Merge remote-tracking branch 'origin/main' into gatekeeper/fix-2325-1

# Conflicts: # python/python/lancedb/remote/table.py
Merge branch 'main' into gatekeeper/fix-2325-1
2026-09-04 04:28:44 +00:00 · 2026-08-28 09:45:02 +00:00 · 2026-08-27 18:13:07 +08:00 · 2026-08-22 07:07:55 +00:00 · 2026-08-22 07:07:52 +00:00 · 2026-08-21 09:23:38 +00:00
51 changed files with 11861 additions and 1060 deletions
@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.38.0-beta.14"
+current_version = "0.38.0-beta.12"
 parse = """(?x)
    (?P<major>0|[1-9]\\d*)\\.
    (?P<minor>0|[1-9]\\d*)\\.
@@ -44,27 +44,3 @@ updates:
      python-deps:
        patterns:
          - "*"
-
-  # The npm ecosystem covers pnpm lockfiles. There are two separate installs:
-  # the bindings themselves and the examples, which have their own lockfile.
-  # As with cargo and pip above, only bump the lockfile — the version ranges
-  # in package.json are our consumers' constraints, not ours.
-  - package-ecosystem: npm
-    directory: /nodejs
-    schedule:
-      interval: weekly
-    versioning-strategy: lockfile-only
-    groups:
-      nodejs-deps:
-        patterns:
-          - "*"
-
-  - package-ecosystem: npm
-    directory: /nodejs/examples
-    schedule:
-      interval: weekly
-    versioning-strategy: lockfile-only
-    groups:
-      nodejs-examples-deps:
-        patterns:
-          - "*"
@@ -29,14 +29,12 @@ jobs:
    steps:
      - uses: actions/setup-node@v6
        with:
-          node-version: "24"
-      - uses: pnpm/action-setup@v6
-        with:
-          version: 11.1.1
+          node-version: "18"
      # These rules are disabled because Github will always ensure there
      # is a blank line between the title and the body and Github will
      # word wrap the description field to ensure a reasonable max line
      # length.
+      - run: npm install @commitlint/config-conventional
      - run: >
          echo 'module.exports = {
            "rules": {
@@ -45,11 +43,7 @@ jobs:
              "body-leading-blank": [0, "always"]
            }
          }' > .commitlintrc.js
-      - run: >
-          pnpm dlx
-          --package @commitlint/cli@21.2.2
-          --package @commitlint/config-conventional@21.2.2
-          commitlint --extends @commitlint/config-conventional --verbose <<< $COMMIT_MSG
+      - run: npx commitlint --extends @commitlint/config-conventional --verbose <<< $COMMIT_MSG
        env:
          COMMIT_MSG: >
            ${{ github.event.pull_request.title }}
@@ -60,7 +54,7 @@ jobs:
        with:
          script: |
            const message = `**ACTION NEEDED**
-
+              
              Lance follows the [Conventional Commits specification](https://www.conventionalcommits.org/en/v1.0.0/) for release automation.

              The PR title and description are used as the merge commit message.\
@@ -56,7 +56,7 @@ jobs:
        uses: lycheeverse/lychee-action@e7477775783ea5526144ba13e8db5eec57747ce8 # v2.9.0
        with:
          # Restricted to http(s) on purpose. Much of docs/src is generated
-          # API reference (the js/ tree comes from `pnpm run docs` in nodejs)
+          # API reference (the js/ tree comes from `npm run docs` in nodejs)
          # and the hand-written pages use mkdocstrings cross-references and
          # nav-relative paths that only resolve in the site mkdocs builds,
          # not in this checkout, so relative links would be reported as
@@ -55,7 +55,9 @@ jobs:
      - name: Set up node
        uses: actions/setup-node@v6
        with:
-          node-version: 24
+          node-version: 20
+          cache: 'npm'
+          cache-dependency-path: docs/package-lock.json
      - name: Install node dependencies
        working-directory: nodejs
        run: |
@@ -47,8 +47,9 @@ jobs:
        version: 11.1.1
    - uses: actions/setup-node@v6
      with:
-        # Build on a supported LTS; the matrix job below covers every
-        # Node version the library claims to support.
+        # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
+        # in October. The library itself still supports Node >= 18
+        # (see test matrix below).
        node-version: 24
        cache: 'pnpm'
        cache-dependency-path: nodejs/pnpm-lock.yaml
@@ -83,7 +84,7 @@ jobs:
    timeout-minutes: 30
    strategy:
      matrix:
-        node-version: [ "22", "24", "26" ]
+        node-version: [ "18", "20" ]
    runs-on: "ubuntu-22.04"
    defaults:
      run:
@@ -100,9 +101,9 @@ jobs:
    - uses: actions/setup-node@v6
      name: Setup Node.js 24 for build
      with:
-        # Build and install once on a fixed version so the generated docs
-        # are identical across matrix legs; the tests below then run on each
-        # supported Node version.
+        # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
+        # in October. Build/install runs on Node 24; tests run on the
+        # matrix version below using direct jest invocation.
        node-version: 24
        cache: 'pnpm'
        cache-dependency-path: nodejs/pnpm-lock.yaml
@@ -151,9 +152,9 @@ jobs:
        S3_TEST: "1"
        # Newer @smithy/core uses dynamic ESM imports.
        NODE_OPTIONS: "--experimental-vm-modules"
-      # Invoke the installed jest binary directly; the pnpm shim is set up
-      # against the build-phase Node, not the version selected above.
-      run: node_modules/.bin/jest --verbose
+      # Invoke jest directly because pnpm 11 itself requires Node 22+
+      # while the matrix tests on older Node versions.
+      run: npx jest --verbose
    - name: Test examples
      working-directory: ./
      env:
@@ -163,7 +164,7 @@ jobs:
      run: |
        python ci/mock_openai.py &
        cd nodejs/examples
-        node_modules/.bin/jest --testEnvironment jest-environment-node-single-context --verbose
+        npx jest --testEnvironment jest-environment-node-single-context --verbose
  macos:
    timeout-minutes: 30
    # macos-15 ships a newer linker; the older macos-14 linker fails to insert
@@ -184,7 +185,8 @@ jobs:
        version: 11.1.1
    - uses: actions/setup-node@v6
      with:
-        # pnpm 11 requires Node >= 22.13.
+        # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
+        # in October.
        node-version: 24
        cache: 'pnpm'
        cache-dependency-path: nodejs/pnpm-lock.yaml
@@ -168,7 +168,8 @@ jobs:
      - name: Setup node
        uses: actions/setup-node@v6
        with:
-          # pnpm 11 requires Node >= 22.13.
+          # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
+          # in October.
          node-version: 24
          cache: pnpm
          cache-dependency-path: nodejs/pnpm-lock.yaml
@@ -250,7 +251,7 @@ jobs:
          run: |
            set -e
            ${{ matrix.settings.pre_build }}
-            node_modules/.bin/napi build --platform --release \
+            npx napi build --platform --release \
              --features ${{ matrix.settings.features }} \
              --target ${{ matrix.settings.target }} \
              --dts ../lancedb/native.d.ts \
@@ -270,7 +271,7 @@ jobs:
      - name: Build
        run: |
          ${{ matrix.settings.pre_build }}
-          node_modules/.bin/napi build --platform --release \
+          npx napi build --platform --release \
              --features ${{ matrix.settings.features }} \
              --target ${{ matrix.settings.target }} \
              --dts ../lancedb/native.d.ts \
@@ -338,7 +339,7 @@ jobs:
          - target: aarch64-unknown-linux-gnu
            host: ubuntu-2404-8x-arm64
        node:
-          - '22'
+          - '20'
    runs-on: ${{ matrix.settings.host }}
    defaults:
      run:
@@ -384,9 +385,9 @@ jobs:
      - name: Move built files
        run: cp dist/native.d.ts dist/native.js dist/*.node lancedb/
      - name: Test bindings
-        # Invoke the installed jest binary directly; the pnpm shim is set up
-        # against the install-phase Node, not the version selected above.
-        run: node_modules/.bin/jest --verbose
+        # Invoke jest directly because pnpm 11 itself requires Node 22+
+        # while the matrix tests on older Node versions.
+        run: npx jest --verbose
  publish:
    name: Publish
    runs-on: ubuntu-latest
@@ -0,0 +1,22 @@
+name: Update package-lock.json
+
+on:
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v6
+        with:
+          ref: main
+          persist-credentials: false
+          fetch-depth: 0
+          lfs: true
+      - uses: ./.github/workflows/update_package_lock
+        with:
+          github_token: ${{ secrets.LANCEDB_RELEASE_TOKEN }}
@@ -0,0 +1,22 @@
+name: Update NodeJs package-lock.json
+
+on:
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v6
+        with:
+          ref: main
+          persist-credentials: false
+          fetch-depth: 0
+          lfs: true
+      - uses: ./.github/workflows/update_package_lock_nodejs
+        with:
+          github_token: ${{ secrets.LANCEDB_RELEASE_TOKEN }}
@@ -20,10 +20,7 @@ repos:
    hooks:
      - id: local-biome-check
        name: biome check
-        # Use the biome from nodejs/package.json rather than a separately
-        # pinned one: the two drifted apart and disagreed on formatting, so
-        # this hook rejected code that `pnpm lint` accepted.
-        entry: nodejs/node_modules/.bin/biome check --config-path nodejs/biome.json nodejs/
+        entry: npx @biomejs/biome@1.8.3 check --config-path nodejs/biome.json nodejs/
        language: system
        types: [text]
        files: "nodejs/.*"
@@ -38,7 +38,7 @@ Before committing changes, run formatting for every language you touched. At min
 * Rust changes: run `cargo fmt --all`.
 * Python changes: run `ruff format .` and `ruff check .` from the repository root,
  and run targeted tests through `cd python && uv run ...`.
-* TypeScript changes: run the relevant `pnpm` lint, format, build, and docs commands in `nodejs`.
+* TypeScript changes: run the relevant `npm`/`pnpm` lint, format, build, and docs commands in `nodejs`.

 Before creating a PR, the exact value passed to `gh pr create --title` must follow
 Conventional Commits, such as `fix: support nested field paths in native index creation`
@@ -101,12 +101,12 @@ Python bindings changes:
 TypeScript bindings changes:

 1. Add napi-rs method binding on `Table` in `nodejs/src/table.rs`.
-2. Run `pnpm build` to generate TypeScript definitions.
+2. Run `npm run build` to generate TypeScript definitions.
 3. Add typescript method on abstract class `Table` in `nodejs/src/table.ts`.
 4. Add concrete method on `LocalTable` class in `nodejs/src/native_table.ts`.
    * Note: despite the name, this class is also used for remote tables.
 5. Add test in `nodejs/__test__/table.test.ts`.
-6. Run `pnpm run docs` to generate TypeScript documentation.
+6. Run `npm run docs` to generate TypeScript documentation.

 ## Python API reference

@@ -3455,8 +3455,8 @@ checksum = "42703706b716c37f96a77aea830392ad231f44c9e9a67872fa5548707e11b11c"

 [[package]]
 name = "fsst"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow-array",
 "rand 0.9.5",
@@ -4815,8 +4815,8 @@ checksum = "e037a2e1d8d5fdbd49b16a4ea09d5d6401c1f29eca5ff29d03d3824dba16256a"

 [[package]]
 name = "lance"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arc-swap",
 "arrow",
@@ -4888,8 +4888,8 @@ dependencies = [

 [[package]]
 name = "lance-arrow"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow-array",
 "arrow-buffer",
@@ -4911,7 +4911,7 @@ dependencies = [
 [[package]]
 name = "lance-arrow-scalar"
 version = "58.0.0"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow-array",
 "arrow-buffer",
@@ -4925,7 +4925,7 @@ dependencies = [
 [[package]]
 name = "lance-arrow-stats"
 version = "58.0.0"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow-array",
 "arrow-schema",
@@ -4934,8 +4934,8 @@ dependencies = [

 [[package]]
 name = "lance-bitpacking"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrayref",
 "crunchy",
@@ -4945,8 +4945,8 @@ dependencies = [

 [[package]]
 name = "lance-core"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow-array",
 "arrow-buffer",
@@ -4983,8 +4983,8 @@ dependencies = [

 [[package]]
 name = "lance-datafusion"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow",
 "arrow-array",
@@ -5013,8 +5013,8 @@ dependencies = [

 [[package]]
 name = "lance-datagen"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow",
 "arrow-array",
@@ -5031,8 +5031,8 @@ dependencies = [

 [[package]]
 name = "lance-derive"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "proc-macro2",
 "quote",
@@ -5041,8 +5041,8 @@ dependencies = [

 [[package]]
 name = "lance-encoding"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow-arith",
 "arrow-array",
@@ -5075,8 +5075,8 @@ dependencies = [

 [[package]]
 name = "lance-file"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow-arith",
 "arrow-array",
@@ -5107,8 +5107,8 @@ dependencies = [

 [[package]]
 name = "lance-index"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arc-swap",
 "arrow",
@@ -5172,8 +5172,8 @@ dependencies = [

 [[package]]
 name = "lance-index-core"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow-array",
 "arrow-schema",
@@ -5195,8 +5195,8 @@ dependencies = [

 [[package]]
 name = "lance-io"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow",
 "arrow-array",
@@ -5236,8 +5236,8 @@ dependencies = [

 [[package]]
 name = "lance-linalg"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow-array",
 "arrow-schema",
@@ -5251,8 +5251,8 @@ dependencies = [

 [[package]]
 name = "lance-namespace"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow",
 "async-trait",
@@ -5264,8 +5264,8 @@ dependencies = [

 [[package]]
 name = "lance-namespace-impls"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow",
 "arrow-ipc",
@@ -5318,8 +5318,8 @@ dependencies = [

 [[package]]
 name = "lance-select"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow-array",
 "arrow-buffer",
@@ -5333,8 +5333,8 @@ dependencies = [

 [[package]]
 name = "lance-table"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow",
 "arrow-array",
@@ -5374,8 +5374,8 @@ dependencies = [

 [[package]]
 name = "lance-testing"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "arrow-array",
 "arrow-schema",
@@ -5388,8 +5388,8 @@ dependencies = [

 [[package]]
 name = "lance-tokenizer"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
 "frostem",
 "icu_segmenter",
@@ -5402,7 +5402,7 @@ dependencies = [

 [[package]]
 name = "lancedb"
-version = "0.38.0-beta.14"
+version = "0.38.0-beta.12"
 dependencies = [
 "ahash",
 "anyhow",
@@ -5490,7 +5490,7 @@ dependencies = [

 [[package]]
 name = "lancedb-nodejs"
-version = "0.38.0-beta.14"
+version = "0.38.0-beta.12"
 dependencies = [
 "arrow-array",
 "arrow-buffer",
@@ -5515,7 +5515,7 @@ dependencies = [

 [[package]]
 name = "lancedb-python"
-version = "0.38.0-beta.14"
+version = "0.38.0-beta.12"
 dependencies = [
 "arrow",
 "async-trait",
@@ -13,20 +13,20 @@ categories = ["database-implementations"]
 rust-version = "1.91.0"

 [workspace.dependencies]
-lance = { "version" = "=12.0.0-beta.5", default-features = false, "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-core = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-datagen = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-file = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-io = { "version" = "=12.0.0-beta.5", default-features = false, "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-index = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-linalg = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-namespace = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-namespace-impls = { "version" = "=12.0.0-beta.5", default-features = false, "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-table = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-testing = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-datafusion = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-encoding = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-arrow = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
+lance = { "version" = "=12.0.0-beta.2", default-features = false, "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-core = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-datagen = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-file = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-io = { "version" = "=12.0.0-beta.2", default-features = false, "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-index = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-linalg = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-namespace = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-namespace-impls = { "version" = "=12.0.0-beta.2", default-features = false, "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-table = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-testing = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-datafusion = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-encoding = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-arrow = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
 lancedb = { path = "rust/lancedb", default-features = false }
 ahash = "0.8"
 # Note that this one does not include pyarrow
@@ -5,5 +5,5 @@ licenses:
 	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 && pnpm dlx license-checker@25 --markdown --out NODEJS_THIRD_PARTY_LICENSES.md
+	cd nodejs && npx license-checker --markdown --out NODEJS_THIRD_PARTY_LICENSES.md
 	cd java && ./mvnw license:aggregate-add-third-party -q
@@ -12,12 +12,16 @@ done
 # This updates the lockfile without building
 cargo metadata --quiet > /dev/null

+pushd nodejs || exit 1
+npm install --package-lock-only --silent
+popd
+
 if git diff --quiet --exit-code; then
  echo "No lockfile changes to commit; skipping amend."
 elif $AMEND; then
-  git add Cargo.lock
+  git add Cargo.lock nodejs/package-lock.json
  git commit --amend --no-edit
 else
-  git add Cargo.lock
+  git add Cargo.lock nodejs/package-lock.json
  git commit -m "Update lockfiles"
 fi
@@ -47,24 +47,22 @@ pytest -vv python/tests/docs

 ### Checking typescript examples

-The examples depend on `@lancedb/lancedb` at `file:../dist`, so the package must be
-built before running the tests. This uses pnpm; see the
-[Typescript contributing guide](../nodejs/CONTRIBUTING.md) for the toolchain setup.
+The `@lancedb/lancedb` package must be built before running the tests:

 ```shell
 pushd nodejs
-pnpm install
-pnpm build
+npm ci
+npm run build
 popd
 ```

-Then you can run the examples by going to the `nodejs/examples` directory, which is a
-separate pnpm package with its own lockfile:
+Then you can run the examples by going to the `nodejs/examples` directory and
+running the tests like a normal npm package:

 ```shell
 pushd nodejs/examples
-pnpm install
-pnpm test
+npm ci
+npm test
 popd
 ```

@@ -86,7 +84,6 @@ The new files should be checked into the repository.

 ```shell
 pushd nodejs
-# `pnpm docs` would invoke pnpm's built-in `docs` command, not the script.
-pnpm run docs
+npm run docs
 popd
 ```
@@ -0,0 +1,135 @@
+{
+  "name": "lancedb-docs-test",
+  "version": "1.0.0",
+  "lockfileVersion": 3,
+  "requires": true,
+  "packages": {
+    "": {
+      "name": "lancedb-docs-test",
+      "version": "1.0.0",
+      "license": "Apache 2",
+      "dependencies": {
+        "apache-arrow": "file:../node/node_modules/apache-arrow",
+        "vectordb": "file:../node"
+      },
+      "devDependencies": {
+        "@types/node": "^20.11.8",
+        "typescript": "^5.3.3"
+      }
+    },
+    "../node": {
+      "name": "vectordb",
+      "version": "0.21.2-beta.0",
+      "cpu": [
+        "x64",
+        "arm64"
+      ],
+      "license": "Apache-2.0",
+      "os": [
+        "darwin",
+        "linux",
+        "win32"
+      ],
+      "dependencies": {
+        "@neon-rs/load": "^0.0.74",
+        "axios": "^1.4.0"
+      },
+      "devDependencies": {
+        "@neon-rs/cli": "^0.0.160",
+        "@types/chai": "^4.3.4",
+        "@types/chai-as-promised": "^7.1.5",
+        "@types/mocha": "^10.0.1",
+        "@types/node": "^18.16.2",
+        "@types/sinon": "^10.0.15",
+        "@types/temp": "^0.9.1",
+        "@types/uuid": "^9.0.3",
+        "@typescript-eslint/eslint-plugin": "^5.59.1",
+        "apache-arrow-old": "npm:apache-arrow@13.0.0",
+        "cargo-cp-artifact": "^0.1",
+        "chai": "^4.3.7",
+        "chai-as-promised": "^7.1.1",
+        "eslint": "^8.39.0",
+        "eslint-config-standard-with-typescript": "^34.0.1",
+        "eslint-plugin-import": "^2.26.0",
+        "eslint-plugin-n": "^15.7.0",
+        "eslint-plugin-promise": "^6.1.1",
+        "mocha": "^10.2.0",
+        "openai": "^4.24.1",
+        "sinon": "^15.1.0",
+        "temp": "^0.9.4",
+        "ts-node": "^10.9.1",
+        "ts-node-dev": "^2.0.0",
+        "typedoc": "^0.24.7",
+        "typedoc-plugin-markdown": "^3.15.3",
+        "typescript": "^5.1.0",
+        "uuid": "^9.0.0"
+      },
+      "optionalDependencies": {
+        "@lancedb/vectordb-darwin-arm64": "0.21.2-beta.0",
+        "@lancedb/vectordb-darwin-x64": "0.21.2-beta.0",
+        "@lancedb/vectordb-linux-arm64-gnu": "0.21.2-beta.0",
+        "@lancedb/vectordb-linux-x64-gnu": "0.21.2-beta.0",
+        "@lancedb/vectordb-win32-x64-msvc": "0.21.2-beta.0"
+      },
+      "peerDependencies": {
+        "@apache-arrow/ts": "^14.0.2",
+        "apache-arrow": "^14.0.2"
+      }
+    },
+    "../node/node_modules/apache-arrow": {
+      "version": "14.0.2",
+      "license": "Apache-2.0",
+      "dependencies": {
+        "@types/command-line-args": "5.2.0",
+        "@types/command-line-usage": "5.0.2",
+        "@types/node": "20.3.0",
+        "@types/pad-left": "2.1.1",
+        "command-line-args": "5.2.1",
+        "command-line-usage": "7.0.1",
+        "flatbuffers": "23.5.26",
+        "json-bignum": "^0.0.3",
+        "pad-left": "^2.1.0",
+        "tslib": "^2.5.3"
+      },
+      "bin": {
+        "arrow2csv": "bin/arrow2csv.js"
+      }
+    },
+    "node_modules/@types/node": {
+      "version": "20.11.8",
+      "resolved": "https://registry.npmjs.org/@types/node/-/node-20.11.8.tgz",
+      "integrity": "sha512-i7omyekpPTNdv4Jb/Rgqg0RU8YqLcNsI12quKSDkRXNfx7Wxdm6HhK1awT3xTgEkgxPn3bvnSpiEAc7a7Lpyow==",
+      "dev": true,
+      "dependencies": {
+        "undici-types": "~5.26.4"
+      }
+    },
+    "node_modules/apache-arrow": {
+      "resolved": "../node/node_modules/apache-arrow",
+      "link": true
+    },
+    "node_modules/typescript": {
+      "version": "5.3.3",
+      "resolved": "https://registry.npmjs.org/typescript/-/typescript-5.3.3.tgz",
+      "integrity": "sha512-pXWcraxM0uxAS+tN0AG/BF2TyqmHO014Z070UsJ+pFvYuRSq8KH8DmWpnbXe0pEPDHXZV3FcAbJkijJ5oNEnWw==",
+      "dev": true,
+      "bin": {
+        "tsc": "bin/tsc",
+        "tsserver": "bin/tsserver"
+      },
+      "engines": {
+        "node": ">=14.17"
+      }
+    },
+    "node_modules/undici-types": {
+      "version": "5.26.5",
+      "resolved": "https://registry.npmjs.org/undici-types/-/undici-types-5.26.5.tgz",
+      "integrity": "sha512-JlCMO+ehdEIKqlFxk6IfVoAUVmgz7cU7zD/h9XZ0qzeosSHmUJVOzSQvvYSYWXkFXC+IfLKSIffhv0sVZup6pA==",
+      "dev": true
+    },
+    "node_modules/vectordb": {
+      "resolved": "../node",
+      "link": true
+    }
+  }
+}
@@ -0,0 +1,20 @@
+{
+  "name": "lancedb-docs-test",
+  "version": "1.0.0",
+  "description": "auto-generated tests from doc",
+  "author": "dev@lancedb.com",
+  "license": "Apache 2",
+  "dependencies": {
+    "apache-arrow": "file:../node/node_modules/apache-arrow",
+    "vectordb": "file:../node"
+  },
+  "scripts": {
+    "build": "tsc -b && cd ../node && npm run build-release",
+    "example": "npm run build && node",
+    "test": "npm run build && ls dist/*.js | xargs -n 1 node"
+  },
+  "devDependencies": {
+    "@types/node": "^20.11.8",
+    "typescript": "^5.3.3"
+  }
+}
@@ -14,7 +14,7 @@ Add the following dependency to your `pom.xml`:
 <dependency>
    <groupId>com.lancedb</groupId>
    <artifactId>lancedb-core</artifactId>
-    <version>0.38.0-beta.14</version>
+    <version>0.38.0-beta.12</version>
 </dependency>
 ```

@@ -42,6 +42,8 @@ listing a storage directory.

 ::: lancedb.table.Table

+::: lancedb.table.CompactionOptions
+
 ::: lancedb.table.FragmentStatistics

 ::: lancedb.table.FragmentSummaryStats
@@ -0,0 +1,17 @@
+{
+  "include": [
+    "src/*.ts",
+  ],
+  "compilerOptions": {
+    "target": "es2022",
+    "module": "nodenext",
+    "declaration": true,
+    "outDir": "./dist",
+    "strict": true,
+    "allowJs": true,
+    "resolveJsonModule": true,
+  },
+  "exclude": [
+    "./dist/*",
+  ]
+}
@@ -8,7 +8,7 @@
    <parent>
      <groupId>com.lancedb</groupId>
      <artifactId>lancedb-parent</artifactId>
-      <version>0.38.0-beta.14</version>
+      <version>0.38.0-beta.12</version>
      <relativePath>../pom.xml</relativePath>
    </parent>

@@ -6,7 +6,7 @@

    <groupId>com.lancedb</groupId>
    <artifactId>lancedb-parent</artifactId>
-    <version>0.38.0-beta.14</version>
+    <version>0.38.0-beta.12</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>12.0.0-beta.5</lance-core.version>
+        <lance-core.version>12.0.0-beta.2</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>
@@ -1,7 +1,7 @@
 [package]
 name = "lancedb-nodejs"
 edition.workspace = true
-version = "0.38.0-beta.14"
+version = "0.38.0-beta.12"
 publish = false
 license.workspace = true
 description.workspace = true
@@ -5,8 +5,8 @@ import packageJson = require("../package.json");

 describe("package metadata", () => {
  it("requires Node.js type declarations compatible with the runtime", () => {
-    expect(packageJson.engines.node).toBe(">= 22");
-    expect(packageJson.peerDependencies["@types/node"]).toBe(">=22");
+    expect(packageJson.engines.node).toBe(">= 18");
+    expect(packageJson.peerDependencies["@types/node"]).toBe(">=18");
    expect(packageJson.peerDependenciesMeta["@types/node"]).toEqual({
      optional: true,
    });
@@ -3,7 +3,6 @@

 import * as http from "http";
 import { RequestListener } from "http";
-import packageJson = require("../package.json");
 import {
  ClientConfig,
  Connection,
@@ -71,13 +70,7 @@ async function withMockDatabase(
  try {
    await callback(db);
  } finally {
-    // `close()` alone leaves the port bound until keep-alive sockets drain, so
-    // a single failing test would cascade into EADDRINUSE for every test after
-    // it. Destroy the connections and wait for the port to actually be free.
-    await new Promise<void>((resolve) => {
-      server.closeAllConnections();
-      server.close(() => resolve());
-    });
+    server.close();
  }
 }

@@ -138,7 +131,7 @@ describe("remote connection", () => {
      (req, res) => {
        expect(req.headers["x-api-key"]).toEqual("fake");
        expect(req.headers["user-agent"]).toEqual(
-          `LanceDB-Node-Client/${packageJson.version}`,
+          `LanceDB-Node-Client/${process.env.npm_package_version}`,
        );

        const body = JSON.stringify({ tables: [] });
@@ -8,8 +8,7 @@
    "//1": "--experimental-vm-modules is needed to run jest with sentence-transformers",
    "//2": "--testEnvironment is needed to run jest with sentence-transformers",
    "//3": "See: https://github.com/huggingface/transformers.js/issues/57",
-    "//4": "jest is invoked by its JS entry, not node_modules/.bin/jest: under pnpm that path is a shell shim, which `node` cannot execute",
-    "test": "node --experimental-vm-modules node_modules/jest/bin/jest.js --testEnvironment jest-environment-node-single-context --verbose",
+    "test": "node --experimental-vm-modules node_modules/.bin/jest --testEnvironment jest-environment-node-single-context --verbose",
    "lint": "biome check *.ts && biome format *.ts",
    "lint-ci": "biome ci .",
    "lint-fix": "biome check --write *.ts && pnpm format",
@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-darwin-arm64",
-	"version": "0.38.0-beta.14",
+	"version": "0.38.0-beta.12",
 	"os": ["darwin"],
 	"cpu": ["arm64"],
 	"main": "lancedb.darwin-arm64.node",
@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-gnu",
-	"version": "0.38.0-beta.14",
+	"version": "0.38.0-beta.12",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-gnu.node",
@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-musl",
-	"version": "0.38.0-beta.14",
+	"version": "0.38.0-beta.12",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-musl.node",
@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-gnu",
-	"version": "0.38.0-beta.14",
+	"version": "0.38.0-beta.12",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-gnu.node",
@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-musl",
-	"version": "0.38.0-beta.14",
+	"version": "0.38.0-beta.12",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-musl.node",
@@ -1,6 +1,6 @@
 {
  "name": "@lancedb/lancedb-win32-arm64-msvc",
-  "version": "0.38.0-beta.14",
+  "version": "0.38.0-beta.12",
  "os": [
    "win32"
  ],
@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-win32-x64-msvc",
-	"version": "0.38.0-beta.14",
+	"version": "0.38.0-beta.12",
 	"os": ["win32"],
 	"cpu": ["x64"],
 	"main": "lancedb.win32-x64-msvc.node",
@@ -11,7 +11,7 @@
    "ann"
  ],
  "private": false,
-  "version": "0.38.0-beta.14",
+  "version": "0.38.0-beta.12",
  "main": "dist/index.js",
  "exports": {
    ".": "./dist/index.js",
@@ -67,7 +67,7 @@
    "timeout": "3m"
  },
  "engines": {
-    "node": ">= 22"
+    "node": ">= 18"
  },
  "packageManager": "pnpm@11.1.1",
  "cpu": ["x64", "arm64"],
@@ -101,7 +101,7 @@
    "openai": "4.29.2"
  },
  "peerDependencies": {
-    "@types/node": ">=22",
+    "@types/node": ">=18",
    "apache-arrow": ">=15.0.0 <=18.1.0"
  },
  "peerDependenciesMeta": {
@@ -1,6 +1,6 @@
 [package]
 name = "lancedb-python"
-version = "0.38.0-beta.14"
+version = "0.38.0-beta.12"
 publish = false
 edition.workspace = true
 description = "Python bindings for LanceDB"
@@ -38,7 +38,7 @@ from .materialized_view import (
    MaterializedView,
    MaterializedViewDefinition,
 )
-from .table import AsyncTable, Table
+from .table import AsyncTable, CompactionOptions, Table
 from .types import BaseTokenizerType
 from ._lancedb import Session
 from .namespace import (
@@ -558,6 +558,7 @@ __all__ = [
    "AsyncJob",
    "AsyncLanceNamespaceDBConnection",
    "AsyncTable",
+    "CompactionOptions",
    "FtsToken",
    "col",
    "Expr",
@@ -374,6 +374,7 @@ class Table:
        *,
        cleanup_since_ms: Optional[int] = None,
        delete_unverified: Optional[bool] = None,
+        compaction_options: Optional[Dict[str, Any]] = None,
    ) -> OptimizeStats: ...
    async def uri(self) -> str: ...
    async def initial_storage_options(self) -> Optional[Dict[str, str]]: ...
@@ -54,18 +54,6 @@ _UInt32 = conint(strict=True, ge=0, le=2**32 - 1)
 _UInt64 = conint(strict=True, ge=0, le=2**64 - 1)


-def _validate_gpu_wire_marker(value: Any) -> bool:
-    if value is not True:
-        raise ValueError("runtime.gpu must be true")
-    return True
-
-
-def _normalize_gpu_marker(value: bool) -> Optional[bool]:
-    if not isinstance(value, bool):
-        raise ValueError("gpu must be a boolean")
-    return True if value else None
-
-
 class _FrozenDict(dict):
    def _immutable(self, *args, **kwargs):
        raise TypeError("remote canonical values are immutable")
@@ -251,23 +239,6 @@ class PythonRuntimeSpec(_RemoteValue):
    python_version: Optional[str] = None
    environment: Optional[PythonEnvironmentSpec] = None
    env: Optional[Mapping[str, str]] = None
-    gpu: Optional[bool] = None
-
-    @model_validator(mode="before")
-    @classmethod
-    def _discard_unknown_runtime_payload(cls, value):
-        if isinstance(value, Mapping):
-            kind = value.get("kind")
-            if isinstance(kind, str) and kind not in {"python", "python_v2"}:
-                return {"kind": kind}
-        return value
-
-    @field_validator("gpu", mode="before")
-    @classmethod
-    def _validate_gpu_marker(cls, value):
-        if value is None:
-            return None
-        return _validate_gpu_wire_marker(value)

    @model_validator(mode="after")
    def _validate_runtime_kind(self):
@@ -276,28 +247,18 @@ class PythonRuntimeSpec(_RemoteValue):
                raise ValueError("python runtime requires python_version")
            if self.environment is None:
                raise ValueError("python runtime requires environment")
-            if self.gpu is not None:
-                raise ValueError("python runtime with gpu requires kind='python_v2'")
-        elif self.kind == "python_v2":
-            if self.python_version is None:
-                raise ValueError("python_v2 runtime requires python_version")
-            if self.environment is None:
-                raise ValueError("python_v2 runtime requires environment")
-            if self.gpu is None:
-                raise ValueError("python_v2 runtime requires gpu")
        else:
            object.__setattr__(self, "python_version", None)
            object.__setattr__(self, "environment", None)
            object.__setattr__(self, "env", None)
-            object.__setattr__(self, "gpu", None)
        return self


 class FunctionVersion(_RemoteValue):
    """An exact immutable Function version returned by Enterprise.

-    The GPU execution requirement is part of this identity. CPU and memory sizing,
-    priority, concurrency, and retry policy belong to the execution platform.
+    Scheduling resources, priority, concurrency, and retry policy belong to
+    the submitting Job and are not part of this identity.
    """

    name: str
@@ -541,107 +502,31 @@ _GRAMMAR_PRIMITIVES = (


 def _canonical_arrow_type(data_type: pa.DataType) -> str:
-    """The compact Function grammar, or canonical exact JSON for nested types."""
-    grammar = _grammar_arrow_type(data_type)
-    if grammar is not None:
-        return grammar
-    exact = _exact_arrow_type(data_type)
-    return json.dumps(exact, ensure_ascii=False, sort_keys=True, separators=(",", ":"))
-
-
-def _grammar_arrow_type(data_type: pa.DataType) -> Optional[str]:
+    """The server's V1 Function type grammar. Anything outside it is rejected
+    here rather than at registration."""
    for candidate, name in _GRAMMAR_PRIMITIVES:
        if data_type == candidate:
            return name
    if pa.types.is_list(data_type) or pa.types.is_large_list(data_type):
-        item = _grammar_list_item(data_type)
-        if item is None:
-            return None
        prefix = "list" if pa.types.is_list(data_type) else "large_list"
-        return f"{prefix}<{item}>"
+        return f"{prefix}<{_canonical_list_item(data_type)}>"
    if pa.types.is_fixed_size_list(data_type) and data_type.list_size > 0:
-        item = _grammar_list_item(data_type)
-        if item is not None:
-            return f"fixed_size_list<{item}, {data_type.list_size}>"
-    return None
+        return (
+            f"fixed_size_list<{_canonical_list_item(data_type)}, {data_type.list_size}>"
+        )
+    raise TypeError(f"unsupported Arrow type for Function signature: {data_type}")


-def _grammar_list_item(data_type: pa.DataType) -> Optional[str]:
+def _canonical_list_item(data_type: pa.DataType) -> str:
    """The grammar names only the item type; it always means a non-nullable
-    child called `item`, so other child properties require exact JSON."""
+    child called `item`, so any other child metadata cannot be represented."""
    child = data_type.value_field
    if child.name != "item" or child.nullable or child.metadata:
-        return None
-    return _grammar_arrow_type(child.type)
-
-
-def _validate_exact_arrow_field(field: pa.Field) -> None:
-    if not field.name:
        raise TypeError(
-            "unsupported Arrow type for Function signature: field names "
-            "must not be empty"
+            "unsupported Arrow type for Function signature: list items must be a "
+            f"non-nullable field named 'item', got {child}"
        )
-    if field.metadata:
-        raise TypeError(
-            "unsupported Arrow type for Function signature: field metadata "
-            f"is not supported, got {field}"
-        )
-
-
-def _exact_arrow_field(field: pa.Field) -> dict[str, Any]:
-    _validate_exact_arrow_field(field)
-    return {
-        "name": field.name,
-        "nullable": field.nullable,
-        "type": _exact_arrow_type(field.type),
-    }
-
-
-def _exact_arrow_type(data_type: pa.DataType) -> dict[str, Any]:
-    for candidate, name in _GRAMMAR_PRIMITIVES:
-        if data_type == candidate:
-            return {"type": name}
-    if pa.types.is_struct(data_type):
-        fields = list(data_type)
-        names = [field.name for field in fields]
-        if not fields or len(set(names)) != len(names):
-            raise TypeError(
-                "unsupported Arrow type for Function signature: structs must have "
-                "non-empty, uniquely named fields"
-            )
-        return {
-            "type": "struct",
-            "fields": [_exact_arrow_field(field) for field in fields],
-        }
-    if (
-        pa.types.is_list(data_type)
-        or pa.types.is_large_list(data_type)
-        or pa.types.is_fixed_size_list(data_type)
-    ):
-        if pa.types.is_fixed_size_list(data_type):
-            if data_type.value_field.name != "item":
-                raise TypeError(
-                    "unsupported Arrow type for Function signature: fixed-size list "
-                    "items must be named 'item'"
-                )
-            if data_type.list_size <= 0:
-                raise TypeError(
-                    f"unsupported Arrow type for Function signature: {data_type}"
-                )
-        value: dict[str, Any] = {
-            "type": (
-                "list"
-                if pa.types.is_list(data_type)
-                else "large_list"
-                if pa.types.is_large_list(data_type)
-                else "fixed_size_list"
-            ),
-            "fields": [_exact_arrow_field(data_type.value_field)],
-        }
-        if pa.types.is_fixed_size_list(data_type):
-            value["length"] = data_type.list_size
-        return value
-    raise TypeError(f"unsupported Arrow type for Function signature: {data_type}")
+    return _canonical_arrow_type(child.type)


 def _list_of(item: pa.DataType) -> pa.DataType:
@@ -715,11 +600,8 @@ def _callable_parameters(function: Callable[..., Any]) -> tuple[inspect.Paramete

 def _function_output(output: pa.DataType | pa.Field | pa.Schema) -> FunctionOutput:
    if isinstance(output, pa.Schema):
-        if output.metadata:
-            raise TypeError("Function output schema metadata is not supported")
        fields = tuple(output)
    elif isinstance(output, pa.Field) and pa.types.is_struct(output.type):
-        _validate_exact_arrow_field(output)
        if output.nullable:
            raise ValueError("Function output must be non-nullable")
        fields = tuple(output.type)
@@ -735,7 +617,6 @@ def _function_output(output: pa.DataType | pa.Field | pa.Schema) -> FunctionOutp
            raise TypeError(
                "output_schema must be a PyArrow DataType, Field, or Schema"
            )
-        _validate_exact_arrow_field(field)
        if field.nullable:
            raise ValueError("Function output must be non-nullable")
        return FunctionOutput(
@@ -748,8 +629,6 @@ def _function_output(output: pa.DataType | pa.Field | pa.Schema) -> FunctionOutp
        raise ValueError("named-struct Function output must contain at least one field")
    if any(field.nullable for field in fields):
        raise ValueError("Function output fields must be non-nullable")
-    for field in fields:
-        _validate_exact_arrow_field(field)
    names = [field.name for field in fields]
    if len(set(names)) != len(names):
        raise ValueError("Function output field names must be unique")
@@ -778,10 +657,6 @@ def _infer_signature(
    if input_schema is not None:
        if not isinstance(input_schema, pa.Schema):
            raise TypeError("input_schema must be a PyArrow Schema")
-        if input_schema.metadata:
-            raise TypeError("Function input schema metadata is not supported")
-        for field in input_schema:
-            _validate_exact_arrow_field(field)
        expected = tuple(parameter.name for parameter in parameters)
        actual = tuple(input_schema.names)
        if actual != expected:
@@ -1035,7 +910,6 @@ class UdfDefinition:
        pip: tuple[str, ...],
        env: Mapping[str, str],
        python_version: Optional[str],
-        gpu: bool = False,
        conda: tuple[str, ...] = (),
        conda_channels: tuple[str, ...] = (),
    ):
@@ -1064,14 +938,12 @@ class UdfDefinition:
        signature = _infer_signature(function, input_schema, output_schema)
        source = _package_source(function)
        digest = f"sha256:{hashlib.sha256(source).hexdigest()}"
-        gpu_marker = _normalize_gpu_marker(gpu)
        runtime = PythonRuntimeSpec(
-            kind="python_v2" if gpu_marker is not None else "python",
+            kind="python",
            python_version=python_version
            or f"{sys.version_info.major}.{sys.version_info.minor}",
            environment=environment_spec,
            env=environment,
-            gpu=gpu_marker,
        )
        self._function = function
        self._request = FunctionRegistrationRequest(
@@ -1117,7 +989,6 @@ def udf(
    pip: tuple[str, ...] | list[str] = (),
    env: Optional[Mapping[str, str]] = None,
    python_version: Optional[str] = None,
-    gpu: bool = False,
    conda: tuple[str, ...] | list[str] = (),
    conda_channels: tuple[str, ...] | list[str] = (),
 ) -> Callable[[Callable[..., Any]], UdfDefinition]: ...
@@ -1132,7 +1003,6 @@ def udf(
    pip: tuple[str, ...] | list[str] = (),
    env: Optional[Mapping[str, str]] = None,
    python_version: Optional[str] = None,
-    gpu: bool = False,
    conda: tuple[str, ...] | list[str] = (),
    conda_channels: tuple[str, ...] | list[str] = (),
 ):
@@ -1165,10 +1035,6 @@ def udf(
        Environment variables included in the Function definition.
    python_version : str, optional
        Remote Python major/minor version. Defaults to the client version.
-    gpu : bool, default False
-        Whether every remote execution requires a GPU. The execution platform
-        selects one compatible GPU for each worker. The requirement is part of
-        the immutable Function version.

    The packaged artifact is a snapshot: the function source plus exactly
    the module-level names it references (modules as imports, importable
@@ -1193,11 +1059,6 @@ def udf(
    ...     return value * 2
    >>> score(1.5)
    3.0
-    >>> @udf(pip=["cupy-cuda12x"], gpu=True)
-    ... def gpu_score(value: int) -> int:
-    ...     return value * 2
-    >>> gpu_score.registration_request.runtime.gpu
-    True
    """

    def decorate(target: Callable[..., Any]) -> UdfDefinition:
@@ -1209,7 +1070,6 @@ def udf(
            pip=tuple(pip),
            env={} if env is None else env,
            python_version=python_version,
-            gpu=gpu,
            conda=tuple(conda),
            conda_channels=tuple(conda_channels),
        )
@@ -71,6 +71,7 @@ from ..table import (
    AsyncTable,
    BlobMode,
    Branches,
+    CompactionOptions,
    IndexStatistics,
    Query,
    Table,
@@ -961,6 +962,7 @@ class RemoteTable(Table):
        *,
        cleanup_older_than: Optional[timedelta] = None,
        delete_unverified: bool = False,
+        compaction_options: Optional[CompactionOptions] = None,
    ):
        """
        optimize() is a no-op on LanceDB Cloud.
@@ -22,6 +22,7 @@ from typing import (
    Optional,
    Sequence,
    Tuple,
+    TypedDict,
    Union,
    overload,
 )
@@ -227,6 +228,88 @@ IndexConfigType = Union[
    FTS,
 ]

+
+class CompactionOptions(TypedDict, total=False):
+    """Options that control file compaction during table optimization.
+
+    Unspecified options use Lance's defaults.
+
+    Compaction planning is row based. Lowering ``target_rows_per_fragment``
+    based on the expected row size can bound later compaction passes once
+    oversized fragments have been rewritten. It does not split an existing
+    fragment, so the first pass over an oversized fragment is not subject to
+    that bound. ``max_bytes_per_file`` limits output file size, not compaction
+    memory. Source budgets keep whole planned tasks; if the first task exceeds
+    a budget, that run performs no compaction work.
+
+    Examples
+    --------
+    Derive a steady-state row target from the expected row size:
+
+    >>> desired_fragment_bytes = 750 * 1024 * 1024
+    >>> average_row_bytes = 1_500_000
+    >>> options: CompactionOptions = {
+    ...     "target_rows_per_fragment": max(
+    ...         1, desired_fragment_bytes // average_row_bytes
+    ...     ),
+    ... }
+    >>> await table.optimize(compaction_options=options)  # doctest: +SKIP
+    """
+
+    target_rows_per_fragment: int
+    """Target rows per fragment; existing oversized fragments are not split."""
+
+    max_rows_per_group: int
+    """Maximum number of rows per row group (default: 1,024)."""
+
+    max_bytes_per_file: Optional[int]
+    """Maximum output data-file size; this does not bound compaction memory."""
+
+    materialize_deletions: bool
+    """Whether to rewrite fragments containing deleted rows (default: True)."""
+
+    materialize_deletions_threshold: float
+    """Minimum deleted-row fraction that makes a fragment eligible (default: 0.1)."""
+
+    num_threads: Optional[int]
+    """Number of compaction tasks to run in parallel."""
+
+    batch_size: Optional[int]
+    """Number of rows per input scan batch."""
+
+    io_buffer_size: Optional[int]
+    """Maximum number of bytes queued in the input scan I/O buffer."""
+
+    defer_index_remap: bool
+    """Whether to defer index remapping during compaction (default: False)."""
+
+    index_remap_mode: Literal["direct", "compact"]
+    """How to construct the old-to-new row-address mapping."""
+
+    compaction_mode: Optional[
+        Literal["reencode", "try_binary_copy", "force_binary_copy"]
+    ]
+    """Whether compaction re-encodes data or uses binary copying."""
+
+    binary_copy_read_batch_bytes: Optional[int]
+    """Number of bytes read per batch during binary-copy compaction."""
+
+    max_source_fragments: Optional[int]
+    """Maximum number of source fragments compacted in one run."""
+
+    max_source_rows: Optional[int]
+    """Maximum live source rows per run, applied to whole planned tasks."""
+
+    max_source_bytes: Optional[int]
+    """Maximum source bytes per run, applied to whole planned tasks."""
+
+    excluded_fragment_ids: List[int]
+    """Fragment IDs to leave unchanged and use as planning boundaries."""
+
+    max_overlays_per_fragment: Optional[int]
+    """Maximum overlays before a fragment is fully compacted."""
+
+
 # Known distance metrics for legacy API detection
 KNOWN_METRICS = {"l2", "cosine", "dot", "hamming"}

@@ -2049,6 +2132,7 @@ class Table(ABC):
        cleanup_older_than: Optional[timedelta] = None,
        delete_unverified: bool = False,
        retrain: bool = False,
+        compaction_options: Optional[CompactionOptions] = None,
    ):
        """
        Optimize the on-disk data and indices for better performance.
@@ -2081,6 +2165,11 @@ class Table(ABC):

        retrain: bool, default False
            This parameter is no longer used and is deprecated.
+        compaction_options: CompactionOptions, optional
+            Options that control file compaction. For large rows, derive a lower
+            ``target_rows_per_fragment`` from the expected row size to bound later
+            passes. This does not cap the first pass over an existing oversized
+            fragment; see [CompactionOptions][lancedb.table.CompactionOptions].

        Notes
        -----
@@ -4201,6 +4290,7 @@ class LanceTable(Table):
        cleanup_older_than: Optional[timedelta] = None,
        delete_unverified: bool = False,
        retrain: bool = False,
+        compaction_options: Optional[CompactionOptions] = None,
    ):
        """
        Optimize the on-disk data and indices for better performance.
@@ -4233,6 +4323,11 @@ class LanceTable(Table):

        retrain: bool, default False
            This parameter is no longer used and is deprecated.
+        compaction_options: CompactionOptions, optional
+            Options that control file compaction. For large rows, derive a lower
+            ``target_rows_per_fragment`` from the expected row size to bound later
+            passes. This does not cap the first pass over an existing oversized
+            fragment; see [CompactionOptions][lancedb.table.CompactionOptions].

        Notes
        -----
@@ -4247,6 +4342,7 @@ class LanceTable(Table):
                cleanup_older_than=cleanup_older_than,
                delete_unverified=delete_unverified,
                retrain=retrain,
+                compaction_options=compaction_options,
            )
        )

@@ -6648,6 +6744,7 @@ class AsyncTable:
        cleanup_older_than: Optional[timedelta] = None,
        delete_unverified: bool = False,
        retrain=False,
+        compaction_options: Optional[CompactionOptions] = None,
    ) -> OptimizeStats:
        """
        Optimize the on-disk data and indices for better performance.
@@ -6680,6 +6777,11 @@ class AsyncTable:

        retrain: bool, default False
            This parameter is no longer used and is deprecated.
+        compaction_options: CompactionOptions, optional
+            Options that control file compaction. For large rows, derive a lower
+            ``target_rows_per_fragment`` from the expected row size to bound later
+            passes. This does not cap the first pass over an existing oversized
+            fragment; see [CompactionOptions][lancedb.table.CompactionOptions].

        Notes
        -----
@@ -6706,6 +6808,7 @@ class AsyncTable:
        return await self._inner.optimize(
            cleanup_since_ms=cleanup_since_ms,
            delete_unverified=delete_unverified,
+            compaction_options=compaction_options,
        )

    async def list_indices(self) -> Iterable[IndexConfig]:
@@ -19,7 +19,7 @@ import pyarrow as pa
 import pytest

 import lancedb
-from lancedb.functions import PythonRuntimeSpec, UdfDefinition, udf
+from lancedb.functions import UdfDefinition, udf

 THRESHOLD = 20
 _CACHE = None
@@ -89,58 +89,6 @@ def test_udf_conda_environment():
        udf(name="channels", conda_channels=["conda-forge"])(lambda value: value)


-def test_udf_gpu_marker_uses_gpu_runtime():
-    @udf(pip=["cupy-cuda12x"], gpu=True)
-    def double_on_gpu(value: int) -> int:
-        return value * 2
-
-    request = json.loads(double_on_gpu.registration_request.to_canonical_json())
-    assert request["runtime"]["kind"] == "python_v2"
-    assert request["runtime"]["gpu"] is True
-
-    @udf(pip=["pyarrow"])
-    def cpu_function(value: int) -> int:
-        return value
-
-    cpu_runtime = json.loads(cpu_function.registration_request.to_canonical_json())[
-        "runtime"
-    ]
-    assert cpu_runtime["kind"] == "python"
-    assert "gpu" not in cpu_runtime
-
-    def identity(value: int) -> int:
-        return value
-
-    for invalid in [None, 0, 1, -1, 1.5, "", "true", "1", "H100"]:
-        with pytest.raises(ValueError, match="gpu must be a boolean"):
-            udf(name="invalid_gpu", gpu=invalid)(identity)
-
-    base_runtime = {
-        "kind": "python_v2",
-        "python_version": "3.12",
-        "environment": {"kind": "pip"},
-    }
-    runtime = PythonRuntimeSpec.model_validate({**base_runtime, "gpu": True})
-    assert runtime.gpu is True
-    for invalid in [False, 1, 0, "", "true", "1", "H100"]:
-        with pytest.raises(ValueError, match="runtime.gpu must be true"):
-            PythonRuntimeSpec.model_validate({**base_runtime, "gpu": invalid})
-
-
-def test_unknown_runtime_discards_payload_before_known_field_validation():
-    for payload in [
-        {"kind": "python_v3", "gpu": {"model": "H100"}},
-        {"kind": "python_v3", "resources": []},
-        {
-            "kind": "python_v3",
-            "environment": {"kind": []},
-            "python_version": 3.15,
-        },
-    ]:
-        runtime = PythonRuntimeSpec.model_validate(payload)
-        assert runtime.to_canonical_json() == '{"kind":"python_v3"}'
-
-
 def test_udf_packages_attribute_access_and_body_imports():
    @udf
    def word_norm(body: str) -> float:
@@ -220,7 +168,7 @@ def test_udf_resolves_module_globals_before_builtins(tmp_path):
        udf(module.uses_callable_shadow)


-def test_canonical_arrow_type_prefers_the_compact_grammar():
+def test_canonical_arrow_type_is_exactly_the_grammar():
    from lancedb.functions import _GRAMMAR_PRIMITIVES, _canonical_arrow_type

    golden = json.loads(
@@ -233,13 +181,6 @@ def test_canonical_arrow_type_prefers_the_compact_grammar():
        case["arrow_type"] for case in golden["valid"] if "<" not in case["arrow_type"]
    ]
    assert [name for _, name in _GRAMMAR_PRIMITIVES] == primitives
-    assert _canonical_arrow_type(pa.list_(pa.field("item", pa.float32(), False))) == (
-        "list<float32>"
-    )
-    assert (
-        _canonical_arrow_type(pa.large_list(pa.field("item", pa.float32(), False)))
-        == "large_list<float32>"
-    )
    for outside in [
        pa.timestamp("us"),
        pa.decimal128(10, 2),
@@ -247,6 +188,7 @@ def test_canonical_arrow_type_prefers_the_compact_grammar():
        pa.large_binary(),
        pa.binary(4),
        pa.duration("s"),
+        pa.struct([pa.field("a", pa.int32())]),
        pa.list_(pa.float32(), 0),
        pa.list_(pa.timestamp("us")),
    ]:
@@ -436,29 +378,14 @@ def test_udf_recursion_versus_a_rebound_module_name(tmp_path):
        udf(raw_fact)


-def test_canonical_arrow_type_uses_exact_json_for_list_child_properties():
+def test_canonical_arrow_type_rejects_unrepresentable_list_children():
    from lancedb.functions import _canonical_arrow_type

-    nullable = pa.list_(pa.float32())
-    assert json.loads(_canonical_arrow_type(nullable)) == {
-        "type": "list",
-        "fields": [
-            {
-                "name": "item",
-                "nullable": True,
-                "type": {"type": "float32"},
-            }
-        ],
-    }
-    named = pa.list_(pa.field("custom", pa.float32(), nullable=False))
-    assert json.loads(_canonical_arrow_type(named))["fields"][0]["name"] == "custom"
    for outside in [
+        pa.list_(pa.float32()),  # pyarrow default: nullable child
+        pa.list_(pa.field("custom", pa.float32(), nullable=False)),
        pa.list_(pa.field("item", pa.float32(), nullable=False, metadata={"k": "v"})),
        pa.list_(pa.field("item", pa.float32(), nullable=False), 0),
-        pa.list_(
-            pa.field("item", pa.float32(), nullable=False, metadata={"k": "v"}), 3
-        ),
-        pa.list_(pa.field("custom", pa.float32(), nullable=False), 3),
    ]:
        with pytest.raises(TypeError, match="unsupported Arrow type"):
            _canonical_arrow_type(outside)
@@ -468,29 +395,6 @@ def test_canonical_arrow_type_uses_exact_json_for_list_child_properties():
        )
        == "fixed_size_list<float32, 3>"
    )
-    fixed = json.loads(_canonical_arrow_type(pa.list_(pa.float32(), 3)))
-    assert fixed == {
-        "type": "fixed_size_list",
-        "fields": [
-            {
-                "name": "item",
-                "nullable": True,
-                "type": {"type": "float32"},
-            }
-        ],
-        "length": 3,
-    }
-    large = json.loads(_canonical_arrow_type(pa.large_list(pa.float32())))
-    assert large["type"] == "large_list"
-    assert large["fields"][0]["nullable"] is True
-
-    for invalid_struct in [
-        pa.struct([]),
-        pa.struct([pa.field("a", pa.int32()), pa.field("a", pa.int64())]),
-        pa.struct([pa.field("", pa.int32())]),
-    ]:
-        with pytest.raises(TypeError, match="unsupported Arrow type"):
-            _canonical_arrow_type(invalid_struct)


 def _calls_missing(value: int) -> int:
@@ -578,105 +482,6 @@ def test_explicit_arrow_schema_is_deterministic():
    assert signature.output.nullable is False


-def test_nested_struct_output_uses_canonical_exact_json():
-    token = pa.struct(
-        [
-            pa.field("position", pa.int32(), nullable=False),
-            pa.field("value", pa.string(), nullable=False),
-            pa.field("length", pa.int32(), nullable=False),
-        ]
-    )
-    analysis = pa.struct(
-        [
-            pa.field("normalized_text", pa.string(), nullable=False),
-            pa.field("has_content", pa.bool_(), nullable=False),
-            pa.field(
-                "metrics",
-                pa.struct(
-                    [
-                        pa.field("character_count", pa.int64(), nullable=False),
-                        pa.field("word_count", pa.int32(), nullable=False),
-                        pa.field("average_word_length", pa.float64(), nullable=False),
-                    ]
-                ),
-                nullable=False,
-            ),
-            pa.field(
-                "diagnostics",
-                pa.struct(
-                    [
-                        pa.field("status", pa.string(), nullable=False),
-                        pa.field(
-                            "normalization",
-                            pa.struct(
-                                [
-                                    pa.field("changed", pa.bool_(), nullable=False),
-                                    pa.field(
-                                        "original_length", pa.int64(), nullable=False
-                                    ),
-                                ]
-                            ),
-                            nullable=False,
-                        ),
-                    ]
-                ),
-                nullable=False,
-            ),
-            pa.field(
-                "token_preview",
-                pa.list_(pa.field("item", token, nullable=False)),
-                nullable=False,
-            ),
-        ]
-    )
-
-    @udf(
-        input_schema=pa.schema([pa.field("text", pa.string(), nullable=False)]),
-        output_schema=pa.field("analysis", analysis, nullable=False),
-    )
-    def analyze(text):
-        return {"normalized_text": text}
-
-    output = analyze.registration_request.signature.output
-    assert output.kind == "named_struct"
-    assert [field.name for field in output.fields] == [
-        "normalized_text",
-        "has_content",
-        "metrics",
-        "diagnostics",
-        "token_preview",
-    ]
-    metrics = json.loads(output.fields[2].arrow_type)
-    assert metrics == {
-        "type": "struct",
-        "fields": [
-            {
-                "name": "character_count",
-                "nullable": False,
-                "type": {"type": "int64"},
-            },
-            {
-                "name": "word_count",
-                "nullable": False,
-                "type": {"type": "int32"},
-            },
-            {
-                "name": "average_word_length",
-                "nullable": False,
-                "type": {"type": "float64"},
-            },
-        ],
-    }
-    preview = json.loads(output.fields[4].arrow_type)
-    assert preview["type"] == "list"
-    assert preview["fields"][0]["type"]["type"] == "struct"
-    assert [field["name"] for field in preview["fields"][0]["type"]["fields"]] == [
-        "position",
-        "value",
-        "length",
-    ]
-
-
 def test_annotation_and_explicit_schema_validation_fail_closed():
    with pytest.raises(TypeError, match="missing Function annotations"):

@@ -720,72 +525,6 @@ def test_annotation_and_explicit_schema_validation_fail_closed():
        def nullable_explicit(value):
            return value

-    for invalid_field in [
-        pa.field("", pa.int32(), nullable=False),
-        pa.field("result", pa.int32(), nullable=False, metadata={"k": "v"}),
-    ]:
-        with pytest.raises(TypeError, match="unsupported Arrow type"):
-
-            @udf(
-                input_schema=pa.schema([pa.field("value", pa.int64())]),
-                output_schema=pa.schema([invalid_field]),
-            )
-            def invalid_explicit_field(value):
-                return value
-
-    with pytest.raises(TypeError, match="unsupported Arrow type"):
-
-        @udf(
-            input_schema=pa.schema(
-                [pa.field("value", pa.int64(), metadata={"k": "v"})]
-            ),
-            output_schema=pa.int64(),
-        )
-        def input_field_metadata(value):
-            return value
-
-    with pytest.raises(TypeError, match="unsupported Arrow type"):
-
-        @udf(
-            input_schema=pa.schema([pa.field("value", pa.int64())]),
-            output_schema=pa.field(
-                "result", pa.int64(), nullable=False, metadata={"k": "v"}
-            ),
-        )
-        def scalar_output_field_metadata(value):
-            return value
-
-    struct_type = pa.struct([pa.field("value", pa.int64(), nullable=False)])
-    with pytest.raises(TypeError, match="unsupported Arrow type"):
-
-        @udf(
-            input_schema=pa.schema([pa.field("value", pa.int64())]),
-            output_schema=pa.field(
-                "result", struct_type, nullable=False, metadata={"k": "v"}
-            ),
-        )
-        def struct_output_field_metadata(value):
-            return {"value": value}
-
-    for input_schema, output_schema in [
-        (
-            pa.schema([pa.field("value", pa.int64())], metadata={"k": "v"}),
-            pa.int64(),
-        ),
-        (
-            pa.schema([pa.field("value", pa.int64())]),
-            pa.schema(
-                [pa.field("result", pa.int64(), nullable=False)],
-                metadata={"k": "v"},
-            ),
-        ),
-    ]:
-        with pytest.raises(TypeError, match="schema metadata"):
-
-            @udf(input_schema=input_schema, output_schema=output_schema)
-            def schema_metadata(value):
-                return value
-

 def test_local_function_catalog_operations_are_not_supported(tmp_path):
    db = lancedb.connect(tmp_path)
@@ -13,7 +13,7 @@ from concurrent.futures import ThreadPoolExecutor
 from datetime import date, datetime, timedelta
 from decimal import Decimal
 from time import sleep
-from typing import List
+from typing import Any, List
 from unittest.mock import patch

 import lancedb
@@ -3876,6 +3876,100 @@ async def test_optimize(mem_db_async: AsyncConnection):
    assert await table.query().to_arrow() == pa.table({"x": [[1], [2]]})


+@pytest.mark.asyncio
+async def test_optimize_compaction_options(mem_db_async: AsyncConnection):
+    table = await mem_db_async.create_table("test", data=[{"x": 1}])
+    await table.add([{"x": 2}])
+
+    stats = await table.optimize(
+        compaction_options={
+            "target_rows_per_fragment": 1,
+            "batch_size": 1,
+            "num_threads": 1,
+        }
+    )
+    assert stats.compaction.fragments_removed == 0
+    assert stats.compaction.fragments_added == 0
+
+    stats = await table.optimize(compaction_options={"target_rows_per_fragment": 3})
+    assert stats.compaction.fragments_removed == 2
+    assert stats.compaction.fragments_added == 1
+
+    with pytest.raises(ValueError, match="Invalid compaction option: unknown"):
+        await table.optimize(compaction_options={"unknown": 1})
+
+
+@pytest.mark.parametrize("option", ["max_source_rows", "max_source_bytes"])
+@pytest.mark.asyncio
+async def test_optimize_compaction_source_limits(
+    mem_db_async: AsyncConnection, option: str
+):
+    table = await mem_db_async.create_table("test", data=[{"x": 1}])
+    await table.add([{"x": 2}])
+
+    stats = await table.optimize(
+        compaction_options={
+            "target_rows_per_fragment": 3,
+            option: 1,
+        }
+    )
+    assert stats.compaction.fragments_removed == 0
+    assert stats.compaction.fragments_added == 0
+
+
+@pytest.mark.asyncio
+async def test_optimize_compaction_excluded_fragments(mem_db_async: AsyncConnection):
+    table = await mem_db_async.create_table("test", data=[{"x": 1}])
+    await table.add([{"x": 2}])
+
+    stats = await table.optimize(
+        compaction_options={
+            "target_rows_per_fragment": 3,
+            "excluded_fragment_ids": [0],
+        }
+    )
+    assert stats.compaction.fragments_removed == 0
+    assert stats.compaction.fragments_added == 0
+
+    stats = await table.optimize(compaction_options={"target_rows_per_fragment": 3})
+    assert stats.compaction.fragments_removed == 2
+    assert stats.compaction.fragments_added == 1
+
+
+@pytest.mark.parametrize(
+    ("option", "value", "message"),
+    [
+        ("target_rows_per_fragment", 0, "must be between 1 and 4294967295"),
+        ("max_rows_per_group", 0, "must be between 1 and 4294967295"),
+        ("batch_size", 0, "must be between 1 and 4294967295"),
+        ("num_threads", 0, "must be greater than 0"),
+        ("target_rows_per_fragment", 2**32, "must be between 1 and 4294967295"),
+        ("max_rows_per_group", 2**32, "must be between 1 and 4294967295"),
+        ("batch_size", 2**32, "must be between 1 and 4294967295"),
+        ("io_buffer_size", 2**63, "must be at most 9223372036854775807"),
+        ("max_source_rows", 0, "must be greater than 0"),
+        ("max_source_bytes", 0, "must be greater than 0"),
+        (
+            "excluded_fragment_ids",
+            [-1],
+            "must contain values between 0 and 4294967295",
+        ),
+        (
+            "excluded_fragment_ids",
+            [2**32],
+            "must contain values between 0 and 4294967295",
+        ),
+    ],
+)
+@pytest.mark.asyncio
+async def test_optimize_compaction_options_validation(
+    mem_db_async: AsyncConnection, option: str, value: Any, message: str
+):
+    table = await mem_db_async.create_table("test", data=[{"x": 1}])
+    with pytest.raises(ValueError, match=message):
+        await table.optimize(compaction_options={option: value})
+
+
@pytest.mark.asyncio
 async def test_optimize_delete_unverified(tmp_db_async: AsyncConnection, tmp_path):
    table = await tmp_db_async.create_table(
@@ -20,8 +20,9 @@ use arrow::{
 use lancedb::blob::{BlobFile, BlobRangeRequest};
 use lancedb::index::scalar::FtsIndexBuilder;
 use lancedb::table::{
-    AddDataMode, ColumnAlteration, Duration, FieldMetadataUpdate, FtsToken as LanceDbFtsToken,
-    NewColumnTransform, OptimizeAction, OptimizeOptions, Ref, Table as LanceDbTable,
+    AddDataMode, ColumnAlteration, CompactionMode, CompactionOptions, Duration,
+    FieldMetadataUpdate, FtsToken as LanceDbFtsToken, IndexRemapMode, NewColumnTransform,
+    OptimizeAction, OptimizeOptions, Ref, Table as LanceDbTable,
 };
 use lancedb::tokenize as lancedb_tokenize;
 use pyo3::{
@@ -100,6 +101,128 @@ enum PredicateArg {
    Sql(String),
 }

+fn validate_positive_u32(value: u64, name: &str) -> PyResult<usize> {
+    if !(1..=u32::MAX as u64).contains(&value) {
+        return Err(PyValueError::new_err(format!(
+            "{name} must be between 1 and {}",
+            u32::MAX
+        )));
+    }
+    Ok(value as usize)
+}
+
+fn positive_u32(value: &Bound<'_, PyAny>, name: &str) -> PyResult<usize> {
+    validate_positive_u32(value.extract()?, name)
+}
+
+fn optional_positive_u32(value: &Bound<'_, PyAny>, name: &str) -> PyResult<Option<usize>> {
+    value
+        .extract::<Option<u64>>()?
+        .map(|value| validate_positive_u32(value, name))
+        .transpose()
+}
+
+fn optional_positive_usize(value: &Bound<'_, PyAny>, name: &str) -> PyResult<Option<usize>> {
+    let value: Option<usize> = value.extract()?;
+    if value == Some(0) {
+        return Err(PyValueError::new_err(format!(
+            "{name} must be greater than 0"
+        )));
+    }
+    Ok(value)
+}
+
+fn optional_positive_u64(value: &Bound<'_, PyAny>, name: &str) -> PyResult<Option<u64>> {
+    let value: Option<u64> = value.extract()?;
+    if value == Some(0) {
+        return Err(PyValueError::new_err(format!(
+            "{name} must be greater than 0"
+        )));
+    }
+    Ok(value)
+}
+
+fn u32_list(value: &Bound<'_, PyAny>, name: &str) -> PyResult<Vec<u32>> {
+    value
+        .extract::<Vec<i64>>()?
+        .into_iter()
+        .map(|value| {
+            u32::try_from(value).map_err(|_| {
+                PyValueError::new_err(format!(
+                    "{name} must contain values between 0 and {}",
+                    u32::MAX
+                ))
+            })
+        })
+        .collect()
+}
+
+fn optional_i64_bounded_u64(value: &Bound<'_, PyAny>, name: &str) -> PyResult<Option<u64>> {
+    let value: Option<u64> = value.extract()?;
+    if value.is_some_and(|value| value > i64::MAX as u64) {
+        return Err(PyValueError::new_err(format!(
+            "{name} must be at most {}",
+            i64::MAX
+        )));
+    }
+    Ok(value)
+}
+
+fn parse_compaction_options(options: Option<&Bound<'_, PyDict>>) -> PyResult<CompactionOptions> {
+    let mut parsed = CompactionOptions::default();
+    let Some(options) = options else {
+        return Ok(parsed);
+    };
+
+    for (key, value) in options.iter() {
+        let key: String = key.extract()?;
+        match key.as_str() {
+            "target_rows_per_fragment" => {
+                parsed.target_rows_per_fragment = positive_u32(&value, &key)?
+            }
+            "max_rows_per_group" => parsed.max_rows_per_group = positive_u32(&value, &key)?,
+            "max_bytes_per_file" => parsed.max_bytes_per_file = value.extract()?,
+            "materialize_deletions" => parsed.materialize_deletions = value.extract()?,
+            "materialize_deletions_threshold" => {
+                parsed.materialize_deletions_threshold = value.extract()?
+            }
+            "num_threads" => parsed.num_threads = optional_positive_usize(&value, &key)?,
+            "batch_size" => parsed.batch_size = optional_positive_u32(&value, &key)?,
+            "io_buffer_size" => parsed.io_buffer_size = optional_i64_bounded_u64(&value, &key)?,
+            "defer_index_remap" => parsed.defer_index_remap = value.extract()?,
+            "index_remap_mode" => {
+                let mode: String = value.extract()?;
+                parsed.index_remap_mode = IndexRemapMode::try_from(mode.as_str())
+                    .map_err(|err| PyValueError::new_err(err.to_string()))?;
+            }
+            "compaction_mode" => {
+                let mode: Option<String> = value.extract()?;
+                parsed.compaction_mode = mode
+                    .map(|mode| {
+                        CompactionMode::try_from(mode.as_str())
+                            .map_err(|err| PyValueError::new_err(err.to_string()))
+                    })
+                    .transpose()?;
+            }
+            "binary_copy_read_batch_bytes" => {
+                parsed.binary_copy_read_batch_bytes = value.extract()?
+            }
+            "max_source_fragments" => parsed.max_source_fragments = value.extract()?,
+            "max_source_rows" => parsed.max_source_rows = optional_positive_usize(&value, &key)?,
+            "max_source_bytes" => parsed.max_source_bytes = optional_positive_u64(&value, &key)?,
+            "excluded_fragment_ids" => parsed.excluded_fragment_ids = u32_list(&value, &key)?,
+            "max_overlays_per_fragment" => parsed.max_overlays_per_fragment = value.extract()?,
+            _ => {
+                return Err(PyValueError::new_err(format!(
+                    "Invalid compaction option: {key}"
+                )));
+            }
+        }
+    }
+
+    Ok(parsed)
+}
+
 /// Statistics about a compaction operation.
 #[pyclass(get_all, from_py_object)]
 #[derive(Clone, Debug)]
@@ -1340,13 +1463,15 @@ impl Table {
    }

    /// Optimize the on-disk data by compacting and pruning old data, for better performance.
-    #[pyo3(signature = (cleanup_since_ms=None, delete_unverified=None))]
-    pub fn optimize(
-        self_: PyRef<'_, Self>,
+    #[pyo3(signature = (cleanup_since_ms=None, delete_unverified=None, compaction_options=None))]
+    pub fn optimize<'py>(
+        self_: PyRef<'py, Self>,
        cleanup_since_ms: Option<u64>,
        delete_unverified: Option<bool>,
-    ) -> PyResult<Bound<'_, PyAny>> {
+        compaction_options: Option<&Bound<'py, PyDict>>,
+    ) -> PyResult<Bound<'py, PyAny>> {
        let inner = self_.inner_ref()?.clone();
+        let compaction_options = parse_compaction_options(compaction_options)?;
        let older_than = if let Some(ms) = cleanup_since_ms {
            if ms > i64::MAX as u64 {
                return Err(PyValueError::new_err(format!(
@@ -1362,7 +1487,7 @@ impl Table {
        future_into_py(self_.py(), async move {
            let compaction_stats = inner
                .optimize(OptimizeAction::Compact {
-                    options: lancedb::table::CompactionOptions::default(),
+                    options: compaction_options,
                    remap_options: None,
                })
                .await
@@ -1,6 +1,6 @@
 [package]
 name = "lancedb"
-version = "0.38.0-beta.14"
+version = "0.38.0-beta.12"
 edition.workspace = true
 description = "LanceDB: A serverless, low-latency vector database for AI applications"
 license.workspace = true
@@ -207,33 +207,6 @@ pub enum PythonRuntimeSpec {
        environment: PythonEnvironmentSpec,
        env: BTreeMap<String, String>,
    },
-    /// The GPU-enabled Sophon-managed Python runtime.
-    ///
-    /// # Examples
-    ///
-    /// ```
-    /// use std::collections::BTreeMap;
-    /// use lancedb::function::{PythonEnvironmentSpec, PythonRuntimeSpec};
-    ///
-    /// let runtime = PythonRuntimeSpec::PythonV2 {
-    ///     python_version: "3.12".to_string(),
-    ///     environment: PythonEnvironmentSpec {
-    ///         kind: "pip".to_string(),
-    ///         packages: vec!["cupy-cuda12x".to_string()],
-    ///         channels: Vec::new(),
-    ///         path: None,
-    ///         modules: Vec::new(),
-    ///         image: None,
-    ///     },
-    ///     env: BTreeMap::new(),
-    /// };
-    /// assert!(runtime.requires_gpu());
-    /// ```
-    PythonV2 {
-        python_version: String,
-        environment: PythonEnvironmentSpec,
-        env: BTreeMap<String, String>,
-    },
    /// A runtime kind introduced by a newer server.
    ///
    /// Unknown payload fields are intentionally not retained because the
@@ -246,27 +219,22 @@ impl PythonRuntimeSpec {
    pub fn kind(&self) -> &str {
        match self {
            Self::Python { .. } => "python",
-            Self::PythonV2 { .. } => "python_v2",
            Self::Unrecognized { kind } => kind,
        }
    }

-    /// The Python version for a known Python runtime, or `None` for an unknown kind.
+    /// The Python version for the V1 runtime, or `None` for an unknown kind.
    pub fn python_version(&self) -> Option<&str> {
        match self {
-            Self::Python { python_version, .. } | Self::PythonV2 { python_version, .. } => {
-                Some(python_version)
-            }
+            Self::Python { python_version, .. } => Some(python_version),
            Self::Unrecognized { .. } => None,
        }
    }

-    /// The Python environment for a known Python runtime, or `None` for an unknown kind.
+    /// The Python environment for the V1 runtime, or `None` for an unknown kind.
    pub fn environment(&self) -> Option<&PythonEnvironmentSpec> {
        match self {
-            Self::Python { environment, .. } | Self::PythonV2 { environment, .. } => {
-                Some(environment)
-            }
+            Self::Python { environment, .. } => Some(environment),
            Self::Unrecognized { .. } => None,
        }
    }
@@ -274,73 +242,38 @@ impl PythonRuntimeSpec {
    /// Environment variables, or `None` for an unknown kind.
    pub fn env(&self) -> Option<&BTreeMap<String, String>> {
        match self {
-            Self::Python { env, .. } | Self::PythonV2 { env, .. } => Some(env),
+            Self::Python { env, .. } => Some(env),
            Self::Unrecognized { .. } => None,
        }
    }
-
-    /// Whether the runtime requires a GPU selected by the execution platform.
-    pub fn requires_gpu(&self) -> bool {
-        matches!(self, Self::PythonV2 { .. })
-    }
 }

 #[derive(Deserialize)]
-struct PythonRuntimeV1Wire {
-    python_version: String,
-    environment: PythonEnvironmentSpec,
+struct PythonRuntimeWire {
+    kind: String,
+    #[serde(default)]
+    python_version: Option<String>,
+    #[serde(default)]
+    environment: Option<PythonEnvironmentSpec>,
    #[serde(default)]
    env: BTreeMap<String, String>,
-    #[serde(default)]
-    gpu: Option<Value>,
-}
-
-#[derive(Deserialize)]
-struct PythonRuntimeV2Wire {
-    python_version: String,
-    environment: PythonEnvironmentSpec,
-    #[serde(default)]
-    env: BTreeMap<String, String>,
-    gpu: bool,
 }

 impl<'de> Deserialize<'de> for PythonRuntimeSpec {
    fn deserialize<D: Deserializer<'de>>(deserializer: D) -> std::result::Result<Self, D::Error> {
-        let value = Value::deserialize(deserializer)?;
-        let kind = value
-            .get("kind")
-            .ok_or_else(|| de::Error::missing_field("kind"))?
-            .as_str()
-            .ok_or_else(|| de::Error::custom("runtime.kind must be a string"))?
-            .to_string();
-        match kind.as_str() {
-            "python" => {
-                let wire: PythonRuntimeV1Wire =
-                    serde_json::from_value(value).map_err(de::Error::custom)?;
-                if wire.gpu.is_some() {
-                    return Err(de::Error::custom(
-                        "python runtime with gpu requires kind='python_v2'",
-                    ));
-                }
-                Ok(Self::Python {
-                    python_version: wire.python_version,
-                    environment: wire.environment,
-                    env: wire.env,
-                })
-            }
-            "python_v2" => {
-                let wire: PythonRuntimeV2Wire =
-                    serde_json::from_value(value).map_err(de::Error::custom)?;
-                if !wire.gpu {
-                    return Err(de::Error::custom("runtime.gpu must be true"));
-                }
-                Ok(Self::PythonV2 {
-                    python_version: wire.python_version,
-                    environment: wire.environment,
-                    env: wire.env,
-                })
-            }
-            _ => Ok(Self::Unrecognized { kind }),
+        let wire = PythonRuntimeWire::deserialize(deserializer)?;
+        if wire.kind == "python" {
+            Ok(Self::Python {
+                python_version: wire
+                    .python_version
+                    .ok_or_else(|| de::Error::missing_field("python_version"))?,
+                environment: wire
+                    .environment
+                    .ok_or_else(|| de::Error::missing_field("environment"))?,
+                env: wire.env,
+            })
+        } else {
+            Ok(Self::Unrecognized { kind: wire.kind })
        }
    }
 }
@@ -354,8 +287,6 @@ impl Serialize for PythonRuntimeSpec {
            environment: &'a PythonEnvironmentSpec,
            #[serde(skip_serializing_if = "BTreeMap::is_empty")]
            env: &'a BTreeMap<String, String>,
-            #[serde(skip_serializing_if = "Option::is_none")]
-            gpu: Option<bool>,
        }

        #[derive(Serialize)]
@@ -373,19 +304,6 @@ impl Serialize for PythonRuntimeSpec {
                python_version,
                environment,
                env,
-                gpu: None,
-            }
-            .serialize(serializer),
-            Self::PythonV2 {
-                python_version,
-                environment,
-                env,
-            } => PythonRuntimeRef {
-                kind: "python_v2",
-                python_version,
-                environment,
-                env,
-                gpu: Some(true),
            }
            .serialize(serializer),
            Self::Unrecognized { kind } => UnrecognizedRuntimeRef { kind }.serialize(serializer),
@@ -395,8 +313,8 @@ impl Serialize for PythonRuntimeSpec {

 /// Immutable Function version returned by the Enterprise catalog.
 ///
-/// The GPU execution requirement is part of this identity. CPU and memory sizing,
-/// priority, concurrency, and retry policy belong to the execution platform.
+/// Scheduling resources, priority, concurrency, and retry policy belong to
+/// the submitting Job and are not part of this identity.
 #[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
 pub struct FunctionVersion {
    name: String,
@@ -671,7 +589,7 @@ impl_json!(RefreshColumnResult);

 #[cfg(test)]
 mod conda_environment_tests {
-    use super::{PythonEnvironmentSpec, PythonRuntimeSpec};
+    use super::PythonEnvironmentSpec;

    #[test]
    fn conda_channels_round_trip_and_pip_stays_bare() {
@@ -690,45 +608,4 @@ mod conda_environment_tests {
            serde_json::from_str(r#"{"kind":"pip","packages":["numpy"]}"#).unwrap();
        assert!(!serde_json::to_string(&pip).unwrap().contains("channels"));
    }
-
-    #[test]
-    fn gpu_python_runtime_marker_round_trips_and_validates() {
-        let runtime: PythonRuntimeSpec = serde_json::from_str(
-            r#"{"kind":"python_v2","python_version":"3.12","environment":{"kind":"pip"},"gpu":true}"#,
-        )
-        .unwrap();
-        assert_eq!(runtime.kind(), "python_v2");
-        assert!(runtime.requires_gpu());
-        assert_eq!(
-            super::canonical_json(&runtime).unwrap(),
-            r#"{"environment":{"kind":"pip"},"gpu":true,"kind":"python_v2","python_version":"3.12"}"#
-        );
-
-        for invalid in [
-            r#"{"kind":"python","python_version":"3.12","environment":{"kind":"pip"},"gpu":true}"#,
-            r#"{"kind":"python_v2","python_version":"3.12","environment":{"kind":"pip"}}"#,
-            r#"{"kind":"python_v2","python_version":"3.12","environment":{"kind":"pip"},"gpu":1}"#,
-            r#"{"kind":"python_v2","python_version":"3.12","environment":{"kind":"pip"},"gpu":false}"#,
-            r#"{"kind":"python_v2","python_version":"3.12","environment":{"kind":"pip"},"gpu":"true"}"#,
-            r#"{"kind":"python_v2","python_version":"3.12","environment":{"kind":"pip"},"gpu":"H100"}"#,
-        ] {
-            assert!(serde_json::from_str::<PythonRuntimeSpec>(invalid).is_err());
-        }
-    }
-
-    #[test]
-    fn unknown_runtime_discards_payload_before_known_field_validation() {
-        for encoded in [
-            r#"{"kind":"python_v3","gpu":{"model":"H100"}}"#,
-            r#"{"kind":"python_v3","resources":[]}"#,
-            r#"{"kind":"python_v3","python_version":3.15,"environment":{"kind":[]}}"#,
-        ] {
-            let runtime: PythonRuntimeSpec = serde_json::from_str(encoded).unwrap();
-            assert_eq!(runtime.kind(), "python_v3");
-            assert_eq!(
-                super::canonical_json(&runtime).unwrap(),
-                r#"{"kind":"python_v3"}"#
-            );
-        }
-    }
 }
@@ -5734,8 +5734,9 @@ mod tests {
            ))
            .execute()
            .await;
-        let Err(err) = result else {
-            panic!("legacy remote query unexpectedly succeeded")
+        let err = match result {
+            Ok(_) => panic!("legacy remote query unexpectedly succeeded"),
+            Err(err) => err,
        };

        assert!(
@@ -7464,93 +7465,6 @@ mod tests {
        assert_eq!(result.version, 8);
    }

-    #[tokio::test]
-    async fn test_add_function_column_allows_an_existing_binding() {
-        let binding = crate::function::FunctionBinding::from_json(include_str!(
-            "../../tests/fixtures/first_class_functions/v1/remote_function_binding.json"
-        ))
-        .unwrap();
-        let binding_metadata = crate::table::computed_columns::function_bindings_metadata(
-            std::slice::from_ref(&binding),
-        )
-        .unwrap();
-        let mut fields = vec![
-            Field::new("title", DataType::Utf8, true),
-            Field::new("body", DataType::Utf8, true),
-        ];
-        fields.extend(binding.outputs().iter().map(|output| {
-            let data_type = match output.arrow_type.as_str() {
-                "utf8" => DataType::Utf8,
-                "int64" => DataType::Int64,
-                other => panic!("unexpected fixture output type {other}"),
-            };
-            Field::new(&output.output_name, data_type, true).with_metadata(
-                crate::table::computed_columns::function_computed_column_metadata(
-                    binding.binding_id(),
-                    output.output_ordinal,
-                    &["title".into(), "body".into()],
-                ),
-            )
-        }));
-        let schema = Schema::new_with_metadata(
-            fields,
-            HashMap::from([(
-                crate::table::computed_columns::FUNCTION_BINDINGS_META_KEY.to_string(),
-                binding_metadata,
-            )]),
-        );
-        let table =
-            Table::new_with_handler("my_table", move |request| match request.url().path() {
-                "/v1/table/my_table/describe/" => http::Response::builder()
-                    .status(200)
-                    .body(describe_response(&schema))
-                    .unwrap(),
-                "/v1/table/my_table/add_columns/" => {
-                    let actual: serde_json::Value =
-                        serde_json::from_slice(request.body().unwrap().as_bytes().unwrap())
-                            .unwrap();
-                    assert_eq!(
-                        actual["new_columns"],
-                        serde_json::json!([
-                            {"name":"secondary_text","all_null":true},
-                            {"name":"secondary_token_count","all_null":true}
-                        ])
-                    );
-                    http::Response::builder()
-                        .status(200)
-                        .body(r#"{"version":10}"#.to_string())
-                        .unwrap()
-                }
-                path => panic!("Unexpected path: {path}"),
-            });
-        let application = crate::function::FunctionApplication::from_json(
-            r#"{
-                "function":{"name":"text_features","version":"fv_01K3TEXT"},
-                "inputs":[
-                    {"parameter":"title","kind":"column","value":{"path":"title"}},
-                    {"parameter":"body","kind":"column","value":{"path":"body"}}
-                ],
-                "output":{"kind":"named_struct","fields":[
-                    {"name":"normalized_text","arrow_type":"utf8","nullable":false},
-                    {"name":"token_count","arrow_type":"int64","nullable":false}
-                ]},
-                "columns":{
-                    "normalized_text":"secondary_text",
-                    "token_count":"secondary_token_count"
-                }
-            }"#,
-        )
-        .unwrap();
-
-        let result = table
-            .add_columns()
-            .function(application)
-            .execute()
-            .await
-            .unwrap();
-        assert_eq!(result.version, 10);
-    }
-
    #[tokio::test]
    async fn test_add_fixed_size_list_function_column_declares_the_vector_type() {
        let table = Table::new_with_handler("my_table", |request| {
@@ -103,7 +103,9 @@ pub use lance::dataset::refs::{BranchContents, Ref, TagContents, Tags as LanceTa
 pub use lance::dataset::scanner::DatasetRecordBatchStream;
 pub use lance_index::optimize::OptimizeOptions;
 pub use lsm_stats::{BucketStats, GenerationStats, LsmStats, MemtableStats};
-pub use optimize::{CompactionOptions, OptimizeAction, OptimizeStats};
+pub use optimize::{
+    CompactionMode, CompactionOptions, IndexRemapMode, OptimizeAction, OptimizeStats,
+};
 pub use refresh::RefreshColumnResult;
 pub use schema_evolution::{
    AddColumnsResult, AlterColumnsResult, DropColumnsResult, FieldMetadataUpdate,
@@ -5763,13 +5765,6 @@ mod tests {
        assert!(index_bytes > 0);
        assert_eq!(with_index, data_only + index_bytes);

-        // Release builds reject unstable overlay datasets unless explicitly opted in.
-        if !lance_table::feature_flags::can_read_dataset(
-            lance_table::feature_flags::FLAG_UNSTABLE_DATA_OVERLAY_FILES,
-        ) {
-            return;
-        }
-
        // Commit an overlay file supplying new `foo` values for the first three
        // rows of fragment 0. There is no high-level API that writes overlays
        // yet, so write the overlay's data file and commit the `DataOverlay`
@@ -547,12 +547,7 @@ fn ensure_known_binding_shape(value: &Value) -> Result<()> {
    Ok(())
 }

-struct ResolvedFieldPath<'a> {
-    root: &'a ArrowField,
-    leaf: &'a ArrowField,
-}
-
-fn resolve_field_path<'a>(schema: &'a ArrowSchema, path: &str) -> Result<ResolvedFieldPath<'a>> {
+fn resolve_field_path<'a>(schema: &'a ArrowSchema, path: &str) -> Result<&'a ArrowField> {
    let parts = lance_core::datatypes::parse_field_path(path).map_err(|e| {
        invalid_function(format!("invalid Function input field path '{path}': {e}"))
    })?;
@@ -561,23 +556,22 @@ fn resolve_field_path<'a>(schema: &'a ArrowSchema, path: &str) -> Result<Resolve
            "Function input field path cannot be empty",
        ));
    };
-    let root = schema
+    let mut field = schema
        .field_with_name(root)
        .map_err(|_| invalid_function(format!("unknown Function input column '{path}'")))?;
-    let mut leaf = root;
    for child in children {
-        let DataType::Struct(fields) = leaf.data_type() else {
+        let DataType::Struct(fields) = field.data_type() else {
            return Err(invalid_function(format!(
                "Function input field path '{path}' traverses a non-struct field"
            )));
        };
-        leaf = fields
+        field = fields
            .iter()
            .find(|field| field.name() == child)
            .map(AsRef::as_ref)
            .ok_or_else(|| invalid_function(format!("unknown Function input column '{path}'")))?;
    }
-    Ok(ResolvedFieldPath { root, leaf })
+    Ok(field)
 }

 fn canonical_input_arrow_type(field: &JsonArrowField) -> Result<String> {
@@ -672,8 +666,7 @@ fn parse_output_arrow_type(raw: &str) -> Result<JsonArrowDataType> {
 fn ensure_binding_matches_schema(schema: &ArrowSchema, binding: &FunctionBinding) -> Result<()> {
    let mut input_fields = Vec::with_capacity(binding.inputs().len());
    for input in binding.inputs() {
-        let resolved = resolve_field_path(schema, &input.field_path)?;
-        let field = resolved.leaf;
+        let field = resolve_field_path(schema, &input.field_path)?;
        if field
            .metadata()
            .get(COMPUTED_COLUMN_META_KEY)
@@ -728,11 +721,6 @@ fn ensure_binding_matches_schema(schema: &ArrowSchema, binding: &FunctionBinding
        )));
    }

-    let expected_inputs = binding
-        .inputs()
-        .iter()
-        .map(|input| input.field_path.clone())
-        .collect::<Vec<_>>();
    let mut output_fields = Vec::with_capacity(binding.outputs().len());
    for output in binding.outputs() {
        let field = schema.field_with_name(&output.output_name).map_err(|_| {
@@ -759,28 +747,6 @@ fn ensure_binding_matches_schema(schema: &ArrowSchema, binding: &FunctionBinding
                binding.binding_id()
            )));
        }
-        let metadata = field.metadata();
-        let declared_inputs = metadata
-            .get(INPUTS_META_KEY)
-            .and_then(|raw| serde_json::from_str::<Vec<String>>(raw).ok());
-        if metadata.get(COMPUTED_COLUMN_META_KEY).map(String::as_str) != Some("true")
-            || metadata.get(KIND_META_KEY).map(String::as_str) != Some(FUNCTION_KIND)
-            || metadata
-                .get(FUNCTION_BINDING_ID_META_KEY)
-                .map(String::as_str)
-                != Some(binding.binding_id())
-            || metadata
-                .get(FUNCTION_OUTPUT_ORDINAL_META_KEY)
-                .and_then(|value| value.parse::<u32>().ok())
-                != Some(output.output_ordinal)
-            || declared_inputs.as_deref() != Some(expected_inputs.as_slice())
-        {
-            return Err(invalid_function(format!(
-                "Function output '{}' declaration metadata does not match binding '{}'",
-                output.output_name,
-                binding.binding_id()
-            )));
-        }
        output_fields.push(ArrowField::new(
            field.name().clone(),
            field.data_type().clone(),
@@ -812,7 +778,7 @@ pub(crate) fn plan_function_application(
    application: &FunctionApplication,
    output_name: Option<&str>,
 ) -> Result<FunctionDeclarationPlan> {
-    ensure_supported_function_metadata(schema)?;
+    ensure_no_function_bindings_for_mutation(schema, "Function binding declaration")?;
    if application.has_unknown_fields() {
        return Err(Error::NotSupported {
            message: "Function application contains fields from a newer contract".into(),
@@ -862,9 +828,8 @@ pub(crate) fn plan_function_application(
                input.parameter
            ))
        })?;
-        let resolved = resolve_field_path(schema, path)?;
-        if resolved
-            .root
+        let field = resolve_field_path(schema, path)?;
+        if field
            .metadata()
            .get(COMPUTED_COLUMN_META_KEY)
            .map(String::as_str)
@@ -874,7 +839,6 @@ pub(crate) fn plan_function_application(
                "Function input '{path}' is computed; computed-on-computed bindings are not supported"
            )));
        }
-        let field = resolved.leaf;
        let parameter_field = ArrowField::new(
            input.parameter.clone(),
            field.data_type().clone(),
@@ -1498,7 +1462,9 @@ fn plan_declarations(schema: SchemaRef, columns: &[(String, String)]) -> Result<

    for (name, expression) in columns {
        if schema.field_with_name(name).is_ok() {
-            return Err(Error::ColumnAlreadyExists { name: name.clone() });
+            return Err(Error::ColumnAlreadyExists {
+                name: name.to_string(),
+            });
        }

        let bound = bind(schema.clone(), name, expression)?;
@@ -2615,37 +2581,6 @@ mod tests {
        ])
    }

-    fn valid_function_binding_schema(
-        title_nullable: bool,
-        body_nullable: bool,
-        binding: &FunctionBinding,
-    ) -> ArrowSchema {
-        let mut fields = function_binding_schema(title_nullable, body_nullable)
-            .fields()
-            .iter()
-            .map(|field| field.as_ref().clone())
-            .collect::<Vec<_>>();
-        let inputs = binding
-            .inputs()
-            .iter()
-            .map(|input| input.field_path.clone())
-            .collect::<Vec<_>>();
-        for output in binding.outputs() {
-            let index = fields
-                .iter()
-                .position(|field| field.name() == &output.output_name)
-                .unwrap();
-            fields[index] = fields[index]
-                .clone()
-                .with_metadata(function_computed_column_metadata(
-                    binding.binding_id(),
-                    output.output_ordinal,
-                    &inputs,
-                ));
-        }
-        ArrowSchema::new(fields)
-    }
-
    #[test]
    fn test_non_nullable_function_inputs_can_bind_to_nullable_parameters() {
        let binding = FunctionBinding::from_json(include_str!(
@@ -2653,11 +2588,7 @@ mod tests {
        ))
        .unwrap();

-        ensure_binding_matches_schema(
-            &valid_function_binding_schema(false, false, &binding),
-            &binding,
-        )
-        .unwrap();
+        ensure_binding_matches_schema(&function_binding_schema(false, false), &binding).unwrap();
    }

    #[test]
@@ -2670,11 +2601,8 @@ mod tests {
        raw_binding["input_schema"]["fields"][0]["nullable"] = Value::Bool(false);
        let binding: FunctionBinding = serde_json::from_value(raw_binding).unwrap();

-        let err = ensure_binding_matches_schema(
-            &valid_function_binding_schema(true, false, &binding),
-            &binding,
-        )
-        .unwrap_err();
+        let err = ensure_binding_matches_schema(&function_binding_schema(true, false), &binding)
+            .unwrap_err();
        assert!(
            matches!(&err, Error::InvalidInput { message }
                if message.contains("input column 'title' is nullable")
@@ -2685,73 +2613,6 @@ mod tests {
        );
    }

-    #[test]
-    fn test_second_binding_rejects_outputs_without_reciprocal_metadata() {
-        let binding = FunctionBinding::from_json(include_str!(
-            "../../tests/fixtures/first_class_functions/v1/remote_function_binding.json"
-        ))
-        .unwrap();
-        let schema = ArrowSchema::new_with_metadata(
-            function_binding_schema(true, true).fields().to_vec(),
-            HashMap::from([(
-                FUNCTION_BINDINGS_META_KEY.to_string(),
-                function_bindings_metadata(std::slice::from_ref(&binding)).unwrap(),
-            )]),
-        );
-
-        let err = plan_function_application(
-            &schema,
-            &named_struct_application(
-                r#"{"normalized_text":"secondary_text","token_count":"secondary_token_count"}"#,
-            ),
-            None,
-        )
-        .unwrap_err();
-        assert!(
-            matches!(&err, Error::InvalidInput { message }
-                if message.contains("declaration metadata")
-                    && message.contains("fb_01K3TEXT")),
-            "{err:?}"
-        );
-    }
-
-    #[test]
-    fn test_persisted_nested_input_keeps_leaf_level_validation() {
-        let mut raw_binding: Value = serde_json::from_str(include_str!(
-            "../../tests/fixtures/first_class_functions/v1/remote_function_binding.json"
-        ))
-        .unwrap();
-        raw_binding["inputs"][0]["field_path"] = Value::String("title.value".to_string());
-        let binding: FunctionBinding = serde_json::from_value(raw_binding).unwrap();
-        let title = ArrowField::new(
-            "title",
-            DataType::Struct(vec![ArrowField::new("value", DataType::Utf8, true)].into()),
-            true,
-        )
-        .with_metadata(HashMap::from([
-            (COMPUTED_COLUMN_META_KEY.to_string(), "true".to_string()),
-            (KIND_META_KEY.to_string(), SQL_KIND.to_string()),
-            (EXPRESSION_META_KEY.to_string(), "title".to_string()),
-        ]));
-        let mut fields = vec![title, ArrowField::new("body", DataType::Utf8, true)];
-        fields.extend(binding.outputs().iter().map(|output| {
-            let data_type = match output.arrow_type.as_str() {
-                "utf8" => DataType::Utf8,
-                "int64" => DataType::Int64,
-                other => panic!("unexpected fixture output type {other}"),
-            };
-            ArrowField::new(&output.output_name, data_type, true).with_metadata(
-                function_computed_column_metadata(
-                    binding.binding_id(),
-                    output.output_ordinal,
-                    &["title.value".into(), "body".into()],
-                ),
-            )
-        }));
-
-        ensure_binding_matches_schema(&ArrowSchema::new(fields), &binding).unwrap();
-    }
-
    #[test]
    fn test_function_binding_metadata_survives_schema_round_trip() {
        let binding = FunctionBinding::from_json(include_str!(
@@ -2800,36 +2661,9 @@ mod tests {
                output_ordinal: 1,
            } if binding_id == "fb_01K3TEXT"
        ));
-        let dependent_application = FunctionApplication::from_json(
-            r#"{
-                "function":{"name":"dependent","version":"fv_dependent"},
-                "inputs":[
-                    {"parameter":"text","kind":"column","value":{"path":"search_text"}}
-                ],
-                "output":{"kind":"scalar","arrow_type":"int64","nullable":false}
-            }"#,
-        )
-        .unwrap();
-        let err = plan_function_application(&reopened, &dependent_application, Some("dependent"))
+        let err = plan_function_application(&reopened, &named_struct_application("{}"), None)
            .unwrap_err();
-        assert!(
-            matches!(&err, Error::InvalidInput { message } if message.contains("computed-on-computed"))
-        );
-        let plan = plan_function_application(
-            &reopened,
-            &named_struct_application(
-                r#"{"normalized_text":"secondary_text","token_count":"secondary_token_count"}"#,
-            ),
-            None,
-        )
-        .unwrap();
-        assert_eq!(
-            plan.outputs
-                .iter()
-                .map(|output| output.output_name.as_str())
-                .collect::<Vec<_>>(),
-            ["secondary_text", "secondary_token_count"]
-        );
+        assert!(matches!(err, Error::NotSupported { .. }));
    }

    #[test]
@@ -2985,38 +2819,5 @@ mod tests {
        assert!(
            matches!(&err, Error::InvalidInput { message } if message.contains("computed-on-computed"))
        );
-
-        let nested_title = ArrowField::new(
-            "title",
-            DataType::Struct(vec![ArrowField::new("value", DataType::Utf8, true)].into()),
-            true,
-        )
-        .with_metadata(HashMap::from([
-            (COMPUTED_COLUMN_META_KEY.to_string(), "true".to_string()),
-            (KIND_META_KEY.to_string(), SQL_KIND.to_string()),
-            (
-                EXPRESSION_META_KEY.to_string(),
-                "struct('value')".to_string(),
-            ),
-        ]));
-        let nested_schema = ArrowSchema::new(vec![nested_title, schema.field(1).as_ref().clone()]);
-        let nested_application = FunctionApplication::from_json(
-            r#"{
-                "function":{"name":"text_features","version":"fv_exact"},
-                "inputs":[
-                    {"parameter":"title","kind":"column","value":{"path":"title.value"}},
-                    {"parameter":"body","kind":"column","value":{"path":"body"}}
-                ],
-                "output":{"kind":"named_struct","fields":[
-                    {"name":"normalized_text","arrow_type":"utf8","nullable":false},
-                    {"name":"token_count","arrow_type":"int64","nullable":false}
-                ]}
-            }"#,
-        )
-        .unwrap();
-        let err = plan_function_application(&nested_schema, &nested_application, None).unwrap_err();
-        assert!(
-            matches!(&err, Error::InvalidInput { message } if message.contains("computed-on-computed"))
-        );
    }
 }
@@ -15,7 +15,7 @@ use lance_index::optimize::OptimizeOptions;
 use log::info;

 pub use chrono::Duration;
-pub use lance::dataset::optimize::CompactionOptions;
+pub use lance::dataset::optimize::{CompactionMode, CompactionOptions, IndexRemapMode};

 use super::NativeTable;
 use crate::error::Result;
Author	SHA1	Message	Date
Gatefixer	9bfd475702	Merge remote-tracking branch 'origin/main' into gatekeeper/fix-2325-1 # Conflicts: # python/python/lancedb/remote/table.py	2026-08-28 09:45:02 +00:00
Xuanwo	670bda8725	Merge branch 'main' into gatekeeper/fix-2325-1	2026-08-27 18:13:07 +08:00
Gatefixer	10151b2dc8	Merge remote-tracking branch 'origin/main' into gatekeeper/fix-2325-1	2026-08-22 07:07:55 +00:00
Gatefixer	53b4c3b715	Merge origin/main into gatekeeper/fix-2325-1 # Conflicts: # python/python/lancedb/__init__.py	2026-08-22 07:07:52 +00:00
Gatefixer	638430fdb4	Merge origin/main into gatekeeper/fix-2325-1	2026-08-21 09:23:38 +00:00
Gatefixer	2221b8df6a	Merge origin/main into gatekeeper/fix-2325-1	2026-08-21 08:56:03 +00:00
Gatefixer	14eeae4bd4	Merge origin/main into gatekeeper/fix-2325-1	2026-08-21 08:47:29 +00:00
Gatefixer	320755ed55	Merge origin/main into gatekeeper/fix-2325-1 # Conflicts: # python/python/lancedb/__init__.py	2026-08-21 08:13:23 +00:00
Gatefixer	e55c2da7b1	fix(python): clarify compaction limits	2026-08-20 11:45:51 +00:00
Gatefixer	d33b05328c	Merge remote-tracking branch 'origin/main' into gatekeeper/fix-2325-1	2026-08-20 11:09:16 +00:00
Gatefixer	82f5355b71	fix(python): expose compaction source limits	2026-08-14 22:41:06 +00:00
Gatefixer	40cff9b644	Merge origin/main into gatekeeper/fix-2325-1	2026-08-14 21:59:28 +00:00
Gatefixer	edf95e53fc	Merge origin/main into gatekeeper/fix-2325-1	2026-08-07 19:46:40 +00:00
Gatefixer	0b5eba085d	Merge origin/main into gatekeeper/fix-2325-1	2026-08-07 19:32:06 +00:00
Gatefixer	21bf859c0b	fix(python): validate compaction option bounds	2026-08-06 06:30:38 +00:00
Gatefixer	e0499de959	fix(python): expose optimize compaction options	2026-08-06 05:01:42 +00:00