field("vector", pa.list_(pa.float32(), 2)),
+          pa.field("item", pa.string()),
+          pa.field("price", pa.float32()),
+      ])
+    tbl = db.create_table("table5", schema=schema)
+    data = [
+        {"vector": [3.1, 4.1], "item": "foo", "price": 10.0},
+        {"vector": [5.9, 26.5], "item": "bar", "price": 20.0},
+    ]
+    tbl.add(data=data)
+    ```
+
+    You can also use Pydantic to specify the schema
+
+    ```python
+    import lancedb
+    from lancedb.pydantic import LanceModel, vector
+
+    class Model(LanceModel):
+          vector: Vector(2)
+
+    tbl = db.create_table("table5", schema=Model.to_arrow_schema())
+    ```
+
+=== "Javascript/Typescript"
+
+    ### VectorDB Connection
+
+    ```javascript
+    const lancedb = require("vectordb");
+
+    const uri = "data/sample-lancedb";
+    const db = await lancedb.connect(uri);
+    ```
+
+    ### Creating a Table
+
+    You can create a LanceDB table in javascript using an array of records.
+
+    ```javascript
+    data
+    const tb = await db.createTable("my_table",
+                  data=[{"vector": [3.1, 4.1], "item": "foo", "price": 10.0},
+                        {"vector": [5.9, 26.5], "item": "bar", "price": 20.0}])
+    ```
+
+    !!! info "Note"
+    If the table already exists, LanceDB will raise an error by default. If you want to overwrite the table, you need to specify the `WriteMode` in the createTable function.
+
+    ```javascript
+    const table = await con.createTable(tableName, data, { writeMode: WriteMode.Overwrite })
+    ```
+
+## Open existing tables
+
+If you forget the name of your table, you can always get a listing of all table names:
+
+
+=== "Python"
+    ### Get a list of existing Tables
+
+    ```python
+    print(db.table_names())
+    ```
+=== "Javascript/Typescript"
+
+    ```javascript
+    console.log(await db.tableNames());
+    ```
+
+Then, you can open any existing tables
+
+=== "Python"
+
+    ```python
+    tbl = db.open_table("my_table")
+    ```
+=== "Javascript/Typescript"
+
+    ```javascript
+    const tbl = await db.openTable("my_table");
+    ```
+
+## Adding to a Table
+After a table has been created, you can always add more data to it using
+
+=== "Python"
+    You can add any of the valid data structures accepted by LanceDB table, i.e, `dict`, `list[dict]`, `pd.DataFrame`, or a `Iterator[pa.RecordBatch]`. Here are some examples.
+
+    ### Adding Pandas DataFrame
+
+    ```python
+    df = pd.DataFrame([{"vector": [1.3, 1.4], "item": "fizz", "price": 100.0},
+                  {"vector": [9.5, 56.2], "item": "buzz", "price": 200.0}])
+    tbl.add(df)
+    ```
+
+    You can also add a large dataset batch in one go using Iterator of any supported data types.
+
+    ### Adding to table using Iterator
+
+    ```python
+    import pandas as pd
+
+    def make_batches():
+        for i in range(5):
+            yield pd.DataFrame(
+                {
+                    "vector": [[3.1, 4.1], [1, 1]],
+                    "item": ["foo", "bar"],
+                    "price": [10.0, 20.0],
+                })
+
+    tbl.add(make_batches())
+    ```
+
+    The other arguments accepted:
+
+    | Name | Type | Description | Default |
+    |---|---|---|---|
+    | data | DATA | The data to insert into the table. | required |
+    | mode | str | The mode to use when writing the data. Valid values are "append" and "overwrite". | append |
+    | on_bad_vectors | str | What to do if any of the vectors are not the same size or contains NaNs. One of "error", "drop", "fill". | drop |
+    | fill value | float | The value to use when filling vectors: Only used if on_bad_vectors="fill". | 0.0 |
+
+
+=== "Javascript/Typescript"
+
+    ```javascript
+    await tbl.add([{vector: [1.3, 1.4], item: "fizz", price: 100.0},
+        {vector: [9.5, 56.2], item: "buzz", price: 200.0}])
+    ```
+
+## Deleting from a Table
+
+Use the `delete()` method on tables to delete rows from a table. To choose which rows to delete, provide a filter that matches on the metadata columns. This can delete any number of rows that match the filter.
+
+=== "Python"
+
+    ```python
+    tbl.delete('item = "fizz"')
+    ```
+
+    ### Deleting row with specific column value
+
+    ```python
+    import lancedb
+    import pandas as pd
+
+    data = pd.DataFrame({"x": [1, 2, 3], "vector": [[1, 2], [3, 4], [5, 6]]})
+    db = lancedb.connect("./.lancedb")
+    table = db.create_table("my_table", data)
+    table.to_pandas()
+    #   x      vector
+    # 0  1  [1.0, 2.0]
+    # 1  2  [3.0, 4.0]
+    # 2  3  [5.0, 6.0]
+
+    table.delete("x = 2")
+    table.to_pandas()
+    #   x      vector
+    # 0  1  [1.0, 2.0]
+    # 1  3  [5.0, 6.0]
+    ```
+
+    ### Delete from a list of values
+
+    ```python
+    to_remove = [1, 5]
+    to_remove = ", ".join(str(v) for v in to_remove)
+
+    table.delete(f"x IN ({to_remove})")
+    table.to_pandas()
+    #   x      vector
+    # 0  3  [5.0, 6.0]
+    ```
+
+=== "Javascript/Typescript"
+
+    ```javascript
+    await tbl.delete('item = "fizz"')
+    ```
+
+    ### Deleting row with specific column value
+
+    ```javascript
+    const con = await lancedb.connect("./.lancedb")
+    const data = [
+      {id: 1, vector: [1, 2]},
+      {id: 2, vector: [3, 4]},
+      {id: 3, vector: [5, 6]},
+    ];
+    const tbl = await con.createTable("my_table", data)
+    await tbl.delete("id = 2")
+    await tbl.countRows() // Returns 2
+    ```
+
+    ### Delete from a list of values
+
+    ```javascript
+    const to_remove = [1, 5];
+    await tbl.delete(`id IN (${to_remove.join(",")})`)
+    await tbl.countRows() // Returns 1
+    ```
+
+### Updating a Table [Experimental]
+EXPERIMENTAL: Update rows in the table (not threadsafe).
+
+This can be used to update zero to all rows depending on how many rows match the where clause.
+
+| Parameter | Type | Description |
+|---|---|---|
+| `where` | `str` | The SQL where clause to use when updating rows. For example, `'x = 2'` or `'x IN (1, 2, 3)'`. The filter must not be empty, or it will error. |
+| `values` | `dict` | The values to update. The keys are the column names and the values are the values to set. |
+
+
+=== "Python"
+
+    ```python
+    import lancedb
+    import pandas as pd
+
+    # Create a lancedb connection
+    db = lancedb.connect("./.lancedb")
+
+    # Create a table from a pandas DataFrame
+    data = pd.DataFrame({"x": [1, 2, 3], "vector": [[1, 2], [3, 4], [5, 6]]})
+    table = db.create_table("my_table", data)
+
+    # Update the table where x = 2
+    table.update(where="x = 2", values={"vector": [10, 10]})
+
+    # Get the updated table as a pandas DataFrame
+    df = table.to_pandas()
+
+    # Print the DataFrame
+    print(df)
+    ```
+
+    Output
+    ```shell
+        x  vector
+    0  1  [1.0, 2.0]
+    1  3  [5.0, 6.0]
+    2  2  [10.0, 10.0]
+    ```
+
+## What's Next?
+
+Learn how to Query your tables and create indices

docs/src/index.md

@@ -1,20 +1,23 @@
-# Welcome to LanceDB's Documentation
+# LanceDB
 
-LanceDB is an open-source database for vector-search built with persistent storage, which greatly simplifies retrevial, filtering and management of embeddings.
+LanceDB is an open-source database for vector-search built with persistent storage, which greatly simplifies retrieval, filtering and management of embeddings.
+
+![Illustration](/lancedb/assets/ecosystem-illustration.png)
 
 The key features of LanceDB include:
 
-* Production-scale vector search with no servers to manage.
-
 * Store, query and filter vectors, metadata and multi-modal data (text, images, videos, point clouds, and more).
 
-* Support for vector similarity search, full-text search and SQL.
+* Support for production-scale vector similarity search, full-text search and SQL, with no servers to manage.
 
 * Native Python and Javascript/Typescript support.
 
 * Zero-copy, automatic versioning, manage versions of your data without needing extra infrastructure.
 
-* Ecosystem integrations with [LangChain 🦜️🔗](https://python.langchain.com/en/latest/modules/indexes/vectorstores/examples/lancedb.html), [LlamaIndex 🦙](https://gpt-index.readthedocs.io/en/latest/examples/vector_stores/LanceDBIndexDemo.html), Apache-Arrow, Pandas, Polars, DuckDB and more on the way.
+* Persisted on HDD, allowing scalability without breaking the bank.
+
+* Ingest your favorite data formats directly, like pandas DataFrames, Pydantic objects and more.
+
 
 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.
 
@@ -33,7 +36,7 @@ LanceDB's core is written in Rust 🦀 and is built using <a href="https://githu
       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_df()
+      result = table.search([100, 100]).limit(2).to_list()
       ```
 
 === "Javascript"
@@ -69,4 +72,4 @@ LanceDB's core is written in Rust 🦀 and is built using <a href="https://githu
 * [`Full text search`](fts.md) - [EXPERIMENTAL] full-text search API
 * [`Ecosystem Integrations`](python/integration.md) - integrating LanceDB with python data tooling ecosystem.
 * [`Python API Reference`](python/python.md) - detailed documentation for the LanceDB Python SDK.
-* [`Node API Reference`](javascript/modules.md) - detailed documentation for the LanceDB Python SDK.
+* [`Node API Reference`](javascript/modules.md) - detailed documentation for the LanceDB Node SDK.

docs/src/integrations/index.md

@@ -0,0 +1,21 @@
+# Integrations
+
+## Data Formats
+
+LanceDB supports ingesting from your favorite data tools.
+
+![Illustration](/lancedb/assets/ecosystem-illustration.png)
+
+
+## Tools
+
+LanceDB is integrated with most of the popular AI tools, with more coming soon.
+Get started using these examples and quick links.
+
+| Integrations | |
+|---|---:|
+| <h3> LlamaIndex </h3>LlamaIndex is a simple, flexible data framework for connecting custom data sources to large language models. Llama index integrates with LanceDB as the serverless VectorDB. <h3>[Lean More](https://gpt-index.readthedocs.io/en/latest/examples/vector_stores/LanceDBIndexDemo.html) </h3> |<img src="../assets/llama-index.jpg" alt="image" width="150" height="auto">|
+| <h3>Langchain</h3>Langchain allows building applications with LLMs through composability <h3>[Lean More](https://python.langchain.com/en/latest/modules/indexes/vectorstores/examples/lancedb.html) | <img src="../assets/langchain.png" alt="image" width="150" height="auto">|
+| <h3>Langchain TS</h3> Javascript bindings for Langchain. It integrates with LanceDB's serverless vectordb allowing you to build powerful AI applications through composibility using only serverless functions. <h3>[Learn More]( https://js.langchain.com/docs/modules/data_connection/vectorstores/integrations/lancedb) | <img src="../assets/langchain.png" alt="image" width="150" height="auto">|
+| <h3>Voxel51</h3>  It is an open source toolkit that enables you to build better computer vision workflows by improving the quality of your datasets and delivering insights about your models.<h3>[Learn More](./voxel51.md) | <img src="../assets/voxel.gif" alt="image" width="150" height="auto">|
+| <h3>PromptTools</h3>  Offers a set of free, open-source tools for testing and experimenting with models, prompts, and configurations. The core idea is to enable developers to evaluate prompts using familiar interfaces like code and notebooks. You can use it to experiment with different configurations of LanceDB, and test how LanceDB integrates with the LLM of your choice.<h3>[Learn More](./prompttools.md) | <img src="../assets/prompttools.jpeg" alt="image" width="150" height="auto">|

docs/src/integrations/prompttools.md

@@ -0,0 +1,7 @@
+
+[PromptTools](https://github.com/hegelai/prompttools) offers a set of free, open-source tools for testing and experimenting with models, prompts, and configurations. The core idea is to enable developers to evaluate prompts using familiar interfaces like code and notebooks. You can use it to experiment with different configurations of LanceDB, and test how LanceDB integrates with the LLM of your choice.
+
+[Evaluating Prompts with PromptTools](./examples/prompttools-eval-prompts/) | <a href="https://colab.research.google.com/github/lancedb/vectordb-recipes/blob/main/examples/prompttools-eval-prompts/main.ipynb"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a>
+
+![Alt text](https://prompttools.readthedocs.io/en/latest/_images/demo.gif "a title")
+

docs/src/notebooks/code_qa_bot.ipynb

@@ -144,7 +144,7 @@
    "source": [
     "# Pre-processing and loading the documentation\n",
     "\n",
-    "Next, let's pre-process and load the documentation. To make sure we don't need to do this repeatedly if we were updating code, we're caching it using pickle so we can retrieve it again (this could take a few minutes to run the first time yyou do it). We'll also add some more metadata to the docs here such as the title and version of the code:"
+    "Next, let's pre-process and load the documentation. To make sure we don't need to do this repeatedly if we were updating code, we're caching it using pickle so we can retrieve it again (this could take a few minutes to run the first time you do it). We'll also add some more metadata to the docs here such as the title and version of the code:"
    ]
   },
   {
@@ -255,7 +255,7 @@
    "id": "28d93b85",
    "metadata": {},
    "source": [
-    "And thats it! We're all setup. The next step is to run some queries, let's try a few:"
+    "And that's it! We're all set up. The next step is to run some queries, let's try a few:"
    ]
   },
   {

docs/src/notebooks/multimodal_search.ipynb

@@ -19,11 +19,11 @@
      "output_type": "stream",
      "text": [
       "\n",
-      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m A new release of pip available: \u001b[0m\u001b[31;49m22.3.1\u001b[0m\u001b[39;49m -> \u001b[0m\u001b[32;49m23.1.2\u001b[0m\n",
-      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m To update, run: \u001b[0m\u001b[32;49mpip install --upgrade pip\u001b[0m\n",
+      "\u001B[1m[\u001B[0m\u001B[34;49mnotice\u001B[0m\u001B[1;39;49m]\u001B[0m\u001B[39;49m A new release of pip available: \u001B[0m\u001B[31;49m22.3.1\u001B[0m\u001B[39;49m -> \u001B[0m\u001B[32;49m23.1.2\u001B[0m\n",
+      "\u001B[1m[\u001B[0m\u001B[34;49mnotice\u001B[0m\u001B[1;39;49m]\u001B[0m\u001B[39;49m To update, run: \u001B[0m\u001B[32;49mpip install --upgrade pip\u001B[0m\n",
       "\n",
-      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m A new release of pip available: \u001b[0m\u001b[31;49m22.3.1\u001b[0m\u001b[39;49m -> \u001b[0m\u001b[32;49m23.1.2\u001b[0m\n",
-      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m To update, run: \u001b[0m\u001b[32;49mpip install --upgrade pip\u001b[0m\n"
+      "\u001B[1m[\u001B[0m\u001B[34;49mnotice\u001B[0m\u001B[1;39;49m]\u001B[0m\u001B[39;49m A new release of pip available: \u001B[0m\u001B[31;49m22.3.1\u001B[0m\u001B[39;49m -> \u001B[0m\u001B[32;49m23.1.2\u001B[0m\n",
+      "\u001B[1m[\u001B[0m\u001B[34;49mnotice\u001B[0m\u001B[1;39;49m]\u001B[0m\u001B[39;49m To update, run: \u001B[0m\u001B[32;49mpip install --upgrade pip\u001B[0m\n"
      ]
     }
    ],
@@ -39,6 +39,7 @@
    "outputs": [],
    "source": [
     "import io\n",
+    "\n",
     "import PIL\n",
     "import duckdb\n",
     "import lancedb"
@@ -158,18 +159,18 @@
     "        \"db = lancedb.connect('~/datasets/demo')\\n\"\n",
     "        \"tbl = db.open_table('diffusiondb')\\n\\n\"\n",
     "        f\"embedding = embed_func('{query}')\\n\"\n",
-    "        \"tbl.search(embedding).limit(9).to_df()\"\n",
+    "        \"tbl.search(embedding).limit(9).to_pandas()\"\n",
     "    )\n",
-    "    return (_extract(tbl.search(emb).limit(9).to_df()), code)\n",
+    "    return (_extract(tbl.search(emb).limit(9).to_pandas()), code)\n",
     "\n",
     "def find_image_keywords(query):\n",
     "    code = (\n",
     "        \"import lancedb\\n\"\n",
     "        \"db = lancedb.connect('~/datasets/demo')\\n\"\n",
     "        \"tbl = db.open_table('diffusiondb')\\n\\n\"\n",
-    "        f\"tbl.search('{query}').limit(9).to_df()\"\n",
+    "        f\"tbl.search('{query}').limit(9).to_pandas()\"\n",
     "    )\n",
-    "    return (_extract(tbl.search(query).limit(9).to_df()), code)\n",
+    "    return (_extract(tbl.search(query).limit(9).to_pandas()), code)\n",
     "\n",
     "def find_image_sql(query):\n",
     "    code = (\n",

docs/src/notebooks/tables_guide.ipynb

@@ -0,0 +1,831 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "d24eb4c6-e246-44ca-ba7c-6eae7923bd4c",
+   "metadata": {},
+   "source": [
+    "## LanceDB Tables\n",
+    "A Table is a collection of Records in a LanceDB Database.\n",
+    "\n",
+    "![illustration](../assets/ecosystem-illustration.png)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 50,
+   "id": "c1b4e34b-a49c-471d-a343-a5940bb5138a",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "!pip install lancedb -qq"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "4e5a8d07-d9a1-48c1-913a-8e0629289579",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import lancedb\n",
+    "db = lancedb.connect(\"./.lancedb\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "66fb93d5-3551-406b-99b2-488442d61d06",
+   "metadata": {},
+   "source": [
+    "LanceDB allows ingesting data from various sources - `dict`, `list[dict]`, `pd.DataFrame`, `pa.Table` or a `Iterator[pa.RecordBatch]`. Let's take a look at some of the these.\n",
+    "\n",
+    " ### From list of tuples or dictionaries"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "5df12f66-8d99-43ad-8d0b-22189ec0a6b9",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "pyarrow.Table\n",
+       "vector: fixed_size_list<item: float>[2]\n",
+       "  child 0, item: float\n",
+       "lat: double\n",
+       "long: double\n",
+       "----\n",
+       "vector: [[[1.1,1.2],[0.2,1.8]]]\n",
+       "lat: [[45.5,40.1]]\n",
+       "long: [[-122.7,-74.1]]"
+      ]
+     },
+     "execution_count": 2,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "import lancedb\n",
+    "\n",
+    "db = lancedb.connect(\"./.lancedb\")\n",
+    "\n",
+    "data = [{\"vector\": [1.1, 1.2], \"lat\": 45.5, \"long\": -122.7},\n",
+    "        {\"vector\": [0.2, 1.8], \"lat\": 40.1, \"long\": -74.1}]\n",
+    "\n",
+    "db.create_table(\"my_table\", data)\n",
+    "\n",
+    "db[\"my_table\"].head()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "10ce802f-1a10-49ee-8ee3-a9bfb302d86c",
+   "metadata": {},
+   "source": [
+    "## From pandas DataFrame\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "id": "f4d87ae9-0ccb-48eb-b31d-bb8f2370e47e",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "pyarrow.Table\n",
+       "vector: fixed_size_list<item: float>[2]\n",
+       "  child 0, item: float\n",
+       "lat: double\n",
+       "long: double\n",
+       "----\n",
+       "vector: [[[1.1,1.2],[0.2,1.8]]]\n",
+       "lat: [[45.5,40.1]]\n",
+       "long: [[-122.7,-74.1]]"
+      ]
+     },
+     "execution_count": 3,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "import pandas as pd\n",
+    "\n",
+    "data = pd.DataFrame({\n",
+    "    \"vector\": [[1.1, 1.2], [0.2, 1.8]],\n",
+    "    \"lat\": [45.5, 40.1],\n",
+    "    \"long\": [-122.7, -74.1]\n",
+    "})\n",
+    "\n",
+    "db.create_table(\"table2\", data)\n",
+    "\n",
+    "db[\"table2\"].head() "
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "4be81469-5b57-4f78-9c72-3938c0378d9d",
+   "metadata": {},
+   "source": [
+    "Data is converted to Arrow before being written to disk. For maximum control over how data is saved, either provide the PyArrow schema to convert to or else provide a PyArrow Table directly.\n",
+    "  "
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "25f34bcf-fca0-4431-8601-eac95d1bd347",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "vector: fixed_size_list<item: float>[2]\n",
+       "  child 0, item: float\n",
+       "lat: float\n",
+       "long: float"
+      ]
+     },
+     "execution_count": 8,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "import pyarrow as pa\n",
+    "\n",
+    "custom_schema = pa.schema([\n",
+    "pa.field(\"vector\", pa.list_(pa.float32(), 2)),\n",
+    "pa.field(\"lat\", pa.float32()),\n",
+    "pa.field(\"long\", pa.float32())\n",
+    "])\n",
+    "\n",
+    "table = db.create_table(\"table3\", data, schema=custom_schema, mode=\"overwrite\")\n",
+    "table.schema"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "4df51925-7ca2-4005-9c72-38b3d26240c6",
+   "metadata": {},
+   "source": [
+    "### From PyArrow Tables\n",
+    "\n",
+    "You can also create LanceDB tables directly from pyarrow tables"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "id": "90a880f6-be43-4c9d-ba65-0b05197c0f6f",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "vector: fixed_size_list<item: float>[2]\n",
+       "  child 0, item: float\n",
+       "item: string\n",
+       "price: double"
+      ]
+     },
+     "execution_count": 12,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "table = pa.Table.from_arrays(\n",
+    "        [\n",
+    "            pa.array([[3.1, 4.1], [5.9, 26.5]],\n",
+    "                    pa.list_(pa.float32(), 2)),\n",
+    "            pa.array([\"foo\", \"bar\"]),\n",
+    "            pa.array([10.0, 20.0]),\n",
+    "        ],\n",
+    "        [\"vector\", \"item\", \"price\"],\n",
+    "    )\n",
+    "\n",
+    "db = lancedb.connect(\"db\")\n",
+    "\n",
+    "tbl = db.create_table(\"test1\", table, mode=\"overwrite\")\n",
+    "tbl.schema"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "0f36c51c-d902-449d-8292-700e53990c32",
+   "metadata": {},
+   "source": [
+    "### From Pydantic Models\n",
+    "\n",
+    "LanceDB supports to create Apache Arrow Schema from a Pydantic BaseModel."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "id": "d81121d7-e4b7-447c-a48c-974b6ebb464a",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "movie_id: int64 not null\n",
+       "vector: fixed_size_list<item: float>[128] not null\n",
+       "  child 0, item: float\n",
+       "genres: string not null\n",
+       "title: string not null\n",
+       "imdb_id: int64 not null"
+      ]
+     },
+     "execution_count": 13,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "from lancedb.pydantic import Vector, LanceModel\n",
+    "\n",
+    "class Content(LanceModel):\n",
+    "    movie_id: int\n",
+    "    vector: Vector(128)\n",
+    "    genres: str\n",
+    "    title: str\n",
+    "    imdb_id: int\n",
+    "        \n",
+    "    @property\n",
+    "    def imdb_url(self) -> str:\n",
+    "        return f\"https://www.imdb.com/title/tt{self.imdb_id}\"\n",
+    "\n",
+    "import pyarrow as pa\n",
+    "db = lancedb.connect(\"~/.lancedb\")\n",
+    "table_name = \"movielens_small\"\n",
+    "table = db.create_table(table_name, schema=Content)\n",
+    "table.schema"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "860e1f77-e860-46a9-98b7-b2979092ccd6",
+   "metadata": {},
+   "source": [
+    "### Using Iterators / Writing Large Datasets\n",
+    "\n",
+    "It is recommended to use itertators to add large datasets in batches when creating your table in one go. This does not create multiple versions of your dataset unlike manually adding batches using `table.add()`\n",
+    "\n",
+    "LanceDB additionally supports pyarrow's `RecordBatch` Iterators or other generators producing supported data types.\n",
+    "\n",
+    "## Here's an example using using `RecordBatch` iterator for creating tables."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "id": "bc247142-4e3c-41a2-b94c-8e00d2c2a508",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "LanceTable(table4)"
+      ]
+     },
+     "execution_count": 14,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "import pyarrow as pa\n",
+    "\n",
+    "def make_batches():\n",
+    "    for i in range(5):\n",
+    "        yield pa.RecordBatch.from_arrays(\n",
+    "            [\n",
+    "                pa.array([[3.1, 4.1], [5.9, 26.5]],\n",
+    "                        pa.list_(pa.float32(), 2)),\n",
+    "                pa.array([\"foo\", \"bar\"]),\n",
+    "                pa.array([10.0, 20.0]),\n",
+    "            ],\n",
+    "            [\"vector\", \"item\", \"price\"],\n",
+    "        )\n",
+    "\n",
+    "schema = pa.schema([\n",
+    "    pa.field(\"vector\", pa.list_(pa.float32(), 2)),\n",
+    "    pa.field(\"item\", pa.utf8()),\n",
+    "    pa.field(\"price\", pa.float32()),\n",
+    "])\n",
+    "\n",
+    "db.create_table(\"table4\", make_batches(), schema=schema)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "94f7dd2b-bae4-4bdf-8534-201437c31027",
+   "metadata": {},
+   "source": [
+    "### Using pandas `DataFrame` Iterator and Pydantic Schema\n",
+    "\n",
+    "You can set the schema via pyarrow schema object or using Pydantic object"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 16,
+   "id": "25ad3523-e0c9-4c28-b3df-38189c4e0e5f",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "vector: fixed_size_list<item: float>[2] not null\n",
+       "  child 0, item: float\n",
+       "item: string not null\n",
+       "price: double not null"
+      ]
+     },
+     "execution_count": 16,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "import pyarrow as pa\n",
+    "import pandas as pd\n",
+    "\n",
+    "class PydanticSchema(LanceModel):\n",
+    "    vector: Vector(2)\n",
+    "    item: str\n",
+    "    price: float\n",
+    "\n",
+    "def make_batches():\n",
+    "    for i in range(5):\n",
+    "        yield pd.DataFrame(\n",
+    "                    {\n",
+    "                        \"vector\": [[3.1, 4.1], [1, 1]],\n",
+    "                        \"item\": [\"foo\", \"bar\"],\n",
+    "                        \"price\": [10.0, 20.0],\n",
+    "                    })\n",
+    "\n",
+    "tbl = db.create_table(\"table5\", make_batches(), schema=PydanticSchema)\n",
+    "tbl.schema"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "4aa955e9-fcd0-4c99-b644-f218f3bb3f1a",
+   "metadata": {},
+   "source": [
+    "## Creating Empty Table\n",
+    "\n",
+    "You can create an empty table by just passing the schema and later add to it using `table.add()`"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 17,
+   "id": "2814173a-eacc-4dd8-a64d-6312b44582cc",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import lancedb\n",
+    "from lancedb.pydantic import LanceModel, Vector\n",
+    "\n",
+    "class Model(LanceModel):\n",
+    "      vector: Vector(2)\n",
+    "\n",
+    "tbl = db.create_table(\"table6\", schema=Model.to_arrow_schema())"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "1d1b0f5c-a1d9-459f-8614-8376b6f577e1",
+   "metadata": {},
+   "source": [
+    "## Open Existing Tables\n",
+    "\n",
+    "If you forget the name of your table, you can always get a listing of all table names:\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 18,
+   "id": "df9e13c0-41f6-437f-9dfa-2fd71d3d9c45",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "['table6', 'table4', 'table5', 'movielens_small']"
+      ]
+     },
+     "execution_count": 18,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "db.table_names()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 20,
+   "id": "9343f5ad-6024-42ee-ac2f-6c1471df8679",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>vector</th>\n",
+       "      <th>item</th>\n",
+       "      <th>price</th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "    <tr>\n",
+       "      <th>0</th>\n",
+       "      <td>[3.1, 4.1]</td>\n",
+       "      <td>foo</td>\n",
+       "      <td>10.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1</th>\n",
+       "      <td>[5.9, 26.5]</td>\n",
+       "      <td>bar</td>\n",
+       "      <td>20.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>2</th>\n",
+       "      <td>[3.1, 4.1]</td>\n",
+       "      <td>foo</td>\n",
+       "      <td>10.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>3</th>\n",
+       "      <td>[5.9, 26.5]</td>\n",
+       "      <td>bar</td>\n",
+       "      <td>20.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>4</th>\n",
+       "      <td>[3.1, 4.1]</td>\n",
+       "      <td>foo</td>\n",
+       "      <td>10.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>5</th>\n",
+       "      <td>[5.9, 26.5]</td>\n",
+       "      <td>bar</td>\n",
+       "      <td>20.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>6</th>\n",
+       "      <td>[3.1, 4.1]</td>\n",
+       "      <td>foo</td>\n",
+       "      <td>10.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>7</th>\n",
+       "      <td>[5.9, 26.5]</td>\n",
+       "      <td>bar</td>\n",
+       "      <td>20.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>8</th>\n",
+       "      <td>[3.1, 4.1]</td>\n",
+       "      <td>foo</td>\n",
+       "      <td>10.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>9</th>\n",
+       "      <td>[5.9, 26.5]</td>\n",
+       "      <td>bar</td>\n",
+       "      <td>20.0</td>\n",
+       "    </tr>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "        vector item  price\n",
+       "0   [3.1, 4.1]  foo   10.0\n",
+       "1  [5.9, 26.5]  bar   20.0\n",
+       "2   [3.1, 4.1]  foo   10.0\n",
+       "3  [5.9, 26.5]  bar   20.0\n",
+       "4   [3.1, 4.1]  foo   10.0\n",
+       "5  [5.9, 26.5]  bar   20.0\n",
+       "6   [3.1, 4.1]  foo   10.0\n",
+       "7  [5.9, 26.5]  bar   20.0\n",
+       "8   [3.1, 4.1]  foo   10.0\n",
+       "9  [5.9, 26.5]  bar   20.0"
+      ]
+     },
+     "execution_count": 20,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "tbl = db.open_table(\"table4\")\n",
+    "tbl.to_pandas()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "5019246f-12e3-4f78-88a8-9f4939802c76",
+   "metadata": {},
+   "source": [
+    "## Adding to table\n",
+    "After a table has been created, you can always add more data to it using\n",
+    "\n",
+    "You can add any of the valid data structures accepted by LanceDB table, i.e, `dict`, `list[dict]`, `pd.DataFrame`, or a `Iterator[pa.RecordBatch]`. Here are some examples."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 21,
+   "id": "8a56250f-73a1-4c26-a6ad-5c7a0ce3a9ab",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df = pd.DataFrame([{\"vector\": [1.3, 1.4], \"item\": \"fizz\", \"price\": 100.0},\n",
+    "              {\"vector\": [9.5, 56.2], \"item\": \"buzz\", \"price\": 200.0}])\n",
+    "tbl.add(df)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "9985f6ee-67e1-45a9-b233-94e3d121ecbf",
+   "metadata": {},
+   "source": [
+    "You can also add a large dataset batch in one go using Iterator of supported data types\n",
+    "\n",
+    "### Adding via Iterator\n",
+    "\n",
+    "here, we'll use pandas DataFrame Iterator"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 22,
+   "id": "030c7057-b98e-4e2f-be14-b8c1f927f83c",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "\n",
+    "import pandas as pd\n",
+    "\n",
+    "def make_batches():\n",
+    "    for i in range(5):\n",
+    "        yield pd.DataFrame(\n",
+    "                {\n",
+    "                    \"vector\": [[3.1, 4.1], [1, 1]],\n",
+    "                    \"item\": [\"foo\", \"bar\"],\n",
+    "                    \"price\": [10.0, 20.0],\n",
+    "                })\n",
+    "tbl.add(make_batches())"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "b8316d5d-0a23-4675-b0ee-178711db873a",
+   "metadata": {},
+   "source": [
+    "## Deleting from a Table\n",
+    "\n",
+    "Use the `delete()` method on tables to delete rows from a table. To choose which rows to delete, provide a filter that matches on the metadata columns. This can delete any number of rows that match the filter, like:\n",
+    "\n",
+    "\n",
+    "```python\n",
+    "tbl.delete('item = \"fizz\"')\n",
+    "```\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 24,
+   "id": "e7a17de2-08d2-41b7-bd05-f63d1045ab1f",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "32\n"
+     ]
+    },
+    {
+     "data": {
+      "text/plain": [
+       "17"
+      ]
+     },
+     "execution_count": 24,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "print(len(tbl))\n",
+    "      \n",
+    "tbl.delete(\"price = 20.0\")\n",
+    "      \n",
+    "len(tbl)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "74ac180b-5432-4c14-b1a8-22c35ac83af8",
+   "metadata": {},
+   "source": [
+    "### Delete from a list of values"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 30,
+   "id": "fe3310bd-08f4-4a22-a63b-b3127d22f9f7",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "         vector  item  price\n",
+      "0    [3.1, 4.1]   foo   10.0\n",
+      "1    [3.1, 4.1]   foo   10.0\n",
+      "2    [3.1, 4.1]   foo   10.0\n",
+      "3    [3.1, 4.1]   foo   10.0\n",
+      "4    [3.1, 4.1]   foo   10.0\n",
+      "5    [1.3, 1.4]  fizz  100.0\n",
+      "6   [9.5, 56.2]  buzz  200.0\n",
+      "7    [3.1, 4.1]   foo   10.0\n",
+      "8    [3.1, 4.1]   foo   10.0\n",
+      "9    [3.1, 4.1]   foo   10.0\n",
+      "10   [3.1, 4.1]   foo   10.0\n",
+      "11   [3.1, 4.1]   foo   10.0\n",
+      "12   [3.1, 4.1]   foo   10.0\n",
+      "13   [3.1, 4.1]   foo   10.0\n",
+      "14   [3.1, 4.1]   foo   10.0\n",
+      "15   [3.1, 4.1]   foo   10.0\n",
+      "16   [3.1, 4.1]   foo   10.0\n"
+     ]
+    },
+    {
+     "ename": "OSError",
+     "evalue": "LanceError(IO): Error during planning: column foo does not exist",
+     "output_type": "error",
+     "traceback": [
+      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
+      "\u001b[0;31mOSError\u001b[0m                                   Traceback (most recent call last)",
+      "Cell \u001b[0;32mIn[30], line 4\u001b[0m\n\u001b[1;32m      2\u001b[0m to_remove \u001b[38;5;241m=\u001b[39m \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m, \u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;241m.\u001b[39mjoin(\u001b[38;5;28mstr\u001b[39m(v) \u001b[38;5;28;01mfor\u001b[39;00m v \u001b[38;5;129;01min\u001b[39;00m to_remove)\n\u001b[1;32m      3\u001b[0m \u001b[38;5;28mprint\u001b[39m(tbl\u001b[38;5;241m.\u001b[39mto_pandas())\n\u001b[0;32m----> 4\u001b[0m \u001b[43mtbl\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mdelete\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;124;43mf\u001b[39;49m\u001b[38;5;124;43m\"\u001b[39;49m\u001b[38;5;124;43mitem IN (\u001b[39;49m\u001b[38;5;132;43;01m{\u001b[39;49;00m\u001b[43mto_remove\u001b[49m\u001b[38;5;132;43;01m}\u001b[39;49;00m\u001b[38;5;124;43m)\u001b[39;49m\u001b[38;5;124;43m\"\u001b[39;49m\u001b[43m)\u001b[49m\n\u001b[1;32m      5\u001b[0m tbl\u001b[38;5;241m.\u001b[39mto_pandas()\n",
+      "File \u001b[0;32m~/Documents/lancedb/lancedb/python/lancedb/table.py:610\u001b[0m, in \u001b[0;36mLanceTable.delete\u001b[0;34m(self, where)\u001b[0m\n\u001b[1;32m    609\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21mdelete\u001b[39m(\u001b[38;5;28mself\u001b[39m, where: \u001b[38;5;28mstr\u001b[39m):\n\u001b[0;32m--> 610\u001b[0m     \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_dataset\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mdelete\u001b[49m\u001b[43m(\u001b[49m\u001b[43mwhere\u001b[49m\u001b[43m)\u001b[49m\n",
+      "File \u001b[0;32m~/Documents/lancedb/lancedb/env/lib/python3.11/site-packages/lance/dataset.py:489\u001b[0m, in \u001b[0;36mLanceDataset.delete\u001b[0;34m(self, predicate)\u001b[0m\n\u001b[1;32m    487\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28misinstance\u001b[39m(predicate, pa\u001b[38;5;241m.\u001b[39mcompute\u001b[38;5;241m.\u001b[39mExpression):\n\u001b[1;32m    488\u001b[0m     predicate \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mstr\u001b[39m(predicate)\n\u001b[0;32m--> 489\u001b[0m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_ds\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mdelete\u001b[49m\u001b[43m(\u001b[49m\u001b[43mpredicate\u001b[49m\u001b[43m)\u001b[49m\n",
+      "\u001b[0;31mOSError\u001b[0m: LanceError(IO): Error during planning: column foo does not exist"
+     ]
+    }
+   ],
+   "source": [
+    "to_remove = [\"foo\", \"buzz\"]\n",
+    "to_remove = \", \".join(str(v) for v in to_remove)\n",
+    "print(tbl.to_pandas())\n",
+    "tbl.delete(f\"item IN ({to_remove})\")\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 43,
+   "id": "87d5bc21-847f-4c81-b56e-f6dbe5d05aac",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df = pd.DataFrame(\n",
+    "                    {\n",
+    "                        \"vector\": [[3.1, 4.1], [1, 1]],\n",
+    "                        \"item\": [\"foo\", \"bar\"],\n",
+    "                        \"price\": [10.0, 20.0],\n",
+    "                    })\n",
+    "\n",
+    "tbl = db.create_table(\"table7\", data=df, mode=\"overwrite\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 44,
+   "id": "9cba4519-eb3a-4941-ab7e-873d762e750f",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "to_remove = [10.0, 20.0]\n",
+    "to_remove = \", \".join(str(v) for v in to_remove)\n",
+    "\n",
+    "tbl.delete(f\"price IN ({to_remove})\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 46,
+   "id": "5bdc9801-d5ed-4871-92d0-88b27108e788",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>vector</th>\n",
+       "      <th>item</th>\n",
+       "      <th>price</th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "Empty DataFrame\n",
+       "Columns: [vector, item, price]\n",
+       "Index: []"
+      ]
+     },
+     "execution_count": 46,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "tbl.to_pandas()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "752d33d4-ce1c-48e5-90d2-c85f0982182d",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.11.4"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

docs/src/notebooks/youtube_transcript_search.ipynb

@@ -27,11 +27,11 @@
      "output_type": "stream",
      "text": [
       "\n",
-      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m A new release of pip is available: \u001b[0m\u001b[31;49m23.0\u001b[0m\u001b[39;49m -> \u001b[0m\u001b[32;49m23.1.1\u001b[0m\n",
-      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m To update, run: \u001b[0m\u001b[32;49mpip install --upgrade pip\u001b[0m\n",
+      "\u001B[1m[\u001B[0m\u001B[34;49mnotice\u001B[0m\u001B[1;39;49m]\u001B[0m\u001B[39;49m A new release of pip is available: \u001B[0m\u001B[31;49m23.0\u001B[0m\u001B[39;49m -> \u001B[0m\u001B[32;49m23.1.1\u001B[0m\n",
+      "\u001B[1m[\u001B[0m\u001B[34;49mnotice\u001B[0m\u001B[1;39;49m]\u001B[0m\u001B[39;49m To update, run: \u001B[0m\u001B[32;49mpip install --upgrade pip\u001B[0m\n",
       "\n",
-      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m A new release of pip is available: \u001b[0m\u001b[31;49m23.0\u001b[0m\u001b[39;49m -> \u001b[0m\u001b[32;49m23.1.1\u001b[0m\n",
-      "\u001b[1m[\u001b[0m\u001b[34;49mnotice\u001b[0m\u001b[1;39;49m]\u001b[0m\u001b[39;49m To update, run: \u001b[0m\u001b[32;49mpip install --upgrade pip\u001b[0m\n"
+      "\u001B[1m[\u001B[0m\u001B[34;49mnotice\u001B[0m\u001B[1;39;49m]\u001B[0m\u001B[39;49m A new release of pip is available: \u001B[0m\u001B[31;49m23.0\u001B[0m\u001B[39;49m -> \u001B[0m\u001B[32;49m23.1.1\u001B[0m\n",
+      "\u001B[1m[\u001B[0m\u001B[34;49mnotice\u001B[0m\u001B[1;39;49m]\u001B[0m\u001B[39;49m To update, run: \u001B[0m\u001B[32;49mpip install --upgrade pip\u001B[0m\n"
      ]
     }
    ],
@@ -184,7 +184,7 @@
     "df = (contextualize(data.to_pandas())\n",
     "      .groupby(\"title\").text_col(\"text\")\n",
     "      .window(20).stride(4)\n",
-    "      .to_df())\n",
+    "      .to_pandas())\n",
     "df.head(1)"
    ]
   },
@@ -603,7 +603,7 @@
    "outputs": [],
    "source": [
     "# Use LanceDB to get top 3 most relevant context\n",
-    "context = tbl.search(emb).limit(3).to_df()"
+    "context = tbl.search(emb).limit(3).to_pandas()"
    ]
   },
   {

docs/src/python/arrow.md

@@ -74,12 +74,12 @@ table = db.open_table("pd_table")
 
 query_vector = [100, 100]
 # Pandas DataFrame
-df = table.search(query_vector).limit(1).to_df()
+df = table.search(query_vector).limit(1).to_pandas()
 print(df)
 ```
 
 ```
-    vector     item  price        score
+    vector     item  price    _distance
 0  [5.9, 26.5]  bar   20.0  14257.05957
 ```
 
@@ -89,12 +89,12 @@ If you have more complex criteria, you can always apply the filter to the result
 ```python
 
 # Apply the filter via LanceDB
-results = table.search([100, 100]).where("price < 15").to_df()
+results = table.search([100, 100]).where("price < 15").to_pandas()
 assert len(results) == 1
 assert results["item"].iloc[0] == "foo"
 
 # Apply the filter via Pandas
-df = results = table.search([100, 100]).to_df()
+df = results = table.search([100, 100]).to_pandas()
 results = df[df.price < 15]
 assert len(results) == 1
 assert results["item"].iloc[0] == "foo"

docs/src/python/pydantic.md

@@ -3,7 +3,6 @@
 [Pydantic](https://docs.pydantic.dev/latest/) is a data validation library in Python.
 LanceDB integrates with Pydantic for schema inference, data ingestion, and query result casting.
 
-
 ## Schema
 
 LanceDB supports to create Apache Arrow Schema from a
@@ -14,10 +13,10 @@ via [pydantic_to_schema()](python.md##lancedb.pydantic.pydantic_to_schema) metho
 
 ## Vector Field
 
-LanceDB provides a [`vector(dim)`](python.md#lancedb.pydantic.vector) method to define a
+LanceDB provides a [`Vector(dim)`](python.md#lancedb.pydantic.Vector) method to define a
 vector Field in a Pydantic Model.
 
-::: lancedb.pydantic.vector
+::: lancedb.pydantic.Vector
 
 ## Type Conversion
 
@@ -34,4 +33,4 @@ Current supported type conversions:
 | `str`               | `pyarrow.utf8()`    |
 | `list`              | `pyarrow.List`    |
 | `BaseModel`         | `pyarrow.Struct`    |
-| `vector(n)`         | `pyarrow.FixedSizeList(float32, n)` |
+| `Vector(n)`         | `pyarrow.FixedSizeList(float32, n)` |

docs/src/python/python.md

@@ -26,9 +26,19 @@ pip install lancedb
 
 ## Embeddings
 
-::: lancedb.embeddings.with_embeddings
+::: lancedb.embeddings.functions.EmbeddingFunctionRegistry
 
-::: lancedb.embeddings.EmbeddingFunction
+::: lancedb.embeddings.functions.EmbeddingFunction
+
+::: lancedb.embeddings.functions.TextEmbeddingFunction
+
+::: lancedb.embeddings.functions.SentenceTransformerEmbeddings
+
+::: lancedb.embeddings.functions.OpenAIEmbeddings
+
+::: lancedb.embeddings.functions.OpenClipEmbeddings
+
+::: lancedb.embeddings.with_embeddings
 
 ## Context
 
@@ -56,4 +66,4 @@ pip install lancedb
 
 ::: lancedb.pydantic.vector
 
-
+::: lancedb.pydantic.LanceModel

docs/src/search.md

@@ -25,8 +25,8 @@ Currently, we support the following metrics:
 
 ### Flat Search
 
-If LanceDB does not create a vector index, LanceDB would need to scan (`Flat Search`) the entire vector column
-and compute the distance for each vector in order to find the closest matches.
+If you do not create a vector index, LanceDB would need to exhaustively scan the entire vector column (via `Flat Search`)
+and compute the distance for *every* vector in order to find the closest matches. This is effectively a KNN search.
 
 
 <!-- Setup Code
@@ -67,7 +67,7 @@ await db_setup.createTable('my_vectors', data)
 
     df = tbl.search(np.random.random((1536))) \
         .limit(10) \
-        .to_df()
+        .to_list()
     ```
 
 === "JavaScript"
@@ -92,7 +92,7 @@ as well.
     df = tbl.search(np.random.random((1536))) \
         .metric("cosine") \
         .limit(10) \
-        .to_df()
+        .to_list()
     ```
 
 
@@ -110,7 +110,7 @@ as well.
 
 To accelerate vector retrievals, it is common to build vector indices.
 A vector index is a data structure specifically designed to efficiently organize and
-search vector data based on their similarity or distance metrics.
+search vector data based on their similarity via the chosen distance metric.
 By constructing a vector index, you can reduce the search space and avoid the need
 for brute-force scanning of the entire vector column.
 

docs/src/styles/global.css

@@ -4,3 +4,12 @@
     --md-text-font: ui-sans-serif, system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", Roboto, "Helvetica Neue", Arial, "Noto Sans", sans-serif, "Apple Color Emoji", "Segoe UI Emoji", "Segoe UI Symbol", "Noto Color Emoji";
     --md-code-font: ui-monospace, SFMono-Regular, Menlo, Monaco, Consolas, "Liberation Mono", "Courier New", monospace;
 }
+
+.md-nav__item, .md-tabs__item {
+    font-size: large;
+}
+
+/* Maximum space for text block */
+.md-grid {
+    max-width: 90%;
+  }

docs/test/md_testing.js

@@ -2,18 +2,17 @@ const glob = require("glob");
 const fs = require("fs");
 const path = require("path");
 
-const excludedFiles = [
+const globString = "../src/**/*.md";
+const excludedGlobs = [
   "../src/fts.md",
   "../src/embedding.md",
-  "../src/examples/serverless_lancedb_with_s3_and_lambda.md",
-  "../src/examples/serverless_qa_bot_with_modal_and_langchain.md",
-  "../src/examples/transformerjs_embedding_search_nodejs.md",
-  "../src/examples/youtube_transcript_bot_with_nodejs.md",
+  "../src/examples/*.md",
+  "../src/guides/tables.md",
 ];
+
 const nodePrefix = "javascript";
 const nodeFile = ".js";
 const nodeFolder = "node";
-const globString = "../src/**/*.md";
 const asyncPrefix = "(async () => {\n";
 const asyncSuffix = "})();";
 
@@ -32,6 +31,7 @@ function* yieldLines(lines, prefix, suffix) {
 }
 
 const files = glob.sync(globString, { recursive: true });
+const excludedFiles = glob.sync(excludedGlobs, { recursive: true });
 
 for (const file of files.filter((file) => !excludedFiles.includes(file))) {
   const lines = [];

docs/test/md_testing.py

@@ -2,19 +2,22 @@ import glob
 from typing import Iterator
 from pathlib import Path
 
-excluded_files = [
+glob_string = "../src/**/*.md"
+excluded_globs = [
     "../src/fts.md",
     "../src/embedding.md",
-    "../src/examples/serverless_lancedb_with_s3_and_lambda.md",
-    "../src/examples/serverless_qa_bot_with_modal_and_langchain.md",
-    "../src/examples/youtube_transcript_bot_with_nodejs.md",
+    "../src/examples/*.md",
     "../src/integrations/voxel51.md",
+    "../src/guides/tables.md",
+    "../src/python/duckdb.md",
 ]
 
 python_prefix = "py"
 python_file = ".py"
 python_folder = "python"
-glob_string = "../src/**/*.md"
+
+files = glob.glob(glob_string, recursive=True)
+excluded_files = [f for excluded_glob in excluded_globs for f in glob.glob(excluded_glob, recursive=True)]
 
 def yield_lines(lines: Iterator[str], prefix: str, suffix: str):
     in_code_block = False
@@ -30,7 +33,7 @@ def yield_lines(lines: Iterator[str], prefix: str, suffix: str):
         elif in_code_block:
             yield line[strip_length:]
 
-for file in filter(lambda file: file not in excluded_files, glob.glob(glob_string, recursive=True)):
+for file in filter(lambda file: file not in excluded_files, files):
     with open(file, "r") as f:
         lines = list(yield_lines(iter(f), "```", "```"))
 

docs/test/requirements.txt

@@ -1,5 +1,8 @@
-lancedb @ git+https://github.com/lancedb/lancedb.git#egg=subdir&subdirectory=python
+-e ../../python
 numpy
 pandas
 pylance
 duckdb
+--extra-index-url https://download.pytorch.org/whl/cpu
+torch
+

node/examples/js-transformers/index.js

@@ -50,7 +50,7 @@ async function example() {
         { id: 5, text: 'Banana', type: 'fruit' }
     ]
 
-    const table = await db.createTable('food_table', data, "create", embed_fun)
+    const table = await db.createTable('food_table', data, embed_fun)
 
 
     // Query the table

node/examples/js-transformers/package.json

@@ -10,7 +10,7 @@
   "license": "Apache-2.0",
   "dependencies": {
     "@xenova/transformers": "^2.4.1",
-    "vectordb": "^0.1.12"
+    "vectordb": "file:../.."
   }
 
 }

node/native.js

@@ -12,26 +12,25 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-const { currentTarget } = require('@neon-rs/load');
+const { currentTarget } = require('@neon-rs/load')
 
-let nativeLib;
+let nativeLib
 
 try {
-    nativeLib = require(`@lancedb/vectordb-${currentTarget()}`);
-} catch (e) {
+  // When developing locally, give preference to the local built library
+  nativeLib = require('./index.node')
+} catch {
   try {
-        // Might be developing locally, so try that. But don't expose that error
-        // to the user.
-        nativeLib = require("./index.node");
-    } catch {
+    nativeLib = require(`@lancedb/vectordb-${currentTarget()}`)
+  } catch (e) {
     throw new Error(`vectordb: failed to load native library.
   You may need to run \`npm install @lancedb/vectordb-${currentTarget()}\`.
 
   If that does not work, please file a bug report at https://github.com/lancedb/lancedb/issues
       
-  Source error: ${e}`);
+  Source error: ${e}`)
   }
 }
 
 // Dynamic require for runtime.
-module.exports = nativeLib;
+module.exports = nativeLib

node/package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "vectordb",
-  "version": "0.1.19",
+  "version": "0.3.0",
   "lockfileVersion": 2,
   "requires": true,
   "packages": {
     "": {
       "name": "vectordb",
-      "version": "0.1.19",
+      "version": "0.3.0",
       "cpu": [
         "x64",
         "arm64"
@@ -31,6 +31,7 @@
         "@types/node": "^18.16.2",
         "@types/sinon": "^10.0.15",
         "@types/temp": "^0.9.1",
+        "@types/uuid": "^9.0.3",
         "@typescript-eslint/eslint-plugin": "^5.59.1",
         "cargo-cp-artifact": "^0.1",
         "chai": "^4.3.7",
@@ -48,14 +49,15 @@
         "ts-node-dev": "^2.0.0",
         "typedoc": "^0.24.7",
         "typedoc-plugin-markdown": "^3.15.3",
-        "typescript": "*"
+        "typescript": "*",
+        "uuid": "^9.0.0"
       },
       "optionalDependencies": {
-        "@lancedb/vectordb-darwin-arm64": "0.1.19",
-        "@lancedb/vectordb-darwin-x64": "0.1.19",
-        "@lancedb/vectordb-linux-arm64-gnu": "0.1.19",
-        "@lancedb/vectordb-linux-x64-gnu": "0.1.19",
-        "@lancedb/vectordb-win32-x64-msvc": "0.1.19"
+        "@lancedb/vectordb-darwin-arm64": "0.3.0",
+        "@lancedb/vectordb-darwin-x64": "0.3.0",
+        "@lancedb/vectordb-linux-arm64-gnu": "0.3.0",
+        "@lancedb/vectordb-linux-x64-gnu": "0.3.0",
+        "@lancedb/vectordb-win32-x64-msvc": "0.3.0"
       }
     },
     "node_modules/@apache-arrow/ts": {
@@ -315,9 +317,9 @@
       }
     },
     "node_modules/@lancedb/vectordb-darwin-arm64": {
-      "version": "0.1.19",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-arm64/-/vectordb-darwin-arm64-0.1.19.tgz",
-      "integrity": "sha512-efQhJkBKvMNhjFq3Sw3/qHo9D9gb9UqiIr98n3STsbNxBQjMnWemXn91Ckl40siRG1O8qXcINW7Qs/EGmus+kg==",
+      "version": "0.3.0",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-arm64/-/vectordb-darwin-arm64-0.3.0.tgz",
+      "integrity": "sha512-Fg+k/cSnqmNQlSWyDp0PpaAJ67kAISfZAD+zZ3mcE8/3ml2I/wM/GVjPy2zeiQX9aR93lG1mZXFSNTDUc74tWQ==",
       "cpu": [
         "arm64"
       ],
@@ -327,9 +329,9 @@
       ]
     },
     "node_modules/@lancedb/vectordb-darwin-x64": {
-      "version": "0.1.19",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-x64/-/vectordb-darwin-x64-0.1.19.tgz",
-      "integrity": "sha512-r6OZNVyemAssABz2w7CRhe7dyREwBEfTytn+ux1zzTnzsgMgDovCQ0rQ3WZcxWvcy7SFCxiemA9IP1b/lsb4tQ==",
+      "version": "0.3.0",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-x64/-/vectordb-darwin-x64-0.3.0.tgz",
+      "integrity": "sha512-CXp4b/brMbnBPZuGzKIOskd9uD90R73rWubaJ0du/Kt6fcyQX1dM1wEhWTLxI6eKf8IDL/R9QLL2cIahm1J86w==",
       "cpu": [
         "x64"
       ],
@@ -339,9 +341,9 @@
       ]
     },
     "node_modules/@lancedb/vectordb-linux-arm64-gnu": {
-      "version": "0.1.19",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-arm64-gnu/-/vectordb-linux-arm64-gnu-0.1.19.tgz",
-      "integrity": "sha512-mL/hRmZp6Kw7hmGJBdOZfp/tTYiCdlOcs8DA/+nr2eiXERv0gIhyiKvr2P5DwbBmut3qXEkDalMHTo95BSdL2A==",
+      "version": "0.3.0",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-arm64-gnu/-/vectordb-linux-arm64-gnu-0.3.0.tgz",
+      "integrity": "sha512-1bjaRzYcDsWIRUbO2K/f+ohNmNvCgKcrrOhmiXSHVlYY8kH1LUMFZj+BhqBC0Ea0Stt7/1rsRLMRXRtaeVOEHw==",
       "cpu": [
         "arm64"
       ],
@@ -351,9 +353,9 @@
       ]
     },
     "node_modules/@lancedb/vectordb-linux-x64-gnu": {
-      "version": "0.1.19",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-x64-gnu/-/vectordb-linux-x64-gnu-0.1.19.tgz",
-      "integrity": "sha512-AG0FHksbbr+cHVKPi4B8cmBtqb6T9E0uaK4kyZkXrX52/xtv9RYVZcykaB/tSSm0XNFPWWRnx9R8UqNZV/hxMA==",
+      "version": "0.3.0",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-x64-gnu/-/vectordb-linux-x64-gnu-0.3.0.tgz",
+      "integrity": "sha512-BEDIJ6ReGAi+tLTS/RzxIw621yo1UUUiVNTzPGV2didyiJCr1chIGbES+39d/wiFQM43Xs3CBZLNzp+jKkv0/w==",
       "cpu": [
         "x64"
       ],
@@ -363,9 +365,9 @@
       ]
     },
     "node_modules/@lancedb/vectordb-win32-x64-msvc": {
-      "version": "0.1.19",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-win32-x64-msvc/-/vectordb-win32-x64-msvc-0.1.19.tgz",
-      "integrity": "sha512-PDWZ2hvLVXH4Z4WIO1rsWY8ev3NpNm7aXlaey32P+l1Iz9Hia9+F2GBpp2UiEQKfvbk82ucAvBLRmpSsHY8Tlw==",
+      "version": "0.3.0",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-win32-x64-msvc/-/vectordb-win32-x64-msvc-0.3.0.tgz",
+      "integrity": "sha512-7K2kbWbShuifQF/6L/tWSz2DhKfIreHKlBdVOuBTYYOReQMHn5cJxgwuFgQHqMubZ9zcagtHpmo+Wtqd034OKQ==",
       "cpu": [
         "x64"
       ],
@@ -596,6 +598,12 @@
         "@types/node": "*"
       }
     },
+    "node_modules/@types/uuid": {
+      "version": "9.0.3",
+      "resolved": "https://registry.npmjs.org/@types/uuid/-/uuid-9.0.3.tgz",
+      "integrity": "sha512-taHQQH/3ZyI3zP8M/puluDEIEvtQHVYcC6y3N8ijFtAd28+Ey/G4sg1u2gB01S8MwybLOKAp9/yCMu/uR5l3Ug==",
+      "dev": true
+    },
     "node_modules/@typescript-eslint/eslint-plugin": {
       "version": "5.59.1",
       "resolved": "https://registry.npmjs.org/@typescript-eslint/eslint-plugin/-/eslint-plugin-5.59.1.tgz",
@@ -4451,6 +4459,15 @@
         "punycode": "^2.1.0"
       }
     },
+    "node_modules/uuid": {
+      "version": "9.0.0",
+      "resolved": "https://registry.npmjs.org/uuid/-/uuid-9.0.0.tgz",
+      "integrity": "sha512-MXcSTerfPa4uqyzStbRoTgt5XIe3x5+42+q1sDuy3R5MDk66URdLMOZe5aPX/SQd+kuYAh0FdP/pO28IkQyTeg==",
+      "dev": true,
+      "bin": {
+        "uuid": "dist/bin/uuid"
+      }
+    },
     "node_modules/v8-compile-cache-lib": {
       "version": "3.0.1",
       "resolved": "https://registry.npmjs.org/v8-compile-cache-lib/-/v8-compile-cache-lib-3.0.1.tgz",
@@ -4852,33 +4869,33 @@
       }
     },
     "@lancedb/vectordb-darwin-arm64": {
-      "version": "0.1.19",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-arm64/-/vectordb-darwin-arm64-0.1.19.tgz",
-      "integrity": "sha512-efQhJkBKvMNhjFq3Sw3/qHo9D9gb9UqiIr98n3STsbNxBQjMnWemXn91Ckl40siRG1O8qXcINW7Qs/EGmus+kg==",
+      "version": "0.3.0",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-arm64/-/vectordb-darwin-arm64-0.3.0.tgz",
+      "integrity": "sha512-Fg+k/cSnqmNQlSWyDp0PpaAJ67kAISfZAD+zZ3mcE8/3ml2I/wM/GVjPy2zeiQX9aR93lG1mZXFSNTDUc74tWQ==",
       "optional": true
     },
     "@lancedb/vectordb-darwin-x64": {
-      "version": "0.1.19",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-x64/-/vectordb-darwin-x64-0.1.19.tgz",
-      "integrity": "sha512-r6OZNVyemAssABz2w7CRhe7dyREwBEfTytn+ux1zzTnzsgMgDovCQ0rQ3WZcxWvcy7SFCxiemA9IP1b/lsb4tQ==",
+      "version": "0.3.0",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-darwin-x64/-/vectordb-darwin-x64-0.3.0.tgz",
+      "integrity": "sha512-CXp4b/brMbnBPZuGzKIOskd9uD90R73rWubaJ0du/Kt6fcyQX1dM1wEhWTLxI6eKf8IDL/R9QLL2cIahm1J86w==",
       "optional": true
     },
     "@lancedb/vectordb-linux-arm64-gnu": {
-      "version": "0.1.19",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-arm64-gnu/-/vectordb-linux-arm64-gnu-0.1.19.tgz",
-      "integrity": "sha512-mL/hRmZp6Kw7hmGJBdOZfp/tTYiCdlOcs8DA/+nr2eiXERv0gIhyiKvr2P5DwbBmut3qXEkDalMHTo95BSdL2A==",
+      "version": "0.3.0",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-arm64-gnu/-/vectordb-linux-arm64-gnu-0.3.0.tgz",
+      "integrity": "sha512-1bjaRzYcDsWIRUbO2K/f+ohNmNvCgKcrrOhmiXSHVlYY8kH1LUMFZj+BhqBC0Ea0Stt7/1rsRLMRXRtaeVOEHw==",
       "optional": true
     },
     "@lancedb/vectordb-linux-x64-gnu": {
-      "version": "0.1.19",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-x64-gnu/-/vectordb-linux-x64-gnu-0.1.19.tgz",
-      "integrity": "sha512-AG0FHksbbr+cHVKPi4B8cmBtqb6T9E0uaK4kyZkXrX52/xtv9RYVZcykaB/tSSm0XNFPWWRnx9R8UqNZV/hxMA==",
+      "version": "0.3.0",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-linux-x64-gnu/-/vectordb-linux-x64-gnu-0.3.0.tgz",
+      "integrity": "sha512-BEDIJ6ReGAi+tLTS/RzxIw621yo1UUUiVNTzPGV2didyiJCr1chIGbES+39d/wiFQM43Xs3CBZLNzp+jKkv0/w==",
       "optional": true
     },
     "@lancedb/vectordb-win32-x64-msvc": {
-      "version": "0.1.19",
-      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-win32-x64-msvc/-/vectordb-win32-x64-msvc-0.1.19.tgz",
-      "integrity": "sha512-PDWZ2hvLVXH4Z4WIO1rsWY8ev3NpNm7aXlaey32P+l1Iz9Hia9+F2GBpp2UiEQKfvbk82ucAvBLRmpSsHY8Tlw==",
+      "version": "0.3.0",
+      "resolved": "https://registry.npmjs.org/@lancedb/vectordb-win32-x64-msvc/-/vectordb-win32-x64-msvc-0.3.0.tgz",
+      "integrity": "sha512-7K2kbWbShuifQF/6L/tWSz2DhKfIreHKlBdVOuBTYYOReQMHn5cJxgwuFgQHqMubZ9zcagtHpmo+Wtqd034OKQ==",
       "optional": true
     },
     "@neon-rs/cli": {
@@ -5093,6 +5110,12 @@
         "@types/node": "*"
       }
     },
+    "@types/uuid": {
+      "version": "9.0.3",
+      "resolved": "https://registry.npmjs.org/@types/uuid/-/uuid-9.0.3.tgz",
+      "integrity": "sha512-taHQQH/3ZyI3zP8M/puluDEIEvtQHVYcC6y3N8ijFtAd28+Ey/G4sg1u2gB01S8MwybLOKAp9/yCMu/uR5l3Ug==",
+      "dev": true
+    },
     "@typescript-eslint/eslint-plugin": {
       "version": "5.59.1",
       "resolved": "https://registry.npmjs.org/@typescript-eslint/eslint-plugin/-/eslint-plugin-5.59.1.tgz",
@@ -7844,6 +7867,12 @@
         "punycode": "^2.1.0"
       }
     },
+    "uuid": {
+      "version": "9.0.0",
+      "resolved": "https://registry.npmjs.org/uuid/-/uuid-9.0.0.tgz",
+      "integrity": "sha512-MXcSTerfPa4uqyzStbRoTgt5XIe3x5+42+q1sDuy3R5MDk66URdLMOZe5aPX/SQd+kuYAh0FdP/pO28IkQyTeg==",
+      "dev": true
+    },
     "v8-compile-cache-lib": {
       "version": "3.0.1",
       "resolved": "https://registry.npmjs.org/v8-compile-cache-lib/-/v8-compile-cache-lib-3.0.1.tgz",

node/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vectordb",
-  "version": "0.1.19",
+  "version": "0.3.0",
   "description": " Serverless, low-latency vector database for AI applications",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",
@@ -9,7 +9,8 @@
     "build": "cargo-cp-artifact --artifact cdylib vectordb-node index.node -- cargo build --message-format=json",
     "build-release": "npm run build -- --release",
     "test": "npm run tsc && mocha -recursive dist/test",
-    "lint": "eslint src --ext .js,.ts",
+    "integration-test": "npm run tsc && mocha -recursive dist/integration_test",
+    "lint": "eslint native.js src --ext .js,.ts",
     "clean": "rm -rf node_modules *.node dist/",
     "pack-build": "neon pack-build",
     "check-npm": "printenv && which node && which npm && npm --version"
@@ -34,6 +35,7 @@
     "@types/node": "^18.16.2",
     "@types/sinon": "^10.0.15",
     "@types/temp": "^0.9.1",
+    "@types/uuid": "^9.0.3",
     "@typescript-eslint/eslint-plugin": "^5.59.1",
     "cargo-cp-artifact": "^0.1",
     "chai": "^4.3.7",
@@ -51,7 +53,8 @@
     "ts-node-dev": "^2.0.0",
     "typedoc": "^0.24.7",
     "typedoc-plugin-markdown": "^3.15.3",
-    "typescript": "*"
+    "typescript": "*",
+    "uuid": "^9.0.0"
   },
   "dependencies": {
     "@apache-arrow/ts": "^12.0.0",
@@ -78,10 +81,10 @@
     }
   },
   "optionalDependencies": {
-    "@lancedb/vectordb-darwin-arm64": "0.1.19",
-    "@lancedb/vectordb-darwin-x64": "0.1.19",
-    "@lancedb/vectordb-linux-arm64-gnu": "0.1.19",
-    "@lancedb/vectordb-linux-x64-gnu": "0.1.19",
-    "@lancedb/vectordb-win32-x64-msvc": "0.1.19"
+    "@lancedb/vectordb-darwin-arm64": "0.3.0",
+    "@lancedb/vectordb-darwin-x64": "0.3.0",
+    "@lancedb/vectordb-linux-arm64-gnu": "0.3.0",
+    "@lancedb/vectordb-linux-x64-gnu": "0.3.0",
+    "@lancedb/vectordb-win32-x64-msvc": "0.3.0"
   }
 }

node/src/arrow.ts

@@ -13,18 +13,19 @@
 // limitations under the License.
 
 import {
-  Field,
+  Field, type FixedSizeListBuilder,
   Float32,
-  List, type ListBuilder,
   makeBuilder,
   RecordBatchFileWriter,
-  Table, Utf8,
+  Utf8,
   type Vector,
-  vectorFromArray
+  FixedSizeList,
+  vectorFromArray, type Schema, Table as ArrowTable
 } from 'apache-arrow'
 import { type EmbeddingFunction } from './index'
 
-export async function convertToTable<T> (data: Array<Record<string, unknown>>, embeddings?: EmbeddingFunction<T>): Promise<Table> {
+// Converts an Array of records into an Arrow Table, optionally applying an embeddings function to it.
+export async function convertToTable<T> (data: Array<Record<string, unknown>>, embeddings?: EmbeddingFunction<T>): Promise<ArrowTable> {
   if (data.length === 0) {
     throw new Error('At least one record needs to be provided')
   }
@@ -34,8 +35,8 @@ export async function convertToTable<T> (data: Array<Record<string, unknown>>, e
 
   for (const columnsKey of columns) {
     if (columnsKey === 'vector') {
-      const listBuilder = newVectorListBuilder()
       const vectorSize = (data[0].vector as any[]).length
+      const listBuilder = newVectorBuilder(vectorSize)
       for (const datum of data) {
         if ((datum[columnsKey] as any[]).length !== vectorSize) {
           throw new Error(`Invalid vector size, expected ${vectorSize}`)
@@ -52,9 +53,7 @@ export async function convertToTable<T> (data: Array<Record<string, unknown>>, e
 
       if (columnsKey === embeddings?.sourceColumn) {
         const vectors = await embeddings.embed(values as T[])
-        const listBuilder = newVectorListBuilder()
-        vectors.map(v => listBuilder.append(v))
-        records.vector = listBuilder.finish().toVector()
+        records.vector = vectorFromArray(vectors, newVectorType(vectors[0].length))
       }
 
       if (typeof values[0] === 'string') {
@@ -66,20 +65,47 @@ export async function convertToTable<T> (data: Array<Record<string, unknown>>, e
     }
   }
 
-  return new Table(records)
+  return new ArrowTable(records)
 }
 
 // Creates a new Arrow ListBuilder that stores a Vector column
-function newVectorListBuilder (): ListBuilder<Float32, any> {
-  const children = new Field<Float32>('item', new Float32())
-  const list = new List(children)
+function newVectorBuilder (dim: number): FixedSizeListBuilder<Float32> {
   return makeBuilder({
-    type: list
+    type: newVectorType(dim)
   })
 }
 
+// Creates the Arrow Type for a Vector column with dimension `dim`
+function newVectorType (dim: number): FixedSizeList<Float32> {
+  const children = new Field<Float32>('item', new Float32())
+  return new FixedSizeList(dim, children)
+}
+
+// Converts an Array of records into Arrow IPC format
 export async function fromRecordsToBuffer<T> (data: Array<Record<string, unknown>>, embeddings?: EmbeddingFunction<T>): Promise<Buffer> {
   const table = await convertToTable(data, embeddings)
   const writer = RecordBatchFileWriter.writeAll(table)
   return Buffer.from(await writer.toUint8Array())
 }
+
+// Converts an Arrow Table into Arrow IPC format
+export async function fromTableToBuffer<T> (table: ArrowTable, embeddings?: EmbeddingFunction<T>): Promise<Buffer> {
+  if (embeddings !== undefined) {
+    const source = table.getChild(embeddings.sourceColumn)
+
+    if (source === null) {
+      throw new Error(`The embedding source column ${embeddings.sourceColumn} was not found in the Arrow Table`)
+    }
+
+    const vectors = await embeddings.embed(source.toArray() as T[])
+    const column = vectorFromArray(vectors, newVectorType(vectors[0].length))
+    table = table.assign(new ArrowTable({ vector: column }))
+  }
+  const writer = RecordBatchFileWriter.writeAll(table)
+  return Buffer.from(await writer.toUint8Array())
+}
+
+// Creates an empty Arrow Table
+export function createEmptyTable (schema: Schema): ArrowTable {
+  return new ArrowTable(schema)
+}

node/src/index.ts

@@ -13,17 +13,17 @@
 // limitations under the License.
 
 import {
-  RecordBatchFileWriter,
-  type Table as ArrowTable
+  type Schema,
+  Table as ArrowTable
 } from 'apache-arrow'
-import { fromRecordsToBuffer } from './arrow'
+import { createEmptyTable, fromRecordsToBuffer, fromTableToBuffer } from './arrow'
 import type { EmbeddingFunction } from './embedding/embedding_function'
 import { RemoteConnection } from './remote'
 import { Query } from './query'
 import { isEmbeddingFunction } from './embedding/embedding_function'
 
 // eslint-disable-next-line @typescript-eslint/no-var-requires
-const { databaseNew, databaseTableNames, databaseOpenTable, databaseDropTable, tableCreate, tableAdd, tableCreateVectorIndex, tableCountRows, tableDelete } = require('../native.js')
+const { databaseNew, databaseTableNames, databaseOpenTable, databaseDropTable, tableCreate, tableAdd, tableCreateVectorIndex, tableCountRows, tableDelete, tableCleanupOldVersions, tableCompactFiles } = require('../native.js')
 
 export { Query }
 export type { EmbeddingFunction }
@@ -42,6 +42,8 @@ export interface ConnectionOptions {
 
   awsCredentials?: AwsCredentials
 
+  awsRegion?: string
+
   // API key for the remote connections
   apiKey?: string
   // Region to connect
@@ -51,6 +53,40 @@ export interface ConnectionOptions {
   hostOverride?: string
 }
 
+function getAwsArgs (opts: ConnectionOptions): any[] {
+  const callArgs = []
+  const awsCredentials = opts.awsCredentials
+  if (awsCredentials !== undefined) {
+    callArgs.push(awsCredentials.accessKeyId)
+    callArgs.push(awsCredentials.secretKey)
+    callArgs.push(awsCredentials.sessionToken)
+  } else {
+    callArgs.push(undefined)
+    callArgs.push(undefined)
+    callArgs.push(undefined)
+  }
+
+  callArgs.push(opts.awsRegion)
+  return callArgs
+}
+
+export interface CreateTableOptions<T> {
+  // Name of Table
+  name: string
+
+  // Data to insert into the Table
+  data?: Array<Record<string, unknown>> | ArrowTable | undefined
+
+  // Optional Arrow Schema for this table
+  schema?: Schema | undefined
+
+  // Optional embedding function used to create embeddings
+  embeddingFunction?: EmbeddingFunction<T> | undefined
+
+  // WriteOptions for this operation
+  writeOptions?: WriteOptions | undefined
+}
+
 /**
  * Connect to a LanceDB instance at the given URI
  * @param uri The uri of the database.
@@ -97,6 +133,17 @@ export interface Connection {
    */
   openTable<T>(name: string, embeddings?: EmbeddingFunction<T>): Promise<Table<T>>
 
+  /**
+   * Creates a new Table, optionally initializing it with new data.
+   *
+   * @param {string} name - The name of the table.
+   * @param data - Array of Records to be inserted into the table
+   * @param schema - An Arrow Schema that describe this table columns
+   * @param {EmbeddingFunction} embeddings - An embedding function to use on this table
+   * @param {WriteOptions} writeOptions - The write options to use when creating the table.
+   */
+  createTable<T> ({ name, data, schema, embeddingFunction, writeOptions }: CreateTableOptions<T>): Promise<Table<T>>
+
   /**
    * Creates a new Table and initialize it with new data.
    *
@@ -132,8 +179,6 @@ export interface Connection {
    */
   createTable<T> (name: string, data: Array<Record<string, unknown>>, embeddings: EmbeddingFunction<T>, options: WriteOptions): Promise<Table<T>>
 
-  createTableArrow(name: string, table: ArrowTable): Promise<Table>
-
   /**
    * Drop an existing table.
    * @param name The name of the table to drop.
@@ -221,16 +266,16 @@ export interface Table<T = number[]> {
  * A connection to a LanceDB database.
  */
 export class LocalConnection implements Connection {
-  private readonly _options: ConnectionOptions
+  private readonly _options: () => ConnectionOptions
   private readonly _db: any
 
   constructor (db: any, options: ConnectionOptions) {
-    this._options = options
+    this._options = () => options
     this._db = db
   }
 
   get uri (): string {
-    return this._options.uri
+    return this._options().uri
   }
 
   /**
@@ -256,15 +301,16 @@ export class LocalConnection implements Connection {
   async openTable<T> (name: string, embeddings: EmbeddingFunction<T>): Promise<Table<T>>
   async openTable<T> (name: string, embeddings?: EmbeddingFunction<T>): Promise<Table<T>>
   async openTable<T> (name: string, embeddings?: EmbeddingFunction<T>): Promise<Table<T>> {
-    const tbl = await databaseOpenTable.call(this._db, name)
+    const tbl = await databaseOpenTable.call(this._db, name, ...getAwsArgs(this._options()))
     if (embeddings !== undefined) {
-      return new LocalTable(tbl, name, this._options, embeddings)
+      return new LocalTable(tbl, name, this._options(), embeddings)
     } else {
-      return new LocalTable(tbl, name, this._options)
+      return new LocalTable(tbl, name, this._options())
     }
   }
 
-  async createTable<T> (name: string, data: Array<Record<string, unknown>>, optsOrEmbedding?: WriteOptions | EmbeddingFunction<T>, opt?: WriteOptions): Promise<Table<T>> {
+  async createTable<T> (name: string | CreateTableOptions<T>, data?: Array<Record<string, unknown>>, optsOrEmbedding?: WriteOptions | EmbeddingFunction<T>, opt?: WriteOptions): Promise<Table<T>> {
+    if (typeof name === 'string') {
       let writeOptions: WriteOptions = new DefaultWriteOptions()
       if (opt !== undefined && isWriteOptions(opt)) {
         writeOptions = opt
@@ -276,28 +322,45 @@ export class LocalConnection implements Connection {
       if (optsOrEmbedding !== undefined && isEmbeddingFunction(optsOrEmbedding)) {
         embeddings = optsOrEmbedding
       }
-    const createArgs = [this._db, name, await fromRecordsToBuffer(data, embeddings), writeOptions.writeMode?.toString()]
-    if (this._options.awsCredentials !== undefined) {
-      createArgs.push(this._options.awsCredentials.accessKeyId)
-      createArgs.push(this._options.awsCredentials.secretKey)
-      if (this._options.awsCredentials.sessionToken !== undefined) {
-        createArgs.push(this._options.awsCredentials.sessionToken)
+      return await this.createTableImpl({ name, data, embeddingFunction: embeddings, writeOptions })
     }
+    return await this.createTableImpl(name)
   }
 
-    const tbl = await tableCreate.call(...createArgs)
+  private async createTableImpl<T> ({ name, data, schema, embeddingFunction, writeOptions = new DefaultWriteOptions() }: {
+    name: string
+    data?: Array<Record<string, unknown>> | ArrowTable | undefined
+    schema?: Schema | undefined
+    embeddingFunction?: EmbeddingFunction<T> | undefined
+    writeOptions?: WriteOptions | undefined
+  }): Promise<Table<T>> {
+    let buffer: Buffer
 
-    if (embeddings !== undefined) {
-      return new LocalTable(tbl, name, this._options, embeddings)
+    function isEmpty (data: Array<Record<string, unknown>> | ArrowTable<any>): boolean {
+      if (data instanceof ArrowTable) {
+        return data.data.length === 0
+      }
+      return data.length === 0
+    }
+
+    if ((data === undefined) || isEmpty(data)) {
+      if (schema === undefined) {
+        throw new Error('Either data or schema needs to defined')
+      }
+      buffer = await fromTableToBuffer(createEmptyTable(schema))
+    } else if (data instanceof ArrowTable) {
+      buffer = await fromTableToBuffer(data, embeddingFunction)
     } else {
-      return new LocalTable(tbl, name, this._options)
-    }
+      // data is Array<Record<...>>
+      buffer = await fromRecordsToBuffer(data, embeddingFunction)
     }
 
-  async createTableArrow (name: string, table: ArrowTable): Promise<Table> {
-    const writer = RecordBatchFileWriter.writeAll(table)
-    await tableCreate.call(this._db, name, Buffer.from(await writer.toUint8Array()))
-    return await this.openTable(name)
+    const tbl = await tableCreate.call(this._db, name, buffer, writeOptions?.writeMode?.toString(), ...getAwsArgs(this._options()))
+    if (embeddingFunction !== undefined) {
+      return new LocalTable(tbl, name, this._options(), embeddingFunction)
+    } else {
+      return new LocalTable(tbl, name, this._options())
+    }
   }
 
   /**
@@ -310,10 +373,10 @@ export class LocalConnection implements Connection {
 }
 
 export class LocalTable<T = number[]> implements Table<T> {
-  private readonly _tbl: any
+  private _tbl: any
   private readonly _name: string
   private readonly _embeddings?: EmbeddingFunction<T>
-  private readonly _options: ConnectionOptions
+  private readonly _options: () => ConnectionOptions
 
   constructor (tbl: any, name: string, options: ConnectionOptions)
   /**
@@ -327,7 +390,7 @@ export class LocalTable<T = number[]> implements Table<T> {
     this._tbl = tbl
     this._name = name
     this._embeddings = embeddings
-    this._options = options
+    this._options = () => options
   }
 
   get name (): string {
@@ -349,15 +412,12 @@ export class LocalTable<T = number[]> implements Table<T> {
    * @return The number of rows added to the table
    */
   async add (data: Array<Record<string, unknown>>): Promise<number> {
-    const callArgs = [this._tbl, await fromRecordsToBuffer(data, this._embeddings), WriteMode.Append.toString()]
-    if (this._options.awsCredentials !== undefined) {
-      callArgs.push(this._options.awsCredentials.accessKeyId)
-      callArgs.push(this._options.awsCredentials.secretKey)
-      if (this._options.awsCredentials.sessionToken !== undefined) {
-        callArgs.push(this._options.awsCredentials.sessionToken)
-      }
-    }
-    return tableAdd.call(...callArgs)
+    return tableAdd.call(
+      this._tbl,
+      await fromRecordsToBuffer(data, this._embeddings),
+      WriteMode.Append.toString(),
+      ...getAwsArgs(this._options())
+    ).then((newTable: any) => { this._tbl = newTable })
   }
 
   /**
@@ -367,15 +427,12 @@ export class LocalTable<T = number[]> implements Table<T> {
    * @return The number of rows added to the table
    */
   async overwrite (data: Array<Record<string, unknown>>): Promise<number> {
-    const callArgs = [this._tbl, await fromRecordsToBuffer(data, this._embeddings), WriteMode.Overwrite.toString()]
-    if (this._options.awsCredentials !== undefined) {
-      callArgs.push(this._options.awsCredentials.accessKeyId)
-      callArgs.push(this._options.awsCredentials.secretKey)
-      if (this._options.awsCredentials.sessionToken !== undefined) {
-        callArgs.push(this._options.awsCredentials.sessionToken)
-      }
-    }
-    return tableAdd.call(this._tbl, await fromRecordsToBuffer(data, this._embeddings), WriteMode.Overwrite.toString())
+    return tableAdd.call(
+      this._tbl,
+      await fromRecordsToBuffer(data, this._embeddings),
+      WriteMode.Overwrite.toString(),
+      ...getAwsArgs(this._options())
+    ).then((newTable: any) => { this._tbl = newTable })
   }
 
   /**
@@ -384,7 +441,7 @@ export class LocalTable<T = number[]> implements Table<T> {
    * @param indexParams The parameters of this Index, @see VectorIndexParams.
    */
   async createIndex (indexParams: VectorIndexParams): Promise<any> {
-    return tableCreateVectorIndex.call(this._tbl, indexParams)
+    return tableCreateVectorIndex.call(this._tbl, indexParams).then((newTable: any) => { this._tbl = newTable })
   }
 
   /**
@@ -400,8 +457,113 @@ export class LocalTable<T = number[]> implements Table<T> {
    * @param filter A filter in the same format used by a sql WHERE clause.
    */
   async delete (filter: string): Promise<void> {
-    return tableDelete.call(this._tbl, filter)
+    return tableDelete.call(this._tbl, filter).then((newTable: any) => { this._tbl = newTable })
   }
+
+  /**
+   * Clean up old versions of the table, freeing disk space.
+   *
+   * @param olderThan The minimum age in minutes of the versions to delete. If not
+   *                  provided, defaults to two weeks.
+   * @param deleteUnverified Because they may be part of an in-progress
+   *                  transaction, uncommitted files newer than 7 days old are
+   *                  not deleted by default. This means that failed transactions
+   *                  can leave around data that takes up disk space for up to
+   *                  7 days. You can override this safety mechanism by setting
+   *                 this option to `true`, only if you promise there are no
+   *                 in progress writes while you run this operation. Failure to
+   *                 uphold this promise can lead to corrupted tables.
+   * @returns
+   */
+  async cleanupOldVersions (olderThan?: number, deleteUnverified?: boolean): Promise<CleanupStats> {
+    return tableCleanupOldVersions.call(this._tbl, olderThan, deleteUnverified)
+      .then((res: { newTable: any, metrics: CleanupStats }) => {
+        this._tbl = res.newTable
+        return res.metrics
+      })
+  }
+
+  /**
+   * Run the compaction process on the table.
+   *
+   * This can be run after making several small appends to optimize the table
+   * for faster reads.
+   *
+   * @param options Advanced options configuring compaction. In most cases, you
+   *               can omit this arguments, as the default options are sensible
+   *               for most tables.
+   * @returns Metrics about the compaction operation.
+   */
+  async compactFiles (options?: CompactionOptions): Promise<CompactionMetrics> {
+    const optionsArg = options ?? {}
+    return tableCompactFiles.call(this._tbl, optionsArg)
+      .then((res: { newTable: any, metrics: CompactionMetrics }) => {
+        this._tbl = res.newTable
+        return res.metrics
+      })
+  }
+}
+
+export interface CleanupStats {
+  /**
+   * The number of bytes removed from disk.
+   */
+  bytesRemoved: number
+  /**
+   * The number of old table versions removed.
+   */
+  oldVersions: number
+}
+
+export interface CompactionOptions {
+  /**
+   * The number of rows per fragment to target. Fragments that have fewer rows
+   * will be compacted into adjacent fragments to produce larger fragments.
+   * Defaults to 1024 * 1024.
+   */
+  targetRowsPerFragment?: number
+  /**
+   * The maximum number of rows per group. Defaults to 1024.
+   */
+  maxRowsPerGroup?: number
+  /**
+   * If true, fragments that have rows that are deleted may be compacted to
+   * remove the deleted rows. This can improve the performance of queries.
+   * Default is true.
+   */
+  materializeDeletions?: boolean
+  /**
+   * A number between 0 and 1, representing the proportion of rows that must be
+   * marked deleted before a fragment is a candidate for compaction to remove
+   * the deleted rows. Default is 10%.
+   */
+  materializeDeletionsThreshold?: number
+  /**
+   * The number of threads to use for compaction. If not provided, defaults to
+   * the number of cores on the machine.
+   */
+  numThreads?: number
+}
+
+export interface CompactionMetrics {
+  /**
+   * The number of fragments that were removed.
+   */
+  fragmentsRemoved: number
+  /**
+   * The number of new fragments that were created.
+   */
+  fragmentsAdded: number
+  /**
+   * The number of files that were removed. Each fragment may have more than one
+   * file.
+   */
+  filesRemoved: number
+  /**
+   * The number of files added. This is typically equal to the number of
+   * fragments added.
+   */
+  filesAdded: number
 }
 
 /// Config to build IVF_PQ index.

node/src/integration_test/test.ts

@@ -0,0 +1,173 @@
+// Copyright 2023 LanceDB Developers.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+import { describe } from 'mocha'
+import * as chai from 'chai'
+import * as chaiAsPromised from 'chai-as-promised'
+import { v4 as uuidv4 } from 'uuid'
+
+import * as lancedb from '../index'
+import { tmpdir } from 'os'
+import * as fs from 'fs'
+import * as path from 'path'
+
+const assert = chai.assert
+chai.use(chaiAsPromised)
+
+describe('LanceDB AWS Integration test', function () {
+  it('s3+ddb schema is processed correctly', async function () {
+    this.timeout(15000)
+
+    // WARNING: specifying engine is NOT a publicly supported feature in lancedb yet
+    // THE API WILL CHANGE
+    const conn = await lancedb.connect('s3://lancedb-integtest?engine=ddb&ddbTableName=lancedb-integtest')
+    const data = [{ vector: Array(128).fill(1.0) }]
+
+    const tableName = uuidv4()
+    let table = await conn.createTable(tableName, data, { writeMode: lancedb.WriteMode.Overwrite })
+
+    const futs = [table.add(data), table.add(data), table.add(data), table.add(data), table.add(data)]
+    await Promise.allSettled(futs)
+
+    table = await conn.openTable(tableName)
+    assert.equal(await table.countRows(), 6)
+  })
+})
+
+describe('LanceDB Mirrored Store Integration test', function () {
+  it('s3://...?mirroredStore=... param is processed correctly', async function () {
+    this.timeout(600000)
+
+    const dir = tmpdir()
+    console.log(dir)
+    const conn = await lancedb.connect(`s3://lancedb-integtest?mirroredStore=${dir}`)
+    const data = Array(200).fill({ vector: Array(128).fill(1.0), id: 0 })
+    data.push(...Array(200).fill({ vector: Array(128).fill(1.0), id: 1 }))
+    data.push(...Array(200).fill({ vector: Array(128).fill(1.0), id: 2 }))
+    data.push(...Array(200).fill({ vector: Array(128).fill(1.0), id: 3 }))
+
+    const tableName = uuidv4()
+
+    // try create table and check if it's mirrored
+    const t = await conn.createTable(tableName, data, { writeMode: lancedb.WriteMode.Overwrite })
+
+    const mirroredPath = path.join(dir, `${tableName}.lance`)
+    fs.readdir(mirroredPath, { withFileTypes: true }, (err, files) => {
+      if (err != null) throw err
+      // there should be two dirs
+      assert.equal(files.length, 2)
+      assert.isTrue(files[0].isDirectory())
+      assert.isTrue(files[1].isDirectory())
+
+      fs.readdir(path.join(mirroredPath, '_transactions'), { withFileTypes: true }, (err, files) => {
+        if (err != null) throw err
+        assert.equal(files.length, 1)
+        assert.isTrue(files[0].name.endsWith('.txn'))
+      })
+
+      fs.readdir(path.join(mirroredPath, 'data'), { withFileTypes: true }, (err, files) => {
+        if (err != null) throw err
+        assert.equal(files.length, 1)
+        assert.isTrue(files[0].name.endsWith('.lance'))
+      })
+    })
+
+    // try create index and check if it's mirrored
+    await t.createIndex({ column: 'vector', type: 'ivf_pq' })
+
+    fs.readdir(mirroredPath, { withFileTypes: true }, (err, files) => {
+      if (err != null) throw err
+      // there should be two dirs
+      assert.equal(files.length, 3)
+      assert.isTrue(files[0].isDirectory())
+      assert.isTrue(files[1].isDirectory())
+      assert.isTrue(files[2].isDirectory())
+
+      // Two TXs now
+      fs.readdir(path.join(mirroredPath, '_transactions'), { withFileTypes: true }, (err, files) => {
+        if (err != null) throw err
+        assert.equal(files.length, 2)
+        assert.isTrue(files[0].name.endsWith('.txn'))
+        assert.isTrue(files[1].name.endsWith('.txn'))
+      })
+
+      fs.readdir(path.join(mirroredPath, 'data'), { withFileTypes: true }, (err, files) => {
+        if (err != null) throw err
+        assert.equal(files.length, 1)
+        assert.isTrue(files[0].name.endsWith('.lance'))
+      })
+
+      fs.readdir(path.join(mirroredPath, '_indices'), { withFileTypes: true }, (err, files) => {
+        if (err != null) throw err
+        assert.equal(files.length, 1)
+        assert.isTrue(files[0].isDirectory())
+
+        fs.readdir(path.join(mirroredPath, '_indices', files[0].name), { withFileTypes: true }, (err, files) => {
+          if (err != null) throw err
+
+          assert.equal(files.length, 1)
+          assert.isTrue(files[0].isFile())
+          assert.isTrue(files[0].name.endsWith('.idx'))
+        })
+      })
+    })
+
+    // try delete and check if it's mirrored
+    await t.delete('id = 0')
+
+    fs.readdir(mirroredPath, { withFileTypes: true }, (err, files) => {
+      if (err != null) throw err
+      // there should be two dirs
+      assert.equal(files.length, 4)
+      assert.isTrue(files[0].isDirectory())
+      assert.isTrue(files[1].isDirectory())
+      assert.isTrue(files[2].isDirectory())
+      assert.isTrue(files[3].isDirectory())
+
+      // Three TXs now
+      fs.readdir(path.join(mirroredPath, '_transactions'), { withFileTypes: true }, (err, files) => {
+        if (err != null) throw err
+        assert.equal(files.length, 3)
+        assert.isTrue(files[0].name.endsWith('.txn'))
+        assert.isTrue(files[1].name.endsWith('.txn'))
+      })
+
+      fs.readdir(path.join(mirroredPath, 'data'), { withFileTypes: true }, (err, files) => {
+        if (err != null) throw err
+        assert.equal(files.length, 1)
+        assert.isTrue(files[0].name.endsWith('.lance'))
+      })
+
+      fs.readdir(path.join(mirroredPath, '_indices'), { withFileTypes: true }, (err, files) => {
+        if (err != null) throw err
+        assert.equal(files.length, 1)
+        assert.isTrue(files[0].isDirectory())
+
+        fs.readdir(path.join(mirroredPath, '_indices', files[0].name), { withFileTypes: true }, (err, files) => {
+          if (err != null) throw err
+
+          assert.equal(files.length, 1)
+          assert.isTrue(files[0].isFile())
+          assert.isTrue(files[0].name.endsWith('.idx'))
+        })
+      })
+
+      fs.readdir(path.join(mirroredPath, '_deletions'), { withFileTypes: true }, (err, files) => {
+        if (err != null) throw err
+        assert.equal(files.length, 1)
+        assert.isTrue(files[0].name.endsWith('.arrow'))
+      })
+    })
+  })
+})

node/src/query.ts

@@ -112,7 +112,8 @@ export class Query<T = number[]> {
       this._queryVector = this._query as number[]
     }
 
-    const buffer = await tableSearch.call(this._tbl, this)
+    const isElectron = this.isElectron()
+    const buffer = await tableSearch.call(this._tbl, this, isElectron)
     const data = tableFromIPC(buffer)
 
     return data.toArray().map((entry: Record<string, unknown>) => {
@@ -127,4 +128,14 @@ export class Query<T = number[]> {
       return newObject as unknown as T
     })
   }
+
+  // See https://github.com/electron/electron/issues/2288
+  private isElectron (): boolean {
+    try {
+      // eslint-disable-next-line no-prototype-builtins
+      return (process?.versions?.hasOwnProperty('electron') || navigator?.userAgent?.toLowerCase()?.includes(' electron'))
+    } catch (e) {
+      return false
+    }
+  }
 }

node/src/remote/client.ts

@@ -104,4 +104,34 @@ export class HttpLancedbClient {
     }
     return response
   }
+
+  /**
+   * Sent POST request.
+   */
+  public async post (path: string, data?: any, params?: Record<string, string | number>): Promise<AxiosResponse> {
+    const response = await axios.post(
+        `${this._url}${path}`,
+        data,
+        {
+          headers: {
+            'Content-Type': 'application/json',
+            'x-api-key': this._apiKey(),
+            ...(this._dbName !== undefined ? { 'x-lancedb-database': this._dbName } : {})
+          },
+          params,
+          timeout: 30000
+        }
+    ).catch((err) => {
+      console.error('error: ', err)
+      return err.response
+    })
+    if (response.status !== 200) {
+      const errorData = new TextDecoder().decode(response.data)
+      throw new Error(
+          `Server Error, status: ${response.status as number}, ` +
+          `message: ${response.statusText as string}: ${errorData}`
+      )
+    }
+    return response
+  }
 }

node/src/remote/index.ts

@@ -14,11 +14,11 @@
 
 import {
   type EmbeddingFunction, type Table, type VectorIndexParams, type Connection,
-  type ConnectionOptions
+  type ConnectionOptions, type CreateTableOptions, type WriteOptions
 } from '../index'
 import { Query } from '../query'
 
-import { type Table as ArrowTable, Vector } from 'apache-arrow'
+import { Vector } from 'apache-arrow'
 import { HttpLancedbClient } from './client'
 
 /**
@@ -66,18 +66,12 @@ export class RemoteConnection implements Connection {
     }
   }
 
-  async createTable (name: string, data: Array<Record<string, unknown>>): Promise<Table>
-  async createTable<T> (name: string, data: Array<Record<string, unknown>>, embeddings: EmbeddingFunction<T>): Promise<Table<T>>
-  async createTable<T> (name: string, data: Array<Record<string, unknown>>, embeddings?: EmbeddingFunction<T>): Promise<Table<T>> {
-    throw new Error('Not implemented')
-  }
-
-  async createTableArrow (name: string, table: ArrowTable): Promise<Table> {
+  async createTable<T> (name: string | CreateTableOptions<T>, data?: Array<Record<string, unknown>>, optsOrEmbedding?: WriteOptions | EmbeddingFunction<T>, opt?: WriteOptions): Promise<Table<T>> {
     throw new Error('Not implemented')
   }
 
   async dropTable (name: string): Promise<void> {
-    throw new Error('Not implemented')
+    await this._client.post(`/v1/table/${name}/drop/`)
   }
 }
 

node/src/test/io.ts

@@ -47,7 +47,9 @@ describe('LanceDB S3 client', function () {
         }
       }
       const table = await createTestDB(opts, 2, 20)
+      console.log(table)
       const con = await lancedb.connect(opts)
+      console.log(con)
       assert.equal(con.uri, opts.uri)
 
       const results = await table.search([0.1, 0.3]).limit(5).execute()
@@ -70,5 +72,5 @@ async function createTestDB (opts: ConnectionOptions, numDimensions: number = 2,
     data.push({ id: i + 1, name: `name_${i}`, price: i + 10, is_active: (i % 2 === 0), vector })
   }
 
-  return await con.createTable('vectors', data)
+  return await con.createTable('vectors_2', data)
 }

node/src/test/test.ts

@@ -18,7 +18,8 @@ import * as chai from 'chai'
 import * as chaiAsPromised from 'chai-as-promised'
 
 import * as lancedb from '../index'
-import { type AwsCredentials, type EmbeddingFunction, MetricType, Query, WriteMode, DefaultWriteOptions, isWriteOptions } from '../index'
+import { type AwsCredentials, type EmbeddingFunction, MetricType, Query, WriteMode, DefaultWriteOptions, isWriteOptions, type LocalTable } from '../index'
+import { FixedSizeList, Field, Int32, makeVector, Schema, Utf8, Table as ArrowTable, vectorFromArray, Float32 } from 'apache-arrow'
 
 const expect = chai.expect
 const assert = chai.assert
@@ -107,9 +108,9 @@ describe('LanceDB client', function () {
       const table = await con.openTable('vectors')
       const results = await table.search([0.1, 0.1]).select(['is_active']).execute()
       assert.equal(results.length, 2)
-      // vector and score are always returned
+      // vector and _distance are always returned
       assert.isDefined(results[0].vector)
-      assert.isDefined(results[0].score)
+      assert.isDefined(results[0]._distance)
       assert.isDefined(results[0].is_active)
 
       assert.isUndefined(results[0].id)
@@ -119,6 +120,45 @@ describe('LanceDB client', function () {
   })
 
   describe('when creating a new dataset', function () {
+    it('create an empty table', async function () {
+      const dir = await track().mkdir('lancejs')
+      const con = await lancedb.connect(dir)
+
+      const schema = new Schema(
+        [new Field('id', new Int32()), new Field('name', new Utf8())]
+      )
+      const table = await con.createTable({ name: 'vectors', schema })
+      assert.equal(table.name, 'vectors')
+      assert.deepEqual(await con.tableNames(), ['vectors'])
+    })
+
+    it('create a table with a empty data array', async function () {
+      const dir = await track().mkdir('lancejs')
+      const con = await lancedb.connect(dir)
+
+      const schema = new Schema(
+        [new Field('id', new Int32()), new Field('name', new Utf8())]
+      )
+      const table = await con.createTable({ name: 'vectors', schema, data: [] })
+      assert.equal(table.name, 'vectors')
+      assert.deepEqual(await con.tableNames(), ['vectors'])
+    })
+
+    it('create a table from an Arrow Table', async function () {
+      const dir = await track().mkdir('lancejs')
+      const con = await lancedb.connect(dir)
+
+      const i32s = new Int32Array(new Array<number>(10))
+      const i32 = makeVector(i32s)
+
+      const data = new ArrowTable({ vector: i32 })
+
+      const table = await con.createTable({ name: 'vectors', data })
+      assert.equal(table.name, 'vectors')
+      assert.equal(await table.countRows(), 10)
+      assert.deepEqual(await con.tableNames(), ['vectors'])
+    })
+
     it('creates a new table from javascript objects', async function () {
       const dir = await track().mkdir('lancejs')
       const con = await lancedb.connect(dir)
@@ -218,6 +258,36 @@ describe('LanceDB client', function () {
     })
   })
 
+  describe('when searching an empty dataset', function () {
+    it('should not fail', async function () {
+      const dir = await track().mkdir('lancejs')
+      const con = await lancedb.connect(dir)
+
+      const schema = new Schema(
+        [new Field('vector', new FixedSizeList(128, new Field('float32', new Float32())))]
+      )
+      const table = await con.createTable({ name: 'vectors', schema })
+      const result = await table.search(Array(128).fill(0.1)).execute()
+      assert.isEmpty(result)
+    })
+  })
+
+  describe('when searching an empty-after-delete dataset', function () {
+    it('should not fail', async function () {
+      const dir = await track().mkdir('lancejs')
+      const con = await lancedb.connect(dir)
+
+      const schema = new Schema(
+        [new Field('vector', new FixedSizeList(128, new Field('float32', new Float32())))]
+      )
+      const table = await con.createTable({ name: 'vectors', schema })
+      await table.add([{ vector: Array(128).fill(0.1) }])
+      await table.delete('vector IS NOT NULL')
+      const result = await table.search(Array(128).fill(0.1)).execute()
+      assert.isEmpty(result)
+    })
+  })
+
   describe('when creating a vector index', function () {
     it('overwrite all records in a table', async function () {
       const uri = await createTestDB(32, 300)
@@ -291,6 +361,20 @@ describe('LanceDB client', function () {
       const results = await table.search('foo').execute()
       assert.equal(results.length, 2)
     })
+
+    it('should create embeddings for Arrow Table', async function () {
+      const dir = await track().mkdir('lancejs')
+      const con = await lancedb.connect(dir)
+      const embeddingFunction = new TextEmbedding('name')
+
+      const names = vectorFromArray(['foo', 'bar'], new Utf8())
+      const data = new ArrowTable({ name: names })
+
+      const table = await con.createTable({ name: 'vectors', data, embeddingFunction })
+      assert.equal(table.name, 'vectors')
+      const results = await table.search('foo').execute()
+      assert.equal(results.length, 2)
+    })
   })
 })
 
@@ -362,3 +446,45 @@ describe('WriteOptions', function () {
     })
   })
 })
+
+describe('Compact and cleanup', function () {
+  it('can cleanup after compaction', async function () {
+    const dir = await track().mkdir('lancejs')
+    const con = await lancedb.connect(dir)
+
+    const data = [
+      { price: 10, name: 'foo', vector: [1, 2, 3] },
+      { price: 50, name: 'bar', vector: [4, 5, 6] }
+    ]
+    const table = await con.createTable('t1', data) as LocalTable
+
+    const newData = [
+      { price: 30, name: 'baz', vector: [7, 8, 9] }
+    ]
+    await table.add(newData)
+
+    const compactionMetrics = await table.compactFiles({
+      numThreads: 2
+    })
+    assert.equal(compactionMetrics.fragmentsRemoved, 2)
+    assert.equal(compactionMetrics.fragmentsAdded, 1)
+    assert.equal(await table.countRows(), 3)
+
+    await table.cleanupOldVersions()
+    assert.equal(await table.countRows(), 3)
+
+    // should have no effect, but this validates the arguments are parsed.
+    await table.compactFiles({
+      targetRowsPerFragment: 1024 * 10,
+      maxRowsPerGroup: 1024,
+      materializeDeletions: true,
+      materializeDeletionsThreshold: 0.5,
+      numThreads: 2
+    })
+
+    const cleanupMetrics = await table.cleanupOldVersions(0, true)
+    assert.isAtLeast(cleanupMetrics.bytesRemoved, 1)
+    assert.isAtLeast(cleanupMetrics.oldVersions, 1)
+    assert.equal(await table.countRows(), 3)
+  })
+})

python/.bumpversion.cfg

@@ -1,5 +1,5 @@
 [bumpversion]
-current_version = 0.1.16
+current_version = 0.3.1
 commit = True
 message = [python] Bump version: {current_version} → {new_version}
 tag = True

python/README.md

@@ -16,7 +16,7 @@ pip install lancedb
 import lancedb
 db = lancedb.connect('<PATH_TO_LANCEDB_DATASET>')
 table = db.open_table('my_table')
-results = table.search([0.1, 0.3]).limit(20).to_df()
+results = table.search([0.1, 0.3]).limit(20).to_list()
 print(results)
 ```
 

python/lancedb/__init__.py

@@ -11,11 +11,15 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 
+import importlib.metadata
 from typing import Optional
 
+__version__ = importlib.metadata.version("lancedb")
+
 from .db import URI, DBConnection, LanceDBConnection
 from .remote.db import RemoteDBConnection
 from .schema import vector
+from .utils import sentry_log
 
 
 def connect(
@@ -31,9 +35,13 @@ def connect(
     ----------
     uri: str or Path
         The uri of the database.
-    api_token: str, optional
+    api_key: str, optional
         If presented, connect to LanceDB cloud.
         Otherwise, connect to a database on file system or cloud storage.
+    region: str, default "us-west-2"
+        The region to use for LanceDB Cloud.
+    host_override: str, optional
+        The override url for LanceDB Cloud.
 
     Examples
     --------

python/lancedb/cli/__init__.py

@@ -0,0 +1,12 @@
+#  Copyright 2023 LanceDB Developers
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.

python/lancedb/cli/cli.py

@@ -0,0 +1,46 @@
+#  Copyright 2023 LanceDB Developers
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+
+import click
+
+from lancedb.utils import CONFIG
+
+
+@click.group()
+@click.version_option(help="LanceDB command line interface entry point")
+def cli():
+    "LanceDB command line interface"
+
+
+diagnostics_help = """
+Enable or disable LanceDB diagnostics. When enabled, LanceDB will send anonymous events to help us improve LanceDB.
+These diagnostics are used only for error reporting and no data is collected. You can find more about diagnosis on
+our docs: https://lancedb.github.io/lancedb/cli_config/
+"""
+
+
+@cli.command(help=diagnostics_help)
+@click.option("--enabled/--disabled", default=True)
+def diagnostics(enabled):
+    CONFIG.update({"diagnostics": True if enabled else False})
+    click.echo("LanceDB diagnostics is %s" % ("enabled" if enabled else "disabled"))
+
+
+@cli.command(help="Show current LanceDB configuration")
+def config():
+    # TODO: pretty print as table with colors and formatting
+    click.echo("Current LanceDB configuration:")
+    cfg = CONFIG.copy()
+    cfg.pop("uuid")  # Don't show uuid as it is not configurable
+    for item, amount in cfg.items():
+        click.echo("{} ({})".format(item, amount))

python/lancedb/conftest.py

@@ -1,7 +1,10 @@
 import os
 
+import numpy as np
 import pytest
 
+from .embeddings import EmbeddingFunctionRegistry, TextEmbeddingFunction
+
 # import lancedb so we don't have to in every example
 
 
@@ -14,3 +17,24 @@ def doctest_setup(monkeypatch, tmpdir):
     monkeypatch.setitem(os.environ, "COLUMNS", "80")
     # Work in a temporary directory
     monkeypatch.chdir(tmpdir)
+
+
+registry = EmbeddingFunctionRegistry.get_instance()
+
+
+@registry.register("test")
+class MockTextEmbeddingFunction(TextEmbeddingFunction):
+    """
+    Return the hash of the first 10 characters
+    """
+
+    def generate_embeddings(self, texts):
+        return [self._compute_one_embedding(row) for row in texts]
+
+    def _compute_one_embedding(self, row):
+        emb = np.array([float(hash(c)) for c in row[:10]])
+        emb /= np.linalg.norm(emb)
+        return emb
+
+    def ndims(self):
+        return 10

python/lancedb/context.py

@@ -12,6 +12,9 @@
 #  limitations under the License.
 from __future__ import annotations
 
+import deprecation
+
+from . import __version__
 from .exceptions import MissingColumnError, MissingValueError
 from .util import safe_import_pandas
 
@@ -43,7 +46,7 @@ def contextualize(raw_df: "pd.DataFrame") -> Contextualizer:
     this how many tokens, but depending on the input data, it could be sentences,
     paragraphs, messages, etc.
 
-    >>> contextualize(data).window(3).stride(1).text_col('token').to_df()
+    >>> contextualize(data).window(3).stride(1).text_col('token').to_pandas()
                     token  document_id
     0     The quick brown            1
     1     quick brown fox            1
@@ -56,7 +59,7 @@ def contextualize(raw_df: "pd.DataFrame") -> Contextualizer:
     8          dog I love            1
     9   I love sandwiches            2
     10    love sandwiches            2
-    >>> contextualize(data).window(7).stride(1).min_window_size(7).text_col('token').to_df()
+    >>> contextualize(data).window(7).stride(1).min_window_size(7).text_col('token').to_pandas()
                                       token  document_id
     0   The quick brown fox jumped over the            1
     1  quick brown fox jumped over the lazy            1
@@ -68,7 +71,7 @@ def contextualize(raw_df: "pd.DataFrame") -> Contextualizer:
     ``stride`` determines how many rows to skip between each window start. This can
     be used to reduce the total number of windows generated.
 
-    >>> contextualize(data).window(4).stride(2).text_col('token').to_df()
+    >>> contextualize(data).window(4).stride(2).text_col('token').to_pandas()
                         token  document_id
     0     The quick brown fox            1
     2   brown fox jumped over            1
@@ -81,7 +84,7 @@ def contextualize(raw_df: "pd.DataFrame") -> Contextualizer:
     context windows that don't cross document boundaries. In this case, we can
     pass ``document_id`` as the group by.
 
-    >>> contextualize(data).window(4).stride(2).text_col('token').groupby('document_id').to_df()
+    >>> contextualize(data).window(4).stride(2).text_col('token').groupby('document_id').to_pandas()
                        token  document_id
     0    The quick brown fox            1
     2  brown fox jumped over            1
@@ -93,14 +96,14 @@ def contextualize(raw_df: "pd.DataFrame") -> Contextualizer:
     This can be used to trim the last few context windows which have size less than
     ``min_window_size``. By default context windows of size 1 are skipped.
 
-    >>> contextualize(data).window(6).stride(3).text_col('token').groupby('document_id').to_df()
+    >>> contextualize(data).window(6).stride(3).text_col('token').groupby('document_id').to_pandas()
                                  token  document_id
     0  The quick brown fox jumped over            1
     3     fox jumped over the lazy dog            1
     6                     the lazy dog            1
     9                I love sandwiches            2
 
-    >>> contextualize(data).window(6).stride(3).min_window_size(4).text_col('token').groupby('document_id').to_df()
+    >>> contextualize(data).window(6).stride(3).min_window_size(4).text_col('token').groupby('document_id').to_pandas()
                                  token  document_id
     0  The quick brown fox jumped over            1
     3     fox jumped over the lazy dog            1
@@ -176,7 +179,16 @@ class Contextualizer:
         self._min_window_size = min_window_size
         return self
 
+    @deprecation.deprecated(
+        deprecated_in="0.3.1",
+        removed_in="0.4.0",
+        current_version=__version__,
+        details="Use the bar function instead",
+    )
     def to_df(self) -> "pd.DataFrame":
+        return self.to_pandas()
+
+    def to_pandas(self) -> "pd.DataFrame":
         """Create the context windows and return a DataFrame."""
         if pd is None:
             raise ImportError(

python/lancedb/db.py

@@ -16,12 +16,14 @@ from __future__ import annotations
 import os
 from abc import ABC, abstractmethod
 from pathlib import Path
-from typing import Optional
+from typing import List, Optional, Union
 
 import pyarrow as pa
 from pyarrow import fs
 
 from .common import DATA, URI
+from .embeddings import EmbeddingFunctionConfig
+from .pydantic import LanceModel
 from .table import LanceTable, Table
 from .util import fs_from_uri, get_uri_location, get_uri_scheme
 
@@ -39,7 +41,7 @@ class DBConnection(ABC):
         self,
         name: str,
         data: Optional[DATA] = None,
-        schema: Optional[pa.Schema] = None,
+        schema: Optional[Union[pa.Schema, LanceModel]] = None,
         mode: str = "create",
         on_bad_vectors: str = "error",
         fill_value: float = 0.0,
@@ -52,7 +54,7 @@ class DBConnection(ABC):
             The name of the table.
         data: list, tuple, dict, pd.DataFrame; optional
             The data to initialize the table. User must provide at least one of `data` or `schema`.
-        schema: pyarrow.Schema; optional
+        schema: pyarrow.Schema or LanceModel; optional
             The schema of the table.
         mode: str; default "create"
             The mode to use when creating the table. Can be either "create" or "overwrite".
@@ -148,14 +150,14 @@ class DBConnection(ABC):
         ...     for i in range(5):
         ...         yield pa.RecordBatch.from_arrays(
         ...             [
-        ...                 pa.array([[3.1, 4.1], [5.9, 26.5]]),
+        ...                 pa.array([[3.1, 4.1], [5.9, 26.5]], pa.list_(pa.float32(), 2)),
         ...                 pa.array(["foo", "bar"]),
         ...                 pa.array([10.0, 20.0]),
         ...             ],
         ...             ["vector", "item", "price"],
         ...         )
         >>> schema=pa.schema([
-        ...     pa.field("vector", pa.list_(pa.float32())),
+        ...     pa.field("vector", pa.list_(pa.float32(), 2)),
         ...     pa.field("item", pa.utf8()),
         ...     pa.field("price", pa.float32()),
         ... ])
@@ -192,6 +194,13 @@ class DBConnection(ABC):
         """
         raise NotImplementedError
 
+    def drop_database(self):
+        """
+        Drop database
+        This is the same thing as dropping all the tables
+        """
+        raise NotImplementedError
+
 
 class LanceDBConnection(DBConnection):
     """
@@ -277,10 +286,11 @@ class LanceDBConnection(DBConnection):
         self,
         name: str,
         data: Optional[DATA] = None,
-        schema: pa.Schema = None,
+        schema: Optional[Union[pa.Schema, LanceModel]] = None,
         mode: str = "create",
         on_bad_vectors: str = "error",
         fill_value: float = 0.0,
+        embedding_functions: Optional[List[EmbeddingFunctionConfig]] = None,
     ) -> LanceTable:
         """Create a table in the database.
 
@@ -299,6 +309,7 @@ class LanceDBConnection(DBConnection):
             mode=mode,
             on_bad_vectors=on_bad_vectors,
             fill_value=fill_value,
+            embedding_functions=embedding_functions,
         )
         return tbl
 
@@ -333,3 +344,7 @@ class LanceDBConnection(DBConnection):
         except FileNotFoundError:
             if not ignore_missing:
                 raise
+
+    def drop_database(self):
+        filesystem, path = fs_from_uri(self.uri)
+        filesystem.delete_dir(path)

python/lancedb/embeddings/__init__.py

@@ -0,0 +1,25 @@
+#  Copyright (c) 2023. LanceDB Developers
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+
+
+from .cohere import CohereEmbeddingFunction
+from .functions import (
+    EmbeddingFunction,
+    EmbeddingFunctionConfig,
+    EmbeddingFunctionRegistry,
+    OpenAIEmbeddings,
+    OpenClipEmbeddings,
+    SentenceTransformerEmbeddings,
+    TextEmbeddingFunction,
+)
+from .utils import with_embeddings

python/lancedb/embeddings/cohere.py

@@ -0,0 +1,86 @@
+#  Copyright (c) 2023. LanceDB Developers
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+
+import os
+from typing import ClassVar, List, Union
+
+import numpy as np
+
+from .functions import TextEmbeddingFunction, register
+from .utils import api_key_not_found_help
+
+
+@register("cohere")
+class CohereEmbeddingFunction(TextEmbeddingFunction):
+    """
+    An embedding function that uses the Cohere API
+
+    https://docs.cohere.com/docs/multilingual-language-models
+
+    Parameters
+    ----------
+    name: str, default "embed-multilingual-v2.0"
+        The name of the model to use. See the Cohere documentation for a list of available models.
+
+    Examples
+    --------
+    import lancedb
+    from lancedb.pydantic import LanceModel, Vector
+    from lancedb.embeddings import EmbeddingFunctionRegistry
+
+    cohere = EmbeddingFunctionRegistry.get_instance().get("cohere").create(name="embed-multilingual-v2.0")
+
+    class TextModel(LanceModel):
+        text: str = cohere.SourceField()
+        vector: Vector(cohere.ndims()) =  cohere.VectorField()
+
+    data = [ { "text": "hello world" },
+            { "text": "goodbye world" }]
+
+    db = lancedb.connect("~/.lancedb")
+    tbl = db.create_table("test", schema=TextModel, mode="overwrite")
+
+    tbl.add(data)
+
+    """
+
+    name: str = "embed-multilingual-v2.0"
+    client: ClassVar = None
+
+    def ndims(self):
+        # TODO: fix hardcoding
+        return 768
+
+    def generate_embeddings(
+        self, texts: Union[List[str], np.ndarray]
+    ) -> List[np.array]:
+        """
+        Get the embeddings for the given texts
+
+        Parameters
+        ----------
+        texts: list[str] or np.ndarray (of str)
+            The texts to embed
+        """
+        # TODO retry, rate limit, token limit
+        self._init_client()
+        rs = CohereEmbeddingFunction.client.embed(texts=texts, model=self.name)
+
+        return [emb for emb in rs.embeddings]
+
+    def _init_client(self):
+        cohere = self.safe_import("cohere")
+        if CohereEmbeddingFunction.client is None:
+            if os.environ.get("COHERE_API_KEY") is None:
+                api_key_not_found_help("cohere")
+            CohereEmbeddingFunction.client = cohere.Client(os.environ["COHERE_API_KEY"])

python/lancedb/embeddings/functions.py

@@ -0,0 +1,578 @@
+#  Copyright (c) 2023. LanceDB Developers
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+import concurrent.futures
+import importlib
+import io
+import json
+import os
+import socket
+import urllib.error
+import urllib.parse as urlparse
+import urllib.request
+from abc import ABC, abstractmethod
+from typing import Dict, List, Optional, Union
+
+import numpy as np
+import pyarrow as pa
+from cachetools import cached
+from pydantic import BaseModel, Field, PrivateAttr
+from tqdm import tqdm
+
+
+class EmbeddingFunctionRegistry:
+    """
+    This is a singleton class used to register embedding functions
+    and fetch them by name. It also handles serializing and deserializing.
+    You can implement your own embedding function by subclassing EmbeddingFunction
+    or TextEmbeddingFunction and registering it with the registry.
+
+    Examples
+    --------
+    >>> registry = EmbeddingFunctionRegistry.get_instance()
+    >>> @registry.register("my-embedding-function")
+    ... class MyEmbeddingFunction(EmbeddingFunction):
+    ...     def ndims(self) -> int:
+    ...         return 128
+    ...
+    ...     def compute_query_embeddings(self, query: str, *args, **kwargs) -> List[np.array]:
+    ...         return self.compute_source_embeddings(query, *args, **kwargs)
+    ...
+    ...     def compute_source_embeddings(self, texts: TEXT, *args, **kwargs) -> List[np.array]:
+    ...         return [np.random.rand(self.ndims()) for _ in range(len(texts))]
+    ...
+    >>> registry.get("my-embedding-function")
+    <class 'lancedb.embeddings.functions.MyEmbeddingFunction'>
+    """
+
+    @classmethod
+    def get_instance(cls):
+        return __REGISTRY__
+
+    def __init__(self):
+        self._functions = {}
+
+    def register(self, alias: str = None):
+        """
+        This creates a decorator that can be used to register
+        an EmbeddingFunction.
+
+        Parameters
+        ----------
+        alias : Optional[str]
+            a human friendly name for the embedding function. If not
+            provided, the class name will be used.
+        """
+
+        # This is a decorator for a class that inherits from BaseModel
+        # It adds the class to the registry
+        def decorator(cls):
+            if not issubclass(cls, EmbeddingFunction):
+                raise TypeError("Must be a subclass of EmbeddingFunction")
+            if cls.__name__ in self._functions:
+                raise KeyError(f"{cls.__name__} was already registered")
+            key = alias or cls.__name__
+            self._functions[key] = cls
+            cls.__embedding_function_registry_alias__ = alias
+            return cls
+
+        return decorator
+
+    def reset(self):
+        """
+        Reset the registry to its initial state
+        """
+        self._functions = {}
+
+    def get(self, name: str):
+        """
+        Fetch an embedding function class by name
+
+        Parameters
+        ----------
+        name : str
+            The name of the embedding function to fetch
+            Either the alias or the class name if no alias was provided
+            during registration
+        """
+        return self._functions[name]
+
+    def parse_functions(
+        self, metadata: Optional[Dict[bytes, bytes]]
+    ) -> Dict[str, "EmbeddingFunctionConfig"]:
+        """
+        Parse the metadata from an arrow table and
+        return a mapping of the vector column to the
+        embedding function and source column
+
+        Parameters
+        ----------
+        metadata : Optional[Dict[bytes, bytes]]
+            The metadata from an arrow table. Note that
+            the keys and values are bytes (pyarrow api)
+
+        Returns
+        -------
+        functions : dict
+            A mapping of vector column name to embedding function.
+            An empty dict is returned if input is None or does not
+            contain b"embedding_functions".
+        """
+        if metadata is None or b"embedding_functions" not in metadata:
+            return {}
+        serialized = metadata[b"embedding_functions"]
+        raw_list = json.loads(serialized.decode("utf-8"))
+        return {
+            obj["vector_column"]: EmbeddingFunctionConfig(
+                vector_column=obj["vector_column"],
+                source_column=obj["source_column"],
+                function=self.get(obj["name"])(**obj["model"]),
+            )
+            for obj in raw_list
+        }
+
+    def function_to_metadata(self, conf: "EmbeddingFunctionConfig"):
+        """
+        Convert the given embedding function and source / vector column configs
+        into a config dictionary that can be serialized into arrow metadata
+        """
+        func = conf.function
+        name = getattr(
+            func, "__embedding_function_registry_alias__", func.__class__.__name__
+        )
+        json_data = func.safe_model_dump()
+        return {
+            "name": name,
+            "model": json_data,
+            "source_column": conf.source_column,
+            "vector_column": conf.vector_column,
+        }
+
+    def get_table_metadata(self, func_list):
+        """
+        Convert a list of embedding functions and source / vector configs
+        into a config dictionary that can be serialized into arrow metadata
+        """
+        if func_list is None or len(func_list) == 0:
+            return None
+        json_data = [self.function_to_metadata(func) for func in func_list]
+        # Note that metadata dictionary values must be bytes
+        # so we need to json dump then utf8 encode
+        metadata = json.dumps(json_data, indent=2).encode("utf-8")
+        return {"embedding_functions": metadata}
+
+
+# Global instance
+__REGISTRY__ = EmbeddingFunctionRegistry()
+
+
+TEXT = Union[str, List[str], pa.Array, pa.ChunkedArray, np.ndarray]
+IMAGES = Union[
+    str, bytes, List[str], List[bytes], pa.Array, pa.ChunkedArray, np.ndarray
+]
+
+
+class EmbeddingFunction(BaseModel, ABC):
+    """
+    An ABC for embedding functions.
+
+    All concrete embedding functions must implement the following:
+    1. compute_query_embeddings() which takes a query and returns a list of embeddings
+    2. get_source_embeddings() which returns a list of embeddings for the source column
+    For text data, the two will be the same. For multi-modal data, the source column
+    might be images and the vector column might be text.
+    3. ndims method which returns the number of dimensions of the vector column
+    """
+
+    _ndims: int = PrivateAttr()
+
+    @classmethod
+    def create(cls, **kwargs):
+        """
+        Create an instance of the embedding function
+        """
+        return cls(**kwargs)
+
+    @abstractmethod
+    def compute_query_embeddings(self, *args, **kwargs) -> List[np.array]:
+        """
+        Compute the embeddings for a given user query
+        """
+        pass
+
+    @abstractmethod
+    def compute_source_embeddings(self, *args, **kwargs) -> List[np.array]:
+        """
+        Compute the embeddings for the source column in the database
+        """
+        pass
+
+    def sanitize_input(self, texts: TEXT) -> Union[List[str], np.ndarray]:
+        """
+        Sanitize the input to the embedding function.
+        """
+        if isinstance(texts, str):
+            texts = [texts]
+        elif isinstance(texts, pa.Array):
+            texts = texts.to_pylist()
+        elif isinstance(texts, pa.ChunkedArray):
+            texts = texts.combine_chunks().to_pylist()
+        return texts
+
+    @classmethod
+    def safe_import(cls, module: str, mitigation=None):
+        """
+        Import the specified module. If the module is not installed,
+        raise an ImportError with a helpful message.
+
+        Parameters
+        ----------
+        module : str
+            The name of the module to import
+        mitigation : Optional[str]
+            The package(s) to install to mitigate the error.
+            If not provided then the module name will be used.
+        """
+        try:
+            return importlib.import_module(module)
+        except ImportError:
+            raise ImportError(f"Please install {mitigation or module}")
+
+    def safe_model_dump(self):
+        from ..pydantic import PYDANTIC_VERSION
+
+        if PYDANTIC_VERSION.major < 2:
+            return dict(self)
+        return self.model_dump()
+
+    @abstractmethod
+    def ndims(self):
+        """
+        Return the dimensions of the vector column
+        """
+        pass
+
+    def SourceField(self, **kwargs):
+        """
+        Creates a pydantic Field that can automatically annotate
+        the source column for this embedding function
+        """
+        return Field(json_schema_extra={"source_column_for": self}, **kwargs)
+
+    def VectorField(self, **kwargs):
+        """
+        Creates a pydantic Field that can automatically annotate
+        the target vector column for this embedding function
+        """
+        return Field(json_schema_extra={"vector_column_for": self}, **kwargs)
+
+
+class EmbeddingFunctionConfig(BaseModel):
+    """
+    This model encapsulates the configuration for a embedding function
+    in a lancedb table. It holds the embedding function, the source column,
+    and the vector column
+    """
+
+    vector_column: str
+    source_column: str
+    function: EmbeddingFunction
+
+
+class TextEmbeddingFunction(EmbeddingFunction):
+    """
+    A callable ABC for embedding functions that take text as input
+    """
+
+    def compute_query_embeddings(self, query: str, *args, **kwargs) -> List[np.array]:
+        return self.compute_source_embeddings(query, *args, **kwargs)
+
+    def compute_source_embeddings(self, texts: TEXT, *args, **kwargs) -> List[np.array]:
+        texts = self.sanitize_input(texts)
+        return self.generate_embeddings(texts)
+
+    @abstractmethod
+    def generate_embeddings(
+        self, texts: Union[List[str], np.ndarray]
+    ) -> List[np.array]:
+        """
+        Generate the embeddings for the given texts
+        """
+        pass
+
+
+# @EmbeddingFunctionRegistry.get_instance().register(name) doesn't work in 3.8
+register = lambda name: EmbeddingFunctionRegistry.get_instance().register(name)
+
+
+@register("sentence-transformers")
+class SentenceTransformerEmbeddings(TextEmbeddingFunction):
+    """
+    An embedding function that uses the sentence-transformers library
+
+    https://huggingface.co/sentence-transformers
+    """
+
+    name: str = "all-MiniLM-L6-v2"
+    device: str = "cpu"
+    normalize: bool = True
+
+    def __init__(self, **kwargs):
+        super().__init__(**kwargs)
+        self._ndims = None
+
+    @property
+    def embedding_model(self):
+        """
+        Get the sentence-transformers embedding model specified by the
+        name and device. This is cached so that the model is only loaded
+        once per process.
+        """
+        return self.__class__.get_embedding_model(self.name, self.device)
+
+    def ndims(self):
+        if self._ndims is None:
+            self._ndims = len(self.generate_embeddings("foo")[0])
+        return self._ndims
+
+    def generate_embeddings(
+        self, texts: Union[List[str], np.ndarray]
+    ) -> List[np.array]:
+        """
+        Get the embeddings for the given texts
+
+        Parameters
+        ----------
+        texts: list[str] or np.ndarray (of str)
+            The texts to embed
+        """
+        return self.embedding_model.encode(
+            list(texts),
+            convert_to_numpy=True,
+            normalize_embeddings=self.normalize,
+        ).tolist()
+
+    @classmethod
+    @cached(cache={})
+    def get_embedding_model(cls, name, device):
+        """
+        Get the sentence-transformers embedding model specified by the
+        name and device. This is cached so that the model is only loaded
+        once per process.
+
+        Parameters
+        ----------
+        name : str
+            The name of the model to load
+        device : str
+            The device to load the model on
+
+        TODO: use lru_cache instead with a reasonable/configurable maxsize
+        """
+        sentence_transformers = cls.safe_import(
+            "sentence_transformers", "sentence-transformers"
+        )
+        return sentence_transformers.SentenceTransformer(name, device=device)
+
+
+@register("openai")
+class OpenAIEmbeddings(TextEmbeddingFunction):
+    """
+    An embedding function that uses the OpenAI API
+
+    https://platform.openai.com/docs/guides/embeddings
+    """
+
+    name: str = "text-embedding-ada-002"
+
+    def ndims(self):
+        # TODO don't hardcode this
+        return 1536
+
+    def generate_embeddings(
+        self, texts: Union[List[str], np.ndarray]
+    ) -> List[np.array]:
+        """
+        Get the embeddings for the given texts
+
+        Parameters
+        ----------
+        texts: list[str] or np.ndarray (of str)
+            The texts to embed
+        """
+        # TODO retry, rate limit, token limit
+        openai = self.safe_import("openai")
+        rs = openai.Embedding.create(input=texts, model=self.name)["data"]
+        return [v["embedding"] for v in rs]
+
+
+@register("open-clip")
+class OpenClipEmbeddings(EmbeddingFunction):
+    """
+    An embedding function that uses the OpenClip API
+    For multi-modal text-to-image search
+
+    https://github.com/mlfoundations/open_clip
+    """
+
+    name: str = "ViT-B-32"
+    pretrained: str = "laion2b_s34b_b79k"
+    device: str = "cpu"
+    batch_size: int = 64
+    normalize: bool = True
+    _model = PrivateAttr()
+    _preprocess = PrivateAttr()
+    _tokenizer = PrivateAttr()
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        open_clip = self.safe_import("open_clip", "open-clip")
+        model, _, preprocess = open_clip.create_model_and_transforms(
+            self.name, pretrained=self.pretrained
+        )
+        model.to(self.device)
+        self._model, self._preprocess = model, preprocess
+        self._tokenizer = open_clip.get_tokenizer(self.name)
+        self._ndims = None
+
+    def ndims(self):
+        if self._ndims is None:
+            self._ndims = self.generate_text_embeddings("foo").shape[0]
+        return self._ndims
+
+    def compute_query_embeddings(
+        self, query: Union[str, "PIL.Image.Image"], *args, **kwargs
+    ) -> List[np.ndarray]:
+        """
+        Compute the embeddings for a given user query
+
+        Parameters
+        ----------
+        query : Union[str, PIL.Image.Image]
+            The query to embed. A query can be either text or an image.
+        """
+        if isinstance(query, str):
+            return [self.generate_text_embeddings(query)]
+        else:
+            PIL = self.safe_import("PIL", "pillow")
+            if isinstance(query, PIL.Image.Image):
+                return [self.generate_image_embedding(query)]
+            else:
+                raise TypeError("OpenClip supports str or PIL Image as query")
+
+    def generate_text_embeddings(self, text: str) -> np.ndarray:
+        torch = self.safe_import("torch")
+        text = self.sanitize_input(text)
+        text = self._tokenizer(text)
+        text.to(self.device)
+        with torch.no_grad():
+            text_features = self._model.encode_text(text.to(self.device))
+            if self.normalize:
+                text_features /= text_features.norm(dim=-1, keepdim=True)
+            return text_features.cpu().numpy().squeeze()
+
+    def sanitize_input(self, images: IMAGES) -> Union[List[bytes], np.ndarray]:
+        """
+        Sanitize the input to the embedding function.
+        """
+        if isinstance(images, (str, bytes)):
+            images = [images]
+        elif isinstance(images, pa.Array):
+            images = images.to_pylist()
+        elif isinstance(images, pa.ChunkedArray):
+            images = images.combine_chunks().to_pylist()
+        return images
+
+    def compute_source_embeddings(
+        self, images: IMAGES, *args, **kwargs
+    ) -> List[np.array]:
+        """
+        Get the embeddings for the given images
+        """
+        images = self.sanitize_input(images)
+        embeddings = []
+        for i in range(0, len(images), self.batch_size):
+            j = min(i + self.batch_size, len(images))
+            batch = images[i:j]
+            embeddings.extend(self._parallel_get(batch))
+        return embeddings
+
+    def _parallel_get(self, images: Union[List[str], List[bytes]]) -> List[np.ndarray]:
+        """
+        Issue concurrent requests to retrieve the image data
+        """
+        with concurrent.futures.ThreadPoolExecutor() as executor:
+            futures = [
+                executor.submit(self.generate_image_embedding, image)
+                for image in images
+            ]
+            return [future.result() for future in tqdm(futures)]
+
+    def generate_image_embedding(
+        self, image: Union[str, bytes, "PIL.Image.Image"]
+    ) -> np.ndarray:
+        """
+        Generate the embedding for a single image
+
+        Parameters
+        ----------
+        image : Union[str, bytes, PIL.Image.Image]
+            The image to embed. If the image is a str, it is treated as a uri.
+            If the image is bytes, it is treated as the raw image bytes.
+        """
+        torch = self.safe_import("torch")
+        # TODO handle retry and errors for https
+        image = self._to_pil(image)
+        image = self._preprocess(image).unsqueeze(0)
+        with torch.no_grad():
+            return self._encode_and_normalize_image(image)
+
+    def _to_pil(self, image: Union[str, bytes]):
+        PIL = self.safe_import("PIL", "pillow")
+        if isinstance(image, bytes):
+            return PIL.Image.open(io.BytesIO(image))
+        if isinstance(image, PIL.Image.Image):
+            return image
+        elif isinstance(image, str):
+            parsed = urlparse.urlparse(image)
+            # TODO handle drive letter on windows.
+            if parsed.scheme == "file":
+                return PIL.Image.open(parsed.path)
+            elif parsed.scheme == "":
+                return PIL.Image.open(image if os.name == "nt" else parsed.path)
+            elif parsed.scheme.startswith("http"):
+                return PIL.Image.open(io.BytesIO(url_retrieve(image)))
+            else:
+                raise NotImplementedError("Only local and http(s) urls are supported")
+
+    def _encode_and_normalize_image(self, image_tensor: "torch.Tensor"):
+        """
+        encode a single image tensor and optionally normalize the output
+        """
+        image_features = self._model.encode_image(image_tensor.to(self.device))
+        if self.normalize:
+            image_features /= image_features.norm(dim=-1, keepdim=True)
+        return image_features.cpu().numpy().squeeze()
+
+
+def url_retrieve(url: str):
+    """
+    Parameters
+    ----------
+    url: str
+        URL to download from
+    """
+    try:
+        with urllib.request.urlopen(url) as conn:
+            return conn.read()
+    except (socket.gaierror, urllib.error.URLError) as err:
+        raise ConnectionError("could not download {} due to {}".format(url, err))

python/lancedb/embeddings/utils.py

@@ -1,4 +1,4 @@
-#  Copyright 2023 LanceDB Developers
+#  Copyright (c) 2023. LanceDB Developers
 #
 #  Licensed under the Apache License, Version 2.0 (the "License");
 #  you may not use this file except in compliance with the License.
@@ -20,7 +20,8 @@ import pyarrow as pa
 from lance.vector import vec_to_table
 from retry import retry
 
-from .util import safe_import_pandas
+from ..util import safe_import_pandas
+from ..utils.general import LOGGER
 
 pd = safe_import_pandas()
 DATA = Union[pa.Table, "pd.DataFrame"]
@@ -58,7 +59,7 @@ def with_embeddings(
     pa.Table
         The input table with a new column called "vector" containing the embeddings.
     """
-    func = EmbeddingFunction(func)
+    func = FunctionWrapper(func)
     if wrap_api:
         func = func.retry().rate_limit()
     func = func.batch_size(batch_size)
@@ -71,7 +72,11 @@ def with_embeddings(
     return data.append_column("vector", table["vector"])
 
 
-class EmbeddingFunction:
+class FunctionWrapper:
+    """
+    A wrapper for embedding functions that adds rate limiting, retries, and batching.
+    """
+
     def __init__(self, func: Callable):
         self.func = func
         self.rate_limiter_kwargs = {}
@@ -148,3 +153,8 @@ class EmbeddingFunction:
             yield from tqdm(_chunker(arr), total=math.ceil(length / self._batch_size))
         else:
             yield from _chunker(arr)
+
+
+def api_key_not_found_help(provider):
+    LOGGER.error(f"Could not find API key for {provider}.")
+    raise ValueError(f"Please set the {provider.upper()}_API_KEY environment variable.")

python/lancedb/pydantic.py

@@ -26,6 +26,8 @@ import pyarrow as pa
 import pydantic
 import semver
 
+from .embeddings import EmbeddingFunctionRegistry
+
 PYDANTIC_VERSION = semver.Version.parse(pydantic.__version__)
 try:
     from pydantic_core import CoreSchema, core_schema
@@ -46,7 +48,19 @@ class FixedSizeListMixin(ABC):
         raise NotImplementedError
 
 
-def vector(
+def vector(dim: int, value_type: pa.DataType = pa.float32()):
+    # TODO: remove in future release
+    from warnings import warn
+
+    warn(
+        "lancedb.pydantic.vector() is deprecated, use lancedb.pydantic.Vector instead."
+        "This function will be removed in future release",
+        DeprecationWarning,
+    )
+    return Vector(dim, value_type)
+
+
+def Vector(
     dim: int, value_type: pa.DataType = pa.float32()
 ) -> Type[FixedSizeListMixin]:
     """Pydantic Vector Type.
@@ -65,12 +79,12 @@ def vector(
     --------
 
     >>> import pydantic
-    >>> from lancedb.pydantic import vector
+    >>> from lancedb.pydantic import Vector
     ...
     >>> class MyModel(pydantic.BaseModel):
     ...     id: int
     ...     url: str
-    ...     embeddings: vector(768)
+    ...     embeddings: Vector(768)
     >>> schema = pydantic_to_schema(MyModel)
     >>> assert schema == pa.schema([
     ...     pa.field("id", pa.int64(), False),
@@ -114,7 +128,7 @@ def vector(
         def validate(cls, v):
             if not isinstance(v, (list, range, np.ndarray)) or len(v) != dim:
                 raise TypeError("A list of numbers or numpy.ndarray is needed")
-            return v
+            return cls(v)
 
         if PYDANTIC_VERSION < (2, 0):
 
@@ -224,27 +238,18 @@ def pydantic_to_schema(model: Type[pydantic.BaseModel]) -> pa.Schema:
     >>> from typing import List, Optional
     >>> import pydantic
     >>> from lancedb.pydantic import pydantic_to_schema
-    ...
-    >>> class InnerModel(pydantic.BaseModel):
-    ...     a: str
-    ...     b: Optional[float]
-    >>>
     >>> class FooModel(pydantic.BaseModel):
     ...     id: int
-    ...     s: Optional[str] = None
+    ...     s: str
     ...     vec: List[float]
     ...     li: List[int]
-    ...     inner: InnerModel
+    ...
     >>> schema = pydantic_to_schema(FooModel)
     >>> assert schema == pa.schema([
     ...     pa.field("id", pa.int64(), False),
-    ...     pa.field("s", pa.utf8(), True),
+    ...     pa.field("s", pa.utf8(), False),
     ...     pa.field("vec", pa.list_(pa.float64()), False),
     ...     pa.field("li", pa.list_(pa.int64()), False),
-    ...     pa.field("inner", pa.struct([
-    ...         pa.field("a", pa.utf8(), False),
-    ...         pa.field("b", pa.float64(), True),
-    ...     ]), False),
     ... ])
     """
     fields = _pydantic_model_to_fields(model)
@@ -258,11 +263,11 @@ class LanceModel(pydantic.BaseModel):
     Examples
     --------
     >>> import lancedb
-    >>> from lancedb.pydantic import LanceModel, vector
+    >>> from lancedb.pydantic import LanceModel, Vector
     >>>
     >>> class TestModel(LanceModel):
     ...     name: str
-    ...     vector: vector(2)
+    ...     vector: Vector(2)
     ...
     >>> db = lancedb.connect("/tmp")
     >>> table = db.create_table("test", schema=TestModel.to_arrow_schema())
@@ -275,10 +280,61 @@ class LanceModel(pydantic.BaseModel):
 
     @classmethod
     def to_arrow_schema(cls):
-        return pydantic_to_schema(cls)
+        """
+        Get the Arrow Schema for this model.
+        """
+        schema = pydantic_to_schema(cls)
+        functions = cls.parse_embedding_functions()
+        if len(functions) > 0:
+            metadata = EmbeddingFunctionRegistry.get_instance().get_table_metadata(
+                functions
+            )
+            schema = schema.with_metadata(metadata)
+        return schema
 
     @classmethod
     def field_names(cls) -> List[str]:
+        """
+        Get the field names of this model.
+        """
+        return list(cls.safe_get_fields().keys())
+
+    @classmethod
+    def safe_get_fields(cls):
         if PYDANTIC_VERSION.major < 2:
-            return list(cls.__fields__.keys())
-        return list(cls.model_fields.keys())
+            return cls.__fields__
+        return cls.model_fields
+
+    @classmethod
+    def parse_embedding_functions(cls) -> List["EmbeddingFunctionConfig"]:
+        """
+        Parse the embedding functions from this model.
+        """
+        from .embeddings import EmbeddingFunctionConfig
+
+        vec_and_function = []
+        for name, field_info in cls.safe_get_fields().items():
+            func = get_extras(field_info, "vector_column_for")
+            if func is not None:
+                vec_and_function.append([name, func])
+
+        configs = []
+        for vec, func in vec_and_function:
+            for source, field_info in cls.safe_get_fields().items():
+                src_func = get_extras(field_info, "source_column_for")
+                if src_func == func:
+                    configs.append(
+                        EmbeddingFunctionConfig(
+                            source_column=source, vector_column=vec, function=func
+                        )
+                    )
+        return configs
+
+
+def get_extras(field_info: pydantic.fields.FieldInfo, key: str) -> Any:
+    """
+    Get the extra metadata from a Pydantic FieldInfo.
+    """
+    if PYDANTIC_VERSION.major >= 2:
+        return (field_info.json_schema_extra or {}).get(key)
+    return (field_info.field_info.extra or {}).get("json_schema_extra", {}).get(key)

python/lancedb/query.py

@@ -13,12 +13,15 @@
 
 from __future__ import annotations
 
+from abc import ABC, abstractmethod
 from typing import List, Literal, Optional, Type, Union
 
+import deprecation
 import numpy as np
 import pyarrow as pa
 import pydantic
 
+from . import __version__
 from .common import VECTOR_COLUMN_NAME
 from .pydantic import LanceModel
 from .util import safe_import_pandas
@@ -37,6 +40,9 @@ class Query(pydantic.BaseModel):
     # sql filter to refine the query with
     filter: Optional[str] = None
 
+    # if True then apply the filter before vector search
+    prefilter: bool = False
+
     # top k results to return
     k: int
 
@@ -54,44 +60,139 @@ class Query(pydantic.BaseModel):
     refine_factor: Optional[int] = None
 
 
-class LanceQueryBuilder:
-    """
-    A builder for nearest neighbor queries for LanceDB.
-
-    Examples
-    --------
-    >>> import lancedb
-    >>> data = [{"vector": [1.1, 1.2], "b": 2},
-    ...         {"vector": [0.5, 1.3], "b": 4},
-    ...         {"vector": [0.4, 0.4], "b": 6},
-    ...         {"vector": [0.4, 0.4], "b": 10}]
-    >>> db = lancedb.connect("./.lancedb")
-    >>> table = db.create_table("my_table", data=data)
-    >>> (table.search([0.4, 0.4])
-    ...       .metric("cosine")
-    ...       .where("b < 10")
-    ...       .select(["b"])
-    ...       .limit(2)
-    ...       .to_df())
-       b      vector  score
-    0  6  [0.4, 0.4]    0.0
-    """
-
-    def __init__(
-        self,
+class LanceQueryBuilder(ABC):
+    @classmethod
+    def create(
+        cls,
         table: "lancedb.table.Table",
-        query: Union[np.ndarray, str],
-        vector_column: str = VECTOR_COLUMN_NAME,
-    ):
-        self._metric = "L2"
-        self._nprobes = 20
-        self._refine_factor = None
+        query: Optional[Union[np.ndarray, str, "PIL.Image.Image"]],
+        query_type: str,
+        vector_column_name: str,
+    ) -> LanceQueryBuilder:
+        if query is None:
+            return LanceEmptyQueryBuilder(table)
+
+        # convert "auto" query_type to "vector" or "fts"
+        # and convert the query to vector if needed
+        query, query_type = cls._resolve_query(
+            table, query, query_type, vector_column_name
+        )
+
+        if isinstance(query, str):
+            # fts
+            return LanceFtsQueryBuilder(table, query)
+
+        if isinstance(query, list):
+            query = np.array(query, dtype=np.float32)
+        elif isinstance(query, np.ndarray):
+            query = query.astype(np.float32)
+        else:
+            raise TypeError(f"Unsupported query type: {type(query)}")
+
+        return LanceVectorQueryBuilder(table, query, vector_column_name)
+
+    @classmethod
+    def _resolve_query(cls, table, query, query_type, vector_column_name):
+        # If query_type is fts, then query must be a string.
+        # otherwise raise TypeError
+        if query_type == "fts":
+            if not isinstance(query, str):
+                raise TypeError(f"'fts' queries must be a string: {type(query)}")
+            return query, query_type
+        elif query_type == "vector":
+            if not isinstance(query, (list, np.ndarray)):
+                conf = table.embedding_functions.get(vector_column_name)
+                if conf is not None:
+                    query = conf.function.compute_query_embeddings(query)[0]
+                else:
+                    msg = f"No embedding function for {vector_column_name}"
+                    raise ValueError(msg)
+            return query, query_type
+        elif query_type == "auto":
+            if isinstance(query, (list, np.ndarray)):
+                return query, "vector"
+            else:
+                conf = table.embedding_functions.get(vector_column_name)
+                if conf is not None:
+                    query = conf.function.compute_query_embeddings(query)[0]
+                    return query, "vector"
+                else:
+                    return query, "fts"
+        else:
+            raise ValueError(
+                f"Invalid query_type, must be 'vector', 'fts', or 'auto': {query_type}"
+            )
+
+    def __init__(self, table: "lancedb.table.Table"):
         self._table = table
-        self._query = query
         self._limit = 10
         self._columns = None
         self._where = None
-        self._vector_column = vector_column
+
+    @deprecation.deprecated(
+        deprecated_in="0.3.1",
+        removed_in="0.4.0",
+        current_version=__version__,
+        details="Use the bar function instead",
+    )
+    def to_df(self) -> "pd.DataFrame":
+        """
+        Deprecated alias for `to_pandas()`. Please use `to_pandas()` instead.
+
+        Execute the query and return the results as a pandas DataFrame.
+        In addition to the selected columns, LanceDB also returns a vector
+        and also the "_distance" column which is the distance between the query
+        vector and the returned vector.
+        """
+        return self.to_pandas()
+
+    def to_pandas(self) -> "pd.DataFrame":
+        """
+        Execute the query and return the results as a pandas DataFrame.
+        In addition to the selected columns, LanceDB also returns a vector
+        and also the "_distance" column which is the distance between the query
+        vector and the returned vector.
+        """
+        return self.to_arrow().to_pandas()
+
+    @abstractmethod
+    def to_arrow(self) -> pa.Table:
+        """
+        Execute the query and return the results as an
+        [Apache Arrow Table](https://arrow.apache.org/docs/python/generated/pyarrow.Table.html#pyarrow.Table).
+
+        In addition to the selected columns, LanceDB also returns a vector
+        and also the "_distance" column which is the distance between the query
+        vector and the returned vectors.
+        """
+        raise NotImplementedError
+
+    def to_list(self) -> List[dict]:
+        """
+        Execute the query and return the results as a list of dictionaries.
+
+        Each list entry is a dictionary with the selected column names as keys,
+        or all table columns if `select` is not called. The vector and the "_distance"
+        fields are returned whether or not they're explicitly selected.
+        """
+        return self.to_arrow().to_pylist()
+
+    def to_pydantic(self, model: Type[LanceModel]) -> List[LanceModel]:
+        """Return the table as a list of pydantic models.
+
+        Parameters
+        ----------
+        model: Type[LanceModel]
+            The pydantic model to use.
+
+        Returns
+        -------
+        List[LanceModel]
+        """
+        return [
+            model(**{k: v for k, v in row.items() if k in model.field_names()})
+            for row in self.to_arrow().to_pylist()
+        ]
 
     def limit(self, limit: int) -> LanceQueryBuilder:
         """Set the maximum number of results to return.
@@ -125,7 +226,7 @@ class LanceQueryBuilder:
         self._columns = columns
         return self
 
-    def where(self, where: str) -> LanceQueryBuilder:
+    def where(self, where) -> LanceQueryBuilder:
         """Set the where clause.
 
         Parameters
@@ -141,7 +242,45 @@ class LanceQueryBuilder:
         self._where = where
         return self
 
-    def metric(self, metric: Literal["L2", "cosine"]) -> LanceQueryBuilder:
+
+class LanceVectorQueryBuilder(LanceQueryBuilder):
+    """
+    A builder for nearest neighbor queries for LanceDB.
+
+    Examples
+    --------
+    >>> import lancedb
+    >>> data = [{"vector": [1.1, 1.2], "b": 2},
+    ...         {"vector": [0.5, 1.3], "b": 4},
+    ...         {"vector": [0.4, 0.4], "b": 6},
+    ...         {"vector": [0.4, 0.4], "b": 10}]
+    >>> db = lancedb.connect("./.lancedb")
+    >>> table = db.create_table("my_table", data=data)
+    >>> (table.search([0.4, 0.4])
+    ...       .metric("cosine")
+    ...       .where("b < 10")
+    ...       .select(["b"])
+    ...       .limit(2)
+    ...       .to_pandas())
+       b      vector  _distance
+    0  6  [0.4, 0.4]        0.0
+    """
+
+    def __init__(
+        self,
+        table: "lancedb.table.Table",
+        query: Union[np.ndarray, list, "PIL.Image.Image"],
+        vector_column: str = VECTOR_COLUMN_NAME,
+    ):
+        super().__init__(table)
+        self._query = query
+        self._metric = "L2"
+        self._nprobes = 20
+        self._refine_factor = None
+        self._vector_column = vector_column
+        self._prefilter = False
+
+    def metric(self, metric: Literal["L2", "cosine"]) -> LanceVectorQueryBuilder:
         """Set the distance metric to use.
 
         Parameters
@@ -151,13 +290,13 @@ class LanceQueryBuilder:
 
         Returns
         -------
-        LanceQueryBuilder
+        LanceVectorQueryBuilder
             The LanceQueryBuilder object.
         """
         self._metric = metric
         return self
 
-    def nprobes(self, nprobes: int) -> LanceQueryBuilder:
+    def nprobes(self, nprobes: int) -> LanceVectorQueryBuilder:
         """Set the number of probes to use.
 
         Higher values will yield better recall (more likely to find vectors if
@@ -173,13 +312,13 @@ class LanceQueryBuilder:
 
         Returns
         -------
-        LanceQueryBuilder
+        LanceVectorQueryBuilder
             The LanceQueryBuilder object.
         """
         self._nprobes = nprobes
         return self
 
-    def refine_factor(self, refine_factor: int) -> LanceQueryBuilder:
+    def refine_factor(self, refine_factor: int) -> LanceVectorQueryBuilder:
         """Set the refine factor to use, increasing the number of vectors sampled.
 
         As an example, a refine factor of 2 will sample 2x as many vectors as
@@ -195,35 +334,26 @@ class LanceQueryBuilder:
 
         Returns
         -------
-        LanceQueryBuilder
+        LanceVectorQueryBuilder
             The LanceQueryBuilder object.
         """
         self._refine_factor = refine_factor
         return self
 
-    def to_df(self) -> "pd.DataFrame":
-        """
-        Execute the query and return the results as a pandas DataFrame.
-        In addition to the selected columns, LanceDB also returns a vector
-        and also the "score" column which is the distance between the query
-        vector and the returned vector.
-        """
-
-        return self.to_arrow().to_pandas()
-
     def to_arrow(self) -> pa.Table:
         """
         Execute the query and return the results as an
         [Apache Arrow Table](https://arrow.apache.org/docs/python/generated/pyarrow.Table.html#pyarrow.Table).
 
         In addition to the selected columns, LanceDB also returns a vector
-        and also the "score" column which is the distance between the query
+        and also the "_distance" column which is the distance between the query
         vector and the returned vectors.
         """
         vector = self._query if isinstance(self._query, list) else self._query.tolist()
         query = Query(
             vector=vector,
             filter=self._where,
+            prefilter=self._prefilter,
             k=self._limit,
             metric=self._metric,
             columns=self._columns,
@@ -233,25 +363,36 @@ class LanceQueryBuilder:
         )
         return self._table._execute_query(query)
 
-    def to_pydantic(self, model: Type[LanceModel]) -> List[LanceModel]:
-        """Return the table as a list of pydantic models.
+    def where(self, where: str, prefilter: bool = False) -> LanceVectorQueryBuilder:
+        """Set the where clause.
 
         Parameters
         ----------
-        model: Type[LanceModel]
-            The pydantic model to use.
+        where: str
+            The where clause.
+        prefilter: bool, default False
+            If True, apply the filter before vector search, otherwise the
+            filter is applied on the result of vector search.
+            This feature is **EXPERIMENTAL** and may be removed and modified
+            without warning in the future. Currently this is only supported
+            in OSS and can only be used with a table that does not have an ANN
+            index.
 
         Returns
         -------
-        List[LanceModel]
+        LanceQueryBuilder
+            The LanceQueryBuilder object.
         """
-        return [
-            model(**{k: v for k, v in row.items() if k in model.field_names()})
-            for row in self.to_arrow().to_pylist()
-        ]
+        self._where = where
+        self._prefilter = prefilter
+        return self
 
 
 class LanceFtsQueryBuilder(LanceQueryBuilder):
+    def __init__(self, table: "lancedb.table.Table", query: str):
+        super().__init__(table)
+        self._query = query
+
     def to_arrow(self) -> pa.Table:
         try:
             import tantivy
@@ -275,3 +416,13 @@ class LanceFtsQueryBuilder(LanceQueryBuilder):
         output_tbl = self._table.to_lance().take(row_ids, columns=self._columns)
         output_tbl = output_tbl.append_column("score", scores)
         return output_tbl
+
+
+class LanceEmptyQueryBuilder(LanceQueryBuilder):
+    def to_arrow(self) -> pa.Table:
+        ds = self._table.to_lance()
+        return ds.to_table(
+            columns=self._columns,
+            filter=self._where,
+            limit=self._limit,
+        )

python/lancedb/remote/__init__.py

@@ -14,7 +14,7 @@
 import abc
 from typing import List, Optional
 
-import attr
+import attrs
 import pyarrow as pa
 from pydantic import BaseModel
 
@@ -44,7 +44,7 @@ class VectorQuery(BaseModel):
     refine_factor: Optional[int] = None
 
 
-@attr.define
+@attrs.define
 class VectorQueryResult:
     # for now the response is directly seralized into a pandas dataframe
     tbl: pa.Table

python/lancedb/remote/client.py

@@ -16,7 +16,7 @@ import functools
 from typing import Any, Callable, Dict, Optional, Union
 
 import aiohttp
-import attr
+import attrs
 import pyarrow as pa
 from pydantic import BaseModel
 
@@ -43,14 +43,14 @@ async def _read_ipc(resp: aiohttp.ClientResponse) -> pa.Table:
         return reader.read_all()
 
 
-@attr.define(slots=False)
+@attrs.define(slots=False)
 class RestfulLanceDBClient:
     db_name: str
     region: str
     api_key: Credential
-    host_override: Optional[str] = attr.field(default=None)
+    host_override: Optional[str] = attrs.field(default=None)
 
-    closed: bool = attr.field(default=False, init=False)
+    closed: bool = attrs.field(default=False, init=False)
 
     @functools.cached_property
     def session(self) -> aiohttp.ClientSession:
@@ -97,7 +97,12 @@ class RestfulLanceDBClient:
         """Send a GET request and returns the deserialized response payload."""
         if isinstance(params, BaseModel):
             params: Dict[str, Any] = params.dict(exclude_none=True)
-        async with self.session.get(uri, params=params, headers=self.headers) as resp:
+        async with self.session.get(
+            uri,
+            params=params,
+            headers=self.headers,
+            timeout=aiohttp.ClientTimeout(total=30),
+        ) as resp:
             await self._check_status(resp)
             return await resp.json()
 
@@ -109,6 +114,7 @@ class RestfulLanceDBClient:
         params: Optional[Dict[str, Any]] = None,
         content_type: Optional[str] = None,
         deserialize: Callable = lambda resp: resp.json(),
+        request_id: Optional[str] = None,
     ) -> Dict[str, Any]:
         """Send a POST request and returns the deserialized response payload.
 
@@ -117,6 +123,8 @@ class RestfulLanceDBClient:
         uri : str
             The uri to send the POST request to.
         data: Union[Dict[str, Any], BaseModel]
+        request_id: Optional[str]
+            Optional client side request id to be sent in the request headers.
 
         """
         if isinstance(data, BaseModel):
@@ -129,10 +137,13 @@ class RestfulLanceDBClient:
         headers = self.headers.copy()
         if content_type is not None:
             headers["content-type"] = content_type
+        if request_id is not None:
+            headers["x-request-id"] = request_id
         async with self.session.post(
             uri,
             headers=headers,
             params=params,
+            timeout=aiohttp.ClientTimeout(total=30),
             **req_kwargs,
         ) as resp:
             resp: aiohttp.ClientResponse = resp

python/lancedb/remote/db.py

@@ -18,10 +18,9 @@ from urllib.parse import urlparse
 
 import pyarrow as pa
 
-from lancedb.common import DATA
-from lancedb.db import DBConnection
-from lancedb.table import Table, _sanitize_data
-
+from ..common import DATA
+from ..db import DBConnection
+from ..table import Table, _sanitize_data
 from .arrow import to_ipc_binary
 from .client import ARROW_STREAM_CONTENT_TYPE, RestfulLanceDBClient
 
@@ -104,8 +103,22 @@ class RemoteDBConnection(DBConnection):
             self._client.post(
                 f"/v1/table/{name}/create/",
                 data=data,
-                params={"request_id": request_id},
+                request_id=request_id,
                 content_type=ARROW_STREAM_CONTENT_TYPE,
             )
         )
         return RemoteTable(self, name)
+
+    def drop_table(self, name: str):
+        """Drop a table from the database.
+
+        Parameters
+        ----------
+        name: str
+            The name of the table.
+        """
+        self._loop.run_until_complete(
+            self._client.post(
+                f"/v1/table/{name}/drop/",
+            )
+        )

python/lancedb/remote/table.py

@@ -13,14 +13,14 @@
 
 import uuid
 from functools import cached_property
-from typing import Union
+from typing import Optional, Union
 
 import pyarrow as pa
 from lance import json_to_schema
 
 from lancedb.common import DATA, VEC, VECTOR_COLUMN_NAME
 
-from ..query import LanceQueryBuilder
+from ..query import LanceVectorQueryBuilder
 from ..table import Query, Table, _sanitize_data
 from .arrow import to_ipc_binary
 from .client import ARROW_STREAM_CONTENT_TYPE
@@ -45,7 +45,15 @@ class RemoteTable(Table):
         return schema
 
     def to_arrow(self) -> pa.Table:
-        raise NotImplementedError
+        """Return the table as an Arrow table."""
+        raise NotImplementedError("to_arrow() is not supported on the LanceDB cloud")
+
+    def to_pandas(self):
+        """Return the table as a Pandas DataFrame.
+
+        Intercept `to_arrow()` for better error message.
+        """
+        return NotImplementedError("to_pandas() is not supported on the LanceDB cloud")
 
     def create_index(
         self,
@@ -54,6 +62,7 @@ class RemoteTable(Table):
         num_sub_vectors=96,
         vector_column_name: str = VECTOR_COLUMN_NAME,
         replace: bool = True,
+        accelerator: Optional[str] = None,
     ):
         raise NotImplementedError
 
@@ -65,7 +74,11 @@ class RemoteTable(Table):
         fill_value: float = 0.0,
     ) -> int:
         data = _sanitize_data(
-            data, self.schema, on_bad_vectors=on_bad_vectors, fill_value=fill_value
+            data,
+            self.schema,
+            metadata=None,
+            on_bad_vectors=on_bad_vectors,
+            fill_value=fill_value,
         )
         payload = to_ipc_binary(data)
 
@@ -81,11 +94,13 @@ class RemoteTable(Table):
         )
 
     def search(
-        self, query: Union[VEC, str], vector_column: str = VECTOR_COLUMN_NAME
-    ) -> LanceQueryBuilder:
-        return LanceQueryBuilder(self, query, vector_column)
+        self, query: Union[VEC, str], vector_column_name: str = VECTOR_COLUMN_NAME
+    ) -> LanceVectorQueryBuilder:
+        return LanceVectorQueryBuilder(self, query, vector_column_name)
 
     def _execute_query(self, query: Query) -> pa.Table:
+        if query.prefilter:
+            raise NotImplementedError("Cloud support for prefiltering is coming soon")
         result = self._conn._client.query(self._name, query)
         return self._conn._loop.run_until_complete(result).to_arrow()
 

python/lancedb/table.py

@@ -13,51 +13,106 @@
 
 from __future__ import annotations
 
+import inspect
 import os
 from abc import ABC, abstractmethod
+from datetime import timedelta
 from functools import cached_property
-from typing import Iterable, List, Union
+from typing import Any, Iterable, List, Optional, Union
 
 import lance
 import numpy as np
 import pyarrow as pa
 import pyarrow.compute as pc
 from lance import LanceDataset
+from lance.dataset import CleanupStats, ReaderLike
 from lance.vector import vec_to_table
 
 from .common import DATA, VEC, VECTOR_COLUMN_NAME
+from .embeddings import EmbeddingFunctionRegistry
+from .embeddings.functions import EmbeddingFunctionConfig
 from .pydantic import LanceModel
-from .query import LanceFtsQueryBuilder, LanceQueryBuilder, Query
+from .query import LanceQueryBuilder, Query
 from .util import fs_from_uri, safe_import_pandas
+from .utils.events import register_event
 
 pd = safe_import_pandas()
 
 
-def _sanitize_data(data, schema, on_bad_vectors, fill_value):
+def _sanitize_data(
+    data,
+    schema: Optional[pa.Schema],
+    metadata: Optional[dict],
+    on_bad_vectors: str,
+    fill_value: Any,
+):
     if isinstance(data, list):
         # convert to list of dict if data is a bunch of LanceModels
         if isinstance(data[0], LanceModel):
             schema = data[0].__class__.to_arrow_schema()
             data = [dict(d) for d in data]
         data = pa.Table.from_pylist(data)
-        data = _sanitize_schema(
-            data, schema=schema, on_bad_vectors=on_bad_vectors, fill_value=fill_value
-        )
-    if isinstance(data, dict):
+    elif isinstance(data, dict):
         data = vec_to_table(data)
-    if pd is not None and isinstance(data, pd.DataFrame):
-        data = pa.Table.from_pandas(data)
+    elif pd is not None and isinstance(data, pd.DataFrame):
+        data = pa.Table.from_pandas(data, preserve_index=False)
+        # Do not serialize Pandas metadata
+        meta = data.schema.metadata if data.schema.metadata is not None else {}
+        meta = {k: v for k, v in meta.items() if k != b"pandas"}
+        data = data.replace_schema_metadata(meta)
+
+    if isinstance(data, pa.Table):
+        if metadata:
+            data = _append_vector_col(data, metadata, schema)
+            metadata.update(data.schema.metadata or {})
+            data = data.replace_schema_metadata(metadata)
         data = _sanitize_schema(
             data, schema=schema, on_bad_vectors=on_bad_vectors, fill_value=fill_value
         )
-    if not isinstance(data, (pa.Table, Iterable)):
+    elif isinstance(data, Iterable):
+        data = _to_record_batch_generator(
+            data, schema, metadata, on_bad_vectors, fill_value
+        )
+    else:
         raise TypeError(f"Unsupported data type: {type(data)}")
     return data
 
 
+def _append_vector_col(data: pa.Table, metadata: dict, schema: Optional[pa.Schema]):
+    """
+    Use the embedding function to automatically embed the source column and add the
+    vector column to the table.
+    """
+    functions = EmbeddingFunctionRegistry.get_instance().parse_functions(metadata)
+    for vector_column, conf in functions.items():
+        func = conf.function
+        if vector_column not in data.column_names:
+            col_data = func.compute_source_embeddings(data[conf.source_column])
+            if schema is not None:
+                dtype = schema.field(vector_column).type
+            else:
+                dtype = pa.list_(pa.float32(), len(col_data[0]))
+            data = data.append_column(
+                pa.field(vector_column, type=dtype), pa.array(col_data, type=dtype)
+            )
+    return data
+
+
+def _to_record_batch_generator(
+    data: Iterable, schema, metadata, on_bad_vectors, fill_value
+):
+    for batch in data:
+        if not isinstance(batch, pa.RecordBatch):
+            table = _sanitize_data(batch, schema, metadata, on_bad_vectors, fill_value)
+            for batch in table.to_batches():
+                yield batch
+        else:
+            yield batch
+
+
 class Table(ABC):
     """
-    A [Table](Table) is a collection of Records in a LanceDB [Database](Database).
+    A Table is a collection of Records in a LanceDB Database.
 
     Examples
     --------
@@ -83,8 +138,8 @@ class Table(ABC):
 
     Can query the table with [Table.search][lancedb.table.Table.search].
 
-    >>> table.search([0.4, 0.4]).select(["b"]).to_df()
-       b      vector  score
+    >>> table.search([0.4, 0.4]).select(["b"]).to_pandas()
+       b      vector  _distance
     0  4  [0.5, 1.3]       0.82
     1  2  [1.1, 1.2]       1.13
 
@@ -92,9 +147,10 @@ class Table(ABC):
     [Table.create_index][lancedb.table.Table.create_index].
     """
 
+    @property
     @abstractmethod
     def schema(self) -> pa.Schema:
-        """Return the [Arrow Schema](https://arrow.apache.org/docs/python/api/datatypes.html#) of
+        """The [Arrow Schema](https://arrow.apache.org/docs/python/api/datatypes.html#) of
         this [Table](Table)
 
         """
@@ -126,6 +182,7 @@ class Table(ABC):
         num_sub_vectors=96,
         vector_column_name: str = VECTOR_COLUMN_NAME,
         replace: bool = True,
+        accelerator: Optional[str] = None,
     ):
         """Create an index on the table.
 
@@ -146,6 +203,9 @@ class Table(ABC):
         replace: bool, default True
             If True, replace the existing index if it exists.
             If False, raise an error if duplicate index exists.
+        accelerator: str, default None
+            If set, use the given accelerator to create the index.
+            Only support "cuda" for now.
         """
         raise NotImplementedError
 
@@ -177,24 +237,37 @@ class Table(ABC):
 
     @abstractmethod
     def search(
-        self, query: Union[VEC, str], vector_column: str = VECTOR_COLUMN_NAME
+        self,
+        query: Optional[Union[VEC, str, "PIL.Image.Image"]] = None,
+        vector_column_name: str = VECTOR_COLUMN_NAME,
+        query_type: str = "auto",
     ) -> LanceQueryBuilder:
         """Create a search query to find the nearest neighbors
         of the given query vector.
 
         Parameters
         ----------
-        query: list, np.ndarray
-            The query vector.
-        vector_column: str, default "vector"
+        query: str, list, np.ndarray, PIL.Image.Image, default None
+            The query to search for. If None then
+            the select/where/limit clauses are applied to filter
+            the table
+        vector_column_name: str, default "vector"
             The name of the vector column to search.
+        query_type: str, default "auto"
+            "vector", "fts", or "auto"
+            If "auto" then the query type is inferred from the query;
+            If `query` is a list/np.ndarray then the query type is "vector";
+            If `query` is a PIL.Image.Image then either do vector search
+            or raise an error if no corresponding embedding function is found.
+            If `query` is a string, then the query type is "vector" if the
+            table has embedding functions else the query type is "fts"
 
         Returns
         -------
         LanceQueryBuilder
             A query builder object representing the query.
             Once executed, the query returns selected columns, the vector,
-            and also the "score" column which is the distance between the query
+            and also the "_distance" column which is the distance between the query
             vector and the returned vector.
         """
         raise NotImplementedError
@@ -261,10 +334,11 @@ class LanceTable(Table):
         self.name = name
         self._version = version
 
-    def _reset_dataset(self):
+    def _reset_dataset(self, version=None):
         try:
             if "_dataset" in self.__dict__:
                 del self.__dict__["_dataset"]
+            self._version = version
         except AttributeError:
             pass
 
@@ -290,7 +364,9 @@ class LanceTable(Table):
     def checkout(self, version: int):
         """Checkout a version of the table. This is an in-place operation.
 
-        This allows viewing previous versions of the table.
+        This allows viewing previous versions of the table. If you wish to
+        keep writing to the dataset starting from an old version, then use
+        the `restore` function.
 
         Parameters
         ----------
@@ -303,14 +379,14 @@ class LanceTable(Table):
         >>> db = lancedb.connect("./.lancedb")
         >>> table = db.create_table("my_table", [{"vector": [1.1, 0.9], "type": "vector"}])
         >>> table.version
-        1
+        2
         >>> table.to_pandas()
                vector    type
         0  [1.1, 0.9]  vector
         >>> table.add([{"vector": [0.5, 0.2], "type": "vector"}])
         >>> table.version
-        2
-        >>> table.checkout(1)
+        3
+        >>> table.checkout(2)
         >>> table.to_pandas()
                vector    type
         0  [1.1, 0.9]  vector
@@ -318,7 +394,65 @@ class LanceTable(Table):
         max_ver = max([v["version"] for v in self._dataset.versions()])
         if version < 1 or version > max_ver:
             raise ValueError(f"Invalid version {version}")
-        self._version = version
+        self._reset_dataset(version=version)
+
+        try:
+            # Accessing the property updates the cached value
+            _ = self._dataset
+        except Exception as e:
+            if "not found" in str(e):
+                raise ValueError(
+                    f"Version {version} no longer exists. Was it cleaned up?"
+                )
+            else:
+                raise e
+
+    def restore(self, version: int = None):
+        """Restore a version of the table. This is an in-place operation.
+
+        This creates a new version where the data is equivalent to the
+        specified previous version. Data is not copied (as of python-v0.2.1).
+
+        Parameters
+        ----------
+        version : int, default None
+            The version to restore. If unspecified then restores the currently
+            checked out version. If the currently checked out version is the
+            latest version then this is a no-op.
+
+        Examples
+        --------
+        >>> import lancedb
+        >>> db = lancedb.connect("./.lancedb")
+        >>> table = db.create_table("my_table", [{"vector": [1.1, 0.9], "type": "vector"}])
+        >>> table.version
+        2
+        >>> table.to_pandas()
+               vector    type
+        0  [1.1, 0.9]  vector
+        >>> table.add([{"vector": [0.5, 0.2], "type": "vector"}])
+        >>> table.version
+        3
+        >>> table.restore(2)
+        >>> table.to_pandas()
+               vector    type
+        0  [1.1, 0.9]  vector
+        >>> len(table.list_versions())
+        4
+        """
+        max_ver = max([v["version"] for v in self._dataset.versions()])
+        if version is None:
+            version = self.version
+        elif version < 1 or version > max_ver:
+            raise ValueError(f"Invalid version {version}")
+        else:
+            self.checkout(version)
+
+        if version == max_ver:
+            # no-op if restoring the latest version
+            return
+
+        self._dataset.restore()
         self._reset_dataset()
 
     def __len__(self):
@@ -362,6 +496,7 @@ class LanceTable(Table):
         num_sub_vectors=96,
         vector_column_name=VECTOR_COLUMN_NAME,
         replace: bool = True,
+        accelerator: Optional[str] = None,
     ):
         """Create an index on the table."""
         self._dataset.create_index(
@@ -371,8 +506,10 @@ class LanceTable(Table):
             num_partitions=num_partitions,
             num_sub_vectors=num_sub_vectors,
             replace=replace,
+            accelerator=accelerator,
         )
         self._reset_dataset()
+        register_event("create_index")
 
     def create_fts_index(self, field_names: Union[str, List[str]]):
         """Create a full-text search index on the table.
@@ -391,6 +528,7 @@ class LanceTable(Table):
             field_names = [field_names]
         index = create_index(self._get_fts_index_path(), field_names)
         populate_index(index, self, field_names)
+        register_event("create_fts_index")
 
     def _get_fts_index_path(self):
         return os.path.join(self._dataset_uri, "_indices", "tantivy")
@@ -411,6 +549,9 @@ class LanceTable(Table):
         fill_value: float = 0.0,
     ):
         """Add data to the table.
+        If vector columns are missing and the table
+        has embedding functions, then the vector columns
+        are automatically computed and added.
 
         Parameters
         ----------
@@ -432,43 +573,133 @@ class LanceTable(Table):
         """
         # TODO: manage table listing and metadata separately
         data = _sanitize_data(
-            data, self.schema, on_bad_vectors=on_bad_vectors, fill_value=fill_value
+            data,
+            self.schema,
+            metadata=self.schema.metadata,
+            on_bad_vectors=on_bad_vectors,
+            fill_value=fill_value,
         )
-        lance.write_dataset(data, self._dataset_uri, mode=mode)
+        lance.write_dataset(data, self._dataset_uri, schema=self.schema, mode=mode)
         self._reset_dataset()
+        register_event("add")
+
+    def merge(
+        self,
+        other_table: Union[LanceTable, ReaderLike],
+        left_on: str,
+        right_on: Optional[str] = None,
+        schema: Optional[Union[pa.Schema, LanceModel]] = None,
+    ):
+        """Merge another table into this table.
+
+        Performs a left join, where the dataset is the left side and other_table
+        is the right side. Rows existing in the dataset but not on the left will
+        be filled with null values, unless Lance doesn't support null values for
+        some types, in which case an error will be raised. The only overlapping
+        column allowed is the join column. If other overlapping columns exist,
+        an error will be raised.
+
+        Parameters
+        ----------
+        other_table: LanceTable or Reader-like
+            The data to be merged. Acceptable types are:
+            - Pandas DataFrame, Pyarrow Table, Dataset, Scanner,
+            Iterator[RecordBatch], or RecordBatchReader
+            - LanceTable
+        left_on: str
+            The name of the column in the dataset to join on.
+        right_on: str or None
+            The name of the column in other_table to join on. If None, defaults to
+            left_on.
+        schema: pa.Schema or LanceModel, optional
+            The schema of the other_table.
+            If not provided, the schema is inferred from the data.
+
+        Examples
+        --------
+        >>> import lancedb
+        >>> import pyarrow as pa
+        >>> df = pa.table({'x': [1, 2, 3], 'y': ['a', 'b', 'c']})
+        >>> db = lancedb.connect("./.lancedb")
+        >>> table = db.create_table("dataset", df)
+        >>> table.to_pandas()
+           x  y
+        0  1  a
+        1  2  b
+        2  3  c
+        >>> new_df = pa.table({'x': [1, 2, 3], 'z': ['d', 'e', 'f']})
+        >>> table.merge(new_df, 'x')
+        >>> table.to_pandas()
+           x  y  z
+        0  1  a  d
+        1  2  b  e
+        2  3  c  f
+        """
+        if isinstance(schema, LanceModel):
+            schema = schema.to_arrow_schema()
+        if isinstance(other_table, LanceTable):
+            other_table = other_table.to_lance()
+        if isinstance(other_table, LanceDataset):
+            other_table = other_table.to_table()
+        self._dataset.merge(
+            other_table, left_on=left_on, right_on=right_on, schema=schema
+        )
+        self._reset_dataset()
+        register_event("merge")
+
+    @cached_property
+    def embedding_functions(self) -> dict:
+        """
+        Get the embedding functions for the table
+
+        Returns
+        -------
+        funcs: dict
+            A mapping of the vector column to the embedding function
+            or empty dict if not configured.
+        """
+        return EmbeddingFunctionRegistry.get_instance().parse_functions(
+            self.schema.metadata
+        )
 
     def search(
-        self, query: Union[VEC, str], vector_column_name=VECTOR_COLUMN_NAME
+        self,
+        query: Optional[Union[VEC, str, "PIL.Image.Image"]] = None,
+        vector_column_name: str = VECTOR_COLUMN_NAME,
+        query_type: str = "auto",
     ) -> LanceQueryBuilder:
         """Create a search query to find the nearest neighbors
         of the given query vector.
 
         Parameters
         ----------
-        query: list, np.ndarray
-            The query vector.
+        query: str, list, np.ndarray, a PIL Image or None
+            The query to search for. If None then
+            the select/where/limit clauses are applied to filter
+            the table
         vector_column_name: str, default "vector"
             The name of the vector column to search.
+        query_type: str, default "auto"
+            "vector", "fts", or "auto"
+            If "auto" then the query type is inferred from the query;
+            If `query` is a list/np.ndarray then the query type is "vector";
+            If `query` is a PIL.Image.Image then either do vector search
+            or raise an error if no corresponding embedding function is found.
+            If the query is a string, then the query type is "vector" if the
+            table has embedding functions, else the query type is "fts"
 
         Returns
         -------
         LanceQueryBuilder
             A query builder object representing the query.
             Once executed, the query returns selected columns, the vector,
-            and also the "score" column which is the distance between the query
+            and also the "_distance" column which is the distance between the query
             vector and the returned vector.
         """
-        if isinstance(query, str):
-            # fts
-            return LanceFtsQueryBuilder(self, query, vector_column_name)
-
-        if isinstance(query, list):
-            query = np.array(query)
-        if isinstance(query, np.ndarray):
-            query = query.astype(np.float32)
-        else:
-            raise TypeError(f"Unsupported query type: {type(query)}")
-        return LanceQueryBuilder(self, query, vector_column_name)
+        register_event("search")
+        return LanceQueryBuilder.create(
+            self, query, query_type, vector_column_name=vector_column_name
+        )
 
     @classmethod
     def create(
@@ -480,6 +711,7 @@ class LanceTable(Table):
         mode="create",
         on_bad_vectors: str = "error",
         fill_value: float = 0.0,
+        embedding_functions: List[EmbeddingFunctionConfig] = None,
     ):
         """
         Create a new table.
@@ -506,7 +738,7 @@ class LanceTable(Table):
         data: list-of-dict, dict, pd.DataFrame, default None
             The data to insert into the table.
             At least one of `data` or `schema` must be provided.
-        schema: dict, optional
+        schema: pa.Schema or LanceModel, optional
             The schema of the table. If not provided, the schema is inferred from the data.
             At least one of `data` or `schema` must be provided.
         mode: str, default "create"
@@ -517,18 +749,59 @@ class LanceTable(Table):
             One of "error", "drop", "fill".
         fill_value: float, default 0.
             The value to use when filling vectors. Only used if on_bad_vectors="fill".
+        embedding_functions: list of EmbeddingFunctionModel, default None
+            The embedding functions to use when creating the table.
         """
         tbl = LanceTable(db, name)
+        if inspect.isclass(schema) and issubclass(schema, LanceModel):
+            # convert LanceModel to pyarrow schema
+            # note that it's possible this contains
+            # embedding function metadata already
+            schema = schema.to_arrow_schema()
+
+        metadata = None
+        if embedding_functions is not None:
+            # If we passed in embedding functions explicitly
+            # then we'll override any schema metadata that
+            # may was implicitly specified by the LanceModel schema
+            registry = EmbeddingFunctionRegistry.get_instance()
+            metadata = registry.get_table_metadata(embedding_functions)
+
         if data is not None:
             data = _sanitize_data(
-                data, schema, on_bad_vectors=on_bad_vectors, fill_value=fill_value
+                data,
+                schema,
+                metadata=metadata,
+                on_bad_vectors=on_bad_vectors,
+                fill_value=fill_value,
             )
-        else:
+
         if schema is None:
+            if data is None:
                 raise ValueError("Either data or schema must be provided")
-            data = pa.Table.from_pylist([], schema=schema)
-        lance.write_dataset(data, tbl._dataset_uri, schema=schema, mode=mode)
-        return LanceTable(db, name)
+            elif hasattr(data, "schema"):
+                schema = data.schema
+            elif isinstance(data, Iterable):
+                if metadata:
+                    raise TypeError(
+                        (
+                            "Persistent embedding functions not yet "
+                            "supported for generator data input"
+                        )
+                    )
+
+        if metadata:
+            schema = schema.with_metadata(metadata)
+
+        empty = pa.Table.from_pylist([], schema=schema)
+        lance.write_dataset(empty, tbl._dataset_uri, schema=schema, mode=mode)
+        table = LanceTable(db, name)
+
+        if data is not None:
+            table.add(data)
+
+        register_event("create_table")
+        return table
 
     @classmethod
     def open(cls, db, name):
@@ -544,11 +817,69 @@ class LanceTable(Table):
     def delete(self, where: str):
         self._dataset.delete(where)
 
+    def update(self, where: str, values: dict):
+        """
+        EXPERIMENTAL: Update rows in the table (not threadsafe).
+
+        This can be used to update zero to all rows depending on how many
+        rows match the where clause.
+
+        Parameters
+        ----------
+        where: str
+            The SQL where clause to use when updating rows. For example, 'x = 2'
+            or 'x IN (1, 2, 3)'. The filter must not be empty, or it will error.
+        values: dict
+            The values to update. The keys are the column names and the values
+            are the values to set.
+
+        Examples
+        --------
+        >>> import lancedb
+        >>> import pandas as pd
+        >>> data = pd.DataFrame({"x": [1, 2, 3], "vector": [[1, 2], [3, 4], [5, 6]]})
+        >>> db = lancedb.connect("./.lancedb")
+        >>> table = db.create_table("my_table", data)
+        >>> table.to_pandas()
+           x      vector
+        0  1  [1.0, 2.0]
+        1  2  [3.0, 4.0]
+        2  3  [5.0, 6.0]
+        >>> table.update(where="x = 2", values={"vector": [10, 10]})
+        >>> table.to_pandas()
+           x        vector
+        0  1    [1.0, 2.0]
+        1  3    [5.0, 6.0]
+        2  2  [10.0, 10.0]
+
+        """
+        orig_data = self._dataset.to_table(filter=where).combine_chunks()
+        if len(orig_data) == 0:
+            return
+        for col, val in values.items():
+            i = orig_data.column_names.index(col)
+            if i < 0:
+                raise ValueError(f"Column {col} does not exist")
+            orig_data = orig_data.set_column(
+                i, col, pa.array([val] * len(orig_data), type=orig_data[col].type)
+            )
+        self.delete(where)
+        self.add(orig_data, mode="append")
+        self._reset_dataset()
+        register_event("update")
+
     def _execute_query(self, query: Query) -> pa.Table:
         ds = self.to_lance()
+        if query.prefilter:
+            for idx in ds.list_indices():
+                if query.vector_column in idx["fields"]:
+                    raise NotImplementedError(
+                        "Prefiltering for indexed vector column is coming soon."
+                    )
         return ds.to_table(
             columns=query.columns,
             filter=query.filter,
+            prefilter=query.prefilter,
             nearest={
                 "column": query.vector_column,
                 "q": query.vector,
@@ -559,6 +890,48 @@ class LanceTable(Table):
             },
         )
 
+    def cleanup_old_versions(
+        self,
+        older_than: Optional[timedelta] = None,
+        *,
+        delete_unverified: bool = False,
+    ) -> CleanupStats:
+        """
+        Clean up old versions of the table, freeing disk space.
+
+        Parameters
+        ----------
+        older_than: timedelta, default None
+            The minimum age of the version to delete. If None, then this defaults
+            to two weeks.
+        delete_unverified: bool, default False
+            Because they may be part of an in-progress transaction, files newer
+            than 7 days old are not deleted by default. If you are sure that
+            there are no in-progress transactions, then you can set this to True
+            to delete all files older than `older_than`.
+
+        Returns
+        -------
+        CleanupStats
+            The stats of the cleanup operation, including how many bytes were
+            freed.
+        """
+        return self.to_lance().cleanup_old_versions(
+            older_than, delete_unverified=delete_unverified
+        )
+
+    def compact_files(self, *args, **kwargs):
+        """
+        Run the compaction process on the table.
+
+        This can be run after making several small appends to optimize the table
+        for faster reads.
+
+        Arguments are passed onto :meth:`lance.dataset.DatasetOptimizer.compact_files`.
+        For most cases, the default should be fine.
+        """
+        return self.to_lance().optimize.compact_files(*args, **kwargs)
+
 
 def _sanitize_schema(
     data: pa.Table,
@@ -586,16 +959,30 @@ def _sanitize_schema(
             return data
         # cast the columns to the expected types
         data = data.combine_chunks()
+        for field in schema:
+            # TODO: we're making an assumption that fixed size list of 10 or more
+            # is a vector column. This is definitely a bit hacky.
+            likely_vector_col = (
+                pa.types.is_fixed_size_list(field.type)
+                and pa.types.is_float32(field.type.value_type)
+                and field.type.list_size >= 10
+            )
+            is_default_vector_col = field.name == VECTOR_COLUMN_NAME
+            if field.name in data.column_names and (
+                likely_vector_col or is_default_vector_col
+            ):
                 data = _sanitize_vector_column(
                     data,
-            vector_column_name=VECTOR_COLUMN_NAME,
+                    vector_column_name=field.name,
                     on_bad_vectors=on_bad_vectors,
                     fill_value=fill_value,
                 )
         return pa.Table.from_arrays(
             [data[name] for name in schema.names], schema=schema
         )
+
     # just check the vector column
+    if VECTOR_COLUMN_NAME in data.column_names:
         return _sanitize_vector_column(
             data,
             vector_column_name=VECTOR_COLUMN_NAME,
@@ -603,6 +990,8 @@ def _sanitize_schema(
             fill_value=fill_value,
         )
 
+    return data
+
 
 def _sanitize_vector_column(
     data: pa.Table,
@@ -625,8 +1014,6 @@ def _sanitize_vector_column(
     fill_value: float, default 0.0
         The value to use when filling vectors. Only used if on_bad_vectors="fill".
     """
-    if vector_column_name not in data.column_names:
-        raise ValueError(f"Missing vector column: {vector_column_name}")
     # ChunkedArray is annoying to work with, so we combine chunks here
     vec_arr = data[vector_column_name].combine_chunks()
     if pa.types.is_list(data[vector_column_name].type):

python/lancedb/util.py

@@ -70,7 +70,11 @@ def fs_from_uri(uri: str) -> Tuple[pa_fs.FileSystem, str]:
     Get a PyArrow FileSystem from a URI, handling extra environment variables.
     """
     if get_uri_scheme(uri) == "s3":
-        fs = pa_fs.S3FileSystem(endpoint_override=os.environ.get("AWS_ENDPOINT"))
+        fs = pa_fs.S3FileSystem(
+            endpoint_override=os.environ.get("AWS_ENDPOINT"),
+            request_timeout=30,
+            connect_timeout=30,
+        )
         path = get_uri_location(uri)
         return fs, path
 

python/lancedb/utils/__init__.py

@@ -0,0 +1,15 @@
+#  Copyright 2023 LanceDB Developers
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+from .config import Config
+
+CONFIG = Config()

python/lancedb/utils/config.py

@@ -0,0 +1,116 @@
+#  Copyright 2023 LanceDB Developers
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+
+import copy
+import hashlib
+import os
+import platform
+import uuid
+from pathlib import Path
+
+from .general import LOGGER, is_dir_writeable, yaml_load, yaml_save
+
+
+def get_user_config_dir(sub_dir="lancedb"):
+    """
+    Get the user config directory.
+
+    Args:
+        sub_dir (str): The name of the subdirectory to create.
+
+    Returns:
+        (Path): The path to the user config directory.
+    """
+    # Return the appropriate config directory for each operating system
+    if platform.system() == "Windows":
+        path = Path.home() / "AppData" / "Roaming" / sub_dir
+    elif platform.system() == "Darwin":
+        path = Path.home() / "Library" / "Application Support" / sub_dir
+    elif platform.system() == "Linux":
+        path = Path.home() / ".config" / sub_dir
+    else:
+        raise ValueError(f"Unsupported operating system: {platform.system()}")
+
+    # GCP and AWS lambda fix, only /tmp is writeable
+    if not is_dir_writeable(path.parent):
+        LOGGER.warning(
+            f"WARNING ⚠️ user config directory '{path}' is not writeable, defaulting to '/tmp' or CWD."
+            "Alternatively you can define a LANCEDB_CONFIG_DIR environment variable for this path."
+        )
+        path = (
+            Path("/tmp") / sub_dir
+            if is_dir_writeable("/tmp")
+            else Path().cwd() / sub_dir
+        )
+
+    # Create the subdirectory if it does not exist
+    path.mkdir(parents=True, exist_ok=True)
+
+    return path
+
+
+USER_CONFIG_DIR = Path(os.getenv("LANCEDB_CONFIG_DIR") or get_user_config_dir())
+CONFIG_FILE = USER_CONFIG_DIR / "config.yaml"
+
+
+class Config(dict):
+    """
+    Manages lancedb config stored in a YAML file.
+
+    Args:
+        file (str | Path): Path to the lancedb config YAML file. Default is USER_CONFIG_DIR / 'config.yaml'.
+    """
+
+    def __init__(self, file=CONFIG_FILE):
+        self.file = Path(file)
+        self.defaults = {  # Default global config values
+            "diagnostics": True,
+            "uuid": hashlib.sha256(str(uuid.getnode()).encode()).hexdigest(),
+        }
+
+        super().__init__(copy.deepcopy(self.defaults))
+
+        if not self.file.exists():
+            self.save()
+
+        self.load()
+        correct_keys = self.keys() == self.defaults.keys()
+        correct_types = all(
+            type(a) is type(b) for a, b in zip(self.values(), self.defaults.values())
+        )
+        if not (correct_keys and correct_types):
+            LOGGER.warning(
+                "WARNING ⚠️ LanceDB settings reset to default values. This may be due to a possible problem "
+                "with your settings or a recent package update. "
+                f"\nView settings & usage with 'lancedb settings' or at '{self.file}'"
+            )
+            self.reset()
+
+    def load(self):
+        """Loads settings from the YAML file."""
+        super().update(yaml_load(self.file))
+
+    def save(self):
+        """Saves the current settings to the YAML file."""
+        yaml_save(self.file, dict(self))
+
+    def update(self, *args, **kwargs):
+        """Updates a setting value in the current settings."""
+        super().update(*args, **kwargs)
+        self.save()
+
+    def reset(self):
+        """Resets the settings to default and saves them."""
+        self.clear()
+        self.update(self.defaults)
+        self.save()

python/lancedb/utils/events.py

@@ -0,0 +1,161 @@
+#  Copyright 2023 LanceDB Developers
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+
+import datetime
+import importlib.metadata
+import platform
+import random
+import sys
+import time
+
+from lancedb.utils import CONFIG
+from lancedb.utils.general import TryExcept
+
+from .general import (
+    PLATFORMS,
+    get_git_origin_url,
+    is_git_dir,
+    is_github_actions_ci,
+    is_online,
+    is_pip_package,
+    is_pytest_running,
+    threaded_request,
+)
+
+
+class _Events:
+    """
+    A class for collecting anonymous event analytics. Event analytics are enabled when ``diagnostics=True`` in config and
+    disabled when ``diagnostics=False``.
+
+    You can enable or disable diagnostics by running ``lancedb diagnostics --enabled`` or ``lancedb diagnostics --disabled``.
+
+    Attributes
+    ----------
+    url : str
+        The URL to send anonymous events.
+    rate_limit : float
+        The rate limit in seconds for sending events.
+    metadata : dict
+        A dictionary containing metadata about the environment.
+    enabled : bool
+        A flag to enable or disable Events based on certain conditions.
+    """
+
+    _instance = None
+
+    url = "https://app.posthog.com/capture/"
+    headers = {"Content-Type": "application/json"}
+    api_key = "phc_oENDjGgHtmIDrV6puUiFem2RB4JA8gGWulfdulmMdZP"
+    # This api-key is write only and is safe to expose in the codebase.
+
+    def __init__(self):
+        """
+        Initializes the Events object with default values for events, rate_limit, and metadata.
+        """
+        self.events = []  # events list
+        self.max_events = 25  # max events to store in memory
+        self.rate_limit = 60.0  # rate limit (seconds)
+        self.time = 0.0
+
+        if is_git_dir():
+            install = "git"
+        elif is_pip_package():
+            install = "pip"
+        else:
+            install = "other"
+        self.metadata = {
+            "cli": sys.argv[0],
+            "install": install,
+            "python": ".".join(platform.python_version_tuple()[:2]),
+            "version": importlib.metadata.version("lancedb"),
+            "platforms": PLATFORMS,
+            "session_id": round(random.random() * 1e15),
+            # 'engagement_time_msec': 1000 # TODO: In future we might be interested in this metric
+        }
+
+        TESTS_RUNNING = is_pytest_running() or is_github_actions_ci()
+        ONLINE = is_online()
+        self.enabled = (
+            CONFIG["diagnostics"]
+            and not TESTS_RUNNING
+            and ONLINE
+            and (
+                is_pip_package()
+                or get_git_origin_url() == "https://github.com/lancedb/lancedb.git"
+            )
+        )
+
+    def __call__(self, event_name, params={}):
+        """
+        Attempts to add a new event to the events list and send events if the rate limit is reached.
+
+        Args
+        ----
+        event_name : str
+            The name of the event to be logged.
+        params : dict, optional
+            A dictionary of additional parameters to be logged with the event.
+        """
+        ### NOTE: We might need a way to tag a session with a label to check usage from a source. Setting label should be exposed to the user.
+        if not self.enabled:
+            return
+        if (
+            len(self.events) < self.max_events
+        ):  # Events list limited to 25 events (drop any events past this)
+            params.update(self.metadata)
+            self.events.append(
+                {
+                    "event": event_name,
+                    "properties": params,
+                    "timestamp": datetime.datetime.now(
+                        tz=datetime.timezone.utc
+                    ).isoformat(),
+                    "distinct_id": CONFIG["uuid"],
+                }
+            )
+
+        # Check rate limit
+        t = time.time()
+        if (t - self.time) < self.rate_limit:
+            return
+        # Time is over rate limiter, send now
+        data = {
+            "api_key": self.api_key,
+            "distinct_id": CONFIG["uuid"],  # posthog needs this to accepts the event
+            "batch": self.events,
+        }
+
+        # POST equivalent to requests.post(self.url, json=data).
+        # threaded request is used to avoid blocking, retries are disabled, and verbose is disabled
+        # to avoid any possible disruption in the console.
+        threaded_request(
+            method="post",
+            url=self.url,
+            headers=self.headers,
+            json=data,
+            retry=0,
+            verbose=False,
+        )
+
+        # Flush & Reset
+        self.events = []
+        self.time = t
+
+
+@TryExcept(verbose=False)
+def register_event(name: str, **kwargs):
+    if _Events._instance is None:
+        _Events._instance = _Events()
+
+    _Events._instance(name, **kwargs)

python/lancedb/utils/general.py

@@ -0,0 +1,445 @@
+#  Copyright 2023 LanceDB Developers
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+
+import contextlib
+import importlib
+import logging.config
+import os
+import platform
+import subprocess
+import sys
+import threading
+import time
+from pathlib import Path
+from typing import Union
+
+import requests
+import yaml
+
+LOGGING_NAME = "lancedb"
+VERBOSE = (
+    str(os.getenv("LANCEDB_VERBOSE", True)).lower() == "true"
+)  # global verbose mode
+
+
+def set_logging(name=LOGGING_NAME, verbose=True):
+    """Sets up logging for the given name.
+
+    Parameters
+    ----------
+    name : str, optional
+        The name of the logger. Default is 'lancedb'.
+    verbose : bool, optional
+        Whether to enable verbose logging. Default is True.
+    """
+
+    rank = int(os.getenv("RANK", -1))  # rank in world for Multi-GPU trainings
+    level = logging.INFO if verbose and rank in {-1, 0} else logging.ERROR
+    logging.config.dictConfig(
+        {
+            "version": 1,
+            "disable_existing_loggers": False,
+            "formatters": {name: {"format": "%(message)s"}},
+            "handlers": {
+                name: {
+                    "class": "logging.StreamHandler",
+                    "formatter": name,
+                    "level": level,
+                }
+            },
+            "loggers": {name: {"level": level, "handlers": [name], "propagate": False}},
+        }
+    )
+
+
+set_logging(LOGGING_NAME, verbose=VERBOSE)
+LOGGER = logging.getLogger(LOGGING_NAME)
+
+
+def is_pip_package(filepath: str = __name__) -> bool:
+    """Determines if the file at the given filepath is part of a pip package.
+
+    Parameters
+    ----------
+    filepath : str, optional
+        The filepath to check. Default is the current file.
+
+    Returns
+    -------
+    bool
+        True if the file is part of a pip package, False otherwise.
+    """
+    # Get the spec for the module
+    spec = importlib.util.find_spec(filepath)
+
+    # Return whether the spec is not None and the origin is not None (indicating it is a package)
+    return spec is not None and spec.origin is not None
+
+
+def is_pytest_running():
+    """Determines whether pytest is currently running or not.
+
+    Returns
+    -------
+    bool
+        True if pytest is running, False otherwise.
+    """
+    return (
+        ("PYTEST_CURRENT_TEST" in os.environ)
+        or ("pytest" in sys.modules)
+        or ("pytest" in Path(sys.argv[0]).stem)
+    )
+
+
+def is_github_actions_ci() -> bool:
+    """
+    Determine if the current environment is a GitHub Actions CI Python runner.
+
+    Returns
+    -------
+    bool
+        True if the current environment is a GitHub Actions CI Python runner, False otherwise.
+    """
+
+    return (
+        "GITHUB_ACTIONS" in os.environ
+        and "RUNNER_OS" in os.environ
+        and "RUNNER_TOOL_CACHE" in os.environ
+    )
+
+
+def is_git_dir():
+    """
+    Determines whether the current file is part of a git repository.
+    If the current file is not part of a git repository, returns None.
+
+    Returns
+    -------
+    bool
+        True if current file is part of a git repository.
+    """
+    return get_git_dir() is not None
+
+
+def is_online() -> bool:
+    """
+    Check internet connectivity by attempting to connect to a known online host.
+
+    Returns
+    -------
+    bool
+        True if connection is successful, False otherwise.
+    """
+    import socket
+
+    for host in "1.1.1.1", "8.8.8.8", "223.5.5.5":  # Cloudflare, Google, AliDNS:
+        try:
+            test_connection = socket.create_connection(address=(host, 53), timeout=2)
+        except (socket.timeout, socket.gaierror, OSError):
+            continue
+        else:
+            # If the connection was successful, close it to avoid a ResourceWarning
+            test_connection.close()
+            return True
+    return False
+
+
+def is_dir_writeable(dir_path: Union[str, Path]) -> bool:
+    """Check if a directory is writeable.
+
+    Parameters
+    ----------
+    dir_path : Union[str, Path]
+        The path to the directory.
+
+    Returns
+    -------
+    bool
+        True if the directory is writeable, False otherwise.
+    """
+    return os.access(str(dir_path), os.W_OK)
+
+
+def is_colab():
+    """Check if the current script is running inside a Google Colab notebook.
+
+    Returns
+    -------
+    bool
+        True if running inside a Colab notebook, False otherwise.
+    """
+    return "COLAB_RELEASE_TAG" in os.environ or "COLAB_BACKEND_VERSION" in os.environ
+
+
+def is_kaggle():
+    """Check if the current script is running inside a Kaggle kernel.
+
+    Returns
+    -------
+    bool
+        True if running inside a Kaggle kernel, False otherwise.
+    """
+    return (
+        os.environ.get("PWD") == "/kaggle/working"
+        and os.environ.get("KAGGLE_URL_BASE") == "https://www.kaggle.com"
+    )
+
+
+def is_jupyter():
+    """Check if the current script is running inside a Jupyter Notebook.
+
+    Returns
+    -------
+    bool
+        True if running inside a Jupyter Notebook, False otherwise.
+    """
+    with contextlib.suppress(Exception):
+        from IPython import get_ipython
+
+        return get_ipython() is not None
+    return False
+
+
+def is_docker() -> bool:
+    """Determine if the script is running inside a Docker container.
+
+    Returns
+    -------
+    bool
+        True if the script is running inside a Docker container, False otherwise.
+    """
+    file = Path("/proc/self/cgroup")
+    if file.exists():
+        with open(file) as f:
+            return "docker" in f.read()
+    else:
+        return False
+
+
+def get_git_dir():
+    """Determine whether the current file is part of a git repository and if so, returns the repository root directory.
+    If the current file is not part of a git repository, returns None.
+
+    Returns
+    -------
+    Path | None
+        Git root directory if found or None if not found.
+    """
+    for d in Path(__file__).parents:
+        if (d / ".git").is_dir():
+            return d
+
+
+def get_git_origin_url():
+    """Retrieve the origin URL of a git repository.
+
+    Returns
+    -------
+    str | None
+        The origin URL of the git repository or None if not git directory.
+    """
+    if is_git_dir():
+        with contextlib.suppress(subprocess.CalledProcessError):
+            origin = subprocess.check_output(
+                ["git", "config", "--get", "remote.origin.url"]
+            )
+            return origin.decode().strip()
+
+
+def yaml_save(file="data.yaml", data=None, header=""):
+    """Save YAML data to a file.
+
+    Parameters
+    ----------
+    file : str, optional
+        File name, by default 'data.yaml'.
+    data : dict, optional
+        Data to save in YAML format, by default None.
+    header : str, optional
+        YAML header to add, by default "".
+    """
+    if data is None:
+        data = {}
+    file = Path(file)
+    if not file.parent.exists():
+        # Create parent directories if they don't exist
+        file.parent.mkdir(parents=True, exist_ok=True)
+
+    # Convert Path objects to strings
+    for k, v in data.items():
+        if isinstance(v, Path):
+            data[k] = str(v)
+
+    # Dump data to file in YAML format
+    with open(file, "w", errors="ignore", encoding="utf-8") as f:
+        if header:
+            f.write(header)
+        yaml.safe_dump(data, f, sort_keys=False, allow_unicode=True)
+
+
+def yaml_load(file="data.yaml", append_filename=False):
+    """
+    Load YAML data from a file.
+
+    Parameters
+    ----------
+    file : str, optional
+        File name. Default is 'data.yaml'.
+    append_filename : bool, optional
+        Add the YAML filename to the YAML dictionary. Default is False.
+
+    Returns
+    -------
+    dict
+        YAML data and file name.
+    """
+    assert Path(file).suffix in (
+        ".yaml",
+        ".yml",
+    ), f"Attempting to load non-YAML file {file} with yaml_load()"
+    with open(file, errors="ignore", encoding="utf-8") as f:
+        s = f.read()  # string
+
+        # Add YAML filename to dict and return
+        data = (
+            yaml.safe_load(s) or {}
+        )  # always return a dict (yaml.safe_load() may return None for empty files)
+        if append_filename:
+            data["yaml_file"] = str(file)
+        return data
+
+
+def yaml_print(yaml_file: Union[str, Path, dict]) -> None:
+    """
+    Pretty prints a YAML file or a YAML-formatted dictionary.
+
+    Parameters
+    ----------
+    yaml_file : Union[str, Path, dict]
+        The file path of the YAML file or a YAML-formatted dictionary.
+
+    Returns
+    -------
+    None
+    """
+    yaml_dict = (
+        yaml_load(yaml_file) if isinstance(yaml_file, (str, Path)) else yaml_file
+    )
+    dump = yaml.dump(yaml_dict, sort_keys=False, allow_unicode=True)
+    LOGGER.info(f"Printing '{yaml_file}'\n\n{dump}")
+
+
+PLATFORMS = [platform.system()]
+if is_colab():
+    PLATFORMS.append("Colab")
+if is_kaggle():
+    PLATFORMS.append("Kaggle")
+if is_jupyter():
+    PLATFORMS.append("Jupyter")
+if is_docker():
+    PLATFORMS.append("Docker")
+
+PLATFORMS = "|".join(PLATFORMS)
+
+
+class TryExcept(contextlib.ContextDecorator):
+    """
+    TryExcept context manager.
+    Usage: @TryExcept() decorator or 'with TryExcept():' context manager.
+    """
+
+    def __init__(self, msg="", verbose=True):
+        """
+        Parameters
+        ----------
+        msg : str, optional
+            Custom message to display in case of exception, by default "".
+        verbose : bool, optional
+            Whether to display the message, by default True.
+        """
+        self.msg = msg
+        self.verbose = verbose
+
+    def __enter__(self):
+        pass
+
+    def __exit__(self, exc_type, value, traceback):
+        if self.verbose and value:
+            LOGGER.info(f"{self.msg}{': ' if self.msg else ''}{value}")
+        return True
+
+
+def threaded_request(
+    method, url, retry=3, timeout=30, thread=True, code=-1, verbose=True, **kwargs
+):
+    """
+    Makes an HTTP request using the 'requests' library, with exponential backoff retries up to a specified timeout.
+
+    Parameters
+    ----------
+    method : str
+        The HTTP method to use for the request. Choices are 'post' and 'get'.
+    url : str
+        The URL to make the request to.
+    retry : int, optional
+        Number of retries to attempt before giving up, by default 3.
+    timeout : int, optional
+        Timeout in seconds after which the function will give up retrying, by default 30.
+    thread : bool, optional
+        Whether to execute the request in a separate daemon thread, by default True.
+    code : int, optional
+        An identifier for the request, used for logging purposes, by default -1.
+    verbose : bool, optional
+        A flag to determine whether to print out to console or not, by default True.
+
+    Returns
+    -------
+    requests.Response
+        The HTTP response object. If the request is executed in a separate thread, returns the thread itself.
+    """
+    retry_codes = ()  # retry only these codes TODO: add codes if needed in future (500, 408)
+
+    @TryExcept(verbose=verbose)
+    def func(method, url, **kwargs):
+        """Make HTTP requests with retries and timeouts, with optional progress tracking."""
+        response = None
+        t0 = time.time()
+        for i in range(retry + 1):
+            if (time.time() - t0) > timeout:
+                break
+            response = requests.request(method, url, **kwargs)
+            if response.status_code < 300:  # good return codes in the 2xx range
+                break
+            try:
+                m = response.json().get("message", "No JSON message.")
+            except AttributeError:
+                m = "Unable to read JSON."
+            if i == 0:
+                if response.status_code in retry_codes:
+                    m += f" Retrying {retry}x for {timeout}s." if retry else ""
+                elif response.status_code == 429:  # rate limit
+                    m = f"Rate limit reached"
+                if verbose:
+                    LOGGER.warning(f"{response.status_code} #{code}")
+                if response.status_code not in retry_codes:
+                    return response
+            time.sleep(2**i)  # exponential standoff
+        return response
+
+    args = method, url
+    if thread:
+        return threading.Thread(
+            target=func, args=args, kwargs=kwargs, daemon=True
+        ).start()
+    else:
+        return func(*args, **kwargs)

python/lancedb/utils/sentry_log.py

@@ -0,0 +1,112 @@
+#  Copyright 2023 LanceDB Developers
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+
+import bdb
+import importlib.metadata
+import logging
+import sys
+from pathlib import Path
+
+from lancedb.utils import CONFIG
+
+from .general import (
+    PLATFORMS,
+    TryExcept,
+    is_git_dir,
+    is_github_actions_ci,
+    is_online,
+    is_pip_package,
+    is_pytest_running,
+)
+
+
+@TryExcept(verbose=False)
+def set_sentry():
+    """
+    Initialize the Sentry SDK for error tracking and reporting. Only used if sentry_sdk package is installed and
+    sync=True in settings. Run 'lancedb settings' to see and update settings YAML file.
+
+    Conditions required to send errors (ALL conditions must be met or no errors will be reported):
+        - sentry_sdk package is installed
+        - sync=True in  settings
+        - pytest is not running
+        - running in a pip package installation
+        - running in a non-git directory
+        - online environment
+
+    The function also configures Sentry SDK to ignore KeyboardInterrupt and FileNotFoundError
+    exceptions for now.
+
+    Additionally, the function sets custom tags and user information for Sentry events.
+    """
+
+    def before_send(event, hint):
+        """
+        Modify the event before sending it to Sentry based on specific exception types and messages.
+
+        Args:
+            event (dict): The event dictionary containing information about the error.
+            hint (dict): A dictionary containing additional information about the error.
+
+        Returns:
+            dict: The modified event or None if the event should not be sent to Sentry.
+        """
+        if "exc_info" in hint:
+            exc_type, exc_value, tb = hint["exc_info"]
+            if "out of memory" in str(exc_value).lower():
+                return None
+
+        if is_git_dir():
+            install = "git"
+        elif is_pip_package():
+            install = "pip"
+        else:
+            install = "other"
+
+        event["tags"] = {
+            "sys_argv": sys.argv[0],
+            "sys_argv_name": Path(sys.argv[0]).name,
+            "install": install,
+            "platforms": PLATFORMS,
+            "version": importlib.metadata.version("lancedb"),
+        }
+        return event
+
+    TESTS_RUNNING = is_pytest_running() or is_github_actions_ci()
+    ONLINE = is_online()
+    if CONFIG["diagnostics"] and not TESTS_RUNNING and ONLINE and is_pip_package():
+        # and not is_git_dir(): # not running inside a git dir. Maybe too restrictive?
+
+        # If sentry_sdk package is not installed then return and do not use Sentry
+        try:
+            import sentry_sdk  # noqa
+        except ImportError:
+            return
+
+        sentry_sdk.init(
+            dsn="https://c63ef8c64e05d1aa1a96513361f3ca2f@o4505950840946688.ingest.sentry.io/4505950933614592",
+            debug=False,
+            include_local_variables=False,
+            traces_sample_rate=1.0,
+            environment="production",  # 'dev' or 'production'
+            before_send=before_send,
+            ignore_errors=[KeyboardInterrupt, FileNotFoundError, bdb.BdbQuit],
+        )
+        sentry_sdk.set_user({"id": CONFIG["uuid"]})  # SHA-256 anonymized UUID hash
+
+        # Disable all sentry logging
+        for logger in "sentry_sdk", "sentry_sdk.errors":
+            logging.getLogger(logger).setLevel(logging.CRITICAL)
+
+
+set_sentry()

python/pyproject.toml

@@ -1,11 +1,23 @@
 [project]
 name = "lancedb"
-version = "0.1.16"
-dependencies = ["pylance==0.5.10", "ratelimiter", "retry", "tqdm", "aiohttp", "pydantic", "attr", "semver"]
-description = "lancedb"
-authors = [
-    { name = "LanceDB Devs", email = "dev@lancedb.com" },
+version = "0.3.1"
+dependencies = [
+    "deprecation",
+    "pylance==0.8.3",
+    "ratelimiter~=1.0",
+    "retry>=0.9.2",
+    "tqdm>=4.1.0",
+    "aiohttp",
+    "pydantic>=1.10",
+    "attrs>=21.3.0",
+    "semver>=3.0",
+    "cachetools",
+    "pyyaml>=6.0",
+    "click>=8.1.7",
+    "requests>=2.31.0"
 ]
+description = "lancedb"
+authors = [{ name = "LanceDB Devs", email = "dev@lancedb.com" }]
 license = { file = "LICENSE" }
 readme = "README.md"
 requires-python = ">=3.8"
@@ -36,22 +48,25 @@ classifiers = [
 repository = "https://github.com/lancedb/lancedb"
 
 [project.optional-dependencies]
-tests = [
-    "pandas>=1.4", "pytest", "pytest-mock", "pytest-asyncio"
-]
-dev = [
-    "ruff", "pre-commit", "black"
-]
-docs = [
-    "mkdocs", "mkdocs-jupyter", "mkdocs-material", "mkdocstrings[python]"
-]
+tests = ["pandas>=1.4", "pytest", "pytest-mock", "pytest-asyncio", "requests"]
+dev = ["ruff", "pre-commit", "black"]
+docs = ["mkdocs", "mkdocs-jupyter", "mkdocs-material", "mkdocstrings[python]"]
+clip = ["torch", "pillow", "open-clip"]
+embeddings = ["openai", "sentence-transformers", "torch", "pillow", "open-clip", "cohere"]
+
+[project.scripts]
+lancedb = "lancedb.cli.cli:cli"
 
 [build-system]
-requires = [
-  "setuptools",
-  "wheel",
-]
+requires = ["setuptools", "wheel"]
 build-backend = "setuptools.build_meta"
 
 [tool.isort]
 profile = "black"
+
+[tool.pytest.ini_options]
+addopts = "--strict-markers"
+markers = [
+    "slow: marks tests as slow (deselect with '-m \"not slow\"')",
+    "asyncio"
+]

python/tests/test_cli.py

@@ -0,0 +1,35 @@
+from click.testing import CliRunner
+
+from lancedb.cli.cli import cli
+from lancedb.utils import CONFIG
+
+
+def test_entry():
+    runner = CliRunner()
+    result = runner.invoke(cli)
+    assert result.exit_code == 0  # Main check
+    assert "lancedb" in result.output.lower()  # lazy check
+
+
+def test_diagnostics():
+    runner = CliRunner()
+    result = runner.invoke(cli, ["diagnostics", "--disabled"])
+    assert result.exit_code == 0  # Main check
+    assert CONFIG["diagnostics"] == False
+
+    result = runner.invoke(cli, ["diagnostics", "--enabled"])
+    assert result.exit_code == 0  # Main check
+    assert CONFIG["diagnostics"] == True
+
+
+def test_config():
+    runner = CliRunner()
+    result = runner.invoke(cli, ["config"])
+    assert result.exit_code == 0  # Main check
+    cfg = CONFIG.copy()
+    cfg.pop("uuid")
+    for (
+        item,
+        _,
+    ) in cfg.items():  # check for keys only as formatting is subject to change
+        assert item in result.output

python/tests/test_context.py

@@ -47,7 +47,7 @@ def test_contextualizer(raw_df: pd.DataFrame):
         .stride(3)
         .text_col("token")
         .groupby("document_id")
-        .to_df()["token"]
+        .to_pandas()["token"]
         .to_list()
     )
 
@@ -67,7 +67,7 @@ def test_contextualizer_with_threshold(raw_df: pd.DataFrame):
         .text_col("token")
         .groupby("document_id")
         .min_window_size(4)
-        .to_df()["token"]
+        .to_pandas()["token"]
         .to_list()
     )
 

python/tests/test_db.py

@@ -17,6 +17,7 @@ import pyarrow as pa
 import pytest
 
 import lancedb
+from lancedb.pydantic import LanceModel, Vector
 
 
 def test_basic(tmp_path):
@@ -32,11 +33,11 @@ def test_basic(tmp_path):
             {"vector": [5.9, 26.5], "item": "bar", "price": 20.0},
         ],
     )
-    rs = table.search([100, 100]).limit(1).to_df()
+    rs = table.search([100, 100]).limit(1).to_pandas()
     assert len(rs) == 1
     assert rs["item"].iloc[0] == "bar"
 
-    rs = table.search([100, 100]).where("price < 15").limit(2).to_df()
+    rs = table.search([100, 100]).where("price < 15").limit(2).to_pandas()
     assert len(rs) == 1
     assert rs["item"].iloc[0] == "foo"
 
@@ -61,11 +62,11 @@ def test_ingest_pd(tmp_path):
         }
     )
     table = db.create_table("test", data=data)
-    rs = table.search([100, 100]).limit(1).to_df()
+    rs = table.search([100, 100]).limit(1).to_pandas()
     assert len(rs) == 1
     assert rs["item"].iloc[0] == "bar"
 
-    rs = table.search([100, 100]).where("price < 15").limit(2).to_df()
+    rs = table.search([100, 100]).where("price < 15").limit(2).to_pandas()
     assert len(rs) == 1
     assert rs["item"].iloc[0] == "foo"
 
@@ -76,30 +77,77 @@ def test_ingest_pd(tmp_path):
     assert db.open_table("test").name == db["test"].name
 
 
-def test_ingest_record_batch_iterator(tmp_path):
-    def batch_reader():
-        for i in range(5):
-            yield pa.RecordBatch.from_arrays(
+def test_ingest_iterator(tmp_path):
+    class PydanticSchema(LanceModel):
+        vector: Vector(2)
+        item: str
+        price: float
+
+    arrow_schema = pa.schema(
         [
-                    pa.array([[3.1, 4.1], [5.9, 26.5]]),
+            pa.field("vector", pa.list_(pa.float32(), 2)),
+            pa.field("item", pa.utf8()),
+            pa.field("price", pa.float32()),
+        ]
+    )
+
+    def make_batches():
+        for _ in range(5):
+            yield from [
+                # pandas
+                pd.DataFrame(
+                    {
+                        "vector": [[3.1, 4.1], [1, 1]],
+                        "item": ["foo", "bar"],
+                        "price": [10.0, 20.0],
+                    }
+                ),
+                # pylist
+                [
+                    {"vector": [3.1, 4.1], "item": "foo", "price": 10.0},
+                    {"vector": [5.9, 26.5], "item": "bar", "price": 20.0},
+                ],
+                # recordbatch
+                pa.RecordBatch.from_arrays(
+                    [
+                        pa.array([[3.1, 4.1], [5.9, 26.5]], pa.list_(pa.float32(), 2)),
                         pa.array(["foo", "bar"]),
                         pa.array([10.0, 20.0]),
                     ],
                     ["vector", "item", "price"],
-            )
-
-    db = lancedb.connect(tmp_path)
-    tbl = db.create_table(
-        "test",
-        batch_reader(),
-        schema=pa.schema(
-            [
-                pa.field("vector", pa.list_(pa.float32())),
-                pa.field("item", pa.utf8()),
-                pa.field("price", pa.float32()),
-            ]
                 ),
-    )
+                # pa Table
+                pa.Table.from_arrays(
+                    [
+                        pa.array([[3.1, 4.1], [5.9, 26.5]], pa.list_(pa.float32(), 2)),
+                        pa.array(["foo", "bar"]),
+                        pa.array([10.0, 20.0]),
+                    ],
+                    ["vector", "item", "price"],
+                ),
+                # pydantic list
+                [
+                    PydanticSchema(vector=[3.1, 4.1], item="foo", price=10.0),
+                    PydanticSchema(vector=[5.9, 26.5], item="bar", price=20.0),
+                ]
+                # TODO: test pydict separately. it is unique column number and names contraint
+            ]
+
+    def run_tests(schema):
+        db = lancedb.connect(tmp_path)
+        tbl = db.create_table("table2", make_batches(), schema=schema, mode="overwrite")
+        tbl.to_pandas()
+        assert tbl.search([3.1, 4.1]).limit(1).to_pandas()["_distance"][0] == 0.0
+        assert tbl.search([5.9, 26.5]).limit(1).to_pandas()["_distance"][0] == 0.0
+        tbl_len = len(tbl)
+        tbl.add(make_batches())
+        assert tbl_len == 50
+        assert len(tbl) == tbl_len * 2
+        assert len(tbl.list_versions()) == 3
+        db.drop_database()
+
+    run_tests(arrow_schema)
+    run_tests(PydanticSchema)
 
 
 def test_create_mode(tmp_path):
@@ -154,6 +202,47 @@ def test_delete_table(tmp_path):
     db.drop_table("does_not_exist", ignore_missing=True)
 
 
+def test_drop_database(tmp_path):
+    db = lancedb.connect(tmp_path)
+    data = pd.DataFrame(
+        {
+            "vector": [[3.1, 4.1], [5.9, 26.5]],
+            "item": ["foo", "bar"],
+            "price": [10.0, 20.0],
+        }
+    )
+    new_data = pd.DataFrame(
+        {
+            "vector": [[5.1, 4.1], [5.9, 10.5]],
+            "item": ["kiwi", "avocado"],
+            "price": [12.0, 17.0],
+        }
+    )
+    db.create_table("test", data=data)
+    with pytest.raises(Exception):
+        db.create_table("test", data=data)
+
+    assert db.table_names() == ["test"]
+
+    db.create_table("new_test", data=new_data)
+    db.drop_database()
+    assert db.table_names() == []
+
+    # it should pass when no tables are present
+    db.create_table("test", data=new_data)
+    db.drop_table("test")
+    assert db.table_names() == []
+    db.drop_database()
+    assert db.table_names() == []
+
+    # creating an empty database with schema
+    schema = pa.schema([pa.field("vector", pa.list_(pa.float32(), list_size=2))])
+    db.create_table("empty_table", schema=schema)
+    # dropping a empty database should pass
+    db.drop_database()
+    assert db.table_names() == []
+
+
 def test_empty_or_nonexistent_table(tmp_path):
     db = lancedb.connect(tmp_path)
     with pytest.raises(Exception):
@@ -162,8 +251,14 @@ def test_empty_or_nonexistent_table(tmp_path):
     with pytest.raises(Exception):
         db.open_table("does_not_exist")
 
-    schema = pa.schema([pa.field("a", pa.int32())])
-    db.create_table("test", schema=schema)
+    schema = pa.schema([pa.field("a", pa.int64(), nullable=False)])
+    test = db.create_table("test", schema=schema)
+
+    class TestModel(LanceModel):
+        a: int
+
+    test2 = db.create_table("test2", schema=TestModel)
+    assert test.schema == test2.schema
 
 
 def test_replace_index(tmp_path):

python/tests/test_e2e_remote_db.py

@@ -23,5 +23,5 @@ from lancedb import LanceDBConnection
 def test_against_local_server():
     conn = LanceDBConnection("lancedb+http://localhost:10024")
     table = conn.open_table("sift1m_ivf1024_pq16")
-    df = table.search(np.random.rand(128)).to_df()
+    df = table.search(np.random.rand(128)).to_pandas()
     assert len(df) == 10

python/tests/test_embeddings.py

@@ -12,10 +12,16 @@
 #  limitations under the License.
 import sys
 
+import lance
 import numpy as np
 import pyarrow as pa
 
-from lancedb.embeddings import with_embeddings
+from lancedb.conftest import MockTextEmbeddingFunction
+from lancedb.embeddings import (
+    EmbeddingFunctionConfig,
+    EmbeddingFunctionRegistry,
+    with_embeddings,
+)
 
 
 def mock_embed_func(input_data):
@@ -40,3 +46,40 @@ def test_with_embeddings():
         assert data.column_names == ["text", "price", "vector"]
         assert data.column("text").to_pylist() == ["foo", "bar"]
         assert data.column("price").to_pylist() == [10.0, 20.0]
+
+
+def test_embedding_function(tmp_path):
+    registry = EmbeddingFunctionRegistry.get_instance()
+
+    # let's create a table
+    table = pa.table(
+        {
+            "text": pa.array(["hello world", "goodbye world"]),
+            "vector": [np.random.randn(10), np.random.randn(10)],
+        }
+    )
+    conf = EmbeddingFunctionConfig(
+        source_column="text",
+        vector_column="vector",
+        function=MockTextEmbeddingFunction(),
+    )
+    metadata = registry.get_table_metadata([conf])
+    table = table.replace_schema_metadata(metadata)
+
+    # Write it to disk
+    lance.write_dataset(table, tmp_path / "test.lance")
+
+    # Load this back
+    ds = lance.dataset(tmp_path / "test.lance")
+
+    # can we get the serialized version back out?
+    configs = registry.parse_functions(ds.schema.metadata)
+
+    conf = configs["vector"]
+    func = conf.function
+    actual = func.compute_query_embeddings("hello world")
+
+    # And we make sure we can call it
+    expected = func.compute_query_embeddings("hello world")
+
+    assert np.allclose(actual, expected)

python/tests/test_embeddings_slow.py

@@ -0,0 +1,149 @@
+#  Copyright (c) 2023. LanceDB Developers
+#
+#  Licensed under the Apache License, Version 2.0 (the "License");
+#  you may not use this file except in compliance with the License.
+#  You may obtain a copy of the License at
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+#  Unless required by applicable law or agreed to in writing, software
+#  distributed under the License is distributed on an "AS IS" BASIS,
+#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#  See the License for the specific language governing permissions and
+#  limitations under the License.
+import io
+import os
+
+import numpy as np
+import pandas as pd
+import pytest
+import requests
+
+import lancedb
+from lancedb.embeddings import EmbeddingFunctionRegistry
+from lancedb.pydantic import LanceModel, Vector
+
+# These are integration tests for embedding functions.
+# They are slow because they require downloading models
+# or connection to external api
+
+
+@pytest.mark.slow
+@pytest.mark.parametrize("alias", ["sentence-transformers", "openai"])
+def test_sentence_transformer(alias, tmp_path):
+    db = lancedb.connect(tmp_path)
+    registry = EmbeddingFunctionRegistry.get_instance()
+    func = registry.get(alias).create()
+
+    class Words(LanceModel):
+        text: str = func.SourceField()
+        vector: Vector(func.ndims()) = func.VectorField()
+
+    table = db.create_table("words", schema=Words)
+    table.add(
+        pd.DataFrame(
+            {
+                "text": [
+                    "hello world",
+                    "goodbye world",
+                    "fizz",
+                    "buzz",
+                    "foo",
+                    "bar",
+                    "baz",
+                ]
+            }
+        )
+    )
+
+    query = "greetings"
+    actual = table.search(query).limit(1).to_pydantic(Words)[0]
+
+    vec = func.compute_query_embeddings(query)[0]
+    expected = table.search(vec).limit(1).to_pydantic(Words)[0]
+    assert actual.text == expected.text
+    assert actual.text == "hello world"
+
+
+@pytest.mark.slow
+def test_openclip(tmp_path):
+    from PIL import Image
+
+    db = lancedb.connect(tmp_path)
+    registry = EmbeddingFunctionRegistry.get_instance()
+    func = registry.get("open-clip").create()
+
+    class Images(LanceModel):
+        label: str
+        image_uri: str = func.SourceField()
+        image_bytes: bytes = func.SourceField()
+        vector: Vector(func.ndims()) = func.VectorField()
+        vec_from_bytes: Vector(func.ndims()) = func.VectorField()
+
+    table = db.create_table("images", schema=Images)
+    labels = ["cat", "cat", "dog", "dog", "horse", "horse"]
+    uris = [
+        "http://farm1.staticflickr.com/53/167798175_7c7845bbbd_z.jpg",
+        "http://farm1.staticflickr.com/134/332220238_da527d8140_z.jpg",
+        "http://farm9.staticflickr.com/8387/8602747737_2e5c2a45d4_z.jpg",
+        "http://farm5.staticflickr.com/4092/5017326486_1f46057f5f_z.jpg",
+        "http://farm9.staticflickr.com/8216/8434969557_d37882c42d_z.jpg",
+        "http://farm6.staticflickr.com/5142/5835678453_4f3a4edb45_z.jpg",
+    ]
+    # get each uri as bytes
+    image_bytes = [requests.get(uri).content for uri in uris]
+    table.add(
+        pd.DataFrame({"label": labels, "image_uri": uris, "image_bytes": image_bytes})
+    )
+
+    # text search
+    actual = table.search("man's best friend").limit(1).to_pydantic(Images)[0]
+    assert actual.label == "dog"
+    frombytes = (
+        table.search("man's best friend", vector_column_name="vec_from_bytes")
+        .limit(1)
+        .to_pydantic(Images)[0]
+    )
+    assert actual.label == frombytes.label
+    assert np.allclose(actual.vector, frombytes.vector)
+
+    # image search
+    query_image_uri = "http://farm1.staticflickr.com/200/467715466_ed4a31801f_z.jpg"
+    image_bytes = requests.get(query_image_uri).content
+    query_image = Image.open(io.BytesIO(image_bytes))
+    actual = table.search(query_image).limit(1).to_pydantic(Images)[0]
+    assert actual.label == "dog"
+    other = (
+        table.search(query_image, vector_column_name="vec_from_bytes")
+        .limit(1)
+        .to_pydantic(Images)[0]
+    )
+    assert actual.label == other.label
+
+    arrow_table = table.search().select(["vector", "vec_from_bytes"]).to_arrow()
+    assert np.allclose(
+        arrow_table["vector"].combine_chunks().values.to_numpy(),
+        arrow_table["vec_from_bytes"].combine_chunks().values.to_numpy(),
+    )
+
+
+@pytest.mark.slow
+@pytest.mark.skipif(
+    os.environ.get("COHERE_API_KEY") is None, reason="COHERE_API_KEY not set"
+)  # also skip if cohere not installed
+def test_cohere_embedding_function():
+    cohere = (
+        EmbeddingFunctionRegistry.get_instance()
+        .get("cohere")
+        .create(name="embed-multilingual-v2.0")
+    )
+
+    class TextModel(LanceModel):
+        text: str = cohere.SourceField()
+        vector: Vector(cohere.ndims()) = cohere.VectorField()
+
+    df = pd.DataFrame({"text": ["hello world", "goodbye world"]})
+    db = lancedb.connect("~/lancedb")
+    tbl = db.create_table("test", schema=TextModel, mode="overwrite")
+
+    tbl.add(df)
+    assert len(tbl.to_pandas()["vector"][0]) == cohere.ndims()

python/tests/test_fts.py

@@ -66,19 +66,19 @@ def test_search_index(tmp_path, table):
     results = ldb.fts.search_index(index, query="puppy", limit=10)
     assert len(results) == 2
     assert len(results[0]) == 10  # row_ids
-    assert len(results[1]) == 10  # scores
+    assert len(results[1]) == 10  # _distance
 
 
 def test_create_index_from_table(tmp_path, table):
     table.create_fts_index("text")
-    df = table.search("puppy").limit(10).select(["text"]).to_df()
+    df = table.search("puppy").limit(10).select(["text"]).to_pandas()
     assert len(df) == 10
     assert "text" in df.columns
 
 
 def test_create_index_multiple_columns(tmp_path, table):
     table.create_fts_index(["text", "text2"])
-    df = table.search("puppy").limit(10).to_df()
+    df = table.search("puppy").limit(10).to_pandas()
     assert len(df) == 10
     assert "text" in df.columns
     assert "text2" in df.columns
@@ -87,5 +87,5 @@ def test_create_index_multiple_columns(tmp_path, table):
 def test_empty_rs(tmp_path, table, mocker):
     table.create_fts_index(["text", "text2"])
     mocker.patch("lancedb.fts.search_index", return_value=([], []))
-    df = table.search("puppy").limit(10).to_df()
+    df = table.search("puppy").limit(10).to_pandas()
     assert len(df) == 0

python/tests/test_io.py

@@ -36,11 +36,11 @@ def test_s3_io():
             {"vector": [5.9, 26.5], "item": "bar", "price": 20.0},
         ],
     )
-    rs = table.search([100, 100]).limit(1).to_df()
+    rs = table.search([100, 100]).limit(1).to_pandas()
     assert len(rs) == 1
     assert rs["item"].iloc[0] == "bar"
 
-    rs = table.search([100, 100]).where("price < 15").limit(2).to_df()
+    rs = table.search([100, 100]).where("price < 15").limit(2).to_pandas()
     assert len(rs) == 1
     assert rs["item"].iloc[0] == "foo"
 

python/tests/test_pydantic.py

@@ -19,8 +19,9 @@ from typing import List, Optional
 import pyarrow as pa
 import pydantic
 import pytest
+from pydantic import Field
 
-from lancedb.pydantic import PYDANTIC_VERSION, LanceModel, pydantic_to_schema, vector
+from lancedb.pydantic import PYDANTIC_VERSION, LanceModel, Vector, pydantic_to_schema
 
 
 @pytest.mark.skipif(
@@ -107,7 +108,7 @@ def test_pydantic_to_arrow_py38():
 
 def test_fixed_size_list_field():
     class TestModel(pydantic.BaseModel):
-        vec: vector(16)
+        vec: Vector(16)
         li: List[int]
 
     data = TestModel(vec=list(range(16)), li=[1, 2, 3])
@@ -154,7 +155,7 @@ def test_fixed_size_list_field():
 
 def test_fixed_size_list_validation():
     class TestModel(pydantic.BaseModel):
-        vec: vector(8)
+        vec: Vector(8)
 
     with pytest.raises(pydantic.ValidationError):
         TestModel(vec=range(9))
@@ -167,9 +168,12 @@ def test_fixed_size_list_validation():
 
 def test_lance_model():
     class TestModel(LanceModel):
-        vec: vector(16)
-        li: List[int]
+        vector: Vector(16) = Field(default=[0.0] * 16)
+        li: List[int] = Field(default=[1, 2, 3])
 
     schema = pydantic_to_schema(TestModel)
     assert schema == TestModel.to_arrow_schema()
-    assert TestModel.field_names() == ["vec", "li"]
+    assert TestModel.field_names() == ["vector", "li"]
+
+    t = TestModel()
+    assert t == TestModel(vec=[0.0] * 16, li=[1, 2, 3])

python/tests/test_query.py

@@ -20,8 +20,8 @@ import pyarrow as pa
 import pytest
 
 from lancedb.db import LanceDBConnection
-from lancedb.pydantic import LanceModel, vector
-from lancedb.query import LanceQueryBuilder, Query
+from lancedb.pydantic import LanceModel, Vector
+from lancedb.query import LanceVectorQueryBuilder, Query
 from lancedb.table import LanceTable
 
 
@@ -38,6 +38,7 @@ class MockTable:
         return ds.to_table(
             columns=query.columns,
             filter=query.filter,
+            prefilter=query.prefilter,
             nearest={
                 "column": query.vector_column,
                 "q": query.vector,
@@ -67,12 +68,12 @@ def table(tmp_path) -> MockTable:
 
 def test_cast(table):
     class TestModel(LanceModel):
-        vector: vector(2)
+        vector: Vector(2)
         id: int
         str_field: str
         float_field: float
 
-    q = LanceQueryBuilder(table, [0, 0], "vector").limit(1)
+    q = LanceVectorQueryBuilder(table, [0, 0], "vector").limit(1)
     results = q.to_pydantic(TestModel)
     assert len(results) == 1
     r0 = results[0]
@@ -84,13 +85,37 @@ def test_cast(table):
 
 
 def test_query_builder(table):
-    df = LanceQueryBuilder(table, [0, 0], "vector").limit(1).select(["id"]).to_df()
-    assert df["id"].values[0] == 1
-    assert all(df["vector"].values[0] == [1, 2])
+    rs = (
+        LanceVectorQueryBuilder(table, [0, 0], "vector")
+        .limit(1)
+        .select(["id"])
+        .to_list()
+    )
+    assert rs[0]["id"] == 1
+    assert all(np.array(rs[0]["vector"]) == [1, 2])
 
 
 def test_query_builder_with_filter(table):
-    df = LanceQueryBuilder(table, [0, 0], "vector").where("id = 2").to_df()
+    rs = LanceVectorQueryBuilder(table, [0, 0], "vector").where("id = 2").to_list()
+    assert rs[0]["id"] == 2
+    assert all(np.array(rs[0]["vector"]) == [3, 4])
+
+
+def test_query_builder_with_prefilter(table):
+    df = (
+        LanceVectorQueryBuilder(table, [0, 0], "vector")
+        .where("id = 2")
+        .limit(1)
+        .to_pandas()
+    )
+    assert len(df) == 0
+
+    df = (
+        LanceVectorQueryBuilder(table, [0, 0], "vector")
+        .where("id = 2", prefilter=True)
+        .limit(1)
+        .to_pandas()
+    )
     assert df["id"].values[0] == 2
     assert all(df["vector"].values[0] == [3, 4])
 
@@ -98,21 +123,25 @@ def test_query_builder_with_filter(table):
 def test_query_builder_with_metric(table):
     query = [4, 8]
     vector_column_name = "vector"
-    df_default = LanceQueryBuilder(table, query, vector_column_name).to_df()
-    df_l2 = LanceQueryBuilder(table, query, vector_column_name).metric("L2").to_df()
+    df_default = LanceVectorQueryBuilder(table, query, vector_column_name).to_pandas()
+    df_l2 = (
+        LanceVectorQueryBuilder(table, query, vector_column_name)
+        .metric("L2")
+        .to_pandas()
+    )
     tm.assert_frame_equal(df_default, df_l2)
 
     df_cosine = (
-        LanceQueryBuilder(table, query, vector_column_name)
+        LanceVectorQueryBuilder(table, query, vector_column_name)
         .metric("cosine")
         .limit(1)
-        .to_df()
+        .to_pandas()
     )
-    assert df_cosine.score[0] == pytest.approx(
+    assert df_cosine._distance[0] == pytest.approx(
         cosine_distance(query, df_cosine.vector[0]),
         abs=1e-6,
     )
-    assert 0 <= df_cosine.score[0] <= 1
+    assert 0 <= df_cosine._distance[0] <= 1
 
 
 def test_query_builder_with_different_vector_column():
@@ -120,7 +149,7 @@ def test_query_builder_with_different_vector_column():
     query = [4, 8]
     vector_column_name = "foo_vector"
     builder = (
-        LanceQueryBuilder(table, query, vector_column_name)
+        LanceVectorQueryBuilder(table, query, vector_column_name)
         .metric("cosine")
         .where("b < 10")
         .select(["b"])

python/tests/test_remote_client.py

@@ -11,7 +11,7 @@
 #  See the License for the specific language governing permissions and
 #  limitations under the License.
 
-import attr
+import attrs
 import numpy as np
 import pandas as pd
 import pyarrow as pa
@@ -21,10 +21,10 @@ from aiohttp import web
 from lancedb.remote.client import RestfulLanceDBClient, VectorQuery
 
 
-@attr.define
+@attrs.define
 class MockLanceDBServer:
-    runner: web.AppRunner = attr.field(init=False)
-    site: web.TCPSite = attr.field(init=False)
+    runner: web.AppRunner = attrs.field(init=False)
+    site: web.TCPSite = attrs.field(init=False)
 
     async def query_handler(self, request: web.Request) -> web.Response:
         table_name = request.match_info["table_name"]
@@ -86,7 +86,7 @@ async def test_e2e_with_mock_server():
                     columns=["id", "vector"],
                 ),
             )
-        ).to_df()
+        ).to_pandas()
 
         assert "vector" in df.columns
         assert "id" in df.columns

python/tests/test_remote_db.py

@@ -32,4 +32,4 @@ def test_remote_db():
     setattr(conn, "_client", FakeLanceDBClient())
 
     table = conn["test"]
-    table.search([1.0, 2.0]).to_df()
+    table.search([1.0, 2.0]).to_pandas()

python/tests/test_table.py

@@ -12,17 +12,21 @@
 #  limitations under the License.
 
 import functools
+from datetime import timedelta
 from pathlib import Path
 from typing import List
 from unittest.mock import PropertyMock, patch
 
+import lance
 import numpy as np
 import pandas as pd
 import pyarrow as pa
 import pytest
 
+from lancedb.conftest import MockTextEmbeddingFunction
 from lancedb.db import LanceDBConnection
-from lancedb.pydantic import LanceModel, vector
+from lancedb.embeddings import EmbeddingFunctionConfig, EmbeddingFunctionRegistry
+from lancedb.pydantic import LanceModel, Vector
 from lancedb.table import LanceTable
 
 
@@ -138,7 +142,7 @@ def test_add(db):
 
 def test_add_pydantic_model(db):
     class TestModel(LanceModel):
-        vector: vector(16)
+        vector: Vector(16)
         li: List[int]
 
     data = TestModel(vector=list(range(16)), li=[1, 2, 3])
@@ -177,16 +181,16 @@ def test_versioning(db):
         ],
     )
 
-    assert len(table.list_versions()) == 1
-    assert table.version == 1
-
-    table.add([{"vector": [6.3, 100.5], "item": "new", "price": 30.0}])
     assert len(table.list_versions()) == 2
     assert table.version == 2
+
+    table.add([{"vector": [6.3, 100.5], "item": "new", "price": 30.0}])
+    assert len(table.list_versions()) == 3
+    assert table.version == 3
     assert len(table) == 3
 
-    table.checkout(1)
-    assert table.version == 1
+    table.checkout(2)
+    assert table.version == 2
     assert len(table) == 2
 
 
@@ -220,6 +224,7 @@ def test_create_index_method():
                 num_partitions=256,
                 num_sub_vectors=96,
                 replace=True,
+                accelerator=None,
             )
 
 
@@ -268,3 +273,201 @@ def test_add_with_nans(db):
     arrow_tbl = table.to_lance().to_table(filter="item == 'bar'")
     v = arrow_tbl["vector"].to_pylist()[0]
     assert np.allclose(v, np.array([0.0, 0.0]))
+
+
+def test_restore(db):
+    table = LanceTable.create(
+        db,
+        "my_table",
+        data=[{"vector": [1.1, 0.9], "type": "vector"}],
+    )
+    table.add([{"vector": [0.5, 0.2], "type": "vector"}])
+    table.restore(2)
+    assert len(table.list_versions()) == 4
+    assert len(table) == 1
+
+    expected = table.to_arrow()
+    table.checkout(2)
+    table.restore()
+    assert len(table.list_versions()) == 5
+    assert table.to_arrow() == expected
+
+    table.restore(5)  # latest version should be no-op
+    assert len(table.list_versions()) == 5
+
+    with pytest.raises(ValueError):
+        table.restore(6)
+
+    with pytest.raises(ValueError):
+        table.restore(0)
+
+
+def test_merge(db, tmp_path):
+    table = LanceTable.create(
+        db,
+        "my_table",
+        data=[{"vector": [1.1, 0.9], "id": 0}, {"vector": [1.2, 1.9], "id": 1}],
+    )
+    other_table = pa.table({"document": ["foo", "bar"], "id": [0, 1]})
+    table.merge(other_table, left_on="id")
+    assert len(table.list_versions()) == 3
+    expected = pa.table(
+        {"vector": [[1.1, 0.9], [1.2, 1.9]], "id": [0, 1], "document": ["foo", "bar"]},
+        schema=table.schema,
+    )
+    assert table.to_arrow() == expected
+
+    other_dataset = lance.write_dataset(other_table, tmp_path / "other_table.lance")
+    table.restore(1)
+    table.merge(other_dataset, left_on="id")
+
+
+def test_delete(db):
+    table = LanceTable.create(
+        db,
+        "my_table",
+        data=[{"vector": [1.1, 0.9], "id": 0}, {"vector": [1.2, 1.9], "id": 1}],
+    )
+    assert len(table) == 2
+    assert len(table.list_versions()) == 2
+    table.delete("id=0")
+    assert len(table.list_versions()) == 3
+    assert table.version == 3
+    assert len(table) == 1
+    assert table.to_pandas()["id"].tolist() == [1]
+
+
+def test_update(db):
+    table = LanceTable.create(
+        db,
+        "my_table",
+        data=[{"vector": [1.1, 0.9], "id": 0}, {"vector": [1.2, 1.9], "id": 1}],
+    )
+    assert len(table) == 2
+    assert len(table.list_versions()) == 2
+    table.update(where="id=0", values={"vector": [1.1, 1.1]})
+    assert len(table.list_versions()) == 4
+    assert table.version == 4
+    assert len(table) == 2
+    v = table.to_arrow()["vector"].combine_chunks()
+    v = v.values.to_numpy().reshape(2, 2)
+    assert np.allclose(v, np.array([[1.2, 1.9], [1.1, 1.1]]))
+
+
+def test_create_with_embedding_function(db):
+    class MyTable(LanceModel):
+        text: str
+        vector: Vector(10)
+
+    func = MockTextEmbeddingFunction()
+    texts = ["hello world", "goodbye world", "foo bar baz fizz buzz"]
+    df = pd.DataFrame({"text": texts, "vector": func.compute_source_embeddings(texts)})
+
+    conf = EmbeddingFunctionConfig(
+        source_column="text", vector_column="vector", function=func
+    )
+    table = LanceTable.create(
+        db,
+        "my_table",
+        schema=MyTable,
+        embedding_functions=[conf],
+    )
+    table.add(df)
+
+    query_str = "hi how are you?"
+    query_vector = func.compute_query_embeddings(query_str)[0]
+    expected = table.search(query_vector).limit(2).to_arrow()
+
+    actual = table.search(query_str).limit(2).to_arrow()
+    assert actual == expected
+
+
+def test_add_with_embedding_function(db):
+    emb = EmbeddingFunctionRegistry.get_instance().get("test")()
+
+    class MyTable(LanceModel):
+        text: str = emb.SourceField()
+        vector: Vector(emb.ndims()) = emb.VectorField()
+
+    table = LanceTable.create(db, "my_table", schema=MyTable)
+
+    texts = ["hello world", "goodbye world", "foo bar baz fizz buzz"]
+    df = pd.DataFrame({"text": texts})
+    table.add(df)
+
+    texts = ["the quick brown fox", "jumped over the lazy dog"]
+    table.add([{"text": t} for t in texts])
+
+    query_str = "hi how are you?"
+    query_vector = emb.compute_query_embeddings(query_str)[0]
+    expected = table.search(query_vector).limit(2).to_arrow()
+
+    actual = table.search(query_str).limit(2).to_arrow()
+    assert actual == expected
+
+
+def test_multiple_vector_columns(db):
+    class MyTable(LanceModel):
+        text: str
+        vector1: Vector(10)
+        vector2: Vector(10)
+
+    table = LanceTable.create(
+        db,
+        "my_table",
+        schema=MyTable,
+    )
+
+    v1 = np.random.randn(10)
+    v2 = np.random.randn(10)
+    data = [
+        {"vector1": v1, "vector2": v2, "text": "foo"},
+        {"vector1": v2, "vector2": v1, "text": "bar"},
+    ]
+    df = pd.DataFrame(data)
+    table.add(df)
+
+    q = np.random.randn(10)
+    result1 = table.search(q, vector_column_name="vector1").limit(1).to_pandas()
+    result2 = table.search(q, vector_column_name="vector2").limit(1).to_pandas()
+
+    assert result1["text"].iloc[0] != result2["text"].iloc[0]
+
+
+def test_empty_query(db):
+    table = LanceTable.create(
+        db,
+        "my_table",
+        data=[{"text": "foo", "id": 0}, {"text": "bar", "id": 1}],
+    )
+    df = table.search().select(["id"]).where("text='bar'").limit(1).to_pandas()
+    val = df.id.iloc[0]
+    assert val == 1
+
+
+def test_compact_cleanup(db):
+    table = LanceTable.create(
+        db,
+        "my_table",
+        data=[{"text": "foo", "id": 0}, {"text": "bar", "id": 1}],
+    )
+
+    table.add([{"text": "baz", "id": 2}])
+    assert len(table) == 3
+    assert table.version == 3
+
+    stats = table.compact_files()
+    assert len(table) == 3
+    assert table.version == 4
+    assert stats.fragments_removed > 0
+    assert stats.fragments_added == 1
+
+    stats = table.cleanup_old_versions()
+    assert stats.bytes_removed == 0
+
+    stats = table.cleanup_old_versions(older_than=timedelta(0), delete_unverified=True)
+    assert stats.bytes_removed > 0
+    assert table.version == 4
+
+    with pytest.raises(Exception, match="Version 3 no longer exists"):
+        table.checkout(3)

python/tests/test_telemetry.py

@@ -0,0 +1,60 @@
+import json
+
+import pytest
+
+import lancedb
+from lancedb.utils.events import _Events
+
+
+@pytest.fixture(autouse=True)
+def request_log_path(tmp_path):
+    return tmp_path / "request.json"
+
+
+def mock_register_event(name: str, **kwargs):
+    if _Events._instance is None:
+        _Events._instance = _Events()
+
+    _Events._instance.enabled = True
+    _Events._instance.rate_limit = 0
+    _Events._instance(name, **kwargs)
+
+
+def test_event_reporting(monkeypatch, request_log_path, tmp_path) -> None:
+    def mock_request(**kwargs):
+        json_data = kwargs.get("json", {})
+        with open(request_log_path, "w") as f:
+            json.dump(json_data, f)
+
+    monkeypatch.setattr(
+        lancedb.table, "register_event", mock_register_event
+    )  # Force enable registering events and strip exception handling
+    monkeypatch.setattr(lancedb.utils.events, "threaded_request", mock_request)
+
+    db = lancedb.connect(tmp_path)
+    db.create_table(
+        "test",
+        data=[
+            {"vector": [3.1, 4.1], "item": "foo", "price": 10.0},
+            {"vector": [5.9, 26.5], "item": "bar", "price": 20.0},
+        ],
+        mode="overwrite",
+    )
+
+    assert request_log_path.exists()  # test if event was registered
+
+    with open(request_log_path, "r") as f:
+        json_data = json.load(f)
+
+    # TODO: don't hardcode these here. Instead create a module level json scehma in lancedb.utils.events for better evolvability
+    batch_keys = ["api_key", "distinct_id", "batch"]
+    event_keys = ["event", "properties", "timestamp", "distinct_id"]
+    property_keys = ["cli", "install", "platforms", "version", "session_id"]
+
+    assert all([key in json_data for key in batch_keys])
+    assert all([key in json_data["batch"][0] for key in event_keys])
+    assert all([key in json_data["batch"][0]["properties"] for key in property_keys])
+
+    # cleanup & reset
+    monkeypatch.undo()
+    _Events._instance = None

rust/ffi/node/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "vectordb-node"
-version = "0.1.19"
+version = "0.3.0"
 description = "Serverless, low-latency vector database for AI applications"
 license = "Apache-2.0"
 edition = "2018"
@@ -13,11 +13,13 @@ crate-type = ["cdylib"]
 arrow-array = { workspace = true }
 arrow-ipc = { workspace = true }
 arrow-schema = { workspace = true }
+chrono = { workspace = true }
 conv = "0.3.3"
 once_cell = "1"
 futures = "0.3"
 half = { workspace = true }
 lance = { workspace = true }
+lance-linalg = { workspace = true }
 vectordb = { path = "../../vectordb" }
 tokio = { version = "1.23", features = ["rt-multi-thread"] }
 neon = {version = "0.10.1", default-features = false, features = ["channel-api", "napi-6", "promise-api", "task-api"] }

rust/ffi/node/src/arrow.rs

@@ -14,60 +14,35 @@
 
 use std::io::Cursor;
 use std::ops::Deref;
-use std::sync::Arc;
 
-use arrow_array::cast::as_list_array;
-use arrow_array::{Array, ArrayRef, FixedSizeListArray, RecordBatch};
+use arrow_array::RecordBatch;
 use arrow_ipc::reader::FileReader;
 use arrow_ipc::writer::FileWriter;
-use arrow_schema::{DataType, Field, Schema};
-use lance::arrow::{FixedSizeListArrayExt, RecordBatchExt};
+use arrow_schema::SchemaRef;
 use vectordb::table::VECTOR_COLUMN_NAME;
 
 use crate::error::{MissingColumnSnafu, Result};
 use snafu::prelude::*;
 
-pub(crate) fn convert_record_batch(record_batch: RecordBatch) -> Result<RecordBatch> {
-    let column = get_column(VECTOR_COLUMN_NAME, &record_batch)?;
-
-    // TODO: we should just consume the underlying js buffer in the future instead of this arrow around a bunch of times
-    let arr = as_list_array(column.as_ref());
-    let list_size = arr.values().len() / record_batch.num_rows();
-    let r = FixedSizeListArray::try_new_from_values(arr.values().to_owned(), list_size as i32)?;
-
-    let schema = Arc::new(Schema::new(vec![Field::new(
-        VECTOR_COLUMN_NAME,
-        DataType::FixedSizeList(
-            Arc::new(Field::new("item", DataType::Float32, true)),
-            list_size as i32,
-        ),
-        true,
-    )]));
-
-    let mut new_batch = RecordBatch::try_new(schema.clone(), vec![Arc::new(r)])?;
-
-    if record_batch.num_columns() > 1 {
-        let rb = record_batch.drop_column(VECTOR_COLUMN_NAME)?;
-        new_batch = new_batch.merge(&rb)?;
-    }
-    Ok(new_batch)
-}
-
-fn get_column(column_name: &str, record_batch: &RecordBatch) -> Result<ArrayRef> {
+fn validate_vector_column(record_batch: &RecordBatch) -> Result<()> {
     record_batch
-        .column_by_name(column_name)
-        .cloned()
-        .context(MissingColumnSnafu { name: column_name })
+        .column_by_name(VECTOR_COLUMN_NAME)
+        .map(|_| ())
+        .context(MissingColumnSnafu {
+            name: VECTOR_COLUMN_NAME,
+        })
 }
 
-pub(crate) fn arrow_buffer_to_record_batch(slice: &[u8]) -> Result<Vec<RecordBatch>> {
+pub(crate) fn arrow_buffer_to_record_batch(slice: &[u8]) -> Result<(Vec<RecordBatch>, SchemaRef)> {
     let mut batches: Vec<RecordBatch> = Vec::new();
     let file_reader = FileReader::try_new(Cursor::new(slice), None)?;
+    let schema = file_reader.schema().clone();
     for b in file_reader {
-        let record_batch = convert_record_batch(b?)?;
+        let record_batch = b?;
+        validate_vector_column(&record_batch)?;
         batches.push(record_batch);
     }
-    Ok(batches)
+    Ok((batches, schema))
 }
 
 pub(crate) fn record_batch_to_buffer(batches: Vec<RecordBatch>) -> Result<Vec<u8>> {

rust/ffi/node/src/error.rs

@@ -67,6 +67,14 @@ impl From<neon::result::Throw> for Error {
     }
 }
 
+impl<T> From<std::sync::mpsc::SendError<T>> for Error {
+    fn from(value: std::sync::mpsc::SendError<T>) -> Self {
+        Self::Neon {
+            message: value.to_string(),
+        }
+    }
+}
+
 /// ResultExt is used to transform a [`Result`] into a [`NeonResult`],
 /// so it can be returned as a JavaScript error
 /// Copied from [Neon](https://github.com/neon-bindings/neon/blob/4c2e455a9e6814f1ba0178616d63caec7f4df317/crates/neon/src/result/mod.rs#L88)

rust/ffi/node/src/index/vector.rs

@@ -12,20 +12,19 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use std::convert::TryFrom;
-
-use lance::index::vector::ivf::IvfBuildParams;
-use lance::index::vector::pq::PQBuildParams;
-use lance::index::vector::MetricType;
+use lance::index::vector::{ivf::IvfBuildParams, pq::PQBuildParams};
+use lance_linalg::distance::MetricType;
 use neon::context::FunctionContext;
 use neon::prelude::*;
+use std::convert::TryFrom;
 
 use vectordb::index::vector::{IvfPQIndexBuilder, VectorIndexBuilder};
 
 use crate::error::Error::InvalidIndexType;
 use crate::error::ResultExt;
 use crate::neon_ext::js_object_ext::JsObjectExt;
-use crate::{runtime, JsTable};
+use crate::runtime;
+use crate::table::JsTable;
 
 pub(crate) fn table_create_vector_index(mut cx: FunctionContext) -> JsResult<JsPromise> {
     let js_table = cx.this().downcast_or_throw::<JsBox<JsTable>, _>(&mut cx)?;
@@ -33,22 +32,17 @@ pub(crate) fn table_create_vector_index(mut cx: FunctionContext) -> JsResult<JsP
     let index_params_builder = get_index_params_builder(&mut cx, index_params).or_throw(&mut cx)?;
 
     let rt = runtime(&mut cx)?;
-    let channel = cx.channel();
 
     let (deferred, promise) = cx.promise();
-    let table = js_table.table.clone();
+    let channel = cx.channel();
+    let mut table = js_table.table.clone();
 
-    rt.block_on(async move {
-        let add_result = table
-            .lock()
-            .unwrap()
-            .create_index(&index_params_builder)
-            .await;
+    rt.spawn(async move {
+        let idx_result = table.create_index(&index_params_builder).await;
 
         deferred.settle_with(&channel, move |mut cx| {
-            add_result
-                .map(|_| cx.undefined())
-                .or_else(|err| cx.throw_error(err.to_string()))
+            idx_result.or_throw(&mut cx)?;
+            Ok(cx.boxed(JsTable::from(table)))
         });
     });
     Ok(promise)
@@ -87,7 +81,7 @@ fn get_index_params_builder(
                 let ivf_params = IvfBuildParams {
                     num_partitions: np,
                     max_iters,
-                    centroids: None,
+                    ..Default::default()
                 };
                 index_builder.ivf_params(ivf_params)
             });

rust/ffi/node/src/lib.rs

@@ -12,36 +12,30 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-use std::collections::HashMap;
-use std::convert::TryFrom;
-use std::ops::Deref;
-use std::sync::{Arc, Mutex};
+use std::sync::Arc;
 
-use arrow_array::{Float32Array, RecordBatchIterator};
 use async_trait::async_trait;
-use futures::{TryFutureExt, TryStreamExt};
-use lance::dataset::{WriteMode, WriteParams};
-use lance::index::vector::MetricType;
 use lance::io::object_store::ObjectStoreParams;
 use neon::prelude::*;
-use neon::types::buffer::TypedArray;
 use object_store::aws::{AwsCredential, AwsCredentialProvider};
 use object_store::CredentialProvider;
 use once_cell::sync::OnceCell;
 use tokio::runtime::Runtime;
 
 use vectordb::database::Database;
-use vectordb::table::{ReadParams, Table};
+use vectordb::table::ReadParams;
 
-use crate::arrow::{arrow_buffer_to_record_batch, record_batch_to_buffer};
 use crate::error::ResultExt;
-use crate::neon_ext::js_object_ext::JsObjectExt;
+use crate::query::JsQuery;
+use crate::table::JsTable;
 
 mod arrow;
 mod convert;
 mod error;
 mod index;
 mod neon_ext;
+mod query;
+mod table;
 
 struct JsDatabase {
     database: Arc<Database>,
@@ -49,12 +43,6 @@ struct JsDatabase {
 
 impl Finalize for JsDatabase {}
 
-struct JsTable {
-    table: Arc<Mutex<Table>>,
-}
-
-impl Finalize for JsTable {}
-
 // TODO: object_store didn't export this type so I copied it.
 // Make a request to object_store to export this type
 #[derive(Debug)]
@@ -133,26 +121,28 @@ fn database_table_names(mut cx: FunctionContext) -> JsResult<JsPromise> {
     Ok(promise)
 }
 
-fn get_aws_creds<T>(
+/// Get AWS creds arguments from the context
+/// Consumes 3 arguments
+fn get_aws_creds(
     cx: &mut FunctionContext,
     arg_starting_location: i32,
-) -> Result<Option<AwsCredentialProvider>, NeonResult<T>> {
+) -> NeonResult<Option<AwsCredentialProvider>> {
     let secret_key_id = cx
         .argument_opt(arg_starting_location)
-        .map(|arg| arg.downcast_or_throw::<JsString, FunctionContext>(cx).ok())
-        .flatten()
+        .filter(|arg| arg.is_a::<JsString, _>(cx))
+        .and_then(|arg| arg.downcast_or_throw::<JsString, FunctionContext>(cx).ok())
         .map(|v| v.value(cx));
 
     let secret_key = cx
         .argument_opt(arg_starting_location + 1)
-        .map(|arg| arg.downcast_or_throw::<JsString, FunctionContext>(cx).ok())
-        .flatten()
+        .filter(|arg| arg.is_a::<JsString, _>(cx))
+        .and_then(|arg| arg.downcast_or_throw::<JsString, FunctionContext>(cx).ok())
         .map(|v| v.value(cx));
 
     let temp_token = cx
         .argument_opt(arg_starting_location + 2)
-        .map(|arg| arg.downcast_or_throw::<JsString, FunctionContext>(cx).ok())
-        .flatten()
+        .filter(|arg| arg.is_a::<JsString, _>(cx))
+        .and_then(|arg| arg.downcast_or_throw::<JsString, FunctionContext>(cx).ok())
         .map(|v| v.value(cx));
 
     match (secret_key_id, secret_key, temp_token) {
@@ -164,7 +154,21 @@ fn get_aws_creds<T>(
             }),
         ))),
         (None, None, None) => Ok(None),
-        _ => Err(cx.throw_error("Invalid credentials configuration")),
+        _ => cx.throw_error("Invalid credentials configuration"),
+    }
+}
+
+/// Get AWS region arguments from the context
+fn get_aws_region(cx: &mut FunctionContext, arg_location: i32) -> NeonResult<Option<String>> {
+    let region = cx
+        .argument_opt(arg_location)
+        .filter(|arg| arg.is_a::<JsString, _>(cx))
+        .map(|arg| arg.downcast_or_throw::<JsString, FunctionContext>(cx));
+
+    match region {
+        Some(Ok(region)) => Ok(Some(region.value(cx))),
+        None => Ok(None),
+        Some(Err(e)) => Err(e),
     }
 }
 
@@ -174,16 +178,14 @@ fn database_open_table(mut cx: FunctionContext) -> JsResult<JsPromise> {
         .downcast_or_throw::<JsBox<JsDatabase>, _>(&mut cx)?;
     let table_name = cx.argument::<JsString>(0)?.value(&mut cx);
 
-    let aws_creds = match get_aws_creds(&mut cx, 1) {
-        Ok(creds) => creds,
-        Err(err) => return err,
-    };
+    let aws_creds = get_aws_creds(&mut cx, 1)?;
+
+    let aws_region = get_aws_region(&mut cx, 4)?;
 
     let params = ReadParams {
-        store_options: Some(ObjectStoreParams {
-            aws_credentials: aws_creds,
-            ..ObjectStoreParams::default()
-        }),
+        store_options: Some(ObjectStoreParams::with_aws_credentials(
+            aws_creds, aws_region,
+        )),
         ..ReadParams::default()
     };
 
@@ -193,11 +195,11 @@ fn database_open_table(mut cx: FunctionContext) -> JsResult<JsPromise> {
 
     let (deferred, promise) = cx.promise();
     rt.spawn(async move {
-        let table_rst = database.open_table_with_params(&table_name, &params).await;
+        let table_rst = database.open_table_with_params(&table_name, params).await;
 
         deferred.settle_with(&channel, move |mut cx| {
-            let table = Arc::new(Mutex::new(table_rst.or_throw(&mut cx)?));
-            Ok(cx.boxed(JsTable { table }))
+            let js_table = JsTable::from(table_rst.or_throw(&mut cx)?);
+            Ok(cx.boxed(js_table))
         });
     });
     Ok(promise)
@@ -224,221 +226,19 @@ fn database_drop_table(mut cx: FunctionContext) -> JsResult<JsPromise> {
     Ok(promise)
 }
 
-fn table_search(mut cx: FunctionContext) -> JsResult<JsPromise> {
-    let js_table = cx.this().downcast_or_throw::<JsBox<JsTable>, _>(&mut cx)?;
-    let query_obj = cx.argument::<JsObject>(0)?;
-
-    let limit = query_obj
-        .get::<JsNumber, _, _>(&mut cx, "_limit")?
-        .value(&mut cx);
-    let select = query_obj
-        .get_opt::<JsArray, _, _>(&mut cx, "_select")?
-        .map(|arr| {
-            let js_array = arr.deref();
-            let mut projection_vec: Vec<String> = Vec::new();
-            for i in 0..js_array.len(&mut cx) {
-                let entry: Handle<JsString> = js_array.get(&mut cx, i).unwrap();
-                projection_vec.push(entry.value(&mut cx));
-            }
-            projection_vec
-        });
-    let filter = query_obj
-        .get_opt::<JsString, _, _>(&mut cx, "_filter")?
-        .map(|s| s.value(&mut cx));
-    let refine_factor = query_obj
-        .get_opt_u32(&mut cx, "_refineFactor")
-        .or_throw(&mut cx)?;
-    let nprobes = query_obj.get_usize(&mut cx, "_nprobes").or_throw(&mut cx)?;
-    let metric_type = query_obj
-        .get_opt::<JsString, _, _>(&mut cx, "_metricType")?
-        .map(|s| s.value(&mut cx))
-        .map(|s| MetricType::try_from(s.as_str()).unwrap());
-
-    let rt = runtime(&mut cx)?;
-    let channel = cx.channel();
-
-    let (deferred, promise) = cx.promise();
-    let table = js_table.table.clone();
-    let query_vector = query_obj.get::<JsArray, _, _>(&mut cx, "_queryVector")?;
-    let query = convert::js_array_to_vec(query_vector.deref(), &mut cx);
-
-    rt.spawn(async move {
-        let builder = table
-            .lock()
-            .unwrap()
-            .search(Float32Array::from(query))
-            .limit(limit as usize)
-            .refine_factor(refine_factor)
-            .nprobes(nprobes)
-            .filter(filter)
-            .metric_type(metric_type)
-            .select(select);
-        let record_batch_stream = builder.execute();
-        let results = record_batch_stream
-            .and_then(|stream| {
-                stream
-                    .try_collect::<Vec<_>>()
-                    .map_err(vectordb::error::Error::from)
-            })
-            .await;
-
-        deferred.settle_with(&channel, move |mut cx| {
-            let results = results.or_throw(&mut cx)?;
-            let buffer = record_batch_to_buffer(results).or_throw(&mut cx)?;
-            Ok(JsBuffer::external(&mut cx, buffer))
-        });
-    });
-    Ok(promise)
-}
-
-fn table_create(mut cx: FunctionContext) -> JsResult<JsPromise> {
-    let db = cx
-        .this()
-        .downcast_or_throw::<JsBox<JsDatabase>, _>(&mut cx)?;
-    let table_name = cx.argument::<JsString>(0)?.value(&mut cx);
-    let buffer = cx.argument::<JsBuffer>(1)?;
-    let batches = arrow_buffer_to_record_batch(buffer.as_slice(&mut cx)).or_throw(&mut cx)?;
-    let schema = batches[0].schema();
-
-    // Write mode
-    let mode = match cx.argument::<JsString>(2)?.value(&mut cx).as_str() {
-        "overwrite" => WriteMode::Overwrite,
-        "append" => WriteMode::Append,
-        "create" => WriteMode::Create,
-        _ => return cx.throw_error("Table::create only supports 'overwrite' and 'create' modes"),
-    };
-
-    let rt = runtime(&mut cx)?;
-    let channel = cx.channel();
-
-    let (deferred, promise) = cx.promise();
-    let database = db.database.clone();
-
-    let aws_creds = match get_aws_creds(&mut cx, 3) {
-        Ok(creds) => creds,
-        Err(err) => return err,
-    };
-
-    let params = WriteParams {
-        store_params: Some(ObjectStoreParams {
-            aws_credentials: aws_creds,
-            ..ObjectStoreParams::default()
-        }),
-        mode: mode,
-        ..WriteParams::default()
-    };
-
-    rt.block_on(async move {
-        let batch_reader = RecordBatchIterator::new(batches.into_iter().map(Ok), schema);
-        let table_rst = database
-            .create_table(&table_name, batch_reader, Some(params))
-            .await;
-
-        deferred.settle_with(&channel, move |mut cx| {
-            let table = Arc::new(Mutex::new(table_rst.or_throw(&mut cx)?));
-            Ok(cx.boxed(JsTable { table }))
-        });
-    });
-    Ok(promise)
-}
-
-fn table_add(mut cx: FunctionContext) -> JsResult<JsPromise> {
-    let write_mode_map: HashMap<&str, WriteMode> = HashMap::from([
-        ("create", WriteMode::Create),
-        ("append", WriteMode::Append),
-        ("overwrite", WriteMode::Overwrite),
-    ]);
-
-    let js_table = cx.this().downcast_or_throw::<JsBox<JsTable>, _>(&mut cx)?;
-    let buffer = cx.argument::<JsBuffer>(0)?;
-    let write_mode = cx.argument::<JsString>(1)?.value(&mut cx);
-
-    let batches = arrow_buffer_to_record_batch(buffer.as_slice(&mut cx)).or_throw(&mut cx)?;
-    let schema = batches[0].schema();
-
-    let rt = runtime(&mut cx)?;
-    let channel = cx.channel();
-
-    let (deferred, promise) = cx.promise();
-    let table = js_table.table.clone();
-    let write_mode = write_mode_map.get(write_mode.as_str()).cloned();
-
-    let aws_creds = match get_aws_creds(&mut cx, 2) {
-        Ok(creds) => creds,
-        Err(err) => return err,
-    };
-
-    let params = WriteParams {
-        store_params: Some(ObjectStoreParams {
-            aws_credentials: aws_creds,
-            ..ObjectStoreParams::default()
-        }),
-        mode: write_mode.unwrap_or(WriteMode::Append),
-        ..WriteParams::default()
-    };
-
-    rt.block_on(async move {
-        let batch_reader = RecordBatchIterator::new(batches.into_iter().map(Ok), schema);
-        let add_result = table.lock().unwrap().add(batch_reader, Some(params)).await;
-
-        deferred.settle_with(&channel, move |mut cx| {
-            let _added = add_result.or_throw(&mut cx)?;
-            Ok(cx.boolean(true))
-        });
-    });
-    Ok(promise)
-}
-
-fn table_count_rows(mut cx: FunctionContext) -> JsResult<JsPromise> {
-    let js_table = cx.this().downcast_or_throw::<JsBox<JsTable>, _>(&mut cx)?;
-    let rt = runtime(&mut cx)?;
-    let channel = cx.channel();
-
-    let (deferred, promise) = cx.promise();
-    let table = js_table.table.clone();
-
-    rt.block_on(async move {
-        let num_rows_result = table.lock().unwrap().count_rows().await;
-
-        deferred.settle_with(&channel, move |mut cx| {
-            let num_rows = num_rows_result.or_throw(&mut cx)?;
-            Ok(cx.number(num_rows as f64))
-        });
-    });
-    Ok(promise)
-}
-
-fn table_delete(mut cx: FunctionContext) -> JsResult<JsPromise> {
-    let js_table = cx.this().downcast_or_throw::<JsBox<JsTable>, _>(&mut cx)?;
-    let rt = runtime(&mut cx)?;
-    let channel = cx.channel();
-
-    let (deferred, promise) = cx.promise();
-    let table = js_table.table.clone();
-
-    let predicate = cx.argument::<JsString>(0)?.value(&mut cx);
-
-    let delete_result = rt.block_on(async move { table.lock().unwrap().delete(&predicate).await });
-
-    deferred.settle_with(&channel, move |mut cx| {
-        delete_result.or_throw(&mut cx)?;
-        Ok(cx.undefined())
-    });
-
-    Ok(promise)
-}
-
 #[neon::main]
 fn main(mut cx: ModuleContext) -> NeonResult<()> {
     cx.export_function("databaseNew", database_new)?;
     cx.export_function("databaseTableNames", database_table_names)?;
     cx.export_function("databaseOpenTable", database_open_table)?;
     cx.export_function("databaseDropTable", database_drop_table)?;
-    cx.export_function("tableSearch", table_search)?;
-    cx.export_function("tableCreate", table_create)?;
-    cx.export_function("tableAdd", table_add)?;
-    cx.export_function("tableCountRows", table_count_rows)?;
-    cx.export_function("tableDelete", table_delete)?;
+    cx.export_function("tableSearch", JsQuery::js_search)?;
+    cx.export_function("tableCreate", JsTable::js_create)?;
+    cx.export_function("tableAdd", JsTable::js_add)?;
+    cx.export_function("tableCountRows", JsTable::js_count_rows)?;
+    cx.export_function("tableDelete", JsTable::js_delete)?;
+    cx.export_function("tableCleanupOldVersions", JsTable::js_cleanup)?;
+    cx.export_function("tableCompactFiles", JsTable::js_compact)?;
     cx.export_function(
         "tableCreateVectorIndex",
         index::vector::table_create_vector_index,

rust/ffi/node/src/query.rs

@@ -0,0 +1,107 @@
+use std::convert::TryFrom;
+use std::ops::Deref;
+
+use arrow_array::Float32Array;
+use futures::{TryFutureExt, TryStreamExt};
+use lance_linalg::distance::MetricType;
+use neon::context::FunctionContext;
+use neon::handle::Handle;
+use neon::prelude::*;
+use neon::types::buffer::TypedArray;
+
+use crate::arrow::record_batch_to_buffer;
+use crate::error::ResultExt;
+use crate::neon_ext::js_object_ext::JsObjectExt;
+use crate::table::JsTable;
+use crate::{convert, runtime};
+
+pub(crate) struct JsQuery {}
+
+impl JsQuery {
+    pub(crate) fn js_search(mut cx: FunctionContext) -> JsResult<JsPromise> {
+        let js_table = cx.this().downcast_or_throw::<JsBox<JsTable>, _>(&mut cx)?;
+        let query_obj = cx.argument::<JsObject>(0)?;
+
+        let limit = query_obj
+            .get::<JsNumber, _, _>(&mut cx, "_limit")?
+            .value(&mut cx);
+        let select = query_obj
+            .get_opt::<JsArray, _, _>(&mut cx, "_select")?
+            .map(|arr| {
+                let js_array = arr.deref();
+                let mut projection_vec: Vec<String> = Vec::new();
+                for i in 0..js_array.len(&mut cx) {
+                    let entry: Handle<JsString> = js_array.get(&mut cx, i).unwrap();
+                    projection_vec.push(entry.value(&mut cx));
+                }
+                projection_vec
+            });
+        let filter = query_obj
+            .get_opt::<JsString, _, _>(&mut cx, "_filter")?
+            .map(|s| s.value(&mut cx));
+        let refine_factor = query_obj
+            .get_opt_u32(&mut cx, "_refineFactor")
+            .or_throw(&mut cx)?;
+        let nprobes = query_obj.get_usize(&mut cx, "_nprobes").or_throw(&mut cx)?;
+        let metric_type = query_obj
+            .get_opt::<JsString, _, _>(&mut cx, "_metricType")?
+            .map(|s| s.value(&mut cx))
+            .map(|s| MetricType::try_from(s.as_str()).unwrap());
+
+        let is_electron = cx
+            .argument::<JsBoolean>(1)
+            .or_throw(&mut cx)?
+            .value(&mut cx);
+
+        let rt = runtime(&mut cx)?;
+
+        let (deferred, promise) = cx.promise();
+        let channel = cx.channel();
+        let query_vector = query_obj.get::<JsArray, _, _>(&mut cx, "_queryVector")?;
+        let query = convert::js_array_to_vec(query_vector.deref(), &mut cx);
+        let table = js_table.table.clone();
+
+        rt.spawn(async move {
+            let builder = table
+                .search(Float32Array::from(query))
+                .limit(limit as usize)
+                .refine_factor(refine_factor)
+                .nprobes(nprobes)
+                .filter(filter)
+                .metric_type(metric_type)
+                .select(select);
+            let record_batch_stream = builder.execute();
+            let results = record_batch_stream
+                .and_then(|stream| {
+                    stream
+                        .try_collect::<Vec<_>>()
+                        .map_err(vectordb::error::Error::from)
+                })
+                .await;
+
+            deferred.settle_with(&channel, move |mut cx| {
+                let results = results.or_throw(&mut cx)?;
+                let buffer = record_batch_to_buffer(results).or_throw(&mut cx)?;
+                Self::new_js_buffer(buffer, &mut cx, is_electron)
+            });
+        });
+        Ok(promise)
+    }
+
+    // Creates a new JsBuffer from a rust buffer with a special logic for electron
+    fn new_js_buffer<'a>(
+        buffer: Vec<u8>,
+        cx: &mut TaskContext<'a>,
+        is_electron: bool,
+    ) -> NeonResult<Handle<'a, JsBuffer>> {
+        if is_electron {
+            // Electron does not support `external`: https://github.com/neon-bindings/neon/pull/937
+            let mut js_buffer = JsBuffer::new(cx, buffer.len()).or_throw(cx)?;
+            let buffer_data = js_buffer.as_mut_slice(cx);
+            buffer_data.copy_from_slice(buffer.as_slice());
+            Ok(js_buffer)
+        } else {
+            Ok(JsBuffer::external(cx, buffer))
+        }
+    }
+}

rust/ffi/node/src/table.rs

@@ -0,0 +1,279 @@
+// Copyright 2023 Lance Developers.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+use arrow_array::RecordBatchIterator;
+use lance::dataset::optimize::CompactionOptions;
+use lance::dataset::{WriteMode, WriteParams};
+use lance::io::object_store::ObjectStoreParams;
+
+use crate::arrow::arrow_buffer_to_record_batch;
+use neon::prelude::*;
+use neon::types::buffer::TypedArray;
+use vectordb::Table;
+
+use crate::error::ResultExt;
+use crate::{get_aws_creds, get_aws_region, runtime, JsDatabase};
+
+pub(crate) struct JsTable {
+    pub table: Table,
+}
+
+impl Finalize for JsTable {}
+
+impl From<Table> for JsTable {
+    fn from(table: Table) -> Self {
+        JsTable { table }
+    }
+}
+
+impl JsTable {
+    pub(crate) fn js_create(mut cx: FunctionContext) -> JsResult<JsPromise> {
+        let db = cx
+            .this()
+            .downcast_or_throw::<JsBox<JsDatabase>, _>(&mut cx)?;
+        let table_name = cx.argument::<JsString>(0)?.value(&mut cx);
+        let buffer = cx.argument::<JsBuffer>(1)?;
+        let (batches, schema) =
+            arrow_buffer_to_record_batch(buffer.as_slice(&mut cx)).or_throw(&mut cx)?;
+
+        // Write mode
+        let mode = match cx.argument::<JsString>(2)?.value(&mut cx).as_str() {
+            "overwrite" => WriteMode::Overwrite,
+            "append" => WriteMode::Append,
+            "create" => WriteMode::Create,
+            _ => {
+                return cx.throw_error("Table::create only supports 'overwrite' and 'create' modes")
+            }
+        };
+
+        let rt = runtime(&mut cx)?;
+        let channel = cx.channel();
+
+        let (deferred, promise) = cx.promise();
+        let database = db.database.clone();
+
+        let aws_creds = get_aws_creds(&mut cx, 3)?;
+        let aws_region = get_aws_region(&mut cx, 6)?;
+
+        let params = WriteParams {
+            store_params: Some(ObjectStoreParams::with_aws_credentials(
+                aws_creds, aws_region,
+            )),
+            mode: mode,
+            ..WriteParams::default()
+        };
+
+        rt.spawn(async move {
+            let batch_reader = RecordBatchIterator::new(batches.into_iter().map(Ok), schema);
+            let table_rst = database
+                .create_table(&table_name, batch_reader, Some(params))
+                .await;
+
+            deferred.settle_with(&channel, move |mut cx| {
+                let table = table_rst.or_throw(&mut cx)?;
+                Ok(cx.boxed(JsTable::from(table)))
+            });
+        });
+        Ok(promise)
+    }
+
+    pub(crate) fn js_add(mut cx: FunctionContext) -> JsResult<JsPromise> {
+        let js_table = cx.this().downcast_or_throw::<JsBox<JsTable>, _>(&mut cx)?;
+        let buffer = cx.argument::<JsBuffer>(0)?;
+        let write_mode = cx.argument::<JsString>(1)?.value(&mut cx);
+        let (batches, schema) =
+            arrow_buffer_to_record_batch(buffer.as_slice(&mut cx)).or_throw(&mut cx)?;
+        let rt = runtime(&mut cx)?;
+        let channel = cx.channel();
+        let mut table = js_table.table.clone();
+
+        let (deferred, promise) = cx.promise();
+        let write_mode = match write_mode.as_str() {
+            "create" => WriteMode::Create,
+            "append" => WriteMode::Append,
+            "overwrite" => WriteMode::Overwrite,
+            s => return cx.throw_error(format!("invalid write mode {}", s)),
+        };
+        let aws_creds = get_aws_creds(&mut cx, 2)?;
+        let aws_region = get_aws_region(&mut cx, 5)?;
+
+        let params = WriteParams {
+            store_params: Some(ObjectStoreParams::with_aws_credentials(
+                aws_creds, aws_region,
+            )),
+            mode: write_mode,
+            ..WriteParams::default()
+        };
+
+        rt.spawn(async move {
+            let batch_reader = RecordBatchIterator::new(batches.into_iter().map(Ok), schema);
+            let add_result = table.add(batch_reader, Some(params)).await;
+
+            deferred.settle_with(&channel, move |mut cx| {
+                let _added = add_result.or_throw(&mut cx)?;
+                Ok(cx.boxed(JsTable::from(table)))
+            });
+        });
+        Ok(promise)
+    }
+
+    pub(crate) fn js_count_rows(mut cx: FunctionContext) -> JsResult<JsPromise> {
+        let js_table = cx.this().downcast_or_throw::<JsBox<JsTable>, _>(&mut cx)?;
+        let rt = runtime(&mut cx)?;
+        let (deferred, promise) = cx.promise();
+        let channel = cx.channel();
+        let table = js_table.table.clone();
+
+        rt.spawn(async move {
+            let num_rows_result = table.count_rows().await;
+
+            deferred.settle_with(&channel, move |mut cx| {
+                let num_rows = num_rows_result.or_throw(&mut cx)?;
+                Ok(cx.number(num_rows as f64))
+            });
+        });
+        Ok(promise)
+    }
+
+    pub(crate) fn js_delete(mut cx: FunctionContext) -> JsResult<JsPromise> {
+        let js_table = cx.this().downcast_or_throw::<JsBox<JsTable>, _>(&mut cx)?;
+        let rt = runtime(&mut cx)?;
+        let (deferred, promise) = cx.promise();
+        let predicate = cx.argument::<JsString>(0)?.value(&mut cx);
+        let channel = cx.channel();
+        let mut table = js_table.table.clone();
+
+        rt.spawn(async move {
+            let delete_result = table.delete(&predicate).await;
+
+            deferred.settle_with(&channel, move |mut cx| {
+                delete_result.or_throw(&mut cx)?;
+                Ok(cx.boxed(JsTable::from(table)))
+            })
+        });
+        Ok(promise)
+    }
+
+    pub(crate) fn js_cleanup(mut cx: FunctionContext) -> JsResult<JsPromise> {
+        let js_table = cx.this().downcast_or_throw::<JsBox<JsTable>, _>(&mut cx)?;
+        let rt = runtime(&mut cx)?;
+        let (deferred, promise) = cx.promise();
+        let table = js_table.table.clone();
+        let channel = cx.channel();
+
+        let older_than: i64 = cx
+            .argument_opt(0)
+            .and_then(|val| val.downcast::<JsNumber, _>(&mut cx).ok())
+            .map(|val| val.value(&mut cx) as i64)
+            .unwrap_or_else(|| 2 * 7 * 24 * 60); // 2 weeks
+        let older_than = chrono::Duration::minutes(older_than);
+        let delete_unverified: bool = cx
+            .argument_opt(1)
+            .and_then(|val| val.downcast::<JsBoolean, _>(&mut cx).ok())
+            .map(|val| val.value(&mut cx))
+            .unwrap_or_default();
+
+        rt.spawn(async move {
+            let stats = table
+                .cleanup_old_versions(older_than, Some(delete_unverified))
+                .await;
+
+            deferred.settle_with(&channel, move |mut cx| {
+                let stats = stats.or_throw(&mut cx)?;
+
+                let output_metrics = JsObject::new(&mut cx);
+                let bytes_removed = cx.number(stats.bytes_removed as f64);
+                output_metrics.set(&mut cx, "bytesRemoved", bytes_removed)?;
+
+                let old_versions = cx.number(stats.old_versions as f64);
+                output_metrics.set(&mut cx, "oldVersions", old_versions)?;
+
+                let output_table = cx.boxed(JsTable::from(table));
+
+                let output = JsObject::new(&mut cx);
+                output.set(&mut cx, "metrics", output_metrics)?;
+                output.set(&mut cx, "newTable", output_table)?;
+
+                Ok(output)
+            })
+        });
+        Ok(promise)
+    }
+
+    pub(crate) fn js_compact(mut cx: FunctionContext) -> JsResult<JsPromise> {
+        let js_table = cx.this().downcast_or_throw::<JsBox<JsTable>, _>(&mut cx)?;
+        let rt = runtime(&mut cx)?;
+        let (deferred, promise) = cx.promise();
+        let mut table = js_table.table.clone();
+        let channel = cx.channel();
+
+        let js_options = cx.argument::<JsObject>(0)?;
+        let mut options = CompactionOptions::default();
+
+        if let Some(target_rows) =
+            js_options.get_opt::<JsNumber, _, _>(&mut cx, "targetRowsPerFragment")?
+        {
+            options.target_rows_per_fragment = target_rows.value(&mut cx) as usize;
+        }
+        if let Some(max_per_group) =
+            js_options.get_opt::<JsNumber, _, _>(&mut cx, "maxRowsPerGroup")?
+        {
+            options.max_rows_per_group = max_per_group.value(&mut cx) as usize;
+        }
+        if let Some(materialize_deletions) =
+            js_options.get_opt::<JsBoolean, _, _>(&mut cx, "materializeDeletions")?
+        {
+            options.materialize_deletions = materialize_deletions.value(&mut cx);
+        }
+        if let Some(materialize_deletions_threshold) =
+            js_options.get_opt::<JsNumber, _, _>(&mut cx, "materializeDeletionsThreshold")?
+        {
+            options.materialize_deletions_threshold =
+                materialize_deletions_threshold.value(&mut cx) as f32;
+        }
+        if let Some(num_threads) = js_options.get_opt::<JsNumber, _, _>(&mut cx, "numThreads")? {
+            options.num_threads = num_threads.value(&mut cx) as usize;
+        }
+
+        rt.spawn(async move {
+            let stats = table.compact_files(options).await;
+
+            deferred.settle_with(&channel, move |mut cx| {
+                let stats = stats.or_throw(&mut cx)?;
+
+                let output_metrics = JsObject::new(&mut cx);
+                let fragments_removed = cx.number(stats.fragments_removed as f64);
+                output_metrics.set(&mut cx, "fragmentsRemoved", fragments_removed)?;
+
+                let fragments_added = cx.number(stats.fragments_added as f64);
+                output_metrics.set(&mut cx, "fragmentsAdded", fragments_added)?;
+
+                let files_removed = cx.number(stats.files_removed as f64);
+                output_metrics.set(&mut cx, "filesRemoved", files_removed)?;
+
+                let files_added = cx.number(stats.files_added as f64);
+                output_metrics.set(&mut cx, "filesAdded", files_added)?;
+
+                let output_table = cx.boxed(JsTable::from(table));
+
+                let output = JsObject::new(&mut cx);
+                output.set(&mut cx, "metrics", output_metrics)?;
+                output.set(&mut cx, "newTable", output_table)?;
+
+                Ok(output)
+            })
+        });
+        Ok(promise)
+    }
+}