Bump version: 0.26.2 → 0.27.0-beta.0

When a table has a read consistency interval, queries within the
interval skip the version check. Once the interval expires, a list call
checks for new versions. If the version hasn't changed, the timer should
reset so the next interval begins, but it didn't. The timer stayed
expired, so every query after that triggered a list call, even though
nothing changed.

This affects all read operations (queries, schema lookups, searches) on
tables with read_consistency_interval set. Each operation adds a
list("_versions/") call to object storage, adding latency proportional
to the store's list performance. For high-QPS workloads, this can
saturate object store list throughput and significantly degrade query
latency.

Bug flow:
1. Every read operation (query, schema, search) calls
ensure_up_to_date()
2. ensure_up_to_date() calls is_up_to_date(), which compares
last_consistency_check.elapsed() against
   read_consistency_interval
  3. If the interval has expired, it calls reload()
4. reload() calls need_reload(), which calls latest_version_id() — this
is the list IOP
  (list("_versions/"))
5. If no new version, reload() returns early without resetting
last_consistency_check
6. On the next query, step 2 sees the stale timer again → step 3 → step
4 → another list IOP
  7. This repeats on every query forever

.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.26.1"
+current_version = "0.27.0-beta.0"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

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

@@ -0,0 +1,173 @@
+name: Codex Fix CI
+
+on:
+  workflow_dispatch:
+    inputs:
+      workflow_run_url:
+        description: "Failing CI workflow run URL (e.g., https://github.com/lancedb/lancedb/actions/runs/12345678)"
+        required: true
+        type: string
+      branch:
+        description: "Branch to fix (e.g., main, release/v2.0, or feature-branch)"
+        required: true
+        type: string
+      guidelines:
+        description: "Additional guidelines for the fix (optional)"
+        required: false
+        type: string
+
+permissions:
+  contents: write
+  pull-requests: write
+  actions: read
+
+jobs:
+  fix-ci:
+    runs-on: warp-ubuntu-latest-x64-4x
+    timeout-minutes: 60
+    env:
+      CC: clang
+      CXX: clang++
+    steps:
+      - name: Show inputs
+        run: |
+          echo "workflow_run_url = ${{ inputs.workflow_run_url }}"
+          echo "branch = ${{ inputs.branch }}"
+          echo "guidelines = ${{ inputs.guidelines }}"
+
+      - name: Checkout Repo
+        uses: actions/checkout@v4
+        with:
+          ref: ${{ inputs.branch }}
+          fetch-depth: 0
+          persist-credentials: true
+
+      - name: Set up Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: 20
+
+      - name: Install Codex CLI
+        run: npm install -g @openai/codex
+
+      - name: Install Rust toolchain
+        uses: dtolnay/rust-toolchain@stable
+        with:
+          toolchain: stable
+          components: clippy, rustfmt
+
+      - uses: Swatinem/rust-cache@v2
+
+      - name: Install system dependencies
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y protobuf-compiler libssl-dev
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+
+      - name: Install Python dependencies
+        run: |
+          pip install maturin ruff pytest pyarrow pandas polars
+
+      - name: Set up Java
+        uses: actions/setup-java@v4
+        with:
+          distribution: temurin
+          java-version: '11'
+          cache: maven
+
+      - name: Install Node.js dependencies for TypeScript bindings
+        run: |
+          cd nodejs
+          npm ci
+
+      - name: Configure git user
+        run: |
+          git config user.name "lancedb automation"
+          git config user.email "robot@lancedb.com"
+
+      - name: Run Codex to fix CI failure
+        env:
+          WORKFLOW_RUN_URL: ${{ inputs.workflow_run_url }}
+          BRANCH: ${{ inputs.branch }}
+          GUIDELINES: ${{ inputs.guidelines }}
+          GITHUB_TOKEN: ${{ secrets.ROBOT_TOKEN }}
+          GH_TOKEN: ${{ secrets.ROBOT_TOKEN }}
+          OPENAI_API_KEY: ${{ secrets.CODEX_TOKEN }}
+        run: |
+          set -euo pipefail
+
+          cat <<EOF >/tmp/codex-prompt.txt
+          You are running inside the lancedb repository on a GitHub Actions runner. Your task is to fix a CI failure.
+
+          Input parameters:
+          - Failing workflow run URL: ${WORKFLOW_RUN_URL}
+          - Branch to fix: ${BRANCH}
+          - Additional guidelines: ${GUIDELINES:-"None provided"}
+
+          Follow these steps exactly:
+
+          1. Extract the run ID from the workflow URL. The URL format is https://github.com/lancedb/lancedb/actions/runs/<run_id>.
+
+          2. Use "gh run view <run_id> --json jobs,conclusion,name" to get information about the failed run.
+
+          3. Identify which jobs failed. For each failed job, use "gh run view <run_id> --job <job_id> --log-failed" to get the failure logs.
+
+          4. Analyze the failure logs to understand what went wrong. Common failures include:
+             - Compilation errors
+             - Test failures
+             - Clippy warnings treated as errors
+             - Formatting issues
+             - Dependency issues
+
+          5. Based on the analysis, fix the issues in the codebase:
+             - For compilation errors: Fix the code that doesn't compile
+             - For test failures: Fix the failing tests or the code they test
+             - For clippy warnings: Apply the suggested fixes
+             - For formatting issues: Run "cargo fmt --all"
+             - For other issues: Apply appropriate fixes
+
+          6. After making fixes, verify them locally:
+             - Run "cargo fmt --all" to ensure formatting is correct
+             - Run "cargo clippy --workspace --tests --all-features -- -D warnings" to check for issues
+             - Run ONLY the specific failing tests to confirm they pass now:
+               - For Rust test failures: Run the specific test with "cargo test -p <crate> <test_name>"
+               - For Python test failures: Build with "cd python && maturin develop" then run "pytest <specific_test_file>::<test_name>"
+               - For Java test failures: Run "cd java && mvn test -Dtest=<TestClass>#<testMethod>"
+               - For TypeScript test failures: Run "cd nodejs && npm run build && npm test -- --testNamePattern='<test_name>'"
+               - Do NOT run the full test suite - only run the tests that were failing
+
+          7. If the additional guidelines are provided, follow them as well.
+
+          8. Inspect "git status --short" and "git diff" to review your changes.
+
+          9. Create a fix branch: "git checkout -b codex/fix-ci-<run_id>".
+
+          10. Stage all changes with "git add -A" and commit with message "fix: resolve CI failures from run <run_id>".
+
+          11. Push the branch: "git push origin codex/fix-ci-<run_id>". If the remote branch exists, delete it first with "gh api -X DELETE repos/lancedb/lancedb/git/refs/heads/codex/fix-ci-<run_id>" then push. Do NOT use "git push --force" or "git push -f".
+
+          12. Create a pull request targeting "${BRANCH}":
+              - Title: "ci: <short summary describing the fix>" (e.g., "ci: fix clippy warnings in lancedb" or "ci: resolve test flakiness in vector search")
+              - First, write the PR body to /tmp/pr-body.md using a heredoc (cat <<'PREOF' > /tmp/pr-body.md). The body should include:
+                - Link to the failing workflow run
+                - Summary of what failed
+                - Description of the fixes applied
+              - Then run "gh pr create --base ${BRANCH} --body-file /tmp/pr-body.md".
+
+          13. Display the new PR URL, "git status --short", and a summary of what was fixed.
+
+          Constraints:
+          - Use bash commands for all operations.
+          - Do not merge the PR.
+          - Do not modify GitHub workflow files unless they are the cause of the failure.
+          - If any command fails, diagnose and attempt to fix the issue instead of aborting immediately.
+          - If you cannot fix the issue automatically, create the PR anyway with a clear explanation of what you tried and what remains to be fixed.
+          - env "GH_TOKEN" is available, use "gh" tools for GitHub-related operations.
+          EOF
+
+          printenv OPENAI_API_KEY | codex login --with-api-key
+          codex --config shell_environment_policy.ignore_default_excludes=true exec --dangerously-bypass-approvals-and-sandbox "$(cat /tmp/codex-prompt.txt)"

Cargo.lock

@@ -3072,9 +3072,9 @@ checksum = "42703706b716c37f96a77aea830392ad231f44c9e9a67872fa5548707e11b11c"
 
 [[package]]
 name = "fsst"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0f03a771ab914e207dd26bd2f12666839555ec8ecc7e1770e1ed6f9900d899a4"
+checksum = "5f9e5c0b1c67a38cb92b41535d44623483beb9511592ae23a3bf42ddec758690"
 dependencies = [
  "arrow-array",
  "rand 0.9.2",
@@ -4405,9 +4405,9 @@ dependencies = [
 
 [[package]]
 name = "lance"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "47b685aca3f97ee02997c83ded16f59c747ccb69e74c8abbbae4aa3d22cf1301"
+checksum = "2b7f07b905df393a5554eba19055c620f9ea25a3e40a013bda4bd8dc4ca66f01"
 dependencies = [
  "arrow",
  "arrow-arith",
@@ -4472,9 +4472,9 @@ dependencies = [
 
 [[package]]
 name = "lance-arrow"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "daf00c7537df524cc518a089f0d156a036d95ca3f5bc2bc1f0a9f9293e9b62ef"
+checksum = "100e076cb81c8f0c24cd2881c706fc53e037c7d6e81eb320e929e265d157effb"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -4493,9 +4493,9 @@ dependencies = [
 
 [[package]]
 name = "lance-bitpacking"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "46752e4ac8fc5590a445e780b63a8800adc7a770bd74770a8dc66963778e4e77"
+checksum = "588318d3d1ba0f97162fab39a323a0a49866bb35b32af42572c6b6a12296fa27"
 dependencies = [
  "arrayref",
  "paste",
@@ -4504,9 +4504,9 @@ dependencies = [
 
 [[package]]
 name = "lance-core"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3d13d87d07305c6d4b4dc7780fb1107babf782a0e5b1dc7872e17ae1f8fd11ca"
+checksum = "6fa01d1cf490ccfd3b8eaeee2781415d0419e6be8366040e57e43677abf2644e"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -4543,9 +4543,9 @@ dependencies = [
 
 [[package]]
 name = "lance-datafusion"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6451b5af876eaef8bec4b38a39dadac9d44621e1ecf85d0cdf6097a5d0aa8721"
+checksum = "ef89a39e3284eef76f79e63f23de8881a0583ad6feb20ed39f47eadd847a2b88"
 dependencies = [
  "arrow",
  "arrow-array",
@@ -4575,9 +4575,9 @@ dependencies = [
 
 [[package]]
 name = "lance-datagen"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e1736708dd7867dfbab8fcc930b21c96717c6c00be73b7d9a240336a4ed80375"
+checksum = "fc2a60eef5c47e65d91e2ffa8e7e1629c52e7190c8b88a371a1a60601dc49371"
 dependencies = [
  "arrow",
  "arrow-array",
@@ -4595,9 +4595,9 @@ dependencies = [
 
 [[package]]
 name = "lance-encoding"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d6b6ca4ff94833240d5ba4a94a742cba786d1949b3c3fa7e11d6f0050443432a"
+checksum = "95ce4a6631308aa681b2671af8f2a845ff781f8d4e755a2a7ccd012379467094"
 dependencies = [
  "arrow-arith",
  "arrow-array",
@@ -4634,9 +4634,9 @@ dependencies = [
 
 [[package]]
 name = "lance-file"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "55fbe959bffe185543aed3cbeb14484f1aa2e55886034fdb1ea3d8cc9b70aad8"
+checksum = "e2d4d82357cbfaa1a18494226c15b1cb3c8ed0b6c84b91146323c82047ede419"
 dependencies = [
  "arrow-arith",
  "arrow-array",
@@ -4668,9 +4668,9 @@ dependencies = [
 
 [[package]]
 name = "lance-geo"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a52b0adabc953d457f336a784a3b37353a180e6a79905f544949746e0d4c6483"
+checksum = "a7183fc870da62826f0f97df8007b634da053eb310157856efe1dc74f446951c"
 dependencies = [
  "datafusion",
  "geo-traits",
@@ -4684,9 +4684,9 @@ dependencies = [
 
 [[package]]
 name = "lance-index"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6b67654bf86fd942dd2cf08294ee7e91053427cd148225f49c9ff398ff9a40fd"
+checksum = "20e9c5aa7024a63af9ae89ee8c0f23c8421b7896742e5cd4a271a60f9956cb80"
 dependencies = [
  "arrow",
  "arrow-arith",
@@ -4753,9 +4753,9 @@ dependencies = [
 
 [[package]]
 name = "lance-io"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8eb0ccc1c414e31687d83992d546af0a0237c8d2f4bf2ae3d347d539fd0fc141"
+checksum = "c7d2af0b17fb374a8181bcf1a10bce5703ae3ee4373c1587ce4bba23e15e45c8"
 dependencies = [
  "arrow",
  "arrow-arith",
@@ -4795,9 +4795,9 @@ dependencies = [
 
 [[package]]
 name = "lance-linalg"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "083404cf12dcdb1a7df98fb58f9daf626b6e43a2f794b37b6b89b4012a0e1f78"
+checksum = "5125aa62696e75a7475807564b4921f252d8815be606b84bc00e6def0f5c24bb"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -4813,9 +4813,9 @@ dependencies = [
 
 [[package]]
 name = "lance-namespace"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c12778d2aabf9c2bfd16e2509ebe120e562a288d8ae630ec6b6b204868df41b2"
+checksum = "70545c2676ce954dfd801da5c6a631a70bba967826cd3a8f31b47d1f04bbfed3"
 dependencies = [
  "arrow",
  "async-trait",
@@ -4827,9 +4827,9 @@ dependencies = [
 
 [[package]]
 name = "lance-namespace-impls"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8863aababdd13a6d2c8d6179dc6981f4f8f49d8b66a00c5dd75115aec4cadc99"
+checksum = "92519f9f27d62655030aac62ea0db9614b65f086ebe651c1b0a96e351b668022"
 dependencies = [
  "arrow",
  "arrow-ipc",
@@ -4872,9 +4872,9 @@ dependencies = [
 
 [[package]]
 name = "lance-table"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f0fcc83f197ce2000c4abe4f5e0873490ab1f41788fa76571c4209b87d4daf50"
+checksum = "b06ad37bd90045de8ef533df170c6098e6ff6ecb427aade47d7db8e2c86f2678"
 dependencies = [
  "arrow",
  "arrow-array",
@@ -4913,9 +4913,9 @@ dependencies = [
 
 [[package]]
 name = "lance-testing"
-version = "2.0.0"
+version = "2.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7fb1f7c7e06f91360e141ecee1cf2110f858c231705f69f2cd2fda9e30c1e9f4"
+checksum = "cd7f13b0f2b6337af015dcb1519645388dca08c970037aa77aff517687c4019f"
 dependencies = [
  "arrow-array",
  "arrow-schema",
@@ -4926,7 +4926,7 @@ dependencies = [
 
 [[package]]
 name = "lancedb"
-version = "0.26.1"
+version = "0.26.2"
 dependencies = [
  "ahash",
  "anyhow",
@@ -5006,7 +5006,7 @@ dependencies = [
 
 [[package]]
 name = "lancedb-nodejs"
-version = "0.26.1"
+version = "0.26.2"
 dependencies = [
  "arrow-array",
  "arrow-ipc",
@@ -5026,7 +5026,7 @@ dependencies = [
 
 [[package]]
 name = "lancedb-python"
-version = "0.29.1"
+version = "0.29.2"
 dependencies = [
  "arrow",
  "async-trait",

Cargo.toml

@@ -15,20 +15,20 @@ categories = ["database-implementations"]
 rust-version = "1.88.0"
 
 [workspace.dependencies]
-lance = { "version" = "=2.0.0", default-features = false }
-lance-core = "=2.0.0"
-lance-datagen = "=2.0.0"
-lance-file = "=2.0.0"
-lance-io = { "version" = "=2.0.0", default-features = false }
-lance-index = "=2.0.0"
-lance-linalg = "=2.0.0"
-lance-namespace = "=2.0.0"
-lance-namespace-impls = { "version" = "=2.0.0", default-features = false }
-lance-table = "=2.0.0"
-lance-testing = "=2.0.0"
-lance-datafusion = "=2.0.0"
-lance-encoding = "=2.0.0"
-lance-arrow = "=2.0.0"
+lance = { "version" = "=2.0.1", default-features = false }
+lance-core = "=2.0.1"
+lance-datagen = "=2.0.1"
+lance-file = "=2.0.1"
+lance-io = { "version" = "=2.0.1", default-features = false }
+lance-index = "=2.0.1"
+lance-linalg = "=2.0.1"
+lance-namespace = "=2.0.1"
+lance-namespace-impls = { "version" = "=2.0.1", default-features = false }
+lance-table = "=2.0.1"
+lance-testing = "=2.0.1"
+lance-datafusion = "=2.0.1"
+lance-encoding = "=2.0.1"
+lance-arrow = "=2.0.1"
 ahash = "0.8"
 # Note that this one does not include pyarrow
 arrow = { version = "57.2", optional = false }

Makefile

@@ -0,0 +1,9 @@
+.PHONY: licenses
+
+licenses:
+	cargo about generate about.hbs -o RUST_THIRD_PARTY_LICENSES.html -c about.toml
+	cd python && cargo about generate ../about.hbs -o RUST_THIRD_PARTY_LICENSES.html -c ../about.toml
+	cd python && uv sync --all-extras && uv tool run pip-licenses --python .venv/bin/python --format=markdown --with-urls --output-file=PYTHON_THIRD_PARTY_LICENSES.md
+	cd nodejs && cargo about generate ../about.hbs -o RUST_THIRD_PARTY_LICENSES.html -c ../about.toml
+	cd nodejs && npx license-checker --markdown --out NODEJS_THIRD_PARTY_LICENSES.md
+	cd java && ./mvnw license:aggregate-add-third-party -q

about.hbs

@@ -0,0 +1,70 @@
+<html>
+
+<head>
+    <style>
+        @media (prefers-color-scheme: dark) {
+            body {
+                background: #333;
+                color: white;
+            }
+            a {
+                color: skyblue;
+            }
+        }
+        .container {
+            font-family: sans-serif;
+            max-width: 800px;
+            margin: 0 auto;
+        }
+        .intro {
+            text-align: center;
+        }
+        .licenses-list {
+            list-style-type: none;
+            margin: 0;
+            padding: 0;
+        }
+        .license-used-by {
+            margin-top: -10px;
+        }
+        .license-text {
+            max-height: 200px;
+            overflow-y: scroll;
+            white-space: pre-wrap;
+        }
+    </style>
+</head>
+
+<body>
+    <main class="container">
+        <div class="intro">
+            <h1>Third Party Licenses</h1>
+            <p>This page lists the licenses of the projects used in cargo-about.</p>
+        </div>
+    
+        <h2>Overview of licenses:</h2>
+        <ul class="licenses-overview">
+            {{#each overview}}
+            <li><a href="#{{id}}">{{name}}</a> ({{count}})</li>
+            {{/each}}
+        </ul>
+
+        <h2>All license text:</h2>
+        <ul class="licenses-list">
+            {{#each licenses}}
+            <li class="license">
+                <h3 id="{{id}}">{{name}}</h3>
+                <h4>Used by:</h4>
+                <ul class="license-used-by">
+                    {{#each used_by}}
+                    <li><a href="{{#if crate.repository}} {{crate.repository}} {{else}} https://crates.io/crates/{{crate.name}} {{/if}}">{{crate.name}} {{crate.version}}</a></li>
+                    {{/each}}
+                </ul>
+                <pre class="license-text">{{text}}</pre>
+            </li>
+            {{/each}}
+        </ul>
+    </main>
+</body>
+
+</html>

about.toml

@@ -0,0 +1,18 @@
+accepted = [
+    "0BSD",
+    "Apache-2.0",
+    "Apache-2.0 WITH LLVM-exception",
+    "BSD-2-Clause",
+    "BSD-3-Clause",
+    "BSL-1.0",
+    "bzip2-1.0.6",
+    "CC0-1.0",
+    "CDDL-1.0",
+    "CDLA-Permissive-2.0",
+    "ISC",
+    "MIT",
+    "MPL-2.0",
+    "OpenSSL",
+    "Unicode-3.0",
+    "Zlib",
+]

docs/src/java/java.md

@@ -14,7 +14,7 @@ Add the following dependency to your `pom.xml`:
 <dependency>
     <groupId>com.lancedb</groupId>
     <artifactId>lancedb-core</artifactId>
-    <version>0.26.1</version>
+    <version>0.27.0-beta.0</version>
 </dependency>
 ```
 

java/JAVA_THIRD_PARTY_LICENSES.md

@@ -0,0 +1,71 @@
+
+List of third-party dependencies grouped by their license type.
+
+    Apache 2.0:
+
+        * error-prone annotations (com.google.errorprone:error_prone_annotations:2.28.0 - https://errorprone.info/error_prone_annotations)
+
+    Apache License 2.0:
+
+        * JsonNullable Jackson module (org.openapitools:jackson-databind-nullable:0.2.6 - https://github.com/OpenAPITools/jackson-databind-nullable)
+
+    Apache License V2.0:
+
+        * FlatBuffers Java API (com.google.flatbuffers:flatbuffers-java:23.5.26 - https://github.com/google/flatbuffers)
+
+    Apache License, Version 2.0:
+
+        * Apache Commons Codec (commons-codec:commons-codec:1.15 - https://commons.apache.org/proper/commons-codec/)
+        * Apache HttpClient (org.apache.httpcomponents.client5:httpclient5:5.2.1 - https://hc.apache.org/httpcomponents-client-5.0.x/5.2.1/httpclient5/)
+        * Apache HttpComponents Core HTTP/1.1 (org.apache.httpcomponents.core5:httpcore5:5.2 - https://hc.apache.org/httpcomponents-core-5.2.x/5.2/httpcore5/)
+        * Apache HttpComponents Core HTTP/2 (org.apache.httpcomponents.core5:httpcore5-h2:5.2 - https://hc.apache.org/httpcomponents-core-5.2.x/5.2/httpcore5-h2/)
+        * Arrow Format (org.apache.arrow:arrow-format:15.0.0 - https://arrow.apache.org/arrow-format/)
+        * Arrow Java C Data Interface (org.apache.arrow:arrow-c-data:15.0.0 - https://arrow.apache.org/arrow-c-data/)
+        * Arrow Java Dataset (org.apache.arrow:arrow-dataset:15.0.0 - https://arrow.apache.org/arrow-dataset/)
+        * Arrow Memory - Core (org.apache.arrow:arrow-memory-core:15.0.0 - https://arrow.apache.org/arrow-memory/arrow-memory-core/)
+        * Arrow Memory - Netty (org.apache.arrow:arrow-memory-netty:15.0.0 - https://arrow.apache.org/arrow-memory/arrow-memory-netty/)
+        * Arrow Vectors (org.apache.arrow:arrow-vector:15.0.0 - https://arrow.apache.org/arrow-vector/)
+        * Guava: Google Core Libraries for Java (com.google.guava:guava:33.3.1-jre - https://github.com/google/guava)
+        * J2ObjC Annotations (com.google.j2objc:j2objc-annotations:3.0.0 - https://github.com/google/j2objc/)
+        * Netty/Buffer (io.netty:netty-buffer:4.1.104.Final - https://netty.io/netty-buffer/)
+        * Netty/Common (io.netty:netty-common:4.1.104.Final - https://netty.io/netty-common/)
+
+    Apache-2.0:
+
+        * Apache Commons Lang (org.apache.commons:commons-lang3:3.18.0 - https://commons.apache.org/proper/commons-lang/)
+        * lance-namespace-apache-client (org.lance:lance-namespace-apache-client:0.4.5 - https://github.com/openapitools/openapi-generator)
+        * lance-namespace-core (org.lance:lance-namespace-core:0.4.5 - https://lance.org/format/namespace/lance-namespace-core/)
+
+    EDL 1.0:
+
+        * Jakarta Activation API jar (jakarta.activation:jakarta.activation-api:1.2.2 - https://github.com/eclipse-ee4j/jaf/jakarta.activation-api)
+
+    Eclipse Distribution License - v 1.0:
+
+        * Eclipse Collections API (org.eclipse.collections:eclipse-collections-api:11.1.0 - https://github.com/eclipse/eclipse-collections/eclipse-collections-api)
+        * Eclipse Collections Main Library (org.eclipse.collections:eclipse-collections:11.1.0 - https://github.com/eclipse/eclipse-collections/eclipse-collections)
+        * Jakarta XML Binding API (jakarta.xml.bind:jakarta.xml.bind-api:2.3.3 - https://github.com/eclipse-ee4j/jaxb-api/jakarta.xml.bind-api)
+
+    Eclipse Public License - v 1.0:
+
+        * Eclipse Collections API (org.eclipse.collections:eclipse-collections-api:11.1.0 - https://github.com/eclipse/eclipse-collections/eclipse-collections-api)
+        * Eclipse Collections Main Library (org.eclipse.collections:eclipse-collections:11.1.0 - https://github.com/eclipse/eclipse-collections/eclipse-collections)
+
+    The Apache Software License, Version 2.0:
+
+        * FindBugs-jsr305 (com.google.code.findbugs:jsr305:3.0.2 - http://findbugs.sourceforge.net/)
+        * Guava InternalFutureFailureAccess and InternalFutures (com.google.guava:failureaccess:1.0.2 - https://github.com/google/guava/failureaccess)
+        * Guava ListenableFuture only (com.google.guava:listenablefuture:9999.0-empty-to-avoid-conflict-with-guava - https://github.com/google/guava/listenablefuture)
+        * Jackson datatype: JSR310 (com.fasterxml.jackson.datatype:jackson-datatype-jsr310:2.16.0 - https://github.com/FasterXML/jackson-modules-java8/jackson-datatype-jsr310)
+        * Jackson module: Old JAXB Annotations (javax.xml.bind) (com.fasterxml.jackson.module:jackson-module-jaxb-annotations:2.17.1 - https://github.com/FasterXML/jackson-modules-base)
+        * Jackson-annotations (com.fasterxml.jackson.core:jackson-annotations:2.16.0 - https://github.com/FasterXML/jackson)
+        * Jackson-core (com.fasterxml.jackson.core:jackson-core:2.16.0 - https://github.com/FasterXML/jackson-core)
+        * jackson-databind (com.fasterxml.jackson.core:jackson-databind:2.15.2 - https://github.com/FasterXML/jackson)
+        * Jackson-JAXRS: base (com.fasterxml.jackson.jaxrs:jackson-jaxrs-base:2.17.1 - https://github.com/FasterXML/jackson-jaxrs-providers/jackson-jaxrs-base)
+        * Jackson-JAXRS: JSON (com.fasterxml.jackson.jaxrs:jackson-jaxrs-json-provider:2.17.1 - https://github.com/FasterXML/jackson-jaxrs-providers/jackson-jaxrs-json-provider)
+        * JAR JNI Loader (org.questdb:jar-jni:1.1.1 - https://github.com/questdb/rust-maven-plugin)
+        * Lance Core (org.lance:lance-core:2.0.0 - https://lance.org/)
+
+    The MIT License:
+
+        * Checker Qual (org.checkerframework:checker-qual:3.43.0 - https://checkerframework.org/)

java/lancedb-core/JAVA_THIRD_PARTY_LICENSES.md

@@ -0,0 +1,71 @@
+
+List of third-party dependencies grouped by their license type.
+
+    Apache 2.0:
+
+        * error-prone annotations (com.google.errorprone:error_prone_annotations:2.28.0 - https://errorprone.info/error_prone_annotations)
+
+    Apache License 2.0:
+
+        * JsonNullable Jackson module (org.openapitools:jackson-databind-nullable:0.2.6 - https://github.com/OpenAPITools/jackson-databind-nullable)
+
+    Apache License V2.0:
+
+        * FlatBuffers Java API (com.google.flatbuffers:flatbuffers-java:23.5.26 - https://github.com/google/flatbuffers)
+
+    Apache License, Version 2.0:
+
+        * Apache Commons Codec (commons-codec:commons-codec:1.15 - https://commons.apache.org/proper/commons-codec/)
+        * Apache HttpClient (org.apache.httpcomponents.client5:httpclient5:5.2.1 - https://hc.apache.org/httpcomponents-client-5.0.x/5.2.1/httpclient5/)
+        * Apache HttpComponents Core HTTP/1.1 (org.apache.httpcomponents.core5:httpcore5:5.2 - https://hc.apache.org/httpcomponents-core-5.2.x/5.2/httpcore5/)
+        * Apache HttpComponents Core HTTP/2 (org.apache.httpcomponents.core5:httpcore5-h2:5.2 - https://hc.apache.org/httpcomponents-core-5.2.x/5.2/httpcore5-h2/)
+        * Arrow Format (org.apache.arrow:arrow-format:15.0.0 - https://arrow.apache.org/arrow-format/)
+        * Arrow Java C Data Interface (org.apache.arrow:arrow-c-data:15.0.0 - https://arrow.apache.org/arrow-c-data/)
+        * Arrow Java Dataset (org.apache.arrow:arrow-dataset:15.0.0 - https://arrow.apache.org/arrow-dataset/)
+        * Arrow Memory - Core (org.apache.arrow:arrow-memory-core:15.0.0 - https://arrow.apache.org/arrow-memory/arrow-memory-core/)
+        * Arrow Memory - Netty (org.apache.arrow:arrow-memory-netty:15.0.0 - https://arrow.apache.org/arrow-memory/arrow-memory-netty/)
+        * Arrow Vectors (org.apache.arrow:arrow-vector:15.0.0 - https://arrow.apache.org/arrow-vector/)
+        * Guava: Google Core Libraries for Java (com.google.guava:guava:33.3.1-jre - https://github.com/google/guava)
+        * J2ObjC Annotations (com.google.j2objc:j2objc-annotations:3.0.0 - https://github.com/google/j2objc/)
+        * Netty/Buffer (io.netty:netty-buffer:4.1.104.Final - https://netty.io/netty-buffer/)
+        * Netty/Common (io.netty:netty-common:4.1.104.Final - https://netty.io/netty-common/)
+
+    Apache-2.0:
+
+        * Apache Commons Lang (org.apache.commons:commons-lang3:3.18.0 - https://commons.apache.org/proper/commons-lang/)
+        * lance-namespace-apache-client (org.lance:lance-namespace-apache-client:0.4.5 - https://github.com/openapitools/openapi-generator)
+        * lance-namespace-core (org.lance:lance-namespace-core:0.4.5 - https://lance.org/format/namespace/lance-namespace-core/)
+
+    EDL 1.0:
+
+        * Jakarta Activation API jar (jakarta.activation:jakarta.activation-api:1.2.2 - https://github.com/eclipse-ee4j/jaf/jakarta.activation-api)
+
+    Eclipse Distribution License - v 1.0:
+
+        * Eclipse Collections API (org.eclipse.collections:eclipse-collections-api:11.1.0 - https://github.com/eclipse/eclipse-collections/eclipse-collections-api)
+        * Eclipse Collections Main Library (org.eclipse.collections:eclipse-collections:11.1.0 - https://github.com/eclipse/eclipse-collections/eclipse-collections)
+        * Jakarta XML Binding API (jakarta.xml.bind:jakarta.xml.bind-api:2.3.3 - https://github.com/eclipse-ee4j/jaxb-api/jakarta.xml.bind-api)
+
+    Eclipse Public License - v 1.0:
+
+        * Eclipse Collections API (org.eclipse.collections:eclipse-collections-api:11.1.0 - https://github.com/eclipse/eclipse-collections/eclipse-collections-api)
+        * Eclipse Collections Main Library (org.eclipse.collections:eclipse-collections:11.1.0 - https://github.com/eclipse/eclipse-collections/eclipse-collections)
+
+    The Apache Software License, Version 2.0:
+
+        * FindBugs-jsr305 (com.google.code.findbugs:jsr305:3.0.2 - http://findbugs.sourceforge.net/)
+        * Guava InternalFutureFailureAccess and InternalFutures (com.google.guava:failureaccess:1.0.2 - https://github.com/google/guava/failureaccess)
+        * Guava ListenableFuture only (com.google.guava:listenablefuture:9999.0-empty-to-avoid-conflict-with-guava - https://github.com/google/guava/listenablefuture)
+        * Jackson datatype: JSR310 (com.fasterxml.jackson.datatype:jackson-datatype-jsr310:2.16.0 - https://github.com/FasterXML/jackson-modules-java8/jackson-datatype-jsr310)
+        * Jackson module: Old JAXB Annotations (javax.xml.bind) (com.fasterxml.jackson.module:jackson-module-jaxb-annotations:2.17.1 - https://github.com/FasterXML/jackson-modules-base)
+        * Jackson-annotations (com.fasterxml.jackson.core:jackson-annotations:2.16.0 - https://github.com/FasterXML/jackson)
+        * Jackson-core (com.fasterxml.jackson.core:jackson-core:2.16.0 - https://github.com/FasterXML/jackson-core)
+        * jackson-databind (com.fasterxml.jackson.core:jackson-databind:2.15.2 - https://github.com/FasterXML/jackson)
+        * Jackson-JAXRS: base (com.fasterxml.jackson.jaxrs:jackson-jaxrs-base:2.17.1 - https://github.com/FasterXML/jackson-jaxrs-providers/jackson-jaxrs-base)
+        * Jackson-JAXRS: JSON (com.fasterxml.jackson.jaxrs:jackson-jaxrs-json-provider:2.17.1 - https://github.com/FasterXML/jackson-jaxrs-providers/jackson-jaxrs-json-provider)
+        * JAR JNI Loader (org.questdb:jar-jni:1.1.1 - https://github.com/questdb/rust-maven-plugin)
+        * Lance Core (org.lance:lance-core:2.0.0 - https://lance.org/)
+
+    The MIT License:
+
+        * Checker Qual (org.checkerframework:checker-qual:3.43.0 - https://checkerframework.org/)

java/lancedb-core/pom.xml

@@ -8,7 +8,7 @@
     <parent>
       <groupId>com.lancedb</groupId>
       <artifactId>lancedb-parent</artifactId>
-      <version>0.26.1-final.0</version>
+      <version>0.27.0-beta.0</version>
       <relativePath>../pom.xml</relativePath>
     </parent>
 

java/pom.xml

@@ -6,7 +6,7 @@
 
     <groupId>com.lancedb</groupId>
     <artifactId>lancedb-parent</artifactId>
-    <version>0.26.1-final.0</version>
+    <version>0.27.0-beta.0</version>
     <packaging>pom</packaging>
     <name>${project.artifactId}</name>
     <description>LanceDB Java SDK Parent POM</description>
@@ -28,7 +28,7 @@
     <properties>
         <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
         <arrow.version>15.0.0</arrow.version>
-        <lance-core.version>2.0.0</lance-core.version>
+        <lance-core.version>2.0.1</lance-core.version>
         <spotless.skip>false</spotless.skip>
         <spotless.version>2.30.0</spotless.version>
         <spotless.java.googlejavaformat.version>1.7</spotless.java.googlejavaformat.version>
@@ -160,6 +160,19 @@
                 <groupId>com.diffplug.spotless</groupId>
                 <artifactId>spotless-maven-plugin</artifactId>
             </plugin>
+            <plugin>
+                <groupId>org.codehaus.mojo</groupId>
+                <artifactId>license-maven-plugin</artifactId>
+                <version>2.4.0</version>
+                <configuration>
+                    <outputDirectory>${project.basedir}</outputDirectory>
+                    <thirdPartyFilename>JAVA_THIRD_PARTY_LICENSES.md</thirdPartyFilename>
+                    <fileTemplate>/org/codehaus/mojo/license/third-party-file-groupByLicense.ftl</fileTemplate>
+                    <includedScopes>compile,runtime</includedScopes>
+                    <excludedScopes>test,provided</excludedScopes>
+                    <sortArtifactByName>true</sortArtifactByName>
+                </configuration>
+            </plugin>
         </plugins>
         <pluginManagement>
             <plugins>

nodejs/Cargo.toml

@@ -1,7 +1,7 @@
 [package]
 name = "lancedb-nodejs"
 edition.workspace = true
-version = "0.26.1"
+version = "0.27.0-beta.0"
 license.workspace = true
 description.workspace = true
 repository.workspace = true

nodejs/NODEJS_THIRD_PARTY_LICENSES.md

@@ -0,0 +1,668 @@
+[@75lb/deep-merge@1.1.2](https://github.com/75lb/deep-merge) - MIT
+[@aashutoshrathi/word-wrap@1.2.6](https://github.com/aashutoshrathi/word-wrap) - MIT
+[@ampproject/remapping@2.2.1](https://github.com/ampproject/remapping) - Apache-2.0
+[@aws-crypto/crc32@3.0.0](https://github.com/aws/aws-sdk-js-crypto-helpers) - Apache-2.0
+[@aws-crypto/crc32c@3.0.0](https://github.com/aws/aws-sdk-js-crypto-helpers) - Apache-2.0
+[@aws-crypto/ie11-detection@3.0.0](https://github.com/aws/aws-sdk-js-crypto-helpers) - Apache-2.0
+[@aws-crypto/sha1-browser@3.0.0](https://github.com/aws/aws-sdk-js-crypto-helpers) - Apache-2.0
+[@aws-crypto/sha256-browser@3.0.0](https://github.com/aws/aws-sdk-js-crypto-helpers) - Apache-2.0
+[@aws-crypto/sha256-browser@5.2.0](https://github.com/aws/aws-sdk-js-crypto-helpers) - Apache-2.0
+[@aws-crypto/sha256-js@3.0.0](https://github.com/aws/aws-sdk-js-crypto-helpers) - Apache-2.0
+[@aws-crypto/sha256-js@5.2.0](https://github.com/aws/aws-sdk-js-crypto-helpers) - Apache-2.0
+[@aws-crypto/supports-web-crypto@3.0.0](https://github.com/aws/aws-sdk-js-crypto-helpers) - Apache-2.0
+[@aws-crypto/supports-web-crypto@5.2.0](https://github.com/aws/aws-sdk-js-crypto-helpers) - Apache-2.0
+[@aws-crypto/util@3.0.0](https://github.com/aws/aws-sdk-js-crypto-helpers) - Apache-2.0
+[@aws-crypto/util@5.2.0](https://github.com/aws/aws-sdk-js-crypto-helpers) - Apache-2.0
+[@aws-sdk/client-dynamodb@3.602.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/client-kms@3.549.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/client-s3@3.550.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/client-sso-oidc@3.549.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/client-sso-oidc@3.600.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/client-sso@3.549.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/client-sso@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/client-sts@3.549.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/client-sts@3.600.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/core@3.549.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/core@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/credential-provider-env@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/credential-provider-env@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/credential-provider-http@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/credential-provider-http@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/credential-provider-ini@3.549.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/credential-provider-ini@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/credential-provider-node@3.549.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/credential-provider-node@3.600.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/credential-provider-process@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/credential-provider-process@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/credential-provider-sso@3.549.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/credential-provider-sso@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/credential-provider-web-identity@3.549.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/credential-provider-web-identity@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/endpoint-cache@3.572.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/middleware-bucket-endpoint@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/middleware-endpoint-discovery@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/middleware-expect-continue@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/middleware-flexible-checksums@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/middleware-host-header@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/middleware-host-header@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/middleware-location-constraint@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/middleware-logger@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/middleware-logger@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/middleware-recursion-detection@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/middleware-recursion-detection@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/middleware-sdk-s3@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/middleware-signing@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/middleware-ssec@3.537.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/middleware-user-agent@3.540.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/middleware-user-agent@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/region-config-resolver@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/region-config-resolver@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/signature-v4-multi-region@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/token-providers@3.549.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/token-providers@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/types@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/types@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/util-arn-parser@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/util-endpoints@3.540.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/util-endpoints@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/util-locate-window@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/util-user-agent-browser@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/util-user-agent-browser@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/util-user-agent-node@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/util-user-agent-node@3.598.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/util-utf8-browser@3.259.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@aws-sdk/xml-builder@3.535.0](https://github.com/aws/aws-sdk-js-v3) - Apache-2.0
+[@babel/code-frame@7.26.2](https://github.com/babel/babel) - MIT
+[@babel/compat-data@7.23.5](https://github.com/babel/babel) - MIT
+[@babel/core@7.23.7](https://github.com/babel/babel) - MIT
+[@babel/generator@7.23.6](https://github.com/babel/babel) - MIT
+[@babel/helper-compilation-targets@7.23.6](https://github.com/babel/babel) - MIT
+[@babel/helper-environment-visitor@7.22.20](https://github.com/babel/babel) - MIT
+[@babel/helper-function-name@7.23.0](https://github.com/babel/babel) - MIT
+[@babel/helper-hoist-variables@7.22.5](https://github.com/babel/babel) - MIT
+[@babel/helper-module-imports@7.22.15](https://github.com/babel/babel) - MIT
+[@babel/helper-module-transforms@7.23.3](https://github.com/babel/babel) - MIT
+[@babel/helper-plugin-utils@7.22.5](https://github.com/babel/babel) - MIT
+[@babel/helper-simple-access@7.22.5](https://github.com/babel/babel) - MIT
+[@babel/helper-split-export-declaration@7.22.6](https://github.com/babel/babel) - MIT
+[@babel/helper-string-parser@7.25.9](https://github.com/babel/babel) - MIT
+[@babel/helper-validator-identifier@7.25.9](https://github.com/babel/babel) - MIT
+[@babel/helper-validator-option@7.23.5](https://github.com/babel/babel) - MIT
+[@babel/helpers@7.27.0](https://github.com/babel/babel) - MIT
+[@babel/parser@7.27.0](https://github.com/babel/babel) - MIT
+[@babel/plugin-syntax-async-generators@7.8.4](https://github.com/babel/babel/tree/master/packages/babel-plugin-syntax-async-generators) - MIT
+[@babel/plugin-syntax-bigint@7.8.3](https://github.com/babel/babel/tree/master/packages/babel-plugin-syntax-bigint) - MIT
+[@babel/plugin-syntax-class-properties@7.12.13](https://github.com/babel/babel) - MIT
+[@babel/plugin-syntax-import-meta@7.10.4](https://github.com/babel/babel) - MIT
+[@babel/plugin-syntax-json-strings@7.8.3](https://github.com/babel/babel/tree/master/packages/babel-plugin-syntax-json-strings) - MIT
+[@babel/plugin-syntax-jsx@7.23.3](https://github.com/babel/babel) - MIT
+[@babel/plugin-syntax-logical-assignment-operators@7.10.4](https://github.com/babel/babel) - MIT
+[@babel/plugin-syntax-nullish-coalescing-operator@7.8.3](https://github.com/babel/babel/tree/master/packages/babel-plugin-syntax-nullish-coalescing-operator) - MIT
+[@babel/plugin-syntax-numeric-separator@7.10.4](https://github.com/babel/babel) - MIT
+[@babel/plugin-syntax-object-rest-spread@7.8.3](https://github.com/babel/babel/tree/master/packages/babel-plugin-syntax-object-rest-spread) - MIT
+[@babel/plugin-syntax-optional-catch-binding@7.8.3](https://github.com/babel/babel/tree/master/packages/babel-plugin-syntax-optional-catch-binding) - MIT
+[@babel/plugin-syntax-optional-chaining@7.8.3](https://github.com/babel/babel/tree/master/packages/babel-plugin-syntax-optional-chaining) - MIT
+[@babel/plugin-syntax-top-level-await@7.14.5](https://github.com/babel/babel) - MIT
+[@babel/plugin-syntax-typescript@7.23.3](https://github.com/babel/babel) - MIT
+[@babel/template@7.27.0](https://github.com/babel/babel) - MIT
+[@babel/traverse@7.23.7](https://github.com/babel/babel) - MIT
+[@babel/types@7.27.0](https://github.com/babel/babel) - MIT
+[@bcoe/v8-coverage@0.2.3](https://github.com/demurgos/v8-coverage) - MIT
+[@biomejs/biome@1.8.3](https://github.com/biomejs/biome) - MIT OR Apache-2.0
+[@biomejs/cli-darwin-arm64@1.8.3](https://github.com/biomejs/biome) - MIT OR Apache-2.0
+[@eslint-community/eslint-utils@4.4.0](https://github.com/eslint-community/eslint-utils) - MIT
+[@eslint-community/regexpp@4.10.0](https://github.com/eslint-community/regexpp) - MIT
+[@eslint/eslintrc@2.1.4](https://github.com/eslint/eslintrc) - MIT
+[@eslint/js@8.57.0](https://github.com/eslint/eslint) - MIT
+[@huggingface/jinja@0.3.2](https://github.com/huggingface/huggingface.js) - MIT
+[@huggingface/transformers@3.0.2](https://github.com/huggingface/transformers.js) - Apache-2.0
+[@humanwhocodes/config-array@0.11.14](https://github.com/humanwhocodes/config-array) - Apache-2.0
+[@humanwhocodes/module-importer@1.0.1](https://github.com/humanwhocodes/module-importer) - Apache-2.0
+[@humanwhocodes/object-schema@2.0.2](https://github.com/humanwhocodes/object-schema) - BSD-3-Clause
+[@img/sharp-darwin-arm64@0.33.5](https://github.com/lovell/sharp) - Apache-2.0
+[@img/sharp-libvips-darwin-arm64@1.0.4](https://github.com/lovell/sharp-libvips) - LGPL-3.0-or-later
+[@isaacs/cliui@8.0.2](https://github.com/yargs/cliui) - ISC
+[@isaacs/fs-minipass@4.0.1](https://github.com/npm/fs-minipass) - ISC
+[@istanbuljs/load-nyc-config@1.1.0](https://github.com/istanbuljs/load-nyc-config) - ISC
+[@istanbuljs/schema@0.1.3](https://github.com/istanbuljs/schema) - MIT
+[@jest/console@29.7.0](https://github.com/jestjs/jest) - MIT
+[@jest/core@29.7.0](https://github.com/jestjs/jest) - MIT
+[@jest/environment@29.7.0](https://github.com/jestjs/jest) - MIT
+[@jest/expect-utils@29.7.0](https://github.com/jestjs/jest) - MIT
+[@jest/expect@29.7.0](https://github.com/jestjs/jest) - MIT
+[@jest/fake-timers@29.7.0](https://github.com/jestjs/jest) - MIT
+[@jest/globals@29.7.0](https://github.com/jestjs/jest) - MIT
+[@jest/reporters@29.7.0](https://github.com/jestjs/jest) - MIT
+[@jest/schemas@29.6.3](https://github.com/jestjs/jest) - MIT
+[@jest/source-map@29.6.3](https://github.com/jestjs/jest) - MIT
+[@jest/test-result@29.7.0](https://github.com/jestjs/jest) - MIT
+[@jest/test-sequencer@29.7.0](https://github.com/jestjs/jest) - MIT
+[@jest/transform@29.7.0](https://github.com/jestjs/jest) - MIT
+[@jest/types@29.6.3](https://github.com/jestjs/jest) - MIT
+[@jridgewell/gen-mapping@0.3.3](https://github.com/jridgewell/gen-mapping) - MIT
+[@jridgewell/resolve-uri@3.1.1](https://github.com/jridgewell/resolve-uri) - MIT
+[@jridgewell/set-array@1.1.2](https://github.com/jridgewell/set-array) - MIT
+[@jridgewell/sourcemap-codec@1.4.15](https://github.com/jridgewell/sourcemap-codec) - MIT
+[@jridgewell/trace-mapping@0.3.22](https://github.com/jridgewell/trace-mapping) - MIT
+[@lancedb/lancedb@0.26.2](https://github.com/lancedb/lancedb) - Apache-2.0
+[@napi-rs/cli@2.18.3](https://github.com/napi-rs/napi-rs) - MIT
+[@nodelib/fs.scandir@2.1.5](https://github.com/nodelib/nodelib/tree/master/packages/fs/fs.scandir) - MIT
+[@nodelib/fs.stat@2.0.5](https://github.com/nodelib/nodelib/tree/master/packages/fs/fs.stat) - MIT
+[@nodelib/fs.walk@1.2.8](https://github.com/nodelib/nodelib/tree/master/packages/fs/fs.walk) - MIT
+[@pkgjs/parseargs@0.11.0](https://github.com/pkgjs/parseargs) - MIT
+[@protobufjs/aspromise@1.1.2](https://github.com/dcodeIO/protobuf.js) - BSD-3-Clause
+[@protobufjs/base64@1.1.2](https://github.com/dcodeIO/protobuf.js) - BSD-3-Clause
+[@protobufjs/codegen@2.0.4](https://github.com/dcodeIO/protobuf.js) - BSD-3-Clause
+[@protobufjs/eventemitter@1.1.0](https://github.com/dcodeIO/protobuf.js) - BSD-3-Clause
+[@protobufjs/fetch@1.1.0](https://github.com/dcodeIO/protobuf.js) - BSD-3-Clause
+[@protobufjs/float@1.0.2](https://github.com/dcodeIO/protobuf.js) - BSD-3-Clause
+[@protobufjs/inquire@1.1.0](https://github.com/dcodeIO/protobuf.js) - BSD-3-Clause
+[@protobufjs/path@1.1.2](https://github.com/dcodeIO/protobuf.js) - BSD-3-Clause
+[@protobufjs/pool@1.1.0](https://github.com/dcodeIO/protobuf.js) - BSD-3-Clause
+[@protobufjs/utf8@1.1.0](https://github.com/dcodeIO/protobuf.js) - BSD-3-Clause
+[@shikijs/core@1.10.3](https://github.com/shikijs/shiki) - MIT
+[@sinclair/typebox@0.27.8](https://github.com/sinclairzx81/typebox) - MIT
+[@sinonjs/commons@3.0.1](https://github.com/sinonjs/commons) - BSD-3-Clause
+[@sinonjs/fake-timers@10.3.0](https://github.com/sinonjs/fake-timers) - BSD-3-Clause
+[@smithy/abort-controller@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/abort-controller@3.1.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/chunked-blob-reader-native@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/chunked-blob-reader@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/config-resolver@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/config-resolver@3.0.3](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/core@1.4.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/core@2.2.3](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/credential-provider-imds@2.3.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/credential-provider-imds@3.1.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/eventstream-codec@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/eventstream-serde-browser@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/eventstream-serde-config-resolver@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/eventstream-serde-node@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/eventstream-serde-universal@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/fetch-http-handler@2.5.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/fetch-http-handler@3.1.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/hash-blob-browser@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/hash-node@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/hash-node@3.0.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/hash-stream-node@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/invalid-dependency@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/invalid-dependency@3.0.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/is-array-buffer@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/is-array-buffer@3.0.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/md5-js@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/middleware-content-length@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/middleware-content-length@3.0.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/middleware-endpoint@2.5.1](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/middleware-endpoint@3.0.3](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/middleware-retry@2.3.1](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/middleware-retry@3.0.6](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/middleware-serde@2.3.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/middleware-serde@3.0.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/middleware-stack@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/middleware-stack@3.0.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/node-config-provider@2.3.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/node-config-provider@3.1.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/node-http-handler@2.5.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/node-http-handler@3.1.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/property-provider@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/property-provider@3.1.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/protocol-http@3.3.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/protocol-http@4.0.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/querystring-builder@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/querystring-builder@3.0.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/querystring-parser@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/querystring-parser@3.0.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/service-error-classification@2.1.5](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/service-error-classification@3.0.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/shared-ini-file-loader@2.4.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/shared-ini-file-loader@3.1.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/signature-v4@2.2.1](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/signature-v4@3.1.1](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/smithy-client@2.5.1](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/smithy-client@3.1.4](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/types@2.12.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/types@3.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/url-parser@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/url-parser@3.0.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-base64@2.3.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-base64@3.0.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-body-length-browser@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-body-length-browser@3.0.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-body-length-node@2.3.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-body-length-node@3.0.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-buffer-from@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-buffer-from@3.0.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-config-provider@2.3.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-config-provider@3.0.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-defaults-mode-browser@2.2.1](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-defaults-mode-browser@3.0.6](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-defaults-mode-node@2.3.1](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-defaults-mode-node@3.0.6](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-endpoints@1.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-endpoints@2.0.3](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-hex-encoding@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-hex-encoding@3.0.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-middleware@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-middleware@3.0.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-retry@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-retry@3.0.2](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-stream@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-stream@3.0.4](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-uri-escape@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-uri-escape@3.0.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-utf8@2.3.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-utf8@3.0.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-waiter@2.2.0](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@smithy/util-waiter@3.1.1](https://github.com/awslabs/smithy-typescript) - Apache-2.0
+[@swc/helpers@0.5.12](https://github.com/swc-project/swc) - Apache-2.0
+[@types/axios@0.14.0](https://github.com/mzabriskie/axios) - MIT
+[@types/babel__core@7.20.5](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/babel__generator@7.6.8](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/babel__template@7.4.4](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/babel__traverse@7.20.5](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/command-line-args@5.2.3](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/command-line-usage@5.0.2](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/command-line-usage@5.0.4](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/graceful-fs@4.1.9](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/hast@3.0.4](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/istanbul-lib-coverage@2.0.6](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/istanbul-lib-report@3.0.3](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/istanbul-reports@3.0.4](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/jest@29.5.12](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/json-schema@7.0.15](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/node-fetch@2.6.11](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/node@18.19.26](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/node@20.16.10](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/node@20.17.9](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/node@22.7.4](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/semver@7.5.6](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/stack-utils@2.0.3](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/tmp@0.2.6](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/unist@3.0.2](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/yargs-parser@21.0.3](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@types/yargs@17.0.32](https://github.com/DefinitelyTyped/DefinitelyTyped) - MIT
+[@typescript-eslint/eslint-plugin@7.1.0](https://github.com/typescript-eslint/typescript-eslint) - MIT
+[@typescript-eslint/parser@7.1.0](https://github.com/typescript-eslint/typescript-eslint) - BSD-2-Clause
+[@typescript-eslint/scope-manager@7.1.0](https://github.com/typescript-eslint/typescript-eslint) - MIT
+[@typescript-eslint/type-utils@7.1.0](https://github.com/typescript-eslint/typescript-eslint) - MIT
+[@typescript-eslint/types@7.1.0](https://github.com/typescript-eslint/typescript-eslint) - MIT
+[@typescript-eslint/typescript-estree@7.1.0](https://github.com/typescript-eslint/typescript-eslint) - BSD-2-Clause
+[@typescript-eslint/utils@7.1.0](https://github.com/typescript-eslint/typescript-eslint) - MIT
+[@typescript-eslint/visitor-keys@7.1.0](https://github.com/typescript-eslint/typescript-eslint) - MIT
+[@ungap/structured-clone@1.2.0](https://github.com/ungap/structured-clone) - ISC
+[abort-controller@3.0.0](https://github.com/mysticatea/abort-controller) - MIT
+[acorn-jsx@5.3.2](https://github.com/acornjs/acorn-jsx) - MIT
+[acorn@8.11.3](https://github.com/acornjs/acorn) - MIT
+[agentkeepalive@4.5.0](https://github.com/node-modules/agentkeepalive) - MIT
+[ajv@6.12.6](https://github.com/ajv-validator/ajv) - MIT
+[ansi-escapes@4.3.2](https://github.com/sindresorhus/ansi-escapes) - MIT
+[ansi-regex@5.0.1](https://github.com/chalk/ansi-regex) - MIT
+[ansi-regex@6.1.0](https://github.com/chalk/ansi-regex) - MIT
+[ansi-styles@4.3.0](https://github.com/chalk/ansi-styles) - MIT
+[ansi-styles@5.2.0](https://github.com/chalk/ansi-styles) - MIT
+[ansi-styles@6.2.1](https://github.com/chalk/ansi-styles) - MIT
+[anymatch@3.1.3](https://github.com/micromatch/anymatch) - ISC
+[apache-arrow@15.0.0](https://github.com/apache/arrow) - Apache-2.0
+[apache-arrow@16.0.0](https://github.com/apache/arrow) - Apache-2.0
+[apache-arrow@17.0.0](https://github.com/apache/arrow) - Apache-2.0
+[apache-arrow@18.0.0](https://github.com/apache/arrow) - Apache-2.0
+[argparse@1.0.10](https://github.com/nodeca/argparse) - MIT
+[argparse@2.0.1](https://github.com/nodeca/argparse) - Python-2.0
+[array-back@3.1.0](https://github.com/75lb/array-back) - MIT
+[array-back@6.2.2](https://github.com/75lb/array-back) - MIT
+[array-union@2.1.0](https://github.com/sindresorhus/array-union) - MIT
+[asynckit@0.4.0](https://github.com/alexindigo/asynckit) - MIT
+[axios@1.8.4](https://github.com/axios/axios) - MIT
+[babel-jest@29.7.0](https://github.com/jestjs/jest) - MIT
+[babel-plugin-istanbul@6.1.1](https://github.com/istanbuljs/babel-plugin-istanbul) - BSD-3-Clause
+[babel-plugin-jest-hoist@29.6.3](https://github.com/jestjs/jest) - MIT
+[babel-preset-current-node-syntax@1.0.1](https://github.com/nicolo-ribaudo/babel-preset-current-node-syntax) - MIT
+[babel-preset-jest@29.6.3](https://github.com/jestjs/jest) - MIT
+[balanced-match@1.0.2](https://github.com/juliangruber/balanced-match) - MIT
+[base-64@0.1.0](https://github.com/mathiasbynens/base64) - MIT
+[bowser@2.11.0](https://github.com/lancedikson/bowser) - MIT
+[brace-expansion@1.1.11](https://github.com/juliangruber/brace-expansion) - MIT
+[brace-expansion@2.0.1](https://github.com/juliangruber/brace-expansion) - MIT
+[braces@3.0.3](https://github.com/micromatch/braces) - MIT
+[browserslist@4.22.2](https://github.com/browserslist/browserslist) - MIT
+[bs-logger@0.2.6](https://github.com/huafu/bs-logger) - MIT
+[bser@2.1.1](https://github.com/facebook/watchman) - Apache-2.0
+[buffer-from@1.1.2](https://github.com/LinusU/buffer-from) - MIT
+[callsites@3.1.0](https://github.com/sindresorhus/callsites) - MIT
+[camelcase@5.3.1](https://github.com/sindresorhus/camelcase) - MIT
+[camelcase@6.3.0](https://github.com/sindresorhus/camelcase) - MIT
+[caniuse-lite@1.0.30001579](https://github.com/browserslist/caniuse-lite) - CC-BY-4.0
+[chalk-template@0.4.0](https://github.com/chalk/chalk-template) - MIT
+[chalk@4.1.2](https://github.com/chalk/chalk) - MIT
+[char-regex@1.0.2](https://github.com/Richienb/char-regex) - MIT
+[charenc@0.0.2](https://github.com/pvorb/node-charenc) - BSD-3-Clause
+[chownr@3.0.0](https://github.com/isaacs/chownr) - BlueOak-1.0.0
+[ci-info@3.9.0](https://github.com/watson/ci-info) - MIT
+[cjs-module-lexer@1.2.3](https://github.com/nodejs/cjs-module-lexer) - MIT
+[cliui@8.0.1](https://github.com/yargs/cliui) - ISC
+[co@4.6.0](https://github.com/tj/co) - MIT
+[collect-v8-coverage@1.0.2](https://github.com/SimenB/collect-v8-coverage) - MIT
+[color-convert@2.0.1](https://github.com/Qix-/color-convert) - MIT
+[color-name@1.1.4](https://github.com/colorjs/color-name) - MIT
+[color-string@1.9.1](https://github.com/Qix-/color-string) - MIT
+[color@4.2.3](https://github.com/Qix-/color) - MIT
+[combined-stream@1.0.8](https://github.com/felixge/node-combined-stream) - MIT
+[command-line-args@5.2.1](https://github.com/75lb/command-line-args) - MIT
+[command-line-usage@7.0.1](https://github.com/75lb/command-line-usage) - MIT
+[concat-map@0.0.1](https://github.com/substack/node-concat-map) - MIT
+[convert-source-map@2.0.0](https://github.com/thlorenz/convert-source-map) - MIT
+[create-jest@29.7.0](https://github.com/jestjs/jest) - MIT
+[cross-spawn@7.0.6](https://github.com/moxystudio/node-cross-spawn) - MIT
+[crypt@0.0.2](https://github.com/pvorb/node-crypt) - BSD-3-Clause
+[debug@4.3.4](https://github.com/debug-js/debug) - MIT
+[dedent@1.5.1](https://github.com/dmnd/dedent) - MIT
+[deep-is@0.1.4](https://github.com/thlorenz/deep-is) - MIT
+[deepmerge@4.3.1](https://github.com/TehShrike/deepmerge) - MIT
+[delayed-stream@1.0.0](https://github.com/felixge/node-delayed-stream) - MIT
+[detect-libc@2.0.3](https://github.com/lovell/detect-libc) - Apache-2.0
+[detect-newline@3.1.0](https://github.com/sindresorhus/detect-newline) - MIT
+[diff-sequences@29.6.3](https://github.com/jestjs/jest) - MIT
+[digest-fetch@1.3.0](https://github.com/devfans/digest-fetch) - ISC
+[dir-glob@3.0.1](https://github.com/kevva/dir-glob) - MIT
+[doctrine@3.0.0](https://github.com/eslint/doctrine) - Apache-2.0
+[eastasianwidth@0.2.0](https://github.com/komagata/eastasianwidth) - MIT
+[electron-to-chromium@1.4.642](https://github.com/kilian/electron-to-chromium) - ISC
+[emittery@0.13.1](https://github.com/sindresorhus/emittery) - MIT
+[emoji-regex@8.0.0](https://github.com/mathiasbynens/emoji-regex) - MIT
+[emoji-regex@9.2.2](https://github.com/mathiasbynens/emoji-regex) - MIT
+[entities@4.5.0](https://github.com/fb55/entities) - BSD-2-Clause
+[error-ex@1.3.2](https://github.com/qix-/node-error-ex) - MIT
+[escalade@3.1.1](https://github.com/lukeed/escalade) - MIT
+[escape-string-regexp@2.0.0](https://github.com/sindresorhus/escape-string-regexp) - MIT
+[escape-string-regexp@4.0.0](https://github.com/sindresorhus/escape-string-regexp) - MIT
+[eslint-scope@7.2.2](https://github.com/eslint/eslint-scope) - BSD-2-Clause
+[eslint-visitor-keys@3.4.3](https://github.com/eslint/eslint-visitor-keys) - Apache-2.0
+[eslint@8.57.0](https://github.com/eslint/eslint) - MIT
+[espree@9.6.1](https://github.com/eslint/espree) - BSD-2-Clause
+[esprima@4.0.1](https://github.com/jquery/esprima) - BSD-2-Clause
+[esquery@1.5.0](https://github.com/estools/esquery) - BSD-3-Clause
+[esrecurse@4.3.0](https://github.com/estools/esrecurse) - BSD-2-Clause
+[estraverse@5.3.0](https://github.com/estools/estraverse) - BSD-2-Clause
+[esutils@2.0.3](https://github.com/estools/esutils) - BSD-2-Clause
+[event-target-shim@5.0.1](https://github.com/mysticatea/event-target-shim) - MIT
+[execa@5.1.1](https://github.com/sindresorhus/execa) - MIT
+[exit@0.1.2](https://github.com/cowboy/node-exit) - MIT
+[expect@29.7.0](https://github.com/jestjs/jest) - MIT
+[fast-deep-equal@3.1.3](https://github.com/epoberezkin/fast-deep-equal) - MIT
+[fast-glob@3.3.2](https://github.com/mrmlnc/fast-glob) - MIT
+[fast-json-stable-stringify@2.1.0](https://github.com/epoberezkin/fast-json-stable-stringify) - MIT
+[fast-levenshtein@2.0.6](https://github.com/hiddentao/fast-levenshtein) - MIT
+[fast-xml-parser@4.2.5](https://github.com/NaturalIntelligence/fast-xml-parser) - MIT
+[fastq@1.16.0](https://github.com/mcollina/fastq) - ISC
+[fb-watchman@2.0.2](https://github.com/facebook/watchman) - Apache-2.0
+[file-entry-cache@6.0.1](https://github.com/royriojas/file-entry-cache) - MIT
+[fill-range@7.1.1](https://github.com/jonschlinkert/fill-range) - MIT
+[find-replace@3.0.0](https://github.com/75lb/find-replace) - MIT
+[find-up@4.1.0](https://github.com/sindresorhus/find-up) - MIT
+[find-up@5.0.0](https://github.com/sindresorhus/find-up) - MIT
+[flat-cache@3.2.0](https://github.com/jaredwray/flat-cache) - MIT
+[flatbuffers@1.12.0](https://github.com/google/flatbuffers) - Apache*
+[flatbuffers@23.5.26](https://github.com/google/flatbuffers) - Apache*
+[flatbuffers@24.3.25](https://github.com/google/flatbuffers) - Apache-2.0
+[flatted@3.2.9](https://github.com/WebReflection/flatted) - ISC
+[follow-redirects@1.15.6](https://github.com/follow-redirects/follow-redirects) - MIT
+[foreground-child@3.3.0](https://github.com/tapjs/foreground-child) - ISC
+[form-data-encoder@1.7.2](https://github.com/octet-stream/form-data-encoder) - MIT
+[form-data@4.0.0](https://github.com/form-data/form-data) - MIT
+[formdata-node@4.4.1](https://github.com/octet-stream/form-data) - MIT
+[fs.realpath@1.0.0](https://github.com/isaacs/fs.realpath) - ISC
+[fsevents@2.3.3](https://github.com/fsevents/fsevents) - MIT
+[function-bind@1.1.2](https://github.com/Raynos/function-bind) - MIT
+[gensync@1.0.0-beta.2](https://github.com/loganfsmyth/gensync) - MIT
+[get-caller-file@2.0.5](https://github.com/stefanpenner/get-caller-file) - ISC
+[get-package-type@0.1.0](https://github.com/cfware/get-package-type) - MIT
+[get-stream@6.0.1](https://github.com/sindresorhus/get-stream) - MIT
+[glob-parent@5.1.2](https://github.com/gulpjs/glob-parent) - ISC
+[glob-parent@6.0.2](https://github.com/gulpjs/glob-parent) - ISC
+[glob@10.4.5](https://github.com/isaacs/node-glob) - ISC
+[glob@7.2.3](https://github.com/isaacs/node-glob) - ISC
+[globals@11.12.0](https://github.com/sindresorhus/globals) - MIT
+[globals@13.24.0](https://github.com/sindresorhus/globals) - MIT
+[globby@11.1.0](https://github.com/sindresorhus/globby) - MIT
+[graceful-fs@4.2.11](https://github.com/isaacs/node-graceful-fs) - ISC
+[graphemer@1.4.0](https://github.com/flmnt/graphemer) - MIT
+[guid-typescript@1.0.9](https://github.com/NicolasDeveloper/guid-typescript) - ISC
+[has-flag@4.0.0](https://github.com/sindresorhus/has-flag) - MIT
+[hasown@2.0.0](https://github.com/inspect-js/hasOwn) - MIT
+[html-escaper@2.0.2](https://github.com/WebReflection/html-escaper) - MIT
+[human-signals@2.1.0](https://github.com/ehmicky/human-signals) - Apache-2.0
+[humanize-ms@1.2.1](https://github.com/node-modules/humanize-ms) - MIT
+[ignore@5.3.0](https://github.com/kaelzhang/node-ignore) - MIT
+[import-fresh@3.3.0](https://github.com/sindresorhus/import-fresh) - MIT
+[import-local@3.1.0](https://github.com/sindresorhus/import-local) - MIT
+[imurmurhash@0.1.4](https://github.com/jensyt/imurmurhash-js) - MIT
+[inflight@1.0.6](https://github.com/npm/inflight) - ISC
+[inherits@2.0.4](https://github.com/isaacs/inherits) - ISC
+[interpret@1.4.0](https://github.com/gulpjs/interpret) - MIT
+[is-arrayish@0.2.1](https://github.com/qix-/node-is-arrayish) - MIT
+[is-arrayish@0.3.2](https://github.com/qix-/node-is-arrayish) - MIT
+[is-buffer@1.1.6](https://github.com/feross/is-buffer) - MIT
+[is-core-module@2.13.1](https://github.com/inspect-js/is-core-module) - MIT
+[is-extglob@2.1.1](https://github.com/jonschlinkert/is-extglob) - MIT
+[is-fullwidth-code-point@3.0.0](https://github.com/sindresorhus/is-fullwidth-code-point) - MIT
+[is-generator-fn@2.1.0](https://github.com/sindresorhus/is-generator-fn) - MIT
+[is-glob@4.0.3](https://github.com/micromatch/is-glob) - MIT
+[is-number@7.0.0](https://github.com/jonschlinkert/is-number) - MIT
+[is-path-inside@3.0.3](https://github.com/sindresorhus/is-path-inside) - MIT
+[is-stream@2.0.1](https://github.com/sindresorhus/is-stream) - MIT
+[isexe@2.0.0](https://github.com/isaacs/isexe) - ISC
+[istanbul-lib-coverage@3.2.2](https://github.com/istanbuljs/istanbuljs) - BSD-3-Clause
+[istanbul-lib-instrument@5.2.1](https://github.com/istanbuljs/istanbuljs) - BSD-3-Clause
+[istanbul-lib-instrument@6.0.1](https://github.com/istanbuljs/istanbuljs) - BSD-3-Clause
+[istanbul-lib-report@3.0.1](https://github.com/istanbuljs/istanbuljs) - BSD-3-Clause
+[istanbul-lib-source-maps@4.0.1](https://github.com/istanbuljs/istanbuljs) - BSD-3-Clause
+[istanbul-reports@3.1.6](https://github.com/istanbuljs/istanbuljs) - BSD-3-Clause
+[jackspeak@3.4.3](https://github.com/isaacs/jackspeak) - BlueOak-1.0.0
+[jest-changed-files@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-circus@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-cli@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-config@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-diff@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-docblock@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-each@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-environment-node@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-get-type@29.6.3](https://github.com/jestjs/jest) - MIT
+[jest-haste-map@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-leak-detector@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-matcher-utils@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-message-util@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-mock@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-pnp-resolver@1.2.3](https://github.com/arcanis/jest-pnp-resolver) - MIT
+[jest-regex-util@29.6.3](https://github.com/jestjs/jest) - MIT
+[jest-resolve-dependencies@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-resolve@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-runner@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-runtime@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-snapshot@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-util@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-validate@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-watcher@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest-worker@29.7.0](https://github.com/jestjs/jest) - MIT
+[jest@29.7.0](https://github.com/jestjs/jest) - MIT
+[js-tokens@4.0.0](https://github.com/lydell/js-tokens) - MIT
+[js-yaml@3.14.1](https://github.com/nodeca/js-yaml) - MIT
+[js-yaml@4.1.0](https://github.com/nodeca/js-yaml) - MIT
+[jsesc@2.5.2](https://github.com/mathiasbynens/jsesc) - MIT
+[json-bignum@0.0.3](https://github.com/datalanche/json-bignum) - MIT
+[json-buffer@3.0.1](https://github.com/dominictarr/json-buffer) - MIT
+[json-parse-even-better-errors@2.3.1](https://github.com/npm/json-parse-even-better-errors) - MIT
+[json-schema-traverse@0.4.1](https://github.com/epoberezkin/json-schema-traverse) - MIT
+[json-stable-stringify-without-jsonify@1.0.1](https://github.com/samn/json-stable-stringify) - MIT
+[json5@2.2.3](https://github.com/json5/json5) - MIT
+[keyv@4.5.4](https://github.com/jaredwray/keyv) - MIT
+[kleur@3.0.3](https://github.com/lukeed/kleur) - MIT
+[leven@3.1.0](https://github.com/sindresorhus/leven) - MIT
+[levn@0.4.1](https://github.com/gkz/levn) - MIT
+[lines-and-columns@1.2.4](https://github.com/eventualbuddha/lines-and-columns) - MIT
+[linkify-it@5.0.0](https://github.com/markdown-it/linkify-it) - MIT
+[locate-path@5.0.0](https://github.com/sindresorhus/locate-path) - MIT
+[locate-path@6.0.0](https://github.com/sindresorhus/locate-path) - MIT
+[lodash.camelcase@4.3.0](https://github.com/lodash/lodash) - MIT
+[lodash.memoize@4.1.2](https://github.com/lodash/lodash) - MIT
+[lodash.merge@4.6.2](https://github.com/lodash/lodash) - MIT
+[lodash@4.17.21](https://github.com/lodash/lodash) - MIT
+[long@5.2.3](https://github.com/dcodeIO/long.js) - Apache-2.0
+[lru-cache@10.4.3](https://github.com/isaacs/node-lru-cache) - ISC
+[lru-cache@5.1.1](https://github.com/isaacs/node-lru-cache) - ISC
+[lunr@2.3.9](https://github.com/olivernn/lunr.js) - MIT
+[make-dir@4.0.0](https://github.com/sindresorhus/make-dir) - MIT
+[make-error@1.3.6](https://github.com/JsCommunity/make-error) - ISC
+[makeerror@1.0.12](https://github.com/daaku/nodejs-makeerror) - BSD-3-Clause
+[markdown-it@14.1.0](https://github.com/markdown-it/markdown-it) - MIT
+[md5@2.3.0](https://github.com/pvorb/node-md5) - BSD-3-Clause
+[mdurl@2.0.0](https://github.com/markdown-it/mdurl) - MIT
+[merge-stream@2.0.0](https://github.com/grncdr/merge-stream) - MIT
+[merge2@1.4.1](https://github.com/teambition/merge2) - MIT
+[micromatch@4.0.8](https://github.com/micromatch/micromatch) - MIT
+[mime-db@1.52.0](https://github.com/jshttp/mime-db) - MIT
+[mime-types@2.1.35](https://github.com/jshttp/mime-types) - MIT
+[mimic-fn@2.1.0](https://github.com/sindresorhus/mimic-fn) - MIT
+[minimatch@3.1.2](https://github.com/isaacs/minimatch) - ISC
+[minimatch@9.0.3](https://github.com/isaacs/minimatch) - ISC
+[minimatch@9.0.5](https://github.com/isaacs/minimatch) - ISC
+[minimist@1.2.8](https://github.com/minimistjs/minimist) - MIT
+[minipass@7.1.2](https://github.com/isaacs/minipass) - ISC
+[minizlib@3.0.1](https://github.com/isaacs/minizlib) - MIT
+[mkdirp@3.0.1](https://github.com/isaacs/node-mkdirp) - MIT
+[mnemonist@0.38.3](https://github.com/yomguithereal/mnemonist) - MIT
+[ms@2.1.2](https://github.com/zeit/ms) - MIT
+[ms@2.1.3](https://github.com/vercel/ms) - MIT
+[natural-compare@1.4.0](https://github.com/litejs/natural-compare-lite) - MIT
+[node-domexception@1.0.0](https://github.com/jimmywarting/node-domexception) - MIT
+[node-fetch@2.7.0](https://github.com/bitinn/node-fetch) - MIT
+[node-int64@0.4.0](https://github.com/broofa/node-int64) - MIT
+[node-releases@2.0.14](https://github.com/chicoxyzzy/node-releases) - MIT
+[normalize-path@3.0.0](https://github.com/jonschlinkert/normalize-path) - MIT
+[npm-run-path@4.0.1](https://github.com/sindresorhus/npm-run-path) - MIT
+[obliterator@1.6.1](https://github.com/yomguithereal/obliterator) - MIT
+[once@1.4.0](https://github.com/isaacs/once) - ISC
+[onetime@5.1.2](https://github.com/sindresorhus/onetime) - MIT
+[onnxruntime-common@1.19.2](https://github.com/Microsoft/onnxruntime) - MIT
+[onnxruntime-common@1.20.0-dev.20241016-2b8fc5529b](https://github.com/Microsoft/onnxruntime) - MIT
+[onnxruntime-node@1.19.2](https://github.com/Microsoft/onnxruntime) - MIT
+[onnxruntime-web@1.21.0-dev.20241024-d9ca84ef96](https://github.com/Microsoft/onnxruntime) - MIT
+[openai@4.29.2](https://github.com/openai/openai-node) - Apache-2.0
+[optionator@0.9.3](https://github.com/gkz/optionator) - MIT
+[p-limit@2.3.0](https://github.com/sindresorhus/p-limit) - MIT
+[p-limit@3.1.0](https://github.com/sindresorhus/p-limit) - MIT
+[p-locate@4.1.0](https://github.com/sindresorhus/p-locate) - MIT
+[p-locate@5.0.0](https://github.com/sindresorhus/p-locate) - MIT
+[p-try@2.2.0](https://github.com/sindresorhus/p-try) - MIT
+[package-json-from-dist@1.0.1](https://github.com/isaacs/package-json-from-dist) - BlueOak-1.0.0
+[parent-module@1.0.1](https://github.com/sindresorhus/parent-module) - MIT
+[parse-json@5.2.0](https://github.com/sindresorhus/parse-json) - MIT
+[path-exists@4.0.0](https://github.com/sindresorhus/path-exists) - MIT
+[path-is-absolute@1.0.1](https://github.com/sindresorhus/path-is-absolute) - MIT
+[path-key@3.1.1](https://github.com/sindresorhus/path-key) - MIT
+[path-parse@1.0.7](https://github.com/jbgutierrez/path-parse) - MIT
+[path-scurry@1.11.1](https://github.com/isaacs/path-scurry) - BlueOak-1.0.0
+[path-type@4.0.0](https://github.com/sindresorhus/path-type) - MIT
+[picocolors@1.0.0](https://github.com/alexeyraspopov/picocolors) - ISC
+[picomatch@2.3.1](https://github.com/micromatch/picomatch) - MIT
+[pirates@4.0.6](https://github.com/danez/pirates) - MIT
+[pkg-dir@4.2.0](https://github.com/sindresorhus/pkg-dir) - MIT
+[platform@1.3.6](https://github.com/bestiejs/platform.js) - MIT
+[prelude-ls@1.2.1](https://github.com/gkz/prelude-ls) - MIT
+[pretty-format@29.7.0](https://github.com/jestjs/jest) - MIT
+[prompts@2.4.2](https://github.com/terkelg/prompts) - MIT
+[protobufjs@7.4.0](https://github.com/protobufjs/protobuf.js) - BSD-3-Clause
+[proxy-from-env@1.1.0](https://github.com/Rob--W/proxy-from-env) - MIT
+[punycode.js@2.3.1](https://github.com/mathiasbynens/punycode.js) - MIT
+[punycode@2.3.1](https://github.com/mathiasbynens/punycode.js) - MIT
+[pure-rand@6.0.4](https://github.com/dubzzz/pure-rand) - MIT
+[queue-microtask@1.2.3](https://github.com/feross/queue-microtask) - MIT
+[react-is@18.2.0](https://github.com/facebook/react) - MIT
+[rechoir@0.6.2](https://github.com/tkellen/node-rechoir) - MIT
+[reflect-metadata@0.2.2](https://github.com/rbuckton/reflect-metadata) - Apache-2.0
+[require-directory@2.1.1](https://github.com/troygoode/node-require-directory) - MIT
+[resolve-cwd@3.0.0](https://github.com/sindresorhus/resolve-cwd) - MIT
+[resolve-from@4.0.0](https://github.com/sindresorhus/resolve-from) - MIT
+[resolve-from@5.0.0](https://github.com/sindresorhus/resolve-from) - MIT
+[resolve.exports@2.0.2](https://github.com/lukeed/resolve.exports) - MIT
+[resolve@1.22.8](https://github.com/browserify/resolve) - MIT
+[reusify@1.0.4](https://github.com/mcollina/reusify) - MIT
+[rimraf@3.0.2](https://github.com/isaacs/rimraf) - ISC
+[rimraf@5.0.10](https://github.com/isaacs/rimraf) - ISC
+[run-parallel@1.2.0](https://github.com/feross/run-parallel) - MIT
+[semver@6.3.1](https://github.com/npm/node-semver) - ISC
+[semver@7.6.3](https://github.com/npm/node-semver) - ISC
+[sharp@0.33.5](https://github.com/lovell/sharp) - Apache-2.0
+[shebang-command@2.0.0](https://github.com/kevva/shebang-command) - MIT
+[shebang-regex@3.0.0](https://github.com/sindresorhus/shebang-regex) - MIT
+[shelljs@0.8.5](https://github.com/shelljs/shelljs) - BSD-3-Clause
+[shiki@1.10.3](https://github.com/shikijs/shiki) - MIT
+[shx@0.3.4](https://github.com/shelljs/shx) - MIT
+[signal-exit@3.0.7](https://github.com/tapjs/signal-exit) - ISC
+[signal-exit@4.1.0](https://github.com/tapjs/signal-exit) - ISC
+[simple-swizzle@0.2.2](https://github.com/qix-/node-simple-swizzle) - MIT
+[sisteransi@1.0.5](https://github.com/terkelg/sisteransi) - MIT
+[slash@3.0.0](https://github.com/sindresorhus/slash) - MIT
+[source-map-support@0.5.13](https://github.com/evanw/node-source-map-support) - MIT
+[source-map@0.6.1](https://github.com/mozilla/source-map) - BSD-3-Clause
+[sprintf-js@1.0.3](https://github.com/alexei/sprintf.js) - BSD-3-Clause
+[stack-utils@2.0.6](https://github.com/tapjs/stack-utils) - MIT
+[stream-read-all@3.0.1](https://github.com/75lb/stream-read-all) - MIT
+[string-length@4.0.2](https://github.com/sindresorhus/string-length) - MIT
+[string-width@4.2.3](https://github.com/sindresorhus/string-width) - MIT
+[string-width@5.1.2](https://github.com/sindresorhus/string-width) - MIT
+[strip-ansi@6.0.1](https://github.com/chalk/strip-ansi) - MIT
+[strip-ansi@7.1.0](https://github.com/chalk/strip-ansi) - MIT
+[strip-bom@4.0.0](https://github.com/sindresorhus/strip-bom) - MIT
+[strip-final-newline@2.0.0](https://github.com/sindresorhus/strip-final-newline) - MIT
+[strip-json-comments@3.1.1](https://github.com/sindresorhus/strip-json-comments) - MIT
+[strnum@1.0.5](https://github.com/NaturalIntelligence/strnum) - MIT
+[supports-color@7.2.0](https://github.com/chalk/supports-color) - MIT
+[supports-color@8.1.1](https://github.com/chalk/supports-color) - MIT
+[supports-preserve-symlinks-flag@1.0.0](https://github.com/inspect-js/node-supports-preserve-symlinks-flag) - MIT
+[table-layout@3.0.2](https://github.com/75lb/table-layout) - MIT
+[tar@7.4.3](https://github.com/isaacs/node-tar) - ISC
+[test-exclude@6.0.0](https://github.com/istanbuljs/test-exclude) - ISC
+[text-table@0.2.0](https://github.com/substack/text-table) - MIT
+[tmp@0.2.3](https://github.com/raszi/node-tmp) - MIT
+[tmpl@1.0.5](https://github.com/daaku/nodejs-tmpl) - BSD-3-Clause
+[to-regex-range@5.0.1](https://github.com/micromatch/to-regex-range) - MIT
+[tr46@0.0.3](https://github.com/Sebmaster/tr46.js) - MIT
+[ts-api-utils@1.0.3](https://github.com/JoshuaKGoldberg/ts-api-utils) - MIT
+[ts-jest@29.1.2](https://github.com/kulshekhar/ts-jest) - MIT
+[tslib@1.14.1](https://github.com/Microsoft/tslib) - 0BSD
+[tslib@2.6.2](https://github.com/Microsoft/tslib) - 0BSD
+[type-check@0.4.0](https://github.com/gkz/type-check) - MIT
+[type-detect@4.0.8](https://github.com/chaijs/type-detect) - MIT
+[type-fest@0.20.2](https://github.com/sindresorhus/type-fest) - (MIT OR CC0-1.0)
+[type-fest@0.21.3](https://github.com/sindresorhus/type-fest) - (MIT OR CC0-1.0)
+[typedoc-plugin-markdown@4.2.1](https://github.com/typedoc2md/typedoc-plugin-markdown) - MIT
+[typedoc@0.26.4](https://github.com/TypeStrong/TypeDoc) - Apache-2.0
+[typescript-eslint@7.1.0](https://github.com/typescript-eslint/typescript-eslint) - MIT
+[typescript@5.5.4](https://github.com/Microsoft/TypeScript) - Apache-2.0
+[typical@4.0.0](https://github.com/75lb/typical) - MIT
+[typical@7.1.1](https://github.com/75lb/typical) - MIT
+[uc.micro@2.1.0](https://github.com/markdown-it/uc.micro) - MIT
+[undici-types@5.26.5](https://github.com/nodejs/undici) - MIT
+[undici-types@6.19.8](https://github.com/nodejs/undici) - MIT
+[update-browserslist-db@1.0.13](https://github.com/browserslist/update-db) - MIT
+[uri-js@4.4.1](https://github.com/garycourt/uri-js) - BSD-2-Clause
+[uuid@9.0.1](https://github.com/uuidjs/uuid) - MIT
+[v8-to-istanbul@9.2.0](https://github.com/istanbuljs/v8-to-istanbul) - ISC
+[walker@1.0.8](https://github.com/daaku/nodejs-walker) - Apache-2.0
+[web-streams-polyfill@3.3.3](https://github.com/MattiasBuelens/web-streams-polyfill) - MIT
+[web-streams-polyfill@4.0.0-beta.3](https://github.com/MattiasBuelens/web-streams-polyfill) - MIT
+[webidl-conversions@3.0.1](https://github.com/jsdom/webidl-conversions) - BSD-2-Clause
+[whatwg-url@5.0.0](https://github.com/jsdom/whatwg-url) - MIT
+[which@2.0.2](https://github.com/isaacs/node-which) - ISC
+[wordwrapjs@5.1.0](https://github.com/75lb/wordwrapjs) - MIT
+[wrap-ansi@7.0.0](https://github.com/chalk/wrap-ansi) - MIT
+[wrap-ansi@8.1.0](https://github.com/chalk/wrap-ansi) - MIT
+[wrappy@1.0.2](https://github.com/npm/wrappy) - ISC
+[write-file-atomic@4.0.2](https://github.com/npm/write-file-atomic) - ISC
+[y18n@5.0.8](https://github.com/yargs/y18n) - ISC
+[yallist@3.1.1](https://github.com/isaacs/yallist) - ISC
+[yallist@5.0.0](https://github.com/isaacs/yallist) - BlueOak-1.0.0
+[yaml@2.4.5](https://github.com/eemeli/yaml) - ISC
+[yargs-parser@21.1.1](https://github.com/yargs/yargs-parser) - ISC
+[yargs@17.7.2](https://github.com/yargs/yargs) - MIT
+[yocto-queue@0.1.0](https://github.com/sindresorhus/yocto-queue) - MIT

nodejs/npm/darwin-arm64/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-darwin-arm64",
-	"version": "0.26.1",
+	"version": "0.27.0-beta.0",
 	"os": ["darwin"],
 	"cpu": ["arm64"],
 	"main": "lancedb.darwin-arm64.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-gnu",
-	"version": "0.26.1",
+	"version": "0.27.0-beta.0",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-gnu.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-musl",
-	"version": "0.26.1",
+	"version": "0.27.0-beta.0",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-musl.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-gnu",
-	"version": "0.26.1",
+	"version": "0.27.0-beta.0",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-gnu.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-musl",
-	"version": "0.26.1",
+	"version": "0.27.0-beta.0",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-musl.node",

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

@@ -1,6 +1,6 @@
 {
   "name": "@lancedb/lancedb-win32-arm64-msvc",
-  "version": "0.26.1",
+  "version": "0.27.0-beta.0",
   "os": [
     "win32"
   ],

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-win32-x64-msvc",
-	"version": "0.26.1",
+	"version": "0.27.0-beta.0",
 	"os": ["win32"],
 	"cpu": ["x64"],
 	"main": "lancedb.win32-x64-msvc.node",

nodejs/package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "@lancedb/lancedb",
-  "version": "0.26.1",
+  "version": "0.26.2",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "@lancedb/lancedb",
-      "version": "0.26.1",
+      "version": "0.26.2",
       "cpu": [
         "x64",
         "arm64"

nodejs/package.json

@@ -11,7 +11,7 @@
     "ann"
   ],
   "private": false,
-  "version": "0.26.1",
+  "version": "0.27.0-beta.0",
   "main": "dist/index.js",
   "exports": {
     ".": "./dist/index.js",

nodejs/src/connection.rs

@@ -13,6 +13,7 @@ use crate::header::JsHeaderProvider;
 use crate::table::Table;
 use crate::ConnectionOptions;
 use lancedb::connection::{ConnectBuilder, Connection as LanceDBConnection};
+
 use lancedb::ipc::{ipc_file_to_batches, ipc_file_to_schema};
 
 #[napi]

python/.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.29.2"
+current_version = "0.30.0-beta.0"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

python/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb-python"
-version = "0.29.2"
+version = "0.30.0-beta.0"
 edition.workspace = true
 description = "Python bindings for LanceDB"
 license.workspace = true

python/PYTHON_THIRD_PARTY_LICENSES.md

@@ -0,0 +1,206 @@
+| Name                           | Version         | License                                                                                          | URL                                                                                               |
+|--------------------------------|-----------------|--------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------|
+| InstructorEmbedding            | 1.0.1           | Apache License 2.0                                                                               | https://github.com/HKUNLP/instructor-embedding                                                    |
+| Jinja2                         | 3.1.6           | BSD License                                                                                      | https://github.com/pallets/jinja/                                                                 |
+| Markdown                       | 3.10.2          | BSD-3-Clause                                                                                     | https://Python-Markdown.github.io/                                                                |
+| MarkupSafe                     | 3.0.3           | BSD-3-Clause                                                                                     | https://github.com/pallets/markupsafe/                                                            |
+| PyJWT                          | 2.11.0          | MIT                                                                                              | https://github.com/jpadilla/pyjwt                                                                 |
+| PyYAML                         | 6.0.3           | MIT License                                                                                      | https://pyyaml.org/                                                                               |
+| Pygments                       | 2.19.2          | BSD License                                                                                      | https://pygments.org                                                                              |
+| accelerate                     | 1.12.0          | Apache Software License                                                                          | https://github.com/huggingface/accelerate                                                         |
+| adlfs                          | 2026.2.0        | BSD License                                                                                      | UNKNOWN                                                                                           |
+| aiohappyeyeballs               | 2.6.1           | Python Software Foundation License                                                               | https://github.com/aio-libs/aiohappyeyeballs                                                      |
+| aiohttp                        | 3.13.3          | Apache-2.0 AND MIT                                                                               | https://github.com/aio-libs/aiohttp                                                               |
+| aiosignal                      | 1.4.0           | Apache Software License                                                                          | https://github.com/aio-libs/aiosignal                                                             |
+| annotated-types                | 0.7.0           | MIT License                                                                                      | https://github.com/annotated-types/annotated-types                                                |
+| anyio                          | 4.12.1          | MIT                                                                                              | https://anyio.readthedocs.io/en/stable/versionhistory.html                                        |
+| appnope                        | 0.1.4           | BSD License                                                                                      | http://github.com/minrk/appnope                                                                   |
+| asttokens                      | 3.0.1           | Apache 2.0                                                                                       | https://github.com/gristlabs/asttokens                                                            |
+| attrs                          | 25.4.0          | MIT                                                                                              | https://www.attrs.org/en/stable/changelog.html                                                    |
+| awscli                         | 1.44.35         | Apache Software License                                                                          | http://aws.amazon.com/cli/                                                                        |
+| azure-core                     | 1.38.0          | MIT License                                                                                      | https://github.com/Azure/azure-sdk-for-python/tree/main/sdk/core/azure-core                       |
+| azure-datalake-store           | 0.0.53          | MIT License                                                                                      | https://github.com/Azure/azure-data-lake-store-python                                             |
+| azure-identity                 | 1.25.1          | MIT                                                                                              | https://github.com/Azure/azure-sdk-for-python                                                     |
+| azure-storage-blob             | 12.28.0         | MIT License                                                                                      | https://github.com/Azure/azure-sdk-for-python/tree/main/sdk/storage/azure-storage-blob            |
+| babel                          | 2.18.0          | BSD License                                                                                      | https://babel.pocoo.org/                                                                          |
+| backrefs                       | 6.1             | MIT                                                                                              | https://github.com/facelessuser/backrefs                                                          |
+| beautifulsoup4                 | 4.14.3          | MIT License                                                                                      | https://www.crummy.com/software/BeautifulSoup/bs4/                                                |
+| bleach                         | 6.3.0           | Apache Software License                                                                          | https://github.com/mozilla/bleach                                                                 |
+| boto3                          | 1.42.45         | Apache-2.0                                                                                       | https://github.com/boto/boto3                                                                     |
+| botocore                       | 1.42.45         | Apache-2.0                                                                                       | https://github.com/boto/botocore                                                                  |
+| cachetools                     | 7.0.0           | MIT                                                                                              | https://github.com/tkem/cachetools/                                                               |
+| certifi                        | 2026.1.4        | Mozilla Public License 2.0 (MPL 2.0)                                                             | https://github.com/certifi/python-certifi                                                         |
+| cffi                           | 2.0.0           | MIT                                                                                              | https://cffi.readthedocs.io/en/latest/whatsnew.html                                               |
+| cfgv                           | 3.5.0           | MIT                                                                                              | https://github.com/asottile/cfgv                                                                  |
+| charset-normalizer             | 3.4.4           | MIT                                                                                              | https://github.com/jawah/charset_normalizer/blob/master/CHANGELOG.md                              |
+| click                          | 8.3.1           | BSD-3-Clause                                                                                     | https://github.com/pallets/click/                                                                 |
+| cohere                         | 5.20.4          | MIT License                                                                                      | https://github.com/cohere-ai/cohere-python                                                        |
+| colorama                       | 0.4.6           | BSD License                                                                                      | https://github.com/tartley/colorama                                                               |
+| colpali_engine                 | 0.3.13          | MIT License                                                                                      | https://github.com/illuin-tech/colpali                                                            |
+| comm                           | 0.2.3           | BSD License                                                                                      | https://github.com/ipython/comm                                                                   |
+| cryptography                   | 46.0.4          | Apache-2.0 OR BSD-3-Clause                                                                       | https://github.com/pyca/cryptography                                                              |
+| datafusion                     | 51.0.0          | Apache Software License                                                                          | https://datafusion.apache.org/python                                                              |
+| debugpy                        | 1.8.20          | MIT License                                                                                      | https://aka.ms/debugpy                                                                            |
+| decorator                      | 5.2.1           | BSD License                                                                                      | UNKNOWN                                                                                           |
+| defusedxml                     | 0.7.1           | Python Software Foundation License                                                               | https://github.com/tiran/defusedxml                                                               |
+| deprecation                    | 2.1.0           | Apache Software License                                                                          | http://deprecation.readthedocs.io/                                                                |
+| distlib                        | 0.4.0           | Python Software Foundation License                                                               | https://github.com/pypa/distlib                                                                   |
+| distro                         | 1.9.0           | Apache Software License                                                                          | https://github.com/python-distro/distro                                                           |
+| docutils                       | 0.19            | BSD License; GNU General Public License (GPL); Public Domain; Python Software Foundation License | https://docutils.sourceforge.io/                                                                  |
+| duckdb                         | 1.4.4           | MIT License                                                                                      | https://github.com/duckdb/duckdb-python                                                           |
+| executing                      | 2.2.1           | MIT License                                                                                      | https://github.com/alexmojaki/executing                                                           |
+| fastavro                       | 1.12.1          | MIT                                                                                              | https://github.com/fastavro/fastavro                                                              |
+| fastjsonschema                 | 2.21.2          | BSD License                                                                                      | https://github.com/horejsek/python-fastjsonschema                                                 |
+| filelock                       | 3.20.3          | Unlicense                                                                                        | https://github.com/tox-dev/py-filelock                                                            |
+| frozenlist                     | 1.8.0           | Apache-2.0                                                                                       | https://github.com/aio-libs/frozenlist                                                            |
+| fsspec                         | 2026.2.0        | BSD-3-Clause                                                                                     | https://github.com/fsspec/filesystem_spec                                                         |
+| ftfy                           | 6.3.1           | Apache-2.0                                                                                       | https://ftfy.readthedocs.io/en/latest/                                                            |
+| ghp-import                     | 2.1.0           | Apache Software License                                                                          | https://github.com/c-w/ghp-import                                                                 |
+| google-ai-generativelanguage   | 0.6.15          | Apache Software License                                                                          | https://github.com/googleapis/google-cloud-python/tree/main/packages/google-ai-generativelanguage |
+| google-api-core                | 2.25.2          | Apache Software License                                                                          | https://github.com/googleapis/python-api-core                                                     |
+| google-api-python-client       | 2.189.0         | Apache Software License                                                                          | https://github.com/googleapis/google-api-python-client/                                           |
+| google-auth                    | 2.48.0          | Apache Software License                                                                          | https://github.com/googleapis/google-auth-library-python                                          |
+| google-auth-httplib2           | 0.3.0           | Apache Software License                                                                          | https://github.com/GoogleCloudPlatform/google-auth-library-python-httplib2                        |
+| google-generativeai            | 0.8.6           | Apache Software License                                                                          | https://github.com/google/generative-ai-python                                                    |
+| googleapis-common-protos       | 1.72.0          | Apache Software License                                                                          | https://github.com/googleapis/google-cloud-python/tree/main/packages/googleapis-common-protos     |
+| griffe                         | 2.0.0           | ISC                                                                                              | https://mkdocstrings.github.io/griffe                                                             |
+| griffecli                      | 2.0.0           | ISC                                                                                              | UNKNOWN                                                                                           |
+| griffelib                      | 2.0.0           | ISC                                                                                              | UNKNOWN                                                                                           |
+| grpcio                         | 1.78.0          | Apache-2.0                                                                                       | https://grpc.io                                                                                   |
+| grpcio-status                  | 1.71.2          | Apache Software License                                                                          | https://grpc.io                                                                                   |
+| h11                            | 0.16.0          | MIT License                                                                                      | https://github.com/python-hyper/h11                                                               |
+| hf-xet                         | 1.2.0           | Apache-2.0                                                                                       | https://github.com/huggingface/xet-core                                                           |
+| httpcore                       | 1.0.9           | BSD-3-Clause                                                                                     | https://www.encode.io/httpcore/                                                                   |
+| httplib2                       | 0.31.2          | MIT License                                                                                      | https://github.com/httplib2/httplib2                                                              |
+| httpx                          | 0.28.1          | BSD License                                                                                      | https://github.com/encode/httpx                                                                   |
+| huggingface_hub                | 0.36.2          | Apache Software License                                                                          | https://github.com/huggingface/huggingface_hub                                                    |
+| ibm-cos-sdk                    | 2.14.3          | Apache Software License                                                                          | https://github.com/ibm/ibm-cos-sdk-python                                                         |
+| ibm-cos-sdk-core               | 2.14.3          | Apache Software License                                                                          | https://github.com/ibm/ibm-cos-sdk-python-core                                                    |
+| ibm-cos-sdk-s3transfer         | 2.14.3          | Apache Software License                                                                          | https://github.com/IBM/ibm-cos-sdk-python-s3transfer                                              |
+| ibm_watsonx_ai                 | 1.5.1           | BSD License                                                                                      | https://ibm.github.io/watsonx-ai-python-sdk/changelog.html                                        |
+| identify                       | 2.6.16          | MIT                                                                                              | https://github.com/pre-commit/identify                                                            |
+| idna                           | 3.11            | BSD-3-Clause                                                                                     | https://github.com/kjd/idna                                                                       |
+| iniconfig                      | 2.3.0           | MIT                                                                                              | https://github.com/pytest-dev/iniconfig                                                           |
+| ipykernel                      | 6.31.0          | BSD-3-Clause                                                                                     | https://ipython.org                                                                               |
+| ipython                        | 9.10.0          | BSD-3-Clause                                                                                     | https://ipython.org                                                                               |
+| ipython_pygments_lexers        | 1.1.1           | BSD License                                                                                      | https://github.com/ipython/ipython-pygments-lexers                                                |
+| isodate                        | 0.7.2           | BSD License                                                                                      | https://github.com/gweis/isodate/                                                                 |
+| jedi                           | 0.19.2          | MIT License                                                                                      | https://github.com/davidhalter/jedi                                                               |
+| jiter                          | 0.13.0          | MIT License                                                                                      | https://github.com/pydantic/jiter/                                                                |
+| jmespath                       | 1.0.1           | MIT License                                                                                      | https://github.com/jmespath/jmespath.py                                                           |
+| joblib                         | 1.5.3           | BSD-3-Clause                                                                                     | https://joblib.readthedocs.io                                                                     |
+| jsonschema                     | 4.26.0          | MIT                                                                                              | https://github.com/python-jsonschema/jsonschema                                                   |
+| jsonschema-specifications      | 2025.9.1        | MIT                                                                                              | https://github.com/python-jsonschema/jsonschema-specifications                                    |
+| jupyter_client                 | 8.8.0           | BSD License                                                                                      | https://jupyter.org                                                                               |
+| jupyter_core                   | 5.9.1           | BSD-3-Clause                                                                                     | https://jupyter.org                                                                               |
+| jupyterlab_pygments            | 0.3.0           | BSD License                                                                                      | https://github.com/jupyterlab/jupyterlab_pygments                                                 |
+| jupytext                       | 1.19.1          | MIT License                                                                                      | https://github.com/mwouts/jupytext                                                                |
+| lance-namespace                | 0.4.5           | Apache-2.0                                                                                       | https://github.com/lance-format/lance-namespace                                                   |
+| lance-namespace-urllib3-client | 0.4.5           | Apache-2.0                                                                                       | https://github.com/lance-format/lance-namespace                                                   |
+| lancedb                        | 0.29.2          | Apache Software License                                                                          | https://github.com/lancedb/lancedb                                                                |
+| lomond                         | 0.3.3           | BSD License                                                                                      | https://github.com/wildfoundry/dataplicity-lomond                                                 |
+| markdown-it-py                 | 4.0.0           | MIT License                                                                                      | https://github.com/executablebooks/markdown-it-py                                                 |
+| matplotlib-inline              | 0.2.1           | UNKNOWN                                                                                          | https://github.com/ipython/matplotlib-inline                                                      |
+| mdit-py-plugins                | 0.5.0           | MIT License                                                                                      | https://github.com/executablebooks/mdit-py-plugins                                                |
+| mdurl                          | 0.1.2           | MIT License                                                                                      | https://github.com/executablebooks/mdurl                                                          |
+| mergedeep                      | 1.3.4           | MIT License                                                                                      | https://github.com/clarketm/mergedeep                                                             |
+| mistune                        | 3.2.0           | BSD License                                                                                      | https://github.com/lepture/mistune                                                                |
+| mkdocs                         | 1.6.1           | BSD-2-Clause                                                                                     | https://github.com/mkdocs/mkdocs                                                                  |
+| mkdocs-autorefs                | 1.4.3           | ISC                                                                                              | https://mkdocstrings.github.io/autorefs                                                           |
+| mkdocs-get-deps                | 0.2.0           | MIT                                                                                              | https://github.com/mkdocs/get-deps                                                                |
+| mkdocs-jupyter                 | 0.25.1          | Apache-2.0                                                                                       | https://github.com/danielfrg/mkdocs-jupyter                                                       |
+| mkdocs-material                | 9.7.1           | MIT                                                                                              | https://github.com/squidfunk/mkdocs-material                                                      |
+| mkdocs-material-extensions     | 1.3.1           | MIT                                                                                              | https://github.com/facelessuser/mkdocs-material-extensions                                        |
+| mkdocstrings                   | 1.0.3           | ISC                                                                                              | https://mkdocstrings.github.io                                                                    |
+| mkdocstrings-python            | 2.0.2           | ISC                                                                                              | https://mkdocstrings.github.io/python                                                             |
+| mpmath                         | 1.3.0           | BSD License                                                                                      | http://mpmath.org/                                                                                |
+| msal                           | 1.34.0          | MIT License                                                                                      | https://github.com/AzureAD/microsoft-authentication-library-for-python                            |
+| msal-extensions                | 1.3.1           | MIT License                                                                                      | https://github.com/AzureAD/microsoft-authentication-extensions-for-python/releases                |
+| multidict                      | 6.7.1           | Apache License 2.0                                                                               | https://github.com/aio-libs/multidict                                                             |
+| nbclient                       | 0.10.4          | BSD License                                                                                      | https://jupyter.org                                                                               |
+| nbconvert                      | 7.17.0          | BSD License                                                                                      | https://jupyter.org                                                                               |
+| nbformat                       | 5.10.4          | BSD License                                                                                      | https://jupyter.org                                                                               |
+| nest-asyncio                   | 1.6.0           | BSD License                                                                                      | https://github.com/erdewit/nest_asyncio                                                           |
+| networkx                       | 3.6.1           | BSD-3-Clause                                                                                     | https://networkx.org/                                                                             |
+| nodeenv                        | 1.10.0          | BSD License                                                                                      | https://github.com/ekalinin/nodeenv                                                               |
+| numpy                          | 2.4.2           | BSD-3-Clause AND 0BSD AND MIT AND Zlib AND CC0-1.0                                               | https://numpy.org                                                                                 |
+| ollama                         | 0.6.1           | MIT                                                                                              | https://ollama.com                                                                                |
+| open_clip_torch                | 3.2.0           | MIT License                                                                                      | https://github.com/mlfoundations/open_clip                                                        |
+| openai                         | 2.18.0          | Apache Software License                                                                          | https://github.com/openai/openai-python                                                           |
+| packaging                      | 26.0            | Apache-2.0 OR BSD-2-Clause                                                                       | https://github.com/pypa/packaging                                                                 |
+| paginate                       | 0.5.7           | MIT License                                                                                      | https://github.com/Signum/paginate                                                                |
+| pandas                         | 2.3.3           | BSD License                                                                                      | https://pandas.pydata.org                                                                         |
+| pandocfilters                  | 1.5.1           | BSD License                                                                                      | http://github.com/jgm/pandocfilters                                                               |
+| parso                          | 0.8.6           | MIT License                                                                                      | https://github.com/davidhalter/parso                                                              |
+| pathspec                       | 1.0.4           | Mozilla Public License 2.0 (MPL 2.0)                                                             | UNKNOWN                                                                                           |
+| peft                           | 0.17.1          | Apache Software License                                                                          | https://github.com/huggingface/peft                                                               |
+| pexpect                        | 4.9.0           | ISC License (ISCL)                                                                               | https://pexpect.readthedocs.io/                                                                   |
+| pillow                         | 12.1.0          | MIT-CMU                                                                                          | https://python-pillow.github.io                                                                   |
+| platformdirs                   | 4.5.1           | MIT                                                                                              | https://github.com/tox-dev/platformdirs                                                           |
+| pluggy                         | 1.6.0           | MIT License                                                                                      | UNKNOWN                                                                                           |
+| polars                         | 1.3.0           | MIT License                                                                                      | https://www.pola.rs/                                                                              |
+| pre_commit                     | 4.5.1           | MIT                                                                                              | https://github.com/pre-commit/pre-commit                                                          |
+| prompt_toolkit                 | 3.0.52          | BSD License                                                                                      | https://github.com/prompt-toolkit/python-prompt-toolkit                                           |
+| propcache                      | 0.4.1           | Apache Software License                                                                          | https://github.com/aio-libs/propcache                                                             |
+| proto-plus                     | 1.27.1          | Apache Software License                                                                          | https://github.com/googleapis/proto-plus-python                                                   |
+| protobuf                       | 5.29.6          | 3-Clause BSD License                                                                             | https://developers.google.com/protocol-buffers/                                                   |
+| psutil                         | 7.2.2           | BSD-3-Clause                                                                                     | https://github.com/giampaolo/psutil                                                               |
+| ptyprocess                     | 0.7.0           | ISC License (ISCL)                                                                               | https://github.com/pexpect/ptyprocess                                                             |
+| pure_eval                      | 0.2.3           | MIT License                                                                                      | http://github.com/alexmojaki/pure_eval                                                            |
+| pyarrow                        | 23.0.0          | Apache-2.0                                                                                       | https://arrow.apache.org/                                                                         |
+| pyarrow-stubs                  | 20.0.0.20251215 | BSD-2-Clause                                                                                     | https://github.com/zen-xu/pyarrow-stubs                                                           |
+| pyasn1                         | 0.6.2           | BSD-2-Clause                                                                                     | https://github.com/pyasn1/pyasn1                                                                  |
+| pyasn1_modules                 | 0.4.2           | BSD License                                                                                      | https://github.com/pyasn1/pyasn1-modules                                                          |
+| pycparser                      | 3.0             | BSD-3-Clause                                                                                     | https://github.com/eliben/pycparser                                                               |
+| pydantic                       | 2.12.5          | MIT                                                                                              | https://github.com/pydantic/pydantic                                                              |
+| pydantic_core                  | 2.41.5          | MIT                                                                                              | https://github.com/pydantic/pydantic-core                                                         |
+| pylance                        | 2.0.0           | Apache Software License                                                                          | UNKNOWN                                                                                           |
+| pymdown-extensions             | 10.20.1         | MIT                                                                                              | https://github.com/facelessuser/pymdown-extensions                                                |
+| pyparsing                      | 3.3.2           | MIT                                                                                              | https://github.com/pyparsing/pyparsing/                                                           |
+| pyright                        | 1.1.408         | MIT                                                                                              | https://github.com/RobertCraigie/pyright-python                                                   |
+| pytest                         | 9.0.2           | MIT                                                                                              | https://docs.pytest.org/en/latest/                                                                |
+| pytest-asyncio                 | 1.3.0           | Apache-2.0                                                                                       | https://github.com/pytest-dev/pytest-asyncio                                                      |
+| pytest-mock                    | 3.15.1          | MIT License                                                                                      | https://github.com/pytest-dev/pytest-mock/                                                        |
+| python-dateutil                | 2.9.0.post0     | Apache Software License; BSD License                                                             | https://github.com/dateutil/dateutil                                                              |
+| pytz                           | 2025.2          | MIT License                                                                                      | http://pythonhosted.org/pytz                                                                      |
+| pyyaml_env_tag                 | 1.1             | MIT                                                                                              | https://github.com/waylan/pyyaml-env-tag                                                          |
+| pyzmq                          | 27.1.0          | BSD License                                                                                      | https://pyzmq.readthedocs.org                                                                     |
+| referencing                    | 0.37.0          | MIT                                                                                              | https://github.com/python-jsonschema/referencing                                                  |
+| regex                          | 2026.1.15       | Apache-2.0 AND CNRI-Python                                                                       | https://github.com/mrabarnett/mrab-regex                                                          |
+| requests                       | 2.32.5          | Apache Software License                                                                          | https://requests.readthedocs.io                                                                   |
+| rpds-py                        | 0.30.0          | MIT                                                                                              | https://github.com/crate-py/rpds                                                                  |
+| rsa                            | 4.7.2           | Apache Software License                                                                          | https://stuvel.eu/rsa                                                                             |
+| ruff                           | 0.15.0          | MIT License                                                                                      | https://docs.astral.sh/ruff                                                                       |
+| s3transfer                     | 0.16.0          | Apache Software License                                                                          | https://github.com/boto/s3transfer                                                                |
+| safetensors                    | 0.7.0           | Apache Software License                                                                          | https://github.com/huggingface/safetensors                                                        |
+| scikit-learn                   | 1.8.0           | BSD-3-Clause                                                                                     | https://scikit-learn.org                                                                          |
+| scipy                          | 1.17.0          | BSD License                                                                                      | https://scipy.org/                                                                                |
+| sentence-transformers          | 5.2.2           | Apache Software License                                                                          | https://www.SBERT.net                                                                             |
+| sentencepiece                  | 0.2.1           | UNKNOWN                                                                                          | https://github.com/google/sentencepiece                                                           |
+| six                            | 1.17.0          | MIT License                                                                                      | https://github.com/benjaminp/six                                                                  |
+| sniffio                        | 1.3.1           | Apache Software License; MIT License                                                             | https://github.com/python-trio/sniffio                                                            |
+| soupsieve                      | 2.8.3           | MIT                                                                                              | https://github.com/facelessuser/soupsieve                                                         |
+| stack-data                     | 0.6.3           | MIT License                                                                                      | http://github.com/alexmojaki/stack_data                                                           |
+| sympy                          | 1.14.0          | BSD License                                                                                      | https://sympy.org                                                                                 |
+| tabulate                       | 0.9.0           | MIT License                                                                                      | https://github.com/astanin/python-tabulate                                                        |
+| tantivy                        | 0.25.1          | UNKNOWN                                                                                          | UNKNOWN                                                                                           |
+| threadpoolctl                  | 3.6.0           | BSD License                                                                                      | https://github.com/joblib/threadpoolctl                                                           |
+| timm                           | 1.0.24          | Apache Software License                                                                          | https://github.com/huggingface/pytorch-image-models                                               |
+| tinycss2                       | 1.4.0           | BSD License                                                                                      | https://www.courtbouillon.org/tinycss2                                                            |
+| tokenizers                     | 0.22.2          | Apache Software License                                                                          | https://github.com/huggingface/tokenizers                                                         |
+| torch                          | 2.8.0           | BSD License                                                                                      | https://pytorch.org/                                                                              |
+| torchvision                    | 0.23.0          | BSD                                                                                              | https://github.com/pytorch/vision                                                                 |
+| tornado                        | 6.5.4           | Apache Software License                                                                          | http://www.tornadoweb.org/                                                                        |
+| tqdm                           | 4.67.3          | MPL-2.0 AND MIT                                                                                  | https://tqdm.github.io                                                                            |
+| traitlets                      | 5.14.3          | BSD License                                                                                      | https://github.com/ipython/traitlets                                                              |
+| transformers                   | 4.57.6          | Apache Software License                                                                          | https://github.com/huggingface/transformers                                                       |
+| types-requests                 | 2.32.4.20260107 | Apache-2.0                                                                                       | https://github.com/python/typeshed                                                                |
+| typing-inspection              | 0.4.2           | MIT                                                                                              | https://github.com/pydantic/typing-inspection                                                     |
+| typing_extensions              | 4.15.0          | PSF-2.0                                                                                          | https://github.com/python/typing_extensions                                                       |
+| tzdata                         | 2025.3          | Apache-2.0                                                                                       | https://github.com/python/tzdata                                                                  |
+| uritemplate                    | 4.2.0           | BSD 3-Clause OR Apache-2.0                                                                       | https://uritemplate.readthedocs.org                                                               |
+| urllib3                        | 2.6.3           | MIT                                                                                              | https://github.com/urllib3/urllib3/blob/main/CHANGES.rst                                          |
+| virtualenv                     | 20.36.1         | MIT                                                                                              | https://github.com/pypa/virtualenv                                                                |
+| watchdog                       | 6.0.0           | Apache Software License                                                                          | https://github.com/gorakhargosh/watchdog                                                          |
+| webencodings                   | 0.5.1           | BSD License                                                                                      | https://github.com/SimonSapin/python-webencodings                                                 |
+| yarl                           | 1.22.0          | Apache Software License                                                                          | https://github.com/aio-libs/yarl                                                                  |

python/python/lancedb/permutation.py

@@ -9,7 +9,7 @@ import json
 from ._lancedb import async_permutation_builder, PermutationReader
 from .table import LanceTable
 from .background_loop import LOOP
-from .util import batch_to_tensor
+from .util import batch_to_tensor, batch_to_tensor_rows
 from typing import Any, Callable, Iterator, Literal, Optional, TYPE_CHECKING, Union
 
 if TYPE_CHECKING:
@@ -333,7 +333,11 @@ class Transforms:
     """
 
     @staticmethod
-    def arrow2python(batch: pa.RecordBatch) -> dict[str, list[Any]]:
+    def arrow2python(batch: pa.RecordBatch) -> list[dict[str, Any]]:
+        return batch.to_pylist()
+
+    @staticmethod
+    def arrow2pythoncol(batch: pa.RecordBatch) -> dict[str, list[Any]]:
         return batch.to_pydict()
 
     @staticmethod
@@ -687,7 +691,17 @@ class Permutation:
             return
 
     def with_format(
-        self, format: Literal["numpy", "python", "pandas", "arrow", "torch", "polars"]
+        self,
+        format: Literal[
+            "numpy",
+            "python",
+            "python_col",
+            "pandas",
+            "arrow",
+            "torch",
+            "torch_col",
+            "polars",
+        ],
     ) -> "Permutation":
         """
         Set the format for batches
@@ -696,16 +710,18 @@ class Permutation:
 
         The format can be one of:
         - "numpy" - the batch will be a dict of numpy arrays (one per column)
-        - "python" - the batch will be a dict of lists (one per column)
+        - "python" - the batch will be a list of dicts (one per row)
+        - "python_col" - the batch will be a dict of lists (one entry per column)
         - "pandas" - the batch will be a pandas DataFrame
         - "arrow" - the batch will be a pyarrow RecordBatch
-        - "torch" - the batch will be a two dimensional torch tensor
+        - "torch" - the batch will be a list of tensors, one per row
+        - "torch_col" - the batch will be a 2D torch tensor (first dim indexes columns)
         - "polars" - the batch will be a polars DataFrame
 
         Conversion may or may not involve a data copy.  Lance uses Arrow internally
-        and so it is able to zero-copy to the arrow and polars.
+        and so it is able to zero-copy to the arrow and polars formats.
 
-        Conversion to torch will be zero-copy but will only support a subset of data
+        Conversion to torch_col will be zero-copy but will only support a subset of data
         types (numeric types).
 
         Conversion to numpy and/or pandas will typically be zero-copy for numeric
@@ -718,6 +734,8 @@ class Permutation:
         assert format is not None, "format is required"
         if format == "python":
             return self.with_transform(Transforms.arrow2python)
+        if format == "python_col":
+            return self.with_transform(Transforms.arrow2pythoncol)
         elif format == "numpy":
             return self.with_transform(Transforms.arrow2numpy)
         elif format == "pandas":
@@ -725,6 +743,8 @@ class Permutation:
         elif format == "arrow":
             return self.with_transform(Transforms.arrow2arrow)
         elif format == "torch":
+            return self.with_transform(batch_to_tensor_rows)
+        elif format == "torch_col":
             return self.with_transform(batch_to_tensor)
         elif format == "polars":
             return self.with_transform(Transforms.arrow2polars())
@@ -746,15 +766,20 @@ class Permutation:
 
     def __getitem__(self, index: int) -> Any:
         """
-        Return a single row from the permutation
-
-        The output will always be a python dictionary regardless of the format.
-
-        This method is mostly useful for debugging and exploration.  For actual
-        processing use [iter](#iter) or a torch data loader to perform batched
-        processing.
+        Returns a single row from the permutation by offset
         """
-        pass
+        return self.__getitems__([index])
+
+    def __getitems__(self, indices: list[int]) -> Any:
+        """
+        Returns rows from the permutation by offset
+        """
+
+        async def do_getitems():
+            return await self.reader.take_offsets(indices, selection=self.selection)
+
+        batch = LOOP.run(do_getitems())
+        return self.transform_fn(batch)
 
     @deprecated(details="Use with_skip instead")
     def skip(self, skip: int) -> "Permutation":

python/python/lancedb/query.py

@@ -2118,19 +2118,17 @@ class LanceHybridQueryBuilder(LanceQueryBuilder):
         """  # noqa: E501
         self._create_query_builders()
 
-        results = ["Vector Search Plan:"]
-        results.append(
-            self._table._explain_plan(
-                self._vector_query.to_query_object(), verbose=verbose
-            )
+        reranker_label = str(self._reranker) if self._reranker else "No reranker"
+        vector_plan = self._table._explain_plan(
+            self._vector_query.to_query_object(), verbose=verbose
         )
-        results.append("FTS Search Plan:")
-        results.append(
-            self._table._explain_plan(
-                self._fts_query.to_query_object(), verbose=verbose
-            )
+        fts_plan = self._table._explain_plan(
+            self._fts_query.to_query_object(), verbose=verbose
         )
-        return "\n".join(results)
+        # Indent sub-plans under the reranker
+        indented_vector = "\n".join("  " + line for line in vector_plan.splitlines())
+        indented_fts = "\n".join("  " + line for line in fts_plan.splitlines())
+        return f"{reranker_label}\n  {indented_vector}\n  {indented_fts}"
 
     def analyze_plan(self):
         """Execute the query and display with runtime metrics.
@@ -3164,23 +3162,20 @@ class AsyncHybridQuery(AsyncStandardQuery, AsyncVectorQueryBase):
         ...     plan = await table.query().nearest_to([1.0, 2.0]).nearest_to_text("hello").explain_plan(True)
         ...     print(plan)
         >>> asyncio.run(doctest_example()) # doctest: +ELLIPSIS, +NORMALIZE_WHITESPACE
-        Vector Search Plan:
-        ProjectionExec: expr=[vector@0 as vector, text@3 as text, _distance@2 as _distance]
-          Take: columns="vector, _rowid, _distance, (text)"
-            CoalesceBatchesExec: target_batch_size=1024
-              GlobalLimitExec: skip=0, fetch=10
-                FilterExec: _distance@2 IS NOT NULL
-                  SortExec: TopK(fetch=10), expr=[_distance@2 ASC NULLS LAST, _rowid@1 ASC NULLS LAST], preserve_partitioning=[false]
-                    KNNVectorDistance: metric=l2
-                      LanceRead: uri=..., projection=[vector], ...
-        <BLANKLINE>
-        FTS Search Plan:
-        ProjectionExec: expr=[vector@2 as vector, text@3 as text, _score@1 as _score]
-          Take: columns="_rowid, _score, (vector), (text)"
-            CoalesceBatchesExec: target_batch_size=1024
-              GlobalLimitExec: skip=0, fetch=10
-                MatchQuery: column=text, query=hello
-        <BLANKLINE>
+        RRFReranker(K=60)
+            ProjectionExec: expr=[vector@0 as vector, text@3 as text, _distance@2 as _distance]
+              Take: columns="vector, _rowid, _distance, (text)"
+                CoalesceBatchesExec: target_batch_size=1024
+                  GlobalLimitExec: skip=0, fetch=10
+                    FilterExec: _distance@2 IS NOT NULL
+                      SortExec: TopK(fetch=10), expr=[_distance@2 ASC NULLS LAST, _rowid@1 ASC NULLS LAST], preserve_partitioning=[false]
+                        KNNVectorDistance: metric=l2
+                          LanceRead: uri=..., projection=[vector], ...
+            ProjectionExec: expr=[vector@2 as vector, text@3 as text, _score@1 as _score]
+              Take: columns="_rowid, _score, (vector), (text)"
+                CoalesceBatchesExec: target_batch_size=1024
+                  GlobalLimitExec: skip=0, fetch=10
+                    MatchQuery: column=text, query=hello
 
         Parameters
         ----------
@@ -3192,12 +3187,12 @@ class AsyncHybridQuery(AsyncStandardQuery, AsyncVectorQueryBase):
         plan : str
         """  # noqa: E501
 
-        results = ["Vector Search Plan:"]
-        results.append(await self._inner.to_vector_query().explain_plan(verbose))
-        results.append("FTS Search Plan:")
-        results.append(await self._inner.to_fts_query().explain_plan(verbose))
-
-        return "\n".join(results)
+        vector_plan = await self._inner.to_vector_query().explain_plan(verbose)
+        fts_plan = await self._inner.to_fts_query().explain_plan(verbose)
+        # Indent sub-plans under the reranker
+        indented_vector = "\n".join("  " + line for line in vector_plan.splitlines())
+        indented_fts = "\n".join("  " + line for line in fts_plan.splitlines())
+        return f"{self._reranker}\n  {indented_vector}\n  {indented_fts}"
 
     async def analyze_plan(self):
         """

python/python/lancedb/rerankers/answerdotai.py

@@ -42,10 +42,18 @@ class AnswerdotaiRerankers(Reranker):
         rerankers = attempt_import_or_raise(
             "rerankers"
         )  # import here for faster ops later
+        self.model_name = model_name
+        self.model_type = model_type
         self.reranker = rerankers.Reranker(
             model_name=model_name, model_type=model_type, **kwargs
         )
 
+    def __str__(self):
+        return (
+            f"AnswerdotaiRerankers(model_type={self.model_type}, "
+            f"model_name={self.model_name})"
+        )
+
     def _rerank(self, result_set: pa.Table, query: str):
         result_set = self._handle_empty_results(result_set)
         if len(result_set) == 0:

python/python/lancedb/rerankers/base.py

@@ -40,6 +40,9 @@ class Reranker(ABC):
         if ARROW_VERSION.major <= 13:
             self._concat_tables_args = {"promote": True}
 
+    def __str__(self):
+        return self.__class__.__name__
+
     def rerank_vector(
         self,
         query: str,

python/python/lancedb/rerankers/cohere.py

@@ -44,6 +44,9 @@ class CohereReranker(Reranker):
         self.top_n = top_n
         self.api_key = api_key
 
+    def __str__(self):
+        return f"CohereReranker(model_name={self.model_name})"
+
     @cached_property
     def _client(self):
         cohere = attempt_import_or_raise("cohere")

python/python/lancedb/rerankers/cross_encoder.py

@@ -50,6 +50,9 @@ class CrossEncoderReranker(Reranker):
         if self.device is None:
             self.device = "cuda" if torch.cuda.is_available() else "cpu"
 
+    def __str__(self):
+        return f"CrossEncoderReranker(model_name={self.model_name})"
+
     @cached_property
     def model(self):
         sbert = attempt_import_or_raise("sentence_transformers")

python/python/lancedb/rerankers/jinaai.py

@@ -45,6 +45,9 @@ class JinaReranker(Reranker):
         self.top_n = top_n
         self.api_key = api_key
 
+    def __str__(self):
+        return f"JinaReranker(model_name={self.model_name})"
+
     @cached_property
     def _client(self):
         import requests

python/python/lancedb/rerankers/linear_combination.py

@@ -38,6 +38,9 @@ class LinearCombinationReranker(Reranker):
         self.weight = weight
         self.fill = fill
 
+    def __str__(self):
+        return f"LinearCombinationReranker(weight={self.weight}, fill={self.fill})"
+
     def rerank_hybrid(
         self,
         query: str,  # noqa: F821

python/python/lancedb/rerankers/mrr.py

@@ -54,6 +54,12 @@ class MRRReranker(Reranker):
         self.weight_vector = weight_vector
         self.weight_fts = weight_fts
 
+    def __str__(self):
+        return (
+            f"MRRReranker(weight_vector={self.weight_vector}, "
+            f"weight_fts={self.weight_fts})"
+        )
+
     def rerank_hybrid(
         self,
         query: str,  # noqa: F821

python/python/lancedb/rerankers/openai.py

@@ -43,6 +43,9 @@ class OpenaiReranker(Reranker):
         self.column = column
         self.api_key = api_key
 
+    def __str__(self):
+        return f"OpenaiReranker(model_name={self.model_name})"
+
     def _rerank(self, result_set: pa.Table, query: str):
         result_set = self._handle_empty_results(result_set)
         if len(result_set) == 0:

python/python/lancedb/rerankers/rrf.py

@@ -36,6 +36,9 @@ class RRFReranker(Reranker):
         super().__init__(return_score)
         self.K = K
 
+    def __str__(self):
+        return f"RRFReranker(K={self.K})"
+
     def rerank_hybrid(
         self,
         query: str,  # noqa: F821

python/python/lancedb/rerankers/voyageai.py

@@ -52,6 +52,9 @@ class VoyageAIReranker(Reranker):
         self.api_key = api_key
         self.truncation = truncation
 
+    def __str__(self):
+        return f"VoyageAIReranker(model_name={self.model_name})"
+
     @cached_property
     def _client(self):
         voyageai = attempt_import_or_raise("voyageai")

python/python/lancedb/table.py

@@ -904,7 +904,9 @@ class Table(ABC):
         ----------
         field_names: str or list of str
             The name(s) of the field to index.
-            can be only str if use_tantivy=True for now.
+            If ``use_tantivy`` is False (default), only a single field name
+            (str) is supported. To index multiple fields, create a separate
+            FTS index for each field.
         replace: bool, default False
             If True, replace the existing index if it exists. Note that this is
             not yet an atomic operation; the index will be temporarily
@@ -2298,7 +2300,11 @@ class LanceTable(Table):
     ):
         if not use_tantivy:
             if not isinstance(field_names, str):
-                raise ValueError("field_names must be a string when use_tantivy=False")
+                raise ValueError(
+                    "Native FTS indexes can only be created on a single field "
+                    "at a time. To search over multiple text fields, create a "
+                    "separate FTS index for each field."
+                )
 
             if tokenizer_name is None:
                 tokenizer_configs = {

python/python/lancedb/util.py

@@ -419,3 +419,22 @@ def batch_to_tensor(batch: pa.RecordBatch):
     """
     torch = attempt_import_or_raise("torch", "torch")
     return torch.stack([torch.from_dlpack(col) for col in batch.columns])
+
+
+def batch_to_tensor_rows(batch: pa.RecordBatch):
+    """
+    Convert a PyArrow RecordBatch to a list of PyTorch Tensor, one per row
+
+    Each column is converted to a tensor (using zero-copy via DLPack)
+    and the columns are then stacked into a single tensor.  The 2D tensor
+    is then converted to a list of tensors, one per row
+
+    Fails if torch or numpy is not installed.
+    Fails if a column's data type is not supported by PyTorch.
+    """
+    torch = attempt_import_or_raise("torch", "torch")
+    numpy = attempt_import_or_raise("numpy", "numpy")
+    columns = [col.to_numpy(zero_copy_only=False) for col in batch.columns]
+    stacked = torch.tensor(numpy.column_stack(columns))
+    rows = list(stacked.unbind(dim=0))
+    return rows

python/python/tests/test_hybrid_query.py

@@ -163,9 +163,7 @@ async def test_explain_plan(table: AsyncTable):
         table.query().nearest_to_text("dog").nearest_to([0.1, 0.1]).explain_plan(True)
     )
 
-    assert "Vector Search Plan" in plan
     assert "KNNVectorDistance" in plan
-    assert "FTS Search Plan" in plan
     assert "LanceRead" in plan
 
 

python/python/tests/test_permutation.py

@@ -664,23 +664,20 @@ def test_iter_basic(some_permutation: Permutation):
     expected_batches = (950 + batch_size - 1) // batch_size  # ceiling division
     assert len(batches) == expected_batches
 
-    # Check that all batches are dicts (default python format)
-    assert all(isinstance(batch, dict) for batch in batches)
+    # Check that all batches are lists of dicts (default python format)
+    assert all(isinstance(batch, list) for batch in batches)
 
     # Check that batches have the correct structure
     for batch in batches:
-        assert "id" in batch
-        assert "value" in batch
-        assert isinstance(batch["id"], list)
-        assert isinstance(batch["value"], list)
+        assert "id" in batch[0]
+        assert "value" in batch[0]
 
     # Check that all batches except the last have the correct size
     for batch in batches[:-1]:
-        assert len(batch["id"]) == batch_size
-        assert len(batch["value"]) == batch_size
+        assert len(batch) == batch_size
 
     # Last batch might be smaller
-    assert len(batches[-1]["id"]) <= batch_size
+    assert len(batches[-1]) <= batch_size
 
 
 def test_iter_skip_last_batch(some_permutation: Permutation):
@@ -699,11 +696,11 @@ def test_iter_skip_last_batch(some_permutation: Permutation):
     if 950 % batch_size != 0:
         assert len(batches_without_skip) == num_full_batches + 1
         # Last batch should be smaller
-        assert len(batches_without_skip[-1]["id"]) == 950 % batch_size
+        assert len(batches_without_skip[-1]) == 950 % batch_size
 
     # All batches with skip_last_batch should be full size
     for batch in batches_with_skip:
-        assert len(batch["id"]) == batch_size
+        assert len(batch) == batch_size
 
 
 def test_iter_different_batch_sizes(some_permutation: Permutation):
@@ -720,12 +717,12 @@ def test_iter_different_batch_sizes(some_permutation: Permutation):
     # Test with batch size equal to total rows
     single_batch = list(some_permutation.iter(950, skip_last_batch=False))
     assert len(single_batch) == 1
-    assert len(single_batch[0]["id"]) == 950
+    assert len(single_batch[0]) == 950
 
     # Test with batch size larger than total rows
     oversized_batch = list(some_permutation.iter(10000, skip_last_batch=False))
     assert len(oversized_batch) == 1
-    assert len(oversized_batch[0]["id"]) == 950
+    assert len(oversized_batch[0]) == 950
 
 
 def test_dunder_iter(some_permutation: Permutation):
@@ -738,15 +735,13 @@ def test_dunder_iter(some_permutation: Permutation):
 
     # All batches should be full size
     for batch in batches:
-        assert len(batch["id"]) == 100
-        assert len(batch["value"]) == 100
+        assert len(batch) == 100
 
     some_permutation = some_permutation.with_batch_size(400)
     batches = list(some_permutation)
     assert len(batches) == 2  # floor(950 / 400) since skip_last_batch=True
     for batch in batches:
-        assert len(batch["id"]) == 400
-        assert len(batch["value"]) == 400
+        assert len(batch) == 400
 
 
 def test_iter_with_different_formats(some_permutation: Permutation):
@@ -761,7 +756,7 @@ def test_iter_with_different_formats(some_permutation: Permutation):
     # Test with python format (default)
     python_perm = some_permutation.with_format("python")
     python_batches = list(python_perm.iter(batch_size, skip_last_batch=False))
-    assert all(isinstance(batch, dict) for batch in python_batches)
+    assert all(isinstance(batch, list) for batch in python_batches)
 
     # Test with pandas format
     pandas_perm = some_permutation.with_format("pandas")
@@ -780,8 +775,8 @@ def test_iter_with_column_selection(some_permutation: Permutation):
 
     # Check that batches only contain the id column
     for batch in batches:
-        assert "id" in batch
-        assert "value" not in batch
+        assert "id" in batch[0]
+        assert "value" not in batch[0]
 
 
 def test_iter_with_column_rename(some_permutation: Permutation):
@@ -791,9 +786,9 @@ def test_iter_with_column_rename(some_permutation: Permutation):
 
     # Check that batches have the renamed column
     for batch in batches:
-        assert "id" in batch
-        assert "data" in batch
-        assert "value" not in batch
+        assert "id" in batch[0]
+        assert "data" in batch[0]
+        assert "value" not in batch[0]
 
 
 def test_iter_with_limit_offset(some_permutation: Permutation):
@@ -812,14 +807,14 @@ def test_iter_with_limit_offset(some_permutation: Permutation):
     assert len(limit_batches) == 5
 
     no_skip = some_permutation.iter(101, skip_last_batch=False)
-    row_100 = next(no_skip)["id"][100]
+    row_100 = next(no_skip)[100]["id"]
 
     # Test with both limit and offset
     limited_perm = some_permutation.with_skip(100).with_take(300)
     limited_batches = list(limited_perm.iter(100, skip_last_batch=False))
     # Should have 3 batches (300 / 100)
     assert len(limited_batches) == 3
-    assert limited_batches[0]["id"][0] == row_100
+    assert limited_batches[0][0]["id"] == row_100
 
 
 def test_iter_empty_permutation(mem_db):
@@ -842,7 +837,7 @@ def test_iter_single_row(mem_db):
     # With skip_last_batch=False, should get one batch
     batches = list(perm.iter(10, skip_last_batch=False))
     assert len(batches) == 1
-    assert len(batches[0]["id"]) == 1
+    assert len(batches[0]) == 1
 
     # With skip_last_batch=True, should skip the single row (since it's < batch_size)
     batches_skip = list(perm.iter(10, skip_last_batch=True))
@@ -860,8 +855,7 @@ def test_identity_permutation(mem_db):
 
     batches = list(permutation.iter(10, skip_last_batch=False))
     assert len(batches) == 1
-    assert len(batches[0]["id"]) == 10
-    assert len(batches[0]["value"]) == 10
+    assert len(batches[0]) == 10
 
     permutation = permutation.remove_columns(["value"])
     assert permutation.num_columns == 1
@@ -904,10 +898,10 @@ def test_transform_fn(mem_db):
     py_result = list(permutation.with_format("python").iter(10, skip_last_batch=False))[
         0
     ]
-    assert len(py_result) == 2
-    assert len(py_result["id"]) == 10
-    assert len(py_result["value"]) == 10
-    assert isinstance(py_result, dict)
+    assert len(py_result) == 10
+    assert "id" in py_result[0]
+    assert "value" in py_result[0]
+    assert isinstance(py_result, list)
 
     try:
         import torch
@@ -915,9 +909,11 @@ def test_transform_fn(mem_db):
         torch_result = list(
             permutation.with_format("torch").iter(10, skip_last_batch=False)
         )[0]
-        assert torch_result.shape == (2, 10)
-        assert torch_result.dtype == torch.int64
-        assert isinstance(torch_result, torch.Tensor)
+        assert isinstance(torch_result, list)
+        assert len(torch_result) == 10
+        assert isinstance(torch_result[0], torch.Tensor)
+        assert torch_result[0].shape == (2,)
+        assert torch_result[0].dtype == torch.int64
     except ImportError:
         # Skip check if torch is not installed
         pass
@@ -945,3 +941,113 @@ def test_custom_transform(mem_db):
     batch = batches[0]
 
     assert batch == pa.record_batch([range(10)], ["id"])
+
+
+def test_getitems_basic(some_permutation: Permutation):
+    """Test __getitems__ returns correct rows by offset."""
+    result = some_permutation.__getitems__([0, 1, 2])
+    assert isinstance(result, list)
+    assert "id" in result[0]
+    assert "value" in result[0]
+    assert len(result) == 3
+
+
+def test_getitems_single_index(some_permutation: Permutation):
+    """Test __getitems__ with a single index."""
+    result = some_permutation.__getitems__([0])
+    assert len(result) == 1
+
+
+def test_getitems_preserves_order(some_permutation: Permutation):
+    """Test __getitems__ returns rows in the requested order."""
+    # Get rows in forward order
+    forward = some_permutation.__getitems__([0, 1, 2, 3, 4])
+    # Get the same rows in reverse order
+    reverse = some_permutation.__getitems__([4, 3, 2, 1, 0])
+
+    assert [r["id"] for r in forward] == list(reversed([r["id"] for r in reverse]))
+    assert [r["value"] for r in forward] == list(
+        reversed([r["value"] for r in reverse])
+    )
+
+
+def test_getitems_non_contiguous(some_permutation: Permutation):
+    """Test __getitems__ with non-contiguous indices."""
+    result = some_permutation.__getitems__([0, 10, 50, 100, 500])
+    assert len(result) == 5
+
+    # Each id/value pair should match what we'd get individually
+    for i, offset in enumerate([0, 10, 50, 100, 500]):
+        single = some_permutation.__getitems__([offset])
+        assert result[i]["id"] == single[0]["id"]
+        assert result[i]["value"] == single[0]["value"]
+
+
+def test_getitems_with_column_selection(some_permutation: Permutation):
+    """Test __getitems__ respects column selection."""
+    id_only = some_permutation.select_columns(["id"])
+    result = id_only.__getitems__([0, 1, 2])
+    assert "id" in result[0]
+    assert "value" not in result[0]
+    assert len(result) == 3
+
+
+def test_getitems_with_column_rename(some_permutation: Permutation):
+    """Test __getitems__ respects column renames."""
+    renamed = some_permutation.rename_column("value", "data")
+    result = renamed.__getitems__([0, 1])
+    assert "data" in result[0]
+    assert "value" not in result[0]
+    assert len(result) == 2
+
+
+def test_getitems_with_format(some_permutation: Permutation):
+    """Test __getitems__ applies the transform function."""
+    arrow_perm = some_permutation.with_format("arrow")
+    result = arrow_perm.__getitems__([0, 1, 2])
+    assert isinstance(result, pa.RecordBatch)
+    assert result.num_rows == 3
+
+
+def test_getitems_with_custom_transform(some_permutation: Permutation):
+    """Test __getitems__ with a custom transform."""
+
+    def transform(batch: pa.RecordBatch) -> list:
+        return batch.column("id").to_pylist()
+
+    custom = some_permutation.with_transform(transform)
+    result = custom.__getitems__([0, 1, 2])
+    assert isinstance(result, list)
+    assert len(result) == 3
+
+
+def test_getitems_identity_permutation(mem_db):
+    """Test __getitems__ on an identity permutation."""
+    tbl = mem_db.create_table(
+        "test_table", pa.table({"id": range(10), "value": range(10)})
+    )
+    perm = Permutation.identity(tbl)
+
+    result = perm.__getitems__([0, 5, 9])
+    assert [r["id"] for r in result] == [0, 5, 9]
+    assert [r["value"] for r in result] == [0, 5, 9]
+
+
+def test_getitems_with_limit_offset(some_permutation: Permutation):
+    """Test __getitems__ on a permutation with skip/take applied."""
+    limited = some_permutation.with_skip(100).with_take(200)
+
+    # Should be able to access offsets within the limited range
+    result = limited.__getitems__([0, 1, 199])
+    assert len(result) == 3
+
+    # The first item of the limited permutation should match offset 100 of original
+    full_result = some_permutation.__getitems__([100])
+    limited_result = limited.__getitems__([0])
+    assert limited_result[0]["id"] == full_result[0]["id"]
+
+
+def test_getitems_invalid_offset(some_permutation: Permutation):
+    """Test __getitems__ with an out-of-range offset raises an error."""
+    with pytest.raises(Exception):
+        some_permutation.__getitems__([999999])

python/python/tests/test_torch.py

@@ -4,6 +4,7 @@
 import pyarrow as pa
 import pytest
 from lancedb.util import tbl_to_tensor
+from lancedb.permutation import Permutation
 
 torch = pytest.importorskip("torch")
 
@@ -16,3 +17,26 @@ def test_table_dataloader(mem_db):
     for batch in dataloader:
         assert batch.size(0) == 1
         assert batch.size(1) == 10
+
+
+def test_permutation_dataloader(mem_db):
+    table = mem_db.create_table("test_table", pa.table({"a": range(1000)}))
+
+    permutation = Permutation.identity(table)
+    dataloader = torch.utils.data.DataLoader(permutation, batch_size=10, shuffle=True)
+    for batch in dataloader:
+        assert batch["a"].size(0) == 10
+
+    permutation = permutation.with_format("torch")
+    dataloader = torch.utils.data.DataLoader(permutation, batch_size=10, shuffle=True)
+    for batch in dataloader:
+        assert batch.size(0) == 10
+        assert batch.size(1) == 1
+
+    permutation = permutation.with_format("torch_col")
+    dataloader = torch.utils.data.DataLoader(
+        permutation, collate_fn=lambda x: x, batch_size=10, shuffle=True
+    )
+    for batch in dataloader:
+        assert batch.size(0) == 1
+        assert batch.size(1) == 10

python/src/connection.rs

@@ -121,7 +121,8 @@ impl Connection {
 
         let mode = Self::parse_create_mode_str(mode)?;
 
-        let batches = ArrowArrayStreamReader::from_pyarrow_bound(&data)?;
+        let batches: Box<dyn arrow::array::RecordBatchReader + Send> =
+            Box::new(ArrowArrayStreamReader::from_pyarrow_bound(&data)?);
 
         let mut builder = inner.create_table(name, batches).mode(mode);
 

python/src/permutation.rs

@@ -6,7 +6,7 @@ use std::sync::{Arc, Mutex};
 use crate::{
     arrow::RecordBatchStream, connection::Connection, error::PythonErrorExt, table::Table,
 };
-use arrow::pyarrow::ToPyArrow;
+use arrow::pyarrow::{PyArrowType, ToPyArrow};
 use lancedb::{
     dataloader::permutation::{
         builder::{PermutationBuilder as LancePermutationBuilder, ShuffleStrategy},
@@ -328,4 +328,21 @@ impl PyPermutationReader {
             Ok(RecordBatchStream::new(stream))
         })
     }
+
+    #[pyo3(signature = (indices, *, selection=None))]
+    pub fn take_offsets<'py>(
+        slf: PyRef<'py, Self>,
+        indices: Vec<u64>,
+        selection: Option<Bound<'py, PyAny>>,
+    ) -> PyResult<Bound<'py, PyAny>> {
+        let selection = Self::parse_selection(selection)?;
+        let reader = slf.reader.clone();
+        future_into_py(slf.py(), async move {
+            let batch = reader
+                .take_offsets(&indices, selection)
+                .await
+                .infer_error()?;
+            Ok(PyArrowType(batch))
+        })
+    }
 }

python/src/table.rs

@@ -296,7 +296,8 @@ impl Table {
         data: Bound<'_, PyAny>,
         mode: String,
     ) -> PyResult<Bound<'a, PyAny>> {
-        let batches = ArrowArrayStreamReader::from_pyarrow_bound(&data)?;
+        let batches: Box<dyn arrow::array::RecordBatchReader + Send> =
+            Box::new(ArrowArrayStreamReader::from_pyarrow_bound(&data)?);
         let mut op = self_.inner_ref()?.add(batches);
         if mode == "append" {
             op = op.mode(AddDataMode::Append);

rust/lancedb/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb"
-version = "0.26.1"
+version = "0.27.0-beta.0"
 edition.workspace = true
 description = "LanceDB: A serverless, low-latency vector database for AI applications"
 license.workspace = true

rust/lancedb/examples/bedrock.rs

@@ -3,13 +3,12 @@
 
 use std::{iter::once, sync::Arc};
 
-use arrow_array::{Float64Array, Int32Array, RecordBatch, RecordBatchIterator, StringArray};
+use arrow_array::{Float64Array, Int32Array, RecordBatch, StringArray};
 use arrow_schema::{DataType, Field, Schema};
 use aws_config::Region;
 use aws_sdk_bedrockruntime::Client;
 use futures::StreamExt;
 use lancedb::{
-    arrow::IntoArrow,
     connect,
     embeddings::{bedrock::BedrockEmbeddingFunction, EmbeddingDefinition, EmbeddingFunction},
     query::{ExecutableQuery, QueryBase},
@@ -67,7 +66,7 @@ async fn main() -> Result<()> {
     Ok(())
 }
 
-fn make_data() -> impl IntoArrow {
+fn make_data() -> RecordBatch {
     let schema = Schema::new(vec![
         Field::new("id", DataType::Int32, true),
         Field::new("text", DataType::Utf8, false),
@@ -83,10 +82,9 @@ fn make_data() -> impl IntoArrow {
     ]);
     let price = Float64Array::from(vec![10.0, 50.0, 100.0, 30.0]);
     let schema = Arc::new(schema);
-    let rb = RecordBatch::try_new(
+    RecordBatch::try_new(
         schema.clone(),
         vec![Arc::new(id), Arc::new(text), Arc::new(price)],
     )
-    .unwrap();
-    Box::new(RecordBatchIterator::new(vec![Ok(rb)], schema))
+    .unwrap()
 }

rust/lancedb/examples/full_text_search.rs

@@ -3,12 +3,13 @@
 
 use std::sync::Arc;
 
-use arrow_array::{Int32Array, RecordBatch, RecordBatchIterator, RecordBatchReader, StringArray};
+use arrow_array::{Int32Array, RecordBatch, RecordBatchIterator, StringArray};
 use arrow_schema::{DataType, Field, Schema};
 
 use futures::TryStreamExt;
 use lance_index::scalar::FullTextSearchQuery;
 use lancedb::connection::Connection;
+
 use lancedb::index::scalar::FtsIndexBuilder;
 use lancedb::index::Index;
 use lancedb::query::{ExecutableQuery, QueryBase};
@@ -29,7 +30,7 @@ async fn main() -> Result<()> {
     Ok(())
 }
 
-fn create_some_records() -> Result<Box<dyn RecordBatchReader + Send>> {
+fn create_some_records() -> Result<Box<dyn arrow_array::RecordBatchReader + Send>> {
     const TOTAL: usize = 1000;
 
     let schema = Arc::new(Schema::new(vec![
@@ -66,7 +67,7 @@ fn create_some_records() -> Result<Box<dyn RecordBatchReader + Send>> {
 }
 
 async fn create_table(db: &Connection) -> Result<Table> {
-    let initial_data: Box<dyn RecordBatchReader + Send> = create_some_records()?;
+    let initial_data = create_some_records()?;
     let tbl = db.create_table("my_table", initial_data).execute().await?;
     Ok(tbl)
 }

rust/lancedb/examples/hybrid_search.rs

@@ -1,14 +1,13 @@
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
-use arrow_array::{RecordBatch, RecordBatchIterator, StringArray};
+use arrow_array::{RecordBatch, StringArray};
 use arrow_schema::{DataType, Field, Schema};
 use futures::TryStreamExt;
 use lance_index::scalar::FullTextSearchQuery;
 use lancedb::index::scalar::FtsIndexBuilder;
 use lancedb::index::Index;
 use lancedb::{
-    arrow::IntoArrow,
     connect,
     embeddings::{
         sentence_transformers::SentenceTransformersEmbeddings, EmbeddingDefinition,
@@ -70,7 +69,7 @@ async fn main() -> Result<()> {
     Ok(())
 }
 
-fn make_data() -> impl IntoArrow {
+fn make_data() -> RecordBatch {
     let schema = Schema::new(vec![Field::new("facts", DataType::Utf8, false)]);
 
     let facts = StringArray::from_iter_values(vec![
@@ -101,8 +100,7 @@ fn make_data() -> impl IntoArrow {
         "The first chatbot was ELIZA, created in the 1960s.",
     ]);
     let schema = Arc::new(schema);
-    let rb = RecordBatch::try_new(schema.clone(), vec![Arc::new(facts)]).unwrap();
-    Box::new(RecordBatchIterator::new(vec![Ok(rb)], schema))
+    RecordBatch::try_new(schema.clone(), vec![Arc::new(facts)]).unwrap()
 }
 
 async fn create_index(table: &Table) -> Result<()> {

rust/lancedb/examples/ivf_pq.rs

@@ -8,13 +8,12 @@
 use std::sync::Arc;
 
 use arrow_array::types::Float32Type;
-use arrow_array::{
-    FixedSizeListArray, Int32Array, RecordBatch, RecordBatchIterator, RecordBatchReader,
-};
+use arrow_array::{FixedSizeListArray, Int32Array, RecordBatch, RecordBatchIterator};
 use arrow_schema::{DataType, Field, Schema};
 
 use futures::TryStreamExt;
 use lancedb::connection::Connection;
+
 use lancedb::index::vector::IvfPqIndexBuilder;
 use lancedb::index::Index;
 use lancedb::query::{ExecutableQuery, QueryBase};
@@ -34,7 +33,7 @@ async fn main() -> Result<()> {
     Ok(())
 }
 
-fn create_some_records() -> Result<Box<dyn RecordBatchReader + Send>> {
+fn create_some_records() -> Result<Box<dyn arrow_array::RecordBatchReader + Send>> {
     const TOTAL: usize = 1000;
     const DIM: usize = 128;
 
@@ -73,9 +72,9 @@ fn create_some_records() -> Result<Box<dyn RecordBatchReader + Send>> {
 }
 
 async fn create_table(db: &Connection) -> Result<Table> {
-    let initial_data: Box<dyn RecordBatchReader + Send> = create_some_records()?;
+    let initial_data = create_some_records()?;
     let tbl = db
-        .create_table("my_table", Box::new(initial_data))
+        .create_table("my_table", initial_data)
         .execute()
         .await
         .unwrap();

rust/lancedb/examples/openai.rs

@@ -5,11 +5,9 @@
 
 use std::{iter::once, sync::Arc};
 
-use arrow_array::{Float64Array, Int32Array, RecordBatch, RecordBatchIterator, StringArray};
-use arrow_schema::{DataType, Field, Schema};
+use arrow_array::{RecordBatch, StringArray};
 use futures::StreamExt;
 use lancedb::{
-    arrow::IntoArrow,
     connect,
     embeddings::{openai::OpenAIEmbeddingFunction, EmbeddingDefinition, EmbeddingFunction},
     query::{ExecutableQuery, QueryBase},
@@ -64,26 +62,20 @@ async fn main() -> Result<()> {
 }
 // --8<-- [end:openai_embeddings]
 
-fn make_data() -> impl IntoArrow {
-    let schema = Schema::new(vec![
-        Field::new("id", DataType::Int32, true),
-        Field::new("text", DataType::Utf8, false),
-        Field::new("price", DataType::Float64, false),
-    ]);
-
-    let id = Int32Array::from(vec![1, 2, 3, 4]);
-    let text = StringArray::from_iter_values(vec![
-        "Black T-Shirt",
-        "Leather Jacket",
-        "Winter Parka",
-        "Hooded Sweatshirt",
-    ]);
-    let price = Float64Array::from(vec![10.0, 50.0, 100.0, 30.0]);
-    let schema = Arc::new(schema);
-    let rb = RecordBatch::try_new(
-        schema.clone(),
-        vec![Arc::new(id), Arc::new(text), Arc::new(price)],
+fn make_data() -> RecordBatch {
+    arrow_array::record_batch!(
+        ("id", Int32, [1, 2, 3, 4]),
+        (
+            "text",
+            Utf8,
+            [
+                "Black T-Shirt",
+                "Leather Jacket",
+                "Winter Parka",
+                "Hooded Sweatshirt"
+            ]
+        ),
+        ("price", Float64, [10.0, 50.0, 100.0, 30.0])
     )
-    .unwrap();
-    Box::new(RecordBatchIterator::new(vec![Ok(rb)], schema))
+    .unwrap()
 }

rust/lancedb/examples/sentence_transformers.rs

@@ -3,11 +3,10 @@
 
 use std::{iter::once, sync::Arc};
 
-use arrow_array::{RecordBatch, RecordBatchIterator, StringArray};
+use arrow_array::{RecordBatch, StringArray};
 use arrow_schema::{DataType, Field, Schema};
 use futures::StreamExt;
 use lancedb::{
-    arrow::IntoArrow,
     connect,
     embeddings::{
         sentence_transformers::SentenceTransformersEmbeddings, EmbeddingDefinition,
@@ -59,7 +58,7 @@ async fn main() -> Result<()> {
     Ok(())
 }
 
-fn make_data() -> impl IntoArrow {
+fn make_data() -> RecordBatch {
     let schema = Schema::new(vec![Field::new("facts", DataType::Utf8, false)]);
 
     let facts = StringArray::from_iter_values(vec![
@@ -90,6 +89,5 @@ fn make_data() -> impl IntoArrow {
         "The first chatbot was ELIZA, created in the 1960s.",
     ]);
     let schema = Arc::new(schema);
-    let rb = RecordBatch::try_new(schema.clone(), vec![Arc::new(facts)]).unwrap();
-    Box::new(RecordBatchIterator::new(vec![Ok(rb)], schema))
+    RecordBatch::try_new(schema.clone(), vec![Arc::new(facts)]).unwrap()
 }

rust/lancedb/examples/simple.rs

@@ -8,11 +8,9 @@
 use std::sync::Arc;
 
 use arrow_array::types::Float32Type;
-use arrow_array::{FixedSizeListArray, Int32Array, RecordBatch, RecordBatchIterator};
+use arrow_array::{FixedSizeListArray, Int32Array, RecordBatch};
 use arrow_schema::{DataType, Field, Schema};
 use futures::TryStreamExt;
-
-use lancedb::arrow::IntoArrow;
 use lancedb::connection::Connection;
 use lancedb::index::Index;
 use lancedb::query::{ExecutableQuery, QueryBase};
@@ -59,7 +57,7 @@ async fn open_with_existing_tbl() -> Result<()> {
     Ok(())
 }
 
-fn create_some_records() -> Result<impl IntoArrow> {
+fn create_some_records() -> Result<RecordBatch> {
     const TOTAL: usize = 1000;
     const DIM: usize = 128;
 
@@ -76,25 +74,18 @@ fn create_some_records() -> Result<impl IntoArrow> {
     ]));
 
     // Create a RecordBatch stream.
-    let batches = RecordBatchIterator::new(
-        vec![RecordBatch::try_new(
-            schema.clone(),
-            vec![
-                Arc::new(Int32Array::from_iter_values(0..TOTAL as i32)),
-                Arc::new(
-                    FixedSizeListArray::from_iter_primitive::<Float32Type, _, _>(
-                        (0..TOTAL).map(|_| Some(vec![Some(1.0); DIM])),
-                        DIM as i32,
-                    ),
-                ),
-            ],
-        )
-        .unwrap()]
-        .into_iter()
-        .map(Ok),
+    Ok(RecordBatch::try_new(
         schema.clone(),
-    );
-    Ok(Box::new(batches))
+        vec![
+            Arc::new(Int32Array::from_iter_values(0..TOTAL as i32)),
+            Arc::new(
+                FixedSizeListArray::from_iter_primitive::<Float32Type, _, _>(
+                    (0..TOTAL).map(|_| Some(vec![Some(1.0); DIM])),
+                    DIM as i32,
+                ),
+            ),
+        ],
+    )?)
 }
 
 async fn create_table(db: &Connection) -> Result<LanceDbTable> {

rust/lancedb/src/connection.rs

@@ -6,8 +6,8 @@
 use std::collections::HashMap;
 use std::sync::Arc;
 
-use arrow_array::RecordBatchReader;
-use arrow_schema::{Field, SchemaRef};
+use arrow_array::RecordBatch;
+use arrow_schema::SchemaRef;
 use lance::dataset::ReadParams;
 use lance_namespace::models::{
     CreateNamespaceRequest, CreateNamespaceResponse, DescribeNamespaceRequest,
@@ -17,24 +17,20 @@ use lance_namespace::models::{
 #[cfg(feature = "aws")]
 use object_store::aws::AwsCredential;
 
-use crate::arrow::{IntoArrow, IntoArrowStream, SendableRecordBatchStream};
-use crate::database::listing::{
-    ListingDatabase, OPT_NEW_TABLE_STORAGE_VERSION, OPT_NEW_TABLE_V2_MANIFEST_PATHS,
-};
+use crate::connection::create_table::CreateTableBuilder;
+use crate::data::scannable::Scannable;
+use crate::database::listing::ListingDatabase;
 use crate::database::{
-    CloneTableRequest, CreateTableData, CreateTableMode, CreateTableRequest, Database,
-    DatabaseOptions, OpenTableRequest, ReadConsistency, TableNamesRequest,
-};
-use crate::embeddings::{
-    EmbeddingDefinition, EmbeddingFunction, EmbeddingRegistry, MemoryRegistry, WithEmbeddings,
+    CloneTableRequest, Database, DatabaseOptions, OpenTableRequest, ReadConsistency,
+    TableNamesRequest,
 };
+use crate::embeddings::{EmbeddingRegistry, MemoryRegistry};
 use crate::error::{Error, Result};
 #[cfg(feature = "remote")]
 use crate::remote::{
     client::ClientConfig,
     db::{OPT_REMOTE_API_KEY, OPT_REMOTE_HOST_OVERRIDE, OPT_REMOTE_REGION},
 };
-use crate::table::{TableDefinition, WriteOptions};
 use crate::Table;
 use lance::io::ObjectStoreParams;
 pub use lance_encoding::version::LanceFileVersion;
@@ -42,6 +38,8 @@ pub use lance_encoding::version::LanceFileVersion;
 use lance_io::object_store::StorageOptions;
 use lance_io::object_store::{StorageOptionsAccessor, StorageOptionsProvider};
 
+mod create_table;
+
 fn merge_storage_options(
     store_params: &mut ObjectStoreParams,
     pairs: impl IntoIterator<Item = (String, String)>,
@@ -116,337 +114,6 @@ impl TableNamesBuilder {
     }
 }
 
-pub struct NoData {}
-
-impl IntoArrow for NoData {
-    fn into_arrow(self) -> Result<Box<dyn arrow_array::RecordBatchReader + Send>> {
-        unreachable!("NoData should never be converted to Arrow")
-    }
-}
-
-// Stores the value given from the initial CreateTableBuilder::new call
-// and defers errors until `execute` is called
-enum CreateTableBuilderInitialData {
-    None,
-    Iterator(Result<Box<dyn RecordBatchReader + Send>>),
-    Stream(Result<SendableRecordBatchStream>),
-}
-
-/// A builder for configuring a [`Connection::create_table`] operation
-pub struct CreateTableBuilder<const HAS_DATA: bool> {
-    parent: Arc<dyn Database>,
-    embeddings: Vec<(EmbeddingDefinition, Arc<dyn EmbeddingFunction>)>,
-    embedding_registry: Arc<dyn EmbeddingRegistry>,
-    request: CreateTableRequest,
-    // This is a bit clumsy but we defer errors until `execute` is called
-    // to maintain backwards compatibility
-    data: CreateTableBuilderInitialData,
-}
-
-// Builder methods that only apply when we have initial data
-impl CreateTableBuilder<true> {
-    fn new<T: IntoArrow>(
-        parent: Arc<dyn Database>,
-        name: String,
-        data: T,
-        embedding_registry: Arc<dyn EmbeddingRegistry>,
-    ) -> Self {
-        let dummy_schema = Arc::new(arrow_schema::Schema::new(Vec::<Field>::default()));
-        Self {
-            parent,
-            request: CreateTableRequest::new(
-                name,
-                CreateTableData::Empty(TableDefinition::new_from_schema(dummy_schema)),
-            ),
-            embeddings: Vec::new(),
-            embedding_registry,
-            data: CreateTableBuilderInitialData::Iterator(data.into_arrow()),
-        }
-    }
-
-    fn new_streaming<T: IntoArrowStream>(
-        parent: Arc<dyn Database>,
-        name: String,
-        data: T,
-        embedding_registry: Arc<dyn EmbeddingRegistry>,
-    ) -> Self {
-        let dummy_schema = Arc::new(arrow_schema::Schema::new(Vec::<Field>::default()));
-        Self {
-            parent,
-            request: CreateTableRequest::new(
-                name,
-                CreateTableData::Empty(TableDefinition::new_from_schema(dummy_schema)),
-            ),
-            embeddings: Vec::new(),
-            embedding_registry,
-            data: CreateTableBuilderInitialData::Stream(data.into_arrow()),
-        }
-    }
-
-    /// Execute the create table operation
-    pub async fn execute(self) -> Result<Table> {
-        let embedding_registry = self.embedding_registry.clone();
-        let parent = self.parent.clone();
-        let request = self.into_request()?;
-        Ok(Table::new_with_embedding_registry(
-            parent.create_table(request).await?,
-            parent,
-            embedding_registry,
-        ))
-    }
-
-    fn into_request(self) -> Result<CreateTableRequest> {
-        if self.embeddings.is_empty() {
-            match self.data {
-                CreateTableBuilderInitialData::Iterator(maybe_iter) => {
-                    let data = maybe_iter?;
-                    Ok(CreateTableRequest {
-                        data: CreateTableData::Data(data),
-                        ..self.request
-                    })
-                }
-                CreateTableBuilderInitialData::None => {
-                    unreachable!("No data provided for CreateTableBuilder<true>")
-                }
-                CreateTableBuilderInitialData::Stream(maybe_stream) => {
-                    let data = maybe_stream?;
-                    Ok(CreateTableRequest {
-                        data: CreateTableData::StreamingData(data),
-                        ..self.request
-                    })
-                }
-            }
-        } else {
-            let CreateTableBuilderInitialData::Iterator(maybe_iter) = self.data else {
-                return Err(Error::NotSupported { message: "Creating a table with embeddings is currently not support when the input is streaming".to_string() });
-            };
-            let data = maybe_iter?;
-            let data = Box::new(WithEmbeddings::new(data, self.embeddings));
-            Ok(CreateTableRequest {
-                data: CreateTableData::Data(data),
-                ..self.request
-            })
-        }
-    }
-}
-
-// Builder methods that only apply when we do not have initial data
-impl CreateTableBuilder<false> {
-    fn new(
-        parent: Arc<dyn Database>,
-        name: String,
-        schema: SchemaRef,
-        embedding_registry: Arc<dyn EmbeddingRegistry>,
-    ) -> Self {
-        let table_definition = TableDefinition::new_from_schema(schema);
-        Self {
-            parent,
-            request: CreateTableRequest::new(name, CreateTableData::Empty(table_definition)),
-            data: CreateTableBuilderInitialData::None,
-            embeddings: Vec::default(),
-            embedding_registry,
-        }
-    }
-
-    /// Execute the create table operation
-    pub async fn execute(self) -> Result<Table> {
-        let parent = self.parent.clone();
-        let embedding_registry = self.embedding_registry.clone();
-        let request = self.into_request()?;
-        Ok(Table::new_with_embedding_registry(
-            parent.create_table(request).await?,
-            parent,
-            embedding_registry,
-        ))
-    }
-
-    fn into_request(self) -> Result<CreateTableRequest> {
-        if self.embeddings.is_empty() {
-            return Ok(self.request);
-        }
-
-        let CreateTableData::Empty(table_def) = self.request.data else {
-            unreachable!("CreateTableBuilder<false> should always have Empty data")
-        };
-
-        let schema = table_def.schema.clone();
-        let empty_batch = arrow_array::RecordBatch::new_empty(schema.clone());
-
-        let reader = Box::new(std::iter::once(Ok(empty_batch)).collect::<Vec<_>>());
-        let reader = arrow_array::RecordBatchIterator::new(reader.into_iter(), schema);
-        let with_embeddings = WithEmbeddings::new(reader, self.embeddings);
-        let table_definition = with_embeddings.table_definition()?;
-
-        Ok(CreateTableRequest {
-            data: CreateTableData::Empty(table_definition),
-            ..self.request
-        })
-    }
-}
-
-impl<const HAS_DATA: bool> CreateTableBuilder<HAS_DATA> {
-    /// Set the mode for creating the table
-    ///
-    /// This controls what happens if a table with the given name already exists
-    pub fn mode(mut self, mode: CreateTableMode) -> Self {
-        self.request.mode = mode;
-        self
-    }
-
-    /// Apply the given write options when writing the initial data
-    pub fn write_options(mut self, write_options: WriteOptions) -> Self {
-        self.request.write_options = write_options;
-        self
-    }
-
-    /// Set an option for the storage layer.
-    ///
-    /// Options already set on the connection will be inherited by the table,
-    /// but can be overridden here.
-    ///
-    /// See available options at <https://lancedb.com/docs/storage/>
-    pub fn storage_option(mut self, key: impl Into<String>, value: impl Into<String>) -> Self {
-        let store_params = self
-            .request
-            .write_options
-            .lance_write_params
-            .get_or_insert(Default::default())
-            .store_params
-            .get_or_insert(Default::default());
-        merge_storage_options(store_params, [(key.into(), value.into())]);
-        self
-    }
-
-    /// Set multiple options for the storage layer.
-    ///
-    /// Options already set on the connection will be inherited by the table,
-    /// but can be overridden here.
-    ///
-    /// See available options at <https://lancedb.com/docs/storage/>
-    pub fn storage_options(
-        mut self,
-        pairs: impl IntoIterator<Item = (impl Into<String>, impl Into<String>)>,
-    ) -> Self {
-        let store_params = self
-            .request
-            .write_options
-            .lance_write_params
-            .get_or_insert(Default::default())
-            .store_params
-            .get_or_insert(Default::default());
-        let updates = pairs
-            .into_iter()
-            .map(|(key, value)| (key.into(), value.into()));
-        merge_storage_options(store_params, updates);
-        self
-    }
-
-    /// Add an embedding definition to the table.
-    ///
-    /// The `embedding_name` must match the name of an embedding function that
-    /// was previously registered with the connection's [`EmbeddingRegistry`].
-    pub fn add_embedding(mut self, definition: EmbeddingDefinition) -> Result<Self> {
-        // Early verification of the embedding name
-        let embedding_func = self
-            .embedding_registry
-            .get(&definition.embedding_name)
-            .ok_or_else(|| Error::EmbeddingFunctionNotFound {
-                name: definition.embedding_name.clone(),
-                reason: "No embedding function found in the connection's embedding_registry"
-                    .to_string(),
-            })?;
-
-        self.embeddings.push((definition, embedding_func));
-        Ok(self)
-    }
-
-    /// Set whether to use V2 manifest paths for the table. (default: false)
-    ///
-    /// These paths provide more efficient opening of tables with many
-    /// versions on object stores.
-    ///
-    /// <div class="warning">Turning this on will make the dataset unreadable
-    /// for older versions of LanceDB (prior to 0.10.0).</div>
-    ///
-    /// To migrate an existing dataset, instead use the
-    /// [[NativeTable::migrate_manifest_paths_v2]].
-    ///
-    /// This has no effect in LanceDB Cloud.
-    #[deprecated(since = "0.15.1", note = "Use `database_options` instead")]
-    pub fn enable_v2_manifest_paths(mut self, use_v2_manifest_paths: bool) -> Self {
-        let store_params = self
-            .request
-            .write_options
-            .lance_write_params
-            .get_or_insert_with(Default::default)
-            .store_params
-            .get_or_insert_with(Default::default);
-        let value = if use_v2_manifest_paths {
-            "true".to_string()
-        } else {
-            "false".to_string()
-        };
-        merge_storage_options(
-            store_params,
-            [(OPT_NEW_TABLE_V2_MANIFEST_PATHS.to_string(), value)],
-        );
-        self
-    }
-
-    /// Set the data storage version.
-    ///
-    /// The default is `LanceFileVersion::Stable`.
-    #[deprecated(since = "0.15.1", note = "Use `database_options` instead")]
-    pub fn data_storage_version(mut self, data_storage_version: LanceFileVersion) -> Self {
-        let store_params = self
-            .request
-            .write_options
-            .lance_write_params
-            .get_or_insert_with(Default::default)
-            .store_params
-            .get_or_insert_with(Default::default);
-        merge_storage_options(
-            store_params,
-            [(
-                OPT_NEW_TABLE_STORAGE_VERSION.to_string(),
-                data_storage_version.to_string(),
-            )],
-        );
-        self
-    }
-
-    /// Set the namespace for the table
-    pub fn namespace(mut self, namespace: Vec<String>) -> Self {
-        self.request.namespace = namespace;
-        self
-    }
-
-    /// Set a custom location for the table.
-    ///
-    /// If not set, the database will derive a location from its URI and the table name.
-    /// This is useful when integrating with namespace systems that manage table locations.
-    pub fn location(mut self, location: impl Into<String>) -> Self {
-        self.request.location = Some(location.into());
-        self
-    }
-
-    /// Set a storage options provider for automatic credential refresh.
-    ///
-    /// This allows tables to automatically refresh cloud storage credentials
-    /// when they expire, enabling long-running operations on remote storage.
-    pub fn storage_options_provider(mut self, provider: Arc<dyn StorageOptionsProvider>) -> Self {
-        let store_params = self
-            .request
-            .write_options
-            .lance_write_params
-            .get_or_insert(Default::default())
-            .store_params
-            .get_or_insert(Default::default());
-        set_storage_options_provider(store_params, provider);
-        self
-    }
-}
-
 #[derive(Clone, Debug)]
 pub struct OpenTableBuilder {
     parent: Arc<dyn Database>,
@@ -684,35 +351,17 @@ impl Connection {
     ///
     /// * `name` - The name of the table
     /// * `initial_data` - The initial data to write to the table
-    pub fn create_table<T: IntoArrow>(
+    pub fn create_table<T: Scannable + 'static>(
         &self,
         name: impl Into<String>,
         initial_data: T,
-    ) -> CreateTableBuilder<true> {
-        CreateTableBuilder::<true>::new(
+    ) -> CreateTableBuilder {
+        let initial_data = Box::new(initial_data);
+        CreateTableBuilder::new(
             self.internal.clone(),
+            self.embedding_registry.clone(),
             name.into(),
             initial_data,
-            self.embedding_registry.clone(),
-        )
-    }
-
-    /// Create a new table from a stream of data
-    ///
-    /// # Parameters
-    ///
-    /// * `name` - The name of the table
-    /// * `initial_data` - The initial data to write to the table
-    pub fn create_table_streaming<T: IntoArrowStream>(
-        &self,
-        name: impl Into<String>,
-        initial_data: T,
-    ) -> CreateTableBuilder<true> {
-        CreateTableBuilder::<true>::new_streaming(
-            self.internal.clone(),
-            name.into(),
-            initial_data,
-            self.embedding_registry.clone(),
         )
     }
 
@@ -726,13 +375,9 @@ impl Connection {
         &self,
         name: impl Into<String>,
         schema: SchemaRef,
-    ) -> CreateTableBuilder<false> {
-        CreateTableBuilder::<false>::new(
-            self.internal.clone(),
-            name.into(),
-            schema,
-            self.embedding_registry.clone(),
-        )
+    ) -> CreateTableBuilder {
+        let empty_batch = RecordBatch::new_empty(schema);
+        self.create_table(name, empty_batch)
     }
 
     /// Open an existing table in the database
@@ -1349,20 +994,11 @@ mod test_utils {
 
 #[cfg(test)]
 mod tests {
-    use crate::database::listing::{ListingDatabaseOptions, NewTableConfig};
-    use crate::query::QueryBase;
-    use crate::query::{ExecutableQuery, QueryExecutionOptions};
-    use crate::test_utils::connection::new_test_connection;
-    use arrow::compute::concat_batches;
-    use arrow_array::RecordBatchReader;
     use arrow_schema::{DataType, Field, Schema};
-    use datafusion_physical_plan::stream::RecordBatchStreamAdapter;
-    use futures::{stream, TryStreamExt};
-    use lance_core::error::{ArrowResult, DataFusionResult};
     use lance_testing::datagen::{BatchGenerator, IncrementingInt32};
     use tempfile::tempdir;
 
-    use crate::arrow::SimpleRecordBatchStream;
+    use crate::test_utils::connection::new_test_connection;
 
     use super::*;
 
@@ -1478,139 +1114,6 @@ mod tests {
         assert_eq!(tables, vec!["table1".to_owned()]);
     }
 
-    fn make_data() -> Box<dyn RecordBatchReader + Send + 'static> {
-        let id = Box::new(IncrementingInt32::new().named("id".to_string()));
-        Box::new(BatchGenerator::new().col(id).batches(10, 2000))
-    }
-
-    #[tokio::test]
-    async fn test_create_table_v2() {
-        let tmp_dir = tempdir().unwrap();
-        let uri = tmp_dir.path().to_str().unwrap();
-        let db = connect(uri)
-            .database_options(&ListingDatabaseOptions {
-                new_table_config: NewTableConfig {
-                    data_storage_version: Some(LanceFileVersion::Legacy),
-                    ..Default::default()
-                },
-                ..Default::default()
-            })
-            .execute()
-            .await
-            .unwrap();
-
-        let tbl = db
-            .create_table("v1_test", make_data())
-            .execute()
-            .await
-            .unwrap();
-
-        // In v1 the row group size will trump max_batch_length
-        let batches = tbl
-            .query()
-            .limit(20000)
-            .execute_with_options(QueryExecutionOptions {
-                max_batch_length: 50000,
-                ..Default::default()
-            })
-            .await
-            .unwrap()
-            .try_collect::<Vec<_>>()
-            .await
-            .unwrap();
-        assert_eq!(batches.len(), 20);
-
-        let db = connect(uri)
-            .database_options(&ListingDatabaseOptions {
-                new_table_config: NewTableConfig {
-                    data_storage_version: Some(LanceFileVersion::Stable),
-                    ..Default::default()
-                },
-                ..Default::default()
-            })
-            .execute()
-            .await
-            .unwrap();
-
-        let tbl = db
-            .create_table("v2_test", make_data())
-            .execute()
-            .await
-            .unwrap();
-
-        // In v2 the page size is much bigger than 50k so we should get a single batch
-        let batches = tbl
-            .query()
-            .execute_with_options(QueryExecutionOptions {
-                max_batch_length: 50000,
-                ..Default::default()
-            })
-            .await
-            .unwrap()
-            .try_collect::<Vec<_>>()
-            .await
-            .unwrap();
-
-        assert_eq!(batches.len(), 1);
-    }
-
-    #[tokio::test]
-    async fn test_create_table_streaming() {
-        let tmp_dir = tempdir().unwrap();
-
-        let uri = tmp_dir.path().to_str().unwrap();
-        let db = connect(uri).execute().await.unwrap();
-
-        let batches = make_data().collect::<ArrowResult<Vec<_>>>().unwrap();
-
-        let schema = batches.first().unwrap().schema();
-        let one_batch = concat_batches(&schema, batches.iter()).unwrap();
-
-        let ldb_stream = stream::iter(batches.clone().into_iter().map(Result::Ok));
-        let ldb_stream: SendableRecordBatchStream =
-            Box::pin(SimpleRecordBatchStream::new(ldb_stream, schema.clone()));
-
-        let tbl1 = db
-            .create_table_streaming("one", ldb_stream)
-            .execute()
-            .await
-            .unwrap();
-
-        let df_stream = stream::iter(batches.into_iter().map(DataFusionResult::Ok));
-        let df_stream: datafusion_physical_plan::SendableRecordBatchStream =
-            Box::pin(RecordBatchStreamAdapter::new(schema.clone(), df_stream));
-
-        let tbl2 = db
-            .create_table_streaming("two", df_stream)
-            .execute()
-            .await
-            .unwrap();
-
-        let tbl1_data = tbl1
-            .query()
-            .execute()
-            .await
-            .unwrap()
-            .try_collect::<Vec<_>>()
-            .await
-            .unwrap();
-
-        let tbl1_data = concat_batches(&schema, tbl1_data.iter()).unwrap();
-        assert_eq!(tbl1_data, one_batch);
-
-        let tbl2_data = tbl2
-            .query()
-            .execute()
-            .await
-            .unwrap()
-            .try_collect::<Vec<_>>()
-            .await
-            .unwrap();
-
-        let tbl2_data = concat_batches(&schema, tbl2_data.iter()).unwrap();
-        assert_eq!(tbl2_data, one_batch);
-    }
-
     #[tokio::test]
     async fn drop_table() {
         let tc = new_test_connection().await.unwrap();
@@ -1640,41 +1143,6 @@ mod tests {
         assert_eq!(tables.len(), 0);
     }
 
-    #[tokio::test]
-    async fn test_create_table_already_exists() {
-        let tmp_dir = tempdir().unwrap();
-        let uri = tmp_dir.path().to_str().unwrap();
-        let db = connect(uri).execute().await.unwrap();
-        let schema = Arc::new(Schema::new(vec![Field::new("x", DataType::Int32, false)]));
-        db.create_empty_table("test", schema.clone())
-            .execute()
-            .await
-            .unwrap();
-        // TODO: None of the open table options are "inspectable" right now but once one is we
-        // should assert we are passing these options in correctly
-        db.create_empty_table("test", schema)
-            .mode(CreateTableMode::exist_ok(|mut req| {
-                req.index_cache_size = Some(16);
-                req
-            }))
-            .execute()
-            .await
-            .unwrap();
-        let other_schema = Arc::new(Schema::new(vec![Field::new("y", DataType::Int32, false)]));
-        assert!(db
-            .create_empty_table("test", other_schema.clone())
-            .execute()
-            .await
-            .is_err());
-        let overwritten = db
-            .create_empty_table("test", other_schema.clone())
-            .mode(CreateTableMode::Overwrite)
-            .execute()
-            .await
-            .unwrap();
-        assert_eq!(other_schema, overwritten.schema().await.unwrap());
-    }
-
     #[tokio::test]
     async fn test_clone_table() {
         let tmp_dir = tempdir().unwrap();
@@ -1685,7 +1153,8 @@ mod tests {
         let mut batch_gen = BatchGenerator::new()
             .col(Box::new(IncrementingInt32::new().named("id")))
             .col(Box::new(IncrementingInt32::new().named("value")));
-        let reader = batch_gen.batches(5, 100);
+        let reader: Box<dyn arrow_array::RecordBatchReader + Send> =
+            Box::new(batch_gen.batches(5, 100));
 
         let source_table = db
             .create_table("source_table", reader)
@@ -1720,128 +1189,4 @@ mod tests {
         let cloned_count = cloned_table.count_rows(None).await.unwrap();
         assert_eq!(source_count, cloned_count);
     }
-
-    #[tokio::test]
-    async fn test_create_empty_table_with_embeddings() {
-        use crate::embeddings::{EmbeddingDefinition, EmbeddingFunction};
-        use arrow_array::{
-            Array, FixedSizeListArray, Float32Array, RecordBatch, RecordBatchIterator, StringArray,
-        };
-        use std::borrow::Cow;
-
-        #[derive(Debug, Clone)]
-        struct MockEmbedding {
-            dim: usize,
-        }
-
-        impl EmbeddingFunction for MockEmbedding {
-            fn name(&self) -> &str {
-                "test_embedding"
-            }
-
-            fn source_type(&self) -> Result<Cow<'_, DataType>> {
-                Ok(Cow::Owned(DataType::Utf8))
-            }
-
-            fn dest_type(&self) -> Result<Cow<'_, DataType>> {
-                Ok(Cow::Owned(DataType::new_fixed_size_list(
-                    DataType::Float32,
-                    self.dim as i32,
-                    true,
-                )))
-            }
-
-            fn compute_source_embeddings(&self, source: Arc<dyn Array>) -> Result<Arc<dyn Array>> {
-                let len = source.len();
-                let values = vec![1.0f32; len * self.dim];
-                let values = Arc::new(Float32Array::from(values));
-                let field = Arc::new(Field::new("item", DataType::Float32, true));
-                Ok(Arc::new(FixedSizeListArray::new(
-                    field,
-                    self.dim as i32,
-                    values,
-                    None,
-                )))
-            }
-
-            fn compute_query_embeddings(&self, _input: Arc<dyn Array>) -> Result<Arc<dyn Array>> {
-                unimplemented!()
-            }
-        }
-
-        let tmp_dir = tempdir().unwrap();
-        let uri = tmp_dir.path().to_str().unwrap();
-        let db = connect(uri).execute().await.unwrap();
-
-        let embed_func = Arc::new(MockEmbedding { dim: 128 });
-        db.embedding_registry()
-            .register("test_embedding", embed_func.clone())
-            .unwrap();
-
-        let schema = Arc::new(Schema::new(vec![Field::new("name", DataType::Utf8, true)]));
-        let ed = EmbeddingDefinition {
-            source_column: "name".to_owned(),
-            dest_column: Some("name_embedding".to_owned()),
-            embedding_name: "test_embedding".to_owned(),
-        };
-
-        let table = db
-            .create_empty_table("test", schema)
-            .mode(CreateTableMode::Overwrite)
-            .add_embedding(ed)
-            .unwrap()
-            .execute()
-            .await
-            .unwrap();
-
-        let table_schema = table.schema().await.unwrap();
-        assert!(table_schema.column_with_name("name").is_some());
-        assert!(table_schema.column_with_name("name_embedding").is_some());
-
-        let embedding_field = table_schema.field_with_name("name_embedding").unwrap();
-        assert_eq!(
-            embedding_field.data_type(),
-            &DataType::new_fixed_size_list(DataType::Float32, 128, true)
-        );
-
-        let input_schema = Arc::new(Schema::new(vec![Field::new("name", DataType::Utf8, true)]));
-        let input_batch = RecordBatch::try_new(
-            input_schema.clone(),
-            vec![Arc::new(StringArray::from(vec![
-                Some("Alice"),
-                Some("Bob"),
-                Some("Charlie"),
-            ]))],
-        )
-        .unwrap();
-
-        let input_reader = Box::new(RecordBatchIterator::new(
-            vec![Ok(input_batch)].into_iter(),
-            input_schema,
-        ));
-
-        table.add(input_reader).execute().await.unwrap();
-
-        let results = table
-            .query()
-            .execute()
-            .await
-            .unwrap()
-            .try_collect::<Vec<_>>()
-            .await
-            .unwrap();
-
-        assert_eq!(results.len(), 1);
-        let batch = &results[0];
-        assert_eq!(batch.num_rows(), 3);
-        assert!(batch.column_by_name("name_embedding").is_some());
-
-        let embedding_col = batch
-            .column_by_name("name_embedding")
-            .unwrap()
-            .as_any()
-            .downcast_ref::<FixedSizeListArray>()
-            .unwrap();
-        assert_eq!(embedding_col.len(), 3);
-    }
 }

rust/lancedb/src/connection/create_table.rs

@@ -0,0 +1,612 @@
+// SPDX-License-Identifier: Apache-2.0
+// SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+use std::sync::Arc;
+
+use lance_io::object_store::StorageOptionsProvider;
+
+use crate::{
+    connection::{merge_storage_options, set_storage_options_provider},
+    data::scannable::{Scannable, WithEmbeddingsScannable},
+    database::{CreateTableMode, CreateTableRequest, Database},
+    embeddings::{EmbeddingDefinition, EmbeddingFunction, EmbeddingRegistry},
+    table::WriteOptions,
+    Error, Result, Table,
+};
+
+pub struct CreateTableBuilder {
+    parent: Arc<dyn Database>,
+    embeddings: Vec<(EmbeddingDefinition, Arc<dyn EmbeddingFunction>)>,
+    embedding_registry: Arc<dyn EmbeddingRegistry>,
+    request: CreateTableRequest,
+}
+
+impl CreateTableBuilder {
+    pub(super) fn new(
+        parent: Arc<dyn Database>,
+        embedding_registry: Arc<dyn EmbeddingRegistry>,
+        name: String,
+        data: Box<dyn Scannable>,
+    ) -> Self {
+        Self {
+            parent,
+            embeddings: Vec::new(),
+            embedding_registry,
+            request: CreateTableRequest::new(name, data),
+        }
+    }
+
+    /// Set the mode for creating the table
+    ///
+    /// This controls what happens if a table with the given name already exists
+    pub fn mode(mut self, mode: CreateTableMode) -> Self {
+        self.request.mode = mode;
+        self
+    }
+
+    /// Apply the given write options when writing the initial data
+    pub fn write_options(mut self, write_options: WriteOptions) -> Self {
+        self.request.write_options = write_options;
+        self
+    }
+
+    /// Set an option for the storage layer.
+    ///
+    /// Options already set on the connection will be inherited by the table,
+    /// but can be overridden here.
+    ///
+    /// See available options at <https://lancedb.com/docs/storage/>
+    pub fn storage_option(mut self, key: impl Into<String>, value: impl Into<String>) -> Self {
+        let store_params = self
+            .request
+            .write_options
+            .lance_write_params
+            .get_or_insert(Default::default())
+            .store_params
+            .get_or_insert(Default::default());
+        merge_storage_options(store_params, [(key.into(), value.into())]);
+        self
+    }
+
+    /// Set multiple options for the storage layer.
+    ///
+    /// Options already set on the connection will be inherited by the table,
+    /// but can be overridden here.
+    ///
+    /// See available options at <https://lancedb.com/docs/storage/>
+    pub fn storage_options(
+        mut self,
+        pairs: impl IntoIterator<Item = (impl Into<String>, impl Into<String>)>,
+    ) -> Self {
+        let store_params = self
+            .request
+            .write_options
+            .lance_write_params
+            .get_or_insert(Default::default())
+            .store_params
+            .get_or_insert(Default::default());
+        let updates = pairs
+            .into_iter()
+            .map(|(key, value)| (key.into(), value.into()));
+        merge_storage_options(store_params, updates);
+        self
+    }
+
+    /// Add an embedding definition to the table.
+    ///
+    /// The `embedding_name` must match the name of an embedding function that
+    /// was previously registered with the connection's [`EmbeddingRegistry`].
+    pub fn add_embedding(mut self, definition: EmbeddingDefinition) -> Result<Self> {
+        // Early verification of the embedding name
+        let embedding_func = self
+            .embedding_registry
+            .get(&definition.embedding_name)
+            .ok_or_else(|| Error::EmbeddingFunctionNotFound {
+                name: definition.embedding_name.clone(),
+                reason: "No embedding function found in the connection's embedding_registry"
+                    .to_string(),
+            })?;
+
+        self.embeddings.push((definition, embedding_func));
+        Ok(self)
+    }
+
+    /// Set the namespace for the table
+    pub fn namespace(mut self, namespace: Vec<String>) -> Self {
+        self.request.namespace = namespace;
+        self
+    }
+
+    /// Set a custom location for the table.
+    ///
+    /// If not set, the database will derive a location from its URI and the table name.
+    /// This is useful when integrating with namespace systems that manage table locations.
+    pub fn location(mut self, location: impl Into<String>) -> Self {
+        self.request.location = Some(location.into());
+        self
+    }
+
+    /// Set a storage options provider for automatic credential refresh.
+    ///
+    /// This allows tables to automatically refresh cloud storage credentials
+    /// when they expire, enabling long-running operations on remote storage.
+    pub fn storage_options_provider(mut self, provider: Arc<dyn StorageOptionsProvider>) -> Self {
+        let store_params = self
+            .request
+            .write_options
+            .lance_write_params
+            .get_or_insert(Default::default())
+            .store_params
+            .get_or_insert(Default::default());
+        set_storage_options_provider(store_params, provider);
+        self
+    }
+
+    /// Execute the create table operation
+    pub async fn execute(mut self) -> Result<Table> {
+        let embedding_registry = self.embedding_registry.clone();
+        let parent = self.parent.clone();
+
+        // If embeddings were configured via add_embedding(), wrap the data
+        if !self.embeddings.is_empty() {
+            let wrapped_data: Box<dyn Scannable> = Box::new(WithEmbeddingsScannable::try_new(
+                self.request.data,
+                self.embeddings,
+            )?);
+            self.request.data = wrapped_data;
+        }
+
+        Ok(Table::new_with_embedding_registry(
+            parent.create_table(self.request).await?,
+            parent,
+            embedding_registry,
+        ))
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use arrow_array::{
+        record_batch, Array, FixedSizeListArray, Float32Array, RecordBatch, RecordBatchIterator,
+    };
+    use arrow_schema::{ArrowError, DataType, Field, Schema};
+    use futures::TryStreamExt;
+    use lance_file::version::LanceFileVersion;
+    use tempfile::tempdir;
+
+    use crate::{
+        arrow::{SendableRecordBatchStream, SimpleRecordBatchStream},
+        connect,
+        database::listing::{ListingDatabaseOptions, NewTableConfig},
+        embeddings::{EmbeddingDefinition, EmbeddingFunction, MemoryRegistry},
+        query::{ExecutableQuery, QueryBase, Select},
+        test_utils::embeddings::MockEmbed,
+    };
+    use std::borrow::Cow;
+
+    use super::*;
+
+    #[tokio::test]
+    async fn create_empty_table() {
+        let db = connect("memory://").execute().await.unwrap();
+        let schema = Arc::new(Schema::new(vec![
+            Field::new("id", DataType::Int64, false),
+            Field::new("value", DataType::Float64, false),
+        ]));
+        db.create_empty_table("name", schema.clone())
+            .execute()
+            .await
+            .unwrap();
+        let table = db.open_table("name").execute().await.unwrap();
+        assert_eq!(table.schema().await.unwrap(), schema);
+        assert_eq!(table.count_rows(None).await.unwrap(), 0);
+    }
+
+    async fn test_create_table_with_data<T>(data: T)
+    where
+        T: Scannable + 'static,
+    {
+        let db = connect("memory://").execute().await.unwrap();
+        let schema = data.schema();
+        db.create_table("data_table", data).execute().await.unwrap();
+        let table = db.open_table("data_table").execute().await.unwrap();
+        assert_eq!(table.count_rows(None).await.unwrap(), 3);
+        assert_eq!(table.schema().await.unwrap(), schema);
+    }
+
+    #[tokio::test]
+    async fn create_table_with_batch() {
+        let batch = record_batch!(("id", Int64, [1, 2, 3])).unwrap();
+        test_create_table_with_data(batch).await;
+    }
+
+    #[tokio::test]
+    async fn test_create_table_with_vec_batch() {
+        let data = vec![
+            record_batch!(("id", Int64, [1, 2])).unwrap(),
+            record_batch!(("id", Int64, [3])).unwrap(),
+        ];
+        test_create_table_with_data(data).await;
+    }
+
+    #[tokio::test]
+    async fn test_create_table_with_record_batch_reader() {
+        let data = vec![
+            record_batch!(("id", Int64, [1, 2])).unwrap(),
+            record_batch!(("id", Int64, [3])).unwrap(),
+        ];
+        let schema = data[0].schema();
+        let reader: Box<dyn arrow_array::RecordBatchReader + Send> = Box::new(
+            RecordBatchIterator::new(data.into_iter().map(Ok), schema.clone()),
+        );
+        test_create_table_with_data(reader).await;
+    }
+
+    #[tokio::test]
+    async fn test_create_table_with_stream() {
+        let data = vec![
+            record_batch!(("id", Int64, [1, 2])).unwrap(),
+            record_batch!(("id", Int64, [3])).unwrap(),
+        ];
+        let schema = data[0].schema();
+        let inner = futures::stream::iter(data.into_iter().map(Ok));
+        let stream: SendableRecordBatchStream = Box::pin(SimpleRecordBatchStream {
+            schema,
+            stream: inner,
+        });
+        test_create_table_with_data(stream).await;
+    }
+
+    #[derive(Debug)]
+    struct MyError;
+
+    impl std::fmt::Display for MyError {
+        fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+            write!(f, "MyError occurred")
+        }
+    }
+
+    impl std::error::Error for MyError {}
+
+    #[tokio::test]
+    async fn test_create_preserves_reader_error() {
+        let first_batch = record_batch!(("id", Int64, [1, 2])).unwrap();
+        let schema = first_batch.schema();
+        let iterator = vec![
+            Ok(first_batch),
+            Err(ArrowError::ExternalError(Box::new(MyError))),
+        ];
+        let reader: Box<dyn arrow_array::RecordBatchReader + Send> = Box::new(
+            RecordBatchIterator::new(iterator.into_iter(), schema.clone()),
+        );
+
+        let db = connect("memory://").execute().await.unwrap();
+        let result = db.create_table("failing_table", reader).execute().await;
+
+        assert!(result.is_err());
+        // TODO: when we upgrade to Lance 2.0.0, this should pass
+        // assert!(matches!(result, Err(Error::External { source})
+        //     if source.downcast_ref::<MyError>().is_some()
+        // ));
+    }
+
+    #[tokio::test]
+    async fn test_create_preserves_stream_error() {
+        let first_batch = record_batch!(("id", Int64, [1, 2])).unwrap();
+        let schema = first_batch.schema();
+        let iterator = vec![
+            Ok(first_batch),
+            Err(Error::External {
+                source: Box::new(MyError),
+            }),
+        ];
+        let stream = futures::stream::iter(iterator);
+        let stream: SendableRecordBatchStream = Box::pin(SimpleRecordBatchStream {
+            schema: schema.clone(),
+            stream,
+        });
+
+        let db = connect("memory://").execute().await.unwrap();
+        let result = db
+            .create_table("failing_stream_table", stream)
+            .execute()
+            .await;
+
+        assert!(result.is_err());
+        // TODO: when we upgrade to Lance 2.0.0, this should pass
+        // assert!(matches!(result, Err(Error::External { source})
+        //     if source.downcast_ref::<MyError>().is_some()
+        // ));
+    }
+
+    #[tokio::test]
+    #[allow(deprecated)]
+    async fn test_create_table_with_storage_options() {
+        let batch = record_batch!(("id", Int64, [1, 2, 3])).unwrap();
+        let db = connect("memory://").execute().await.unwrap();
+
+        let table = db
+            .create_table("options_table", batch)
+            .storage_option("timeout", "30s")
+            .storage_options([("retry_count", "3")])
+            .execute()
+            .await
+            .unwrap();
+
+        let final_options = table.storage_options().await.unwrap();
+        assert_eq!(final_options.get("timeout"), Some(&"30s".to_string()));
+        assert_eq!(final_options.get("retry_count"), Some(&"3".to_string()));
+    }
+
+    #[tokio::test]
+    async fn test_create_table_unregistered_embedding() {
+        let db = connect("memory://").execute().await.unwrap();
+        let batch = record_batch!(("text", Utf8, ["hello", "world"])).unwrap();
+
+        // Try to add an embedding that doesn't exist in the registry
+        let result = db
+            .create_table("embed_table", batch)
+            .add_embedding(EmbeddingDefinition::new(
+                "text",
+                "nonexistent_embedding_function",
+                None::<&str>,
+            ));
+
+        match result {
+            Err(Error::EmbeddingFunctionNotFound { name, .. }) => {
+                assert_eq!(name, "nonexistent_embedding_function");
+            }
+            Err(other) => panic!("Expected EmbeddingFunctionNotFound error, got: {:?}", other),
+            Ok(_) => panic!("Expected error, but got Ok"),
+        }
+    }
+
+    #[tokio::test]
+    async fn test_create_table_already_exists() {
+        let tmp_dir = tempdir().unwrap();
+        let uri = tmp_dir.path().to_str().unwrap();
+        let db = connect(uri).execute().await.unwrap();
+        let schema = Arc::new(Schema::new(vec![Field::new("x", DataType::Int32, false)]));
+        db.create_empty_table("test", schema.clone())
+            .execute()
+            .await
+            .unwrap();
+        db.create_empty_table("test", schema)
+            .mode(CreateTableMode::exist_ok(|mut req| {
+                req.index_cache_size = Some(16);
+                req
+            }))
+            .execute()
+            .await
+            .unwrap();
+        let other_schema = Arc::new(Schema::new(vec![Field::new("y", DataType::Int32, false)]));
+        assert!(db
+            .create_empty_table("test", other_schema.clone())
+            .execute()
+            .await
+            .is_err()); // TODO: assert what this error is
+        let overwritten = db
+            .create_empty_table("test", other_schema.clone())
+            .mode(CreateTableMode::Overwrite)
+            .execute()
+            .await
+            .unwrap();
+        assert_eq!(other_schema, overwritten.schema().await.unwrap());
+    }
+
+    #[tokio::test]
+    #[rstest::rstest]
+    #[case(LanceFileVersion::Legacy)]
+    #[case(LanceFileVersion::Stable)]
+    async fn test_create_table_with_storage_version(
+        #[case] data_storage_version: LanceFileVersion,
+    ) {
+        let db = connect("memory://")
+            .database_options(&ListingDatabaseOptions {
+                new_table_config: NewTableConfig {
+                    data_storage_version: Some(data_storage_version),
+                    ..Default::default()
+                },
+                ..Default::default()
+            })
+            .execute()
+            .await
+            .unwrap();
+
+        let batch = record_batch!(("id", Int64, [1, 2, 3])).unwrap();
+        let table = db
+            .create_table("legacy_table", batch)
+            .execute()
+            .await
+            .unwrap();
+
+        let native_table = table.as_native().unwrap();
+        let storage_format = native_table
+            .manifest()
+            .await
+            .unwrap()
+            .data_storage_format
+            .lance_file_version()
+            .unwrap();
+        // Compare resolved versions since Stable/Next are aliases that resolve at storage time
+        assert_eq!(storage_format.resolve(), data_storage_version.resolve());
+    }
+
+    #[tokio::test]
+    async fn test_create_table_with_embedding() {
+        // Register the mock embedding function
+        let registry = Arc::new(MemoryRegistry::new());
+        let mock_embedding: Arc<dyn EmbeddingFunction> = Arc::new(MockEmbed::new("mock", 4));
+        registry.register("mock", mock_embedding).unwrap();
+
+        // Connect with the custom registry
+        let conn = connect("memory://")
+            .embedding_registry(registry)
+            .execute()
+            .await
+            .unwrap();
+
+        // Create data without the embedding column
+        let batch = record_batch!(("text", Utf8, ["hello", "world", "test"])).unwrap();
+
+        // Create table with add_embedding - embeddings should be computed automatically
+        let table = conn
+            .create_table("embed_test", batch)
+            .add_embedding(EmbeddingDefinition::new(
+                "text",
+                "mock",
+                Some("text_embedding"),
+            ))
+            .unwrap()
+            .execute()
+            .await
+            .unwrap();
+
+        // Verify row count
+        assert_eq!(table.count_rows(None).await.unwrap(), 3);
+
+        // Verify the schema includes the embedding column
+        let result_schema = table.schema().await.unwrap();
+        assert_eq!(result_schema.fields().len(), 2);
+        assert_eq!(result_schema.field(0).name(), "text");
+        assert_eq!(result_schema.field(1).name(), "text_embedding");
+
+        // Verify the embedding column has the correct type
+        assert!(matches!(
+            result_schema.field(1).data_type(),
+            DataType::FixedSizeList(_, 4)
+        ));
+
+        // Query to verify the embeddings were computed
+        let results: Vec<RecordBatch> = table
+            .query()
+            .select(Select::columns(&["text", "text_embedding"]))
+            .execute()
+            .await
+            .unwrap()
+            .try_collect()
+            .await
+            .unwrap();
+
+        let total_rows: usize = results.iter().map(|b| b.num_rows()).sum();
+        assert_eq!(total_rows, 3);
+
+        // Check that all rows have embedding values (not null)
+        for batch in &results {
+            let embedding_col = batch.column(1);
+            assert_eq!(embedding_col.null_count(), 0);
+            assert_eq!(embedding_col.len(), batch.num_rows());
+        }
+
+        // Verify the schema metadata contains the column definitions
+        assert!(
+            result_schema
+                .metadata
+                .contains_key("lancedb::column_definitions"),
+            "Schema metadata should contain column definitions"
+        );
+    }
+
+    #[tokio::test]
+    async fn test_create_empty_table_with_embeddings() {
+        #[derive(Debug, Clone)]
+        struct MockEmbedding {
+            dim: usize,
+        }
+
+        impl EmbeddingFunction for MockEmbedding {
+            fn name(&self) -> &str {
+                "test_embedding"
+            }
+
+            fn source_type(&self) -> Result<Cow<'_, DataType>> {
+                Ok(Cow::Owned(DataType::Utf8))
+            }
+
+            fn dest_type(&self) -> Result<Cow<'_, DataType>> {
+                Ok(Cow::Owned(DataType::new_fixed_size_list(
+                    DataType::Float32,
+                    self.dim as i32,
+                    true,
+                )))
+            }
+
+            fn compute_source_embeddings(&self, source: Arc<dyn Array>) -> Result<Arc<dyn Array>> {
+                let len = source.len();
+                let values = vec![1.0f32; len * self.dim];
+                let values = Arc::new(Float32Array::from(values));
+                let field = Arc::new(Field::new("item", DataType::Float32, true));
+                Ok(Arc::new(FixedSizeListArray::new(
+                    field,
+                    self.dim as i32,
+                    values,
+                    None,
+                )))
+            }
+
+            fn compute_query_embeddings(&self, _input: Arc<dyn Array>) -> Result<Arc<dyn Array>> {
+                unimplemented!()
+            }
+        }
+
+        let tmp_dir = tempdir().unwrap();
+        let uri = tmp_dir.path().to_str().unwrap();
+        let db = connect(uri).execute().await.unwrap();
+
+        let embed_func = Arc::new(MockEmbedding { dim: 128 });
+        db.embedding_registry()
+            .register("test_embedding", embed_func.clone())
+            .unwrap();
+
+        let schema = Arc::new(Schema::new(vec![Field::new("name", DataType::Utf8, true)]));
+        let ed = EmbeddingDefinition {
+            source_column: "name".to_owned(),
+            dest_column: Some("name_embedding".to_owned()),
+            embedding_name: "test_embedding".to_owned(),
+        };
+
+        let table = db
+            .create_empty_table("test", schema)
+            .mode(CreateTableMode::Overwrite)
+            .add_embedding(ed)
+            .unwrap()
+            .execute()
+            .await
+            .unwrap();
+
+        let table_schema = table.schema().await.unwrap();
+        assert!(table_schema.column_with_name("name").is_some());
+        assert!(table_schema.column_with_name("name_embedding").is_some());
+
+        let embedding_field = table_schema.field_with_name("name_embedding").unwrap();
+        assert_eq!(
+            embedding_field.data_type(),
+            &DataType::new_fixed_size_list(DataType::Float32, 128, true)
+        );
+
+        let input_batch = record_batch!(("name", Utf8, ["Alice", "Bob", "Charlie"])).unwrap();
+        table.add(input_batch).execute().await.unwrap();
+
+        let results = table
+            .query()
+            .execute()
+            .await
+            .unwrap()
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+
+        assert_eq!(results.len(), 1);
+        let batch = &results[0];
+        assert_eq!(batch.num_rows(), 3);
+        assert!(batch.column_by_name("name_embedding").is_some());
+
+        let embedding_col = batch
+            .column_by_name("name_embedding")
+            .unwrap()
+            .as_any()
+            .downcast_ref::<FixedSizeListArray>()
+            .unwrap();
+        assert_eq!(embedding_col.len(), 3);
+    }
+}

rust/lancedb/src/data.rs

@@ -5,3 +5,4 @@
 
 pub mod inspect;
 pub mod sanitize;
+pub mod scannable;

rust/lancedb/src/data/scannable.rs

@@ -0,0 +1,580 @@
+// SPDX-License-Identifier: Apache-2.0
+// SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+//! Data source abstraction for LanceDB.
+//!
+//! This module provides a [`Scannable`] trait that allows input data sources to express
+//! capabilities (row count, rescannability) so the insert pipeline can make
+//! better decisions about write parallelism and retry strategies.
+
+use std::sync::Arc;
+
+use arrow_array::{RecordBatch, RecordBatchIterator, RecordBatchReader};
+use arrow_schema::{ArrowError, SchemaRef};
+use async_trait::async_trait;
+use futures::stream::once;
+use futures::StreamExt;
+use lance_datafusion::utils::StreamingWriteSource;
+
+use crate::arrow::{
+    SendableRecordBatchStream, SendableRecordBatchStreamExt, SimpleRecordBatchStream,
+};
+use crate::embeddings::{
+    compute_embeddings_for_batch, compute_output_schema, EmbeddingDefinition, EmbeddingFunction,
+    EmbeddingRegistry,
+};
+use crate::table::{ColumnDefinition, ColumnKind, TableDefinition};
+use crate::{Error, Result};
+
+pub trait Scannable: Send {
+    /// Returns the schema of the data.
+    fn schema(&self) -> SchemaRef;
+
+    /// Read data as a stream of record batches.
+    ///
+    /// For rescannable sources (in-memory data like RecordBatch, Vec<RecordBatch>),
+    /// this can be called multiple times and returns cloned data each time.
+    ///
+    /// For non-rescannable sources (streams, readers), this can only be called once.
+    /// Calling it a second time returns a stream whose first item is an error.
+    fn scan_as_stream(&mut self) -> SendableRecordBatchStream;
+
+    /// Optional hint about the number of rows.
+    ///
+    /// When available, this allows the pipeline to estimate total data size
+    /// and choose appropriate partitioning.
+    fn num_rows(&self) -> Option<usize> {
+        None
+    }
+
+    /// Whether the source can be re-read from the beginning.
+    ///
+    /// `true` for in-memory data (Tables, DataFrames) and disk-based sources (Datasets).
+    /// `false` for streaming sources (DuckDB results, network streams).
+    ///
+    /// When true, the pipeline can retry failed writes by rescanning.
+    fn rescannable(&self) -> bool {
+        false
+    }
+}
+
+impl std::fmt::Debug for dyn Scannable {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.debug_struct("Scannable")
+            .field("schema", &self.schema())
+            .field("num_rows", &self.num_rows())
+            .field("rescannable", &self.rescannable())
+            .finish()
+    }
+}
+
+impl Scannable for RecordBatch {
+    fn schema(&self) -> SchemaRef {
+        Self::schema(self)
+    }
+
+    fn scan_as_stream(&mut self) -> SendableRecordBatchStream {
+        let batch = self.clone();
+        let schema = batch.schema();
+        Box::pin(SimpleRecordBatchStream {
+            schema,
+            stream: once(async move { Ok(batch) }),
+        })
+    }
+
+    fn num_rows(&self) -> Option<usize> {
+        Some(Self::num_rows(self))
+    }
+
+    fn rescannable(&self) -> bool {
+        true
+    }
+}
+
+impl Scannable for Vec<RecordBatch> {
+    fn schema(&self) -> SchemaRef {
+        if self.is_empty() {
+            Arc::new(arrow_schema::Schema::empty())
+        } else {
+            self[0].schema()
+        }
+    }
+
+    fn scan_as_stream(&mut self) -> SendableRecordBatchStream {
+        if self.is_empty() {
+            let schema = Scannable::schema(self);
+            return Box::pin(SimpleRecordBatchStream {
+                schema,
+                stream: once(async {
+                    Err(Error::InvalidInput {
+                        message: "Cannot scan an empty Vec<RecordBatch>".to_string(),
+                    })
+                }),
+            });
+        }
+        let schema = Scannable::schema(self);
+        let batches = self.clone();
+        let stream = futures::stream::iter(batches.into_iter().map(Ok));
+        Box::pin(SimpleRecordBatchStream { schema, stream })
+    }
+
+    fn num_rows(&self) -> Option<usize> {
+        Some(self.iter().map(|b| b.num_rows()).sum())
+    }
+
+    fn rescannable(&self) -> bool {
+        true
+    }
+}
+
+impl Scannable for Box<dyn RecordBatchReader + Send> {
+    fn schema(&self) -> SchemaRef {
+        RecordBatchReader::schema(self.as_ref())
+    }
+
+    fn scan_as_stream(&mut self) -> SendableRecordBatchStream {
+        let schema = Scannable::schema(self);
+
+        // Swap self with a reader that errors on iteration, so a second call
+        // produces a clear error instead of silently returning empty data.
+        let err_reader: Box<dyn RecordBatchReader + Send> = Box::new(RecordBatchIterator::new(
+            vec![Err(ArrowError::InvalidArgumentError(
+                "Reader has already been consumed".into(),
+            ))],
+            schema.clone(),
+        ));
+        let reader = std::mem::replace(self, err_reader);
+
+        // Bridge the blocking RecordBatchReader to an async stream via a channel.
+        let (tx, rx) = tokio::sync::mpsc::channel::<crate::Result<RecordBatch>>(2);
+        tokio::task::spawn_blocking(move || {
+            for batch_result in reader {
+                let result = batch_result.map_err(Into::into);
+                if tx.blocking_send(result).is_err() {
+                    break;
+                }
+            }
+        });
+
+        let stream = futures::stream::unfold(rx, |mut rx| async move {
+            rx.recv().await.map(|batch| (batch, rx))
+        })
+        .fuse();
+
+        Box::pin(SimpleRecordBatchStream { schema, stream })
+    }
+}
+
+impl Scannable for SendableRecordBatchStream {
+    fn schema(&self) -> SchemaRef {
+        self.as_ref().schema()
+    }
+
+    fn scan_as_stream(&mut self) -> SendableRecordBatchStream {
+        let schema = Scannable::schema(self);
+
+        // Swap self with an error stream so a second call produces a clear error.
+        let error_stream = Box::pin(SimpleRecordBatchStream {
+            schema: schema.clone(),
+            stream: once(async {
+                Err(Error::InvalidInput {
+                    message: "Stream has already been consumed".to_string(),
+                })
+            }),
+        });
+        std::mem::replace(self, error_stream)
+    }
+}
+
+#[async_trait]
+impl StreamingWriteSource for Box<dyn Scannable> {
+    fn arrow_schema(&self) -> SchemaRef {
+        self.schema()
+    }
+
+    fn into_stream(mut self) -> datafusion_physical_plan::SendableRecordBatchStream {
+        self.scan_as_stream().into_df_stream()
+    }
+}
+
+/// A scannable that applies embeddings to the stream.
+pub struct WithEmbeddingsScannable {
+    inner: Box<dyn Scannable>,
+    embeddings: Vec<(EmbeddingDefinition, Arc<dyn EmbeddingFunction>)>,
+    output_schema: SchemaRef,
+}
+
+impl WithEmbeddingsScannable {
+    /// Create a new WithEmbeddingsScannable.
+    ///
+    /// The embeddings are applied to the inner scannable's data as new columns.
+    pub fn try_new(
+        inner: Box<dyn Scannable>,
+        embeddings: Vec<(EmbeddingDefinition, Arc<dyn EmbeddingFunction>)>,
+    ) -> Result<Self> {
+        let output_schema = compute_output_schema(&inner.schema(), &embeddings)?;
+
+        // Build column definitions: Physical for base columns, Embedding for new ones
+        let base_col_count = inner.schema().fields().len();
+        let column_definitions: Vec<ColumnDefinition> = (0..base_col_count)
+            .map(|_| ColumnDefinition {
+                kind: ColumnKind::Physical,
+            })
+            .chain(embeddings.iter().map(|(ed, _)| ColumnDefinition {
+                kind: ColumnKind::Embedding(ed.clone()),
+            }))
+            .collect();
+
+        let table_definition = TableDefinition::new(output_schema, column_definitions);
+        let output_schema = table_definition.into_rich_schema();
+
+        Ok(Self {
+            inner,
+            embeddings,
+            output_schema,
+        })
+    }
+}
+
+impl Scannable for WithEmbeddingsScannable {
+    fn schema(&self) -> SchemaRef {
+        self.output_schema.clone()
+    }
+
+    fn scan_as_stream(&mut self) -> SendableRecordBatchStream {
+        let inner_stream = self.inner.scan_as_stream();
+        let embeddings = self.embeddings.clone();
+        let output_schema = self.output_schema.clone();
+
+        let mapped_stream = inner_stream.then(move |batch_result| {
+            let embeddings = embeddings.clone();
+            async move {
+                let batch = batch_result?;
+                let result = tokio::task::spawn_blocking(move || {
+                    compute_embeddings_for_batch(batch, &embeddings)
+                })
+                .await
+                .map_err(|e| Error::Runtime {
+                    message: format!("Task panicked during embedding computation: {}", e),
+                })??;
+                Ok(result)
+            }
+        });
+
+        Box::pin(SimpleRecordBatchStream {
+            schema: output_schema,
+            stream: mapped_stream,
+        })
+    }
+
+    fn num_rows(&self) -> Option<usize> {
+        self.inner.num_rows()
+    }
+
+    fn rescannable(&self) -> bool {
+        self.inner.rescannable()
+    }
+}
+
+pub fn scannable_with_embeddings(
+    inner: Box<dyn Scannable>,
+    table_definition: &TableDefinition,
+    registry: Option<&Arc<dyn EmbeddingRegistry>>,
+) -> Result<Box<dyn Scannable>> {
+    if let Some(registry) = registry {
+        let mut embeddings = Vec::with_capacity(table_definition.column_definitions.len());
+        for cd in table_definition.column_definitions.iter() {
+            if let ColumnKind::Embedding(embedding_def) = &cd.kind {
+                match registry.get(&embedding_def.embedding_name) {
+                    Some(func) => {
+                        embeddings.push((embedding_def.clone(), func));
+                    }
+                    None => {
+                        return Err(Error::EmbeddingFunctionNotFound {
+                            name: embedding_def.embedding_name.clone(),
+                            reason: format!(
+                                "Table was defined with an embedding column `{}` but no embedding function was found with that name within the registry.",
+                                embedding_def.embedding_name
+                            ),
+                        });
+                    }
+                }
+            }
+        }
+
+        if !embeddings.is_empty() {
+            return Ok(Box::new(WithEmbeddingsScannable::try_new(
+                inner, embeddings,
+            )?));
+        }
+    }
+
+    Ok(inner)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use arrow_array::record_batch;
+    use futures::TryStreamExt;
+
+    #[tokio::test]
+    async fn test_record_batch_rescannable() {
+        let mut batch = record_batch!(("id", Int64, [0, 1, 2])).unwrap();
+
+        let stream1 = batch.scan_as_stream();
+        let batches1: Vec<RecordBatch> = stream1.try_collect().await.unwrap();
+        assert_eq!(batches1.len(), 1);
+        assert_eq!(batches1[0], batch);
+
+        assert!(batch.rescannable());
+        let stream2 = batch.scan_as_stream();
+        let batches2: Vec<RecordBatch> = stream2.try_collect().await.unwrap();
+        assert_eq!(batches2.len(), 1);
+        assert_eq!(batches2[0], batch);
+    }
+
+    #[tokio::test]
+    async fn test_vec_batch_rescannable() {
+        let mut batches = vec![
+            record_batch!(("id", Int64, [0, 1])).unwrap(),
+            record_batch!(("id", Int64, [2, 3, 4])).unwrap(),
+        ];
+
+        let stream1 = batches.scan_as_stream();
+        let result1: Vec<RecordBatch> = stream1.try_collect().await.unwrap();
+        assert_eq!(result1.len(), 2);
+        assert_eq!(result1[0], batches[0]);
+        assert_eq!(result1[1], batches[1]);
+
+        assert!(batches.rescannable());
+        let stream2 = batches.scan_as_stream();
+        let result2: Vec<RecordBatch> = stream2.try_collect().await.unwrap();
+        assert_eq!(result2.len(), 2);
+        assert_eq!(result2[0], batches[0]);
+        assert_eq!(result2[1], batches[1]);
+    }
+
+    #[tokio::test]
+    async fn test_vec_batch_empty_errors() {
+        let mut empty: Vec<RecordBatch> = vec![];
+        let mut stream = empty.scan_as_stream();
+        let result = stream.next().await;
+        assert!(result.is_some());
+        assert!(result.unwrap().is_err());
+    }
+
+    #[tokio::test]
+    async fn test_reader_not_rescannable() {
+        let batch = record_batch!(("id", Int64, [0, 1, 2])).unwrap();
+        let schema = batch.schema();
+        let mut reader: Box<dyn arrow_array::RecordBatchReader + Send> = Box::new(
+            RecordBatchIterator::new(vec![Ok(batch.clone())], schema.clone()),
+        );
+
+        let stream1 = reader.scan_as_stream();
+        let result1: Vec<RecordBatch> = stream1.try_collect().await.unwrap();
+        assert_eq!(result1.len(), 1);
+        assert_eq!(result1[0], batch);
+
+        assert!(!reader.rescannable());
+        // Second call returns a stream whose first item is an error
+        let mut stream2 = reader.scan_as_stream();
+        let result2 = stream2.next().await;
+        assert!(result2.is_some());
+        assert!(result2.unwrap().is_err());
+    }
+
+    #[tokio::test]
+    async fn test_stream_not_rescannable() {
+        let batch = record_batch!(("id", Int64, [0, 1, 2])).unwrap();
+        let schema = batch.schema();
+        let inner_stream = futures::stream::iter(vec![Ok(batch.clone())]);
+        let mut stream: SendableRecordBatchStream = Box::pin(SimpleRecordBatchStream {
+            schema: schema.clone(),
+            stream: inner_stream,
+        });
+
+        let stream1 = stream.scan_as_stream();
+        let result1: Vec<RecordBatch> = stream1.try_collect().await.unwrap();
+        assert_eq!(result1.len(), 1);
+        assert_eq!(result1[0], batch);
+
+        assert!(!stream.rescannable());
+        // Second call returns a stream whose first item is an error
+        let mut stream2 = stream.scan_as_stream();
+        let result2 = stream2.next().await;
+        assert!(result2.is_some());
+        assert!(result2.unwrap().is_err());
+    }
+
+    mod embedding_tests {
+        use super::*;
+        use crate::embeddings::MemoryRegistry;
+        use crate::table::{ColumnDefinition, ColumnKind};
+        use crate::test_utils::embeddings::MockEmbed;
+        use arrow_array::Array as _;
+        use arrow_array::{ArrayRef, StringArray};
+        use arrow_schema::{DataType, Field, Schema};
+
+        #[tokio::test]
+        async fn test_with_embeddings_scannable() {
+            let schema = Arc::new(Schema::new(vec![Field::new("text", DataType::Utf8, false)]));
+            let text_array = StringArray::from(vec!["hello", "world", "test"]);
+            let batch =
+                RecordBatch::try_new(schema.clone(), vec![Arc::new(text_array) as ArrayRef])
+                    .unwrap();
+
+            let mock_embedding: Arc<dyn EmbeddingFunction> = Arc::new(MockEmbed::new("mock", 4));
+            let embedding_def = EmbeddingDefinition::new("text", "mock", Some("text_embedding"));
+
+            let mut scannable = WithEmbeddingsScannable::try_new(
+                Box::new(batch.clone()),
+                vec![(embedding_def, mock_embedding)],
+            )
+            .unwrap();
+
+            // Check that schema has the embedding column
+            let output_schema = scannable.schema();
+            assert_eq!(output_schema.fields().len(), 2);
+            assert_eq!(output_schema.field(0).name(), "text");
+            assert_eq!(output_schema.field(1).name(), "text_embedding");
+
+            // Check num_rows and rescannable are preserved
+            assert_eq!(scannable.num_rows(), Some(3));
+            assert!(scannable.rescannable());
+
+            // Read the data
+            let stream = scannable.scan_as_stream();
+            let results: Vec<RecordBatch> = stream.try_collect().await.unwrap();
+            assert_eq!(results.len(), 1);
+
+            let result_batch = &results[0];
+            assert_eq!(result_batch.num_rows(), 3);
+            assert_eq!(result_batch.num_columns(), 2);
+
+            // Verify the embedding column is present and has the right shape
+            let embedding_col = result_batch.column(1);
+            assert_eq!(embedding_col.len(), 3);
+        }
+
+        #[tokio::test]
+        async fn test_maybe_embedded_scannable_no_embeddings() {
+            let batch = record_batch!(("id", Int64, [1, 2, 3])).unwrap();
+
+            // Create a table definition with no embedding columns
+            let table_def = TableDefinition::new_from_schema(batch.schema());
+
+            // Even with a registry, if there are no embedding columns, it's a passthrough
+            let registry: Arc<dyn EmbeddingRegistry> = Arc::new(MemoryRegistry::new());
+            let mut scannable =
+                scannable_with_embeddings(Box::new(batch.clone()), &table_def, Some(&registry))
+                    .unwrap();
+
+            // Check that data passes through unchanged
+            let stream = scannable.scan_as_stream();
+            let results: Vec<RecordBatch> = stream.try_collect().await.unwrap();
+            assert_eq!(results.len(), 1);
+            assert_eq!(results[0], batch);
+        }
+
+        #[tokio::test]
+        async fn test_maybe_embedded_scannable_with_embeddings() {
+            let schema = Arc::new(Schema::new(vec![Field::new("text", DataType::Utf8, false)]));
+            let text_array = StringArray::from(vec!["hello", "world"]);
+            let batch =
+                RecordBatch::try_new(schema.clone(), vec![Arc::new(text_array) as ArrayRef])
+                    .unwrap();
+
+            // Create a table definition with an embedding column
+            let embedding_def = EmbeddingDefinition::new("text", "mock", Some("text_embedding"));
+            let embedding_schema = Arc::new(Schema::new(vec![
+                Field::new("text", DataType::Utf8, false),
+                Field::new(
+                    "text_embedding",
+                    DataType::FixedSizeList(
+                        Arc::new(Field::new("item", DataType::Float32, true)),
+                        4,
+                    ),
+                    false,
+                ),
+            ]));
+            let table_def = TableDefinition::new(
+                embedding_schema,
+                vec![
+                    ColumnDefinition {
+                        kind: ColumnKind::Physical,
+                    },
+                    ColumnDefinition {
+                        kind: ColumnKind::Embedding(embedding_def.clone()),
+                    },
+                ],
+            );
+
+            // Register the mock embedding function
+            let registry: Arc<dyn EmbeddingRegistry> = Arc::new(MemoryRegistry::new());
+            let mock_embedding: Arc<dyn EmbeddingFunction> = Arc::new(MockEmbed::new("mock", 4));
+            registry.register("mock", mock_embedding).unwrap();
+
+            let mut scannable =
+                scannable_with_embeddings(Box::new(batch), &table_def, Some(&registry)).unwrap();
+
+            // Read and verify the data has embeddings
+            let stream = scannable.scan_as_stream();
+            let results: Vec<RecordBatch> = stream.try_collect().await.unwrap();
+            assert_eq!(results.len(), 1);
+
+            let result_batch = &results[0];
+            assert_eq!(result_batch.num_columns(), 2);
+            assert_eq!(result_batch.schema().field(1).name(), "text_embedding");
+        }
+
+        #[tokio::test]
+        async fn test_maybe_embedded_scannable_missing_function() {
+            let schema = Arc::new(Schema::new(vec![Field::new("text", DataType::Utf8, false)]));
+            let text_array = StringArray::from(vec!["hello"]);
+            let batch =
+                RecordBatch::try_new(schema.clone(), vec![Arc::new(text_array) as ArrayRef])
+                    .unwrap();
+
+            // Create a table definition with an embedding column
+            let embedding_def =
+                EmbeddingDefinition::new("text", "nonexistent", Some("text_embedding"));
+            let embedding_schema = Arc::new(Schema::new(vec![
+                Field::new("text", DataType::Utf8, false),
+                Field::new(
+                    "text_embedding",
+                    DataType::FixedSizeList(
+                        Arc::new(Field::new("item", DataType::Float32, true)),
+                        4,
+                    ),
+                    false,
+                ),
+            ]));
+            let table_def = TableDefinition::new(
+                embedding_schema,
+                vec![
+                    ColumnDefinition {
+                        kind: ColumnKind::Physical,
+                    },
+                    ColumnDefinition {
+                        kind: ColumnKind::Embedding(embedding_def),
+                    },
+                ],
+            );
+
+            // Registry has no embedding functions registered
+            let registry: Arc<dyn EmbeddingRegistry> = Arc::new(MemoryRegistry::new());
+
+            let result = scannable_with_embeddings(Box::new(batch), &table_def, Some(&registry));
+
+            // Should fail because the embedding function is not found
+            assert!(result.is_err());
+            let err = result.err().unwrap();
+            assert!(
+                matches!(err, Error::EmbeddingFunctionNotFound { .. }),
+                "Expected EmbeddingFunctionNotFound"
+            );
+        }
+    }
+}

rust/lancedb/src/database.rs

@@ -18,12 +18,7 @@ use std::collections::HashMap;
 use std::sync::Arc;
 use std::time::Duration;
 
-use arrow_array::RecordBatchReader;
-use async_trait::async_trait;
-use datafusion_physical_plan::stream::RecordBatchStreamAdapter;
-use futures::stream;
 use lance::dataset::ReadParams;
-use lance_datafusion::utils::StreamingWriteSource;
 use lance_namespace::models::{
     CreateNamespaceRequest, CreateNamespaceResponse, DescribeNamespaceRequest,
     DescribeNamespaceResponse, DropNamespaceRequest, DropNamespaceResponse, ListNamespacesRequest,
@@ -31,9 +26,9 @@ use lance_namespace::models::{
 };
 use lance_namespace::LanceNamespace;
 
-use crate::arrow::{SendableRecordBatchStream, SendableRecordBatchStreamExt};
+use crate::data::scannable::Scannable;
 use crate::error::Result;
-use crate::table::{BaseTable, TableDefinition, WriteOptions};
+use crate::table::{BaseTable, WriteOptions};
 
 pub mod listing;
 pub mod namespace;
@@ -115,51 +110,14 @@ impl Default for CreateTableMode {
     }
 }
 
-/// The data to start a table or a schema to create an empty table
-pub enum CreateTableData {
-    /// Creates a table using an iterator of data, the schema will be obtained from the data
-    Data(Box<dyn RecordBatchReader + Send>),
-    /// Creates a table using a stream of data, the schema will be obtained from the data
-    StreamingData(SendableRecordBatchStream),
-    /// Creates an empty table, the definition / schema must be provided separately
-    Empty(TableDefinition),
-}
-
-impl CreateTableData {
-    pub fn schema(&self) -> Arc<arrow_schema::Schema> {
-        match self {
-            Self::Data(reader) => reader.schema(),
-            Self::StreamingData(stream) => stream.schema(),
-            Self::Empty(definition) => definition.schema.clone(),
-        }
-    }
-}
-
-#[async_trait]
-impl StreamingWriteSource for CreateTableData {
-    fn arrow_schema(&self) -> Arc<arrow_schema::Schema> {
-        self.schema()
-    }
-    fn into_stream(self) -> datafusion_physical_plan::SendableRecordBatchStream {
-        match self {
-            Self::Data(reader) => reader.into_stream(),
-            Self::StreamingData(stream) => stream.into_df_stream(),
-            Self::Empty(table_definition) => {
-                let schema = table_definition.schema.clone();
-                Box::pin(RecordBatchStreamAdapter::new(schema, stream::empty()))
-            }
-        }
-    }
-}
-
 /// A request to create a table
 pub struct CreateTableRequest {
     /// The name of the new table
     pub name: String,
     /// The namespace to create the table in. Empty list represents root namespace.
     pub namespace: Vec<String>,
-    /// Initial data to write to the table, can be None to create an empty table
-    pub data: CreateTableData,
+    /// Initial data to write to the table, can be empty.
+    pub data: Box<dyn Scannable>,
     /// The mode to use when creating the table
     pub mode: CreateTableMode,
     /// Options to use when writing data (only used if `data` is not None)
@@ -173,7 +131,7 @@ pub struct CreateTableRequest {
 }
 
 impl CreateTableRequest {
-    pub fn new(name: String, data: CreateTableData) -> Self {
+    pub fn new(name: String, data: Box<dyn Scannable>) -> Self {
         Self {
             name,
             namespace: vec![],

rust/lancedb/src/database/listing.rs

@@ -922,7 +922,7 @@ impl Database for ListingDatabase {
             .with_read_params(read_params.clone())
             .load()
             .await
-            .map_err(|e| Error::Lance { source: e })?;
+            .map_err(|e| -> Error { e.into() })?;
 
         let version_ref = match (request.source_version, request.source_tag) {
             (Some(v), None) => Ok(Ref::Version(None, Some(v))),
@@ -937,7 +937,7 @@ impl Database for ListingDatabase {
         source_dataset
             .shallow_clone(&target_uri, version_ref, Some(storage_params))
             .await
-            .map_err(|e| Error::Lance { source: e })?;
+            .map_err(|e| -> Error { e.into() })?;
 
         let cloned_table = NativeTable::open_with_params(
             &target_uri,
@@ -1098,8 +1098,10 @@ impl Database for ListingDatabase {
 mod tests {
     use super::*;
     use crate::connection::ConnectRequest;
-    use crate::database::{CreateTableData, CreateTableMode, CreateTableRequest, WriteOptions};
-    use crate::table::{Table, TableDefinition};
+    use crate::data::scannable::Scannable;
+    use crate::database::{CreateTableMode, CreateTableRequest};
+    use crate::table::WriteOptions;
+    use crate::Table;
     use arrow_array::{Int32Array, RecordBatch, StringArray};
     use arrow_schema::{DataType, Field, Schema};
     use std::path::PathBuf;
@@ -1139,7 +1141,7 @@ mod tests {
             .create_table(CreateTableRequest {
                 name: "source_table".to_string(),
                 namespace: vec![],
-                data: CreateTableData::Empty(TableDefinition::new_from_schema(schema.clone())),
+                data: Box::new(RecordBatch::new_empty(schema.clone())) as Box<dyn Scannable>,
                 mode: CreateTableMode::Create,
                 write_options: Default::default(),
                 location: None,
@@ -1196,16 +1198,11 @@ mod tests {
         )
         .unwrap();
 
-        let reader = Box::new(arrow_array::RecordBatchIterator::new(
-            vec![Ok(batch)],
-            schema.clone(),
-        ));
-
         let source_table = db
             .create_table(CreateTableRequest {
                 name: "source_with_data".to_string(),
                 namespace: vec![],
-                data: CreateTableData::Data(reader),
+                data: Box::new(batch) as Box<dyn Scannable>,
                 mode: CreateTableMode::Create,
                 write_options: Default::default(),
                 location: None,
@@ -1264,7 +1261,7 @@ mod tests {
         db.create_table(CreateTableRequest {
             name: "source".to_string(),
             namespace: vec![],
-            data: CreateTableData::Empty(TableDefinition::new_from_schema(schema)),
+            data: Box::new(RecordBatch::new_empty(schema)) as Box<dyn Scannable>,
             mode: CreateTableMode::Create,
             write_options: Default::default(),
             location: None,
@@ -1300,7 +1297,7 @@ mod tests {
         db.create_table(CreateTableRequest {
             name: "source".to_string(),
             namespace: vec![],
-            data: CreateTableData::Empty(TableDefinition::new_from_schema(schema)),
+            data: Box::new(RecordBatch::new_empty(schema)) as Box<dyn Scannable>,
             mode: CreateTableMode::Create,
             write_options: Default::default(),
             location: None,
@@ -1340,7 +1337,7 @@ mod tests {
         db.create_table(CreateTableRequest {
             name: "source".to_string(),
             namespace: vec![],
-            data: CreateTableData::Empty(TableDefinition::new_from_schema(schema)),
+            data: Box::new(RecordBatch::new_empty(schema)) as Box<dyn Scannable>,
             mode: CreateTableMode::Create,
             write_options: Default::default(),
             location: None,
@@ -1380,7 +1377,7 @@ mod tests {
         db.create_table(CreateTableRequest {
             name: "source".to_string(),
             namespace: vec![],
-            data: CreateTableData::Empty(TableDefinition::new_from_schema(schema)),
+            data: Box::new(RecordBatch::new_empty(schema)) as Box<dyn Scannable>,
             mode: CreateTableMode::Create,
             write_options: Default::default(),
             location: None,
@@ -1435,7 +1432,7 @@ mod tests {
         db.create_table(CreateTableRequest {
             name: "source".to_string(),
             namespace: vec![],
-            data: CreateTableData::Empty(TableDefinition::new_from_schema(schema)),
+            data: Box::new(RecordBatch::new_empty(schema)) as Box<dyn Scannable>,
             mode: CreateTableMode::Create,
             write_options: Default::default(),
             location: None,
@@ -1484,16 +1481,11 @@ mod tests {
         )
         .unwrap();
 
-        let reader = Box::new(arrow_array::RecordBatchIterator::new(
-            vec![Ok(batch1)],
-            schema.clone(),
-        ));
-
         let source_table = db
             .create_table(CreateTableRequest {
                 name: "versioned_source".to_string(),
                 namespace: vec![],
-                data: CreateTableData::Data(reader),
+                data: Box::new(batch1) as Box<dyn Scannable>,
                 mode: CreateTableMode::Create,
                 write_options: Default::default(),
                 location: None,
@@ -1517,14 +1509,7 @@ mod tests {
 
         let db = Arc::new(db);
         let source_table_obj = Table::new(source_table.clone(), db.clone());
-        source_table_obj
-            .add(Box::new(arrow_array::RecordBatchIterator::new(
-                vec![Ok(batch2)],
-                schema.clone(),
-            )))
-            .execute()
-            .await
-            .unwrap();
+        source_table_obj.add(batch2).execute().await.unwrap();
 
         // Verify source table now has 4 rows
         assert_eq!(source_table.count_rows(None).await.unwrap(), 4);
@@ -1570,16 +1555,11 @@ mod tests {
         )
         .unwrap();
 
-        let reader = Box::new(arrow_array::RecordBatchIterator::new(
-            vec![Ok(batch1)],
-            schema.clone(),
-        ));
-
         let source_table = db
             .create_table(CreateTableRequest {
                 name: "tagged_source".to_string(),
                 namespace: vec![],
-                data: CreateTableData::Data(reader),
+                data: Box::new(batch1),
                 mode: CreateTableMode::Create,
                 write_options: Default::default(),
                 location: None,
@@ -1607,14 +1587,7 @@ mod tests {
         .unwrap();
 
         let source_table_obj = Table::new(source_table.clone(), db.clone());
-        source_table_obj
-            .add(Box::new(arrow_array::RecordBatchIterator::new(
-                vec![Ok(batch2)],
-                schema.clone(),
-            )))
-            .execute()
-            .await
-            .unwrap();
+        source_table_obj.add(batch2).execute().await.unwrap();
 
         // Source table should have 4 rows
         assert_eq!(source_table.count_rows(None).await.unwrap(), 4);
@@ -1657,16 +1630,11 @@ mod tests {
         )
         .unwrap();
 
-        let reader = Box::new(arrow_array::RecordBatchIterator::new(
-            vec![Ok(batch1)],
-            schema.clone(),
-        ));
-
         let source_table = db
             .create_table(CreateTableRequest {
                 name: "independent_source".to_string(),
                 namespace: vec![],
-                data: CreateTableData::Data(reader),
+                data: Box::new(batch1),
                 mode: CreateTableMode::Create,
                 write_options: Default::default(),
                 location: None,
@@ -1706,14 +1674,7 @@ mod tests {
 
         let db = Arc::new(db);
         let cloned_table_obj = Table::new(cloned_table.clone(), db.clone());
-        cloned_table_obj
-            .add(Box::new(arrow_array::RecordBatchIterator::new(
-                vec![Ok(batch_clone)],
-                schema.clone(),
-            )))
-            .execute()
-            .await
-            .unwrap();
+        cloned_table_obj.add(batch_clone).execute().await.unwrap();
 
         // Add different data to the source table
         let batch_source = RecordBatch::try_new(
@@ -1726,14 +1687,7 @@ mod tests {
         .unwrap();
 
         let source_table_obj = Table::new(source_table.clone(), db);
-        source_table_obj
-            .add(Box::new(arrow_array::RecordBatchIterator::new(
-                vec![Ok(batch_source)],
-                schema.clone(),
-            )))
-            .execute()
-            .await
-            .unwrap();
+        source_table_obj.add(batch_source).execute().await.unwrap();
 
         // Verify they have evolved independently
         assert_eq!(source_table.count_rows(None).await.unwrap(), 4); // 2 + 2
@@ -1751,16 +1705,11 @@ mod tests {
             RecordBatch::try_new(schema.clone(), vec![Arc::new(Int32Array::from(vec![1, 2]))])
                 .unwrap();
 
-        let reader = Box::new(arrow_array::RecordBatchIterator::new(
-            vec![Ok(batch1)],
-            schema.clone(),
-        ));
-
         let source_table = db
             .create_table(CreateTableRequest {
                 name: "latest_version_source".to_string(),
                 namespace: vec![],
-                data: CreateTableData::Data(reader),
+                data: Box::new(batch1),
                 mode: CreateTableMode::Create,
                 write_options: Default::default(),
                 location: None,
@@ -1779,14 +1728,7 @@ mod tests {
             .unwrap();
 
             let source_table_obj = Table::new(source_table.clone(), db.clone());
-            source_table_obj
-                .add(Box::new(arrow_array::RecordBatchIterator::new(
-                    vec![Ok(batch)],
-                    schema.clone(),
-                )))
-                .execute()
-                .await
-                .unwrap();
+            source_table_obj.add(batch).execute().await.unwrap();
         }
 
         // Source should have 8 rows total (2 + 2 + 2 + 2)
@@ -1849,16 +1791,11 @@ mod tests {
         )
         .unwrap();
 
-        let reader = Box::new(arrow_array::RecordBatchIterator::new(
-            vec![Ok(batch)],
-            schema.clone(),
-        ));
-
         let table = db
             .create_table(CreateTableRequest {
                 name: "test_stable".to_string(),
                 namespace: vec![],
-                data: CreateTableData::Data(reader),
+                data: Box::new(batch),
                 mode: CreateTableMode::Create,
                 write_options: Default::default(),
                 location: None,
@@ -1887,11 +1824,6 @@ mod tests {
         )
         .unwrap();
 
-        let reader = Box::new(arrow_array::RecordBatchIterator::new(
-            vec![Ok(batch)],
-            schema.clone(),
-        ));
-
         let mut storage_options = HashMap::new();
         storage_options.insert(
             OPT_NEW_TABLE_ENABLE_STABLE_ROW_IDS.to_string(),
@@ -1914,7 +1846,7 @@ mod tests {
             .create_table(CreateTableRequest {
                 name: "test_stable_table_level".to_string(),
                 namespace: vec![],
-                data: CreateTableData::Data(reader),
+                data: Box::new(batch),
                 mode: CreateTableMode::Create,
                 write_options,
                 location: None,
@@ -1963,11 +1895,6 @@ mod tests {
         )
         .unwrap();
 
-        let reader = Box::new(arrow_array::RecordBatchIterator::new(
-            vec![Ok(batch)],
-            schema.clone(),
-        ));
-
         let mut storage_options = HashMap::new();
         storage_options.insert(
             OPT_NEW_TABLE_ENABLE_STABLE_ROW_IDS.to_string(),
@@ -1990,7 +1917,7 @@ mod tests {
             .create_table(CreateTableRequest {
                 name: "test_override".to_string(),
                 namespace: vec![],
-                data: CreateTableData::Data(reader),
+                data: Box::new(batch),
                 mode: CreateTableMode::Create,
                 write_options,
                 location: None,
@@ -2108,7 +2035,7 @@ mod tests {
         db.create_table(CreateTableRequest {
             name: "table1".to_string(),
             namespace: vec![],
-            data: CreateTableData::Empty(TableDefinition::new_from_schema(schema.clone())),
+            data: Box::new(RecordBatch::new_empty(schema.clone())) as Box<dyn Scannable>,
             mode: CreateTableMode::Create,
             write_options: Default::default(),
             location: None,
@@ -2120,7 +2047,7 @@ mod tests {
         db.create_table(CreateTableRequest {
             name: "table2".to_string(),
             namespace: vec![],
-            data: CreateTableData::Empty(TableDefinition::new_from_schema(schema)),
+            data: Box::new(RecordBatch::new_empty(schema)) as Box<dyn Scannable>,
             mode: CreateTableMode::Create,
             write_options: Default::default(),
             location: None,

rust/lancedb/src/database/namespace.rs

@@ -354,15 +354,13 @@ mod tests {
     use super::*;
     use crate::connect_namespace;
     use crate::query::ExecutableQuery;
-    use arrow_array::{Int32Array, RecordBatch, RecordBatchIterator, StringArray};
+    use arrow_array::{Int32Array, RecordBatch, StringArray};
     use arrow_schema::{DataType, Field, Schema};
     use futures::TryStreamExt;
     use tempfile::tempdir;
 
     /// Helper function to create test data
-    fn create_test_data() -> RecordBatchIterator<
-        std::vec::IntoIter<std::result::Result<RecordBatch, arrow_schema::ArrowError>>,
-    > {
+    fn create_test_data() -> RecordBatch {
         let schema = Arc::new(Schema::new(vec![
             Field::new("id", DataType::Int32, false),
             Field::new("name", DataType::Utf8, false),
@@ -371,12 +369,7 @@ mod tests {
         let id_array = Int32Array::from(vec![1, 2, 3, 4, 5]);
         let name_array = StringArray::from(vec!["Alice", "Bob", "Charlie", "David", "Eve"]);
 
-        let batch = RecordBatch::try_new(
-            schema.clone(),
-            vec![Arc::new(id_array), Arc::new(name_array)],
-        )
-        .unwrap();
-        RecordBatchIterator::new(vec![std::result::Result::Ok(batch)].into_iter(), schema)
+        RecordBatch::try_new(schema, vec![Arc::new(id_array), Arc::new(name_array)]).unwrap()
     }
 
     #[tokio::test]
@@ -618,13 +611,7 @@ mod tests {
 
         // Test: Overwrite the table
         let table2 = conn
-            .create_table(
-                "overwrite_test",
-                RecordBatchIterator::new(
-                    vec![std::result::Result::Ok(test_data2)].into_iter(),
-                    schema,
-                ),
-            )
+            .create_table("overwrite_test", test_data2)
             .namespace(vec!["test_ns".into()])
             .mode(CreateTableMode::Overwrite)
             .execute()

rust/lancedb/src/dataloader/permutation/builder.rs

@@ -13,7 +13,7 @@ use lance_datafusion::exec::SessionContextExt;
 use crate::{
     arrow::{SendableRecordBatchStream, SendableRecordBatchStreamExt, SimpleRecordBatchStream},
     connect,
-    database::{CreateTableData, CreateTableRequest, Database},
+    database::{CreateTableRequest, Database},
     dataloader::permutation::{
         shuffle::{Shuffler, ShufflerConfig},
         split::{SplitStrategy, Splitter, SPLIT_ID_COLUMN},
@@ -313,10 +313,8 @@ impl PermutationBuilder {
             }
         };
 
-        let create_table_request = CreateTableRequest::new(
-            name.to_string(),
-            CreateTableData::StreamingData(streaming_data),
-        );
+        let create_table_request =
+            CreateTableRequest::new(name.to_string(), Box::new(streaming_data));
 
         let table = database.create_table(create_table_request).await?;
 
@@ -347,7 +345,7 @@ mod tests {
             .col("col_b", lance_datagen::array::step::<Int32Type>())
             .into_ldb_stream(RowCount::from(100), BatchCount::from(10));
         let data_table = db
-            .create_table_streaming("base_tbl", initial_data)
+            .create_table("base_tbl", initial_data)
             .execute()
             .await
             .unwrap();
@@ -387,7 +385,7 @@ mod tests {
             .col("some_value", lance_datagen::array::step::<Int32Type>())
             .into_ldb_stream(RowCount::from(100), BatchCount::from(10));
         let data_table = db
-            .create_table_streaming("mytbl", initial_data)
+            .create_table("mytbl", initial_data)
             .execute()
             .await
             .unwrap();

rust/lancedb/src/dataloader/permutation/reader.rs

@@ -39,6 +39,9 @@ pub struct PermutationReader {
     limit: Option<u64>,
     available_rows: u64,
     split: u64,
+    // Cached map of offset to row id for the split
+    #[allow(clippy::type_complexity)]
+    offset_map: Arc<tokio::sync::Mutex<Option<Arc<HashMap<u64, u64>>>>>,
 }
 
 impl std::fmt::Debug for PermutationReader {
@@ -72,6 +75,7 @@ impl PermutationReader {
             limit: None,
             available_rows: 0,
             split,
+            offset_map: Arc::new(tokio::sync::Mutex::new(None)),
         };
         slf.validate().await?;
         // Calculate the number of available rows
@@ -157,6 +161,7 @@ impl PermutationReader {
         let available_rows = self.verify_limit_offset(self.limit, Some(offset)).await?;
         self.offset = Some(offset);
         self.available_rows = available_rows;
+        self.offset_map = Arc::new(tokio::sync::Mutex::new(None));
         Ok(self)
     }
 
@@ -164,6 +169,7 @@ impl PermutationReader {
         let available_rows = self.verify_limit_offset(Some(limit), self.offset).await?;
         self.available_rows = available_rows;
         self.limit = Some(limit);
+        self.offset_map = Arc::new(tokio::sync::Mutex::new(None));
         Ok(self)
     }
 
@@ -180,8 +186,9 @@ impl PermutationReader {
         base_table: &Arc<dyn BaseTable>,
         row_ids: RecordBatch,
         selection: Select,
-        has_row_id: bool,
     ) -> Result<RecordBatch> {
+        let has_row_id = Self::has_row_id(&selection)?;
+
         let num_rows = row_ids.num_rows();
         let row_ids = row_ids
             .column(0)
@@ -282,14 +289,13 @@ impl PermutationReader {
         row_ids: DatasetRecordBatchStream,
         selection: Select,
     ) -> Result<SendableRecordBatchStream> {
-        let has_row_id = Self::has_row_id(&selection)?;
         let mut stream = row_ids
             .map_err(Error::from)
             .try_filter_map(move |batch| {
                 let selection = selection.clone();
                 let base_table = base_table.clone();
                 async move {
-                    Self::load_batch(&base_table, batch, selection, has_row_id)
+                    Self::load_batch(&base_table, batch, selection)
                         .await
                         .map(Some)
                 }
@@ -397,6 +403,82 @@ impl PermutationReader {
         Self::row_ids_to_batches(self.base_table.clone(), row_ids, selection).await
     }
 
+    /// If we are going to use `take` then we load the offset -> row id map once for the split and cache it
+    ///
+    /// This method fetches the map with find-or-create semantics.
+    async fn get_offset_map(
+        &self,
+        permutation_table: &Arc<dyn BaseTable>,
+    ) -> Result<Arc<HashMap<u64, u64>>> {
+        let mut offset_map_ref = self.offset_map.lock().await;
+        if let Some(offset_map) = &*offset_map_ref {
+            return Ok(offset_map.clone());
+        }
+        let mut offset_map = HashMap::new();
+        let mut row_ids_query = Table::from(permutation_table.clone())
+            .query()
+            .select(Select::Columns(vec![SRC_ROW_ID_COL.to_string()]))
+            .only_if(format!("{} = {}", SPLIT_ID_COLUMN, self.split));
+        if let Some(offset) = self.offset {
+            row_ids_query = row_ids_query.offset(offset as usize);
+        }
+        if let Some(limit) = self.limit {
+            row_ids_query = row_ids_query.limit(limit as usize);
+        }
+        let mut row_ids = row_ids_query.execute().await?;
+        while let Some(batch) = row_ids.try_next().await? {
+            let row_ids = batch
+                .column(0)
+                .as_primitive::<UInt64Type>()
+                .values()
+                .to_vec();
+            for (i, row_id) in row_ids.iter().enumerate() {
+                offset_map.insert(i as u64, *row_id);
+            }
+        }
+        let offset_map = Arc::new(offset_map);
+        *offset_map_ref = Some(offset_map.clone());
+        Ok(offset_map)
+    }
+
+    pub async fn take_offsets(&self, offsets: &[u64], selection: Select) -> Result<RecordBatch> {
+        if let Some(permutation_table) = &self.permutation_table {
+            let offset_map = self.get_offset_map(permutation_table).await?;
+            let row_ids = offsets
+                .iter()
+                .map(|o| offset_map.get(o).copied().expect_ok().map_err(Error::from))
+                .collect::<Result<Vec<_>>>()?;
+            let row_ids = RecordBatch::try_new(
+                Arc::new(arrow_schema::Schema::new(vec![arrow_schema::Field::new(
+                    "row_id",
+                    arrow_schema::DataType::UInt64,
+                    false,
+                )])),
+                vec![Arc::new(UInt64Array::from(row_ids))],
+            )?;
+            Self::load_batch(&self.base_table, row_ids, selection).await
+        } else {
+            let table = Table::from(self.base_table.clone());
+            let batches = table
+                .take_offsets(offsets.to_vec())
+                .select(selection.clone())
+                .execute()
+                .await?
+                .try_collect::<Vec<_>>()
+                .await?;
+            if let Some(first_batch) = batches.first() {
+                let schema = first_batch.schema();
+                let batch = arrow::compute::concat_batches(&schema, &batches)?;
+                Ok(batch)
+            } else {
+                Ok(RecordBatch::try_new(
+                    self.output_schema(selection).await?,
+                    vec![],
+                )?)
+            }
+        }
+    }
+
     pub async fn output_schema(&self, selection: Select) -> Result<SchemaRef> {
         let table = Table::from(self.base_table.clone());
         table.query().select(selection).output_schema().await
@@ -543,4 +625,224 @@ mod tests {
         check_batch(&mut stream, &row_ids[7..9]).await;
         assert!(stream.try_next().await.unwrap().is_none());
     }
+
+    /// Helper to create a base table and permutation table for take_offsets tests.
+    /// Returns (base_table, row_ids_table, shuffled_row_ids).
+    async fn setup_permutation_tables(num_rows: usize) -> (Table, Table, Vec<u64>) {
+        let base_table = lance_datagen::gen_batch()
+            .col("idx", lance_datagen::array::step::<Int32Type>())
+            .col("other_col", lance_datagen::array::step::<UInt64Type>())
+            .into_mem_table("tbl", RowCount::from(num_rows as u64), BatchCount::from(1))
+            .await;
+
+        let mut row_ids = collect_column::<UInt64Type>(&base_table, "_rowid").await;
+        row_ids.shuffle(&mut rand::rng());
+
+        let split_ids = UInt64Array::from_iter_values(std::iter::repeat_n(0u64, row_ids.len()));
+        let permutation_batch = RecordBatch::try_new(
+            Arc::new(Schema::new(vec![
+                Field::new("row_id", DataType::UInt64, false),
+                Field::new(SPLIT_ID_COLUMN, DataType::UInt64, false),
+            ])),
+            vec![
+                Arc::new(UInt64Array::from(row_ids.clone())),
+                Arc::new(split_ids),
+            ],
+        )
+        .unwrap();
+        let row_ids_table = virtual_table("row_ids", &permutation_batch).await;
+
+        (base_table, row_ids_table, row_ids)
+    }
+
+    #[tokio::test]
+    async fn test_take_offsets_with_permutation_table() {
+        let (base_table, row_ids_table, row_ids) = setup_permutation_tables(10).await;
+
+        let reader = PermutationReader::try_from_tables(
+            base_table.base_table().clone(),
+            row_ids_table.base_table().clone(),
+            0,
+        )
+        .await
+        .unwrap();
+
+        // Take specific offsets and verify the returned rows match the permutation order
+        let offsets = vec![0, 2, 4];
+        let batch = reader.take_offsets(&offsets, Select::All).await.unwrap();
+
+        assert_eq!(batch.num_rows(), 3);
+
+        let idx_values = batch
+            .column(0)
+            .as_primitive::<Int32Type>()
+            .values()
+            .to_vec();
+        let expected: Vec<i32> = offsets
+            .iter()
+            .map(|&o| row_ids[o as usize] as i32)
+            .collect();
+        assert_eq!(idx_values, expected);
+    }
+
+    #[tokio::test]
+    async fn test_take_offsets_preserves_order() {
+        let (base_table, row_ids_table, row_ids) = setup_permutation_tables(10).await;
+
+        let reader = PermutationReader::try_from_tables(
+            base_table.base_table().clone(),
+            row_ids_table.base_table().clone(),
+            0,
+        )
+        .await
+        .unwrap();
+
+        // Take offsets in reverse order and verify returned rows match that order
+        let offsets = vec![5, 3, 1, 0];
+        let batch = reader.take_offsets(&offsets, Select::All).await.unwrap();
+
+        assert_eq!(batch.num_rows(), 4);
+
+        let idx_values = batch
+            .column(0)
+            .as_primitive::<Int32Type>()
+            .values()
+            .to_vec();
+        let expected: Vec<i32> = offsets
+            .iter()
+            .map(|&o| row_ids[o as usize] as i32)
+            .collect();
+        assert_eq!(idx_values, expected);
+    }
+
+    #[tokio::test]
+    async fn test_take_offsets_with_column_selection() {
+        let (base_table, row_ids_table, row_ids) = setup_permutation_tables(10).await;
+
+        let reader = PermutationReader::try_from_tables(
+            base_table.base_table().clone(),
+            row_ids_table.base_table().clone(),
+            0,
+        )
+        .await
+        .unwrap();
+
+        let offsets = vec![1, 3];
+        let batch = reader
+            .take_offsets(&offsets, Select::Columns(vec!["idx".to_string()]))
+            .await
+            .unwrap();
+
+        assert_eq!(batch.num_rows(), 2);
+        assert_eq!(batch.num_columns(), 1);
+        assert_eq!(batch.schema().field(0).name(), "idx");
+
+        let idx_values = batch
+            .column(0)
+            .as_primitive::<Int32Type>()
+            .values()
+            .to_vec();
+        let expected: Vec<i32> = offsets
+            .iter()
+            .map(|&o| row_ids[o as usize] as i32)
+            .collect();
+        assert_eq!(idx_values, expected);
+    }
+
+    #[tokio::test]
+    async fn test_take_offsets_invalid_offset() {
+        let (base_table, row_ids_table, _) = setup_permutation_tables(5).await;
+
+        let reader = PermutationReader::try_from_tables(
+            base_table.base_table().clone(),
+            row_ids_table.base_table().clone(),
+            0,
+        )
+        .await
+        .unwrap();
+
+        // Offset 999 doesn't exist in the offset map
+        let result = reader.take_offsets(&[0, 999], Select::All).await;
+        assert!(result.is_err());
+    }
+
+    #[tokio::test]
+    async fn test_take_offsets_identity_reader() {
+        let base_table = lance_datagen::gen_batch()
+            .col("idx", lance_datagen::array::step::<Int32Type>())
+            .into_mem_table("tbl", RowCount::from(10), BatchCount::from(1))
+            .await;
+
+        let reader = PermutationReader::identity(base_table.base_table().clone()).await;
+
+        // With no permutation table, take_offsets uses the base table directly
+        let offsets = vec![0, 2, 4, 6];
+        let batch = reader.take_offsets(&offsets, Select::All).await.unwrap();
+
+        assert_eq!(batch.num_rows(), 4);
+
+        let idx_values = batch
+            .column(0)
+            .as_primitive::<Int32Type>()
+            .values()
+            .to_vec();
+        assert_eq!(idx_values, vec![0, 2, 4, 6]);
+    }
+
+    #[tokio::test]
+    async fn test_take_offsets_caches_offset_map() {
+        let (base_table, row_ids_table, row_ids) = setup_permutation_tables(10).await;
+
+        let reader = PermutationReader::try_from_tables(
+            base_table.base_table().clone(),
+            row_ids_table.base_table().clone(),
+            0,
+        )
+        .await
+        .unwrap();
+
+        // First call populates the cache
+        let batch1 = reader.take_offsets(&[0, 1], Select::All).await.unwrap();
+
+        // Second call should use the cached offset map and produce consistent results
+        let batch2 = reader.take_offsets(&[0, 1], Select::All).await.unwrap();
+
+        let values1 = batch1
+            .column(0)
+            .as_primitive::<Int32Type>()
+            .values()
+            .to_vec();
+        let values2 = batch2
+            .column(0)
+            .as_primitive::<Int32Type>()
+            .values()
+            .to_vec();
+        assert_eq!(values1, values2);
+
+        let expected: Vec<i32> = vec![row_ids[0] as i32, row_ids[1] as i32];
+        assert_eq!(values1, expected);
+    }
+
+    #[tokio::test]
+    async fn test_take_offsets_single_offset() {
+        let (base_table, row_ids_table, row_ids) = setup_permutation_tables(5).await;
+
+        let reader = PermutationReader::try_from_tables(
+            base_table.base_table().clone(),
+            row_ids_table.base_table().clone(),
+            0,
+        )
+        .await
+        .unwrap();
+
+        let batch = reader.take_offsets(&[2], Select::All).await.unwrap();
+
+        assert_eq!(batch.num_rows(), 1);
+        let idx_values = batch
+            .column(0)
+            .as_primitive::<Int32Type>()
+            .values()
+            .to_vec();
+        assert_eq!(idx_values, vec![row_ids[2] as i32]);
+    }
 }

rust/lancedb/src/embeddings.rs

@@ -18,7 +18,7 @@ use std::{
 };
 
 use arrow_array::{Array, RecordBatch, RecordBatchReader};
-use arrow_schema::{DataType, Field, SchemaBuilder};
+use arrow_schema::{DataType, Field, SchemaBuilder, SchemaRef};
 // use async_trait::async_trait;
 use serde::{Deserialize, Serialize};
 
@@ -190,6 +190,112 @@ impl<R: RecordBatchReader> WithEmbeddings<R> {
     }
 }
 
+/// Compute embedding arrays for a batch.
+///
+/// When multiple embedding functions are defined, they are computed in parallel using
+/// scoped threads. For a single embedding function, computation is done inline.
+fn compute_embedding_arrays(
+    batch: &RecordBatch,
+    embeddings: &[(EmbeddingDefinition, Arc<dyn EmbeddingFunction>)],
+) -> Result<Vec<Arc<dyn Array>>> {
+    if embeddings.len() == 1 {
+        let (fld, func) = &embeddings[0];
+        let src_column =
+            batch
+                .column_by_name(&fld.source_column)
+                .ok_or_else(|| Error::InvalidInput {
+                    message: format!("Source column '{}' not found", fld.source_column),
+                })?;
+        return Ok(vec![func.compute_source_embeddings(src_column.clone())?]);
+    }
+
+    // Parallel path: multiple embeddings
+    std::thread::scope(|s| {
+        let handles: Vec<_> = embeddings
+            .iter()
+            .map(|(fld, func)| {
+                let src_column = batch.column_by_name(&fld.source_column).ok_or_else(|| {
+                    Error::InvalidInput {
+                        message: format!("Source column '{}' not found", fld.source_column),
+                    }
+                })?;
+
+                let handle = s.spawn(move || func.compute_source_embeddings(src_column.clone()));
+
+                Ok(handle)
+            })
+            .collect::<Result<_>>()?;
+
+        handles
+            .into_iter()
+            .map(|h| {
+                h.join().map_err(|e| Error::Runtime {
+                    message: format!("Thread panicked during embedding computation: {:?}", e),
+                })?
+            })
+            .collect()
+    })
+}
+
+/// Compute the output schema when embeddings are applied to a base schema.
+///
+/// This returns the schema with embedding columns appended.
+pub fn compute_output_schema(
+    base_schema: &SchemaRef,
+    embeddings: &[(EmbeddingDefinition, Arc<dyn EmbeddingFunction>)],
+) -> Result<SchemaRef> {
+    let mut sb: SchemaBuilder = base_schema.as_ref().into();
+
+    for (ed, func) in embeddings {
+        let src_field = base_schema
+            .field_with_name(&ed.source_column)
+            .map_err(|_| Error::InvalidInput {
+                message: format!("Source column '{}' not found in schema", ed.source_column),
+            })?;
+
+        let field_name = ed
+            .dest_column
+            .clone()
+            .unwrap_or_else(|| format!("{}_embedding", &ed.source_column));
+
+        sb.push(Field::new(
+            field_name,
+            func.dest_type()?.into_owned(),
+            src_field.is_nullable(),
+        ));
+    }
+
+    Ok(Arc::new(sb.finish()))
+}
+
+/// Compute embeddings for a batch and append as new columns.
+///
+/// This function computes embeddings using the provided embedding functions and
+/// appends them as new columns to the batch.
+pub fn compute_embeddings_for_batch(
+    batch: RecordBatch,
+    embeddings: &[(EmbeddingDefinition, Arc<dyn EmbeddingFunction>)],
+) -> Result<RecordBatch> {
+    let embedding_arrays = compute_embedding_arrays(&batch, embeddings)?;
+
+    let mut result = batch;
+    for ((fld, _), embedding) in embeddings.iter().zip(embedding_arrays.iter()) {
+        let dst_field_name = fld
+            .dest_column
+            .clone()
+            .unwrap_or_else(|| format!("{}_embedding", &fld.source_column));
+
+        let dst_field = Field::new(
+            dst_field_name,
+            embedding.data_type().clone(),
+            embedding.nulls().is_some(),
+        );
+
+        result = result.try_with_column(dst_field, embedding.clone())?;
+    }
+    Ok(result)
+}
+
 impl<R: RecordBatchReader> WithEmbeddings<R> {
     fn dest_fields(&self) -> Result<Vec<Field>> {
         let schema = self.inner.schema();
@@ -240,48 +346,6 @@ impl<R: RecordBatchReader> WithEmbeddings<R> {
             column_definitions,
         })
     }
-
-    fn compute_embeddings_parallel(&self, batch: &RecordBatch) -> Result<Vec<Arc<dyn Array>>> {
-        if self.embeddings.len() == 1 {
-            let (fld, func) = &self.embeddings[0];
-            let src_column =
-                batch
-                    .column_by_name(&fld.source_column)
-                    .ok_or_else(|| Error::InvalidInput {
-                        message: format!("Source column '{}' not found", fld.source_column),
-                    })?;
-            return Ok(vec![func.compute_source_embeddings(src_column.clone())?]);
-        }
-
-        // Parallel path: multiple embeddings
-        std::thread::scope(|s| {
-            let handles: Vec<_> = self
-                .embeddings
-                .iter()
-                .map(|(fld, func)| {
-                    let src_column = batch.column_by_name(&fld.source_column).ok_or_else(|| {
-                        Error::InvalidInput {
-                            message: format!("Source column '{}' not found", fld.source_column),
-                        }
-                    })?;
-
-                    let handle =
-                        s.spawn(move || func.compute_source_embeddings(src_column.clone()));
-
-                    Ok(handle)
-                })
-                .collect::<Result<_>>()?;
-
-            handles
-                .into_iter()
-                .map(|h| {
-                    h.join().map_err(|e| Error::Runtime {
-                        message: format!("Thread panicked during embedding computation: {:?}", e),
-                    })?
-                })
-                .collect()
-        })
-    }
 }
 
 impl<R: RecordBatchReader> Iterator for MaybeEmbedded<R> {
@@ -309,37 +373,13 @@ impl<R: RecordBatchReader> Iterator for WithEmbeddings<R> {
     fn next(&mut self) -> Option<Self::Item> {
         let batch = self.inner.next()?;
         match batch {
-            Ok(batch) => {
-                let embeddings = match self.compute_embeddings_parallel(&batch) {
-                    Ok(emb) => emb,
-                    Err(e) => {
-                        return Some(Err(arrow_schema::ArrowError::ComputeError(format!(
-                            "Error computing embedding: {}",
-                            e
-                        ))))
-                    }
-                };
-
-                let mut batch = batch;
-                for ((fld, _), embedding) in self.embeddings.iter().zip(embeddings.iter()) {
-                    let dst_field_name = fld
-                        .dest_column
-                        .clone()
-                        .unwrap_or_else(|| format!("{}_embedding", &fld.source_column));
-
-                    let dst_field = Field::new(
-                        dst_field_name,
-                        embedding.data_type().clone(),
-                        embedding.nulls().is_some(),
-                    );
-
-                    match batch.try_with_column(dst_field.clone(), embedding.clone()) {
-                        Ok(b) => batch = b,
-                        Err(e) => return Some(Err(e)),
-                    };
-                }
-                Some(Ok(batch))
-            }
+            Ok(batch) => match compute_embeddings_for_batch(batch, &self.embeddings) {
+                Ok(batch_with_embeddings) => Some(Ok(batch_with_embeddings)),
+                Err(e) => Some(Err(arrow_schema::ArrowError::ComputeError(format!(
+                    "Error computing embedding: {}",
+                    e
+                )))),
+            },
             Err(e) => Some(Err(e)),
         }
     }

rust/lancedb/src/error.rs

@@ -6,7 +6,7 @@ use std::sync::PoisonError;
 use arrow_schema::ArrowError;
 use snafu::Snafu;
 
-type BoxError = Box<dyn std::error::Error + Send + Sync>;
+pub(crate) type BoxError = Box<dyn std::error::Error + Send + Sync>;
 
 #[derive(Debug, Snafu)]
 #[snafu(visibility(pub(crate)))]
@@ -80,6 +80,9 @@ pub enum Error {
     Arrow { source: ArrowError },
     #[snafu(display("LanceDBError: not supported: {message}"))]
     NotSupported { message: String },
+    /// External error pass through from user code.
+    #[snafu(transparent)]
+    External { source: BoxError },
     #[snafu(whatever, display("{message}"))]
     Other {
         message: String,
@@ -92,15 +95,26 @@ pub type Result<T> = std::result::Result<T, Error>;
 
 impl From<ArrowError> for Error {
     fn from(source: ArrowError) -> Self {
-        Self::Arrow { source }
+        match source {
+            ArrowError::ExternalError(source) => match source.downcast::<Self>() {
+                Ok(e) => *e,
+                Err(source) => Self::External { source },
+            },
+            _ => Self::Arrow { source },
+        }
     }
 }
 
 impl From<lance::Error> for Error {
     fn from(source: lance::Error) -> Self {
-        // TODO: Once Lance is changed to preserve ObjectStore, DataFusion, and Arrow errors, we can
-        // pass those variants through here as well.
-        Self::Lance { source }
+        // Try to unwrap external errors that were wrapped by lance
+        match source {
+            lance::Error::Wrapped { error, .. } => match error.downcast::<Self>() {
+                Ok(e) => *e,
+                Err(source) => Self::External { source },
+            },
+            _ => Self::Lance { source },
+        }
     }
 }
 

rust/lancedb/src/io/object_store.rs

@@ -218,8 +218,9 @@ mod test {
         datagen = datagen.col(Box::<IncrementingInt32>::default());
         datagen = datagen.col(Box::new(RandomVector::default().named("vector".into())));
 
+        let data: Box<dyn arrow_array::RecordBatchReader + Send> = Box::new(datagen.batch(100));
         let res = db
-            .create_table("test", Box::new(datagen.batch(100)))
+            .create_table("test", data)
             .write_options(WriteOptions {
                 lance_write_params: Some(param),
             })

rust/lancedb/src/ipc.rs

@@ -12,10 +12,10 @@ use arrow_schema::Schema;
 use crate::{Error, Result};
 
 /// Convert a Arrow IPC file to a batch reader
-pub fn ipc_file_to_batches(buf: Vec<u8>) -> Result<impl RecordBatchReader> {
+pub fn ipc_file_to_batches(buf: Vec<u8>) -> Result<Box<dyn RecordBatchReader + Send>> {
     let buf_reader = Cursor::new(buf);
     let reader = FileReader::try_new(buf_reader, None)?;
-    Ok(reader)
+    Ok(Box::new(reader))
 }
 
 /// Convert record batches to Arrow IPC file

rust/lancedb/src/lib.rs

@@ -39,7 +39,6 @@
 //! #### Connect to a database.
 //!
 //! ```rust
-//! # use arrow_schema::{Field, Schema};
 //! # tokio::runtime::Runtime::new().unwrap().block_on(async {
 //! let db = lancedb::connect("data/sample-lancedb").execute().await.unwrap();
 //! # });
@@ -74,7 +73,10 @@
 //!
 //! #### Create a table
 //!
-//! To create a Table, you need to provide a [`arrow_schema::Schema`] and a [`arrow_array::RecordBatch`] stream.
+//! To create a Table, you need to provide an [`arrow_array::RecordBatch`]. The
+//! schema of the `RecordBatch` determines the schema of the table.
+//!
+//! Vector columns should be represented as `FixedSizeList<Float16/Float32>` data type.
 //!
 //! ```rust
 //! # use std::sync::Arc;
@@ -85,34 +87,29 @@
 //! # tokio::runtime::Runtime::new().unwrap().block_on(async {
 //! # let tmpdir = tempfile::tempdir().unwrap();
 //! # let db = lancedb::connect(tmpdir.path().to_str().unwrap()).execute().await.unwrap();
+//! let ndims = 128;
 //! let schema = Arc::new(Schema::new(vec![
 //!     Field::new("id", DataType::Int32, false),
 //!     Field::new(
 //!         "vector",
-//!         DataType::FixedSizeList(Arc::new(Field::new("item", DataType::Float32, true)), 128),
+//!         DataType::FixedSizeList(Arc::new(Field::new("item", DataType::Float32, true)), ndims),
 //!         true,
 //!     ),
 //! ]));
-//! // Create a RecordBatch stream.
-//! let batches = RecordBatchIterator::new(
-//!     vec![RecordBatch::try_new(
+//! let data = RecordBatch::try_new(
 //!         schema.clone(),
 //!         vec![
 //!             Arc::new(Int32Array::from_iter_values(0..256)),
 //!             Arc::new(
 //!                 FixedSizeListArray::from_iter_primitive::<Float32Type, _, _>(
-//!                     (0..256).map(|_| Some(vec![Some(1.0); 128])),
-//!                     128,
+//!                     (0..256).map(|_| Some(vec![Some(1.0); ndims as usize])),
+//!                     ndims,
 //!                 ),
 //!             ),
 //!         ],
 //!     )
-//!     .unwrap()]
-//!     .into_iter()
-//!     .map(Ok),
-//!     schema.clone(),
-//! );
-//! db.create_table("my_table", Box::new(batches))
+//!     .unwrap();
+//! db.create_table("my_table", data)
 //!     .execute()
 //!     .await
 //!     .unwrap();
@@ -151,42 +148,18 @@
 //! #### Open table and search
 //!
 //! ```rust
-//! # use std::sync::Arc;
 //! # use futures::TryStreamExt;
-//! # use arrow_schema::{DataType, Schema, Field};
-//! # use arrow_array::{RecordBatch, RecordBatchIterator};
-//! # use arrow_array::{FixedSizeListArray, Float32Array, Int32Array, types::Float32Type};
 //! # use lancedb::query::{ExecutableQuery, QueryBase};
-//! # tokio::runtime::Runtime::new().unwrap().block_on(async {
-//! # let tmpdir = tempfile::tempdir().unwrap();
-//! # let db = lancedb::connect(tmpdir.path().to_str().unwrap()).execute().await.unwrap();
-//! # let schema = Arc::new(Schema::new(vec![
-//! #  Field::new("id", DataType::Int32, false),
-//! #  Field::new("vector", DataType::FixedSizeList(
-//! #    Arc::new(Field::new("item", DataType::Float32, true)), 128), true),
-//! # ]));
-//! # let batches = RecordBatchIterator::new(vec![
-//! #    RecordBatch::try_new(schema.clone(),
-//! #       vec![
-//! #           Arc::new(Int32Array::from_iter_values(0..10)),
-//! #           Arc::new(FixedSizeListArray::from_iter_primitive::<Float32Type, _, _>(
-//! #               (0..10).map(|_| Some(vec![Some(1.0); 128])), 128)),
-//! #       ]).unwrap()
-//! #   ].into_iter().map(Ok),
-//! #    schema.clone());
-//! # db.create_table("my_table", Box::new(batches)).execute().await.unwrap();
-//! # let table = db.open_table("my_table").execute().await.unwrap();
+//! # async fn example(table: &lancedb::Table) -> lancedb::Result<()> {
 //! let results = table
 //!     .query()
-//!     .nearest_to(&[1.0; 128])
-//!     .unwrap()
+//!     .nearest_to(&[1.0; 128])?
 //!     .execute()
-//!     .await
-//!     .unwrap()
+//!     .await?
 //!     .try_collect::<Vec<_>>()
-//!     .await
-//!     .unwrap();
-//! # });
+//!     .await?;
+//! #   Ok(())
+//! # }
 //! ```
 
 pub mod arrow;

rust/lancedb/src/query.rs

@@ -1381,7 +1381,7 @@ mod tests {
     use arrow::{array::downcast_array, compute::concat_batches, datatypes::Int32Type};
     use arrow_array::{
         cast::AsArray, types::Float32Type, FixedSizeListArray, Float32Array, Int32Array,
-        RecordBatch, RecordBatchIterator, RecordBatchReader, StringArray,
+        RecordBatch, StringArray,
     };
     use arrow_schema::{DataType, Field as ArrowField, Schema as ArrowSchema};
     use futures::{StreamExt, TryStreamExt};
@@ -1402,7 +1402,7 @@ mod tests {
         let batches = make_test_batches();
         let conn = connect(uri).execute().await.unwrap();
         let table = conn
-            .create_table("my_table", Box::new(batches))
+            .create_table("my_table", batches)
             .execute()
             .await
             .unwrap();
@@ -1463,7 +1463,7 @@ mod tests {
         let batches = make_non_empty_batches();
         let conn = connect(uri).execute().await.unwrap();
         let table = conn
-            .create_table("my_table", Box::new(batches))
+            .create_table("my_table", batches)
             .execute()
             .await
             .unwrap();
@@ -1525,7 +1525,7 @@ mod tests {
         let batches = make_non_empty_batches();
         let conn = connect(uri).execute().await.unwrap();
         let table = conn
-            .create_table("my_table", Box::new(batches))
+            .create_table("my_table", batches)
             .execute()
             .await
             .unwrap();
@@ -1578,7 +1578,7 @@ mod tests {
         let batches = make_non_empty_batches();
         let conn = connect(uri).execute().await.unwrap();
         let table = conn
-            .create_table("my_table", Box::new(batches))
+            .create_table("my_table", batches)
             .execute()
             .await
             .unwrap();
@@ -1599,13 +1599,13 @@ mod tests {
         assert!(result.is_err());
     }
 
-    fn make_non_empty_batches() -> impl RecordBatchReader + Send + 'static {
+    fn make_non_empty_batches() -> Box<dyn arrow_array::RecordBatchReader + Send> {
         let vec = Box::new(RandomVector::new().named("vector".to_string()));
         let id = Box::new(IncrementingInt32::new().named("id".to_string()));
-        BatchGenerator::new().col(vec).col(id).batch(512)
+        Box::new(BatchGenerator::new().col(vec).col(id).batch(512))
     }
 
-    fn make_test_batches() -> impl RecordBatchReader + Send + 'static {
+    fn make_test_batches() -> RecordBatch {
         let dim: usize = 128;
         let schema = Arc::new(ArrowSchema::new(vec![
             ArrowField::new("key", DataType::Int32, false),
@@ -1619,12 +1619,7 @@ mod tests {
             ),
             ArrowField::new("uri", DataType::Utf8, true),
         ]));
-        RecordBatchIterator::new(
-            vec![RecordBatch::new_empty(schema.clone())]
-                .into_iter()
-                .map(Ok),
-            schema,
-        )
+        RecordBatch::new_empty(schema)
     }
 
     async fn make_test_table(tmp_dir: &tempfile::TempDir) -> Table {
@@ -1633,7 +1628,7 @@ mod tests {
 
         let batches = make_non_empty_batches();
         let conn = connect(uri).execute().await.unwrap();
-        conn.create_table("my_table", Box::new(batches))
+        conn.create_table("my_table", batches)
             .execute()
             .await
             .unwrap()
@@ -1862,10 +1857,8 @@ mod tests {
 
         let record_batch =
             RecordBatch::try_new(schema.clone(), vec![Arc::new(text), Arc::new(vector)]).unwrap();
-        let record_batch_iter =
-            RecordBatchIterator::new(vec![record_batch].into_iter().map(Ok), schema.clone());
         let table = conn
-            .create_table("my_table", record_batch_iter)
+            .create_table("my_table", record_batch)
             .execute()
             .await
             .unwrap();
@@ -1949,10 +1942,8 @@ mod tests {
             ],
         )
         .unwrap();
-        let record_batch_iter =
-            RecordBatchIterator::new(vec![record_batch].into_iter().map(Ok), schema.clone());
         let table = conn
-            .create_table("my_table", record_batch_iter)
+            .create_table("my_table", record_batch)
             .mode(CreateTableMode::Overwrite)
             .execute()
             .await
@@ -2062,8 +2053,6 @@ mod tests {
     async fn test_pagination_with_fts() {
         let db = connect("memory://test").execute().await.unwrap();
         let data = fts_test_data(400);
-        let schema = data.schema();
-        let data = RecordBatchIterator::new(vec![Ok(data)], schema);
         let table = db.create_table("test_table", data).execute().await.unwrap();
 
         table

rust/lancedb/src/remote/client.rs

@@ -491,7 +491,7 @@ impl<S: HttpSend> RestfulLanceDbClient<S> {
     }
 
     /// Apply dynamic headers from the header provider if configured
-    async fn apply_dynamic_headers(&self, mut request: Request) -> Result<Request> {
+    pub(crate) async fn apply_dynamic_headers(&self, mut request: Request) -> Result<Request> {
         if let Some(ref provider) = self.header_provider {
             let headers = provider.get_headers().await?;
             let request_headers = request.headers_mut();
@@ -555,7 +555,9 @@ impl<S: HttpSend> RestfulLanceDbClient<S> {
             message: "Attempted to retry a request that cannot be cloned".to_string(),
         })?;
         let (_, r) = tmp_req.build_split();
-        let mut r = r.unwrap();
+        let mut r = r.map_err(|e| Error::Runtime {
+            message: format!("Failed to build request: {}", e),
+        })?;
         let request_id = self.extract_request_id(&mut r);
         let mut retry_counter = RetryCounter::new(retry_config, request_id.clone());
 
@@ -571,7 +573,9 @@ impl<S: HttpSend> RestfulLanceDbClient<S> {
             }
 
             let (c, request) = req_builder.build_split();
-            let mut request = request.unwrap();
+            let mut request = request.map_err(|e| Error::Runtime {
+                message: format!("Failed to build request: {}", e),
+            })?;
             self.set_request_id(&mut request, &request_id.clone());
 
             // Apply dynamic headers before each retry attempt
@@ -621,7 +625,7 @@ impl<S: HttpSend> RestfulLanceDbClient<S> {
         }
     }
 
-    fn log_request(&self, request: &Request, request_id: &String) {
+    pub(crate) fn log_request(&self, request: &Request, request_id: &String) {
         if log::log_enabled!(log::Level::Debug) {
             let content_type = request
                 .headers()

rust/lancedb/src/remote/db.rs

@@ -4,13 +4,11 @@
 use std::collections::HashMap;
 use std::sync::Arc;
 
-use arrow_array::RecordBatchIterator;
 use async_trait::async_trait;
 use http::StatusCode;
 use lance_io::object_store::StorageOptions;
 use moka::future::Cache;
 use reqwest::header::CONTENT_TYPE;
-use tokio::task::spawn_blocking;
 
 use lance_namespace::models::{
     CreateNamespaceRequest, CreateNamespaceResponse, DescribeNamespaceRequest,
@@ -19,16 +17,17 @@ use lance_namespace::models::{
 };
 
 use crate::database::{
-    CloneTableRequest, CreateTableData, CreateTableMode, CreateTableRequest, Database,
-    DatabaseOptions, OpenTableRequest, ReadConsistency, TableNamesRequest,
+    CloneTableRequest, CreateTableMode, CreateTableRequest, Database, DatabaseOptions,
+    OpenTableRequest, ReadConsistency, TableNamesRequest,
 };
 use crate::error::Result;
+use crate::remote::util::stream_as_body;
 use crate::table::BaseTable;
 use crate::Error;
 
 use super::client::{ClientConfig, HttpSend, RequestResultExt, RestfulLanceDbClient, Sender};
 use super::table::RemoteTable;
-use super::util::{batches_to_ipc_bytes, parse_server_version};
+use super::util::parse_server_version;
 use super::ARROW_STREAM_CONTENT_TYPE;
 
 // Request structure for the remote clone table API
@@ -436,26 +435,8 @@ impl<S: HttpSend> Database for RemoteDatabase<S> {
         Ok(response)
     }
 
-    async fn create_table(&self, request: CreateTableRequest) -> Result<Arc<dyn BaseTable>> {
-        let data = match request.data {
-            CreateTableData::Data(data) => data,
-            CreateTableData::StreamingData(_) => {
-                return Err(Error::NotSupported {
-                    message: "Creating a remote table from a streaming source".to_string(),
-                })
-            }
-            CreateTableData::Empty(table_definition) => {
-                let schema = table_definition.schema.clone();
-                Box::new(RecordBatchIterator::new(vec![], schema))
-            }
-        };
-
-        // TODO: https://github.com/lancedb/lancedb/issues/1026
-        // We should accept data from an async source.  In the meantime, spawn this as blocking
-        // to make sure we don't block the tokio runtime if the source is slow.
-        let data_buffer = spawn_blocking(move || batches_to_ipc_bytes(data))
-            .await
-            .unwrap()?;
+    async fn create_table(&self, mut request: CreateTableRequest) -> Result<Arc<dyn BaseTable>> {
+        let body = stream_as_body(request.data.scan_as_stream())?;
 
         let identifier =
             build_table_identifier(&request.name, &request.namespace, &self.client.id_delimiter);
@@ -463,7 +444,7 @@ impl<S: HttpSend> Database for RemoteDatabase<S> {
             .client
             .post(&format!("/v1/table/{}/create/", identifier))
             .query(&[("mode", Into::<&str>::into(&request.mode))])
-            .body(data_buffer)
+            .body(body)
             .header(CONTENT_TYPE, ARROW_STREAM_CONTENT_TYPE);
 
         let (request_id, rsp) = self.client.send(req).await?;
@@ -813,7 +794,7 @@ mod tests {
     use std::collections::HashMap;
     use std::sync::{Arc, OnceLock};
 
-    use arrow_array::{Int32Array, RecordBatch, RecordBatchIterator};
+    use arrow_array::{Int32Array, RecordBatch};
     use arrow_schema::{DataType, Field, Schema};
 
     use crate::connection::ConnectBuilder;
@@ -993,8 +974,7 @@ mod tests {
             vec![Arc::new(Int32Array::from(vec![1, 2, 3]))],
         )
         .unwrap();
-        let reader = RecordBatchIterator::new([Ok(data.clone())], data.schema());
-        let table = conn.create_table("table1", reader).execute().await.unwrap();
+        let table = conn.create_table("table1", data).execute().await.unwrap();
         assert_eq!(table.name(), "table1");
     }
 
@@ -1011,8 +991,7 @@ mod tests {
             vec![Arc::new(Int32Array::from(vec![1, 2, 3]))],
         )
         .unwrap();
-        let reader = RecordBatchIterator::new([Ok(data.clone())], data.schema());
-        let result = conn.create_table("table1", reader).execute().await;
+        let result = conn.create_table("table1", data).execute().await;
         assert!(result.is_err());
         assert!(
             matches!(result, Err(crate::Error::TableAlreadyExists { name }) if name == "table1")
@@ -1045,8 +1024,7 @@ mod tests {
                 vec![Arc::new(Int32Array::from(vec![1, 2, 3]))],
             )
             .unwrap();
-            let reader = RecordBatchIterator::new([Ok(data.clone())], data.schema());
-            let mut builder = conn.create_table("table1", reader);
+            let mut builder = conn.create_table("table1", data.clone());
             if let Some(mode) = mode {
                 builder = builder.mode(mode);
             }
@@ -1071,9 +1049,8 @@ mod tests {
         .unwrap();
 
         let called: Arc<OnceLock<bool>> = Arc::new(OnceLock::new());
-        let reader = RecordBatchIterator::new([Ok(data.clone())], data.schema());
         let called_in_cb = called.clone();
-        conn.create_table("table1", reader)
+        conn.create_table("table1", data)
             .mode(CreateTableMode::ExistOk(Box::new(move |b| {
                 called_in_cb.clone().set(true).unwrap();
                 b
@@ -1262,9 +1239,8 @@ mod tests {
             vec![Arc::new(Int32Array::from(vec![1, 2, 3]))],
         )
         .unwrap();
-        let reader = RecordBatchIterator::new([Ok(data.clone())], data.schema());
         let table = conn
-            .create_table("table1", reader)
+            .create_table("table1", data)
             .namespace(vec!["ns1".to_string()])
             .execute()
             .await
@@ -1730,10 +1706,8 @@ mod tests {
                 vec![Arc::new(Int32Array::from(vec![1, 2, 3]))],
             )
             .unwrap();
-            let reader = RecordBatchIterator::new([Ok(data.clone())], schema.clone());
-
             let table = conn
-                .create_table("test_table", reader)
+                .create_table("test_table", data)
                 .namespace(namespace.clone())
                 .execute()
                 .await;
@@ -1806,9 +1780,7 @@ mod tests {
                 let data =
                     RecordBatch::try_new(schema.clone(), vec![Arc::new(Int32Array::from(vec![i]))])
                         .unwrap();
-                let reader = RecordBatchIterator::new([Ok(data.clone())], schema.clone());
-
-                conn.create_table(format!("table{}", i), reader)
+                conn.create_table(format!("table{}", i), data)
                     .namespace(namespace.clone())
                     .execute()
                     .await

rust/lancedb/src/remote/util.rs

@@ -1,29 +1,50 @@
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
-use std::io::Cursor;
-
-use arrow_array::RecordBatchReader;
+use arrow_ipc::CompressionType;
+use futures::{Stream, StreamExt};
 use reqwest::Response;
 
-use crate::Result;
+use crate::{arrow::SendableRecordBatchStream, Result};
 
 use super::db::ServerVersion;
 
-pub fn batches_to_ipc_bytes(batches: impl RecordBatchReader) -> Result<Vec<u8>> {
+pub fn stream_as_ipc(
+    data: SendableRecordBatchStream,
+) -> Result<impl Stream<Item = Result<bytes::Bytes>>> {
+    let options = arrow_ipc::writer::IpcWriteOptions::default()
+        .try_with_compression(Some(CompressionType::LZ4_FRAME))?;
     const WRITE_BUF_SIZE: usize = 4096;
     let buf = Vec::with_capacity(WRITE_BUF_SIZE);
-    let mut buf = Cursor::new(buf);
-    {
-        let mut writer = arrow_ipc::writer::StreamWriter::try_new(&mut buf, &batches.schema())?;
+    let writer =
+        arrow_ipc::writer::StreamWriter::try_new_with_options(buf, &data.schema(), options)?;
+    let stream = futures::stream::try_unfold(
+        (data, writer, false),
+        move |(mut data, mut writer, finished)| async move {
+            if finished {
+                return Ok(None);
+            }
+            match data.next().await {
+                Some(Ok(batch)) => {
+                    writer.write(&batch)?;
+                    let buffer = std::mem::take(writer.get_mut());
+                    Ok(Some((bytes::Bytes::from(buffer), (data, writer, false))))
+                }
+                Some(Err(e)) => Err(e),
+                None => {
+                    writer.finish()?;
+                    let buffer = std::mem::take(writer.get_mut());
+                    Ok(Some((bytes::Bytes::from(buffer), (data, writer, true))))
+                }
+            }
+        },
+    );
+    Ok(stream)
+}
 
-        for batch in batches {
-            let batch = batch?;
-            writer.write(&batch)?;
-        }
-        writer.finish()?;
-    }
-    Ok(buf.into_inner())
+pub fn stream_as_body(data: SendableRecordBatchStream) -> Result<reqwest::Body> {
+    let stream = stream_as_ipc(data)?;
+    Ok(reqwest::Body::wrap_stream(stream))
 }
 
 pub fn parse_server_version(req_id: &str, rsp: &Response) -> Result<ServerVersion> {

rust/lancedb/src/table.rs

@@ -5,7 +5,7 @@
 
 use arrow::array::{AsArray, FixedSizeListBuilder, Float32Builder};
 use arrow::datatypes::{Float32Type, UInt8Type};
-use arrow_array::{RecordBatchIterator, RecordBatchReader};
+use arrow_array::{RecordBatch, RecordBatchReader};
 use arrow_schema::{DataType, Field, Schema, SchemaRef};
 use async_trait::async_trait;
 use datafusion_expr::Expr;
@@ -16,8 +16,6 @@ use datafusion_physical_plan::union::UnionExec;
 use datafusion_physical_plan::ExecutionPlan;
 use futures::{FutureExt, StreamExt, TryFutureExt};
 use lance::dataset::builder::DatasetBuilder;
-use lance::dataset::cleanup::RemovalStats;
-use lance::dataset::optimize::{compact_files, CompactionMetrics, IndexRemapperOptions};
 use lance::dataset::scanner::Scanner;
 pub use lance::dataset::ColumnAlteration;
 pub use lance::dataset::NewColumnTransform;
@@ -46,17 +44,15 @@ use lance_namespace::models::{
 use lance_namespace::LanceNamespace;
 use lance_table::format::Manifest;
 use lance_table::io::commit::ManifestNamingScheme;
-use log::info;
 use serde::{Deserialize, Serialize};
 use std::collections::HashMap;
 use std::format;
 use std::path::Path;
 use std::sync::Arc;
 
-use crate::arrow::IntoArrow;
-use crate::connection::NoData;
+use crate::data::scannable::{scannable_with_embeddings, Scannable};
 use crate::database::Database;
-use crate::embeddings::{EmbeddingDefinition, EmbeddingRegistry, MaybeEmbedded, MemoryRegistry};
+use crate::embeddings::{EmbeddingDefinition, EmbeddingRegistry, MemoryRegistry};
 use crate::error::{Error, Result};
 use crate::index::vector::VectorIndex;
 use crate::index::IndexStatistics;
@@ -75,23 +71,27 @@ use crate::utils::{
 use self::dataset::DatasetConsistencyWrapper;
 use self::merge::MergeInsertBuilder;
 
+mod add_data;
 pub mod datafusion;
 pub(crate) mod dataset;
 pub mod delete;
 pub mod merge;
+pub mod optimize;
 pub mod schema_evolution;
 pub mod update;
 
+pub use add_data::{AddDataBuilder, AddDataMode, AddResult};
+
 use crate::index::waiter::wait_for_index;
 pub use chrono::Duration;
 pub use delete::DeleteResult;
 use futures::future::{join_all, Either};
-pub use lance::dataset::optimize::CompactionOptions;
 pub use lance::dataset::refs::{TagContents, Tags as LanceTags};
 pub use lance::dataset::scanner::DatasetRecordBatchStream;
 use lance::dataset::statistics::DatasetStatisticsExt;
 use lance_index::frag_reuse::FRAG_REUSE_INDEX_NAME;
 pub use lance_index::optimize::OptimizeOptions;
+pub use optimize::{CompactionOptions, OptimizeAction, OptimizeStats};
 pub use schema_evolution::{AddColumnsResult, AlterColumnsResult, DropColumnsResult};
 use serde_with::skip_serializing_none;
 pub use update::{UpdateBuilder, UpdateResult};
@@ -169,85 +169,6 @@ impl TableDefinition {
     }
 }
 
-/// Optimize the dataset.
-///
-/// Similar to `VACUUM` in PostgreSQL, it offers different options to
-/// optimize different parts of the table on disk.
-///
-/// By default, it optimizes everything, as [`OptimizeAction::All`].
-pub enum OptimizeAction {
-    /// Run all optimizations with default values
-    All,
-    /// Compacts files in the dataset
-    ///
-    /// LanceDb uses a readonly filesystem for performance and safe concurrency.  Every time
-    /// new data is added it will be added into new files.  Small files
-    /// can hurt both read and write performance.  Compaction will merge small files
-    /// into larger ones.
-    ///
-    /// All operations that modify data (add, delete, update, merge insert, etc.) will create
-    /// new files.  If these operations are run frequently then compaction should run frequently.
-    ///
-    /// If these operations are never run (search only) then compaction is not necessary.
-    Compact {
-        options: CompactionOptions,
-        remap_options: Option<Arc<dyn IndexRemapperOptions>>,
-    },
-    /// Prune old version of datasets
-    ///
-    /// Every change in LanceDb is additive.  When data is removed from a dataset a new version is
-    /// created that doesn't contain the removed data.  However, the old version, which does contain
-    /// the removed data, is left in place.  This is necessary for consistency and concurrency and
-    /// also enables time travel functionality like the ability to checkout an older version of the
-    /// dataset to undo changes.
-    ///
-    /// Over time, these old versions can consume a lot of disk space.  The prune operation will
-    /// remove versions of the dataset that are older than a certain age.  This will free up the
-    /// space used by that old data.
-    ///
-    /// Once a version is pruned it can no longer be checked out.
-    Prune {
-        /// The duration of time to keep versions of the dataset.
-        older_than: Option<Duration>,
-        /// 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`.
-        delete_unverified: Option<bool>,
-        /// If true, an error will be returned if there are any old versions that are still tagged.
-        error_if_tagged_old_versions: Option<bool>,
-    },
-    /// Optimize the indices
-    ///
-    /// This operation optimizes all indices in the table.  When new data is added to LanceDb
-    /// it is not added to the indices.  However, it can still turn up in searches because the search
-    /// function will scan both the indexed data and the unindexed data in parallel.  Over time, the
-    /// unindexed data can become large enough that the search performance is slow.  This operation
-    /// will add the unindexed data to the indices without rerunning the full index creation process.
-    ///
-    /// Optimizing an index is faster than re-training the index but it does not typically adjust the
-    /// underlying model relied upon by the index.  This can eventually lead to poor search accuracy
-    /// and so users may still want to occasionally retrain the index after adding a large amount of
-    /// data.
-    ///
-    /// For example, when using IVF, an index will create clusters.  Optimizing an index assigns unindexed
-    /// data to the existing clusters, but it does not move the clusters or create new clusters.
-    Index(OptimizeOptions),
-}
-
-impl Default for OptimizeAction {
-    fn default() -> Self {
-        Self::All
-    }
-}
-
-/// Statistics about the optimization.
-pub struct OptimizeStats {
-    /// Stats of the file compaction.
-    pub compaction: Option<CompactionMetrics>,
-
-    /// Stats of the version pruning
-    pub prune: Option<RemovalStats>,
-}
-
 /// Describes what happens when a vector either contains NaN or
 /// does not have enough values
 #[derive(Clone, Debug, Default)]
@@ -277,60 +198,6 @@ pub struct WriteOptions {
     pub lance_write_params: Option<WriteParams>,
 }
 
-#[derive(Debug, Clone, Default)]
-pub enum AddDataMode {
-    /// Rows will be appended to the table (the default)
-    #[default]
-    Append,
-    /// The existing table will be overwritten with the new data
-    Overwrite,
-}
-
-/// A builder for configuring a [`crate::connection::Connection::create_table`] or [`Table::add`]
-/// operation
-pub struct AddDataBuilder<T: IntoArrow> {
-    parent: Arc<dyn BaseTable>,
-    pub(crate) data: T,
-    pub(crate) mode: AddDataMode,
-    pub(crate) write_options: WriteOptions,
-    embedding_registry: Option<Arc<dyn EmbeddingRegistry>>,
-}
-
-impl<T: IntoArrow> std::fmt::Debug for AddDataBuilder<T> {
-    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        f.debug_struct("AddDataBuilder")
-            .field("parent", &self.parent)
-            .field("mode", &self.mode)
-            .field("write_options", &self.write_options)
-            .finish()
-    }
-}
-
-impl<T: IntoArrow> AddDataBuilder<T> {
-    pub fn mode(mut self, mode: AddDataMode) -> Self {
-        self.mode = mode;
-        self
-    }
-
-    pub fn write_options(mut self, options: WriteOptions) -> Self {
-        self.write_options = options;
-        self
-    }
-
-    pub async fn execute(self) -> Result<AddResult> {
-        let parent = self.parent.clone();
-        let data = self.data.into_arrow()?;
-        let without_data = AddDataBuilder::<NoData> {
-            data: NoData {},
-            mode: self.mode,
-            parent: self.parent,
-            write_options: self.write_options,
-            embedding_registry: self.embedding_registry,
-        };
-        parent.add(without_data, data).await
-    }
-}
-
 /// Filters that can be used to limit the rows returned by a query
 pub enum Filter {
     /// A SQL filter string
@@ -364,15 +231,6 @@ pub trait Tags: Send + Sync {
     async fn update(&mut self, tag: &str, version: u64) -> Result<()>;
 }
 
-#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, Default)]
-pub struct AddResult {
-    // The commit version associated with the operation.
-    // A version of `0` indicates compatibility with legacy servers that do not return
-    /// a commit version.
-    #[serde(default)]
-    pub version: u64,
-}
-
 #[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, Default)]
 pub struct MergeResult {
     // The commit version associated with the operation.
@@ -445,11 +303,7 @@ pub trait BaseTable: std::fmt::Display + std::fmt::Debug + Send + Sync {
     ) -> Result<String>;
 
     /// Add new records to the table.
-    async fn add(
-        &self,
-        add: AddDataBuilder<NoData>,
-        data: Box<dyn arrow_array::RecordBatchReader + Send>,
-    ) -> Result<AddResult>;
+    async fn add(&self, add: AddDataBuilder) -> Result<AddResult>;
     /// Delete rows from the table.
     async fn delete(&self, predicate: &str) -> Result<DeleteResult>;
     /// Update rows in the table.
@@ -594,6 +448,30 @@ mod test_utils {
                 embedding_registry: Arc::new(MemoryRegistry::new()),
             }
         }
+
+        pub fn new_with_handler_and_config<T>(
+            name: impl Into<String>,
+            handler: impl Fn(reqwest::Request) -> http::Response<T> + Clone + Send + Sync + 'static,
+            config: crate::remote::ClientConfig,
+        ) -> Self
+        where
+            T: Into<reqwest::Body>,
+        {
+            let inner = Arc::new(crate::remote::table::RemoteTable::new_mock_with_config(
+                name.into(),
+                handler.clone(),
+                config.clone(),
+            ));
+            let database = Arc::new(crate::remote::db::RemoteDatabase::new_mock_with_config(
+                handler, config,
+            ));
+            Self {
+                inner,
+                database: Some(database),
+                // Registry is unused.
+                embedding_registry: Arc::new(MemoryRegistry::new()),
+            }
+        }
     }
 }
 
@@ -694,16 +572,14 @@ impl Table {
     ///
     /// # Arguments
     ///
-    /// * `batches` data to be added to the Table
+    /// * `data` data to be added to the Table
     /// * `options` options to control how data is added
-    pub fn add<T: IntoArrow>(&self, batches: T) -> AddDataBuilder<T> {
-        AddDataBuilder {
-            parent: self.inner.clone(),
-            data: batches,
-            mode: AddDataMode::Append,
-            write_options: WriteOptions::default(),
-            embedding_registry: Some(self.embedding_registry.clone()),
-        }
+    pub fn add<T: Scannable + 'static>(&self, data: T) -> AddDataBuilder {
+        AddDataBuilder::new(
+            self.inner.clone(),
+            Box::new(data),
+            Some(self.embedding_registry.clone()),
+        )
     }
 
     /// Update existing records in the Table
@@ -742,31 +618,26 @@ impl Table {
     ///     .execute()
     ///     .await
     ///     .unwrap();
-    /// # let schema = Arc::new(Schema::new(vec![
-    /// #  Field::new("id", DataType::Int32, false),
-    /// #  Field::new("vector", DataType::FixedSizeList(
-    /// #    Arc::new(Field::new("item", DataType::Float32, true)), 128), true),
-    /// # ]));
-    /// let batches = RecordBatchIterator::new(
-    ///     vec![RecordBatch::try_new(
-    ///         schema.clone(),
-    ///         vec![
-    ///             Arc::new(Int32Array::from_iter_values(0..10)),
-    ///             Arc::new(
-    ///                 FixedSizeListArray::from_iter_primitive::<Float32Type, _, _>(
-    ///                     (0..10).map(|_| Some(vec![Some(1.0); 128])),
-    ///                     128,
-    ///                 ),
-    ///             ),
-    ///         ],
-    ///     )
-    ///     .unwrap()]
-    ///     .into_iter()
-    ///     .map(Ok),
+    /// let schema = Arc::new(Schema::new(vec![
+    ///     Field::new("id", DataType::Int32, false),
+    ///     Field::new("vector", DataType::FixedSizeList(
+    ///         Arc::new(Field::new("item", DataType::Float32, true)), 128), true),
+    /// ]));
+    /// let data = RecordBatch::try_new(
     ///     schema.clone(),
-    /// );
+    ///     vec![
+    ///         Arc::new(Int32Array::from_iter_values(0..10)),
+    ///         Arc::new(
+    ///             FixedSizeListArray::from_iter_primitive::<Float32Type, _, _>(
+    ///                 (0..10).map(|_| Some(vec![Some(1.0); 128])),
+    ///                 128,
+    ///             ),
+    ///         ),
+    ///     ],
+    /// )
+    /// .unwrap();
     /// let tbl = db
-    ///     .create_table("delete_test", Box::new(batches))
+    ///     .create_table("delete_test", data)
     ///     .execute()
     ///     .await
     ///     .unwrap();
@@ -1526,7 +1397,7 @@ impl NativeTable {
                     name: name.to_string(),
                     source: Box::new(e),
                 },
-                source => Error::Lance { source },
+                e => e.into(),
             })?;
 
         let dataset = DatasetConsistencyWrapper::new_latest(dataset, read_consistency_interval);
@@ -1610,7 +1481,7 @@ impl NativeTable {
                 lance::Error::Namespace { source, .. } => Error::Runtime {
                     message: format!("Failed to get table info from namespace: {:?}", source),
                 },
-                source => Error::Lance { source },
+                e => e.into(),
             })?;
 
         let dataset = builder
@@ -1622,7 +1493,7 @@ impl NativeTable {
                     name: name.to_string(),
                     source: Box::new(e),
                 },
-                source => Error::Lance { source },
+                e => e.into(),
             })?;
 
         let uri = dataset.uri().to_string();
@@ -1716,7 +1587,7 @@ impl NativeTable {
                 lance::Error::DatasetAlreadyExists { .. } => Error::TableAlreadyExists {
                     name: name.to_string(),
                 },
-                source => Error::Lance { source },
+                e => e.into(),
             })?;
 
         let id = Self::build_id(&namespace, name);
@@ -1743,12 +1614,12 @@ impl NativeTable {
         read_consistency_interval: Option<std::time::Duration>,
         namespace_client: Option<Arc<dyn LanceNamespace>>,
     ) -> Result<Self> {
-        let batches = RecordBatchIterator::new(vec![], schema);
+        let data: Box<dyn Scannable> = Box::new(RecordBatch::new_empty(schema));
         Self::create(
             uri,
             name,
             namespace,
-            batches,
+            data,
             write_store_wrapper,
             params,
             read_consistency_interval,
@@ -1837,7 +1708,7 @@ impl NativeTable {
                 lance::Error::DatasetAlreadyExists { .. } => Error::TableAlreadyExists {
                     name: name.to_string(),
                 },
-                source => Error::Lance { source },
+                e => e.into(),
             })?;
 
         let id = Self::build_id(&namespace, name);
@@ -1859,16 +1730,6 @@ impl NativeTable {
         })
     }
 
-    async fn optimize_indices(&self, options: &OptimizeOptions) -> Result<()> {
-        info!("LanceDB: optimizing indices: {:?}", options);
-        self.dataset
-            .get_mut()
-            .await?
-            .optimize_indices(options)
-            .await?;
-        Ok(())
-    }
-
     /// Merge new data into this table.
     pub async fn merge(
         &mut self,
@@ -1884,47 +1745,6 @@ impl NativeTable {
         Ok(())
     }
 
-    /// Remove old versions of the dataset from disk.
-    ///
-    /// # Arguments
-    /// * `older_than` - The duration of time to keep versions of the dataset.
-    /// * `delete_unverified` - 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`.
-    ///
-    /// This calls into [lance::dataset::Dataset::cleanup_old_versions] and
-    /// returns the result.
-    async fn cleanup_old_versions(
-        &self,
-        older_than: Duration,
-        delete_unverified: Option<bool>,
-        error_if_tagged_old_versions: Option<bool>,
-    ) -> Result<RemovalStats> {
-        Ok(self
-            .dataset
-            .get_mut()
-            .await?
-            .cleanup_old_versions(older_than, delete_unverified, error_if_tagged_old_versions)
-            .await?)
-    }
-
-    /// Compact files in the dataset.
-    ///
-    /// This can be run after making several small appends to optimize the table
-    /// for faster reads.
-    ///
-    /// This calls into [lance::dataset::optimize::compact_files].
-    async fn compact_files(
-        &self,
-        options: CompactionOptions,
-        remap_options: Option<Arc<dyn IndexRemapperOptions>>,
-    ) -> Result<CompactionMetrics> {
-        let mut dataset_mut = self.dataset.get_mut().await?;
-        let metrics = compact_files(&mut dataset_mut, options, remap_options).await?;
-        Ok(metrics)
-    }
-
     // TODO: why are these individual methods and not some single "get_stats" method?
     pub async fn count_fragments(&self) -> Result<usize> {
         Ok(self.dataset.get().await?.count_fragments())
@@ -2670,17 +2490,7 @@ impl BaseTable for NativeTable {
         }
     }
 
-    async fn add(
-        &self,
-        add: AddDataBuilder<NoData>,
-        data: Box<dyn RecordBatchReader + Send>,
-    ) -> Result<AddResult> {
-        let data = Box::new(MaybeEmbedded::try_new(
-            data,
-            self.table_definition().await?,
-            add.embedding_registry,
-        )?) as Box<dyn RecordBatchReader + Send>;
-
+    async fn add(&self, add: AddDataBuilder) -> Result<AddResult> {
         let lance_params = add.write_options.lance_write_params.unwrap_or(WriteParams {
             mode: match add.mode {
                 AddDataMode::Append => WriteMode::Append,
@@ -2689,6 +2499,11 @@ impl BaseTable for NativeTable {
             ..Default::default()
         });
 
+        // Apply embeddings if configured
+        let table_def = self.table_definition().await?;
+        let data =
+            scannable_with_embeddings(add.data, &table_def, add.embedding_registry.as_ref())?;
+
         let dataset = {
             // Limited scope for the mutable borrow of self.dataset avoids deadlock.
             let ds = self.dataset.get_mut().await?;
@@ -3010,55 +2825,8 @@ impl BaseTable for NativeTable {
     }
 
     async fn optimize(&self, action: OptimizeAction) -> Result<OptimizeStats> {
-        let mut stats = OptimizeStats {
-            compaction: None,
-            prune: None,
-        };
-        match action {
-            OptimizeAction::All => {
-                stats.compaction = self
-                    .optimize(OptimizeAction::Compact {
-                        options: CompactionOptions::default(),
-                        remap_options: None,
-                    })
-                    .await?
-                    .compaction;
-                stats.prune = self
-                    .optimize(OptimizeAction::Prune {
-                        older_than: None,
-                        delete_unverified: None,
-                        error_if_tagged_old_versions: None,
-                    })
-                    .await?
-                    .prune;
-                self.optimize(OptimizeAction::Index(OptimizeOptions::default()))
-                    .await?;
-            }
-            OptimizeAction::Compact {
-                options,
-                remap_options,
-            } => {
-                stats.compaction = Some(self.compact_files(options, remap_options).await?);
-            }
-            OptimizeAction::Prune {
-                older_than,
-                delete_unverified,
-                error_if_tagged_old_versions,
-            } => {
-                stats.prune = Some(
-                    self.cleanup_old_versions(
-                        older_than.unwrap_or(Duration::try_days(7).expect("valid delta")),
-                        delete_unverified,
-                        error_if_tagged_old_versions,
-                    )
-                    .await?,
-                );
-            }
-            OptimizeAction::Index(options) => {
-                self.optimize_indices(&options).await?;
-            }
-        }
-        Ok(stats)
+        // Delegate to the submodule implementation
+        optimize::execute_optimize(self, action).await
     }
 
     async fn add_columns(
@@ -3342,7 +3110,7 @@ mod tests {
     use arrow_array::{BinaryArray, LargeBinaryArray};
     use arrow_data::ArrayDataBuilder;
     use arrow_schema::{DataType, Field, Schema};
-    use lance::dataset::WriteMode;
+    use futures::TryStreamExt;
     use lance::io::{ObjectStoreParams, WrappingObjectStore};
     use lance::Dataset;
     use rand::Rng;
@@ -3353,14 +3121,17 @@ mod tests {
     use crate::connection::ConnectBuilder;
     use crate::index::scalar::{BTreeIndexBuilder, BitmapIndexBuilder};
     use crate::index::vector::{IvfHnswPqIndexBuilder, IvfHnswSqIndexBuilder};
+    use crate::query::{ExecutableQuery, QueryBase};
+    use crate::test_utils::connection::new_test_connection;
 
     #[tokio::test]
     async fn test_open() {
         let tmp_dir = tempdir().unwrap();
         let dataset_path = tmp_dir.path().join("test.lance");
 
-        let batches = make_test_batches();
-        Dataset::write(batches, dataset_path.to_str().unwrap(), None)
+        let batch = make_test_batches();
+        let reader = RecordBatchIterator::new(vec![Ok(batch.clone())], batch.schema());
+        Dataset::write(reader, dataset_path.to_str().unwrap(), None)
             .await
             .unwrap();
 
@@ -3393,9 +3164,12 @@ mod tests {
         let tmp_dir = tempdir().unwrap();
         let uri = tmp_dir.path().to_str().unwrap();
 
-        let batches = make_test_batches();
-        let batches = Box::new(batches) as Box<dyn RecordBatchReader + Send>;
-        let table = NativeTable::create(uri, "test", vec![], batches, None, None, None, None)
+        let batch = make_test_batches();
+        let reader: Box<dyn RecordBatchReader + Send> = Box::new(RecordBatchIterator::new(
+            vec![Ok(batch.clone())],
+            batch.schema(),
+        ));
+        let table = NativeTable::create(uri, "test", vec![], reader, None, None, None, None)
             .await
             .unwrap();
 
@@ -3409,33 +3183,6 @@ mod tests {
         );
     }
 
-    #[tokio::test]
-    async fn test_add() {
-        let tmp_dir = tempdir().unwrap();
-        let uri = tmp_dir.path().to_str().unwrap();
-        let conn = connect(uri).execute().await.unwrap();
-
-        let batches = make_test_batches();
-        let schema = batches.schema().clone();
-        let table = conn.create_table("test", batches).execute().await.unwrap();
-        assert_eq!(table.count_rows(None).await.unwrap(), 10);
-
-        let new_batches = RecordBatchIterator::new(
-            vec![RecordBatch::try_new(
-                schema.clone(),
-                vec![Arc::new(Int32Array::from_iter_values(100..110))],
-            )
-            .unwrap()]
-            .into_iter()
-            .map(Ok),
-            schema.clone(),
-        );
-
-        table.add(new_batches).execute().await.unwrap();
-        assert_eq!(table.count_rows(None).await.unwrap(), 20);
-        assert_eq!(table.name(), "test");
-    }
-
     #[tokio::test]
     async fn test_merge_insert() {
         let tmp_dir = tempdir().unwrap();
@@ -3452,7 +3199,7 @@ mod tests {
         assert_eq!(table.count_rows(None).await.unwrap(), 10);
 
         // Create new data with i=5..15
-        let new_batches = Box::new(merge_insert_test_batches(5, 1));
+        let new_batches = merge_insert_test_batches(5, 1);
 
         // Perform a "insert if not exists"
         let mut merge_insert_builder = table.merge_insert(&["i"]);
@@ -3466,7 +3213,7 @@ mod tests {
         assert_eq!(result.num_attempts, 1);
 
         // Create new data with i=15..25 (no id matches)
-        let new_batches = Box::new(merge_insert_test_batches(15, 2));
+        let new_batches = merge_insert_test_batches(15, 2);
         // Perform a "bulk update" (should not affect anything)
         let mut merge_insert_builder = table.merge_insert(&["i"]);
         merge_insert_builder.when_matched_update_all(None);
@@ -3479,7 +3226,7 @@ mod tests {
         );
 
         // Conditional update that only replaces the age=0 data
-        let new_batches = Box::new(merge_insert_test_batches(5, 3));
+        let new_batches = merge_insert_test_batches(5, 3);
         let mut merge_insert_builder = table.merge_insert(&["i"]);
         merge_insert_builder.when_matched_update_all(Some("target.age = 0".to_string()));
         merge_insert_builder.execute(new_batches).await.unwrap();
@@ -3505,7 +3252,7 @@ mod tests {
         assert_eq!(table.count_rows(None).await.unwrap(), 10);
 
         // Test use_index=true (default behavior)
-        let new_batches = Box::new(merge_insert_test_batches(5, 1));
+        let new_batches = merge_insert_test_batches(5, 1);
         let mut merge_insert_builder = table.merge_insert(&["i"]);
         merge_insert_builder.when_not_matched_insert_all();
         merge_insert_builder.use_index(true);
@@ -3513,7 +3260,7 @@ mod tests {
         assert_eq!(table.count_rows(None).await.unwrap(), 15);
 
         // Test use_index=false (force table scan)
-        let new_batches = Box::new(merge_insert_test_batches(15, 2));
+        let new_batches = merge_insert_test_batches(15, 2);
         let mut merge_insert_builder = table.merge_insert(&["i"]);
         merge_insert_builder.when_not_matched_insert_all();
         merge_insert_builder.use_index(false);
@@ -3521,59 +3268,6 @@ mod tests {
         assert_eq!(table.count_rows(None).await.unwrap(), 25);
     }
 
-    #[tokio::test]
-    async fn test_add_overwrite() {
-        let tmp_dir = tempdir().unwrap();
-        let uri = tmp_dir.path().to_str().unwrap();
-        let conn = connect(uri).execute().await.unwrap();
-
-        let batches = make_test_batches();
-        let schema = batches.schema().clone();
-        let table = conn.create_table("test", batches).execute().await.unwrap();
-        assert_eq!(table.count_rows(None).await.unwrap(), 10);
-
-        let batches = vec![RecordBatch::try_new(
-            schema.clone(),
-            vec![Arc::new(Int32Array::from_iter_values(100..110))],
-        )
-        .unwrap()]
-        .into_iter()
-        .map(Ok);
-
-        let new_batches = RecordBatchIterator::new(batches.clone(), schema.clone());
-
-        // Can overwrite using AddDataOptions::mode
-        table
-            .add(new_batches)
-            .mode(AddDataMode::Overwrite)
-            .execute()
-            .await
-            .unwrap();
-        assert_eq!(table.count_rows(None).await.unwrap(), 10);
-        assert_eq!(table.name(), "test");
-
-        // Can overwrite using underlying WriteParams (which
-        // take precedence over AddDataOptions::mode)
-
-        let param: WriteParams = WriteParams {
-            mode: WriteMode::Overwrite,
-            ..Default::default()
-        };
-
-        let new_batches = RecordBatchIterator::new(batches.clone(), schema.clone());
-        table
-            .add(new_batches)
-            .write_options(WriteOptions {
-                lance_write_params: Some(param),
-            })
-            .mode(AddDataMode::Append)
-            .execute()
-            .await
-            .unwrap();
-        assert_eq!(table.count_rows(None).await.unwrap(), 10);
-        assert_eq!(table.name(), "test");
-    }
-
     #[derive(Default, Debug)]
     struct NoOpCacheWrapper {
         called: AtomicBool,
@@ -3629,35 +3323,25 @@ mod tests {
         assert!(wrapper.called());
     }
 
-    fn merge_insert_test_batches(
-        offset: i32,
-        age: i32,
-    ) -> impl RecordBatchReader + Send + Sync + 'static {
+    fn merge_insert_test_batches(offset: i32, age: i32) -> Box<dyn RecordBatchReader + Send> {
         let schema = Arc::new(Schema::new(vec![
             Field::new("i", DataType::Int32, false),
             Field::new("age", DataType::Int32, false),
         ]));
-        RecordBatchIterator::new(
-            vec![RecordBatch::try_new(
-                schema.clone(),
-                vec![
-                    Arc::new(Int32Array::from_iter_values(offset..(offset + 10))),
-                    Arc::new(Int32Array::from_iter_values(std::iter::repeat_n(age, 10))),
-                ],
-            )],
-            schema,
+        let batch = RecordBatch::try_new(
+            schema.clone(),
+            vec![
+                Arc::new(Int32Array::from_iter_values(offset..(offset + 10))),
+                Arc::new(Int32Array::from_iter_values(std::iter::repeat_n(age, 10))),
+            ],
         )
+        .unwrap();
+        Box::new(RecordBatchIterator::new(vec![Ok(batch)], schema))
     }
 
-    fn make_test_batches() -> impl RecordBatchReader + Send + Sync + 'static {
+    fn make_test_batches() -> RecordBatch {
         let schema = Arc::new(Schema::new(vec![Field::new("i", DataType::Int32, false)]));
-        RecordBatchIterator::new(
-            vec![RecordBatch::try_new(
-                schema.clone(),
-                vec![Arc::new(Int32Array::from_iter_values(0..10))],
-            )],
-            schema,
-        )
+        RecordBatch::try_new(schema, vec![Arc::new(Int32Array::from_iter_values(0..10))]).unwrap()
     }
 
     #[tokio::test]
@@ -3745,14 +3429,9 @@ mod tests {
         );
 
         let vectors = Arc::new(create_fixed_size_list(float_arr, dimension).unwrap());
-        let batches = RecordBatchIterator::new(
-            vec![RecordBatch::try_new(schema.clone(), vec![vectors.clone()]).unwrap()]
-                .into_iter()
-                .map(Ok),
-            schema,
-        );
+        let batch = RecordBatch::try_new(schema.clone(), vec![vectors.clone()]).unwrap();
 
-        let table = conn.create_table("test", batches).execute().await.unwrap();
+        let table = conn.create_table("test", batch).execute().await.unwrap();
 
         assert_eq!(table.index_stats("my_index").await.unwrap(), None);
 
@@ -3783,6 +3462,31 @@ mod tests {
         assert_eq!(table.list_indices().await.unwrap().len(), 0);
     }
 
+    #[tokio::test]
+    async fn test_dynamic_select() {
+        let tc = new_test_connection().await.unwrap();
+        let db = tc.connection;
+
+        let table = db
+            .create_table("test", some_sample_data())
+            .execute()
+            .await
+            .unwrap();
+
+        let query = table.query().select(Select::dynamic(&[("i_alias", "i")]));
+
+        let result = query.execute().await;
+        let batches = result
+            .expect("should have result")
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+
+        for batch in batches {
+            assert!(batch.column_by_name("i_alias").is_some());
+        }
+    }
+
     #[tokio::test]
     async fn test_ivf_pq_uses_default_partition_size_for_num_partitions() {
         use arrow_array::{Float32Array, RecordBatch};
@@ -3809,14 +3513,9 @@ mod tests {
         let float_arr =
             Float32Array::from_iter_values((0..(num_rows * dimension)).map(|v| v as f32));
         let vectors = Arc::new(create_fixed_size_list(float_arr, dimension as i32).unwrap());
-        let batches = RecordBatchIterator::new(
-            vec![RecordBatch::try_new(schema.clone(), vec![vectors]).unwrap()]
-                .into_iter()
-                .map(Ok),
-            schema,
-        );
+        let batch = RecordBatch::try_new(schema.clone(), vec![vectors]).unwrap();
 
-        let table = conn.create_table("test", batches).execute().await.unwrap();
+        let table = conn.create_table("test", batch).execute().await.unwrap();
         let native_table = table.as_native().unwrap();
         let builder = IvfPqIndexBuilder::default();
         table
@@ -3886,14 +3585,9 @@ mod tests {
         );
 
         let vectors = Arc::new(create_fixed_size_list(float_arr, dimension).unwrap());
-        let batches = RecordBatchIterator::new(
-            vec![RecordBatch::try_new(schema.clone(), vec![vectors.clone()]).unwrap()]
-                .into_iter()
-                .map(Ok),
-            schema,
-        );
+        let batch = RecordBatch::try_new(schema.clone(), vec![vectors.clone()]).unwrap();
 
-        let table = conn.create_table("test", batches).execute().await.unwrap();
+        let table = conn.create_table("test", batch).execute().await.unwrap();
 
         let stats = table.index_stats("my_index").await.unwrap();
         assert!(stats.is_none());
@@ -3951,14 +3645,9 @@ mod tests {
         );
 
         let vectors = Arc::new(create_fixed_size_list(float_arr, dimension).unwrap());
-        let batches = RecordBatchIterator::new(
-            vec![RecordBatch::try_new(schema.clone(), vec![vectors.clone()]).unwrap()]
-                .into_iter()
-                .map(Ok),
-            schema,
-        );
+        let batch = RecordBatch::try_new(schema.clone(), vec![vectors.clone()]).unwrap();
 
-        let table = conn.create_table("test", batches).execute().await.unwrap();
+        let table = conn.create_table("test", batch).execute().await.unwrap();
         let stats = table.index_stats("my_index").await.unwrap();
         assert!(stats.is_none());
 
@@ -4001,7 +3690,7 @@ mod tests {
         Ok(FixedSizeListArray::from(data))
     }
 
-    fn some_sample_data() -> Box<dyn RecordBatchReader + Send> {
+    fn some_sample_data() -> Box<dyn arrow_array::RecordBatchReader + Send> {
         let batch = RecordBatch::try_new(
             Arc::new(Schema::new(vec![Field::new("i", DataType::Int32, false)])),
             vec![Arc::new(Int32Array::from(vec![1]))],
@@ -4025,10 +3714,7 @@ mod tests {
         .unwrap();
         let conn = ConnectBuilder::new(uri).execute().await.unwrap();
         let table = conn
-            .create_table(
-                "my_table",
-                RecordBatchIterator::new(vec![Ok(batch.clone())], batch.schema()),
-            )
+            .create_table("my_table", batch.clone())
             .execute()
             .await
             .unwrap();
@@ -4107,10 +3793,7 @@ mod tests {
         .unwrap();
 
         let table = conn
-            .create_table(
-                "test_bitmap",
-                RecordBatchIterator::new(vec![Ok(batch.clone())], batch.schema()),
-            )
+            .create_table("test_bitmap", batch.clone())
             .execute()
             .await
             .unwrap();
@@ -4211,10 +3894,7 @@ mod tests {
         .unwrap();
 
         let table = conn
-            .create_table(
-                "test_bitmap",
-                RecordBatchIterator::new(vec![Ok(batch.clone())], batch.schema()),
-            )
+            .create_table("test_bitmap", batch.clone())
             .execute()
             .await
             .unwrap();
@@ -4274,10 +3954,7 @@ mod tests {
         .unwrap();
 
         let table = conn
-            .create_table(
-                "test_bitmap",
-                RecordBatchIterator::new(vec![Ok(batch.clone())], batch.schema()),
-            )
+            .create_table("test_bitmap", batch.clone())
             .execute()
             .await
             .unwrap();
@@ -4322,7 +3999,7 @@ mod tests {
 
             let conn1 = ConnectBuilder::new(uri).execute().await.unwrap();
             let table1 = conn1
-                .create_empty_table("my_table", data.schema())
+                .create_empty_table("my_table", RecordBatchReader::schema(&data))
                 .execute()
                 .await
                 .unwrap();
@@ -4592,10 +4269,7 @@ mod tests {
         .unwrap();
 
         let table = conn
-            .create_table(
-                "test_stats",
-                RecordBatchIterator::new(vec![Ok(batch.clone())], batch.schema()),
-            )
+            .create_table("test_stats", batch.clone())
             .execute()
             .await
             .unwrap();
@@ -4608,21 +4282,11 @@ mod tests {
                 ],
             )
             .unwrap();
-            table
-                .add(RecordBatchIterator::new(
-                    vec![Ok(batch.clone())],
-                    batch.schema(),
-                ))
-                .execute()
-                .await
-                .unwrap();
+            table.add(batch.clone()).execute().await.unwrap();
         }
 
         let empty_table = conn
-            .create_table(
-                "test_stats_empty",
-                RecordBatchIterator::new(vec![], batch.schema()),
-            )
+            .create_table("test_stats_empty", RecordBatch::new_empty(batch.schema()))
             .execute()
             .await
             .unwrap();
@@ -4696,22 +4360,12 @@ mod tests {
         .unwrap();
 
         let table = conn
-            .create_table(
-                "test_list_indices_skip_frag_reuse",
-                RecordBatchIterator::new(vec![Ok(batch.clone())], batch.schema()),
-            )
+            .create_table("test_list_indices_skip_frag_reuse", batch.clone())
             .execute()
             .await
             .unwrap();
 
-        table
-            .add(RecordBatchIterator::new(
-                vec![Ok(batch.clone())],
-                batch.schema(),
-            ))
-            .execute()
-            .await
-            .unwrap();
+        table.add(batch.clone()).execute().await.unwrap();
 
         table
             .create_index(&["id"], Index::Bitmap(BitmapIndexBuilder {}))
@@ -4741,8 +4395,9 @@ mod tests {
         let tmp_dir = tempdir().unwrap();
         let dataset_path = tmp_dir.path().join("test_ns_query.lance");
 
-        let batches = make_test_batches();
-        Dataset::write(batches, dataset_path.to_str().unwrap(), None)
+        let batch = make_test_batches();
+        let reader = RecordBatchIterator::new(vec![Ok(batch.clone())], batch.schema());
+        Dataset::write(reader, dataset_path.to_str().unwrap(), None)
             .await
             .unwrap();
 
@@ -4794,8 +4449,9 @@ mod tests {
         let tmp_dir = tempdir().unwrap();
         let dataset_path = tmp_dir.path().join("test_ns_plain.lance");
 
-        let batches = make_test_batches();
-        Dataset::write(batches, dataset_path.to_str().unwrap(), None)
+        let batch = make_test_batches();
+        let reader = RecordBatchIterator::new(vec![Ok(batch.clone())], batch.schema());
+        Dataset::write(reader, dataset_path.to_str().unwrap(), None)
             .await
             .unwrap();
 

rust/lancedb/src/table/add_data.rs

@@ -0,0 +1,343 @@
+// SPDX-License-Identifier: Apache-2.0
+// SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+use std::sync::Arc;
+
+use serde::{Deserialize, Serialize};
+
+use crate::data::scannable::Scannable;
+use crate::embeddings::EmbeddingRegistry;
+use crate::Result;
+
+use super::{BaseTable, WriteOptions};
+
+#[derive(Debug, Clone, Default)]
+pub enum AddDataMode {
+    /// Rows will be appended to the table (the default)
+    #[default]
+    Append,
+    /// The existing table will be overwritten with the new data
+    Overwrite,
+}
+
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, Default)]
+pub struct AddResult {
+    // The commit version associated with the operation.
+    // A version of `0` indicates compatibility with legacy servers that do not return
+    /// a commit version.
+    #[serde(default)]
+    pub version: u64,
+}
+
+/// A builder for configuring a [`crate::table::Table::add`] operation
+pub struct AddDataBuilder {
+    pub(crate) parent: Arc<dyn BaseTable>,
+    pub(crate) data: Box<dyn Scannable>,
+    pub(crate) mode: AddDataMode,
+    pub(crate) write_options: WriteOptions,
+    pub(crate) embedding_registry: Option<Arc<dyn EmbeddingRegistry>>,
+}
+
+impl std::fmt::Debug for AddDataBuilder {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.debug_struct("AddDataBuilder")
+            .field("parent", &self.parent)
+            .field("mode", &self.mode)
+            .field("write_options", &self.write_options)
+            .finish()
+    }
+}
+
+impl AddDataBuilder {
+    pub(crate) fn new(
+        parent: Arc<dyn BaseTable>,
+        data: Box<dyn Scannable>,
+        embedding_registry: Option<Arc<dyn EmbeddingRegistry>>,
+    ) -> Self {
+        Self {
+            parent,
+            data,
+            mode: AddDataMode::Append,
+            write_options: WriteOptions::default(),
+            embedding_registry,
+        }
+    }
+
+    pub fn mode(mut self, mode: AddDataMode) -> Self {
+        self.mode = mode;
+        self
+    }
+
+    pub fn write_options(mut self, options: WriteOptions) -> Self {
+        self.write_options = options;
+        self
+    }
+
+    pub async fn execute(self) -> Result<AddResult> {
+        self.parent.clone().add(self).await
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use std::sync::Arc;
+
+    use arrow_array::{record_batch, RecordBatch, RecordBatchIterator};
+    use arrow_schema::{ArrowError, DataType, Field, Schema};
+    use futures::TryStreamExt;
+    use lance::dataset::{WriteMode, WriteParams};
+
+    use crate::arrow::{SendableRecordBatchStream, SimpleRecordBatchStream};
+    use crate::connect;
+    use crate::data::scannable::Scannable;
+    use crate::embeddings::{
+        EmbeddingDefinition, EmbeddingFunction, EmbeddingRegistry, MemoryRegistry,
+    };
+    use crate::query::{ExecutableQuery, QueryBase, Select};
+    use crate::table::{ColumnDefinition, ColumnKind, Table, TableDefinition, WriteOptions};
+    use crate::test_utils::embeddings::MockEmbed;
+    use crate::Error;
+
+    use super::AddDataMode;
+
+    async fn create_test_table() -> Table {
+        let conn = connect("memory://").execute().await.unwrap();
+        let batch = record_batch!(("id", Int64, [1, 2, 3])).unwrap();
+        conn.create_table("test", batch).execute().await.unwrap()
+    }
+
+    async fn test_add_with_data<T>(data: T)
+    where
+        T: Scannable + 'static,
+    {
+        let table = create_test_table().await;
+        let schema = data.schema();
+        table.add(data).execute().await.unwrap();
+        assert_eq!(table.count_rows(None).await.unwrap(), 5); // 3 initial + 2 added
+        assert_eq!(table.schema().await.unwrap(), schema);
+    }
+
+    #[tokio::test]
+    async fn test_add_with_batch() {
+        let batch = record_batch!(("id", Int64, [4, 5])).unwrap();
+        test_add_with_data(batch).await;
+    }
+
+    #[tokio::test]
+    async fn test_add_with_vec_batch() {
+        let data = vec![
+            record_batch!(("id", Int64, [4])).unwrap(),
+            record_batch!(("id", Int64, [5])).unwrap(),
+        ];
+        test_add_with_data(data).await;
+    }
+
+    #[tokio::test]
+    async fn test_add_with_record_batch_reader() {
+        let data = vec![
+            record_batch!(("id", Int64, [4])).unwrap(),
+            record_batch!(("id", Int64, [5])).unwrap(),
+        ];
+        let schema = data[0].schema();
+        let reader: Box<dyn arrow_array::RecordBatchReader + Send> = Box::new(
+            RecordBatchIterator::new(data.into_iter().map(Ok), schema.clone()),
+        );
+        test_add_with_data(reader).await;
+    }
+
+    #[tokio::test]
+    async fn test_add_with_stream() {
+        let data = vec![
+            record_batch!(("id", Int64, [4])).unwrap(),
+            record_batch!(("id", Int64, [5])).unwrap(),
+        ];
+        let schema = data[0].schema();
+        let inner = futures::stream::iter(data.into_iter().map(Ok));
+        let stream: SendableRecordBatchStream = Box::pin(SimpleRecordBatchStream {
+            schema,
+            stream: inner,
+        });
+        test_add_with_data(stream).await;
+    }
+
+    #[derive(Debug)]
+    struct MyError;
+
+    impl std::fmt::Display for MyError {
+        fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+            write!(f, "MyError occurred")
+        }
+    }
+
+    impl std::error::Error for MyError {}
+
+    #[tokio::test]
+    async fn test_add_preserves_reader_error() {
+        let table = create_test_table().await;
+        let first_batch = record_batch!(("id", Int64, [4])).unwrap();
+        let schema = first_batch.schema();
+        let iterator = vec![
+            Ok(first_batch),
+            Err(ArrowError::ExternalError(Box::new(MyError))),
+        ];
+        let reader: Box<dyn arrow_array::RecordBatchReader + Send> = Box::new(
+            RecordBatchIterator::new(iterator.into_iter(), schema.clone()),
+        );
+
+        let result = table.add(reader).execute().await;
+
+        assert!(result.is_err());
+    }
+
+    #[tokio::test]
+    async fn test_add_preserves_stream_error() {
+        let table = create_test_table().await;
+        let first_batch = record_batch!(("id", Int64, [4])).unwrap();
+        let schema = first_batch.schema();
+        let iterator = vec![
+            Ok(first_batch),
+            Err(Error::External {
+                source: Box::new(MyError),
+            }),
+        ];
+        let stream = futures::stream::iter(iterator);
+        let stream: SendableRecordBatchStream = Box::pin(SimpleRecordBatchStream {
+            schema: schema.clone(),
+            stream,
+        });
+
+        let result = table.add(stream).execute().await;
+
+        assert!(result.is_err());
+    }
+
+    #[tokio::test]
+    async fn test_add() {
+        let conn = connect("memory://").execute().await.unwrap();
+
+        let batch = record_batch!(("i", Int32, [0, 1, 2])).unwrap();
+        let table = conn
+            .create_table("test", batch.clone())
+            .execute()
+            .await
+            .unwrap();
+        assert_eq!(table.count_rows(None).await.unwrap(), 3);
+
+        let new_batch = record_batch!(("i", Int32, [3])).unwrap();
+        table.add(new_batch).execute().await.unwrap();
+
+        assert_eq!(table.count_rows(None).await.unwrap(), 4);
+        assert_eq!(table.schema().await.unwrap(), batch.schema());
+    }
+
+    #[tokio::test]
+    async fn test_add_overwrite() {
+        let conn = connect("memory://").execute().await.unwrap();
+
+        let batch = record_batch!(("i", Int32, [0, 1, 2])).unwrap();
+        let table = conn
+            .create_table("test", batch.clone())
+            .execute()
+            .await
+            .unwrap();
+        assert_eq!(table.count_rows(None).await.unwrap(), batch.num_rows());
+
+        let new_batch = record_batch!(("x", Float32, [0.0, 1.0])).unwrap();
+        let res = table
+            .add(new_batch.clone())
+            .mode(AddDataMode::Overwrite)
+            .execute()
+            .await
+            .unwrap();
+        assert_eq!(res.version, table.version().await.unwrap());
+        assert_eq!(table.count_rows(None).await.unwrap(), new_batch.num_rows());
+        assert_eq!(table.schema().await.unwrap(), new_batch.schema());
+
+        // Can overwrite using underlying WriteParams (which
+        // take precedence over AddDataMode)
+        let param: WriteParams = WriteParams {
+            mode: WriteMode::Overwrite,
+            ..Default::default()
+        };
+
+        table
+            .add(new_batch.clone())
+            .write_options(WriteOptions {
+                lance_write_params: Some(param),
+            })
+            .mode(AddDataMode::Append)
+            .execute()
+            .await
+            .unwrap();
+        assert_eq!(table.count_rows(None).await.unwrap(), new_batch.num_rows());
+    }
+
+    #[tokio::test]
+    async fn test_add_with_embeddings() {
+        let registry = Arc::new(MemoryRegistry::new());
+        let mock_embedding: Arc<dyn EmbeddingFunction> = Arc::new(MockEmbed::new("mock", 4));
+        registry.register("mock", mock_embedding).unwrap();
+
+        let conn = connect("memory://")
+            .embedding_registry(registry)
+            .execute()
+            .await
+            .unwrap();
+
+        let schema = Arc::new(Schema::new(vec![
+            Field::new("text", DataType::Utf8, false),
+            Field::new(
+                "text_embedding",
+                DataType::FixedSizeList(Arc::new(Field::new("item", DataType::Float32, true)), 4),
+                false,
+            ),
+        ]));
+
+        // Add embedding metadata to the schema
+        let embedding_def = EmbeddingDefinition::new("text", "mock", Some("text_embedding"));
+        let table_def = TableDefinition::new(
+            schema.clone(),
+            vec![
+                ColumnDefinition {
+                    kind: ColumnKind::Physical,
+                },
+                ColumnDefinition {
+                    kind: ColumnKind::Embedding(embedding_def),
+                },
+            ],
+        );
+        let rich_schema = table_def.into_rich_schema();
+
+        let table = conn
+            .create_empty_table("embed_test", rich_schema)
+            .execute()
+            .await
+            .unwrap();
+
+        // Now add new data WITHOUT the embedding column - it should be computed automatically
+        let new_batch = record_batch!(("text", Utf8, ["hello", "world"])).unwrap();
+        table.add(new_batch).execute().await.unwrap();
+
+        assert_eq!(table.count_rows(None).await.unwrap(), 2);
+
+        // Query to verify the embeddings were computed for the new rows
+        let results: Vec<RecordBatch> = table
+            .query()
+            .select(Select::columns(&["text", "text_embedding"]))
+            .execute()
+            .await
+            .unwrap()
+            .try_collect()
+            .await
+            .unwrap();
+
+        let total_rows: usize = results.iter().map(|b| b.num_rows()).sum();
+        assert_eq!(total_rows, 2);
+
+        // Check that all rows have embedding values (not null)
+        for batch in &results {
+            let embedding_col = batch.column(1);
+            assert_eq!(embedding_col.null_count(), 0);
+        }
+    }
+}

rust/lancedb/src/table/datafusion.rs

@@ -287,8 +287,7 @@ pub mod tests {
 
     use arrow::array::AsArray;
     use arrow_array::{
-        BinaryArray, Float64Array, Int32Array, Int64Array, RecordBatch, RecordBatchIterator,
-        RecordBatchReader, StringArray, UInt32Array,
+        BinaryArray, Float64Array, Int32Array, Int64Array, RecordBatch, StringArray, UInt32Array,
     };
     use arrow_schema::{DataType, Field, Schema};
     use datafusion::{
@@ -308,7 +307,7 @@ pub mod tests {
         table::datafusion::BaseTableAdapter,
     };
 
-    fn make_test_batches() -> impl RecordBatchReader + Send + Sync + 'static {
+    fn make_test_batches() -> RecordBatch {
         let metadata = HashMap::from_iter(vec![("foo".to_string(), "bar".to_string())]);
         let schema = Arc::new(
             Schema::new(vec![
@@ -317,19 +316,17 @@ pub mod tests {
             ])
             .with_metadata(metadata),
         );
-        RecordBatchIterator::new(
-            vec![RecordBatch::try_new(
-                schema.clone(),
-                vec![
-                    Arc::new(Int32Array::from_iter_values(0..10)),
-                    Arc::new(UInt32Array::from_iter_values(0..10)),
-                ],
-            )],
+        RecordBatch::try_new(
             schema,
+            vec![
+                Arc::new(Int32Array::from_iter_values(0..10)),
+                Arc::new(UInt32Array::from_iter_values(0..10)),
+            ],
         )
+        .unwrap()
     }
 
-    fn make_tbl_two_test_batches() -> impl RecordBatchReader + Send + Sync + 'static {
+    fn make_tbl_two_test_batches() -> RecordBatch {
         let metadata = HashMap::from_iter(vec![("foo".to_string(), "bar".to_string())]);
         let schema = Arc::new(
             Schema::new(vec![
@@ -342,28 +339,26 @@ pub mod tests {
             ])
             .with_metadata(metadata),
         );
-        RecordBatchIterator::new(
-            vec![RecordBatch::try_new(
-                schema.clone(),
-                vec![
-                    Arc::new(Int64Array::from_iter_values(0..1000)),
-                    Arc::new(StringArray::from_iter_values(
-                        (0..1000).map(|i| i.to_string()),
-                    )),
-                    Arc::new(Float64Array::from_iter_values((0..1000).map(|i| i as f64))),
-                    Arc::new(StringArray::from_iter_values(
-                        (0..1000).map(|i| format!("{{\"i\":{}}}", i)),
-                    )),
-                    Arc::new(BinaryArray::from_iter_values(
-                        (0..1000).map(|i| (i as u32).to_be_bytes().to_vec()),
-                    )),
-                    Arc::new(StringArray::from_iter_values(
-                        (0..1000).map(|i| i.to_string()),
-                    )),
-                ],
-            )],
+        RecordBatch::try_new(
             schema,
+            vec![
+                Arc::new(Int64Array::from_iter_values(0..1000)),
+                Arc::new(StringArray::from_iter_values(
+                    (0..1000).map(|i| i.to_string()),
+                )),
+                Arc::new(Float64Array::from_iter_values((0..1000).map(|i| i as f64))),
+                Arc::new(StringArray::from_iter_values(
+                    (0..1000).map(|i| format!("{{\"i\":{}}}", i)),
+                )),
+                Arc::new(BinaryArray::from_iter_values(
+                    (0..1000).map(|i| (i as u32).to_be_bytes().to_vec()),
+                )),
+                Arc::new(StringArray::from_iter_values(
+                    (0..1000).map(|i| i.to_string()),
+                )),
+            ],
         )
+        .unwrap()
     }
 
     struct TestFixture {

rust/lancedb/src/table/datafusion/insert.rs

@@ -222,7 +222,7 @@ mod tests {
     use std::vec;
 
     use super::*;
-    use arrow_array::{record_batch, Int32Array, RecordBatchIterator};
+    use arrow_array::{record_batch, RecordBatchIterator};
     use datafusion::prelude::SessionContext;
     use datafusion_catalog::MemTable;
     use tempfile::tempdir;
@@ -238,11 +238,8 @@ mod tests {
 
         // Create initial table
         let batch = record_batch!(("id", Int32, [1, 2, 3])).unwrap();
-        let schema = batch.schema();
-        let reader = RecordBatchIterator::new(vec![Ok(batch)], schema);
-
         let table = db
-            .create_table("test_insert", Box::new(reader))
+            .create_table("test_insert", batch)
             .execute()
             .await
             .unwrap();
@@ -279,11 +276,8 @@ mod tests {
 
         // Create initial table with 3 rows
         let batch = record_batch!(("id", Int32, [1, 2, 3])).unwrap();
-        let schema = batch.schema();
-        let reader = RecordBatchIterator::new(vec![Ok(batch)], schema);
-
         let table = db
-            .create_table("test_overwrite", Box::new(reader))
+            .create_table("test_overwrite", batch)
             .execute()
             .await
             .unwrap();
@@ -318,20 +312,9 @@ mod tests {
         let db = connect(uri).execute().await.unwrap();
 
         // Create initial table
-        let schema = Arc::new(ArrowSchema::new(vec![Field::new(
-            "id",
-            DataType::Int32,
-            false,
-        )]));
-        let batches = vec![RecordBatch::try_new(
-            schema.clone(),
-            vec![Arc::new(Int32Array::from(vec![1, 2, 3]))],
-        )
-        .unwrap()];
-        let reader = RecordBatchIterator::new(batches.into_iter().map(Ok), schema.clone());
-
+        let batch = record_batch!(("id", Int32, [1, 2, 3])).unwrap();
         let table = db
-            .create_table("test_empty", Box::new(reader))
+            .create_table("test_empty", batch)
             .execute()
             .await
             .unwrap();
@@ -352,12 +335,13 @@ mod tests {
             false,
         )]));
         // Empty batches
-        let source_reader = RecordBatchIterator::new(
-            std::iter::empty::<Result<RecordBatch, arrow_schema::ArrowError>>(),
-            source_schema,
-        );
+        let source_reader: Box<dyn arrow_array::RecordBatchReader + Send> =
+            Box::new(RecordBatchIterator::new(
+                std::iter::empty::<Result<RecordBatch, arrow_schema::ArrowError>>(),
+                source_schema,
+            ));
         let source_table = db
-            .create_table("empty_source", Box::new(source_reader))
+            .create_table("empty_source", source_reader)
             .execute()
             .await
             .unwrap();
@@ -389,20 +373,10 @@ mod tests {
         let db = connect(uri).execute().await.unwrap();
 
         // Create initial table
-        let schema = Arc::new(ArrowSchema::new(vec![Field::new(
-            "id",
-            DataType::Int32,
-            true,
-        )]));
-        let batches =
-            vec![
-                RecordBatch::try_new(schema.clone(), vec![Arc::new(Int32Array::from(vec![1]))])
-                    .unwrap(),
-            ];
-        let reader = RecordBatchIterator::new(batches.into_iter().map(Ok), schema.clone());
-
+        let batch = record_batch!(("id", Int32, [1])).unwrap();
+        let schema = batch.schema();
         let table = db
-            .create_table("test_multi_batch", Box::new(reader))
+            .create_table("test_multi_batch", batch)
             .execute()
             .await
             .unwrap();

rust/lancedb/src/table/datafusion/udtf/fts.rs

@@ -97,7 +97,7 @@ mod tests {
         table::datafusion::BaseTableAdapter,
         Connection, Table,
     };
-    use arrow_array::{Int32Array, RecordBatch, RecordBatchIterator, StringArray};
+    use arrow_array::{Int32Array, RecordBatch, StringArray};
     use arrow_schema::{DataType, Field, Schema as ArrowSchema};
     use datafusion::prelude::SessionContext;
 
@@ -173,14 +173,7 @@ mod tests {
 
         // Create LanceDB database and table
         let db = crate::connect("memory://test").execute().await.unwrap();
-        let table = db
-            .create_table(
-                "foo",
-                RecordBatchIterator::new(vec![Ok(batch)].into_iter(), schema),
-            )
-            .execute()
-            .await
-            .unwrap();
+        let table = db.create_table("foo", batch).execute().await.unwrap();
 
         // Create FTS index
         table
@@ -323,13 +316,7 @@ mod tests {
             RecordBatch::try_new(metadata_schema.clone(), vec![metadata_col, extra_col]).unwrap();
 
         let _metadata_table = db
-            .create_table(
-                "metadata",
-                RecordBatchIterator::new(
-                    vec![Ok(metadata_batch.clone())].into_iter(),
-                    metadata_schema.clone(),
-                ),
-            )
+            .create_table("metadata", metadata_batch.clone())
             .execute()
             .await
             .unwrap();
@@ -393,14 +380,7 @@ mod tests {
         let batch =
             RecordBatch::try_new(schema.clone(), vec![id_col, text_col, category_col]).unwrap();
 
-        let table = db
-            .create_table(
-                table_name,
-                RecordBatchIterator::new(vec![Ok(batch)].into_iter(), schema),
-            )
-            .execute()
-            .await
-            .unwrap();
+        let table = db.create_table(table_name, batch).execute().await.unwrap();
 
         // Create FTS index
         table
@@ -546,14 +526,7 @@ mod tests {
         ]));
         let batch = RecordBatch::try_new(schema.clone(), vec![id_col, text_col]).unwrap();
 
-        let table = db
-            .create_table(
-                "docs",
-                RecordBatchIterator::new(vec![Ok(batch)].into_iter(), schema),
-            )
-            .execute()
-            .await
-            .unwrap();
+        let table = db.create_table("docs", batch).execute().await.unwrap();
 
         // Create FTS index with position information for phrase queries
         table
@@ -691,14 +664,7 @@ mod tests {
         let batch =
             RecordBatch::try_new(schema.clone(), vec![id_col, title_col, content_col]).unwrap();
 
-        let table = db
-            .create_table(
-                "multi_col",
-                RecordBatchIterator::new(vec![Ok(batch)].into_iter(), schema),
-            )
-            .execute()
-            .await
-            .unwrap();
+        let table = db.create_table("multi_col", batch).execute().await.unwrap();
 
         // Create FTS indices on both columns
         table
@@ -963,13 +929,7 @@ mod tests {
         let metadata_batch =
             RecordBatch::try_new(metadata_schema.clone(), vec![metadata_id, extra_info]).unwrap();
         let _metadata_table = db
-            .create_table(
-                "metadata",
-                RecordBatchIterator::new(
-                    vec![Ok(metadata_batch.clone())].into_iter(),
-                    metadata_schema,
-                ),
-            )
+            .create_table("metadata", metadata_batch.clone())
             .execute()
             .await
             .unwrap();
@@ -1358,14 +1318,7 @@ mod tests {
         ]));
         let batch = RecordBatch::try_new(schema.clone(), vec![id_col, text_col]).unwrap();
 
-        let table = db
-            .create_table(
-                "docs",
-                RecordBatchIterator::new(vec![Ok(batch)].into_iter(), schema),
-            )
-            .execute()
-            .await
-            .unwrap();
+        let table = db.create_table("docs", batch).execute().await.unwrap();
 
         // Create FTS index with position information
         table
@@ -1510,14 +1463,7 @@ mod tests {
         let batch =
             RecordBatch::try_new(schema.clone(), vec![id_col, title_col, content_col]).unwrap();
 
-        let table = db
-            .create_table(
-                "docs",
-                RecordBatchIterator::new(vec![Ok(batch)].into_iter(), schema),
-            )
-            .execute()
-            .await
-            .unwrap();
+        let table = db.create_table("docs", batch).execute().await.unwrap();
 
         // Create FTS indices on both columns
         table
@@ -1591,14 +1537,7 @@ mod tests {
         let batch =
             RecordBatch::try_new(schema.clone(), vec![id_col, title_col, content_col]).unwrap();
 
-        let table = db
-            .create_table(
-                "docs",
-                RecordBatchIterator::new(vec![Ok(batch)].into_iter(), schema),
-            )
-            .execute()
-            .await
-            .unwrap();
+        let table = db.create_table("docs", batch).execute().await.unwrap();
 
         // Create FTS indices
         table
@@ -1724,36 +1663,23 @@ mod tests {
             .unwrap();
 
         // Create table with simple text for n-gram testing
-        let data = RecordBatchIterator::new(
-            vec![RecordBatch::try_new(
-                Arc::new(ArrowSchema::new(vec![
-                    Field::new("id", DataType::Int32, false),
-                    Field::new("text", DataType::Utf8, false),
-                ])),
-                vec![
-                    Arc::new(Int32Array::from(vec![1, 2, 3])),
-                    Arc::new(StringArray::from(vec![
-                        "hello world",
-                        "lance database",
-                        "lance is cool",
-                    ])),
-                ],
-            )
-            .unwrap()]
-            .into_iter()
-            .map(Ok),
+        let data = RecordBatch::try_new(
             Arc::new(ArrowSchema::new(vec![
                 Field::new("id", DataType::Int32, false),
                 Field::new("text", DataType::Utf8, false),
             ])),
-        );
+            vec![
+                Arc::new(Int32Array::from(vec![1, 2, 3])),
+                Arc::new(StringArray::from(vec![
+                    "hello world",
+                    "lance database",
+                    "lance is cool",
+                ])),
+            ],
+        )
+        .unwrap();
 
-        let table = Arc::new(
-            db.create_table("docs", Box::new(data))
-                .execute()
-                .await
-                .unwrap(),
-        );
+        let table = Arc::new(db.create_table("docs", data).execute().await.unwrap());
 
         // Create FTS index with n-gram tokenizer (default min_ngram_length=3)
         table
@@ -1876,43 +1802,29 @@ mod tests {
             .unwrap();
 
         // Create table with two text columns
-        let data = RecordBatchIterator::new(
-            vec![RecordBatch::try_new(
-                Arc::new(ArrowSchema::new(vec![
-                    Field::new("id", DataType::Int32, false),
-                    Field::new("title", DataType::Utf8, false),
-                    Field::new("content", DataType::Utf8, false),
-                ])),
-                vec![
-                    Arc::new(Int32Array::from(vec![1, 2, 3])),
-                    Arc::new(StringArray::from(vec![
-                        "Important Document",
-                        "Another Document",
-                        "Random Text",
-                    ])),
-                    Arc::new(StringArray::from(vec![
-                        "This is important information",
-                        "This has details",
-                        "Nothing special here",
-                    ])),
-                ],
-            )
-            .unwrap()]
-            .into_iter()
-            .map(Ok),
+        let data = RecordBatch::try_new(
             Arc::new(ArrowSchema::new(vec![
                 Field::new("id", DataType::Int32, false),
                 Field::new("title", DataType::Utf8, false),
                 Field::new("content", DataType::Utf8, false),
             ])),
-        );
+            vec![
+                Arc::new(Int32Array::from(vec![1, 2, 3])),
+                Arc::new(StringArray::from(vec![
+                    "Important Document",
+                    "Another Document",
+                    "Random Text",
+                ])),
+                Arc::new(StringArray::from(vec![
+                    "This is important information",
+                    "This has details",
+                    "Nothing special here",
+                ])),
+            ],
+        )
+        .unwrap();
 
-        let table = Arc::new(
-            db.create_table("docs", Box::new(data))
-                .execute()
-                .await
-                .unwrap(),
-        );
+        let table = Arc::new(db.create_table("docs", data).execute().await.unwrap());
 
         // Create FTS indices on both columns
         table

rust/lancedb/src/table/dataset.rs

@@ -57,15 +57,6 @@ impl DatasetRef {
         matches!(self, Self::Latest { .. })
     }
 
-    async fn need_reload(&self) -> Result<bool> {
-        Ok(match self {
-            Self::Latest { dataset, .. } => {
-                dataset.latest_version_id().await? != dataset.version().version
-            }
-            Self::TimeTravel { dataset, version } => dataset.version().version != *version,
-        })
-    }
-
     async fn as_latest(&mut self, read_consistency_interval: Option<Duration>) -> Result<()> {
         match self {
             Self::Latest { .. } => Ok(()),
@@ -118,6 +109,21 @@ impl DatasetRef {
         Ok(())
     }
 
+    fn is_up_to_date(&self) -> bool {
+        match self {
+            Self::Latest {
+                read_consistency_interval,
+                last_consistency_check,
+                ..
+            } => match (read_consistency_interval, last_consistency_check) {
+                (None, _) => true,
+                (Some(_), None) => false,
+                (Some(interval), Some(last_check)) => last_check.elapsed() < *interval,
+            },
+            Self::TimeTravel { dataset, version } => dataset.version().version == *version,
+        }
+    }
+
     fn time_travel_version(&self) -> Option<u64> {
         match self {
             Self::Latest { .. } => None,
@@ -205,18 +211,7 @@ impl DatasetConsistencyWrapper {
     }
 
     pub async fn reload(&self) -> Result<()> {
-        if !self.0.read().await.need_reload().await? {
-            return Ok(());
-        }
-
-        let mut write_guard = self.0.write().await;
-        // on lock escalation -- check if someone else has already reloaded
-        if !write_guard.need_reload().await? {
-            return Ok(());
-        }
-
-        // actually need reloading
-        write_guard.reload().await
+        self.0.write().await.reload().await
     }
 
     /// Returns the version, if in time travel mode, or None otherwise
@@ -235,35 +230,20 @@ impl DatasetConsistencyWrapper {
         }
     }
 
-    async fn is_up_to_date(&self) -> Result<bool> {
-        let dataset_ref = self.0.read().await;
-        match &*dataset_ref {
-            DatasetRef::Latest {
-                read_consistency_interval,
-                last_consistency_check,
-                ..
-            } => match (read_consistency_interval, last_consistency_check) {
-                (None, _) => Ok(true),
-                (Some(_), None) => Ok(false),
-                (Some(read_consistency_interval), Some(last_consistency_check)) => {
-                    if &last_consistency_check.elapsed() < read_consistency_interval {
-                        Ok(true)
-                    } else {
-                        Ok(false)
-                    }
-                }
-            },
-            DatasetRef::TimeTravel { dataset, version } => {
-                Ok(dataset.version().version == *version)
-            }
-        }
+    async fn is_up_to_date(&self) -> bool {
+        self.0.read().await.is_up_to_date()
     }
 
     /// Ensures that the dataset is loaded and up-to-date with consistency and
     /// version parameters.
     async fn ensure_up_to_date(&self) -> Result<()> {
-        if !self.is_up_to_date().await? {
-            self.reload().await?;
+        if !self.is_up_to_date().await {
+            // Re-check under write lock — another task may have reloaded
+            // while we waited for the lock.
+            let mut write_guard = self.0.write().await;
+            if !write_guard.is_up_to_date() {
+                write_guard.reload().await?;
+            }
         }
         Ok(())
     }
@@ -351,4 +331,60 @@ mod tests {
         let stats = io_stats.incremental_stats();
         assert_eq!(stats.read_iops, 1);
     }
+
+    /// Regression test: before the fix, the reload fast-path (no version change)
+    /// did not reset `last_consistency_check`, causing a list call on every
+    /// subsequent query once the interval expired.
+    #[tokio::test]
+    async fn test_reload_resets_consistency_timer() {
+        let db = connect("memory://")
+            .read_consistency_interval(Duration::from_secs(1))
+            .execute()
+            .await
+            .unwrap();
+        let io_stats = IoStatsHolder::default();
+        let schema = Arc::new(Schema::new(vec![Field::new("id", DataType::Int32, false)]));
+        let table = db
+            .create_empty_table("test", schema)
+            .write_options(WriteOptions {
+                lance_write_params: Some(WriteParams {
+                    store_params: Some(ObjectStoreParams {
+                        object_store_wrapper: Some(Arc::new(io_stats.clone())),
+                        ..Default::default()
+                    }),
+                    ..Default::default()
+                }),
+            })
+            .execute()
+            .await
+            .unwrap();
+
+        let start = Instant::now();
+        io_stats.incremental_stats(); // reset
+
+        // Step 1: within interval — no list
+        table.schema().await.unwrap();
+        let s = io_stats.incremental_stats();
+        assert_eq!(s.read_iops, 0, "step 1, elapsed={:?}", start.elapsed());
+
+        // Step 2: still within interval — no list
+        table.schema().await.unwrap();
+        let s = io_stats.incremental_stats();
+        assert_eq!(s.read_iops, 0, "step 2, elapsed={:?}", start.elapsed());
+
+        // Step 3: sleep past the 1s boundary
+        tokio::time::sleep(Duration::from_secs(1)).await;
+
+        // Step 4: interval expired — exactly 1 list, timer resets
+        table.schema().await.unwrap();
+        let s = io_stats.incremental_stats();
+        assert_eq!(s.read_iops, 1, "step 4, elapsed={:?}", start.elapsed());
+
+        // Step 5: 10 more calls — timer just reset, no lists (THIS is the regression test).
+        for _ in 0..10 {
+            table.schema().await.unwrap();
+        }
+        let s = io_stats.incremental_stats();
+        assert_eq!(s.read_iops, 0, "step 5, elapsed={:?}", start.elapsed());
+    }
 }

rust/lancedb/src/table/delete.rs

@@ -34,7 +34,7 @@ pub(crate) async fn execute_delete(table: &NativeTable, predicate: &str) -> Resu
 #[cfg(test)]
 mod tests {
     use crate::connect;
-    use arrow_array::{record_batch, Int32Array, RecordBatch, RecordBatchIterator};
+    use arrow_array::{record_batch, Int32Array, RecordBatch};
     use arrow_schema::{DataType, Field, Schema};
     use std::sync::Arc;
 
@@ -53,10 +53,7 @@ mod tests {
         .unwrap();
 
         let table = conn
-            .create_table(
-                "test_delete",
-                RecordBatchIterator::new(vec![Ok(batch)], schema),
-            )
+            .create_table("test_delete", batch)
             .execute()
             .await
             .unwrap();
@@ -102,10 +99,7 @@ mod tests {
         let original_schema = batch.schema();
 
         let table = conn
-            .create_table(
-                "test_delete_all",
-                RecordBatchIterator::new(vec![Ok(batch)], original_schema.clone()),
-            )
+            .create_table("test_delete_all", batch)
             .execute()
             .await
             .unwrap();
@@ -126,13 +120,8 @@ mod tests {
         // Create a table with 5 rows
         let batch = record_batch!(("id", Int32, [1, 2, 3, 4, 5])).unwrap();
 
-        let schema = batch.schema();
-
         let table = conn
-            .create_table(
-                "test_delete_noop",
-                RecordBatchIterator::new(vec![Ok(batch)], schema),
-            )
+            .create_table("test_delete_noop", batch)
             .execute()
             .await
             .unwrap();

rust/lancedb/src/table/optimize.rs

@@ -0,0 +1,729 @@
+// SPDX-License-Identifier: Apache-2.0
+// SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+//! Table optimization operations for compaction, pruning, and index optimization.
+//!
+//! This module contains the implementation of optimization operations that help
+//! maintain good performance for LanceDB tables.
+
+use std::sync::Arc;
+
+use lance::dataset::cleanup::RemovalStats;
+use lance::dataset::optimize::{compact_files, CompactionMetrics, IndexRemapperOptions};
+use lance_index::optimize::OptimizeOptions;
+use lance_index::DatasetIndexExt;
+use log::info;
+
+pub use chrono::Duration;
+pub use lance::dataset::optimize::CompactionOptions;
+
+use super::NativeTable;
+use crate::error::Result;
+
+/// Optimize the dataset.
+///
+/// Similar to `VACUUM` in PostgreSQL, it offers different options to
+/// optimize different parts of the table on disk.
+///
+/// By default, it optimizes everything, as [`OptimizeAction::All`].
+pub enum OptimizeAction {
+    /// Run all optimizations with default values
+    All,
+    /// Compacts files in the dataset
+    ///
+    /// LanceDb uses a readonly filesystem for performance and safe concurrency.  Every time
+    /// new data is added it will be added into new files.  Small files
+    /// can hurt both read and write performance.  Compaction will merge small files
+    /// into larger ones.
+    ///
+    /// All operations that modify data (add, delete, update, merge insert, etc.) will create
+    /// new files.  If these operations are run frequently then compaction should run frequently.
+    ///
+    /// If these operations are never run (search only) then compaction is not necessary.
+    Compact {
+        options: CompactionOptions,
+        remap_options: Option<Arc<dyn IndexRemapperOptions>>,
+    },
+    /// Prune old version of datasets
+    ///
+    /// Every change in LanceDb is additive.  When data is removed from a dataset a new version is
+    /// created that doesn't contain the removed data.  However, the old version, which does contain
+    /// the removed data, is left in place.  This is necessary for consistency and concurrency and
+    /// also enables time travel functionality like the ability to checkout an older version of the
+    /// dataset to undo changes.
+    ///
+    /// Over time, these old versions can consume a lot of disk space.  The prune operation will
+    /// remove versions of the dataset that are older than a certain age.  This will free up the
+    /// space used by that old data.
+    ///
+    /// Once a version is pruned it can no longer be checked out.
+    Prune {
+        /// The duration of time to keep versions of the dataset.
+        older_than: Option<Duration>,
+        /// 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`.
+        delete_unverified: Option<bool>,
+        /// If true, an error will be returned if there are any old versions that are still tagged.
+        error_if_tagged_old_versions: Option<bool>,
+    },
+    /// Optimize the indices
+    ///
+    /// This operation optimizes all indices in the table.  When new data is added to LanceDb
+    /// it is not added to the indices.  However, it can still turn up in searches because the search
+    /// function will scan both the indexed data and the unindexed data in parallel.  Over time, the
+    /// unindexed data can become large enough that the search performance is slow.  This operation
+    /// will add the unindexed data to the indices without rerunning the full index creation process.
+    ///
+    /// Optimizing an index is faster than re-training the index but it does not typically adjust the
+    /// underlying model relied upon by the index.  This can eventually lead to poor search accuracy
+    /// and so users may still want to occasionally retrain the index after adding a large amount of
+    /// data.
+    ///
+    /// For example, when using IVF, an index will create clusters.  Optimizing an index assigns unindexed
+    /// data to the existing clusters, but it does not move the clusters or create new clusters.
+    Index(OptimizeOptions),
+}
+
+impl Default for OptimizeAction {
+    fn default() -> Self {
+        Self::All
+    }
+}
+
+/// Statistics about the optimization.
+#[derive(Debug, Default)]
+pub struct OptimizeStats {
+    /// Stats of the file compaction.
+    pub compaction: Option<CompactionMetrics>,
+
+    /// Stats of the version pruning
+    pub prune: Option<RemovalStats>,
+}
+
+/// Internal implementation of optimize_indices
+///
+/// This logic was moved from NativeTable to keep table.rs clean.
+pub(crate) async fn optimize_indices(table: &NativeTable, options: &OptimizeOptions) -> Result<()> {
+    info!("LanceDB: optimizing indices: {:?}", options);
+    table
+        .dataset
+        .get_mut()
+        .await?
+        .optimize_indices(options)
+        .await?;
+    Ok(())
+}
+
+/// Remove old versions of the dataset from disk.
+///
+/// # Arguments
+/// * `older_than` - The duration of time to keep versions of the dataset.
+/// * `delete_unverified` - 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`.
+///
+/// This calls into [lance::dataset::Dataset::cleanup_old_versions] and
+/// returns the result.
+pub(crate) async fn cleanup_old_versions(
+    table: &NativeTable,
+    older_than: Duration,
+    delete_unverified: Option<bool>,
+    error_if_tagged_old_versions: Option<bool>,
+) -> Result<RemovalStats> {
+    Ok(table
+        .dataset
+        .get_mut()
+        .await?
+        .cleanup_old_versions(older_than, delete_unverified, error_if_tagged_old_versions)
+        .await?)
+}
+
+/// Compact files in the dataset.
+///
+/// This can be run after making several small appends to optimize the table
+/// for faster reads.
+///
+/// This calls into [lance::dataset::optimize::compact_files].
+pub(crate) async fn compact_files_impl(
+    table: &NativeTable,
+    options: CompactionOptions,
+    remap_options: Option<Arc<dyn IndexRemapperOptions>>,
+) -> Result<CompactionMetrics> {
+    let mut dataset_mut = table.dataset.get_mut().await?;
+    let metrics = compact_files(&mut dataset_mut, options, remap_options).await?;
+    Ok(metrics)
+}
+
+/// Execute the optimize operation on the table.
+///
+/// This is the main entry point for all optimization operations.
+pub(crate) async fn execute_optimize(
+    table: &NativeTable,
+    action: OptimizeAction,
+) -> Result<OptimizeStats> {
+    let mut stats = OptimizeStats {
+        compaction: None,
+        prune: None,
+    };
+    match action {
+        OptimizeAction::All => {
+            // Call helper functions directly to avoid async recursion issues
+            stats.compaction =
+                Some(compact_files_impl(table, CompactionOptions::default(), None).await?);
+            stats.prune = Some(
+                cleanup_old_versions(
+                    table,
+                    Duration::try_days(7).expect("valid delta"),
+                    None,
+                    None,
+                )
+                .await?,
+            );
+            optimize_indices(table, &OptimizeOptions::default()).await?;
+        }
+        OptimizeAction::Compact {
+            options,
+            remap_options,
+        } => {
+            stats.compaction = Some(compact_files_impl(table, options, remap_options).await?);
+        }
+        OptimizeAction::Prune {
+            older_than,
+            delete_unverified,
+            error_if_tagged_old_versions,
+        } => {
+            stats.prune = Some(
+                cleanup_old_versions(
+                    table,
+                    older_than.unwrap_or(Duration::try_days(7).expect("valid delta")),
+                    delete_unverified,
+                    error_if_tagged_old_versions,
+                )
+                .await?,
+            );
+        }
+        OptimizeAction::Index(options) => {
+            optimize_indices(table, &options).await?;
+        }
+    }
+    Ok(stats)
+}
+
+#[cfg(test)]
+mod tests {
+    use arrow_array::{Int32Array, RecordBatch, StringArray};
+    use arrow_schema::{DataType, Field, Schema};
+    use rstest::rstest;
+    use std::sync::Arc;
+
+    use crate::connect;
+    use crate::index::{scalar::BTreeIndexBuilder, Index};
+    use crate::query::ExecutableQuery;
+    use crate::table::{CompactionOptions, OptimizeAction, OptimizeStats};
+    use futures::TryStreamExt;
+
+    #[tokio::test]
+    async fn test_optimize_compact_simple() {
+        let conn = connect("memory://").execute().await.unwrap();
+
+        // Create a table with initial data
+        let schema = Arc::new(Schema::new(vec![Field::new("i", DataType::Int32, false)]));
+        let batch = RecordBatch::try_new(
+            schema.clone(),
+            vec![Arc::new(Int32Array::from_iter_values(0..100))],
+        )
+        .unwrap();
+
+        let table = conn
+            .create_table("test_compact", batch)
+            .execute()
+            .await
+            .unwrap();
+
+        // Add more data to create multiple fragments
+        for i in 0..5 {
+            let batch = RecordBatch::try_new(
+                schema.clone(),
+                vec![Arc::new(Int32Array::from_iter_values(
+                    (i * 100 + 100)..((i + 1) * 100 + 100),
+                ))],
+            )
+            .unwrap();
+            table.add(batch).execute().await.unwrap();
+        }
+
+        // Verify we have multiple fragments before compaction
+        let initial_row_count = table.count_rows(None).await.unwrap();
+        assert_eq!(initial_row_count, 600);
+
+        // Run compaction
+        let stats = table
+            .optimize(OptimizeAction::Compact {
+                options: CompactionOptions {
+                    target_rows_per_fragment: 1000,
+                    ..Default::default()
+                },
+                remap_options: None,
+            })
+            .await
+            .unwrap();
+
+        // Verify compaction occurred
+        assert!(stats.compaction.is_some());
+        let compaction_metrics = stats.compaction.unwrap();
+        assert!(compaction_metrics.fragments_removed > 0);
+
+        // Verify data integrity after compaction
+        let final_row_count = table.count_rows(None).await.unwrap();
+        assert_eq!(final_row_count, 600);
+
+        // Verify data content is correct
+        let batches = table
+            .query()
+            .execute()
+            .await
+            .unwrap()
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+
+        let total_rows: usize = batches.iter().map(|b| b.num_rows()).sum();
+        assert_eq!(total_rows, 600);
+
+        // Verify the values are as expected
+        let mut all_values: Vec<i32> = Vec::new();
+        for batch in &batches {
+            let array = batch["i"].as_any().downcast_ref::<Int32Array>().unwrap();
+            all_values.extend(array.values().iter().copied());
+        }
+        all_values.sort();
+        let expected: Vec<i32> = (0..600).collect();
+        assert_eq!(all_values, expected);
+    }
+
+    #[tokio::test]
+    async fn test_optimize_prune_versions() {
+        let conn = connect("memory://").execute().await.unwrap();
+
+        // Create a table
+        let schema = Arc::new(Schema::new(vec![Field::new("i", DataType::Int32, false)]));
+        let batch = RecordBatch::try_new(
+            schema.clone(),
+            vec![Arc::new(Int32Array::from_iter_values(0..10))],
+        )
+        .unwrap();
+
+        let table = conn
+            .create_table("test_prune", batch)
+            .execute()
+            .await
+            .unwrap();
+
+        // Make several modifications to create versions
+        for i in 0..5 {
+            let batch = RecordBatch::try_new(
+                schema.clone(),
+                vec![Arc::new(Int32Array::from_iter_values(
+                    (i * 10 + 10)..((i + 1) * 10 + 10),
+                ))],
+            )
+            .unwrap();
+            table.add(batch).execute().await.unwrap();
+        }
+
+        // Verify multiple versions exist
+        let versions = table.list_versions().await.unwrap();
+        assert!(versions.len() > 1);
+
+        // Run prune with a very old cutoff (won't delete recent versions)
+        let stats = table
+            .optimize(OptimizeAction::Prune {
+                older_than: Some(chrono::Duration::try_days(0).unwrap()),
+                delete_unverified: Some(true),
+                error_if_tagged_old_versions: None,
+            })
+            .await
+            .unwrap();
+
+        // Prune-only operation should not have compaction stats
+        assert!(stats.compaction.is_none());
+
+        // Verify prune stats
+        let prune_stats = stats.prune.unwrap();
+        assert!(prune_stats.bytes_removed > 0);
+        assert_eq!(prune_stats.old_versions, 5);
+
+        // Verify data is still intact
+        let final_row_count = table.count_rows(None).await.unwrap();
+        assert_eq!(final_row_count, 60);
+
+        // Verify data content is correct
+        let batches = table
+            .query()
+            .execute()
+            .await
+            .unwrap()
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+
+        let mut all_values: Vec<i32> = Vec::new();
+        for batch in &batches {
+            let array = batch["i"].as_any().downcast_ref::<Int32Array>().unwrap();
+            all_values.extend(array.values().iter().copied());
+        }
+        all_values.sort();
+        let expected: Vec<i32> = (0..60).collect();
+        assert_eq!(all_values, expected);
+    }
+
+    #[tokio::test]
+    async fn test_optimize_index() {
+        let conn = connect("memory://").execute().await.unwrap();
+
+        // Create a table with data
+        let schema = Arc::new(Schema::new(vec![Field::new("i", DataType::Int32, false)]));
+        let batch = RecordBatch::try_new(
+            schema.clone(),
+            vec![Arc::new(Int32Array::from_iter_values(0..100))],
+        )
+        .unwrap();
+
+        let table = conn
+            .create_table("test_index_optimize", batch)
+            .execute()
+            .await
+            .unwrap();
+
+        // Create an index
+        table
+            .create_index(&["i"], Index::BTree(BTreeIndexBuilder::default()))
+            .execute()
+            .await
+            .unwrap();
+
+        // Add more data (unindexed)
+        let batch = RecordBatch::try_new(
+            schema.clone(),
+            vec![Arc::new(Int32Array::from_iter_values(100..200))],
+        )
+        .unwrap();
+        table.add(batch).execute().await.unwrap();
+
+        // Verify index stats before optimization
+        let indices = table.list_indices().await.unwrap();
+        assert_eq!(indices.len(), 1);
+        let index_name = indices[0].name.clone();
+        let stats_before = table.index_stats(&index_name).await.unwrap().unwrap();
+        assert_eq!(stats_before.num_indexed_rows, 100);
+        assert_eq!(stats_before.num_unindexed_rows, 100);
+
+        // Run index optimization
+        let stats = table
+            .optimize(OptimizeAction::Index(Default::default()))
+            .await
+            .unwrap();
+
+        // For index optimization, compaction and prune stats should be None
+        assert!(stats.compaction.is_none());
+        assert!(stats.prune.is_none());
+
+        // Verify index stats after optimization
+        let stats_after = table.index_stats(&index_name).await.unwrap().unwrap();
+        assert_eq!(stats_after.num_indexed_rows, 200);
+        assert_eq!(stats_after.num_unindexed_rows, 0);
+        assert!(stats_after.num_indices.is_some());
+
+        // Verify data integrity
+        let final_row_count = table.count_rows(None).await.unwrap();
+        assert_eq!(final_row_count, 200);
+    }
+
+    #[tokio::test]
+    async fn test_optimize_all() {
+        let conn = connect("memory://").execute().await.unwrap();
+
+        // Create a table with data
+        let schema = Arc::new(Schema::new(vec![Field::new("i", DataType::Int32, false)]));
+        let batch = RecordBatch::try_new(
+            schema.clone(),
+            vec![Arc::new(Int32Array::from_iter_values(0..100))],
+        )
+        .unwrap();
+
+        let table = conn
+            .create_table("test_optimize_all", batch)
+            .execute()
+            .await
+            .unwrap();
+
+        // Add more data
+        for i in 0..3 {
+            let batch = RecordBatch::try_new(
+                schema.clone(),
+                vec![Arc::new(Int32Array::from_iter_values(
+                    (i * 100 + 100)..((i + 1) * 100 + 100),
+                ))],
+            )
+            .unwrap();
+            table.add(batch).execute().await.unwrap();
+        }
+
+        // Run all optimizations
+        let stats = table.optimize(OptimizeAction::All).await.unwrap();
+
+        // Verify stats from both compaction and prune
+        assert!(stats.compaction.is_some());
+        assert!(stats.prune.is_some());
+
+        // Verify data integrity
+        let final_row_count = table.count_rows(None).await.unwrap();
+        assert_eq!(final_row_count, 400);
+
+        // Verify data content
+        let batches = table
+            .query()
+            .execute()
+            .await
+            .unwrap()
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+
+        let mut all_values: Vec<i32> = Vec::new();
+        for batch in &batches {
+            let array = batch["i"].as_any().downcast_ref::<Int32Array>().unwrap();
+            all_values.extend(array.values().iter().copied());
+        }
+        all_values.sort();
+        let expected: Vec<i32> = (0..400).collect();
+        assert_eq!(all_values, expected);
+    }
+
+    #[tokio::test]
+    async fn test_optimize_default_action() {
+        // Verify that default action is All
+        let action: OptimizeAction = Default::default();
+        assert!(matches!(action, OptimizeAction::All));
+    }
+
+    #[tokio::test]
+    async fn test_optimize_stats_default() {
+        // Verify OptimizeStats default values
+        let stats: OptimizeStats = Default::default();
+        assert!(stats.compaction.is_none());
+        assert!(stats.prune.is_none());
+    }
+
+    #[tokio::test]
+    async fn test_compact_with_deferred_index_remap() {
+        // Smoke test: verifies compaction with deferred index remap doesn't error.
+        // We don't currently assert that remap is actually deferred.
+        let conn = connect("memory://").execute().await.unwrap();
+
+        // Create a table with data
+        let schema = Arc::new(Schema::new(vec![Field::new("id", DataType::Int32, false)]));
+        let batch = RecordBatch::try_new(
+            schema.clone(),
+            vec![Arc::new(Int32Array::from_iter_values(0..100))],
+        )
+        .unwrap();
+
+        let table = conn
+            .create_table("test_deferred_remap", batch.clone())
+            .execute()
+            .await
+            .unwrap();
+
+        // Add more data
+        table.add(batch).execute().await.unwrap();
+
+        // Create an index
+        table
+            .create_index(&["id"], Index::BTree(BTreeIndexBuilder::default()))
+            .execute()
+            .await
+            .unwrap();
+
+        // Run compaction with deferred index remap
+        let stats = table
+            .optimize(OptimizeAction::Compact {
+                options: CompactionOptions {
+                    target_rows_per_fragment: 2000,
+                    defer_index_remap: true,
+                    ..Default::default()
+                },
+                remap_options: None,
+            })
+            .await
+            .unwrap();
+
+        assert!(stats.compaction.is_some());
+
+        // Verify data integrity after compaction
+        let final_row_count = table.count_rows(None).await.unwrap();
+        assert_eq!(final_row_count, 200);
+
+        // Verify data content is correct
+        let batches = table
+            .query()
+            .execute()
+            .await
+            .unwrap()
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+
+        let mut all_values: Vec<i32> = Vec::new();
+        for batch in &batches {
+            let array = batch["id"].as_any().downcast_ref::<Int32Array>().unwrap();
+            all_values.extend(array.values().iter().copied());
+        }
+        all_values.sort();
+
+        // Since we added the same data twice (0..100 twice), we expect 200 values
+        // with values 0-99 appearing twice
+        let mut expected: Vec<i32> = (0..100).chain(0..100).collect();
+        expected.sort();
+        assert_eq!(all_values, expected);
+    }
+
+    #[tokio::test]
+    async fn test_compaction_preserves_schema() {
+        let conn = connect("memory://").execute().await.unwrap();
+
+        // Create a table with multiple columns
+        let schema = Arc::new(Schema::new(vec![
+            Field::new("id", DataType::Int32, false),
+            Field::new("name", DataType::Utf8, true),
+        ]));
+
+        let batch = RecordBatch::try_new(
+            schema.clone(),
+            vec![
+                Arc::new(Int32Array::from_iter_values(0..10)),
+                Arc::new(StringArray::from(
+                    (0..10).map(|i| format!("name_{}", i)).collect::<Vec<_>>(),
+                )),
+            ],
+        )
+        .unwrap();
+
+        let original_schema = batch.schema();
+
+        let table = conn
+            .create_table("test_schema_preserved", batch.clone())
+            .execute()
+            .await
+            .unwrap();
+
+        // Add more data
+        table.add(batch).execute().await.unwrap();
+
+        // Run compaction
+        table
+            .optimize(OptimizeAction::Compact {
+                options: CompactionOptions::default(),
+                remap_options: None,
+            })
+            .await
+            .unwrap();
+
+        // Verify schema is preserved
+        let current_schema = table.schema().await.unwrap();
+        assert_eq!(current_schema, original_schema);
+
+        // Verify data is intact and correct
+        let batches = table
+            .query()
+            .execute()
+            .await
+            .unwrap()
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+
+        let total_rows: usize = batches.iter().map(|b| b.num_rows()).sum();
+        assert_eq!(total_rows, 20);
+    }
+
+    #[tokio::test]
+    async fn test_optimize_empty_table() {
+        let conn = connect("memory://").execute().await.unwrap();
+
+        // Create a table and delete all data
+        let schema = Arc::new(Schema::new(vec![Field::new("i", DataType::Int32, false)]));
+        let batch = RecordBatch::try_new(
+            schema.clone(),
+            vec![Arc::new(Int32Array::from_iter_values(0..10))],
+        )
+        .unwrap();
+
+        let table = conn
+            .create_table("test_empty_optimize", batch)
+            .execute()
+            .await
+            .unwrap();
+
+        // Delete all rows
+        table.delete("true").await.unwrap();
+
+        // Verify table is empty
+        assert_eq!(table.count_rows(None).await.unwrap(), 0);
+
+        // Optimize should work on empty table
+        let stats = table.optimize(OptimizeAction::All).await.unwrap();
+        assert!(stats.compaction.is_some());
+        assert!(stats.prune.is_some());
+
+        // Verify table is still empty but schema is preserved
+        assert_eq!(table.count_rows(None).await.unwrap(), 0);
+        let current_schema = table.schema().await.unwrap();
+        assert_eq!(current_schema, schema);
+    }
+
+    #[rstest]
+    #[case::all(OptimizeAction::All)]
+    #[case::compact(OptimizeAction::Compact {
+        options: CompactionOptions::default(),
+        remap_options: None,
+    })]
+    #[case::prune(OptimizeAction::Prune {
+        older_than: Some(chrono::Duration::try_days(0).unwrap()),
+        delete_unverified: Some(true),
+        error_if_tagged_old_versions: None,
+    })]
+    #[case::index(OptimizeAction::Index(Default::default()))]
+    #[tokio::test]
+    async fn test_optimize_fails_on_checked_out_table(#[case] action: OptimizeAction) {
+        let conn = connect("memory://").execute().await.unwrap();
+
+        let schema = Arc::new(Schema::new(vec![Field::new("i", DataType::Int32, false)]));
+        let batch = RecordBatch::try_new(
+            schema.clone(),
+            vec![Arc::new(Int32Array::from_iter_values(0..10))],
+        )
+        .unwrap();
+
+        let table = conn
+            .create_table("test_checkout_optimize", batch.clone())
+            .execute()
+            .await
+            .unwrap();
+
+        table.add(batch).execute().await.unwrap();
+
+        table.checkout(1).await.unwrap();
+
+        let result = table.optimize(action).await;
+        assert!(result.is_err());
+
+        let err_msg = result.unwrap_err().to_string();
+        assert!(
+            err_msg.contains("cannot be modified when a specific version is checked out"),
+            "Expected error message about checked out table, got: {}",
+            err_msg
+        );
+    }
+}

rust/lancedb/src/table/schema_evolution.rs

@@ -89,7 +89,7 @@ pub(crate) async fn execute_drop_columns(
 
 #[cfg(test)]
 mod tests {
-    use arrow_array::{record_batch, Int32Array, RecordBatchIterator, StringArray};
+    use arrow_array::{record_batch, Int32Array, StringArray};
     use arrow_schema::DataType;
     use futures::TryStreamExt;
     use lance::dataset::ColumnAlteration;
@@ -105,13 +105,9 @@ mod tests {
         let conn = connect("memory://").execute().await.unwrap();
 
         let batch = record_batch!(("id", Int32, [1, 2, 3, 4, 5])).unwrap();
-        let schema = batch.schema();
 
         let table = conn
-            .create_table(
-                "test_add_columns",
-                RecordBatchIterator::new(vec![Ok(batch)], schema),
-            )
+            .create_table("test_add_columns", batch)
             .execute()
             .await
             .unwrap();
@@ -169,13 +165,9 @@ mod tests {
         let conn = connect("memory://").execute().await.unwrap();
 
         let batch = record_batch!(("x", Int32, [10, 20, 30])).unwrap();
-        let schema = batch.schema();
 
         let table = conn
-            .create_table(
-                "test_add_multi_columns",
-                RecordBatchIterator::new(vec![Ok(batch)], schema),
-            )
+            .create_table("test_add_multi_columns", batch)
             .execute()
             .await
             .unwrap();
@@ -205,13 +197,9 @@ mod tests {
         let conn = connect("memory://").execute().await.unwrap();
 
         let batch = record_batch!(("id", Int32, [1, 2, 3])).unwrap();
-        let schema = batch.schema();
 
         let table = conn
-            .create_table(
-                "test_add_const_column",
-                RecordBatchIterator::new(vec![Ok(batch)], schema),
-            )
+            .create_table("test_add_const_column", batch)
             .execute()
             .await
             .unwrap();
@@ -255,13 +243,9 @@ mod tests {
         let conn = connect("memory://").execute().await.unwrap();
 
         let batch = record_batch!(("old_name", Int32, [1, 2, 3])).unwrap();
-        let schema = batch.schema();
 
         let table = conn
-            .create_table(
-                "test_alter_rename",
-                RecordBatchIterator::new(vec![Ok(batch)], schema),
-            )
+            .create_table("test_alter_rename", batch)
             .execute()
             .await
             .unwrap();
@@ -304,10 +288,7 @@ mod tests {
         .unwrap();
 
         let table = conn
-            .create_table(
-                "test_alter_nullable",
-                RecordBatchIterator::new(vec![Ok(batch)], schema),
-            )
+            .create_table("test_alter_nullable", batch)
             .execute()
             .await
             .unwrap();
@@ -332,13 +313,9 @@ mod tests {
         let conn = connect("memory://").execute().await.unwrap();
 
         let batch = record_batch!(("num", Int32, [1, 2, 3])).unwrap();
-        let schema = batch.schema();
 
         let table = conn
-            .create_table(
-                "test_cast_type",
-                RecordBatchIterator::new(vec![Ok(batch)], schema),
-            )
+            .create_table("test_cast_type", batch)
             .execute()
             .await
             .unwrap();
@@ -379,13 +356,9 @@ mod tests {
         let conn = connect("memory://").execute().await.unwrap();
 
         let batch = record_batch!(("num", Int32, [1, 2, 3])).unwrap();
-        let schema = batch.schema();
 
         let table = conn
-            .create_table(
-                "test_invalid_cast",
-                RecordBatchIterator::new(vec![Ok(batch)], schema),
-            )
+            .create_table("test_invalid_cast", batch)
             .execute()
             .await
             .unwrap();
@@ -407,13 +380,9 @@ mod tests {
         let conn = connect("memory://").execute().await.unwrap();
 
         let batch = record_batch!(("a", Int32, [1, 2, 3]), ("b", Int32, [4, 5, 6])).unwrap();
-        let schema = batch.schema();
 
         let table = conn
-            .create_table(
-                "test_alter_multi",
-                RecordBatchIterator::new(vec![Ok(batch)], schema),
-            )
+            .create_table("test_alter_multi", batch)
             .execute()
             .await
             .unwrap();
@@ -441,13 +410,9 @@ mod tests {
 
         let batch =
             record_batch!(("keep", Int32, [1, 2, 3]), ("remove", Int32, [4, 5, 6])).unwrap();
-        let schema = batch.schema();
 
         let table = conn
-            .create_table(
-                "test_drop_single",
-                RecordBatchIterator::new(vec![Ok(batch)], schema),
-            )
+            .create_table("test_drop_single", batch)
             .execute()
             .await
             .unwrap();
@@ -478,13 +443,9 @@ mod tests {
             ("d", Int32, [7, 8])
         )
         .unwrap();
-        let schema = batch.schema();
 
         let table = conn
-            .create_table(
-                "test_drop_multi",
-                RecordBatchIterator::new(vec![Ok(batch)], schema),
-            )
+            .create_table("test_drop_multi", batch)
             .execute()
             .await
             .unwrap();
@@ -511,13 +472,9 @@ mod tests {
             ("extra", Int32, [10, 20, 30])
         )
         .unwrap();
-        let schema = batch.schema();
 
         let table = conn
-            .create_table(
-                "test_drop_preserves",
-                RecordBatchIterator::new(vec![Ok(batch)], schema),
-            )
+            .create_table("test_drop_preserves", batch)
             .execute()
             .await
             .unwrap();
@@ -567,13 +524,9 @@ mod tests {
         let conn = connect("memory://").execute().await.unwrap();
 
         let batch = record_batch!(("existing", Int32, [1, 2, 3])).unwrap();
-        let schema = batch.schema();
 
         let table = conn
-            .create_table(
-                "test_drop_nonexistent",
-                RecordBatchIterator::new(vec![Ok(batch)], schema),
-            )
+            .create_table("test_drop_nonexistent", batch)
             .execute()
             .await
             .unwrap();
@@ -593,13 +546,9 @@ mod tests {
         let conn = connect("memory://").execute().await.unwrap();
 
         let batch = record_batch!(("existing", Int32, [1, 2, 3])).unwrap();
-        let schema = batch.schema();
 
         let table = conn
-            .create_table(
-                "test_alter_nonexistent",
-                RecordBatchIterator::new(vec![Ok(batch)], schema),
-            )
+            .create_table("test_alter_nonexistent", batch)
             .execute()
             .await
             .unwrap();
@@ -623,13 +572,8 @@ mod tests {
         let conn = connect("memory://").execute().await.unwrap();
 
         let batch = record_batch!(("a", Int32, [1, 2, 3]), ("b", Int32, [4, 5, 6])).unwrap();
-        let schema = batch.schema();
-
         let table = conn
-            .create_table(
-                "test_version_increment",
-                RecordBatchIterator::new(vec![Ok(batch)], schema),
-            )
+            .create_table("test_version_increment", batch)
             .execute()
             .await
             .unwrap();

rust/lancedb/src/table/update.rs

@@ -117,9 +117,8 @@ mod tests {
     use crate::query::{ExecutableQuery, Select};
     use arrow_array::{
         record_batch, Array, BooleanArray, Date32Array, FixedSizeListArray, Float32Array,
-        Float64Array, Int32Array, Int64Array, LargeStringArray, RecordBatch, RecordBatchIterator,
-        RecordBatchReader, StringArray, TimestampMillisecondArray, TimestampNanosecondArray,
-        UInt32Array,
+        Float64Array, Int32Array, Int64Array, LargeStringArray, RecordBatch, StringArray,
+        TimestampMillisecondArray, TimestampNanosecondArray, UInt32Array,
     };
     use arrow_data::ArrayDataBuilder;
     use arrow_schema::{ArrowError, DataType, Field, Schema, TimeUnit};
@@ -167,51 +166,46 @@ mod tests {
             ),
         ]));
 
-        let record_batch_iter = RecordBatchIterator::new(
-            vec![RecordBatch::try_new(
-                schema.clone(),
-                vec![
-                    Arc::new(Int32Array::from_iter_values(0..10)),
-                    Arc::new(Int64Array::from_iter_values(0..10)),
-                    Arc::new(UInt32Array::from_iter_values(0..10)),
-                    Arc::new(StringArray::from_iter_values(vec![
-                        "a", "b", "c", "d", "e", "f", "g", "h", "i", "j",
-                    ])),
-                    Arc::new(LargeStringArray::from_iter_values(vec![
-                        "a", "b", "c", "d", "e", "f", "g", "h", "i", "j",
-                    ])),
-                    Arc::new(Float32Array::from_iter_values((0..10).map(|i| i as f32))),
-                    Arc::new(Float64Array::from_iter_values((0..10).map(|i| i as f64))),
-                    Arc::new(Into::<BooleanArray>::into(vec![
-                        true, false, true, false, true, false, true, false, true, false,
-                    ])),
-                    Arc::new(Date32Array::from_iter_values(0..10)),
-                    Arc::new(TimestampNanosecondArray::from_iter_values(0..10)),
-                    Arc::new(TimestampMillisecondArray::from_iter_values(0..10)),
-                    Arc::new(
-                        create_fixed_size_list(
-                            Float32Array::from_iter_values((0..20).map(|i| i as f32)),
-                            2,
-                        )
-                        .unwrap(),
-                    ),
-                    Arc::new(
-                        create_fixed_size_list(
-                            Float64Array::from_iter_values((0..20).map(|i| i as f64)),
-                            2,
-                        )
-                        .unwrap(),
-                    ),
-                ],
-            )
-            .unwrap()]
-            .into_iter()
-            .map(Ok),
+        let batch = RecordBatch::try_new(
             schema.clone(),
-        );
+            vec![
+                Arc::new(Int32Array::from_iter_values(0..10)),
+                Arc::new(Int64Array::from_iter_values(0..10)),
+                Arc::new(UInt32Array::from_iter_values(0..10)),
+                Arc::new(StringArray::from_iter_values(vec![
+                    "a", "b", "c", "d", "e", "f", "g", "h", "i", "j",
+                ])),
+                Arc::new(LargeStringArray::from_iter_values(vec![
+                    "a", "b", "c", "d", "e", "f", "g", "h", "i", "j",
+                ])),
+                Arc::new(Float32Array::from_iter_values((0..10).map(|i| i as f32))),
+                Arc::new(Float64Array::from_iter_values((0..10).map(|i| i as f64))),
+                Arc::new(Into::<BooleanArray>::into(vec![
+                    true, false, true, false, true, false, true, false, true, false,
+                ])),
+                Arc::new(Date32Array::from_iter_values(0..10)),
+                Arc::new(TimestampNanosecondArray::from_iter_values(0..10)),
+                Arc::new(TimestampMillisecondArray::from_iter_values(0..10)),
+                Arc::new(
+                    create_fixed_size_list(
+                        Float32Array::from_iter_values((0..20).map(|i| i as f32)),
+                        2,
+                    )
+                    .unwrap(),
+                ),
+                Arc::new(
+                    create_fixed_size_list(
+                        Float64Array::from_iter_values((0..20).map(|i| i as f64)),
+                        2,
+                    )
+                    .unwrap(),
+                ),
+            ],
+        )
+        .unwrap();
 
         let table = conn
-            .create_table("my_table", record_batch_iter)
+            .create_table("my_table", batch)
             .execute()
             .await
             .unwrap();
@@ -338,15 +332,13 @@ mod tests {
         Ok(FixedSizeListArray::from(data))
     }
 
-    fn make_test_batches() -> impl RecordBatchReader + Send + Sync + 'static {
+    fn make_test_batch() -> RecordBatch {
         let schema = Arc::new(Schema::new(vec![Field::new("i", DataType::Int32, false)]));
-        RecordBatchIterator::new(
-            vec![RecordBatch::try_new(
-                schema.clone(),
-                vec![Arc::new(Int32Array::from_iter_values(0..10))],
-            )],
-            schema,
+        RecordBatch::try_new(
+            schema.clone(),
+            vec![Arc::new(Int32Array::from_iter_values(0..10))],
         )
+        .unwrap()
     }
 
     #[tokio::test]
@@ -367,12 +359,8 @@ mod tests {
         )
         .unwrap();
 
-        let schema = batch.schema();
-        // need the iterator for create table
-        let record_batch_iter = RecordBatchIterator::new(vec![Ok(batch)], schema);
-
         let table = conn
-            .create_table("my_table", record_batch_iter)
+            .create_table("my_table", batch)
             .execute()
             .await
             .unwrap();
@@ -430,7 +418,7 @@ mod tests {
             .await
             .unwrap();
         let tbl = conn
-            .create_table("my_table", make_test_batches())
+            .create_table("my_table", make_test_batch())
             .execute()
             .await
             .unwrap();

rust/lancedb/src/test_utils.rs

@@ -3,3 +3,4 @@
 
 pub mod connection;
 pub mod datagen;
+pub mod embeddings;

rust/lancedb/src/test_utils/datagen.rs

@@ -34,10 +34,7 @@ impl LanceDbDatagenExt for BatchGeneratorBuilder {
             schema,
         ));
         let db = connect("memory:///").execute().await.unwrap();
-        db.create_table_streaming(table_name, stream)
-            .execute()
-            .await
-            .unwrap()
+        db.create_table(table_name, stream).execute().await.unwrap()
     }
 }
 
@@ -48,8 +45,5 @@ pub async fn virtual_table(name: &str, values: &RecordBatch) -> Table {
         schema,
     ));
     let db = connect("memory:///").execute().await.unwrap();
-    db.create_table_streaming(name, stream)
-        .execute()
-        .await
-        .unwrap()
+    db.create_table(name, stream).execute().await.unwrap()
 }

rust/lancedb/src/test_utils/embeddings.rs

@@ -0,0 +1,59 @@
+// SPDX-License-Identifier: Apache-2.0
+// SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+use std::{borrow::Cow, sync::Arc};
+
+use arrow_array::{Array, FixedSizeListArray, Float32Array};
+use arrow_schema::{DataType, Field};
+
+use crate::embeddings::EmbeddingFunction;
+use crate::Result;
+
+#[derive(Debug, Clone)]
+pub struct MockEmbed {
+    name: String,
+    dim: usize,
+}
+
+impl MockEmbed {
+    pub fn new(name: impl Into<String>, dim: usize) -> Self {
+        Self {
+            name: name.into(),
+            dim,
+        }
+    }
+}
+
+impl EmbeddingFunction for MockEmbed {
+    fn name(&self) -> &str {
+        &self.name
+    }
+
+    fn source_type(&self) -> Result<Cow<'_, DataType>> {
+        Ok(Cow::Borrowed(&DataType::Utf8))
+    }
+
+    fn dest_type(&self) -> Result<Cow<'_, DataType>> {
+        Ok(Cow::Owned(DataType::new_fixed_size_list(
+            DataType::Float32,
+            self.dim as _,
+            true,
+        )))
+    }
+
+    fn compute_source_embeddings(&self, source: Arc<dyn Array>) -> Result<Arc<dyn Array>> {
+        // We can't use the FixedSizeListBuilder here because it always adds a null bitmap
+        // and we want to explicitly work with non-nullable arrays.
+        let len = source.len();
+        let inner = Arc::new(Float32Array::from(vec![Some(1.0); len * self.dim]));
+        let field = Field::new("item", inner.data_type().clone(), false);
+        let arr = FixedSizeListArray::new(Arc::new(field), self.dim as _, inner, None);
+
+        Ok(Arc::new(arr))
+    }
+
+    #[allow(unused_variables)]
+    fn compute_query_embeddings(&self, input: Arc<dyn Array>) -> Result<Arc<dyn Array>> {
+        todo!()
+    }
+}

rust/lancedb/tests/embedding_registry_test.rs

@@ -15,7 +15,6 @@ use arrow_array::{
 use arrow_schema::{DataType, Field, Schema};
 use futures::StreamExt;
 use lancedb::{
-    arrow::IntoArrow,
     connect,
     embeddings::{EmbeddingDefinition, EmbeddingFunction, EmbeddingRegistry},
     query::ExecutableQuery,
@@ -253,7 +252,7 @@ async fn test_no_func_in_registry_on_add() -> Result<()> {
     Ok(())
 }
 
-fn create_some_records() -> Result<impl IntoArrow> {
+fn create_some_records() -> Result<Box<dyn arrow_array::RecordBatchReader + Send>> {
     const TOTAL: usize = 2;
 
     let schema = Arc::new(Schema::new(vec![

rust/lancedb/tests/object_store_test.rs

@@ -4,7 +4,7 @@
 #![cfg(feature = "s3-test")]
 use std::sync::Arc;
 
-use arrow_array::{Int32Array, RecordBatch, RecordBatchIterator, StringArray};
+use arrow_array::{Int32Array, RecordBatch, StringArray};
 use arrow_schema::{DataType, Field, Schema};
 
 use aws_config::{BehaviorVersion, ConfigLoader, Region, SdkConfig};
@@ -111,7 +111,6 @@ async fn test_minio_lifecycle() -> Result<()> {
         .await?;
 
     let data = test_data();
-    let data = RecordBatchIterator::new(vec![Ok(data.clone())], data.schema());
 
     let table = db.create_table("test_table", data).execute().await?;
 
@@ -127,7 +126,6 @@ async fn test_minio_lifecycle() -> Result<()> {
     assert_eq!(row_count, 3);
 
     let data = test_data();
-    let data = RecordBatchIterator::new(vec![Ok(data.clone())], data.schema());
     table.add(data).execute().await?;
 
     db.drop_table("test_table", &[]).await?;
@@ -247,7 +245,6 @@ async fn test_encryption() -> Result<()> {
 
     // Create a table with encryption
     let data = test_data();
-    let data = RecordBatchIterator::new(vec![Ok(data.clone())], data.schema());
 
     let mut builder = db.create_table("test_table", data);
     for (key, value) in CONFIG {
@@ -274,7 +271,6 @@ async fn test_encryption() -> Result<()> {
     let table = db.open_table("test_table").execute().await?;
 
     let data = test_data();
-    let data = RecordBatchIterator::new(vec![Ok(data.clone())], data.schema());
     table.add(data).execute().await?;
     validate_objects_encrypted(&bucket.0, "test_table", &key.0).await;
 
@@ -300,7 +296,6 @@ async fn test_table_storage_options_override() -> Result<()> {
 
     // Create table overriding with key2 encryption
     let data = test_data();
-    let data = RecordBatchIterator::new(vec![Ok(data.clone())], data.schema());
     let _table = db
         .create_table("test_override", data)
         .storage_option("aws_sse_kms_key_id", &key2.0)
@@ -312,7 +307,6 @@ async fn test_table_storage_options_override() -> Result<()> {
 
     // Also test that a table created without override uses connection settings
     let data = test_data();
-    let data = RecordBatchIterator::new(vec![Ok(data.clone())], data.schema());
     let _table2 = db.create_table("test_inherit", data).execute().await?;
 
     // Verify this table uses key1 from connection
@@ -419,7 +413,6 @@ async fn test_concurrent_dynamodb_commit() {
         .unwrap();
 
     let data = test_data();
-    let data = RecordBatchIterator::new(vec![Ok(data.clone())], data.schema());
 
     let table = db.create_table("test_table", data).execute().await.unwrap();
 
@@ -430,7 +423,6 @@ async fn test_concurrent_dynamodb_commit() {
         let table = db.open_table("test_table").execute().await.unwrap();
         let data = data.clone();
         tasks.push(tokio::spawn(async move {
-            let data = RecordBatchIterator::new(vec![Ok(data.clone())], data.schema());
             table.add(data).execute().await.unwrap();
         }));
     }