Bump version: 0.35.0-beta.2 → 0.35.0-beta.3

feat(fts): add block size configuration (#3691 )
## What changed - add `block_size` to Python FTS configuration and the deprecated local/remote helpers - add `blockSize` to the TypeScript FTS options and propagate it through the NAPI binding - serialize the value as `block_size` for remote index creation - document the existing Rust builder API and generate the TypeScript API reference - add local, remote, metadata, search, and invalid-value regression coverage ## Why Lance supports configuring the number of documents per compressed FTS posting block, but LanceDB's Python and TypeScript APIs did not expose the setting. This made the experimental FTS V3 layout unavailable through those clients and allowed the value to be dropped before index creation. ## How it works The default remains `128`. Supported values are `128` and `256`; selecting `256` uses the experimental FTS V3 format. Invalid values are rejected by the Lance builder and surfaced as Python or JavaScript errors. ## Validation - `cargo check --quiet --features remote --tests --examples` - `cargo +1.94.0 clippy --quiet --features remote --tests --examples -- -D warnings` - targeted Rust local and remote index tests - Rust doctests: 34 passed - Python Ruff checks, doctest, and targeted local/remote tests: 5 passed - TypeScript build, Biome lint, generated docs, and targeted Jest tests: 9 passed - `git diff --check` ## Limitations The Java client remains unchanged because its external remote REST model does not currently expose `block_size`. Co-authored-by: Yang Cen <yangcen@Yangs-Mac-mini.local>
2026-08-26 07:58:31 +00:00 · 2026-07-24 22:03:30 +00:00 · 2026-07-24 15:02:38 -07:00 · 2026-07-24 15:02:23 -07:00 · 2026-07-24 15:01:30 -07:00 · 2026-07-22 14:11:57 -07:00
93 changed files with 4684 additions and 426 deletions
@@ -0,0 +1,20 @@
+{
+  "name": "lancedb",
+  "interface": {
+    "displayName": "LanceDB"
+  },
+  "plugins": [
+    {
+      "name": "lancedb",
+      "source": {
+        "source": "local",
+        "path": "./plugins/lancedb"
+      },
+      "policy": {
+        "installation": "AVAILABLE",
+        "authentication": "ON_INSTALL"
+      },
+      "category": "Developer Tools"
+    }
+  ]
+}
@@ -5,3 +5,7 @@ This directory contains repo-scoped code agent skills for the LanceDB project.
 Each skill is a folder that contains a required `SKILL.md` and optional bundled resources.

 Codex discovers skills from `.agents/skills` in the current working directory and parent directories.
+
+The `lancedb` skill lives in the `plugins/lancedb` plugin (see `plugins/lancedb/skills/lancedb`)
+so it can be installed via the plugin marketplaces (`.claude-plugin/marketplace.json` and
+`.agents/plugins/marketplace.json`); the `lancedb` entry here is a symlink into that plugin.
@@ -0,0 +1 @@
+../../plugins/lancedb/skills/lancedb
@@ -0,0 +1,19 @@
+{
+  "name": "lancedb",
+  "owner": {
+    "name": "LanceDB"
+  },
+  "description": "LanceDB plugins for Claude Code.",
+  "plugins": [
+    {
+      "name": "lancedb",
+      "source": "./plugins/lancedb",
+      "description": "Write, review, debug, and document LanceDB pipelines in Python and TypeScript that work across local LanceDB OSS tables and remote LanceDB Enterprise/Cloud tables.",
+      "version": "0.1.0",
+      "author": {
+        "name": "LanceDB"
+      },
+      "category": "development"
+    }
+  ]
+}
@@ -18,6 +18,14 @@ inputs:
    description: "The manylinux version to build for"
    required: false
    default: "2_17"
+  package-name:
+    description: "Override [project] name in python/pyproject.toml (e.g. 'lancedb-compat'). Default keeps 'lancedb'."
+    required: false
+    default: "lancedb"
+  rustflags:
+    description: "RUSTFLAGS for the build container, as a single whitespace-free token (e.g. '-Ctarget-cpu=x86-64-v2'). Empty leaves RUSTFLAGS unset, keeping the defaults from .cargo/config.toml."
+    required: false
+    default: ""
 runs:
  using: "composite"
  steps:
@@ -27,6 +35,18 @@ runs:
        ARM_BUILD: ${{ inputs.arm-build }}
      run: |
        echo "ARM BUILD: $ARM_BUILD"
+    - name: Patch package name for variant build
+      if: ${{ inputs.package-name != 'lancedb' }}
+      shell: bash
+      env:
+        PACKAGE_NAME: ${{ inputs.package-name }}
+      run: |
+        # Swap the [project] name so this build produces e.g. lancedb-compat
+        # wheels. The package still installs files under the lancedb/
+        # namespace -- import lancedb still works after pip install.
+        sed -i.bak 's/^name = "lancedb"$/name = "'"$PACKAGE_NAME"'"/' python/pyproject.toml
+        rm -f python/pyproject.toml.bak
+        grep '^name = ' python/pyproject.toml
    - name: Build x86_64 Manylinux wheel
      if: ${{ inputs.arm-build == 'false' }}
      uses: PyO3/maturin-action@v1
@@ -34,7 +54,7 @@ runs:
        maturin-version: "1.12.4"
        command: build
        working-directory: python
-        docker-options: "-e PIP_EXTRA_INDEX_URL='https://pypi.fury.io/lance-format/ https://pypi.fury.io/lancedb/' -e PROTOC=/usr/local/bin/protoc"
+        docker-options: "-e PIP_EXTRA_INDEX_URL='https://pypi.fury.io/lance-format/ https://pypi.fury.io/lancedb/' -e PROTOC=/usr/local/bin/protoc ${{ inputs.rustflags != '' && format('-e RUSTFLAGS={0}', inputs.rustflags) || '' }}"
        target: x86_64-unknown-linux-gnu
        manylinux: ${{ inputs.manylinux }}
        args: ${{ inputs.args }}
@@ -51,7 +71,7 @@ runs:
        maturin-version: "1.12.4"
        command: build
        working-directory: python
-        docker-options: "-e PIP_EXTRA_INDEX_URL='https://pypi.fury.io/lance-format/ https://pypi.fury.io/lancedb/' -e PROTOC=/usr/local/bin/protoc"
+        docker-options: "-e PIP_EXTRA_INDEX_URL='https://pypi.fury.io/lance-format/ https://pypi.fury.io/lancedb/' -e PROTOC=/usr/local/bin/protoc ${{ inputs.rustflags != '' && format('-e RUSTFLAGS={0}', inputs.rustflags) || '' }}"
        target: aarch64-unknown-linux-gnu
        manylinux: ${{ inputs.manylinux }}
        args: ${{ inputs.args }}
@@ -87,7 +87,7 @@ jobs:
          bash ci/update_lockfiles.sh --amend
      - name: Push new version tag
        if: ${{ !inputs.dry_run }}
-        uses: ad-m/github-push-action@master
+        uses: ad-m/github-push-action@881a6320fdb16eb5318c5054f31c218aec2b324c # v1.3.0
        with:
          # Need to use PAT here too to trigger next workflow. See comment above.
          github_token: ${{ secrets.LANCEDB_RELEASE_TOKEN }}
@@ -22,7 +22,7 @@ permissions:

 jobs:
  linux:
-    name: Python ${{ matrix.config.platform }} manylinux${{ matrix.config.manylinux }}
+    name: Python ${{ matrix.config.package_name }} ${{ matrix.config.platform }} manylinux${{ matrix.config.manylinux }}
    timeout-minutes: 60
    strategy:
      matrix:
@@ -31,11 +31,28 @@ jobs:
            manylinux: "2_28"
            extra_args: "--features fp16kernels"
            runner: ubuntu-22.04
+            package_name: "lancedb"
+            rustflags: ""
          # For successful fat LTO builds, we need a large runner to avoid OOM errors.
          - platform: aarch64
            manylinux: "2_28"
            extra_args: "--features fp16kernels"
            runner: ubuntu-2404-8x-arm64
+            package_name: "lancedb"
+            rustflags: ""
+          # `lancedb-compat`: pre-Haswell-friendly variant for x86_64 hosts
+          # without AVX2 (Sandy Bridge / Ivy Bridge / Westmere on Intel,
+          # Bulldozer / Piledriver / Steamroller on AMD). Compiled at the
+          # `x86-64-v2` baseline; runtime SIMD dispatch in lance-linalg
+          # picks the appropriate tier (scalar / AVX / AVX+FMA / AVX2+FMA
+          # / AVX-512) at load time. Same import as `lancedb` -- conflicts
+          # at install time, so users pick one.
+          - platform: x86_64
+            manylinux: "2_28"
+            extra_args: ""
+            runner: ubuntu-22.04
+            package_name: "lancedb-compat"
+            rustflags: "-Ctarget-cpu=x86-64-v2"
    runs-on: ${{ matrix.config.runner }}
    steps:
      - uses: actions/checkout@v6
@@ -52,11 +69,13 @@ jobs:
          args: "--release --strip ${{ matrix.config.extra_args }}"
          arm-build: ${{ matrix.config.platform == 'aarch64' }}
          manylinux: ${{ matrix.config.manylinux }}
+          package-name: ${{ matrix.config.package_name }}
+          rustflags: ${{ matrix.config.rustflags }}
      - uses: actions/upload-artifact@v7
        if: startsWith(github.ref, 'refs/tags/python-v')
        with:
-          name: wheels-linux-${{ matrix.config.platform }}-${{ matrix.config.manylinux }}
-          path: target/wheels/lancedb-*.whl
+          name: wheels-linux-${{ matrix.config.package_name }}-${{ matrix.config.platform }}-${{ matrix.config.manylinux }}
+          path: target/wheels/*.whl
          if-no-files-found: error
  mac:
    timeout-minutes: 90
@@ -145,7 +164,7 @@ jobs:
          FURY_TOKEN: ${{ secrets.FURY_TOKEN }}
        run: |
          shopt -s nullglob
-          WHEELS=(target/wheels/lancedb-*.whl)
+          WHEELS=(target/wheels/*.whl)
          if [[ ${#WHEELS[@]} -eq 0 ]]; then
            echo "No wheels found in target/wheels/" >&2
            exit 1
@@ -98,7 +98,7 @@ jobs:
          cargo build --profile ci --benches --all-features --tests

  linux:
-    timeout-minutes: 30
+    timeout-minutes: 60
    # To build all features, we need more disk space than is available
    # on the free OSS github runner. This is mostly due to the the
    # sentence-transformers feature.
@@ -158,7 +158,7 @@ jobs:
        run: CARGO_ARGS="--profile ci" make -C ./lancedb remote-tests

  macos:
-    timeout-minutes: 30
+    timeout-minutes: 60
    strategy:
      matrix:
        mac-runner: ["macos-14", "macos-15"]
@@ -157,9 +157,9 @@ dependencies = [

 [[package]]
 name = "anyhow"
-version = "1.0.103"
+version = "1.0.104"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2a4385e2e34eb35d6b3efe798b9eb88096925d87726c0798709bf56d9ed84af3"
+checksum = "330a5ed07fa54e4702c9d6c4174f74427fc0ef6e214bbd677ae50a5099946470"

 [[package]]
 name = "approx"
@@ -535,13 +535,13 @@ dependencies = [

 [[package]]
 name = "async-trait"
-version = "0.1.89"
+version = "0.1.91"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9035ad2d096bed7955a320ee7e2230574d28fd3c3a0f186cbea1ff3c7eed5dbb"
+checksum = "ae36dc4177970ef04fde5178d3e2429882def40e57a451f919c098f72baa6cec"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn 2.0.117",
+ "syn 3.0.3",
 ]

 [[package]]
@@ -1750,7 +1750,7 @@ version = "3.1.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "faf9468729b8cbcea668e36183cb69d317348c2e08e994829fb56ebfdfbaac34"
 dependencies = [
- "windows-sys 0.59.0",
+ "windows-sys 0.61.2",
 ]

 [[package]]
@@ -2288,9 +2288,9 @@ dependencies = [

 [[package]]
 name = "datafusion"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "997a31e15872606a49478e670c58302094c97cb96abb0a7d60720f8e92170040"
+checksum = "754ef4e8f073922a26f5b23133b9db4829342362b09be0bc94309cf261c2f098"
 dependencies = [
 "arrow",
 "arrow-schema",
@@ -2335,9 +2335,9 @@ dependencies = [

 [[package]]
 name = "datafusion-catalog"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f7dd61161508f8f5fa1107774ea687bd753c22d83a32eebf963549f89de14139"
+checksum = "06afd1e38dd27bbb1258685a1fc6524df6aff4e07b25b393a47de59635178d99"
 dependencies = [
 "arrow",
 "async-trait",
@@ -2360,9 +2360,9 @@ dependencies = [

 [[package]]
 name = "datafusion-catalog-listing"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "897c70f871277f9ce99aa38347be0d679bbe3e617156c4d2a8378cec8a2a0891"
+checksum = "f0668fb32c12065ec242be0e5b4bc62bd7a06a0be3ecd83791ef877e4be67e02"
 dependencies = [
 "arrow",
 "async-trait",
@@ -2383,9 +2383,9 @@ dependencies = [

 [[package]]
 name = "datafusion-common"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "121c9ded5d87d9172319e006f2afdb9928d72dbacd6a90a458d8acb1e3b43a65"
+checksum = "ca43b263cdff57042cfa8fb817fb3469f4878933380dccff25f5e793580abbf9"
 dependencies = [
 "arrow",
 "arrow-ipc",
@@ -2407,9 +2407,9 @@ dependencies = [

 [[package]]
 name = "datafusion-common-runtime"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "981b9dae74f78ee3d9f714fb49b01919eab975461b56149510c3ba9ea11287d1"
+checksum = "05f0ba2b864792bdca4d76c59a1de0ab6e1b61946596b9936888dbd6360035f2"
 dependencies = [
 "futures",
 "log",
@@ -2418,9 +2418,9 @@ dependencies = [

 [[package]]
 name = "datafusion-datasource"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ffd7d295b2ec7c00d8a56562f41ed41062cf0af75549ed891c12a0a09eddfefe"
+checksum = "b840a8bce0bcbf5afad02946d438591e7c373f7afccaf3d874c04485772514dd"
 dependencies = [
 "arrow",
 "async-trait",
@@ -2448,9 +2448,9 @@ dependencies = [

 [[package]]
 name = "datafusion-datasource-arrow"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "552b0b3f342f7ec41b3fbd70f6339dc82a30cfd0349e7f280e7852528085349f"
+checksum = "a24cc0b9cf6e367f27f27406eff13abf48a11b72446aaa40b3105c0ded5c17d9"
 dependencies = [
 "arrow",
 "arrow-ipc",
@@ -2472,9 +2472,9 @@ dependencies = [

 [[package]]
 name = "datafusion-datasource-csv"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "68850aa426b897e879c8b87e512ea8124f1d0a2869a4e51808ddaaddf1bc0ada"
+checksum = "e1abe56b2a7a2d1d6de5117dd1a203181e267f28529faa5da546947621b697d7"
 dependencies = [
 "arrow",
 "async-trait",
@@ -2495,9 +2495,9 @@ dependencies = [

 [[package]]
 name = "datafusion-datasource-json"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "402f93242ae08ef99139ee2c528a49d087efe88d5c7b2c3ff5480855a40ce54f"
+checksum = "9c3e467f0611ad7bdd5aad17c63c9bb6182d04e5282e5496d897ea2b49c024ba"
 dependencies = [
 "arrow",
 "async-trait",
@@ -2518,15 +2518,15 @@ dependencies = [

 [[package]]
 name = "datafusion-doc"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "cb9e7e5d11130c48c8bd4e80c79a9772dd28ce6dc330baca9246205d245b9e2e"
+checksum = "d69bb69d8769e34f76839c960dbde24c1ac0c885a79b6c3c2287bdc56ec67891"

 [[package]]
 name = "datafusion-execution"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "37a8643ab852eb68864e1b72ae789e8066282dce48eea6347ffb0aee33d1ccc0"
+checksum = "d8eac0a09bc8d263f52025cad9e001da4d8138d633fa288edda4d06b1772eae6"
 dependencies = [
 "arrow",
 "arrow-buffer",
@@ -2546,9 +2546,9 @@ dependencies = [

 [[package]]
 name = "datafusion-expr"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6932f4d71eed9c8d9341476a2b845aadfabde5495d08dbcd8fc23881f49fa7a0"
+checksum = "eeb14d374767ee0fc62dc79a5ba8bcf8a63c14e993c7d992d0e63adfa23d77d3"
 dependencies = [
 "arrow",
 "arrow-schema",
@@ -2568,9 +2568,9 @@ dependencies = [

 [[package]]
 name = "datafusion-expr-common"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0225491839a31b1f7d2cb8092c2d50792e2fe1c1724e4e6d08e011f5feaf4ed2"
+checksum = "b7b19a8c95522bee8cbb313d74263b85e355d2b52f42e67ef5694bf5de9e9356"
 dependencies = [
 "arrow",
 "datafusion-common",
@@ -2580,9 +2580,9 @@ dependencies = [

 [[package]]
 name = "datafusion-functions"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "14872c47bfc3d21e53ec82f57074e6987a15941c1e2f43cde4ac6ae2746634e3"
+checksum = "5f64c983bbbdcb729d921a2b2ac3375598719b5cc0c30345ad664936f3176fc7"
 dependencies = [
 "arrow",
 "arrow-buffer",
@@ -2612,9 +2612,9 @@ dependencies = [

 [[package]]
 name = "datafusion-functions-aggregate"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "75a2ca14e1b609be21e657e2d3130b2f446456b08393b377bb721a33952d2e09"
+checksum = "89bc17041e424a47ed062f43df24d84aab8b57c4c3221e5c1a5eef46d6c5718b"
 dependencies = [
 "arrow",
 "datafusion-common",
@@ -2633,9 +2633,9 @@ dependencies = [

 [[package]]
 name = "datafusion-functions-aggregate-common"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1ece74ba09092d2ef9c9b54a38445450aea292a1f8b04faf531936b723a24b3c"
+checksum = "97dd2a9e865c6108059f5b37b77934f84b50bfb108f837bd0e5c9536e03f0545"
 dependencies = [
 "arrow",
 "datafusion-common",
@@ -2645,9 +2645,9 @@ dependencies = [

 [[package]]
 name = "datafusion-functions-nested"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3f3e3f9ee8ca59bf70518802107de6f1b88a9509efdc629fadc5de9d6b2d5ef5"
+checksum = "75f0bdfeef16d96417b9632ef855645376b242e9006a126dfd0bedfc54a93f5f"
 dependencies = [
 "arrow",
 "arrow-ord",
@@ -2670,9 +2670,9 @@ dependencies = [

 [[package]]
 name = "datafusion-functions-table"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "89161dffc22cf2b50f9f4b1bee83b5221d3b4ed7c2e37fd7aa2b22a5297b3a26"
+checksum = "f4e4941673c917819616877e9993da4503e4f4739812be0bc32c5356184c6383"
 dependencies = [
 "arrow",
 "async-trait",
@@ -2686,9 +2686,9 @@ dependencies = [

 [[package]]
 name = "datafusion-functions-window"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d7339345b226b3874037708bf5023ba1c2de705128f8457a095aae5ae9cb9c78"
+checksum = "12dd2e16c12b84b6f6b41b19f55b366dd1c46876bb35b86896c6349067379e8d"
 dependencies = [
 "arrow",
 "datafusion-common",
@@ -2703,9 +2703,9 @@ dependencies = [

 [[package]]
 name = "datafusion-functions-window-common"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fa84836dc2392df6f43d6a29d37fb56a8ebdc8b3f4e10ae8dc15861fd20278fb"
+checksum = "4cdc5e4b6f8b6ef823cc1c761f85088ad4c884fe8df64df3cbcc6b2b84698441"
 dependencies = [
 "datafusion-common",
 "datafusion-physical-expr-common",
@@ -2713,9 +2713,9 @@ dependencies = [

 [[package]]
 name = "datafusion-macros"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "587164e03ad68732aa9e7bfe5686e3f25970d4c64fd4bd80790749840892dae5"
+checksum = "1a3614234dd93578c92428cb4f408e020874f0d2b7e6c90c928d9d28b5df2ceb"
 dependencies = [
 "datafusion-doc",
 "quote",
@@ -2724,9 +2724,9 @@ dependencies = [

 [[package]]
 name = "datafusion-optimizer"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "77f20e8cf9e8654d92f4c16b24c487353ee5bf153ffc12d5772cd399ab8cd281"
+checksum = "a0635620b050b81bb92764e99250868f654e2cd5ad1bece413283b3f73c83179"
 dependencies = [
 "arrow",
 "chrono",
@@ -2743,9 +2743,9 @@ dependencies = [

 [[package]]
 name = "datafusion-physical-expr"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f015a4a82f6f7ff7e1d8d4bf3870a936752fa38b17705dfcc14adef95aa8922c"
+checksum = "8cabf7a86eb70b816729e33c81bf7767c936ee1226f607a114f5dac2decac8d0"
 dependencies = [
 "arrow",
 "datafusion-common",
@@ -2764,9 +2764,9 @@ dependencies = [

 [[package]]
 name = "datafusion-physical-expr-adapter"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "51e6ffff8acdfe54e0ea15ccf38115c4a9184433b0439f42907637928d00a235"
+checksum = "de222e04f7e6744501555a54ab0abe26bfdfebee380af79a9bdc175704246859"
 dependencies = [
 "arrow",
 "datafusion-common",
@@ -2779,9 +2779,9 @@ dependencies = [

 [[package]]
 name = "datafusion-physical-expr-common"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7967a3e171c6a4bf09474b3f7a14f1a3db13ed1714ba12156f33fcce2bba54e8"
+checksum = "72d0d0057fc5a502d45c870cb6d47c66eb7bdd5edb1bd71ad6f3f724975ac2a8"
 dependencies = [
 "arrow",
 "chrono",
@@ -2796,9 +2796,9 @@ dependencies = [

 [[package]]
 name = "datafusion-physical-optimizer"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "59ff803e2a96054cb6d83f35f9e60fd4f42eac515e1932bd1b2dbc91d5fcbf36"
+checksum = "86046eed10950c5f9aaed9acfd148e9bd2e1dfdfe4f9aef607d1447b271e4183"
 dependencies = [
 "arrow",
 "datafusion-common",
@@ -2814,9 +2814,9 @@ dependencies = [

 [[package]]
 name = "datafusion-physical-plan"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "776ee54d47d15bdb126452f9ca17b03761e3b004682914beaedd3f86eb507fbc"
+checksum = "9bc84da934c903407ba297971ebcc020c4c1a38aafd765d6c144c76eff3fa6a1"
 dependencies = [
 "arrow",
 "arrow-data",
@@ -2847,9 +2847,9 @@ dependencies = [

 [[package]]
 name = "datafusion-pruning"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d5fb9e5774660aa69c3ba93c610f175f75b65cb8c3776edb3626de8f3a4f4ee3"
+checksum = "eb63eeac6de19be40f487b65dd84e546195f783c5a9928618e0c4f2a3569b0d7"
 dependencies = [
 "arrow",
 "datafusion-common",
@@ -2863,9 +2863,9 @@ dependencies = [

 [[package]]
 name = "datafusion-session"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "15ce715fa2a61f4623cc234bcc14a3ef6a91f189128d5b14b468a6a17cdfc417"
+checksum = "5f961d209177f91bd014db5cbb2c33b7d28a2597b9003e77f17aeb712964315a"
 dependencies = [
 "async-trait",
 "datafusion-common",
@@ -2877,9 +2877,9 @@ dependencies = [

 [[package]]
 name = "datafusion-sql"
-version = "54.0.0"
+version = "54.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6094ad36a3ed6d7ac87b20b479b2d0b118250f66cf997603829fdc65b44a7099"
+checksum = "b1d71cb454da682b2af7488e1fc1ddd72ee1b28f19297b8ccad73f0a21ee9a69"
 dependencies = [
 "arrow",
 "bigdecimal",
@@ -3006,7 +3006,7 @@ dependencies = [
 "libc",
 "option-ext",
 "redox_users",
- "windows-sys 0.59.0",
+ "windows-sys 0.61.2",
 ]

 [[package]]
@@ -3229,7 +3229,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "39cab71617ae0d63f51a36d69f866391735b51691dbda63cf6f96d042b63efeb"
 dependencies = [
 "libc",
- "windows-sys 0.59.0",
+ "windows-sys 0.61.2",
 ]

 [[package]]
@@ -3421,8 +3421,8 @@ checksum = "42703706b716c37f96a77aea830392ad231f44c9e9a67872fa5548707e11b11c"

 [[package]]
 name = "fsst"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow-array",
 "rand 0.9.5",
@@ -4218,7 +4218,7 @@ dependencies = [
 "libc",
 "percent-encoding",
 "pin-project-lite",
- "socket2 0.5.10",
+ "socket2 0.6.3",
 "system-configuration",
 "tokio",
 "tower-service",
@@ -4517,7 +4517,7 @@ checksum = "3640c1c38b8e4e43584d8df18be5fc6b0aa314ce6ebf51b53313d4306cca8e46"
 dependencies = [
 "hermit-abi",
 "libc",
- "windows-sys 0.59.0",
+ "windows-sys 0.61.2",
 ]

 [[package]]
@@ -4611,7 +4611,7 @@ dependencies = [
 "portable-atomic-util",
 "serde_core",
 "wasm-bindgen",
- "windows-sys 0.59.0",
+ "windows-sys 0.61.2",
 ]

 [[package]]
@@ -4777,8 +4777,8 @@ checksum = "e037a2e1d8d5fdbd49b16a4ea09d5d6401c1f29eca5ff29d03d3824dba16256a"

 [[package]]
 name = "lance"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arc-swap",
 "arrow",
@@ -4852,8 +4852,8 @@ dependencies = [

 [[package]]
 name = "lance-arrow"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow-array",
 "arrow-buffer",
@@ -4875,7 +4875,7 @@ dependencies = [
 [[package]]
 name = "lance-arrow-scalar"
 version = "58.0.0"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow-array",
 "arrow-buffer",
@@ -4889,7 +4889,7 @@ dependencies = [
 [[package]]
 name = "lance-arrow-stats"
 version = "58.0.0"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow-array",
 "arrow-schema",
@@ -4898,8 +4898,8 @@ dependencies = [

 [[package]]
 name = "lance-bitpacking"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrayref",
 "crunchy",
@@ -4909,8 +4909,8 @@ dependencies = [

 [[package]]
 name = "lance-core"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow-array",
 "arrow-buffer",
@@ -4948,8 +4948,8 @@ dependencies = [

 [[package]]
 name = "lance-datafusion"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow",
 "arrow-array",
@@ -4979,8 +4979,8 @@ dependencies = [

 [[package]]
 name = "lance-datagen"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow",
 "arrow-array",
@@ -4997,8 +4997,8 @@ dependencies = [

 [[package]]
 name = "lance-derive"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "proc-macro2",
 "quote",
@@ -5007,8 +5007,8 @@ dependencies = [

 [[package]]
 name = "lance-encoding"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow-arith",
 "arrow-array",
@@ -5043,8 +5043,8 @@ dependencies = [

 [[package]]
 name = "lance-file"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow-arith",
 "arrow-array",
@@ -5074,13 +5074,14 @@ dependencies = [

 [[package]]
 name = "lance-index"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arc-swap",
 "arrow",
 "arrow-arith",
 "arrow-array",
+ "arrow-ipc",
 "arrow-ord",
 "arrow-schema",
 "arrow-select",
@@ -5141,8 +5142,8 @@ dependencies = [

 [[package]]
 name = "lance-index-core"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow-array",
 "arrow-schema",
@@ -5164,8 +5165,8 @@ dependencies = [

 [[package]]
 name = "lance-io"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow",
 "arrow-arith",
@@ -5208,8 +5209,8 @@ dependencies = [

 [[package]]
 name = "lance-linalg"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow-array",
 "arrow-buffer",
@@ -5225,8 +5226,8 @@ dependencies = [

 [[package]]
 name = "lance-namespace"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow",
 "async-trait",
@@ -5238,8 +5239,8 @@ dependencies = [

 [[package]]
 name = "lance-namespace-impls"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow",
 "arrow-ipc",
@@ -5293,8 +5294,8 @@ dependencies = [

 [[package]]
 name = "lance-select"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow-array",
 "arrow-buffer",
@@ -5309,8 +5310,8 @@ dependencies = [

 [[package]]
 name = "lance-table"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow",
 "arrow-array",
@@ -5349,8 +5350,8 @@ dependencies = [

 [[package]]
 name = "lance-testing"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "arrow-array",
 "arrow-schema",
@@ -5363,8 +5364,8 @@ dependencies = [

 [[package]]
 name = "lance-tokenizer"
-version = "9.0.0-rc.1"
-source = "git+https://github.com/lance-format/lance.git?tag=v9.0.0-rc.1#cec0b7dffe2d85c7e66dbe9d1f3891c297903a1d"
+version = "10.0.0-beta.3"
+source = "git+https://github.com/lance-format/lance.git?tag=v10.0.0-beta.3#ed078e7c19c9560906e1eae54b0baf01dc509986"
 dependencies = [
 "icu_segmenter",
 "jieba-rs",
@@ -5432,7 +5433,6 @@ dependencies = [
 "lance-namespace-impls",
 "lance-table",
 "lance-testing",
- "lazy_static",
 "log",
 "metrics",
 "metrics-util",
@@ -5591,9 +5591,9 @@ dependencies = [

 [[package]]
 name = "libc"
-version = "0.2.186"
+version = "0.2.189"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "68ab91017fe16c622486840e4c83c9a37afeff978bd239b5293d61ece587de66"
+checksum = "3eaf3ede3fee6db1a4c2ee091bf8a8b4dccdc6d17f656fb07896ee72867612f2"

 [[package]]
 name = "libloading"
@@ -6064,9 +6064,9 @@ dependencies = [

 [[package]]
 name = "napi"
-version = "3.10.5"
+version = "3.11.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6826e5ddc15589b2d68c8ad5321c18e85d40488e93e32962f362e572669bccf6"
+checksum = "de33522036981030a75c231829566bc63414e08101a6f5ff4ac6cef19c8e0941"
 dependencies = [
 "bitflags 2.11.1",
 "chrono",
@@ -6089,9 +6089,9 @@ checksum = "c9c366d2c8c60b86fa632df75f745509b52f9128f91a6bad4c796e44abb505e1"

 [[package]]
 name = "napi-derive"
-version = "3.5.10"
+version = "3.6.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b0fe526e81c105d3640516fcde83909dd1afe757c0d7a15af58830b5bc0fb9a1"
+checksum = "a49c513341a61a16a10af6efcce46b30d0822ba2d4fb197d24d33dfc199c78d5"
 dependencies = [
 "convert_case",
 "ctor 1.0.5",
@@ -6103,9 +6103,9 @@ dependencies = [

 [[package]]
 name = "napi-derive-backend"
-version = "5.1.2"
+version = "6.0.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "514281397bcddd9ea9a876c7a21a57bff2374237a000ca9a64ea0211ec1993e2"
+checksum = "4747005fa3e2c9989ac45a723a514c5db2411238b72981a3cda4c701a9dfea17"
 dependencies = [
 "convert_case",
 "proc-macro2",
@@ -6116,9 +6116,9 @@ dependencies = [

 [[package]]
 name = "napi-sys"
-version = "3.2.3"
+version = "3.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "73e43cf2eb0bd1bf95a43c07c076ebd2da5d1e015a71c3d201faeffffcc0ecac"
+checksum = "85fbf1fa9f1babfe396d74bbbf52b3643770243e8f5b0b46715d4caf7f0dfc9a"
 dependencies = [
 "libloading",
 ]
@@ -6207,7 +6207,7 @@ version = "0.50.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "7957b9740744892f114936ab4a57b3f487491bbeafaf8083688b16841a4240e5"
 dependencies = [
- "windows-sys 0.59.0",
+ "windows-sys 0.61.2",
 ]

 [[package]]
@@ -7586,8 +7586,8 @@ version = "0.14.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "343d3bd7056eda839b03204e68deff7d1b13aba7af2b2fd16890697274262ee7"
 dependencies = [
- "heck 0.4.1",
- "itertools 0.11.0",
+ "heck 0.5.0",
+ "itertools 0.14.0",
 "log",
 "multimap",
 "petgraph",
@@ -7606,7 +7606,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "27c6023962132f4b30eb4c172c91ce92d933da334c59c23cddee82358ddafb0b"
 dependencies = [
 "anyhow",
- "itertools 0.11.0",
+ "itertools 0.14.0",
 "proc-macro2",
 "quote",
 "syn 2.0.117",
@@ -7815,7 +7815,7 @@ dependencies = [
 "quinn-udp",
 "rustc-hash",
 "rustls 0.23.40",
- "socket2 0.5.10",
+ "socket2 0.6.3",
 "thiserror 2.0.18",
 "tokio",
 "tracing",
@@ -7853,9 +7853,9 @@ dependencies = [
 "cfg_aliases",
 "libc",
 "once_cell",
- "socket2 0.5.10",
+ "socket2 0.6.3",
 "tracing",
- "windows-sys 0.59.0",
+ "windows-sys 0.60.2",
 ]

 [[package]]
@@ -8161,9 +8161,9 @@ dependencies = [

 [[package]]
 name = "regex"
-version = "1.13.0"
+version = "1.13.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2a0e75113e14dc5acb068cd0786884f214f1312650a3d36d269f5c4f3cdee8a2"
+checksum = "f020237b6c8eed93db2e2cb53c00c60a8e1bc73da7d073199a1180401450218d"
 dependencies = [
 "aho-corasick",
 "memchr",
@@ -8173,9 +8173,9 @@ dependencies = [

 [[package]]
 name = "regex-automata"
-version = "0.4.14"
+version = "0.4.16"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6e1dd4122fc1595e8162618945476892eefca7b88c52820e74af6262213cae8f"
+checksum = "8fcfdb36bda0c880c5931cdc7a2bcdc8ba4556847b9d912bca70bc94708711ad"
 dependencies = [
 "aho-corasick",
 "memchr",
@@ -8629,7 +8629,7 @@ dependencies = [
 "errno",
 "libc",
 "linux-raw-sys",
- "windows-sys 0.59.0",
+ "windows-sys 0.61.2",
 ]

 [[package]]
@@ -8700,7 +8700,7 @@ dependencies = [
 "security-framework",
 "security-framework-sys",
 "webpki-root-certs",
- "windows-sys 0.59.0",
+ "windows-sys 0.61.2",
 ]

 [[package]]
@@ -8905,9 +8905,9 @@ checksum = "1bc711410fbe7399f390ca1c3b60ad0f53f80e95c5eb935e52268a0e2cd49acc"

 [[package]]
 name = "serde"
-version = "1.0.228"
+version = "1.0.229"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9a8e94ea7f378bd32cbbd37198a4a91436180c5bb472411e48b5ec2e2124ae9e"
+checksum = "4148590afebada386688f18773da617792bf2ef03ffc1e4cbd2b1d45b023e0ba"
 dependencies = [
 "serde_core",
 "serde_derive",
@@ -8915,29 +8915,29 @@ dependencies = [

 [[package]]
 name = "serde_core"
-version = "1.0.228"
+version = "1.0.229"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "41d385c7d4ca58e59fc732af25c3983b67ac852c1a25000afe1175de458b67ad"
+checksum = "67dca2c9c51e58a4791a4b1ed58308b39c64224d349a935ab5039aa360942a48"
 dependencies = [
 "serde_derive",
 ]

 [[package]]
 name = "serde_derive"
-version = "1.0.228"
+version = "1.0.229"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d540f220d3187173da220f885ab66608367b6574e925011a9353e4badda91d79"
+checksum = "e7a5d71263a5a7d47b41f6b3f06ba276f10cc18b0931f1799f710578e2309348"
 dependencies = [
 "proc-macro2",
 "quote",
- "syn 2.0.117",
+ "syn 3.0.3",
 ]

 [[package]]
 name = "serde_json"
-version = "1.0.150"
+version = "1.0.151"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e8014e44b4736ed0538adeecded0fce2a272f22dc9578a7eb6b2d9993c74cfb9"
+checksum = "c841b55ecdae098c80dcae9cf767f6f8a0c2cdb3416bbef72181df4d0fe73f14"
 dependencies = [
 "itoa",
 "memchr",
@@ -9268,7 +9268,7 @@ version = "0.8.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c1c97747dbf44bb1ca44a561ece23508e99cb592e862f22222dcf42f51d1e451"
 dependencies = [
- "heck 0.4.1",
+ "heck 0.5.0",
 "proc-macro2",
 "quote",
 "syn 2.0.117",
@@ -9280,7 +9280,7 @@ version = "0.9.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "54254b8531cafa275c5e096f62d48c81435d1015405a91198ddb11e967301d40"
 dependencies = [
- "heck 0.4.1",
+ "heck 0.5.0",
 "proc-macro2",
 "quote",
 "syn 2.0.117",
@@ -9596,6 +9596,17 @@ dependencies = [
 "unicode-ident",
 ]

+[[package]]
+name = "syn"
+version = "3.0.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "53e9bae58849f64dfa4f5d5ae372c8341f7305f82a3868709269343628b659a3"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "unicode-ident",
+]
+
 [[package]]
 name = "sync_wrapper"
 version = "1.0.2"
@@ -9713,7 +9724,7 @@ dependencies = [
 "getrandom 0.4.2",
 "once_cell",
 "rustix",
- "windows-sys 0.59.0",
+ "windows-sys 0.61.2",
 ]

 [[package]]
@@ -9916,9 +9927,9 @@ dependencies = [

 [[package]]
 name = "tokio"
-version = "1.52.3"
+version = "1.53.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8fc7f01b389ac15039e4dc9531aa973a135d7a4135281b12d7c1bc79fd57fffe"
+checksum = "202caea871b69668250d242070849eb495be178ed697a3e98aebce5bc81a0bed"
 dependencies = [
 "bytes",
 "libc",
@@ -10403,9 +10414,9 @@ checksum = "06abde3611657adf66d383f00b093d7faecc7fa57071cce2578660c9f1010821"

 [[package]]
 name = "uuid"
-version = "1.23.5"
+version = "1.24.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ea5fab0d6c3c01ae70085a09cb03d4c7a1d6314e2b3e075392783396d724ca0a"
+checksum = "bf3923a6f5c4c6382e0b653c4117f48d631ea17f38ed86e2a828e6f7412f5239"
 dependencies = [
 "getrandom 0.4.2",
 "js-sys",
@@ -10690,7 +10701,7 @@ version = "0.1.11"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c2a7b1c03c876122aa43f3020e6c3c3ee5c05081c9a00739faf7503aeba10d22"
 dependencies = [
- "windows-sys 0.59.0",
+ "windows-sys 0.61.2",
 ]

 [[package]]
@@ -13,20 +13,20 @@ categories = ["database-implementations"]
 rust-version = "1.91.0"

 [workspace.dependencies]
-lance = { "version" = "=9.0.0-rc.1", default-features = false, "tag" = "v9.0.0-rc.1", "git" = "https://github.com/lance-format/lance.git" }
-lance-core = { "version" = "=9.0.0-rc.1", "tag" = "v9.0.0-rc.1", "git" = "https://github.com/lance-format/lance.git" }
-lance-datagen = { "version" = "=9.0.0-rc.1", "tag" = "v9.0.0-rc.1", "git" = "https://github.com/lance-format/lance.git" }
-lance-file = { "version" = "=9.0.0-rc.1", "tag" = "v9.0.0-rc.1", "git" = "https://github.com/lance-format/lance.git" }
-lance-io = { "version" = "=9.0.0-rc.1", default-features = false, "tag" = "v9.0.0-rc.1", "git" = "https://github.com/lance-format/lance.git" }
-lance-index = { "version" = "=9.0.0-rc.1", "tag" = "v9.0.0-rc.1", "git" = "https://github.com/lance-format/lance.git" }
-lance-linalg = { "version" = "=9.0.0-rc.1", "tag" = "v9.0.0-rc.1", "git" = "https://github.com/lance-format/lance.git" }
-lance-namespace = { "version" = "=9.0.0-rc.1", "tag" = "v9.0.0-rc.1", "git" = "https://github.com/lance-format/lance.git" }
-lance-namespace-impls = { "version" = "=9.0.0-rc.1", default-features = false, "tag" = "v9.0.0-rc.1", "git" = "https://github.com/lance-format/lance.git" }
-lance-table = { "version" = "=9.0.0-rc.1", "tag" = "v9.0.0-rc.1", "git" = "https://github.com/lance-format/lance.git" }
-lance-testing = { "version" = "=9.0.0-rc.1", "tag" = "v9.0.0-rc.1", "git" = "https://github.com/lance-format/lance.git" }
-lance-datafusion = { "version" = "=9.0.0-rc.1", "tag" = "v9.0.0-rc.1", "git" = "https://github.com/lance-format/lance.git" }
-lance-encoding = { "version" = "=9.0.0-rc.1", "tag" = "v9.0.0-rc.1", "git" = "https://github.com/lance-format/lance.git" }
-lance-arrow = { "version" = "=9.0.0-rc.1", "tag" = "v9.0.0-rc.1", "git" = "https://github.com/lance-format/lance.git" }
+lance = { "version" = "=10.0.0-beta.3", default-features = false, "tag" = "v10.0.0-beta.3", "git" = "https://github.com/lance-format/lance.git" }
+lance-core = { "version" = "=10.0.0-beta.3", "tag" = "v10.0.0-beta.3", "git" = "https://github.com/lance-format/lance.git" }
+lance-datagen = { "version" = "=10.0.0-beta.3", "tag" = "v10.0.0-beta.3", "git" = "https://github.com/lance-format/lance.git" }
+lance-file = { "version" = "=10.0.0-beta.3", "tag" = "v10.0.0-beta.3", "git" = "https://github.com/lance-format/lance.git" }
+lance-io = { "version" = "=10.0.0-beta.3", default-features = false, "tag" = "v10.0.0-beta.3", "git" = "https://github.com/lance-format/lance.git" }
+lance-index = { "version" = "=10.0.0-beta.3", "tag" = "v10.0.0-beta.3", "git" = "https://github.com/lance-format/lance.git" }
+lance-linalg = { "version" = "=10.0.0-beta.3", "tag" = "v10.0.0-beta.3", "git" = "https://github.com/lance-format/lance.git" }
+lance-namespace = { "version" = "=10.0.0-beta.3", "tag" = "v10.0.0-beta.3", "git" = "https://github.com/lance-format/lance.git" }
+lance-namespace-impls = { "version" = "=10.0.0-beta.3", default-features = false, "tag" = "v10.0.0-beta.3", "git" = "https://github.com/lance-format/lance.git" }
+lance-table = { "version" = "=10.0.0-beta.3", "tag" = "v10.0.0-beta.3", "git" = "https://github.com/lance-format/lance.git" }
+lance-testing = { "version" = "=10.0.0-beta.3", "tag" = "v10.0.0-beta.3", "git" = "https://github.com/lance-format/lance.git" }
+lance-datafusion = { "version" = "=10.0.0-beta.3", "tag" = "v10.0.0-beta.3", "git" = "https://github.com/lance-format/lance.git" }
+lance-encoding = { "version" = "=10.0.0-beta.3", "tag" = "v10.0.0-beta.3", "git" = "https://github.com/lance-format/lance.git" }
+lance-arrow = { "version" = "=10.0.0-beta.3", "tag" = "v10.0.0-beta.3", "git" = "https://github.com/lance-format/lance.git" }
 ahash = "0.8"
 # Note that this one does not include pyarrow
 arrow = { version = "58.0.0", optional = false }
@@ -64,7 +64,6 @@ snafu = "0.8"
 url = "2"
 num-traits = "0.2"
 regex = "1.10"
-lazy_static = "1"
 semver = "1.0.25"
 chrono = "0.4"

@@ -249,6 +249,57 @@ try (BufferAllocator allocator = new RootAllocator();
 }
 ```

+### Creating an Empty Table
+
+To create an empty table, send an Arrow IPC stream that contains the table schema and no record batches.
+The schema in the IPC stream becomes the table schema, and rows can be inserted later.
+
+```java
+import org.lance.namespace.model.CreateTableRequest;
+import org.lance.namespace.model.CreateTableResponse;
+import org.apache.arrow.memory.BufferAllocator;
+import org.apache.arrow.memory.RootAllocator;
+import org.apache.arrow.vector.VectorSchemaRoot;
+import org.apache.arrow.vector.ipc.ArrowStreamWriter;
+import org.apache.arrow.vector.types.FloatingPointPrecision;
+import org.apache.arrow.vector.types.pojo.ArrowType;
+import org.apache.arrow.vector.types.pojo.Field;
+import org.apache.arrow.vector.types.pojo.FieldType;
+import org.apache.arrow.vector.types.pojo.Schema;
+
+import java.io.ByteArrayOutputStream;
+import java.nio.channels.Channels;
+import java.util.Arrays;
+
+Schema schema = new Schema(Arrays.asList(
+    new Field("id", FieldType.nullable(new ArrowType.Int(32, true)), null),
+    new Field("name", FieldType.nullable(new ArrowType.Utf8()), null),
+    new Field("embedding",
+        FieldType.nullable(new ArrowType.FixedSizeList(128)),
+        Arrays.asList(new Field("item",
+            FieldType.nullable(new ArrowType.FloatingPoint(FloatingPointPrecision.SINGLE)),
+            null)))
+));
+
+byte[] emptyTableData;
+try (BufferAllocator allocator = new RootAllocator();
+     VectorSchemaRoot root = VectorSchemaRoot.create(schema, allocator)) {
+    root.setRowCount(0);
+
+    ByteArrayOutputStream out = new ByteArrayOutputStream();
+    try (ArrowStreamWriter writer = new ArrowStreamWriter(root, null, Channels.newChannel(out))) {
+        writer.start();
+        writer.end();
+    }
+    emptyTableData = out.toByteArray();
+}
+
+CreateTableRequest request = new CreateTableRequest();
+request.setId(Arrays.asList("my_namespace", "empty_table"));
+
+CreateTableResponse response = namespaceClient.createTable(request, emptyTableData);
+```
+
 ### Insert

 ```java
@@ -431,9 +482,88 @@ query.setVector(vector);
 byte[] result = namespaceClient.queryTable(query);
 ```

-### Reading Query Results
+## Indexing

-Query results are returned in Apache Arrow IPC file format. Here's how to read them:
+The Java SDK exposes the REST namespace index operations through the same `LanceNamespace` client.
+Index creation runs asynchronously, so use `listTableIndices` or `describeTableIndexStats` to check progress.
+
+### Creating a Vector Index
+
+```java
+import org.lance.namespace.model.CreateTableIndexRequest;
+import org.lance.namespace.model.CreateTableIndexResponse;
+
+CreateTableIndexRequest request = new CreateTableIndexRequest();
+request.setId(Arrays.asList("my_namespace", "my_table"));
+request.setColumn("embedding");
+request.setIndexType("IVF_PQ");
+request.setDistanceType("cosine");
+request.setName("embedding_idx");
+
+CreateTableIndexResponse response = namespaceClient.createTableIndex(request);
+System.out.println("Index transaction: " + response.getTransactionId());
+```
+
+### Creating a Scalar Index
+
+```java
+import org.lance.namespace.model.CreateTableIndexRequest;
+import org.lance.namespace.model.CreateTableScalarIndexResponse;
+
+CreateTableIndexRequest request = new CreateTableIndexRequest();
+request.setId(Arrays.asList("my_namespace", "my_table"));
+request.setColumn("category");
+request.setIndexType("BTREE");
+request.setName("category_idx");
+
+CreateTableScalarIndexResponse response = namespaceClient.createTableScalarIndex(request);
+System.out.println("Index transaction: " + response.getTransactionId());
+```
+
+### Creating a Full Text Search Index
+
+```java
+import org.lance.namespace.model.CreateTableIndexRequest;
+import org.lance.namespace.model.CreateTableScalarIndexResponse;
+
+CreateTableIndexRequest request = new CreateTableIndexRequest();
+request.setId(Arrays.asList("my_namespace", "my_table"));
+request.setColumn("text_column");
+request.setIndexType("FTS");
+request.setName("text_idx");
+request.setBaseTokenizer("simple");
+request.setLowerCase(true);
+request.setWithPosition(true);
+
+CreateTableScalarIndexResponse response = namespaceClient.createTableScalarIndex(request);
+System.out.println("Index transaction: " + response.getTransactionId());
+```
+
+### Listing Indexes
+
+```java
+import org.lance.namespace.model.IndexContent;
+import org.lance.namespace.model.ListTableIndicesRequest;
+import org.lance.namespace.model.ListTableIndicesResponse;
+
+ListTableIndicesRequest request = new ListTableIndicesRequest();
+request.setId(Arrays.asList("my_namespace", "my_table"));
+
+ListTableIndicesResponse response = namespaceClient.listTableIndices(request);
+for (IndexContent index : response.getIndexes()) {
+    System.out.println(index.getIndexName() + ": " + index.getStatus());
+}
+```
+
+!!! note
+    The current Java namespace API exposes index type, index name, distance type, and full text search tokenizer options.
+    IVF training parameters such as `num_partitions` are not exposed by `CreateTableIndexRequest` yet.
+    To make those configurable from Java, the namespace API must add those fields first.
+
+## Reading Query Results
+
+Query results are returned as bytes in Apache Arrow IPC file format. Put the byte-channel
+adapter behind a small helper so query code can work with `ArrowFileReader` directly:

 ```java
 import org.apache.arrow.vector.ipc.ArrowFileReader;
@@ -441,45 +571,50 @@ import org.apache.arrow.vector.VectorSchemaRoot;
 import org.apache.arrow.memory.BufferAllocator;
 import org.apache.arrow.memory.RootAllocator;

+import java.io.IOException;
 import java.nio.ByteBuffer;
 import java.nio.channels.SeekableByteChannel;

-// Helper class to read Arrow data from byte array
-class ByteArraySeekableByteChannel implements SeekableByteChannel {
-    private final byte[] data;
-    private long position = 0;
-    private boolean isOpen = true;
-
-    public ByteArraySeekableByteChannel(byte[] data) {
-        this.data = data;
+final class ArrowIpc {
+    static ArrowFileReader openFileReader(byte[] data, BufferAllocator allocator) throws IOException {
+        return new ArrowFileReader(new ByteArraySeekableByteChannel(data), allocator);
    }

-    @Override
-    public int read(ByteBuffer dst) {
-        int remaining = dst.remaining();
-        int available = (int) (data.length - position);
-        if (available <= 0) return -1;
-        int toRead = Math.min(remaining, available);
-        dst.put(data, (int) position, toRead);
-        position += toRead;
-        return toRead;
-    }
+    private static final class ByteArraySeekableByteChannel implements SeekableByteChannel {
+        private final byte[] data;
+        private long position = 0;
+        private boolean isOpen = true;

-    @Override public long position() { return position; }
-    @Override public SeekableByteChannel position(long newPosition) { position = newPosition; return this; }
-    @Override public long size() { return data.length; }
-    @Override public boolean isOpen() { return isOpen; }
-    @Override public void close() { isOpen = false; }
-    @Override public int write(ByteBuffer src) { throw new UnsupportedOperationException(); }
-    @Override public SeekableByteChannel truncate(long size) { throw new UnsupportedOperationException(); }
+        private ByteArraySeekableByteChannel(byte[] data) {
+            this.data = data;
+        }
+
+        @Override
+        public int read(ByteBuffer dst) {
+            int remaining = dst.remaining();
+            int available = (int) (data.length - position);
+            if (available <= 0) return -1;
+            int toRead = Math.min(remaining, available);
+            dst.put(data, (int) position, toRead);
+            position += toRead;
+            return toRead;
+        }
+
+        @Override public long position() { return position; }
+        @Override public SeekableByteChannel position(long newPosition) { position = newPosition; return this; }
+        @Override public long size() { return data.length; }
+        @Override public boolean isOpen() { return isOpen; }
+        @Override public void close() { isOpen = false; }
+        @Override public int write(ByteBuffer src) { throw new UnsupportedOperationException(); }
+        @Override public SeekableByteChannel truncate(long size) { throw new UnsupportedOperationException(); }
+    }
 }

 // Read query results
 byte[] queryResult = namespaceClient.queryTable(query);

 try (BufferAllocator allocator = new RootAllocator();
-     ArrowFileReader reader = new ArrowFileReader(
-         new ByteArraySeekableByteChannel(queryResult), allocator)) {
+     ArrowFileReader reader = ArrowIpc.openFileReader(queryResult, allocator)) {

    for (int i = 0; i < reader.getRecordBlocks().size(); i++) {
        reader.loadRecordBatch(reader.getRecordBlocks().get(i));
@@ -83,6 +83,24 @@ Delete a branch.

 ***

+### diff()
+
+```ts
+diff(fromBranch): Promise<BranchDiff>
+```
+
+Compare a branch against main without modifying either branch.
+
+#### Parameters
+
+* **fromBranch**: `string`
+
+#### Returns
+
+`Promise`&lt;[`BranchDiff`](../interfaces/BranchDiff.md)&gt;
+
+***
+
 ### list()

 ```ts
@@ -94,3 +112,28 @@ List all branches, mapping name to branch metadata.
 #### Returns

 `Promise`&lt;`Record`&lt;`string`, [`BranchContents`](BranchContents.md)&gt;&gt;
+
+***
+
+### merge()
+
+```ts
+merge(fromBranch, dryRun): Promise<MergeBranchResult>
+```
+
+Merge a branch into main.
+
+Set `dryRun` to `true` to preview the merge. A rejected merge resolves
+with `status: "rejected"` instead of throwing.
+
+#### Parameters
+
+* **fromBranch**: `string`
+    Branch to merge from.
+
+* **dryRun**: `boolean` = `false`
+    When true, only preview the merge. Defaults to false.
+
+#### Returns
+
+`Promise`&lt;[`MergeBranchResult`](../interfaces/MergeBranchResult.md)&gt;
@@ -52,6 +52,11 @@
 - [AddDataOptions](interfaces/AddDataOptions.md)
 - [AddResult](interfaces/AddResult.md)
 - [AlterColumnsResult](interfaces/AlterColumnsResult.md)
+- [BranchColumnChange](interfaces/BranchColumnChange.md)
+- [BranchColumnSummary](interfaces/BranchColumnSummary.md)
+- [BranchDiff](interfaces/BranchDiff.md)
+- [BranchIndexSummary](interfaces/BranchIndexSummary.md)
+- [BranchRowCountSummary](interfaces/BranchRowCountSummary.md)
 - [ClientConfig](interfaces/ClientConfig.md)
 - [ColumnAlteration](interfaces/ColumnAlteration.md)
 - [ColumnOrdering](interfaces/ColumnOrdering.md)
@@ -86,6 +91,9 @@
 - [ListNamespacesOptions](interfaces/ListNamespacesOptions.md)
 - [ListNamespacesResponse](interfaces/ListNamespacesResponse.md)
 - [LsmWriteSpec](interfaces/LsmWriteSpec.md)
+- [MergeBlocker](interfaces/MergeBlocker.md)
+- [MergeBranchResult](interfaces/MergeBranchResult.md)
+- [MergePreview](interfaces/MergePreview.md)
 - [MergeResult](interfaces/MergeResult.md)
 - [NativeOAuthConfig](interfaces/NativeOAuthConfig.md)
 - [OAuthConfig](interfaces/OAuthConfig.md)
@@ -0,0 +1,33 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / BranchColumnChange
+
+# Interface: BranchColumnChange
+
+A column whose definition differs between main and the branch.
+
+## Properties
+
+### branch
+
+```ts
+branch: BranchColumnSummary;
+```
+
+***
+
+### main
+
+```ts
+main: BranchColumnSummary;
+```
+
+***
+
+### name
+
+```ts
+name: string;
+```
@@ -0,0 +1,33 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / BranchColumnSummary
+
+# Interface: BranchColumnSummary
+
+Summary of a column in a branch diff.
+
+## Properties
+
+### dataType
+
+```ts
+dataType: string;
+```
+
+***
+
+### name
+
+```ts
+name: string;
+```
+
+***
+
+### nullable
+
+```ts
+nullable: boolean;
+```
@@ -0,0 +1,129 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / BranchDiff
+
+# Interface: BranchDiff
+
+Read-only comparison of a branch against main.
+
+## Properties
+
+### addedColumns
+
+```ts
+addedColumns: BranchColumnSummary[];
+```
+
+***
+
+### addedIndexes
+
+```ts
+addedIndexes: BranchIndexSummary[];
+```
+
+***
+
+### baseMoved
+
+```ts
+baseMoved: boolean;
+```
+
+***
+
+### branchVersion
+
+```ts
+branchVersion: number;
+```
+
+***
+
+### changedColumns
+
+```ts
+changedColumns: BranchColumnChange[];
+```
+
+***
+
+### fromBranch
+
+```ts
+fromBranch: string;
+```
+
+***
+
+### mainVersion
+
+```ts
+mainVersion: number;
+```
+
+***
+
+### mergeBlockers
+
+```ts
+mergeBlockers: MergeBlocker[];
+```
+
+***
+
+### mergeable
+
+```ts
+mergeable: boolean;
+```
+
+***
+
+### parentVersion
+
+```ts
+parentVersion: number;
+```
+
+***
+
+### removedColumns
+
+```ts
+removedColumns: BranchColumnSummary[];
+```
+
+***
+
+### removedIndexes
+
+```ts
+removedIndexes: BranchIndexSummary[];
+```
+
+***
+
+### rowCountBranch
+
+```ts
+rowCountBranch: number;
+```
+
+***
+
+### rowCountMain
+
+```ts
+rowCountMain: number;
+```
+
+***
+
+### rowSummary
+
+```ts
+rowSummary: BranchRowCountSummary;
+```
@@ -0,0 +1,41 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / BranchIndexSummary
+
+# Interface: BranchIndexSummary
+
+Summary of an index in a branch diff.
+
+## Properties
+
+### columns
+
+```ts
+columns: string[];
+```
+
+***
+
+### indexName
+
+```ts
+indexName: string;
+```
+
+***
+
+### indexType?
+
+```ts
+optional indexType: string;
+```
+
+***
+
+### status
+
+```ts
+status: string;
+```
@@ -0,0 +1,57 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / BranchRowCountSummary
+
+# Interface: BranchRowCountSummary
+
+Row-level comparison between main and the branch.
+
+## Properties
+
+### deltaAvailable
+
+```ts
+deltaAvailable: boolean;
+```
+
+***
+
+### inputsChanged
+
+```ts
+inputsChanged: number;
+```
+
+***
+
+### newOnBase
+
+```ts
+newOnBase: number;
+```
+
+***
+
+### newOnBranch
+
+```ts
+newOnBranch: number;
+```
+
+***
+
+### staleRecompute
+
+```ts
+staleRecompute: number;
+```
+
+***
+
+### unchanged
+
+```ts
+unchanged: number;
+```
@@ -43,6 +43,19 @@ The following tokenizers are available:

 ***

+### blockSize?
+
+```ts
+optional blockSize: 128 | 256;
+```
+
+Number of documents per compressed posting block.
+
+The default is 128. Supported values are 128 and 256. A value of 256 uses
+the experimental FTS V3 format and may introduce breaking changes.
+
+***
+
 ### language?

 ```ts
@@ -0,0 +1,25 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / MergeBlocker
+
+# Interface: MergeBlocker
+
+A reason why a branch cannot currently be merged.
+
+## Properties
+
+### code
+
+```ts
+code: string;
+```
+
+***
+
+### message
+
+```ts
+message: string;
+```
@@ -0,0 +1,46 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / MergeBranchResult
+
+# Interface: MergeBranchResult
+
+Result of previewing or attempting a branch merge.
+
+## Properties
+
+### diff
+
+```ts
+diff: BranchDiff;
+```
+
+***
+
+### mainVersionAfter?
+
+```ts
+optional mainVersionAfter: number;
+```
+
+***
+
+### preview
+
+```ts
+preview: MergePreview;
+```
+
+***
+
+### status
+
+```ts
+status:
+  | "unknown"
+  | "rejected"
+  | "ready"
+  | "notImplemented"
+  | "merged";
+```
@@ -0,0 +1,17 @@
+[**@lancedb/lancedb**](../README.md) • **Docs**
+
+***
+
+[@lancedb/lancedb](../globals.md) / MergePreview
+
+# Interface: MergePreview
+
+Changes that would be, or were, promoted by a branch merge.
+
+## Properties
+
+### promotedColumns
+
+```ts
+promotedColumns: string[];
+```
@@ -28,7 +28,7 @@
    <properties>
        <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
        <arrow.version>15.0.0</arrow.version>
-        <lance-core.version>9.0.0-rc.1</lance-core.version>
+        <lance-core.version>10.0.0-beta.3</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>
@@ -52,6 +52,7 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
      Float64,
      Struct,
      List,
+      Map_,
      Int16,
      Int32,
      Int64,
@@ -69,6 +70,30 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
    type Schema = ApacheArrow["Schema"];
    type Table = ApacheArrow["Table"];

+    function expectValidMapField(
+      // biome-ignore lint/suspicious/noExplicitAny: Arrow Field types vary across supported versions
+      field: any,
+    ): void {
+      expect(DataType.isMap(field.type)).toBe(true);
+      expect(field.type.keysSorted).toBe(true);
+      expect(field.type.children).toHaveLength(1);
+
+      const entries = field.type.children[0];
+      expect(entries.name).toBe("entries");
+      expect(entries.nullable).toBe(false);
+      expect(DataType.isStruct(entries.type)).toBe(true);
+      expect(entries.type.children).toHaveLength(2);
+
+      const [key, value] = entries.type.children;
+      expect([key.name, value.name]).toEqual(["key", "value"]);
+      expect(key.nullable).toBe(false);
+      expect(DataType.isUtf8(key.type)).toBe(true);
+      expect(value.nullable).toBe(true);
+      expect(DataType.isInt(value.type)).toBe(true);
+      expect(value.type.bitWidth).toBe(32);
+      expect(value.type.isSigned).toBe(true);
+    }
+
    // Helper method to verify various ways to create a table
    async function checkTableCreation(
      tableCreationMethod: (
@@ -938,6 +963,34 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
          false,
        );
      });
+
+      it("will make an empty table with a Map field", async function () {
+        const schema = new Schema([
+          new Field(
+            "attributes",
+            new Map_(
+              new Field(
+                "entries",
+                new Struct([
+                  new Field("key", new Utf8(), false),
+                  new Field("value", new Int32(), true),
+                ]),
+                false,
+              ),
+              true,
+            ),
+          ),
+        ]);
+
+        const table = makeEmptyTable(schema);
+
+        expectValidMapField(table.schema.fields[0]);
+
+        const buffer = await fromTableToBuffer(table);
+        const roundTripped = tableFromIPC(buffer);
+
+        expectValidMapField(roundTripped.schema.fields[0]);
+      });
    });

    describe("when using two versions of arrow", function () {
@@ -15,6 +15,7 @@ import {
  OAuthHeaderProvider,
  StaticHeaderProvider,
 } from "../lancedb/header";
+import { Index } from "../lancedb/indices";

 // Test-only header providers
 class CustomProvider extends HeaderProvider {
@@ -225,6 +226,161 @@ describe("remote connection", () => {
    );
  });

+  it("sends the FTS posting block size to remote tables", async () => {
+    let createIndexBody: Record<string, unknown> | undefined;
+
+    await withMockDatabase(
+      (req, res) => {
+        const path = req.url ?? "";
+        if (path.endsWith("/describe/")) {
+          res.writeHead(200, { "Content-Type": "application/json" }).end(
+            JSON.stringify({
+              name: "t",
+              version: 1,
+              schema: {
+                fields: [
+                  { name: "text", type: { type: "string" }, nullable: false },
+                ],
+              },
+            }),
+          );
+          return;
+        }
+
+        if (path.endsWith("/create_index/")) {
+          let raw = "";
+          req.on("data", (chunk) => {
+            raw += chunk;
+          });
+          req.on("end", () => {
+            createIndexBody = JSON.parse(raw);
+            res.writeHead(200).end();
+          });
+          return;
+        }
+
+        res.writeHead(404).end();
+      },
+      async (db) => {
+        const table = await db.openTable("t");
+        await table.createIndex("text", {
+          config: Index.fts({ blockSize: 256 }),
+        });
+      },
+    );
+
+    expect(createIndexBody?.["column"]).toBe("text");
+    expect(createIndexBody?.["index_type"]).toBe("FTS");
+    expect(createIndexBody?.["block_size"]).toBe(256);
+  });
+
+  it("diffs and merges remote branches", async () => {
+    const sampleDiff = {
+      fromBranch: "exp",
+      parentVersion: 1,
+      mainVersion: 2,
+      branchVersion: 3,
+      baseMoved: false,
+      rowCountMain: 3,
+      rowCountBranch: 3,
+      rowSummary: {
+        unchanged: 3,
+        newOnBase: 0,
+        newOnBranch: 0,
+        staleRecompute: 0,
+        inputsChanged: 0,
+        deltaAvailable: false,
+      },
+      addedColumns: [{ name: "tag", dataType: "utf8", nullable: true }],
+      removedColumns: [],
+      changedColumns: [],
+      addedIndexes: [],
+      removedIndexes: [],
+      mergeable: true,
+      mergeBlockers: [],
+    };
+    const mergeBodies: Record<string, unknown>[] = [];
+
+    await withMockDatabase(
+      (req, res) => {
+        const path = req.url ?? "";
+        if (path.endsWith("/describe/")) {
+          res.writeHead(200, { "Content-Type": "application/json" }).end(
+            JSON.stringify({
+              name: "t",
+              version: 2,
+              schema: { fields: [] },
+            }),
+          );
+          return;
+        }
+
+        let raw = "";
+        req.on("data", (chunk) => {
+          raw += chunk;
+        });
+        req.on("end", () => {
+          const body = raw ? JSON.parse(raw) : {};
+          if (path.endsWith("/branches/diff/")) {
+            // biome-ignore lint/style/useNamingConvention: snake_case mandated by the server wire format
+            expect(body).toEqual({ from_branch: "exp" });
+            res
+              .writeHead(200, { "Content-Type": "application/json" })
+              .end(JSON.stringify(sampleDiff));
+            return;
+          }
+          if (path.endsWith("/branches/merge/")) {
+            mergeBodies.push(body);
+            const dryRun = body["dry_run"] === true;
+            const response = {
+              status: dryRun ? "ready" : "rejected",
+              diff: dryRun
+                ? sampleDiff
+                : {
+                    ...sampleDiff,
+                    mergeable: false,
+                    mergeBlockers: [
+                      { code: "baseMoved", message: "main has advanced" },
+                    ],
+                  },
+              preview: { promotedColumns: dryRun ? ["tag"] : [] },
+            };
+            res
+              .writeHead(dryRun ? 200 : 409, {
+                "Content-Type": "application/json",
+              })
+              .end(JSON.stringify(response));
+            return;
+          }
+          res.writeHead(404).end();
+        });
+      },
+      async (db) => {
+        const table = await db.openTable("t");
+        const branches = await table.branches();
+
+        await expect(branches.diff("exp")).resolves.toEqual(sampleDiff);
+
+        const rejected = await branches.merge("exp");
+        expect(rejected.status).toBe("rejected");
+        expect(rejected.diff.mergeBlockers).toEqual([
+          { code: "baseMoved", message: "main has advanced" },
+        ]);
+
+        const preview = await branches.merge("exp", true);
+        expect(preview.status).toBe("ready");
+        expect(preview.preview.promotedColumns).toEqual(["tag"]);
+      },
+    );
+
+    expect(mergeBodies).toEqual([
+      // biome-ignore lint/style/useNamingConvention: snake_case mandated by the server wire format
+      { from_branch: "exp", dry_run: false },
+      // biome-ignore lint/style/useNamingConvention: snake_case mandated by the server wire format
+      { from_branch: "exp", dry_run: true },
+    ]);
+  });
+
  describe("TlsConfig", () => {
    it("should create TlsConfig with all fields", () => {
      const tlsConfig: TlsConfig = {
@@ -2,7 +2,7 @@
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors

 import * as arrow from "../lancedb/arrow";
-import { sanitizeField, sanitizeType } from "../lancedb/sanitize";
+import { sanitizeField, sanitizeMap, sanitizeType } from "../lancedb/sanitize";

 describe("sanitize", function () {
  describe("sanitizeType function", function () {
@@ -181,4 +181,15 @@ describe("sanitize", function () {
      );
    });
  });
+
+  describe("sanitizeMap function", function () {
+    it.each([
+      ["no children", []],
+      ["two children", [{}, {}]],
+    ])("should reject a Map type with %s", function (_, children) {
+      expect(() => sanitizeMap({ children, keysSorted: false })).toThrow(
+        "Expected a Map type to have exactly one child",
+      );
+    });
+  });
 });
@@ -2527,6 +2527,35 @@ describe.each([arrow15, arrow16, arrow17, arrow18])(
      expect(results3.length).toBe(1);
    });

+    test("full text search with custom posting block size", async () => {
+      const db = await connect(tmpDir.name);
+      const data = [
+        { text: "hello world", vector: [0.1, 0.2, 0.3] },
+        { text: "goodbye world", vector: [0.4, 0.5, 0.6] },
+      ];
+      const table = await db.createTable("test", data);
+      await table.createIndex("text", {
+        config: Index.fts({ blockSize: 256 }),
+      });
+
+      const index = (await table.listIndices()).find(
+        (index) => index.indexType === "FTS",
+      );
+      expect(index?.indexVersion).toBe(3);
+      expect(
+        (index?.indexDetails as Record<string, unknown>)["block_size"],
+      ).toBe(256);
+
+      const results = await table.search("hello").toArray();
+      expect(results[0].text).toBe(data[0].text);
+    });
+
+    test("rejects invalid full text posting block size", () => {
+      expect(() => Index.fts({ blockSize: 129 as 128 | 256 })).toThrow(
+        "128 or 256",
+      );
+    });
+
    test("full text search without lowercase", async () => {
      const db = await connect(tmpDir.name);
      const data = [
@@ -124,6 +124,14 @@ export {
 export {
  Table,
  Branches,
+  BranchColumnSummary,
+  BranchColumnChange,
+  BranchIndexSummary,
+  BranchRowCountSummary,
+  MergeBlocker,
+  BranchDiff,
+  MergePreview,
+  MergeBranchResult,
  AddDataOptions,
  UpdateOptions,
  OptimizeOptions,
@@ -572,6 +572,14 @@ export interface FtsOptions {
   * whether to only index the prefix of the token for ngram tokenizer
   */
  prefixOnly?: boolean;
+
+  /**
+   * Number of documents per compressed posting block.
+   *
+   * The default is 128. Supported values are 128 and 256. A value of 256 uses
+   * the experimental FTS V3 format and may introduce breaking changes.
+   */
+  blockSize?: 128 | 256;
 }

 export class Index {
@@ -751,6 +759,7 @@ export class Index {
        options?.ngramMinLength,
        options?.ngramMaxLength,
        options?.prefixOnly,
+        options?.blockSize,
      ),
    );
  }
@@ -288,12 +288,11 @@ export function sanitizeMap(typeLike: object) {
  if (!("keysSorted" in typeLike) || typeof typeLike.keysSorted !== "boolean") {
    throw Error("Expected a Map type to have a `keysSorted` property");
  }
+  if (typeLike.children.length !== 1) {
+    throw Error("Expected a Map type to have exactly one child");
+  }

-  return new Map_(
-    // biome-ignore lint/suspicious/noExplicitAny: skip
-    typeLike.children.map((field) => sanitizeField(field)) as any,
-    typeLike.keysSorted,
-  );
+  return new Map_(sanitizeField(typeLike.children[0]), typeLike.keysSorted);
 }

 export function sanitizeDuration(typeLike: object) {
@@ -1329,6 +1329,76 @@ export interface FieldMetadataUpdate {
  replace?: boolean;
 }

+/** Summary of a column in a branch diff. */
+export interface BranchColumnSummary {
+  name: string;
+  dataType: string;
+  nullable: boolean;
+}
+
+/** A column whose definition differs between main and the branch. */
+export interface BranchColumnChange {
+  name: string;
+  main: BranchColumnSummary;
+  branch: BranchColumnSummary;
+}
+
+/** Summary of an index in a branch diff. */
+export interface BranchIndexSummary {
+  indexName: string;
+  columns: string[];
+  indexType?: string;
+  status: string;
+}
+
+/** Row-level comparison between main and the branch. */
+export interface BranchRowCountSummary {
+  unchanged: number;
+  newOnBase: number;
+  newOnBranch: number;
+  staleRecompute: number;
+  inputsChanged: number;
+  deltaAvailable: boolean;
+}
+
+/** A reason why a branch cannot currently be merged. */
+export interface MergeBlocker {
+  code: string;
+  message: string;
+}
+
+/** Read-only comparison of a branch against main. */
+export interface BranchDiff {
+  fromBranch: string;
+  parentVersion: number;
+  mainVersion: number;
+  branchVersion: number;
+  baseMoved: boolean;
+  rowCountMain: number;
+  rowCountBranch: number;
+  rowSummary: BranchRowCountSummary;
+  addedColumns: BranchColumnSummary[];
+  removedColumns: BranchColumnSummary[];
+  changedColumns: BranchColumnChange[];
+  addedIndexes: BranchIndexSummary[];
+  removedIndexes: BranchIndexSummary[];
+  mergeable: boolean;
+  mergeBlockers: MergeBlocker[];
+}
+
+/** Changes that would be, or were, promoted by a branch merge. */
+export interface MergePreview {
+  promotedColumns: string[];
+}
+
+/** Result of previewing or attempting a branch merge. */
+export interface MergeBranchResult {
+  status: "ready" | "rejected" | "notImplemented" | "merged" | "unknown";
+  diff: BranchDiff;
+  preview: MergePreview;
+  mainVersionAfter?: number;
+}
+
 /**
 * Branch manager for a {@link Table}.
 *
@@ -1381,4 +1451,28 @@ export class Branches {
  async delete(name: string): Promise<void> {
    return await this.#inner.delete(name);
  }
+
+  /** Compare a branch against main without modifying either branch. */
+  async diff(fromBranch: string): Promise<BranchDiff> {
+    return (await this.#inner.diff(fromBranch)) as unknown as BranchDiff;
+  }
+
+  /**
+   * Merge a branch into main.
+   *
+   * Set `dryRun` to `true` to preview the merge. A rejected merge resolves
+   * with `status: "rejected"` instead of throwing.
+   *
+   * @param fromBranch Branch to merge from.
+   * @param dryRun When true, only preview the merge. Defaults to false.
+   */
+  async merge(
+    fromBranch: string,
+    dryRun: boolean = false,
+  ): Promise<MergeBranchResult> {
+    return (await this.#inner.merge(
+      fromBranch,
+      dryRun,
+    )) as unknown as MergeBranchResult;
+  }
 }
@@ -226,7 +226,8 @@ impl Index {
        ngram_min_length: Option<u32>,
        ngram_max_length: Option<u32>,
        prefix_only: Option<bool>,
-    ) -> Self {
+        block_size: Option<u32>,
+    ) -> napi::Result<Self> {
        let mut opts = FtsIndexBuilder::default();
        if let Some(with_position) = with_position {
            opts = opts.with_position(with_position);
@@ -261,10 +262,15 @@ impl Index {
        if let Some(prefix_only) = prefix_only {
            opts = opts.ngram_prefix_only(prefix_only);
        }
-
-        Self {
-            inner: Mutex::new(Some(LanceDbIndex::FTS(opts))),
+        if let Some(block_size) = block_size {
+            opts = opts
+                .block_size(block_size as usize)
+                .map_err(|err| napi::Error::from_reason(err.to_string()))?;
        }
+
+        Ok(Self {
+            inner: Mutex::new(Some(LanceDbIndex::FTS(opts))),
+        })
    }

    #[napi(factory)]
@@ -232,6 +232,10 @@ impl From<ClientConfig> for lancedb::remote::ClientConfig {
            tls_config: config.tls_config.map(Into::into),
            header_provider: None, // the header provider is set separately later
            user_id: config.user_id,
+            // Resolved from LANCE_CLIENT_MAX_BYTES_PER_REQUEST or the default.
+            max_bytes_per_request: None,
+            // Resolved from LANCE_CLIENT_MAX_REQUEST_DURATION or the read timeout.
+            max_request_duration: None,
        }
    }
 }
@@ -1355,4 +1355,28 @@ impl Branches {
    pub async fn delete(&self, name: String) -> napi::Result<()> {
        self.inner.delete_branch(&name).await.default_error()
    }
+
+    #[napi(ts_return_type = "Promise<Record<string, unknown>>")]
+    pub async fn diff(&self, from_branch: String) -> napi::Result<serde_json::Value> {
+        let diff = self.inner.diff_branch(&from_branch).await.default_error()?;
+        serde_json::to_value(diff).map_err(|err| {
+            napi::Error::from_reason(format!("failed to serialize branch diff: {err}"))
+        })
+    }
+
+    #[napi(ts_return_type = "Promise<Record<string, unknown>>")]
+    pub async fn merge(
+        &self,
+        from_branch: String,
+        dry_run: Option<bool>,
+    ) -> napi::Result<serde_json::Value> {
+        let result = self
+            .inner
+            .merge_branch(&from_branch, dry_run.unwrap_or(false))
+            .await
+            .default_error()?;
+        serde_json::to_value(result).map_err(|err| {
+            napi::Error::from_reason(format!("failed to serialize branch merge result: {err}"))
+        })
+    }
 }
@@ -0,0 +1,21 @@
+{
+  "name": "lancedb",
+  "description": "Write, review, debug, and document LanceDB pipelines in Python and TypeScript that work across local LanceDB OSS tables and remote LanceDB Enterprise/Cloud tables, with idiomatic query/search patterns and performance defaults for ingestion, indexing, filtering, and diagnostics.",
+  "version": "0.1.0",
+  "author": {
+    "name": "LanceDB"
+  },
+  "homepage": "https://www.lancedb.com",
+  "keywords": [
+    "lancedb",
+    "vector-search",
+    "full-text-search",
+    "hybrid-search",
+    "python",
+    "typescript",
+    "pipelines",
+    "ingestion",
+    "indexing",
+    "performance"
+  ]
+}
@@ -0,0 +1,33 @@
+{
+  "name": "lancedb",
+  "version": "0.1.0",
+  "description": "Codex plugin for building LanceDB pipelines in Python and TypeScript.",
+  "author": {
+    "name": "LanceDB"
+  },
+  "keywords": [
+    "lancedb",
+    "vector-search",
+    "full-text-search",
+    "hybrid-search",
+    "python",
+    "typescript",
+    "pipelines"
+  ],
+  "skills": "./skills/",
+  "interface": {
+    "displayName": "LanceDB",
+    "shortDescription": "Build LanceDB pipelines in Python and TypeScript.",
+    "longDescription": "Write, review, debug, and document LanceDB pipelines in Python and TypeScript that work across local LanceDB OSS tables and remote LanceDB Enterprise/Cloud tables, with idiomatic query/search patterns and performance defaults for ingestion, indexing, filtering, and diagnostics.",
+    "developerName": "LanceDB",
+    "websiteURL": "https://www.lancedb.com",
+    "category": "Developer Tools",
+    "capabilities": [
+      "Developer Tools"
+    ],
+    "defaultPrompt": "Create a LanceDB table, embed sample text, and run a vector search.",
+    "composerIcon": "./assets/logo.png",
+    "logo": "./assets/logo.png",
+    "logoDark": "./assets/logo-dark.png"
+  }
+}
@@ -1,6 +1,6 @@
 ---
 name: lancedb
-description: Use when writing, reviewing, debugging, or documenting LanceDB pipelines in Python or TypeScript, especially code that should work across local LanceDB OSS tables and remote LanceDB Enterprise/Cloud tables. Helps avoid non-portable full-table materialization, choose idiomatic query/search patterns, and apply LanceDB performance defaults for ingestion, indexing, filtering, and diagnostics.
+description: Use when writing, reviewing, debugging, or documenting LanceDB pipelines in Python or TypeScript, especially code that should work across local LanceDB OSS tables and remote LanceDB Enterprise/Cloud tables. Helps avoid non-portable full-table materialization, choose idiomatic query/search patterns, apply LanceDB performance defaults for ingestion, indexing, filtering, and diagnostics, and resolve connections to the remote server for Enterprise-only operations such as jobs.
 ---

 # Building LanceDB Pipelines
@@ -19,7 +19,7 @@ Do NOT assume local-only table helpers exist on remote tables. If the user asks
 ## Workflow

 1. Identify the SDK: Python, TypeScript, or both.
-2. Identify the table mode: local/embedded OSS, remote Enterprise/Cloud, or portable across both. If the user says "LanceDB Enterprise", choose the remote table path.
+2. Identify the table mode: local/embedded OSS, remote Enterprise/Cloud, or portable across both. If the user says "LanceDB Enterprise", choose the remote table path. If the task involves jobs in any way (listing, inspecting, creating, or canceling jobs), it is always the remote path and requires a remote server connection — see "Connecting to the LanceDB remote server" below before doing anything else.
 3. Read the matching language branch before writing or changing code:
   - Python patterns: `references/python/patterns.md`
   - Python API quick reference: `references/python/api_reference.md`
@@ -29,7 +29,9 @@ Do NOT assume local-only table helpers exist on remote tables. If the user asks
   - TypeScript performance guidance: `references/typescript/performance.md`
   - Column metadata authoring (both SDKs): `references/column_metadata.md`
   - Branch operations (both SDKs): `references/branch_ops.md`
-4. Start with `patterns.md` for the selected SDK. Read `api_reference.md` when choosing method names or return collectors. Read `performance.md` when the task involves ingestion, indexing, filtering, query tuning, diagnostics, or large datasets. Read `column_metadata.md` when the task is documenting, tagging, classifying, or grouping table columns (field descriptions, `lancedb:tag:*` tags, logical column families). Read `branch_ops.md` when the task involves branch lifecycle (list/create/delete), writing to a non-main branch, or verifying a change stayed off main.
+   - Remote server connection resolution (jobs, raw REST): `references/remote_connect.md`
+   - Job operations REST API (list/describe/cancel/query_events): `references/remote_jobs.md`
+4. Start with `patterns.md` for the selected SDK. Read `api_reference.md` when choosing method names or return collectors. Read `performance.md` when the task involves ingestion, indexing, filtering, query tuning, diagnostics, or large datasets. Read `column_metadata.md` when the task is documenting, tagging, classifying, or grouping table columns (field descriptions, `lancedb:tag:*` tags, logical column families). Read `branch_ops.md` when the task involves branch lifecycle (list/create/delete), writing to a non-main branch, or verifying a change stayed off main. Read `remote_connect.md` when the task involves jobs or direct REST access to an Enterprise deployment, and `remote_jobs.md` for the job REST methods themselves (list, describe, cancel, query_events).
 5. For Python schemas, favor Pydantic models and validate records before writing. Use PyArrow schemas when Arrow-native, streaming, or highly dynamic data makes them materially better suited.
 6. Prefer `search()` or `query()` builders with explicit `select()` and `limit()` for reads.
 7. Avoid table-level full materialization in remote or portable code. This is the main local-vs-remote read pitfall.
@@ -70,6 +72,10 @@ Rules for portable Enterprise ingestion:

 This is Enterprise/Cloud-specific. Local/OSS tables have no separate data plane, so `mode="overwrite"` and immediate same-name reuse are fine there.

+## Connecting to the LanceDB remote server
+
+LanceDB Enterprise/Cloud deployments are served by a server implementing the lance-namespace OpenAPI spec (<https://github.com/lance-format/lance-namespace/blob/main/docs/src/spec.yaml>). Every remote (`db://...`) connection talks to such a server, and some operations exist only there. In particular, **all operations around jobs (listing, inspecting, creating, or canceling jobs) run server-side** — there is no local/OSS equivalent. Before any job work, or any direct REST call to an Enterprise deployment, read `references/remote_connect.md` to resolve the base URL, credentials, and database header and to validate the connection. Then use the four job REST methods documented in `references/remote_jobs.md` (list, describe, cancel, query_events).
+
 ## Script

 Run the scanner when reviewing or modifying an existing codebase:
@@ -0,0 +1,6 @@
+interface:
+  display_name: "LanceDB"
+  short_description: "Build LanceDB pipelines in Python and TypeScript"
+  default_prompt: "Use $lancedb to create a table, embed sample text, and run a vector search."
+  icon_small: "./assets/icon.png"
+  icon_large: "./assets/icon.png"
@@ -2,7 +2,7 @@

 Manage branches on a LanceDB table: list what exists, create new ones, delete stale ones, and direct read/write operations at a specific branch without touching main. Use for branch lifecycle tasks, experimental/isolated table versions, targeting an operation at a non-main branch, or confirming a mutation did not affect main.

-Works on local/OSS and remote Enterprise/Cloud tables.
+Works on local/OSS and remote Enterprise/Cloud tables, except merging a branch into main, which is Enterprise-only.

 ## The branch model (important)

@@ -101,6 +101,69 @@ assert b"lancedb:description" not in (table.schema.field("category").metadata or

 Two handles on the same branch see each other's writes (e.g. `table.branches.create("exp")` and `db.open_table(name, branch="exp")`); main stays isolated.

+## Merging a branch into main (Enterprise only)
+
+Merge is available through the SDKs (`table.branches.merge(...)`) on **Enterprise tables only** — it is not supported on Cloud or local/OSS tables, which raise `NotSupported`.
+
+`merge` takes the branch to merge **from** and a `dry_run` flag. Both the SDK method and the underlying REST endpoint **actually merge by default** (`dry_run=False`); pass `dry_run=True` to only preview. A rejected merge is **not an exception** — it returns a result with `status="rejected"` rather than raising, so inspect the return value. Use `branches.diff(from_branch)` to inspect a branch's pending diff without attempting a merge.
+
+```python
+exp = "experiment-reindex"
+
+# preview only — returns status="ready" if it would merge cleanly
+preview = table.branches.merge(exp, dry_run=True)
+
+# actually merge (default)
+result = table.branches.merge(exp)
+if result["status"] == "merged":
+    print("landed at", result["mainVersionAfter"])
+elif result["status"] == "rejected":
+    print(result["diff"]["mergeBlockers"])   # why it was refused
+
+# inspect a branch's pending diff without merging
+diff = table.branches.diff(exp)
+```
+
+Async: `await table.branches.merge(exp)`, `await table.branches.diff(exp)`.
+
+```typescript
+const branches = await table.branches();
+const exp = "experiment-reindex";
+
+// preview only (second arg is dryRun)
+const preview = await branches.merge(exp, true);
+
+// actually merge (default)
+const result = await branches.merge(exp);
+if (result.status === "merged") {
+  console.log("landed at", result.mainVersionAfter);
+} else if (result.status === "rejected") {
+  console.log(result.diff.mergeBlockers);
+}
+
+const diff = await branches.diff(exp);
+```
+
+The result is the wire JSON, containing `status` (`ready` on a passing dry run, `merged` on success, `rejected` when refused — also `notImplemented`/`unknown`), the branch `diff` (including `mergeBlockers` explaining any rejection), a `preview` of the columns that would be promoted, and — after a real merge — `mainVersionAfter`.
+
+### Merge preconditions
+
+Merge only **promotes newly added columns** onto main; it does not replay arbitrary commits. Practically, a branch is mergeable only if it has **exactly one commit since it was created, and that commit added a column**. The merge is rejected (`status: "rejected"`, with `mergeBlockers` set) if:
+
+- the branch was forked from another branch rather than directly from main
+- main has advanced since the branch was forked
+- the branch's rows changed since the fork (row counts must match main exactly)
+- the branch removed columns or changed a column's type/nullability
+- the branch added no columns (index-only changes are not merged)
+
+### Adding a column in a single commit
+
+Because the branch must contain just one column-adding commit, add the column with its values in one operation rather than add-then-backfill:
+
+1. **SQL transformation** — `add_columns` with a SQL expression computed from existing columns, so the column lands populated in one commit.
+2. **Precompute the values** — compute the column's values externally, then add the fully-populated column in a single operation (e.g. via `merge_insert`/`add_columns` with the data ready).
+3. **Lance-format-level data evolution (pylance)** — use Lance's data evolution with backfill, documented at <https://lance.org/guide/data_evolution/#with-data-backfill>.
+
 ## Quick reference

 | Goal | Python | TypeScript |
@@ -113,5 +176,7 @@ Two handles on the same branch see each other's writes (e.g. `table.branches.cre
 | Delete branch | `table.branches.delete(name)` | `await branches.delete(name)` |
 | Which branch is this handle on? | `table.current_branch()` (`None` = main) | `table.currentBranch()` (`null` = main) |
 | Target main | use the original (non-branch) handle | use the original (non-branch) handle |
+| Merge branch into main (Enterprise only) | `table.branches.merge(from_branch, dry_run=False)` | `await branches.merge(fromBranch, dryRun)` |
+| Preview a branch's pending diff (Enterprise only) | `table.branches.diff(from_branch)` | `await branches.diff(fromBranch)` |

 Branch names must be non-empty; empty names raise a validation error.
@@ -4,12 +4,19 @@ Quick method reference for Python LanceDB code. Cross-check source for non-trivi

 ## Connect

+If you're connecting to a remote database, use this:
 ```python
 import lancedb

-db = lancedb.connect("./camelot-db")         # local/OSS
-db = lancedb.connect("db://my-db", api_key=api_key, region=region)  # remote
+db = lancedb.connect("db://my-db", api_key=api_key, host_override=host_override) # remote
 ```
+(values may be found in LANCEDB_API_KEY and LANCEDB_HOST_OVERRIDE, either in env vars or a .env file)
+
+If you're connecting to a local table using OSS LanceDB, use this:
+```python
+db = lancedb.connect("./camelot-db") # local/OSS
+```
+If you're not sure which, or if you can't find the api_key or host_override params, ask the user.

 **Place the local database directory next to the script/entrypoint that opens it** (i.e. resolve the path relative to the script, `Path(__file__).parent / "camelot-db"`), not buried under a shared `data/` folder. The Lance dataset is the database, not a data file — keeping it beside its code makes ownership obvious and paths stable regardless of the working directory the script is launched from.

@@ -0,0 +1,45 @@
+# Connecting to a LanceDB remote server
+
+LanceDB Enterprise/Cloud deployments are served by a server implementing the
+lance-namespace OpenAPI spec
+(<https://github.com/lance-format/lance-namespace/blob/main/docs/src/spec.yaml>).
+Every remote (`db://...`) connection talks to such a server, and some operations
+exist only there. In particular, all operations around jobs (listing, inspecting,
+creating, or canceling jobs) run server-side — there is no local/OSS equivalent, so
+resolve a server connection before attempting any job work. The job REST methods
+themselves are documented in `references/remote_jobs.md`.
+
+Every request needs two things:
+
+1. **Base URL** — the server endpoint
+2. **Credentials** — an API key (`x-api-key` header over REST), and usually a database name (`x-lancedb-database` header)
+
+## Resolution steps
+
+1. If the user already gave a URL and API key (or said which environment they're working against), use that.
+2. Otherwise, look for credentials already available in the environment:
+   - Env vars like `LANCEDB_URI` / `LANCEDB_HOST` / `LANCEDB_API_KEY`
+   - A server endpoint already running or port-forwarded locally (the REST default port is 2333, i.e. `http://localhost:2333`)
+3. If you didn't find both pieces, ask the user directly: **"What's your LanceDB endpoint's URL, and what's your API key?"** Also ask which database to use if it isn't obvious. Don't guess or probe further — the user knows their deployment.
+
+## Validating the connection
+
+Make a cheap authenticated request and check the status before starting real work:
+
+```bash
+curl -s -w "\n%{http_code}" "{base_url}/v1/table/?limit=1" \
+  -H "x-api-key: <key>" \
+  -H "x-lancedb-database: <database>"
+```
+
+- `200` — connection, key, and database header all good
+- `401` — API key missing or wrong
+- `400` mentioning a database header — this deployment expects `x-lancedb-database`
+
+## Non-REST equivalents
+
+The same credentials work through the SDKs and CLI:
+
+- Python SDK: `lancedb.connect("db://<database>", api_key="<key>", host_override="<base_url>")`
+- TypeScript SDK: `await lancedb.connect("db://<database>", { apiKey: "<key>", hostOverride: "<base_url>" })`
+- `lancedb` CLI: a `[profiles.<name>]` entry in `~/.lancedb/config.toml` with `http_server_url`, `api_key`, `database`
@@ -0,0 +1,151 @@
+# Job operations over the LanceDB remote server REST API
+
+Jobs are server-side background operations on LanceDB Enterprise/Cloud — index builds,
+column backfills, materialized view refreshes, and similar async work. Endpoints that
+trigger async work (e.g. the column backfill or materialized view refresh endpoints)
+return a `job_id`; these four methods are how you track and manage those jobs.
+
+Resolve the connection first — see `references/remote_connect.md`. All four methods
+are **POST** requests under `{base_url}/v1/jobs/` with JSON bodies, and take the usual
+`x-api-key` / `x-lancedb-database` headers. If every job call returns `501`, job APIs
+are disabled on that deployment (the server has no job registry configured) — report
+that rather than retrying.
+
+## 1. List jobs — `POST /v1/jobs/list`
+
+The body is optional; an empty body lists everything. All fields are filters:
+
+```json
+{
+  "limit": 100,
+  "table_name": "my_table",
+  "job_type": "...",
+  "job_subtype": "...",
+  "state": "...",
+  "page_token": "..."
+}
+```
+
+```bash
+curl -s -X POST "{base_url}/v1/jobs/list" \
+  -H "x-api-key: <key>" -H "x-lancedb-database: <database>" \
+  -H "content-type: application/json" \
+  -d '{"table_name": "my_table"}'
+```
+
+Response:
+
+```json
+{
+  "jobs": [
+    {
+      "job_id": "...",
+      "table": "my_table",
+      "job_type": "...",
+      "job_subtype": "...",
+      "state": "done",
+      "created_at_millis": 1720000000000
+    }
+  ],
+  "page_token": "..."
+}
+```
+
+A `page_token` in the response means there are more results — pass it back in the next
+request to continue. Note list rows use a lowercase `state` string, while describe uses
+an uppercase `job_state`.
+
+## 2. Describe a job — `POST /v1/jobs/describe`
+
+Body: `{"job_id": "<id>"}`. Returns full detail for one job:
+
+```json
+{
+  "job_id": "...",
+  "job_type": "...",
+  "job_subtype": "...",
+  "job_state": "IN_PROGRESS",
+  "creation_ms": 1720000000000,
+  "spec": {},
+  "status": {}
+}
+```
+
+`job_state` is one of `IN_PROGRESS`, `CANCELLED`, `FAILED`, `DONE`. `spec` and `status`
+are job-type-specific JSON objects (the job's input specification and its current
+progress/status). Returns `404` for an unknown job id.
+
+## 3. Cancel a job — `POST /v1/jobs/cancel`
+
+Body: `{"job_id": "<id>"}`; response echoes `{"job_id": "<id>"}`. Cancellation is a
+service-level operation requiring the same administrative authorization as the
+`/admin` routes — a database-scoped API key that can list and describe jobs may still
+get a permission error here. Other errors: `404` unknown job, `409` state conflict
+(e.g. already in a terminal state), `429` too much write contention (safe to retry).
+
+## 4. Query job event history — `POST /v1/jobs/query_events`
+
+Returns the event history (state transitions, progress updates) for one or more jobs.
+Body: `{"job_id": "<id>"}` for one job, or `{"job_ids": ["<id>", ...]}` for a batch.
+Optional fields: `limit` (max event rows), `limit_per_job` (per job in a batch query),
+and `filter` — a SQL-like expression over the columns `state`, `updated_by`,
+`owner_component`, and `claim_entity`. (`full_text_search` is reserved and currently
+rejected as not implemented.)
+
+The response is **not JSON** — it is an Arrow IPC stream
+(`content-type: application/vnd.apache.arrow.stream`). Decode it, e.g. in Python:
+
+```python
+import pyarrow.ipc
+import requests
+
+resp = requests.post(
+    f"{base_url}/v1/jobs/query_events",
+    headers={"x-api-key": key, "x-lancedb-database": database},
+    json={"job_id": job_id},
+)
+resp.raise_for_status()
+events = pyarrow.ipc.open_stream(resp.content).read_all()
+```
+
+## Feature engineering (Geneva) jobs
+
+Feature engineering jobs — UDF column backfills and materialized view refreshes run
+through Geneva — are tracked **separately** from the `/v1/jobs` registry above. Their
+records live in a `geneva_jobs` table inside the database itself (in the `__system`
+namespace), and you access them through a Python `geneva` connection rather than the
+REST endpoints above:
+
+```python
+import geneva
+from geneva.jobs import JobStateManager
+
+# Same credentials as lancedb.connect / the REST API
+conn = geneva.connect("db://<database>", api_key="<key>", host_override="<base_url>")
+jsm = JobStateManager(conn)
+
+# List jobs. NOTE: status defaults to "RUNNING"; pass status=None for all jobs.
+# Statuses: PENDING | RUNNING | DONE | FAILED | CANCELLED
+jobs = jsm.list_jobs(table_name="my_table", status=None)
+
+# Fetch one job by id (returns a list of JobRecord)
+records = jsm.get("<job_id>")
+```
+
+Each `JobRecord` has `table_name`, `column_name`, `job_id`, `job_type`, `status`,
+`launched_at`, `completed_at`, `config`, `launched_by`, `manifest_id`, `cluster_name`,
+`metrics` (progress counters), `events` (human-readable history), and `updated_at`.
+For filters `list_jobs` doesn't support (e.g. time ranges), query the underlying table
+directly: `jsm.get_table(True).search().where("launched_at >= TIMESTAMP '...'")` —
+pass `True` to check out the latest version, since other processes update job state.
+
+Stale-status caveat: nothing reaps dead Geneva jobs, so a job can sit in
+`RUNNING`/`PENDING` forever if its worker died. Treat a job as effectively `FAILED`
+when it has been running longer than ~36 hours, or its `updated_at` is more than ~2
+hours old (this matches the heuristic the Geneva console UI applies on read).
+
+## Workflow tips
+
+- To wait for async work (a backfill, an index build), poll `describe` until
+  `job_state` leaves `IN_PROGRESS`; on `FAILED`, pull `status` and `query_events` for
+  the failure detail.
@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.35.0-beta.2"
+current_version = "0.35.0-beta.3"
 parse = """(?x)
    (?P<major>0|[1-9]\\d*)\\.
    (?P<minor>0|[1-9]\\d*)\\.
@@ -1,6 +1,6 @@
 [package]
 name = "lancedb-python"
-version = "0.35.0-beta.2"
+version = "0.35.0-beta.3"
 publish = false
 edition.workspace = true
 description = "Python bindings for LanceDB"
@@ -8,6 +8,27 @@ A Python library for [LanceDB](https://github.com/lancedb/lancedb).
 pip install lancedb
 ```

+### Pre-Haswell x86_64 hosts: `lancedb-compat`
+
+The default `lancedb` wheel targets `x86-64-haswell` (AVX2 + FMA + F16C) for full performance on modern hardware. Pre-Haswell hosts — Intel Sandy Bridge / Ivy Bridge / Westmere; AMD Bulldozer / Piledriver / Steamroller — don't have AVX2 and crash with `Illegal instruction` at `import lancedb`.
+
+For those hosts, install the `lancedb-compat` package instead:
+
+```bash
+pip install lancedb-compat
+```
+
+Same Python API (`import lancedb` works as usual). The compat wheel is compiled at the `x86-64-v2` baseline (Nehalem-class) and uses runtime SIMD dispatch in the embedded lance crate to pick the right kernel tier (scalar / AVX / AVX+FMA / AVX2+FMA / AVX-512) at load time, so it still goes fast on modern hardware while running cleanly on the pre-Haswell silicon. Use `lance.simd_info()` from Python to verify which tier was selected.
+
+`lancedb` and `lancedb-compat` install to the same `lancedb/` namespace and conflict at install time. Pick one. To switch, `pip uninstall lancedb` first, then `pip install lancedb-compat` (or vice-versa).
+
+If you need a custom baseline (or `lancedb-compat` isn't yet published for your platform), build from source with the override:
+
+```bash
+RUSTFLAGS="-C target-cpu=x86-64-v2" maturin build --release
+pip install ./target/wheels/lancedb-*.whl
+```
+
 ### Preview Releases

 Stable releases are created about every 2 weeks. For the latest features and bug fixes, you can install the preview release. These releases receive the same level of testing as stable releases, but are not guaranteed to be available for more than 6 months after they are released. Once your application is stable, we recommend switching to stable releases.
@@ -219,6 +219,7 @@ class Table:
        data: pa.RecordBatchReader,
        mode: Literal["append", "overwrite"],
        progress: Optional[Any] = None,
+        write_parallelism: Optional[int] = None,
    ) -> AddResult: ...
    async def update(
        self, updates: Dict[str, str], where: Optional[str]
@@ -318,6 +319,10 @@ class Branches:
    ) -> Table: ...
    async def checkout(self, name: str, version: Optional[int] = None) -> Table: ...
    async def delete(self, name: str) -> None: ...
+    async def diff(self, from_branch: str) -> Dict[str, Any]: ...
+    async def merge(
+        self, from_branch: str, dry_run: bool = False
+    ) -> Dict[str, Any]: ...

 class IndexConfig:
    name: str
@@ -41,6 +41,7 @@ from lance_namespace import (
    ListTablesResponse,
    connect as namespace_connect,
 )
+from lance_namespace.errors import NamespaceNotEmptyError, TableNotFoundError

 from . import __version__
 from ._lancedb import connect as lancedb_connect  # type: ignore
@@ -746,10 +747,12 @@ class LanceDBConnection(DBConnection):
        """
        if namespace_path is None:
            namespace_path = []
-        return self._namespace_conn().list_namespaces(
-            namespace_path=namespace_path,
-            page_token=page_token,
-            limit=limit,
+        return LOOP.run(
+            self._conn.list_namespaces(
+                namespace_path=namespace_path,
+                page_token=page_token,
+                limit=limit,
+            )
        )

    @override
@@ -759,10 +762,12 @@ class LanceDBConnection(DBConnection):
        mode: Optional[str] = None,
        properties: Optional[Dict[str, str]] = None,
    ) -> CreateNamespaceResponse:
-        return self._namespace_conn().create_namespace(
-            namespace_path=namespace_path,
-            mode=mode,
-            properties=properties,
+        return LOOP.run(
+            self._conn.create_namespace(
+                namespace_path=namespace_path,
+                mode=mode,
+                properties=properties,
+            )
        )

    @override
@@ -772,19 +777,24 @@ class LanceDBConnection(DBConnection):
        mode: Optional[str] = None,
        behavior: Optional[str] = None,
    ) -> DropNamespaceResponse:
-        return self._namespace_conn().drop_namespace(
-            namespace_path=namespace_path,
-            mode=mode,
-            behavior=behavior,
-        )
+        try:
+            return LOOP.run(
+                self._conn.drop_namespace(
+                    namespace_path=namespace_path,
+                    mode=mode,
+                    behavior=behavior,
+                )
+            )
+        except RuntimeError as e:
+            if "Namespace not empty" in str(e):
+                raise NamespaceNotEmptyError(str(e)) from e
+            raise

    @override
    def describe_namespace(
        self, namespace_path: List[str]
    ) -> DescribeNamespaceResponse:
-        return self._namespace_conn().describe_namespace(
-            namespace_path=namespace_path,
-        )
+        return LOOP.run(self._conn.describe_namespace(namespace_path=namespace_path))

    @override
    def list_tables(
@@ -813,12 +823,6 @@ class LanceDBConnection(DBConnection):
        """
        if namespace_path is None:
            namespace_path = []
-        if namespace_path:
-            return self._namespace_conn().list_tables(
-                namespace_path=namespace_path,
-                page_token=page_token,
-                limit=limit,
-            )
        return LOOP.run(
            self._conn.list_tables(
                namespace_path=namespace_path, page_token=page_token, limit=limit
@@ -916,22 +920,6 @@ class LanceDBConnection(DBConnection):
            raise ValueError("mode must be either 'create' or 'overwrite'")
        validate_table_name(name)

-        if namespace_path:
-            return self._namespace_conn().create_table(
-                name,
-                data=data,
-                schema=schema,
-                mode=mode,
-                exist_ok=exist_ok,
-                on_bad_vectors=on_bad_vectors,
-                fill_value=fill_value,
-                embedding_functions=embedding_functions,
-                namespace_path=namespace_path,
-                storage_options=storage_options,
-                data_storage_version=data_storage_version,
-                enable_v2_manifest_paths=enable_v2_manifest_paths,
-            )
-
        tbl = LanceTable.create(
            self,
            name,
@@ -944,22 +932,11 @@ class LanceDBConnection(DBConnection):
            embedding_functions=embedding_functions,
            namespace_path=namespace_path,
            storage_options=storage_options,
+            data_storage_version=data_storage_version,
+            enable_v2_manifest_paths=enable_v2_manifest_paths,
        )
        return tbl

-    def _namespace_conn(self) -> DBConnection:
-        """Return a LanceNamespaceDBConnection backed by this connection's
-        directory namespace.  Used to delegate child-namespace operations."""
-        from lancedb.namespace import LanceNamespaceDBConnection
-
-        return LanceNamespaceDBConnection(
-            self.namespace_client(),
-            read_consistency_interval=self.read_consistency_interval,
-            storage_options=self.storage_options,
-            namespace_client_impl=None,
-            namespace_client_properties=None,
-        )
-
    @override
    def open_table(
        self,
@@ -1006,14 +983,7 @@ class LanceDBConnection(DBConnection):
                stacklevel=2,
            )

-        if namespace_path:
-            tbl = self._namespace_conn().open_table(
-                name,
-                namespace_path=namespace_path,
-                storage_options=storage_options,
-                index_cache_size=index_cache_size,
-            )
-        else:
+        try:
            tbl = LanceTable.open(
                self,
                name,
@@ -1021,6 +991,15 @@ class LanceDBConnection(DBConnection):
                storage_options=storage_options,
                index_cache_size=index_cache_size,
            )
+        except (RuntimeError, ValueError) as e:
+            if namespace_path and (
+                "Table not found" in str(e) or "was not found" in str(e)
+            ):
+                table_id = namespace_path + [name]
+                raise TableNotFoundError(
+                    f"Table not found: {'$'.join(table_id)}"
+                ) from e
+            raise

        if branch is not None:
            tbl = tbl.branches.checkout(branch, version)
@@ -1104,9 +1083,6 @@ class LanceDBConnection(DBConnection):
        """
        if namespace_path is None:
            namespace_path = []
-        if namespace_path:
-            self._namespace_conn().drop_table(name, namespace_path=namespace_path)
-            return
        LOOP.run(
            self._conn.drop_table(
                name, namespace_path=namespace_path, ignore_missing=ignore_missing
@@ -14,29 +14,76 @@ import numpy as np

 DEFAULT_WATSONX_URL = "https://us-south.ml.cloud.ibm.com"

-MODELS_DIMS = {
+# Models currently available on the watsonx.ai SaaS platform.
+# These are the IDs advertised to new users via model_names() and shown in
+# validation error messages.  Regional availability and withdrawal dates are
+# documented at:
+# https://www.ibm.com/docs/en/watsonx/saas?topic=models-supported-encoder
+CURRENT_MODELS: dict[str, int] = {
+    "ibm/granite-embedding-278m-multilingual": 768,
+    "ibm/slate-125m-english-rtrvr-v2": 768,
+    "ibm/slate-30m-english-rtrvr-v2": 384,
+    "intfloat/multilingual-e5-large": 1024,
+}
+
+# Full dimension map including legacy model IDs from earlier releases.
+# Kept so that existing tables whose stored metadata uses these names can still
+# resolve dimensions on load without raising an error.  These IDs are NOT
+# advertised to new users.
+MODELS_DIMS: dict[str, int] = {
+    **CURRENT_MODELS,
+    # Deprecated — withdrawal announced but still functional until the dates above.
+    "sentence-transformers/all-minilm-l6-v2": 384,
+    # Pre-v2 legacy names retained for metadata compatibility only.
    "ibm/slate-125m-english-rtrvr": 768,
    "ibm/slate-30m-english-rtrvr": 384,
    "sentence-transformers/all-minilm-l12-v2": 384,
-    "intfloat/multilingual-e5-large": 1024,
 }


@register("watsonx")
 class WatsonxEmbeddings(TextEmbeddingFunction):
    """
+    An embedding function that uses the IBM watsonx.ai Embeddings API.
+
    API Docs:
-    ---------
-    https://cloud.ibm.com/apidocs/watsonx-ai#text-embeddings
+        https://cloud.ibm.com/apidocs/watsonx-ai#text-embeddings

    Supported embedding models:
-    ---------------------------
-    https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models-embed.html?context=wx
+        https://dataplatform.cloud.ibm.com/docs/content/wsj/analyze-data/fm-models-embed.html?context=wx
+
+    Parameters
+    ----------
+    name : str, default "ibm/slate-125m-english-rtrvr"
+        The ID of the embedding model to use.  For new tables,
+        ``"ibm/granite-embedding-278m-multilingual"`` is recommended.
+    api_key : str, optional
+        IBM Cloud API key.  Falls back to the ``WATSONX_API_KEY`` environment
+        variable when not provided.
+    project_id : str, optional
+        watsonx.ai project ID.  Explicit value takes precedence over the
+        ``WATSONX_PROJECT_ID`` environment variable.  Mutually exclusive with
+        ``space_id`` — exactly one must be supplied.
+    space_id : str, optional
+        watsonx.ai deployment space ID.  Explicit value takes precedence over
+        the ``WATSONX_SPACE_ID`` environment variable.  Mutually exclusive with
+        ``project_id`` — exactly one must be supplied.
+    url : str, optional
+        watsonx.ai service URL.  Defaults to
+        ``"https://us-south.ml.cloud.ibm.com"``.
+    params : dict, optional
+        Extra parameters forwarded verbatim to ``Embeddings`` (e.g.
+        ``{"truncate_input_tokens": 512}``).
    """

+    # Intentionally kept at the original pre-PR default so that existing tables
+    # whose stored metadata contains model:{} reload with the same model they
+    # were created with.  New users should pass name= explicitly, e.g.
+    # name="ibm/granite-embedding-278m-multilingual".
    name: str = "ibm/slate-125m-english-rtrvr"
    api_key: Optional[str] = None
    project_id: Optional[str] = None
+    space_id: Optional[str] = None
    url: Optional[str] = None
    params: Optional[Dict] = None

@@ -46,12 +93,13 @@ class WatsonxEmbeddings(TextEmbeddingFunction):

    @staticmethod
    def model_names():
-        return [
-            "ibm/slate-125m-english-rtrvr",
-            "ibm/slate-30m-english-rtrvr",
-            "sentence-transformers/all-minilm-l12-v2",
-            "intfloat/multilingual-e5-large",
-        ]
+        """Return the IDs of models currently available for new tables.
+
+        Legacy / deprecated IDs are intentionally excluded.  They remain
+        resolvable for dimension lookups on existing tables via ``MODELS_DIMS``,
+        but should not be used when creating new tables.
+        """
+        return list(CURRENT_MODELS.keys())

    def ndims(self):
        return self._ndims
@@ -59,7 +107,10 @@ class WatsonxEmbeddings(TextEmbeddingFunction):
    @cached_property
    def _ndims(self):
        if self.name not in MODELS_DIMS:
-            raise ValueError(f"Unknown model name {self.name}")
+            raise ValueError(
+                f"Unknown model '{self.name}'. "
+                f"Available models: {list(CURRENT_MODELS.keys())}"
+            )
        return MODELS_DIMS[self.name]

    def generate_embeddings(
@@ -81,27 +132,45 @@ class WatsonxEmbeddings(TextEmbeddingFunction):
            "ibm_watsonx_ai.foundation_models"
        )

-        kwargs = {"model_id": self.name}
+        # --- credentials ---
+        # Explicit field takes priority; env var is the fallback.
+        api_key = self.api_key or os.environ.get("WATSONX_API_KEY")
+        if not api_key:
+            raise ValueError(
+                "WATSONX_API_KEY not set. Either set it in your environment or "
+                "pass it as `api_key` argument to WatsonxEmbeddings."
+            )
+        credentials = ibm_watsonx_ai.Credentials(
+            api_key=api_key,
+            url=self.url or DEFAULT_WATSONX_URL,
+        )
+
+        # --- project_id / space_id (exactly one required) ---
+        # Explicit field always wins; env var is consulted only when the
+        # corresponding field was not set, so passing project_id= never
+        # conflicts with a stray WATSONX_SPACE_ID env var and vice-versa.
+        space_id, project_id = self.space_id, self.project_id
+
+        if project_id is None and space_id is None:
+            # Neither was passed explicitly — fall back to env vars.
+            project_id = os.environ.get("WATSONX_PROJECT_ID")
+            space_id = os.environ.get("WATSONX_SPACE_ID")
+
+        if project_id and space_id:
+            raise ValueError("Provide either `project_id` or `space_id`, not both.")
+        if not project_id and not space_id:
+            raise ValueError(
+                "Either WATSONX_PROJECT_ID or WATSONX_SPACE_ID must be set. "
+                "Pass one as an argument to WatsonxEmbeddings or set the "
+                "corresponding environment variable."
+            )
+
+        client_kwargs: Dict = dict(model_id=self.name, credentials=credentials)
        if self.params:
-            kwargs["params"] = self.params
-        if self.project_id:
-            kwargs["project_id"] = self.project_id
-        elif "WATSONX_PROJECT_ID" in os.environ:
-            kwargs["project_id"] = os.environ["WATSONX_PROJECT_ID"]
+            client_kwargs["params"] = self.params
+        if project_id:
+            client_kwargs["project_id"] = project_id
        else:
-            raise ValueError("WATSONX_PROJECT_ID must be set or passed")
+            client_kwargs["space_id"] = space_id

-        creds_kwargs = {}
-        if self.api_key:
-            creds_kwargs["api_key"] = self.api_key
-        elif "WATSONX_API_KEY" in os.environ:
-            creds_kwargs["api_key"] = os.environ["WATSONX_API_KEY"]
-        else:
-            raise ValueError("WATSONX_API_KEY must be set or passed")
-        if self.url:
-            creds_kwargs["url"] = self.url
-        else:
-            creds_kwargs["url"] = DEFAULT_WATSONX_URL
-        kwargs["credentials"] = ibm_watsonx_ai.Credentials(**creds_kwargs)
-
-        return ibm_watsonx_ai_foundation_models.Embeddings(**kwargs)
+        return ibm_watsonx_ai_foundation_models.Embeddings(**client_kwargs)
@@ -115,6 +115,12 @@ class FTS:

    For example, it works with `title`, `description`, `content`, etc.

+    Examples
+    --------
+    Create an index configuration that uses 256-document posting blocks:
+
+    >>> config = FTS(block_size=256)
+
    Attributes
    ----------
    with_position : bool, default False
@@ -148,6 +154,10 @@ class FTS:
    ascii_folding : bool, default True
        Whether to fold ASCII characters. This converts accented characters to
        their ASCII equivalent. For example, "café" would be converted to "cafe".
+    block_size : int, default 128
+        The number of documents per compressed posting block. Supported values
+        are 128 and 256. A value of 256 uses the experimental FTS V3 format
+        and may introduce breaking changes.

    Notes
    -----
@@ -168,6 +178,7 @@ class FTS:
    ngram_min_length: int = 3
    ngram_max_length: int = 3
    prefix_only: bool = False
+    block_size: int = 128


@dataclass
@@ -885,7 +885,7 @@ class Permutation:
        This method refines the current selection, potentially removing columns.  It
        will not add back columns that were previously removed.

-        If any of the columns do not exist then an error will be raised
+        If any of the columns do not exist then an error will be raised.

        This does not introduce a post-processing step.  It simply reduces the amount
        of data we read.
@@ -898,9 +898,14 @@ class Permutation:
        for name in columns:
            value = self.selection.get(name, None)
            if value is None:
-                raise ValueError(
-                    f"Cannot select column `{name}` because it does not exist"
-                )
+                if name == "_rowid":
+                    # _rowid is a system column not in the default schema
+                    # but can be explicitly selected
+                    value = "_rowid"
+                else:
+                    raise ValueError(
+                        f"Cannot select column `{name}` because it does not exist"
+                    )
            new_selection[name] = value
        return self._with_selection(new_selection)

@@ -3875,18 +3875,16 @@ class AsyncHybridQuery(AsyncStandardQuery, AsyncVectorQueryBase):
        >>> asyncio.run(doctest_example()) # doctest: +ELLIPSIS, +NORMALIZE_WHITESPACE
        RRFReranker(K=60)
            ProjectionExec: expr=[vector@0 as vector, text@3 as text, _distance@2 as _distance]
-              Take: columns="vector, _rowid, _distance, (text)"
-                CoalesceBatchesExec: target_batch_size=1024
-                  GlobalLimitExec: skip=0, fetch=10
-                    FilterExec: _distance@2 IS NOT NULL
-                      SortExec: TopK(fetch=10), expr=[_distance@2 ASC NULLS LAST, _rowid@1 ASC NULLS LAST], preserve_partitioning=[false]
-                        KNNVectorDistance: metric=l2
-                          LanceRead: uri=..., projection=[vector], ...
+              LanceRead: uri=..., projection=[text], source=stream(_rowid)
+                GlobalLimitExec: skip=0, fetch=10
+                  FilterExec: _distance@2 IS NOT NULL
+                    SortExec: TopK(fetch=10), expr=[_distance@2 ASC NULLS LAST, _rowid@1 ASC NULLS LAST], preserve_partitioning=[false]
+                      KNNVectorDistance: metric=l2
+                        LanceRead: uri=..., projection=[vector], ...
            ProjectionExec: expr=[vector@2 as vector, text@3 as text, _score@1 as _score]
-              Take: columns="_rowid, _score, (vector), (text)"
-                CoalesceBatchesExec: target_batch_size=1024
-                  GlobalLimitExec: skip=0, fetch=10
-                    MatchQuery: column=text, query=hello
+              LanceRead: uri=..., projection=[vector, text], source=stream(_rowid)
+                GlobalLimitExec: skip=0, fetch=10
+                  MatchQuery: column=text, query=[hello]

        Parameters
        ----------
@@ -344,6 +344,7 @@ class RemoteTable(Table):
        ngram_min_length: int = 3,
        ngram_max_length: int = 3,
        prefix_only: bool = False,
+        block_size: int = 128,
        name: Optional[str] = None,
    ):
        """Create a full-text search index on a column.
@@ -364,6 +365,7 @@ class RemoteTable(Table):
            ngram_min_length=ngram_min_length,
            ngram_max_length=ngram_max_length,
            prefix_only=prefix_only,
+            block_size=block_size,
        )
        LOOP.run(
            self._table.create_index(
@@ -574,6 +576,7 @@ class RemoteTable(Table):
        on_bad_vectors: str = "error",
        fill_value: float = 0.0,
        progress: Optional[Union[bool, Callable, Any]] = None,
+        write_parallelism: Optional[int] = None,
    ) -> AddResult:
        """Add more data to the [Table](Table). It has the same API signature as
        the OSS version.
@@ -599,6 +602,12 @@ class RemoteTable(Table):
        progress: bool, callable, or tqdm-like, optional
            A callback or tqdm-compatible progress bar. See
            :meth:`Table.add` for details.
+        write_parallelism: int, optional
+            Number of partitions to write in parallel. Higher values increase
+            throughput but also peak memory use, since each partition buffers
+            data in flight. Defaults to an estimate based on the data size,
+            capped at the number of CPU cores. Lower this if bulk ingestion is
+            using too much memory.

        Returns
        -------
@@ -614,6 +623,7 @@ class RemoteTable(Table):
                    on_bad_vectors=on_bad_vectors,
                    fill_value=fill_value,
                    progress=progress,
+                    write_parallelism=write_parallelism,
                )
            )
        finally:
@@ -23,7 +23,7 @@ class AnswerdotaiRerankers(Reranker):
    column : str, default "text"
        The name of the column to use as input to the cross encoder model.
    return_score : str, default "relevance"
-        options are "relevance" or "all". Only "relevance" is supported for now.
+        options are "relevance" or "all".
    **kwargs
        Additional keyword arguments to pass to the model. For example, 'device'.
        See AnswerDotAI/rerankers for more information.
@@ -77,12 +77,13 @@ class AnswerdotaiRerankers(Reranker):
        vector_results: pa.Table,
        fts_results: pa.Table,
    ):
-        combined_results = self.merge_results(vector_results, fts_results)
+        if self.score == "all":
+            combined_results = self._merge_and_keep_scores(vector_results, fts_results)
+        else:
+            combined_results = self.merge_results(vector_results, fts_results)
        combined_results = self._rerank(combined_results, query)
        if self.score == "relevance":
            combined_results = self._keep_relevance_score(combined_results)
-        elif self.score == "all":
-            combined_results = self._merge_and_keep_scores(vector_results, fts_results)
        combined_results = combined_results.sort_by(
            [("_relevance_score", "descending")]
        )
@@ -16,7 +16,7 @@ class ColbertReranker(AnswerdotaiRerankers):
    column : str, default "text"
        The name of the column to use as input to the cross encoder model.
    return_score : str, default "relevance"
-        options are "relevance" or "all". Only "relevance" is supported for now.
+        options are "relevance" or "all".
    **kwargs
        Additional keyword arguments to pass to the model, for example, 'device'.
        See AnswerDotAI/rerankers for more information.
@@ -40,12 +40,12 @@ class WatsonxReranker(Reranker):
        IBM Cloud API key.  Falls back to the ``WATSONX_API_KEY`` environment
        variable when not provided.
    project_id : str, optional
-        watsonx.ai project ID.  Falls back to the ``WATSONX_PROJECT_ID``
-        environment variable when not provided.  Mutually exclusive with
+        watsonx.ai project ID.  Explicit value takes precedence over the
+        ``WATSONX_PROJECT_ID`` environment variable.  Mutually exclusive with
        ``space_id`` — exactly one must be supplied.
    space_id : str, optional
-        watsonx.ai deployment space ID.  Falls back to the ``WATSONX_SPACE_ID``
-        environment variable when not provided.  Mutually exclusive with
+        watsonx.ai deployment space ID.  Explicit value takes precedence over
+        the ``WATSONX_SPACE_ID`` environment variable.  Mutually exclusive with
        ``project_id`` — exactly one must be supplied.
    url : str, optional
        watsonx.ai service URL.  Defaults to
@@ -100,8 +100,16 @@ class WatsonxReranker(Reranker):
        )

        # --- project_id / space_id (exactly one required) ---
-        project_id = self.project_id or os.environ.get("WATSONX_PROJECT_ID")
-        space_id = self.space_id or os.environ.get("WATSONX_SPACE_ID")
+        # Explicit field always wins; env vars are consulted only when neither
+        # was passed explicitly, so a stray WATSONX_SPACE_ID never overrides an
+        # explicit project_id and vice-versa.
+        project_id = self.project_id
+        space_id = self.space_id
+
+        if project_id is None and space_id is None:
+            # Neither was passed explicitly — fall back to env vars.
+            project_id = os.environ.get("WATSONX_PROJECT_ID")
+            space_id = os.environ.get("WATSONX_SPACE_ID")

        if project_id and space_id:
            raise ValueError("Provide either `project_id` or `space_id`, not both.")
@@ -7,10 +7,158 @@ import sys
 from typing import Callable, Iterator, Optional
 from lancedb.arrow import to_arrow
 import pyarrow as pa
+import pyarrow.compute as pc
 import pyarrow.dataset as ds

 from .pydantic import LanceModel

+# pyarrow's default scanner settings are tuned for narrow rows. For wide rows
+# (e.g. embedding columns) they buffer a huge read-ahead window in host memory
+# and can OOM the client during bulk ingestion. We size the scanner so the
+# estimated in-flight memory stays within a budget, while leaving narrow
+# datasets on pyarrow's defaults (no throughput regression).
+_SCAN_MEMORY_BUDGET_BYTES = 1024 * 1024 * 1024  # ~1 GiB in-flight target
+_TARGET_BATCH_BYTES = 16 * 1024 * 1024  # ~16 MiB per batch
+_MIN_BATCH_ROWS = 512
+# pyarrow defaults (see arrow/dataset ScanOptions); we never exceed these.
+_PA_DEFAULT_BATCH_ROWS = 131_072
+_PA_DEFAULT_BATCH_READAHEAD = 16
+_PA_DEFAULT_FRAGMENT_READAHEAD = 4
+# Read-ahead used for wide rows. pyarrow reads a whole parquet row group at a
+# time and keeps `batch_readahead` of them resident, so read-ahead depth (not
+# batch size) dominates peak memory for wide data; keep both small but leave a
+# little prefetch for throughput. Tuned empirically against embedding datasets.
+_WIDE_BATCH_READAHEAD = 2
+_WIDE_FRAGMENT_READAHEAD = 1
+# Estimate for variable-width columns (string/binary/list) whose true width is
+# unknown from the schema alone. Only needs to be large enough to flag "wide".
+_VARIABLE_WIDTH_ESTIMATE = 128
+# Rows peeked from a rescannable source to refine the list-length guess for
+# variable-length list columns (e.g. embeddings stored as `list<float32>`
+# instead of `list<float32, N>`), whose per-row width the schema can't tell us.
+_SAMPLE_ROWS = 10
+
+
+def _observed_list_length(sample: pa.ChunkedArray) -> Optional[int]:
+    """Average element count per row observed in a list/large_list sample."""
+    if len(sample) == 0:
+        return None
+    mean = pc.mean(pc.list_value_length(sample)).as_py()
+    return None if mean is None else max(1, round(mean))
+
+
+def _estimate_field_width(
+    dtype: pa.DataType, sample: Optional[pa.ChunkedArray] = None
+) -> int:
+    if pa.types.is_fixed_size_list(dtype):
+        return dtype.list_size * _estimate_field_width(dtype.value_type)
+    if pa.types.is_struct(dtype):
+        return sum(
+            _estimate_field_width(
+                dtype.field(i).type,
+                pc.struct_field(sample, [i]) if sample is not None else None,
+            )
+            for i in range(dtype.num_fields)
+        )
+    if pa.types.is_dictionary(dtype):
+        return _estimate_field_width(dtype.value_type)
+    if pa.types.is_fixed_size_binary(dtype):
+        return dtype.byte_width
+    if pa.types.is_boolean(dtype):
+        return 1
+    if (pa.types.is_list(dtype) or pa.types.is_large_list(dtype)) and (
+        sample is not None
+    ):
+        observed_length = _observed_list_length(sample)
+        if observed_length is not None:
+            return observed_length * _estimate_field_width(dtype.value_type)
+    # Fixed-width scalars (ints, floats, temporal, decimal) expose bit_width;
+    # variable-width types (string, binary, list, map, ...) raise ValueError.
+    try:
+        return max(1, dtype.bit_width // 8)
+    except (ValueError, AttributeError):
+        return _VARIABLE_WIDTH_ESTIMATE
+
+
+def _estimate_bytes_per_row(
+    schema: pa.Schema, sample: Optional[pa.Table] = None
+) -> int:
+    return max(
+        1,
+        sum(
+            _estimate_field_width(
+                field.type, sample.column(field.name) if sample is not None else None
+            )
+            for field in schema
+        ),
+    )
+
+
+def _sample_head(head: Callable[..., pa.Table]) -> Optional[pa.Table]:
+    """Best-effort peek at a few rows to refine the bytes-per-row estimate.
+
+    Uses a tight batch size and no read-ahead so the peek itself can't trigger
+    the wide-row memory blowup this module exists to avoid. Returns None (fall
+    back to the schema-only estimate) if sampling isn't possible for any
+    reason, e.g. an empty dataset.
+    """
+    try:
+        sample = head(
+            _SAMPLE_ROWS,
+            batch_size=_SAMPLE_ROWS,
+            batch_readahead=1,
+            fragment_readahead=1,
+        )
+    except Exception:
+        return None
+    return sample if sample.num_rows > 0 else None
+
+
+def _bounded_scanner_kwargs(
+    schema: pa.Schema, sample: Optional[pa.Table] = None
+) -> dict:
+    """Scanner kwargs that cap in-flight memory for wide rows.
+
+    Narrow datasets keep pyarrow's defaults unchanged (no throughput
+    regression). For wide rows (e.g. embedding columns) pyarrow's default
+    read-ahead buffers many large batches/row-groups at once, which can OOM the
+    client during bulk ingestion, so we shrink the batch size and read-ahead to
+    keep the estimated in-flight memory near the budget.
+
+    Read-ahead (not just batch size) has to drop: pyarrow reads a whole parquet
+    row group at a time and keeps `batch_readahead`/`fragment_readahead` of them
+    resident, so a small batch size alone still pins large row-group buffers.
+
+    `sample`, if given, is a small (see `_SAMPLE_ROWS`) table of rows from the
+    source used to refine the estimate for variable-length list columns (e.g.
+    embeddings stored without a fixed size), whose width the schema alone
+    can't tell us.
+    """
+    bytes_per_row = _estimate_bytes_per_row(schema, sample)
+
+    # If pyarrow's defaults already stay within budget, leave them alone so
+    # narrow datasets keep their throughput. A "unit" of in-flight memory is one
+    # default-sized batch, held `batch_readahead + fragment_readahead` deep.
+    default_in_flight = (
+        _PA_DEFAULT_BATCH_ROWS
+        * bytes_per_row
+        * (_PA_DEFAULT_BATCH_READAHEAD + _PA_DEFAULT_FRAGMENT_READAHEAD)
+    )
+    if default_in_flight <= _SCAN_MEMORY_BUDGET_BYTES:
+        return {}
+
+    # Wide rows: cap batch bytes and pull read-ahead down so only a couple of
+    # large row-group buffers are resident at once.
+    batch_size = min(
+        _PA_DEFAULT_BATCH_ROWS,
+        max(_MIN_BATCH_ROWS, _TARGET_BATCH_BYTES // bytes_per_row),
+    )
+    return {
+        "batch_size": batch_size,
+        "batch_readahead": _WIDE_BATCH_READAHEAD,
+        "fragment_readahead": _WIDE_FRAGMENT_READAHEAD,
+    }
+

@dataclass
 class Scannable:
@@ -56,10 +204,12 @@ def _from_table(data: pa.Table) -> Scannable:

@to_scannable.register(ds.Dataset)
 def _from_dataset(data: ds.Dataset) -> Scannable:
+    sample = _sample_head(data.head)
+    scanner_kwargs = _bounded_scanner_kwargs(data.schema, sample)
    return Scannable(
        schema=data.schema,
        num_rows=data.count_rows(),
-        reader=lambda: data.scanner().to_reader(),
+        reader=lambda: data.scanner(**scanner_kwargs).to_reader(),
    )


@@ -206,10 +356,12 @@ def _register_optional_converters():

        @to_scannable.register(lance.LanceDataset)
        def _from_lance(data: lance.LanceDataset) -> Scannable:
+            sample = _sample_head(data.head)
+            scanner_kwargs = _bounded_scanner_kwargs(data.schema, sample)
            return Scannable(
                schema=data.schema,
                num_rows=data.count_rows(),
-                reader=lambda: data.scanner().to_reader(),
+                reader=lambda: data.scanner(**scanner_kwargs).to_reader(),
            )


@@ -1106,6 +1106,7 @@ class Table(ABC):
        ngram_min_length: int = 3,
        ngram_max_length: int = 3,
        prefix_only: bool = False,
+        block_size: int = 128,
        wait_timeout: Optional[timedelta] = None,
        name: Optional[str] = None,
    ):
@@ -1177,6 +1178,10 @@ class Table(ABC):
            The maximum length of an n-gram.
        prefix_only: bool, default False
            Whether to only index the prefix of the token for ngram tokenizer.
+        block_size: int, default 128
+            The number of documents per compressed posting block. Must be 128
+            or 256. A value of 256 uses the experimental FTS V3 format and
+            may introduce breaking changes.
        wait_timeout: timedelta, optional
            The timeout to wait if indexing is asynchronous.
        name: str, optional
@@ -1199,6 +1204,7 @@ class Table(ABC):
        on_bad_vectors: OnBadVectorsType = "error",
        fill_value: float = 0.0,
        progress: Optional[Union[bool, Callable, Any]] = None,
+        write_parallelism: Optional[int] = None,
    ) -> AddResult:
        """Add more data to the [Table](Table).

@@ -1244,6 +1250,13 @@ class Table(ABC):
                with tqdm() as pbar:
                    table.add(data, progress=pbar)

+        write_parallelism: int, optional
+            Number of partitions to write in parallel. Higher values increase
+            throughput but also peak memory use, since each partition buffers
+            data in flight. Defaults to an estimate based on the data size,
+            capped at the number of CPU cores. Lower this if bulk ingestion is
+            using too much memory.
+
        Returns
        -------
        AddResult
@@ -2195,7 +2208,7 @@ class LanceTable(Table):
        namespace_client = self._namespace_client
        if namespace_client is None:
            conn_uri = getattr(self._conn, "uri", "")
-            if get_uri_scheme(conn_uri) == "namespace":
+            if get_uri_scheme(conn_uri) == "namespace" or self._namespace_path:
                namespace_client = self._conn.namespace_client()
                self._namespace_client = namespace_client

@@ -3018,6 +3031,7 @@ class LanceTable(Table):
        ngram_min_length: int = 3,
        ngram_max_length: int = 3,
        prefix_only: bool = False,
+        block_size: int = 128,
        name: Optional[str] = None,
    ):
        """Create a full-text search index on a column.
@@ -3067,9 +3081,7 @@ class LanceTable(Table):
        else:
            tokenizer_configs = self.infer_tokenizer_configs(tokenizer_name)

-        config = FTS(
-            **tokenizer_configs,
-        )
+        config = FTS(block_size=block_size, **tokenizer_configs)

        try:
            LOOP.run(
@@ -3158,6 +3170,7 @@ class LanceTable(Table):
        on_bad_vectors: OnBadVectorsType = "error",
        fill_value: float = 0.0,
        progress: Optional[Union[bool, Callable, Any]] = None,
+        write_parallelism: Optional[int] = None,
    ) -> AddResult:
        """Add data to the table.
        If vector columns are missing and the table
@@ -3179,6 +3192,12 @@ class LanceTable(Table):
        progress: bool, callable, or tqdm-like, optional
            A callback or tqdm-compatible progress bar. See
            :meth:`Table.add` for details.
+        write_parallelism: int, optional
+            Number of partitions to write in parallel. Higher values increase
+            throughput but also peak memory use, since each partition buffers
+            data in flight. Defaults to an estimate based on the data size,
+            capped at the number of CPU cores. Lower this if bulk ingestion is
+            using too much memory.

        Returns
        -------
@@ -3194,6 +3213,7 @@ class LanceTable(Table):
                    on_bad_vectors=on_bad_vectors,
                    fill_value=fill_value,
                    progress=progress,
+                    write_parallelism=write_parallelism,
                )
            )
        finally:
@@ -4936,6 +4956,7 @@ class AsyncTable:
        on_bad_vectors: Optional[OnBadVectorsType] = None,
        fill_value: Optional[float] = None,
        progress: Optional[Union[bool, Callable, Any]] = None,
+        write_parallelism: Optional[int] = None,
    ) -> AddResult:
        """Add more data to the [Table](Table).

@@ -4960,6 +4981,12 @@ class AsyncTable:
        progress: callable or tqdm-like, optional
            A callback or tqdm-compatible progress bar. See
            :meth:`Table.add` for details.
+        write_parallelism: int, optional
+            Number of partitions to write in parallel. Higher values increase
+            throughput but also peak memory use, since each partition buffers
+            data in flight. Defaults to an estimate based on the data size,
+            capped at the number of CPU cores. Lower this if bulk ingestion is
+            using too much memory.

        """
        schema = await self.schema()
@@ -4991,7 +5018,12 @@ class AsyncTable:
        data = to_scannable(data)
        progress, owns = _normalize_progress(progress)
        try:
-            return await self._inner.add(data, mode or "append", progress=progress)
+            return await self._inner.add(
+                data,
+                mode or "append",
+                progress=progress,
+                write_parallelism=write_parallelism,
+            )
        except RuntimeError as e:
            if "Cast error" in str(e):
                raise ValueError(e)
@@ -6239,6 +6271,24 @@ class Branches:
        """Delete a branch."""
        LOOP.run(self._table.branches.delete(name))

+    def diff(self, from_branch: str) -> Dict[str, Any]:
+        """Diff a branch against main."""
+        return LOOP.run(self._table.branches.diff(from_branch))
+
+    def merge(self, from_branch: str, dry_run: bool = False) -> Dict[str, Any]:
+        """Merge a branch into main, or dry-run.
+
+        Parameters
+        ----------
+        from_branch: str
+            Branch to merge from.
+        dry_run: bool, default False
+            When True, only preview. When False, attempt the merge.
+
+        A rejected merge returns ``status="rejected"`` instead of raising.
+        """
+        return LOOP.run(self._table.branches.merge(from_branch, dry_run))
+
    def _wrap(
        self, async_table: "AsyncTable", version: Optional[int] = None
    ) -> "Table":
@@ -6368,3 +6418,14 @@ class AsyncBranches:
    async def delete(self, name: str) -> None:
        """Delete a branch."""
        await self._table.branches.delete(name)
+
+    async def diff(self, from_branch: str) -> Dict[str, Any]:
+        """Diff a branch against main."""
+        return await self._table.branches.diff(from_branch)
+
+    async def merge(self, from_branch: str, dry_run: bool = False) -> Dict[str, Any]:
+        """Merge a branch into main, or dry-run.
+
+        A rejected merge returns ``status="rejected"`` instead of raising.
+        """
+        return await self._table.branches.merge(from_branch, dry_run)
@@ -307,13 +307,19 @@ def infer_vector_column_name(
        # FTS queries do not require a vector column
        return None

-    if query is not None or query_type == "hybrid":
-        try:
-            vector_column_name = inf_vector_column_query(
-                schema, dim=_query_vector_dim(query)
-            )
-        except Exception as e:
-            raise e
+    if query is None and query_type != "hybrid":
+        # No vector search was requested (e.g. a plain scan), so there's
+        # nothing to infer.
+        return None
+
+    vector_column_name = inf_vector_column_query(schema, dim=_query_vector_dim(query))
+
+    if vector_column_name is None:
+        raise ValueError(
+            "No vector column found in the schema. Please specify the "
+            "vector column name explicitly via the `vector_column_name` "
+            "parameter."
+        )

    return vector_column_name

@@ -13,6 +13,7 @@ import numpy as np
 import pandas as pd
 import pyarrow as pa
 import pytest
+from lance_namespace.errors import NamespaceNotEmptyError, TableNotFoundError
 from lancedb.pydantic import LanceModel, Vector


@@ -955,6 +956,47 @@ def test_local_namespace_operations(tmp_path):
    assert db.list_namespaces().namespaces == []


+def test_local_sync_namespace_uses_rust_without_python_client(tmp_path, monkeypatch):
+    """Sync local namespace operations should avoid the Python namespace client."""
+    db = lancedb.connect(tmp_path)
+
+    def fail_namespace_client():
+        raise AssertionError("Python namespace client should not be constructed")
+
+    monkeypatch.setattr(db, "namespace_client", fail_namespace_client)
+
+    db.create_namespace(["child"])
+    assert "child" in db.list_namespaces().namespaces
+
+    schema = pa.schema([pa.field("id", pa.int64())])
+    table = db.create_table("tbl", schema=schema, namespace_path=["child"])
+    assert table.namespace == ["child"]
+    assert "tbl" in db.table_names(namespace_path=["child"])
+    assert db.list_tables(namespace_path=["child"]).tables == ["tbl"]
+
+    opened = db.open_table("tbl", namespace_path=["child"])
+    assert opened.namespace == ["child"]
+
+    db.drop_table("tbl", namespace_path=["child"])
+    assert db.list_tables(namespace_path=["child"]).tables == []
+    db.drop_namespace(["child"])
+    assert db.list_namespaces().namespaces == []
+
+
+def test_local_sync_namespace_preserves_public_errors(tmp_path):
+    db = lancedb.connect(tmp_path)
+    db.create_namespace(["child"])
+    db.create_table(
+        "tbl", schema=pa.schema([pa.field("id", pa.int64())]), namespace_path=["child"]
+    )
+
+    with pytest.raises(TableNotFoundError, match="child\\$missing"):
+        db.open_table("missing", namespace_path=["child"])
+
+    with pytest.raises(NamespaceNotEmptyError):
+        db.drop_namespace(["child"])
+
+
 def test_create_namespace_invalid_mode_raises(tmp_path):
    """Unrecognized create namespace modes raise a clear error."""
    db = lancedb.connect(tmp_path)
@@ -226,6 +226,23 @@ def test_create_inverted_index(table, with_position):
    assert any(i.name == "custom_fts_index" for i in fts_indices)


+@pytest.mark.parametrize("block_size", [128, 256])
+def test_create_inverted_index_block_size(table, block_size):
+    table.create_index("text", config=FTS(block_size=block_size))
+
+    index = next(index for index in table.list_indices() if index.index_type == "FTS")
+    assert index.index_details["block_size"] == block_size
+    assert index.index_version == (2 if block_size == 128 else 3)
+
+    results = table.search("puppy").limit(5).to_list()
+    assert len(results) == 5
+
+
+def test_create_inverted_index_rejects_invalid_block_size(table):
+    with pytest.raises(ValueError, match="128 or 256"):
+        table.create_index("text", config=FTS(block_size=129))
+
+
 def test_search_fts(table):
    table.create_fts_index("text")
    results = table.search("puppy").select(["id", "text"]).limit(5).to_list()
@@ -1136,3 +1136,61 @@ def test_take_offsets_empty_permutation(some_permutation: Permutation):
    result = some_permutation.take_offsets([])

    assert result == []
+
+
+def test_select_rowid(some_permutation: Permutation):
+    """Test that _rowid can be selected alongside regular columns."""
+    perm_with_rowid = some_permutation.select_columns(["_rowid", "id"])
+    assert "_rowid" in perm_with_rowid.column_names
+    batches = list(perm_with_rowid.iter(100, skip_last_batch=False))
+    for batch in batches:
+        assert "_rowid" in batch[0]
+
+
+def test_select_rowid_only(some_permutation: Permutation):
+    """Test that _rowid can be selected as the sole column."""
+    perm_rowid_only = some_permutation.select_columns(["_rowid"])
+    assert perm_rowid_only.column_names == ["_rowid"]
+    batches = list(perm_rowid_only.iter(100, skip_last_batch=False))
+    assert len(batches) > 0
+    for batch in batches:
+        assert list(batch[0].keys()) == ["_rowid"]
+
+
+def test_select_rowid_not_in_default(some_permutation: Permutation):
+    """Test that _rowid is NOT in the default column_names or schema."""
+    assert "_rowid" not in some_permutation.column_names
+    assert "_rowid" not in some_permutation.schema.names
+
+
+def test_select_rowid_identity_permutation(mem_db):
+    """Test that _rowid works with an identity permutation."""
+    tbl = mem_db.create_table(
+        "test_rowid_identity", pa.table({"id": range(10), "value": range(10)})
+    )
+    perm = Permutation.identity(tbl)
+    perm_with_rowid = perm.select_columns(["_rowid", "id"])
+    batches = list(perm_with_rowid.iter(10, skip_last_batch=False))
+    assert len(batches) == 1
+    assert "_rowid" in batches[0][0]
+
+
+def test_rename_rowid(some_permutation: Permutation):
+    """Test that _rowid can be selected and then renamed."""
+    perm_with_rowid = some_permutation.select_columns(["_rowid", "id"])
+    renamed = perm_with_rowid.rename_column("_rowid", "my_row_id")
+    assert "my_row_id" in renamed.column_names
+    assert "_rowid" not in renamed.column_names
+    batches = list(renamed.iter(100, skip_last_batch=False))
+    for batch in batches:
+        assert "my_row_id" in batch[0]
+        assert "_rowid" not in batch[0]
+
+
+def test_remove_rowid_after_select(some_permutation: Permutation):
+    """Test that _rowid can be selected and then removed."""
+    perm_with_rowid = some_permutation.select_columns(["_rowid", "id"])
+    assert "_rowid" in perm_with_rowid.column_names
+    perm_without_rowid = perm_with_rowid.remove_columns(["_rowid"])
+    assert "_rowid" not in perm_without_rowid.column_names
+    assert perm_without_rowid.column_names == ["id"]
@@ -1273,7 +1273,7 @@ async def test_explain_plan_fts(table_async: AsyncTable):
    query = await table_async.search("dog", query_type="fts", fts_columns="text")
    plan = await query.explain_plan()
    # Should show FTS details (issue #2465 is now fixed)
-    assert "MatchQuery: column=text, query=dog" in plan
+    assert "MatchQuery: column=text, query=[dog]" in plan
    assert "GlobalLimitExec" in plan  # Default limit

    # Test FTS query with limit
@@ -1281,7 +1281,7 @@ async def test_explain_plan_fts(table_async: AsyncTable):
        "dog", query_type="fts", fts_columns="text"
    )
    plan_with_limit = await query_with_limit.limit(1).explain_plan()
-    assert "MatchQuery: column=text, query=dog" in plan_with_limit
+    assert "MatchQuery: column=text, query=[dog]" in plan_with_limit
    assert "GlobalLimitExec: skip=0, fetch=1" in plan_with_limit

    # Test FTS query with offset and limit
@@ -1289,7 +1289,7 @@ async def test_explain_plan_fts(table_async: AsyncTable):
        "dog", query_type="fts", fts_columns="text"
    )
    plan_with_offset = await query_with_offset.offset(1).limit(1).explain_plan()
-    assert "MatchQuery: column=text, query=dog" in plan_with_offset
+    assert "MatchQuery: column=text, query=[dog]" in plan_with_offset
    assert "GlobalLimitExec: skip=1, fetch=1" in plan_with_offset


@@ -1333,7 +1333,7 @@ async def test_explain_plan_with_filters(table_async: AsyncTable):
        "dog", query_type="fts", fts_columns="text"
    )
    plan_fts_filter = await query_fts_filter.where("id = 1").explain_plan()
-    assert "MatchQuery: column=text, query=dog" in plan_fts_filter
+    assert "MatchQuery: column=text, query=[dog]" in plan_fts_filter
    assert "LanceRead" in plan_fts_filter
    assert "full_filter=id = Int64(1)" in plan_fts_filter  # Should show filter details

@@ -236,6 +236,65 @@ def test_remote_table_branches_sync():
        table.branches.delete("exp")


+def test_remote_table_branch_merge_defaults_to_execute():
+    merge_bodies = []
+    diff = {
+        "fromBranch": "exp",
+        "parentVersion": 1,
+        "mainVersion": 2,
+        "branchVersion": 3,
+        "baseMoved": False,
+        "rowCountMain": 3,
+        "rowCountBranch": 3,
+        "rowSummary": {
+            "unchanged": 3,
+            "newOnBase": 0,
+            "newOnBranch": 0,
+            "staleRecompute": 0,
+            "inputsChanged": 0,
+            "deltaAvailable": False,
+        },
+        "addedColumns": [],
+        "removedColumns": [],
+        "changedColumns": [],
+        "addedIndexes": [],
+        "removedIndexes": [],
+        "mergeable": True,
+        "mergeBlockers": [],
+    }
+
+    def handler(request):
+        if request.path.endswith("/describe/"):
+            status = 200
+            body = {"version": 2, "schema": {"fields": []}}
+        else:
+            content_len = int(request.headers.get("Content-Length"))
+            request_body = json.loads(request.rfile.read(content_len))
+            merge_bodies.append(request_body)
+            dry_run = request_body["dry_run"]
+            status = 200 if dry_run else 409
+            body = {
+                "status": "ready" if dry_run else "rejected",
+                "diff": diff,
+                "preview": {"promotedColumns": []},
+            }
+
+        request.send_response(status)
+        request.send_header("Content-Type", "application/json")
+        request.end_headers()
+        request.wfile.write(json.dumps(body).encode())
+
+    with mock_lancedb_connection(handler) as db:
+        branches = db.open_table("test").branches
+        assert branches.merge("exp")["status"] == "rejected"
+        assert branches.merge("exp", dry_run=True)["status"] == "ready"
+
+    assert merge_bodies == [
+        {"from_branch": "exp", "dry_run": False},
+        {"from_branch": "exp", "dry_run": True},
+    ]
+
+
@pytest.mark.asyncio
 async def test_async_remote_open_table_branch_and_version():
    async with mock_lancedb_connection_async(_branch_open_handler) as db:
@@ -709,7 +768,10 @@ def test_table_create_indices():
        # Test create_fts_index with custom name (legacy method)
        with pytest.warns(DeprecationWarning, match="create_fts_index"):
            table.create_fts_index(
-                "text", wait_timeout=timedelta(seconds=2), name="custom_fts_idx"
+                "text",
+                wait_timeout=timedelta(seconds=2),
+                block_size=256,
+                name="custom_fts_idx",
            )

        # Test create_index with custom name (legacy form: vector_column_name kwarg)
@@ -732,6 +794,7 @@ def test_table_create_indices():
        fts_req = received_requests[1]
        assert "name" in fts_req
        assert fts_req["name"] == "custom_fts_idx"
+        assert fts_req["block_size"] == 256

        # Check vector index request has custom name
        vector_req = received_requests[2]
@@ -817,7 +880,7 @@ def test_remote_create_index_new_api():
            _warnings.simplefilter("error", DeprecationWarning)
            table.create_index("vector", config=IvfPq(distance_type="l2"))
            table.create_index("category", config=BTree())
-            table.create_index("text", config=FTS())
+            table.create_index("text", config=FTS(block_size=256))
            # IvfRq via new API
            table.create_index("vector", config=IvfRq(distance_type="l2"))

@@ -837,6 +900,7 @@ def test_remote_create_index_new_api():
            "vector",
            "vector",
        ]
+        assert received_requests[2]["block_size"] == 256


 def test_table_wait_for_index_timeout():
@@ -644,6 +644,21 @@ def test_cross_encoder_reranker_return_all(tmp_path):
    assert "_distance" in result.column_names


+def test_answerdotai_reranker_return_all(tmp_path):
+    pytest.importorskip("rerankers")
+    reranker = AnswerdotaiRerankers(return_score="all")
+    table, schema = get_test_table(tmp_path)
+    query = "single player experience"
+    result = (
+        table.search(query, query_type="hybrid", vector_column_name="vector")
+        .rerank(reranker=reranker)
+        .to_arrow()
+    )
+    assert "_relevance_score" in result.column_names
+    assert "_score" in result.column_names
+    assert "_distance" in result.column_names
+
+
 # ---------------------------------------------------------------------------
 # Regression tests for LinearCombinationReranker scoring bugs (issue #3154)
 # ---------------------------------------------------------------------------
@@ -0,0 +1,183 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+import numpy as np
+import pyarrow as pa
+import pyarrow.dataset as ds
+import pyarrow.parquet as pq
+
+from lancedb.scannable import (
+    _PA_DEFAULT_BATCH_ROWS,
+    _SAMPLE_ROWS,
+    _VARIABLE_WIDTH_ESTIMATE,
+    _WIDE_BATCH_READAHEAD,
+    _WIDE_FRAGMENT_READAHEAD,
+    _bounded_scanner_kwargs,
+    _estimate_bytes_per_row,
+    _sample_head,
+    to_scannable,
+)
+
+
+def test_estimate_bytes_per_row():
+    # fixed-width scalars
+    assert _estimate_bytes_per_row(pa.schema([("a", pa.int64())])) == 8
+    assert (
+        _estimate_bytes_per_row(pa.schema([("a", pa.int32()), ("b", pa.float64())]))
+        == 12
+    )
+    assert _estimate_bytes_per_row(pa.schema([("a", pa.bool_())])) == 1
+    # fixed-size list (embedding) dominates
+    assert (
+        _estimate_bytes_per_row(pa.schema([("v", pa.list_(pa.float32(), 768))]))
+        == 768 * 4
+    )
+    # struct sums its children
+    struct = pa.struct([("x", pa.int32()), ("y", pa.int32())])
+    assert _estimate_bytes_per_row(pa.schema([("s", struct)])) == 8
+    # variable-width columns get a flat estimate, not zero
+    assert _estimate_bytes_per_row(pa.schema([("s", pa.string())])) > 0
+
+
+def test_estimate_bytes_per_row_uses_sample_for_variable_length_lists():
+    # A vector column without a fixed size (e.g. `list<float32>` instead of
+    # `list<float32, 768>`) has no width the schema alone can tell us.
+    schema = pa.schema([("v", pa.list_(pa.float32()))])
+    assert _estimate_bytes_per_row(schema) == _VARIABLE_WIDTH_ESTIMATE
+
+    sample = pa.table({"v": pa.array([[0.0] * 768], type=pa.list_(pa.float32()))})
+    assert _estimate_bytes_per_row(schema, sample) == 768 * 4
+
+
+def test_estimate_bytes_per_row_sample_ignores_missing_or_null_lists():
+    schema = pa.schema([("v", pa.list_(pa.float32()))])
+    # an all-null sample column can't tell us anything either
+    sample = pa.table({"v": pa.array([None], type=pa.list_(pa.float32()))})
+    assert _estimate_bytes_per_row(schema, sample) == _VARIABLE_WIDTH_ESTIMATE
+
+
+def test_bounded_scanner_kwargs_narrow_uses_defaults():
+    # Narrow rows stay on pyarrow defaults (empty kwargs) so throughput is
+    # unchanged.
+    for schema in [
+        pa.schema([("a", pa.int64()), ("b", pa.int32()), ("c", pa.string())]),
+        pa.schema([("a", pa.int64()), ("t", pa.string()), ("u", pa.string())]),
+        # a 100-dim float32 vector is still under the per-row budget
+        pa.schema([("id", pa.int64()), ("v", pa.list_(pa.float32(), 100))]),
+    ]:
+        assert _bounded_scanner_kwargs(schema) == {}, schema
+
+
+def test_bounded_scanner_kwargs_wide_is_bounded():
+    schema = pa.schema(
+        [
+            ("uid", pa.string()),
+            ("img", pa.list_(pa.float32(), 768)),
+            ("txt", pa.list_(pa.float32(), 768)),
+        ]
+    )
+    kwargs = _bounded_scanner_kwargs(schema)
+    assert kwargs, "wide schema should be throttled"
+    assert kwargs["batch_readahead"] == _WIDE_BATCH_READAHEAD
+    assert kwargs["fragment_readahead"] == _WIDE_FRAGMENT_READAHEAD
+    # batch is capped well below the pyarrow default for wide rows
+    assert kwargs["batch_size"] < _PA_DEFAULT_BATCH_ROWS
+
+
+def test_bounded_scanner_kwargs_variable_length_list_needs_sample():
+    # Without a sample, a variable-length (not fixed-size) vector column looks
+    # narrow because its true width is unknown from the schema alone.
+    schema = pa.schema([("uid", pa.string()), ("vec", pa.list_(pa.float32()))])
+    assert _bounded_scanner_kwargs(schema) == {}
+
+    sample = pa.table(
+        {
+            "uid": pa.array(["a"]),
+            "vec": pa.array([[0.0] * 768], type=pa.list_(pa.float32())),
+        }
+    )
+    kwargs = _bounded_scanner_kwargs(schema, sample)
+    assert kwargs, "sample should reveal the wide vector column"
+    assert kwargs["batch_size"] < _PA_DEFAULT_BATCH_ROWS
+
+
+def _write_wide_dataset(
+    path, *, files=2, rows_per_file=20_000, dim=768, fixed_size=True
+):
+    rng = np.random.default_rng(0)
+    for i in range(files):
+        emb = rng.standard_normal((rows_per_file, dim), dtype=np.float32)
+        vec_type = pa.list_(pa.float32(), dim) if fixed_size else pa.list_(pa.float32())
+        vec_array = (
+            pa.FixedSizeListArray.from_arrays(pa.array(emb.reshape(-1)), dim)
+            if fixed_size
+            else pa.array(emb.tolist(), type=vec_type)
+        )
+        table = pa.table(
+            {
+                "uid": pa.array([f"{i}_{j}" for j in range(rows_per_file)]),
+                "vec": vec_array,
+            }
+        )
+        pq.write_table(table, f"{path}/part-{i}.parquet")
+
+
+def test_dataset_reader_respects_bounded_batch_size(tmp_path):
+    # The Dataset path should stream small batches for wide rows, not pyarrow's
+    # 131072-row default, and still return every row.
+    _write_wide_dataset(str(tmp_path))
+    dataset = ds.dataset(str(tmp_path), format="parquet")
+
+    expected = _bounded_scanner_kwargs(dataset.schema)["batch_size"]
+
+    scannable = to_scannable(dataset)
+    assert scannable.rescannable
+    assert scannable.num_rows == 40_000
+
+    total = 0
+    for batch in scannable.reader():
+        assert batch.num_rows <= expected
+        total += batch.num_rows
+    assert total == 40_000
+
+    # factory can be called again (rescannable)
+    assert sum(b.num_rows for b in scannable.reader()) == 40_000
+
+
+def test_dataset_reader_samples_variable_length_list_width(tmp_path):
+    # A vector column stored without a fixed size (e.g. produced by tools that
+    # don't tag list columns with their length) is invisible to the
+    # schema-only estimate, so `to_scannable` must peek a sample of rows to
+    # detect that it's wide and bound the scanner accordingly.
+    _write_wide_dataset(str(tmp_path), fixed_size=False)
+    dataset = ds.dataset(str(tmp_path), format="parquet")
+
+    schema_only_kwargs = _bounded_scanner_kwargs(dataset.schema)
+    assert schema_only_kwargs == {}, "schema alone can't see the list width"
+
+    scannable = to_scannable(dataset)
+    assert scannable.rescannable
+    assert scannable.num_rows == 40_000
+
+    total = 0
+    for batch in scannable.reader():
+        assert batch.num_rows < _PA_DEFAULT_BATCH_ROWS
+        total += batch.num_rows
+    assert total == 40_000
+
+
+def test_sample_head_is_bounded_rows(tmp_path):
+    # The peek itself must not read the whole dataset.
+    _write_wide_dataset(str(tmp_path), files=1, rows_per_file=1000, fixed_size=False)
+    dataset = ds.dataset(str(tmp_path), format="parquet")
+
+    sample = _sample_head(dataset.head)
+    assert sample.num_rows == _SAMPLE_ROWS
+
+
+def test_sample_head_returns_none_for_empty_dataset(tmp_path):
+    table = pa.table({"v": pa.array([], type=pa.list_(pa.float32()))})
+    pq.write_table(table, f"{tmp_path}/empty.parquet")
+    dataset = ds.dataset(str(tmp_path), format="parquet")
+
+    assert _sample_head(dataset.head) is None
@@ -434,6 +434,29 @@ def test_add(mem_db: DBConnection):
    _add(table, schema)


+def test_add_write_parallelism(mem_db: DBConnection):
+    schema = pa.schema([pa.field("id", pa.int64())])
+    table = mem_db.create_table("test", schema=schema)
+
+    data = pa.table({"id": list(range(1000))}, schema=schema)
+    table.add(data, write_parallelism=4)
+    assert len(table) == 1000
+
+    # invalid parallelism is rejected
+    with pytest.raises(ValueError, match="write_parallelism"):
+        table.add(data, write_parallelism=0)
+
+
+@pytest.mark.asyncio
+async def test_add_write_parallelism_async(mem_db_async: AsyncConnection):
+    schema = pa.schema([pa.field("id", pa.int64())])
+    table = await mem_db_async.create_table("test", schema=schema)
+
+    data = pa.table({"id": list(range(1000))}, schema=schema)
+    await table.add(data, write_parallelism=4)
+    assert await table.count_rows() == 1000
+
+
 def test_add_struct(mem_db: DBConnection):
    # https://github.com/lancedb/lancedb/issues/2114
    schema = pa.schema(
@@ -924,3 +924,23 @@ def test_sanitize_data_stream():

    with pytest.raises(ValueError):
        next(output)
+
+
+def test_infer_vector_column_raises_clear_error(tmp_path):
+    """Regression: querying a table with no inferable vector column should raise
+    a clear ValueError, not a cryptic TypeError (issue #1653).
+
+    Previously, inf_vector_column_query silently returned None which then caused
+    a confusing TypeError deep in schema lookup. The fix adds a ValueError guard
+    so the user gets a direct, actionable error message.
+    """
+    db = lancedb.connect(tmp_path)
+    table = db.create_table(
+        "no_vec",
+        data=[{"id": 1, "text": "hello"}, {"id": 2, "text": "world"}],
+    )
+
+    with pytest.raises(ValueError, match="vector"):
+        # Plain vector search on a table with no vector column should raise
+        # a clear ValueError, not a cryptic TypeError.
+        table.search([1.0, 2.0]).to_list()
@@ -0,0 +1,506 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+"""Unit tests for WatsonxEmbeddings — no live API calls required."""
+
+import pytest
+from unittest.mock import MagicMock, patch
+
+from lancedb.embeddings import get_registry
+from lancedb.embeddings.watsonx import CURRENT_MODELS, MODELS_DIMS, WatsonxEmbeddings
+
+
+# ---------------------------------------------------------------------------
+# Helpers
+# ---------------------------------------------------------------------------
+
+
+def _make_func(monkeypatch, env=None, **create_kwargs):
+    """
+    Return a WatsonxEmbeddings instance with ibm_watsonx_ai mocked out.
+
+    Parameters
+    ----------
+    env : dict, optional
+        Environment variables to inject (merged on top of an empty env so that
+        no real WATSONX_* vars from the host bleed into the test).
+    create_kwargs :
+        Forwarded to ``WatsonxEmbeddings.create()``.
+    """
+    base_env = {
+        k: "" for k in ("WATSONX_API_KEY", "WATSONX_PROJECT_ID", "WATSONX_SPACE_ID")
+    }
+    base_env.update(env or {})
+    # Only keep keys that have non-empty values so that absent vars are truly absent.
+    clean_env = {k: v for k, v in base_env.items() if v}
+
+    mock_embeddings_instance = MagicMock()
+    mock_foundation = MagicMock()
+    mock_foundation.Embeddings.return_value = mock_embeddings_instance
+    mock_ibm = MagicMock()
+
+    def _fake_import(name):
+        if name == "ibm_watsonx_ai":
+            return mock_ibm
+        if name == "ibm_watsonx_ai.foundation_models":
+            return mock_foundation
+        raise ImportError(name)
+
+    with patch.dict("os.environ", clean_env, clear=True):
+        with patch(
+            "lancedb.embeddings.watsonx.attempt_import_or_raise",
+            side_effect=_fake_import,
+        ):
+            func = get_registry().get("watsonx").create(**create_kwargs)
+            # Force the cached_property to evaluate inside the patch context.
+            _ = func._watsonx_client
+    return func, mock_foundation
+
+
+# ---------------------------------------------------------------------------
+# Registry
+# ---------------------------------------------------------------------------
+
+
+class TestRegistry:
+    def test_watsonx_registered(self):
+        assert get_registry().get("watsonx") is not None
+
+    def test_model_names_returns_only_current_models(self):
+        names = WatsonxEmbeddings.model_names()
+        assert names == list(CURRENT_MODELS.keys())
+        # Current models must all be present.
+        for name in (
+            "ibm/granite-embedding-278m-multilingual",
+            "ibm/slate-125m-english-rtrvr-v2",
+            "ibm/slate-30m-english-rtrvr-v2",
+            "intfloat/multilingual-e5-large",
+        ):
+            assert name in names, f"{name!r} missing from model_names()"
+        # Legacy / deprecated IDs must NOT appear in model_names().
+        for legacy in (
+            "ibm/slate-125m-english-rtrvr",
+            "ibm/slate-30m-english-rtrvr",
+            "sentence-transformers/all-minilm-l12-v2",
+            "sentence-transformers/all-minilm-l6-v2",
+        ):
+            assert legacy not in names, (
+                f"Legacy model {legacy!r} should not appear in model_names()"
+            )
+
+
+# ---------------------------------------------------------------------------
+# Dimensions
+# ---------------------------------------------------------------------------
+
+
+class TestDimensions:
+    @pytest.mark.parametrize(
+        "model_name,expected_dims",
+        [
+            ("ibm/granite-embedding-278m-multilingual", 768),
+            ("ibm/slate-125m-english-rtrvr-v2", 768),
+            ("ibm/slate-30m-english-rtrvr-v2", 384),
+            ("intfloat/multilingual-e5-large", 1024),
+            ("sentence-transformers/all-minilm-l6-v2", 384),
+        ],
+    )
+    def test_current_model_dimensions(self, monkeypatch, model_name, expected_dims):
+        func, _ = _make_func(
+            monkeypatch,
+            env={"WATSONX_API_KEY": "key", "WATSONX_PROJECT_ID": "proj"},
+            name=model_name,
+        )
+        assert func.ndims() == expected_dims
+
+    def test_unknown_model_raises(self):
+        func = WatsonxEmbeddings(name="not/a-real-model")
+        with pytest.raises(ValueError, match="Unknown model"):
+            func.ndims()
+
+    # -- Backward-compat: legacy names must still resolve dims on table load --
+    @pytest.mark.parametrize(
+        "legacy_name,expected_dims",
+        [
+            ("ibm/slate-125m-english-rtrvr", 768),
+            ("ibm/slate-30m-english-rtrvr", 384),
+            ("sentence-transformers/all-minilm-l12-v2", 384),
+        ],
+    )
+    def test_legacy_model_dimensions_still_resolve(self, legacy_name, expected_dims):
+        """Tables written with old model names must not raise on reload."""
+        assert MODELS_DIMS[legacy_name] == expected_dims
+
+
+# ---------------------------------------------------------------------------
+# Scope resolution (project_id / space_id)
+# ---------------------------------------------------------------------------
+
+
+class TestScopeResolution:
+    def test_explicit_project_id(self, monkeypatch):
+        func, mock_foundation = _make_func(
+            monkeypatch,
+            env={"WATSONX_API_KEY": "key"},
+            project_id="explicit-proj",
+        )
+        _, call_kwargs = mock_foundation.Embeddings.call_args
+        assert call_kwargs.get("project_id") == "explicit-proj"
+        assert "space_id" not in call_kwargs
+
+    def test_explicit_space_id(self, monkeypatch):
+        func, mock_foundation = _make_func(
+            monkeypatch,
+            env={"WATSONX_API_KEY": "key"},
+            space_id="explicit-space",
+        )
+        _, call_kwargs = mock_foundation.Embeddings.call_args
+        assert call_kwargs.get("space_id") == "explicit-space"
+        assert "project_id" not in call_kwargs
+
+    def test_env_project_id_fallback(self, monkeypatch):
+        func, mock_foundation = _make_func(
+            monkeypatch,
+            env={"WATSONX_API_KEY": "key", "WATSONX_PROJECT_ID": "env-proj"},
+        )
+        _, call_kwargs = mock_foundation.Embeddings.call_args
+        assert call_kwargs.get("project_id") == "env-proj"
+
+    def test_env_space_id_fallback(self, monkeypatch):
+        func, mock_foundation = _make_func(
+            monkeypatch,
+            env={"WATSONX_API_KEY": "key", "WATSONX_SPACE_ID": "env-space"},
+        )
+        _, call_kwargs = mock_foundation.Embeddings.call_args
+        assert call_kwargs.get("space_id") == "env-space"
+
+    def test_explicit_project_id_wins_over_env_space_id(self, monkeypatch):
+        """Explicit project_id must not be overridden by WATSONX_SPACE_ID in env."""
+        func, mock_foundation = _make_func(
+            monkeypatch,
+            env={"WATSONX_API_KEY": "key", "WATSONX_SPACE_ID": "stray-env-space"},
+            project_id="explicit-proj",
+        )
+        _, call_kwargs = mock_foundation.Embeddings.call_args
+        assert call_kwargs.get("project_id") == "explicit-proj"
+        assert "space_id" not in call_kwargs
+
+    def test_explicit_space_id_wins_over_env_project_id(self, monkeypatch):
+        """Explicit space_id must not be overridden by WATSONX_PROJECT_ID in env."""
+        func, mock_foundation = _make_func(
+            monkeypatch,
+            env={"WATSONX_API_KEY": "key", "WATSONX_PROJECT_ID": "stray-env-proj"},
+            space_id="explicit-space",
+        )
+        _, call_kwargs = mock_foundation.Embeddings.call_args
+        assert call_kwargs.get("space_id") == "explicit-space"
+        assert "project_id" not in call_kwargs
+
+    def test_both_env_vars_raises(self, monkeypatch):
+        """When both WATSONX_PROJECT_ID and WATSONX_SPACE_ID env vars are set
+        (and neither is passed explicitly), it must raise 'not both'."""
+        func = WatsonxEmbeddings(name="ibm/granite-embedding-278m-multilingual")
+        mock_ibm = MagicMock()
+        mock_foundation = MagicMock()
+
+        def _fake_import(name):
+            if name == "ibm_watsonx_ai":
+                return mock_ibm
+            if name == "ibm_watsonx_ai.foundation_models":
+                return mock_foundation
+            raise ImportError(name)
+
+        with patch.dict(
+            "os.environ",
+            {
+                "WATSONX_API_KEY": "key",
+                "WATSONX_PROJECT_ID": "env-proj",
+                "WATSONX_SPACE_ID": "env-space",
+            },
+            clear=True,
+        ):
+            with patch(
+                "lancedb.embeddings.watsonx.attempt_import_or_raise",
+                side_effect=_fake_import,
+            ):
+                with pytest.raises(ValueError, match="not both"):
+                    _ = func._watsonx_client
+
+    def test_both_explicit_raises(self):
+        func = WatsonxEmbeddings(
+            name="ibm/granite-embedding-278m-multilingual",
+            project_id="p",
+            space_id="s",
+        )
+        # The error surfaces when _watsonx_client is first accessed.
+        mock_ibm = MagicMock()
+        mock_foundation = MagicMock()
+
+        def _fake_import(name):
+            if name == "ibm_watsonx_ai":
+                return mock_ibm
+            if name == "ibm_watsonx_ai.foundation_models":
+                return mock_foundation
+            raise ImportError(name)
+
+        with patch.dict("os.environ", {"WATSONX_API_KEY": "key"}, clear=True):
+            with patch(
+                "lancedb.embeddings.watsonx.attempt_import_or_raise",
+                side_effect=_fake_import,
+            ):
+                with pytest.raises(ValueError, match="not both"):
+                    _ = func._watsonx_client
+
+    def test_neither_raises(self):
+        func = WatsonxEmbeddings(name="ibm/granite-embedding-278m-multilingual")
+        mock_ibm = MagicMock()
+        mock_foundation = MagicMock()
+
+        def _fake_import(name):
+            if name == "ibm_watsonx_ai":
+                return mock_ibm
+            if name == "ibm_watsonx_ai.foundation_models":
+                return mock_foundation
+            raise ImportError(name)
+
+        with patch.dict("os.environ", {"WATSONX_API_KEY": "key"}, clear=True):
+            with patch(
+                "lancedb.embeddings.watsonx.attempt_import_or_raise",
+                side_effect=_fake_import,
+            ):
+                with pytest.raises(
+                    ValueError, match="WATSONX_PROJECT_ID or WATSONX_SPACE_ID"
+                ):
+                    _ = func._watsonx_client
+
+    def test_missing_api_key_raises(self):
+        func = WatsonxEmbeddings(name="ibm/granite-embedding-278m-multilingual")
+        mock_ibm = MagicMock()
+        mock_foundation = MagicMock()
+
+        def _fake_import(name):
+            if name == "ibm_watsonx_ai":
+                return mock_ibm
+            if name == "ibm_watsonx_ai.foundation_models":
+                return mock_foundation
+            raise ImportError(name)
+
+        with patch.dict("os.environ", {"WATSONX_PROJECT_ID": "proj"}, clear=True):
+            with patch(
+                "lancedb.embeddings.watsonx.attempt_import_or_raise",
+                side_effect=_fake_import,
+            ):
+                with pytest.raises(ValueError, match="WATSONX_API_KEY"):
+                    _ = func._watsonx_client
+
+
+# ---------------------------------------------------------------------------
+# Metadata round-trip (backward compat)
+# ---------------------------------------------------------------------------
+
+
+class TestMetadataRoundTrip:
+    def test_reload_with_empty_model_metadata_preserves_model(self):
+        """
+        Reproduce the exact deserialization path used by the registry:
+
+            create(**{})  →  safe_model_dump() == {}
+            →  stored as  model: {}
+            →  reloaded via  create(**{})
+
+        The model must be identical before and after — no silent switch.
+        This guards against changing the class-level default between releases.
+        """
+        from lancedb.embeddings.registry import EmbeddingFunctionRegistry
+
+        registry = EmbeddingFunctionRegistry.get_instance()
+
+        # Simulate original table creation with no explicit args.
+        original = registry.get("watsonx").create()
+        stored = original.safe_model_dump()  # what gets written to arrow metadata
+
+        assert stored == {}, (
+            f"Expected empty stored args when create() called with no kwargs; "
+            f"got {stored!r}"
+        )
+
+        # Simulate reload: registry calls create(**stored) == create(**{})
+        reloaded = registry.get("watsonx").create(**stored)
+
+        assert reloaded.name == original.name, (
+            f"Model changed on reload: was {original.name!r}, "
+            f"became {reloaded.name!r}. "
+            "The class-level default must not change without a migration path."
+        )
+
+    def test_reload_from_legacy_metadata_explicit(self):
+        """
+        Deserialize a representative legacy metadata payload and assert the exact
+        model name — this is the real cross-version regression guard.
+
+        Tables created before the v2 rename stored ``model: {"name": ...}`` with
+        the pre-v2 name.  Reloading must produce exactly that model, not silently
+        switch to the current class default.
+        """
+        from lancedb.embeddings.registry import EmbeddingFunctionRegistry
+
+        registry = EmbeddingFunctionRegistry.get_instance()
+
+        # This is what is stored in Arrow metadata for a table created with the
+        # pre-v2 default model name (no explicit name= was passed at the time).
+        legacy_stored = {"name": "ibm/slate-125m-english-rtrvr"}
+
+        reloaded = registry.get("watsonx").create(**legacy_stored)
+
+        assert reloaded.name == "ibm/slate-125m-english-rtrvr", (
+            f"Legacy metadata reload returned {reloaded.name!r} instead of "
+            "'ibm/slate-125m-english-rtrvr'. "
+            "MODELS_DIMS must keep legacy entries for backward compat."
+        )
+
+    def test_legacy_model_names_resolve_dims(self):
+        """Legacy names in MODELS_DIMS so ndims() never raises on old tables."""
+        assert MODELS_DIMS["ibm/slate-125m-english-rtrvr"] == 768
+        assert MODELS_DIMS["ibm/slate-30m-english-rtrvr"] == 384
+        assert MODELS_DIMS["sentence-transformers/all-minilm-l12-v2"] == 384
+
+
+# ---------------------------------------------------------------------------
+# WatsonxReranker — scope resolution (project_id / space_id)
+# ---------------------------------------------------------------------------
+
+
+def _make_reranker(env=None, **init_kwargs):
+    """
+    Return a WatsonxReranker with ibm_watsonx_ai mocked out.
+
+    Scope precedence is tested by inspecting what was passed to Rerank().
+    """
+    from lancedb.rerankers.watsonx import WatsonxReranker
+
+    base_env = {
+        k: "" for k in ("WATSONX_API_KEY", "WATSONX_PROJECT_ID", "WATSONX_SPACE_ID")
+    }
+    base_env.update(env or {})
+    clean_env = {k: v for k, v in base_env.items() if v}
+
+    mock_rerank_instance = MagicMock()
+    mock_foundation = MagicMock()
+    mock_foundation.Rerank.return_value = mock_rerank_instance
+    mock_ibm = MagicMock()
+
+    def _fake_import(name):
+        if name == "ibm_watsonx_ai":
+            return mock_ibm
+        if name == "ibm_watsonx_ai.foundation_models":
+            return mock_foundation
+        raise ImportError(name)
+
+    reranker = WatsonxReranker(**init_kwargs)
+    with patch.dict("os.environ", clean_env, clear=True):
+        with patch(
+            "lancedb.rerankers.watsonx.attempt_import_or_raise",
+            side_effect=_fake_import,
+        ):
+            _ = reranker._client
+    return reranker, mock_foundation
+
+
+class TestRerankerScopeResolution:
+    def test_explicit_project_id(self):
+        _, mock_foundation = _make_reranker(
+            env={"WATSONX_API_KEY": "key"},
+            project_id="explicit-proj",
+        )
+        _, call_kwargs = mock_foundation.Rerank.call_args
+        assert call_kwargs.get("project_id") == "explicit-proj"
+        assert "space_id" not in call_kwargs
+
+    def test_explicit_space_id(self):
+        _, mock_foundation = _make_reranker(
+            env={"WATSONX_API_KEY": "key"},
+            space_id="explicit-space",
+        )
+        _, call_kwargs = mock_foundation.Rerank.call_args
+        assert call_kwargs.get("space_id") == "explicit-space"
+        assert "project_id" not in call_kwargs
+
+    def test_env_project_id_fallback(self):
+        _, mock_foundation = _make_reranker(
+            env={"WATSONX_API_KEY": "key", "WATSONX_PROJECT_ID": "env-proj"},
+        )
+        _, call_kwargs = mock_foundation.Rerank.call_args
+        assert call_kwargs.get("project_id") == "env-proj"
+
+    def test_env_space_id_fallback(self):
+        _, mock_foundation = _make_reranker(
+            env={"WATSONX_API_KEY": "key", "WATSONX_SPACE_ID": "env-space"},
+        )
+        _, call_kwargs = mock_foundation.Rerank.call_args
+        assert call_kwargs.get("space_id") == "env-space"
+
+    def test_explicit_project_id_wins_over_env_space_id(self):
+        """Explicit project_id must not be overridden by WATSONX_SPACE_ID in env."""
+        _, mock_foundation = _make_reranker(
+            env={"WATSONX_API_KEY": "key", "WATSONX_SPACE_ID": "stray-env-space"},
+            project_id="explicit-proj",
+        )
+        _, call_kwargs = mock_foundation.Rerank.call_args
+        assert call_kwargs.get("project_id") == "explicit-proj"
+        assert "space_id" not in call_kwargs
+
+    def test_explicit_space_id_wins_over_env_project_id(self):
+        """Explicit space_id must not be overridden by WATSONX_PROJECT_ID in env."""
+        _, mock_foundation = _make_reranker(
+            env={"WATSONX_API_KEY": "key", "WATSONX_PROJECT_ID": "stray-env-proj"},
+            space_id="explicit-space",
+        )
+        _, call_kwargs = mock_foundation.Rerank.call_args
+        assert call_kwargs.get("space_id") == "explicit-space"
+        assert "project_id" not in call_kwargs
+
+    def test_both_explicit_raises(self):
+        from lancedb.rerankers.watsonx import WatsonxReranker
+
+        reranker = WatsonxReranker(project_id="p", space_id="s")
+        mock_ibm = MagicMock()
+        mock_foundation = MagicMock()
+
+        def _fake_import(name):
+            if name == "ibm_watsonx_ai":
+                return mock_ibm
+            if name == "ibm_watsonx_ai.foundation_models":
+                return mock_foundation
+            raise ImportError(name)
+
+        with patch.dict("os.environ", {"WATSONX_API_KEY": "key"}, clear=True):
+            with patch(
+                "lancedb.rerankers.watsonx.attempt_import_or_raise",
+                side_effect=_fake_import,
+            ):
+                with pytest.raises(ValueError, match="not both"):
+                    _ = reranker._client
+
+    def test_neither_raises(self):
+        from lancedb.rerankers.watsonx import WatsonxReranker
+
+        reranker = WatsonxReranker()
+        mock_ibm = MagicMock()
+        mock_foundation = MagicMock()
+
+        def _fake_import(name):
+            if name == "ibm_watsonx_ai":
+                return mock_ibm
+            if name == "ibm_watsonx_ai.foundation_models":
+                return mock_foundation
+            raise ImportError(name)
+
+        with patch.dict("os.environ", {"WATSONX_API_KEY": "key"}, clear=True):
+            with patch(
+                "lancedb.rerankers.watsonx.attempt_import_or_raise",
+                side_effect=_fake_import,
+            ):
+                with pytest.raises(
+                    ValueError, match="WATSONX_PROJECT_ID or WATSONX_SPACE_ID"
+                ):
+                    _ = reranker._client
@@ -800,6 +800,10 @@ impl From<PyClientConfig> for lancedb::remote::ClientConfig {
            tls_config: value.tls_config.map(Into::into),
            header_provider,
            user_id: value.user_id,
+            // Resolved from LANCE_CLIENT_MAX_BYTES_PER_REQUEST or the default.
+            max_bytes_per_request: None,
+            // Resolved from LANCE_CLIENT_MAX_REQUEST_DURATION or the read timeout.
+            max_request_duration: None,
        }
    }
 }
@@ -60,6 +60,9 @@ pub fn extract_index_params(source: &Option<Bound<'_, PyAny>>) -> PyResult<Lance
                    .ngram_min_length(params.ngram_min_length)
                    .ngram_max_length(params.ngram_max_length)
                    .ngram_prefix_only(params.prefix_only);
+                let inner_opts = inner_opts
+                    .block_size(params.block_size)
+                    .map_err(|err| PyValueError::new_err(err.to_string()))?;
                Ok(LanceDbIndex::FTS(inner_opts))
            }
            "IvfFlat" => {
@@ -207,6 +210,7 @@ struct FtsParams {
    ngram_min_length: u32,
    ngram_max_length: u32,
    prefix_only: bool,
+    block_size: usize,
 }

 #[derive(FromPyObject)]
@@ -625,12 +625,13 @@ impl Table {
        })
    }

-    #[pyo3(signature = (data, mode, progress=None))]
+    #[pyo3(signature = (data, mode, progress=None, write_parallelism=None))]
    pub fn add<'a>(
        self_: PyRef<'a, Self>,
        data: PyScannable,
        mode: String,
        progress: Option<Py<PyAny>>,
+        write_parallelism: Option<usize>,
    ) -> PyResult<Bound<'a, PyAny>> {
        let mut op = self_.inner_ref()?.add(data);
        if mode == "append" {
@@ -640,6 +641,9 @@ impl Table {
        } else {
            return Err(PyValueError::new_err(format!("Invalid mode: {}", mode)));
        }
+        if let Some(write_parallelism) = write_parallelism {
+            op = op.write_parallelism(write_parallelism);
+        }
        if let Some(progress_obj) = progress {
            let is_callable = Python::attach(|py| progress_obj.bind(py).is_callable());
            if is_callable {
@@ -1589,4 +1593,40 @@ impl Branches {
            Ok(())
        })
    }
+
+    pub fn diff(self_: PyRef<'_, Self>, from_branch: String) -> PyResult<Bound<'_, PyAny>> {
+        let inner = self_.inner.clone();
+        future_into_py(self_.py(), async move {
+            let diff = inner.diff_branch(&from_branch).await.infer_error()?;
+            Python::attach(|py| struct_to_wire_py(py, &diff))
+        })
+    }
+
+    #[pyo3(signature = (from_branch, dry_run=false))]
+    pub fn merge(
+        self_: PyRef<'_, Self>,
+        from_branch: String,
+        dry_run: bool,
+    ) -> PyResult<Bound<'_, PyAny>> {
+        let inner = self_.inner.clone();
+        future_into_py(self_.py(), async move {
+            let result = inner
+                .merge_branch(&from_branch, dry_run)
+                .await
+                .infer_error()?;
+            Python::attach(|py| struct_to_wire_py(py, &result))
+        })
+    }
+}
+
+/// Decode a serde value as the wire JSON object (camelCase keys).
+fn struct_to_wire_py(py: Python<'_>, value: &impl serde::Serialize) -> PyResult<Py<PyAny>> {
+    let json = py.import("json")?;
+    Ok(json
+        .call_method1(
+            "loads",
+            (serde_json::to_string(value)
+                .map_err(|e| PyRuntimeError::new_err(format!("failed to serialize json: {e}")))?,),
+        )?
+        .unbind())
 }
@@ -34,7 +34,6 @@ datafusion.workspace = true
 object_store = { workspace = true }
 snafu = { workspace = true }
 half = { workspace = true }
-lazy_static.workspace = true
 lance = { workspace = true }
 lance-core = { workspace = true }
 lance-datafusion.workspace = true
@@ -273,7 +273,11 @@ pub(crate) async fn take_blobs_aligned(
        if *is_null {
            builder.append_null();
        } else {
-            builder.append_value(payloads[payload_idx].data.as_ref());
+            if let Some(data) = &payloads[payload_idx].data {
+                builder.append_value(data);
+            } else {
+                builder.append_null();
+            }
            payload_idx += 1;
        }
    }
@@ -315,7 +319,7 @@ pub(crate) async fn take_blob_files_aligned(
            if *is_null {
                None
            } else {
-                Some(handles.next().unwrap())
+                handles.next().flatten()
            }
        })
        .collect())
@@ -61,29 +61,29 @@ pub fn is_in(expr: Expr, list: Vec<Expr>) -> Expr {
    expr.in_list(list, false)
 }

-lazy_static::lazy_static! {
-    static ref FUNC_REGISTRY: std::sync::RwLock<std::collections::HashMap<String, Arc<ScalarUDF>>> = {
+static FUNC_REGISTRY: std::sync::LazyLock<std::collections::HashMap<String, Arc<ScalarUDF>>> =
+    std::sync::LazyLock::new(|| {
        let mut m = std::collections::HashMap::new();
        m.insert("lower".to_string(), datafusion_functions::string::lower());
        m.insert("upper".to_string(), datafusion_functions::string::upper());
-        m.insert("contains".to_string(), datafusion_functions::string::contains());
+        m.insert(
+            "contains".to_string(),
+            datafusion_functions::string::contains(),
+        );
        m.insert("btrim".to_string(), datafusion_functions::string::btrim());
        m.insert("ltrim".to_string(), datafusion_functions::string::ltrim());
        m.insert("rtrim".to_string(), datafusion_functions::string::rtrim());
        m.insert("concat".to_string(), datafusion_functions::string::concat());
-        m.insert("octet_length".to_string(), datafusion_functions::string::octet_length());
-        std::sync::RwLock::new(m)
-    };
-}
+        m.insert(
+            "octet_length".to_string(),
+            datafusion_functions::string::octet_length(),
+        );
+        m
+    });

 pub fn func(name: impl AsRef<str>, args: Vec<Expr>) -> crate::Result<Expr> {
    let name = name.as_ref();
-    let registry = FUNC_REGISTRY
-        .read()
-        .map_err(|e| crate::Error::InvalidInput {
-            message: format!("lock poisoned: {}", e),
-        })?;
-    let udf = registry
+    let udf = FUNC_REGISTRY
        .get(name)
        .ok_or_else(|| crate::Error::InvalidInput {
            message: format!("unknown function: {}", name),
@@ -54,7 +54,26 @@ pub enum Index {
    /// substrings of the raw bytes, unlike the tokenized [`Index::FTS`] index.
    Fm(FmIndexBuilder),

-    /// Full text search index using bm25.
+    /// Full text search index using BM25.
+    ///
+    /// The posting block size defaults to 128. Supported values are 128 and 256;
+    /// a value of 256 uses the experimental FTS V3 format and may introduce
+    /// breaking changes.
+    ///
+    /// ```
+    /// use lancedb::index::{Index, scalar::FtsIndexBuilder};
+    ///
+    /// # async fn create_fts_index(
+    /// #     table: &lancedb::Table,
+    /// # ) -> Result<(), Box<dyn std::error::Error>> {
+    /// let params = FtsIndexBuilder::default().block_size(256)?;
+    /// table
+    ///     .create_index(&["text"], Index::FTS(params))
+    ///     .execute()
+    ///     .await?;
+    /// # Ok(())
+    /// # }
+    /// ```
    FTS(FtsIndexBuilder),

    /// IVF index
@@ -47,6 +47,20 @@ pub trait HeaderProvider: Send + Sync + std::fmt::Debug {
    async fn get_headers(&self) -> Result<HashMap<String, String>>;
 }

+/// Default maximum bytes per insert request (8 GiB).
+///
+/// Sized so a multipart part can hold at least one full Lance data file (the
+/// default is 1M rows / 90 GB per file), which keeps fragments from being split
+/// into undersized files across parts. The time-based cut
+/// ([`DEFAULT_MAX_REQUEST_DURATION_DIVISOR`]) bounds request duration on slow
+/// uploads, so a large byte budget does not risk the read timeout.
+const DEFAULT_MAX_BYTES_PER_REQUEST: u64 = 8 * 1024 * 1024 * 1024;
+
+/// The default max request duration is the read timeout divided by this, leaving
+/// headroom for the server to finalize and acknowledge a part before the read
+/// timeout (which also covers the request-body upload) fires.
+const DEFAULT_MAX_REQUEST_DURATION_DIVISOR: u32 = 2;
+
 /// Configuration for the LanceDB Cloud HTTP client.
 #[derive(Clone)]
 pub struct ClientConfig {
@@ -71,6 +85,33 @@ pub struct ClientConfig {
    /// Alternatively, set `LANCEDB_USER_ID_ENV_KEY` to specify another environment
    /// variable that contains the user ID value.
    pub user_id: Option<String>,
+    /// Maximum number of bytes to send in a single insert HTTP request.
+    ///
+    /// During a multipart write, each partition's data is split into one or more
+    /// parts of at most this many (Arrow IPC, compressed) bytes, each uploaded as
+    /// a separate request under the shared upload id. This bounds how long any
+    /// one request stays open, so large bulk ingests do not exceed the client
+    /// read timeout while the server streams the part to object storage.
+    ///
+    /// The request body is still streamed (not buffered), so this does not
+    /// increase peak memory. Set to `Some(0)` to disable splitting (one request
+    /// per partition). You can also set the `LANCE_CLIENT_MAX_BYTES_PER_REQUEST`
+    /// environment variable. Defaults to 8 GiB.
+    pub max_bytes_per_request: Option<u64>,
+    /// Maximum wall-clock time to spend uploading a single insert HTTP request.
+    ///
+    /// Complements [`Self::max_bytes_per_request`]: during a multipart write a
+    /// part is cut when it reaches either the byte budget or this duration,
+    /// whichever comes first. The client read timeout also covers the
+    /// request-body upload, so a slow or throttled upload of a large part can
+    /// hit that timeout before the byte budget is reached; cutting by time keeps
+    /// each request short enough that it completes (and the server acknowledges
+    /// the part) within the read timeout.
+    ///
+    /// Set to `Some(Duration::ZERO)` to disable the time-based cut. You can also
+    /// set the `LANCE_CLIENT_MAX_REQUEST_DURATION` environment variable (integer
+    /// seconds). Defaults to half the resolved read timeout.
+    pub max_request_duration: Option<Duration>,
 }

 impl std::fmt::Debug for ClientConfig {
@@ -87,6 +128,8 @@ impl std::fmt::Debug for ClientConfig {
                &self.header_provider.as_ref().map(|_| "Some(...)"),
            )
            .field("user_id", &self.user_id)
+            .field("max_bytes_per_request", &self.max_bytes_per_request)
+            .field("max_request_duration", &self.max_request_duration)
            .finish()
    }
 }
@@ -102,6 +145,8 @@ impl Default for ClientConfig {
            tls_config: None,
            header_provider: None,
            user_id: None,
+            max_bytes_per_request: None,
+            max_request_duration: None,
        }
    }
 }
@@ -248,6 +293,16 @@ pub struct RestfulLanceDbClient<S: HttpSend = Sender> {
    /// Connection-level read consistency interval. Drives the
    /// `x-lancedb-min-timestamp` freshness header sent on read requests.
    pub(crate) read_consistency_interval: Option<Duration>,
+    // Note the `Option` here means the opposite of the same-named
+    // `ClientConfig` fields: those are pre-resolution, where `None` means "fall
+    // back to env var / default". These are post-resolution (see
+    // `resolve_max_bytes_per_request` / `resolve_max_request_duration`), where a
+    // default has already been applied and `None` means the feature is disabled.
+    /// Maximum bytes per insert request. `None` disables request splitting.
+    pub(crate) max_bytes_per_request: Option<u64>,
+    /// Maximum wall-clock time per insert request. `None` disables the
+    /// time-based part cut.
+    pub(crate) max_request_duration: Option<Duration>,
 }

 impl<S: HttpSend> std::fmt::Debug for RestfulLanceDbClient<S> {
@@ -429,6 +484,10 @@ impl RestfulLanceDbClient<Sender> {
        };
        debug!("Created client for host: {}", host);
        let retry_config = client_config.retry_config.clone().try_into()?;
+        let max_bytes_per_request =
+            Self::resolve_max_bytes_per_request(client_config.max_bytes_per_request)?;
+        let max_request_duration =
+            Self::resolve_max_request_duration(client_config.max_request_duration, read_timeout)?;
        Ok(Self {
            client,
            host,
@@ -440,8 +499,52 @@ impl RestfulLanceDbClient<Sender> {
                .unwrap_or("$".to_string()),
            header_provider: client_config.header_provider,
            read_consistency_interval,
+            max_bytes_per_request,
+            max_request_duration,
        })
    }
+
+    /// Resolve the max bytes per insert request from config, environment, or the
+    /// default. A value of `0` (from either source) disables request splitting.
+    fn resolve_max_bytes_per_request(passed: Option<u64>) -> Result<Option<u64>> {
+        let value = if let Some(value) = passed {
+            value
+        } else if let Ok(env) = std::env::var("LANCE_CLIENT_MAX_BYTES_PER_REQUEST") {
+            env.parse::<u64>().map_err(|_| Error::InvalidInput {
+                message: format!(
+                    "LANCE_CLIENT_MAX_BYTES_PER_REQUEST must be a non-negative integer, got '{}'",
+                    env
+                ),
+            })?
+        } else {
+            DEFAULT_MAX_BYTES_PER_REQUEST
+        };
+        Ok((value > 0).then_some(value))
+    }
+
+    /// Resolve the max request duration from config, environment, or a default
+    /// derived from the read timeout. A zero duration (from either source)
+    /// disables the time-based cut.
+    fn resolve_max_request_duration(
+        passed: Option<Duration>,
+        read_timeout: Duration,
+    ) -> Result<Option<Duration>> {
+        let value = if let Some(value) = passed {
+            value
+        } else if let Ok(env) = std::env::var("LANCE_CLIENT_MAX_REQUEST_DURATION") {
+            let secs = env.parse::<u64>().map_err(|_| Error::InvalidInput {
+                message: format!(
+                    "LANCE_CLIENT_MAX_REQUEST_DURATION must be a non-negative integer \
+                     number of seconds, got '{}'",
+                    env
+                ),
+            })?;
+            Duration::from_secs(secs)
+        } else {
+            read_timeout / DEFAULT_MAX_REQUEST_DURATION_DIVISOR
+        };
+        Ok((!value.is_zero()).then_some(value))
+    }
 }

 impl<S: HttpSend> RestfulLanceDbClient<S> {
@@ -449,6 +552,18 @@ impl<S: HttpSend> RestfulLanceDbClient<S> {
        &self.host
    }

+    /// Maximum bytes per insert request, or `None` if request splitting is
+    /// disabled.
+    pub(crate) fn max_bytes_per_request(&self) -> Option<u64> {
+        self.max_bytes_per_request
+    }
+
+    /// Maximum wall-clock time per insert request, or `None` if the time-based
+    /// cut is disabled.
+    pub(crate) fn max_request_duration(&self) -> Option<Duration> {
+        self.max_request_duration
+    }
+
    pub fn default_headers(
        api_key: &str,
        region: &str,
@@ -875,6 +990,8 @@ pub mod test_utils {
            id_delimiter: "$".to_string(),
            header_provider: None,
            read_consistency_interval,
+            max_bytes_per_request: None,
+            max_request_duration: None,
        }
    }

@@ -900,6 +1017,12 @@ pub mod test_utils {
            id_delimiter: config.id_delimiter.unwrap_or_else(|| "$".to_string()),
            header_provider: config.header_provider,
            read_consistency_interval: None,
+            max_bytes_per_request: config
+                .max_bytes_per_request
+                .and_then(|v| (v > 0).then_some(v)),
+            max_request_duration: config
+                .max_request_duration
+                .and_then(|v| (!v.is_zero()).then_some(v)),
        }
    }
 }
@@ -1103,6 +1226,8 @@ mod tests {
            id_delimiter: "+".to_string(),
            header_provider: Some(Arc::new(provider) as Arc<dyn HeaderProvider>),
            read_consistency_interval: None,
+            max_bytes_per_request: None,
+            max_request_duration: None,
        };

        // Apply dynamic headers
@@ -1139,6 +1264,8 @@ mod tests {
            id_delimiter: "+".to_string(),
            header_provider: Some(Arc::new(provider) as Arc<dyn HeaderProvider>),
            read_consistency_interval: None,
+            max_bytes_per_request: None,
+            max_request_duration: None,
        };

        // Apply dynamic headers
@@ -1177,6 +1304,8 @@ mod tests {
            id_delimiter: "+".to_string(),
            header_provider: Some(Arc::new(provider) as Arc<dyn HeaderProvider>),
            read_consistency_interval: None,
+            max_bytes_per_request: None,
+            max_request_duration: None,
        };

        // Header provider errors should fail the request
@@ -1288,4 +1417,193 @@ mod tests {
            std::env::remove_var("LANCEDB_USER_ID");
        }
    }
+
+    #[test]
+    fn test_resolve_max_bytes_passed_value_wins() {
+        // An explicit config value is used verbatim; env/default are not consulted.
+        let resolved =
+            RestfulLanceDbClient::<Sender>::resolve_max_bytes_per_request(Some(1234)).unwrap();
+        assert_eq!(resolved, Some(1234));
+    }
+
+    #[test]
+    fn test_resolve_max_bytes_zero_disables() {
+        let resolved =
+            RestfulLanceDbClient::<Sender>::resolve_max_bytes_per_request(Some(0)).unwrap();
+        assert_eq!(resolved, None);
+    }
+
+    #[test]
+    #[serial(request_limits_env)]
+    fn test_resolve_max_bytes_default_when_unset() {
+        let _guard = lock_env();
+        // SAFETY: This is only called in tests
+        unsafe {
+            std::env::remove_var("LANCE_CLIENT_MAX_BYTES_PER_REQUEST");
+        }
+        let resolved = RestfulLanceDbClient::<Sender>::resolve_max_bytes_per_request(None).unwrap();
+        assert_eq!(resolved, Some(DEFAULT_MAX_BYTES_PER_REQUEST));
+    }
+
+    #[test]
+    #[serial(request_limits_env)]
+    fn test_resolve_max_bytes_from_env() {
+        let _guard = lock_env();
+        // SAFETY: This is only called in tests
+        unsafe {
+            std::env::set_var("LANCE_CLIENT_MAX_BYTES_PER_REQUEST", "4096");
+        }
+        let resolved = RestfulLanceDbClient::<Sender>::resolve_max_bytes_per_request(None).unwrap();
+        // SAFETY: This is only called in tests
+        unsafe {
+            std::env::remove_var("LANCE_CLIENT_MAX_BYTES_PER_REQUEST");
+        }
+        assert_eq!(resolved, Some(4096));
+    }
+
+    #[test]
+    #[serial(request_limits_env)]
+    fn test_resolve_max_bytes_env_zero_disables() {
+        let _guard = lock_env();
+        // SAFETY: This is only called in tests
+        unsafe {
+            std::env::set_var("LANCE_CLIENT_MAX_BYTES_PER_REQUEST", "0");
+        }
+        let resolved = RestfulLanceDbClient::<Sender>::resolve_max_bytes_per_request(None).unwrap();
+        // SAFETY: This is only called in tests
+        unsafe {
+            std::env::remove_var("LANCE_CLIENT_MAX_BYTES_PER_REQUEST");
+        }
+        assert_eq!(resolved, None);
+    }
+
+    #[test]
+    #[serial(request_limits_env)]
+    fn test_resolve_max_bytes_config_overrides_env() {
+        let _guard = lock_env();
+        // SAFETY: This is only called in tests
+        unsafe {
+            std::env::set_var("LANCE_CLIENT_MAX_BYTES_PER_REQUEST", "4096");
+        }
+        // A config value takes precedence over the environment variable.
+        let resolved =
+            RestfulLanceDbClient::<Sender>::resolve_max_bytes_per_request(Some(1234)).unwrap();
+        // SAFETY: This is only called in tests
+        unsafe {
+            std::env::remove_var("LANCE_CLIENT_MAX_BYTES_PER_REQUEST");
+        }
+        assert_eq!(resolved, Some(1234));
+    }
+
+    #[test]
+    #[serial(request_limits_env)]
+    fn test_resolve_max_bytes_invalid_env_errors() {
+        let _guard = lock_env();
+        // SAFETY: This is only called in tests
+        unsafe {
+            std::env::set_var("LANCE_CLIENT_MAX_BYTES_PER_REQUEST", "not-a-number");
+        }
+        let err = RestfulLanceDbClient::<Sender>::resolve_max_bytes_per_request(None).unwrap_err();
+        // SAFETY: This is only called in tests
+        unsafe {
+            std::env::remove_var("LANCE_CLIENT_MAX_BYTES_PER_REQUEST");
+        }
+        assert!(matches!(err, Error::InvalidInput { .. }), "got: {err:?}");
+    }
+
+    #[test]
+    fn test_resolve_max_request_duration_passed_value_wins() {
+        let resolved = RestfulLanceDbClient::<Sender>::resolve_max_request_duration(
+            Some(Duration::from_secs(42)),
+            Duration::from_secs(300),
+        )
+        .unwrap();
+        assert_eq!(resolved, Some(Duration::from_secs(42)));
+    }
+
+    #[test]
+    fn test_resolve_max_request_duration_zero_disables() {
+        let resolved = RestfulLanceDbClient::<Sender>::resolve_max_request_duration(
+            Some(Duration::ZERO),
+            Duration::from_secs(300),
+        )
+        .unwrap();
+        assert_eq!(resolved, None);
+    }
+
+    #[test]
+    #[serial(request_limits_env)]
+    fn test_resolve_max_request_duration_default_is_half_read_timeout() {
+        let _guard = lock_env();
+        // SAFETY: This is only called in tests
+        unsafe {
+            std::env::remove_var("LANCE_CLIENT_MAX_REQUEST_DURATION");
+        }
+        let resolved = RestfulLanceDbClient::<Sender>::resolve_max_request_duration(
+            None,
+            Duration::from_secs(300),
+        )
+        .unwrap();
+        assert_eq!(resolved, Some(Duration::from_secs(150)));
+    }
+
+    #[test]
+    #[serial(request_limits_env)]
+    fn test_resolve_max_request_duration_from_env_seconds() {
+        let _guard = lock_env();
+        // SAFETY: This is only called in tests
+        unsafe {
+            std::env::set_var("LANCE_CLIENT_MAX_REQUEST_DURATION", "30");
+        }
+        let resolved = RestfulLanceDbClient::<Sender>::resolve_max_request_duration(
+            None,
+            Duration::from_secs(300),
+        )
+        .unwrap();
+        // SAFETY: This is only called in tests
+        unsafe {
+            std::env::remove_var("LANCE_CLIENT_MAX_REQUEST_DURATION");
+        }
+        assert_eq!(resolved, Some(Duration::from_secs(30)));
+    }
+
+    #[test]
+    #[serial(request_limits_env)]
+    fn test_resolve_max_request_duration_env_zero_disables() {
+        let _guard = lock_env();
+        // SAFETY: This is only called in tests
+        unsafe {
+            std::env::set_var("LANCE_CLIENT_MAX_REQUEST_DURATION", "0");
+        }
+        let resolved = RestfulLanceDbClient::<Sender>::resolve_max_request_duration(
+            None,
+            Duration::from_secs(300),
+        )
+        .unwrap();
+        // SAFETY: This is only called in tests
+        unsafe {
+            std::env::remove_var("LANCE_CLIENT_MAX_REQUEST_DURATION");
+        }
+        assert_eq!(resolved, None);
+    }
+
+    #[test]
+    #[serial(request_limits_env)]
+    fn test_resolve_max_request_duration_invalid_env_errors() {
+        let _guard = lock_env();
+        // SAFETY: This is only called in tests
+        unsafe {
+            std::env::set_var("LANCE_CLIENT_MAX_REQUEST_DURATION", "12.5");
+        }
+        let err = RestfulLanceDbClient::<Sender>::resolve_max_request_duration(
+            None,
+            Duration::from_secs(300),
+        )
+        .unwrap_err();
+        // SAFETY: This is only called in tests
+        unsafe {
+            std::env::remove_var("LANCE_CLIENT_MAX_REQUEST_DURATION");
+        }
+        assert!(matches!(err, Error::InvalidInput { .. }), "got: {err:?}");
+    }
 }
@@ -18,9 +18,11 @@ use crate::index::waiter::wait_for_index;
 use crate::query::{QueryFilter, QueryRequest, Select, VectorQueryRequest};
 use crate::table::AddColumnsResult;
 use crate::table::AddResult;
+use crate::table::BranchDiff;
 use crate::table::DeleteResult;
 use crate::table::DropColumnsResult;
 use crate::table::LsmWriteSpec;
+use crate::table::MergeBranchResult;
 use crate::table::MergeResult;
 use crate::table::Tags;
 use crate::table::UpdateResult;
@@ -1361,6 +1363,8 @@ impl<S: HttpSend + 'static> RemoteTable<S> {
            upload_id.to_string(),
            output.tracker.clone(),
            self.branch.clone(),
+            self.client.max_bytes_per_request(),
+            self.client.max_request_duration(),
        ));

        let task_ctx = Arc::new(datafusion_execution::TaskContext::default());
@@ -1815,6 +1819,79 @@ impl<S: HttpSend> BaseTable for RemoteTable<S> {
        Ok(())
    }

+    async fn diff_branch(&self, from_branch: &str) -> Result<BranchDiff> {
+        if from_branch.trim().is_empty() {
+            return Err(Error::InvalidInput {
+                message: "from_branch must be a non-empty string".into(),
+            });
+        }
+        let request = self
+            .client
+            .post(&format!("/v1/table/{}/branches/diff/", self.identifier))
+            .json(&serde_json::json!({ "from_branch": from_branch }));
+        let (request_id, response) = self.send(request, true).await?;
+        if response.status() == StatusCode::NOT_FOUND {
+            return Err(Error::TableNotFound {
+                name: format!("{} (branch: {})", self.name, from_branch),
+                source: format!("branch '{}' does not exist", from_branch).into(),
+            });
+        }
+        let response = self.check_table_response(&request_id, response).await?;
+        let body = response.text().await.err_to_http(request_id.clone())?;
+        serde_json::from_str(&body).map_err(|err| Error::Http {
+            source: format!(
+                "Failed to parse diff_branch response: {}, body: {}",
+                err, body
+            )
+            .into(),
+            request_id,
+            status_code: None,
+        })
+    }
+
+    async fn merge_branch(&self, from_branch: &str, dry_run: bool) -> Result<MergeBranchResult> {
+        if from_branch.trim().is_empty() {
+            return Err(Error::InvalidInput {
+                message: "from_branch must be a non-empty string".into(),
+            });
+        }
+        let request = self
+            .client
+            .post(&format!("/v1/table/{}/branches/merge/", self.identifier))
+            .json(&serde_json::json!({
+                "from_branch": from_branch,
+                "dry_run": dry_run,
+            }));
+        // No retry. 409 rejected merge is final and carries a body.
+        let (request_id, response) = self.send(request, false).await?;
+        let status = response.status();
+        if status == StatusCode::NOT_FOUND {
+            return Err(Error::TableNotFound {
+                name: format!("{} (branch: {})", self.name, from_branch),
+                source: format!("branch '{}' does not exist", from_branch).into(),
+            });
+        }
+        // 200 and 409 both carry MergeBranchResult.
+        if status != StatusCode::OK && status != StatusCode::CONFLICT {
+            let body = response.text().await.unwrap_or_default();
+            return Err(Error::Http {
+                source: format!("unexpected status {status} from merge_branch: {body}").into(),
+                request_id,
+                status_code: Some(status),
+            });
+        }
+        let body = response.text().await.err_to_http(request_id.clone())?;
+        serde_json::from_str(&body).map_err(|err| Error::Http {
+            source: format!(
+                "Failed to parse merge_branch response: {}, body: {}",
+                err, body
+            )
+            .into(),
+            request_id,
+            status_code: Some(status),
+        })
+    }
+
    fn current_branch(&self) -> Option<String> {
        self.branch.clone()
    }
@@ -1863,25 +1940,33 @@ impl<S: HttpSend> BaseTable for RemoteTable<S> {
        let table_schema = self.schema().await?;
        let table_def = TableDefinition::try_from_rich_schema(table_schema.clone())?;

-        let num_partitions = if let Some(parallelism) = add.write_parallelism {
-            if parallelism > 1 && self.server_version.support_multipart_write() {
-                parallelism
-            } else {
-                1
-            }
-        } else if self.server_version.support_multipart_write() {
-            // Peek at the first batch to estimate write partitions, same as NativeTable.
+        let num_partitions = if self.server_version.support_multipart_write() {
+            // Peek at the first batch to estimate write partitions (same as
+            // NativeTable) and, regardless of `write_parallelism`, to detect a
+            // fully empty input. A multipart write creates its upload session
+            // before any partition executes; if the input turns out to have no
+            // batches at all, no partition ever stages a part (see
+            // `send_multipart_chunked`), so completing the write has nothing to
+            // commit and e.g. `mode=overwrite` would be silently dropped. Route
+            // empty input through the single-request path instead, which always
+            // sends one schema-only request.
            let mut peeked = PeekedScannable::new(add.data);
-            let n = if let Some(first_batch) = peeked.peek().await {
-                let max_partitions = lance_core::utils::tokio::get_num_compute_intensive_cpus();
-                estimate_write_partitions(
-                    first_batch.get_array_memory_size(),
-                    first_batch.num_rows(),
-                    peeked.num_rows(),
-                    max_partitions,
-                )
-            } else {
-                1
+            let n = match peeked.peek().await {
+                Some(first_batch) => match add.write_parallelism {
+                    Some(parallelism) if parallelism > 1 => parallelism,
+                    Some(_) => 1,
+                    None => {
+                        let max_partitions =
+                            lance_core::utils::tokio::get_num_compute_intensive_cpus();
+                        estimate_write_partitions(
+                            first_batch.get_array_memory_size(),
+                            first_batch.num_rows(),
+                            peeked.num_rows(),
+                            max_partitions,
+                        )
+                    }
+                },
+                None => 1,
            };
            add.data = Box::new(peeked);
            n
@@ -4411,6 +4496,15 @@ mod tests {
                serde_json::to_value(InvertedIndexParams::default()).unwrap(),
                Index::FTS(Default::default()),
            ),
+            (
+                "FTS",
+                {
+                    let mut body = serde_json::to_value(InvertedIndexParams::default()).unwrap();
+                    body["block_size"] = 256.into();
+                    body
+                },
+                Index::FTS(InvertedIndexParams::default().block_size(256).unwrap()),
+            ),
        ];

        for (index_type, expected_body, index) in cases {
@@ -7166,6 +7260,76 @@ mod tests {
        assert_eq!(insert_count.load(Ordering::SeqCst), 1);
    }

+    #[tokio::test]
+    async fn test_multipart_write_empty_overwrite_uses_single_partition() {
+        // A multipart write creates its upload session before any partition
+        // executes. If the input has no batches at all, every partition would
+        // stage nothing (see `send_multipart_chunked`), so completing the write
+        // would have nothing to commit and `mode=overwrite` would be silently
+        // dropped. An explicit `write_parallelism` must not force the multipart
+        // path for empty input; it should fall back to the single-request path,
+        // which always sends one schema-only request and carries `mode=overwrite`.
+        let insert_count = Arc::new(AtomicUsize::new(0));
+        let multipart_count = Arc::new(AtomicUsize::new(0));
+
+        let insert_count_c = insert_count.clone();
+        let multipart_count_c = multipart_count.clone();
+
+        let table = Table::new_with_handler_version(
+            "my_table",
+            semver::Version::new(0, 4, 0),
+            move |request| {
+                let path = request.url().path();
+
+                if path == "/v1/table/my_table/describe/" {
+                    return simple_describe_response();
+                }
+
+                if path.contains("multipart_write") {
+                    multipart_count_c.fetch_add(1, Ordering::SeqCst);
+                    panic!("Should not use multipart write endpoints for empty input");
+                }
+
+                if path == "/v1/table/my_table/insert/" {
+                    let query = request.url().query().unwrap_or("");
+                    assert!(
+                        !query.contains("upload_id"),
+                        "Should not have upload_id for empty input"
+                    );
+                    assert!(
+                        query.contains("mode=overwrite"),
+                        "Should carry mode=overwrite, got query: {}",
+                        query
+                    );
+                    insert_count_c.fetch_add(1, Ordering::SeqCst);
+                    return http::Response::builder()
+                        .status(200)
+                        .body(r#"{"version": 2}"#.to_string())
+                        .unwrap();
+                }
+
+                panic!("Unexpected request path: {}", path);
+            },
+        );
+
+        let schema = Arc::new(Schema::new(vec![Field::new("id", DataType::Int32, true)]));
+        let empty_batches: Vec<std::result::Result<RecordBatch, arrow_schema::ArrowError>> =
+            Vec::new();
+        let data: Box<dyn RecordBatchReader + Send> =
+            Box::new(RecordBatchIterator::new(empty_batches, schema));
+        let result = table
+            .add(data)
+            .mode(AddDataMode::Overwrite)
+            .write_parallelism(4)
+            .execute()
+            .await
+            .unwrap();
+
+        assert_eq!(result.version, 2);
+        assert_eq!(multipart_count.load(Ordering::SeqCst), 0);
+        assert_eq!(insert_count.load(Ordering::SeqCst), 1);
+    }
+
    #[tokio::test]
    async fn test_multipart_write_abort_on_insert_failure() {
        let create_count = Arc::new(AtomicUsize::new(0));
@@ -8131,6 +8295,147 @@ mod tests {
        assert!(matches!(err, Error::TableNotFound { .. }), "got {err:?}");
    }

+    fn sample_branch_diff_json() -> &'static str {
+        r#"{
+            "fromBranch":"exp",
+            "parentVersion":1,
+            "mainVersion":1,
+            "branchVersion":2,
+            "baseMoved":false,
+            "rowCountMain":3,
+            "rowCountBranch":3,
+            "rowSummary":{
+                "unchanged":3,
+                "newOnBase":0,
+                "newOnBranch":0,
+                "staleRecompute":0,
+                "inputsChanged":0,
+                "deltaAvailable":false
+            },
+            "addedColumns":[{"name":"tag","dataType":"utf8","nullable":true}],
+            "removedColumns":[],
+            "changedColumns":[],
+            "addedIndexes":[],
+            "removedIndexes":[],
+            "mergeable":true,
+            "mergeBlockers":[]
+        }"#
+    }
+
+    #[tokio::test]
+    async fn test_diff_branch() {
+        let table = Table::new_with_handler("my_table", |request| {
+            assert_eq!(request.method(), "POST");
+            assert_eq!(request.url().path(), "/v1/table/my_table/branches/diff/");
+            let body = request_body_json(&request);
+            assert_eq!(body["from_branch"], "exp");
+            http::Response::builder()
+                .status(200)
+                .body(sample_branch_diff_json())
+                .unwrap()
+        });
+        let diff = table.diff_branch("exp").await.unwrap();
+        assert_eq!(diff.from_branch, "exp");
+        assert!(diff.mergeable);
+        assert_eq!(diff.added_columns.len(), 1);
+        assert_eq!(diff.added_columns[0].name, "tag");
+    }
+
+    #[tokio::test]
+    async fn test_merge_branch_dry_run() {
+        let table = Table::new_with_handler("my_table", |request| {
+            assert_eq!(request.url().path(), "/v1/table/my_table/branches/merge/");
+            let body = request_body_json(&request);
+            assert_eq!(body["from_branch"], "exp");
+            assert_eq!(body["dry_run"], true);
+            let resp = format!(
+                r#"{{"status":"ready","diff":{},"preview":{{"promotedColumns":["tag"]}}}}"#,
+                sample_branch_diff_json()
+            );
+            http::Response::builder().status(200).body(resp).unwrap()
+        });
+        let result = table.merge_branch("exp", true).await.unwrap();
+        assert_eq!(result.status, crate::table::MergeBranchStatus::Ready);
+        assert_eq!(result.preview.promoted_columns, vec!["tag".to_string()]);
+        assert!(result.main_version_after.is_none());
+    }
+
+    #[tokio::test]
+    async fn test_merge_branch_rejected_returns_ok_with_body() {
+        let table = Table::new_with_handler("my_table", |request| {
+            assert_eq!(request.url().path(), "/v1/table/my_table/branches/merge/");
+            let body = request_body_json(&request);
+            assert_eq!(body["dry_run"], false);
+            let mut diff: serde_json::Value =
+                serde_json::from_str(sample_branch_diff_json()).unwrap();
+            diff["mergeable"] = serde_json::json!(false);
+            diff["mergeBlockers"] = serde_json::json!([{
+                "code": "baseMoved",
+                "message": "main has advanced"
+            }]);
+            let resp = serde_json::json!({
+                "status": "rejected",
+                "diff": diff,
+                "preview": { "promotedColumns": [] }
+            });
+            http::Response::builder()
+                .status(409)
+                .body(resp.to_string())
+                .unwrap()
+        });
+        let result = table.merge_branch("exp", false).await.unwrap();
+        assert_eq!(result.status, crate::table::MergeBranchStatus::Rejected);
+        assert!(!result.diff.mergeable);
+        assert_eq!(result.diff.merge_blockers.len(), 1);
+    }
+
+    #[tokio::test]
+    async fn test_merge_branch_unknown_blocker_code_parses() {
+        let table = Table::new_with_handler("my_table", |_| {
+            let mut diff: serde_json::Value =
+                serde_json::from_str(sample_branch_diff_json()).unwrap();
+            diff["mergeable"] = serde_json::json!(false);
+            diff["mergeBlockers"] = serde_json::json!([{
+                "code": "multipleCommits",
+                "message": "branch has more than one data commit"
+            }]);
+            let resp = serde_json::json!({
+                "status": "rejected",
+                "diff": diff,
+                "preview": { "operation": "append", "rowsAdded": 2 }
+            });
+            http::Response::builder()
+                .status(409)
+                .body(resp.to_string())
+                .unwrap()
+        });
+        let result = table.merge_branch("exp", false).await.unwrap();
+        assert_eq!(result.status, crate::table::MergeBranchStatus::Rejected);
+        assert_eq!(
+            result.diff.merge_blockers[0].code,
+            crate::table::MergeBlockerCode::Unknown
+        );
+        assert!(result.preview.promoted_columns.is_empty());
+    }
+
+    #[tokio::test]
+    async fn test_merge_branch_unexpected_2xx_is_error() {
+        let table = Table::new_with_handler("my_table", |_| {
+            http::Response::builder()
+                .status(204)
+                .body(String::new())
+                .unwrap()
+        });
+        let err = table.merge_branch("exp", false).await.unwrap_err();
+        match err {
+            Error::Http {
+                status_code: Some(code),
+                ..
+            } => assert_eq!(code, reqwest::StatusCode::NO_CONTENT),
+            other => panic!("expected Http error, got {other:?}"),
+        }
+    }
+
    #[tokio::test]
    async fn test_checkout_branch_validates_via_list() {
        let table = Table::new_with_handler("my_table", |request| {
@@ -4,6 +4,7 @@
 //! DataFusion ExecutionPlan for inserting data into remote LanceDB tables.

 use std::sync::{Arc, Mutex};
+use std::time::{Duration, Instant};

 use arrow_array::{ArrayRef, RecordBatch, UInt64Array};
 use arrow_ipc::CompressionType;
@@ -15,7 +16,7 @@ use datafusion_physical_plan::stream::RecordBatchStreamAdapter;
 use datafusion_physical_plan::{
    DisplayAs, DisplayFormatType, ExecutionPlan, ExecutionPlanProperties, PlanProperties,
 };
-use futures::StreamExt;
+use futures::{SinkExt, StreamExt};
 use http::header::CONTENT_TYPE;
 use lance::io::exec::utils::InstrumentedRecordBatchStreamAdapter;

@@ -49,6 +50,15 @@ pub struct RemoteInsertExec<S: HttpSend = Sender> {
    tracker: Option<Arc<WriteProgressTracker>>,
    /// Branch to write to via `?branch=`. `None` targets the main branch.
    branch: Option<String>,
+    /// For multipart writes, split each partition into parts of at most this
+    /// many bytes, each uploaded as a separate request. `None` sends the whole
+    /// partition as a single request.
+    max_bytes_per_request: Option<u64>,
+    /// For multipart writes, also cut a part once it has been uploading for this
+    /// long, even if it has not reached `max_bytes_per_request`. Bounds request
+    /// duration on slow/throttled uploads so no request exceeds the read
+    /// timeout. `None` disables the time-based cut.
+    max_request_duration: Option<Duration>,
 }

 impl<S: HttpSend + 'static> RemoteInsertExec<S> {
@@ -63,7 +73,7 @@ impl<S: HttpSend + 'static> RemoteInsertExec<S> {
        branch: Option<String>,
    ) -> Self {
        Self::new_inner(
-            table_name, identifier, client, input, overwrite, None, tracker, branch,
+            table_name, identifier, client, input, overwrite, None, tracker, branch, None, None,
        )
    }

@@ -82,6 +92,8 @@ impl<S: HttpSend + 'static> RemoteInsertExec<S> {
        upload_id: String,
        tracker: Option<Arc<WriteProgressTracker>>,
        branch: Option<String>,
+        max_bytes_per_request: Option<u64>,
+        max_request_duration: Option<Duration>,
    ) -> Self {
        Self::new_inner(
            table_name,
@@ -92,6 +104,8 @@ impl<S: HttpSend + 'static> RemoteInsertExec<S> {
            Some(upload_id),
            tracker,
            branch,
+            max_bytes_per_request,
+            max_request_duration,
        )
    }

@@ -105,6 +119,8 @@ impl<S: HttpSend + 'static> RemoteInsertExec<S> {
        upload_id: Option<String>,
        tracker: Option<Arc<WriteProgressTracker>>,
        branch: Option<String>,
+        max_bytes_per_request: Option<u64>,
+        max_request_duration: Option<Duration>,
    ) -> Self {
        let num_partitions = if upload_id.is_some() {
            input.output_partitioning().partition_count()
@@ -131,6 +147,8 @@ impl<S: HttpSend + 'static> RemoteInsertExec<S> {
            upload_id,
            tracker,
            branch,
+            max_bytes_per_request,
+            max_request_duration,
        }
    }

@@ -214,6 +232,238 @@ impl<S: HttpSend + 'static> RemoteInsertExec<S> {
    }
 }

+/// Shared context for the requests of a single partition's multipart upload.
+/// These values are identical for every part; only the part id and streamed
+/// body differ between requests. Bundling them keeps the per-part helpers from
+/// each threading the same handful of arguments.
+struct PartRequestCtx<'a, S: HttpSend> {
+    client: &'a RestfulLanceDbClient<S>,
+    identifier: &'a str,
+    table_name: &'a str,
+    upload_id: &'a str,
+    branch: Option<&'a str>,
+    overwrite: bool,
+}
+
+impl<S: HttpSend + 'static> PartRequestCtx<'_, S> {
+    /// Upload a partition as one or more multipart parts, cutting a new part
+    /// whenever the current one reaches `max_bytes` (Arrow IPC, compressed) or
+    /// has been uploading for `max_duration`, whichever comes first.
+    ///
+    /// Each part is a separate `/insert?upload_id=...&upload_part_id=...` request
+    /// whose body is still streamed through a bounded channel, so peak memory
+    /// stays at a couple of batches regardless of `max_bytes`. The server stages
+    /// every part under the shared `upload_id` and merges them atomically when
+    /// the caller completes the multipart write. An empty partition stages
+    /// nothing: the multipart write always has at least one non-empty partition
+    /// to commit.
+    ///
+    /// The byte budget targets a good on-disk fragment size; the duration budget
+    /// bounds request time so a slow or throttled upload does not keep a request
+    /// open past the client read timeout (which also covers the request body).
+    async fn send_multipart_chunked(
+        &self,
+        max_bytes: u64,
+        max_duration: Option<Duration>,
+        mut input: SendableRecordBatchStream,
+        tracker: Option<Arc<WriteProgressTracker>>,
+    ) -> DataFusionResult<()> {
+        let schema = input.schema();
+
+        // A part always starts from a batch we already hold: the first batch of
+        // the partition, or the look-ahead batch from the previous part. This
+        // keeps empty partitions from staging a part and stops a size cut that
+        // lands exactly on the end of input from emitting a trailing empty part.
+        let mut first = match input.next().await {
+            Some(batch) => batch?,
+            None => return Ok(()),
+        };
+
+        loop {
+            let input_ended = self
+                .send_one_part(
+                    &schema,
+                    max_bytes,
+                    max_duration,
+                    first,
+                    &mut input,
+                    &tracker,
+                )
+                .await?;
+
+            if input_ended {
+                break;
+            }
+
+            first = match input.next().await {
+                Some(batch) => batch?,
+                None => break,
+            };
+        }
+
+        Ok(())
+    }
+
+    /// Build the `/insert` request for a single multipart part.
+    fn build_part_request(&self, part_id: &str, body: reqwest::Body) -> reqwest::RequestBuilder {
+        let mut request = self
+            .client
+            .post(&format!("/v1/table/{}/insert/", self.identifier))
+            .header(CONTENT_TYPE, ARROW_STREAM_CONTENT_TYPE)
+            .query(&[("upload_id", self.upload_id)])
+            .query(&[("upload_part_id", part_id)]);
+        // Every part of an overwrite carries `mode=overwrite`. The server records
+        // it against the shared `upload_id` and applies the overwrite once, when
+        // the multipart write is completed, rather than per part.
+        if self.overwrite {
+            request = request.query(&[("mode", "overwrite")]);
+        }
+        if let Some(b) = self.branch {
+            request = request.query(&[("branch", b)]);
+        }
+        request.body(body)
+    }
+
+    /// Send a single part's request and drain the response, mapping HTTP and
+    /// table-not-found errors into `DataFusionError`.
+    async fn send_part_request(&self, request: reqwest::RequestBuilder) -> DataFusionResult<()> {
+        let (request_id, response) = self
+            .client
+            .send(request)
+            .await
+            .map_err(|e| DataFusionError::External(Box::new(e)))?;
+        let response =
+            RemoteTable::<Sender>::handle_table_not_found(self.table_name, response, &request_id)
+                .await
+                .map_err(|e| DataFusionError::External(Box::new(e)))?;
+        let response = self
+            .client
+            .check_response(&request_id, response)
+            .await
+            .map_err(|e| DataFusionError::External(Box::new(e)))?;
+        response.bytes().await.map_err(|e| {
+            DataFusionError::External(Box::new(Error::Http {
+                source: Box::new(e),
+                request_id: request_id.clone(),
+                status_code: None,
+            }))
+        })?;
+        Ok(())
+    }
+
+    /// Stream one part, starting from `first` and pulling from `input` until the
+    /// part reaches `max_bytes`, has been uploading for `max_duration`, or the
+    /// input ends. The body is streamed through a bounded channel concurrently
+    /// with the request, so peak memory stays at a couple of batches. Wire bytes
+    /// are recorded on `tracker` as each chunk is produced, so progress advances
+    /// smoothly rather than jumping once per completed part. Returns whether the
+    /// input was exhausted while filling this part.
+    async fn send_one_part(
+        &self,
+        schema: &arrow_schema::SchemaRef,
+        max_bytes: u64,
+        max_duration: Option<Duration>,
+        first: RecordBatch,
+        input: &mut SendableRecordBatchStream,
+        tracker: &Option<Arc<WriteProgressTracker>>,
+    ) -> DataFusionResult<bool> {
+        let (mut chunk_tx, chunk_rx) =
+            futures::channel::mpsc::channel::<Result<Vec<u8>, std::io::Error>>(2);
+        let body = reqwest::Body::wrap_stream(chunk_rx);
+
+        let part_id = uuid::Uuid::new_v4().to_string();
+        let request = self.build_part_request(&part_id, body);
+
+        // Measured from just before the request is sent, matching the window the
+        // client read timeout applies to the upload.
+        let started = Instant::now();
+        let tracker = tracker.clone();
+        // Unlike `stream_as_http_body`, this producer also cuts the part at the
+        // byte/time budget and reports back whether the input ended, so it drives
+        // its own bounded mpsc channel joined with the request instead of reusing
+        // that helper.
+        let producer = async move {
+            let options = arrow_ipc::writer::IpcWriteOptions::default()
+                .try_with_compression(Some(CompressionType::LZ4_FRAME))
+                .map_err(|e| DataFusionError::External(Box::new(e)))?;
+            let mut writer =
+                arrow_ipc::writer::StreamWriter::try_new_with_options(Vec::new(), schema, options)
+                    .map_err(|e| DataFusionError::External(Box::new(e)))?;
+
+            let mut part_bytes: u64 = 0;
+            let mut input_ended = false;
+            let mut pending = Some(first);
+            loop {
+                let batch = match pending.take() {
+                    Some(batch) => batch,
+                    None => match input.next().await {
+                        Some(Ok(batch)) => batch,
+                        Some(Err(e)) => {
+                            // Abort the body so the server does not treat the
+                            // truncated stream as a successful write; the
+                            // original error is surfaced to the caller.
+                            let _ = chunk_tx
+                                .send(Err(std::io::Error::other("input stream error")))
+                                .await;
+                            return Err(e);
+                        }
+                        None => {
+                            input_ended = true;
+                            break;
+                        }
+                    },
+                };
+                writer
+                    .write(&batch)
+                    .map_err(|e| DataFusionError::External(Box::new(e)))?;
+                let chunk = std::mem::take(writer.get_mut());
+                let chunk_len = chunk.len();
+                part_bytes += chunk_len as u64;
+                if chunk_tx.send(Ok(chunk)).await.is_err() {
+                    // The request finished or failed; stop producing.
+                    break;
+                }
+                if let Some(ref t) = tracker {
+                    t.record_bytes(chunk_len);
+                }
+                if part_bytes >= max_bytes
+                    || max_duration.is_some_and(|limit| started.elapsed() >= limit)
+                {
+                    break;
+                }
+            }
+
+            writer
+                .finish()
+                .map_err(|e| DataFusionError::External(Box::new(e)))?;
+            let tail = std::mem::take(writer.get_mut());
+            if !tail.is_empty() {
+                let tail_len = tail.len();
+                if chunk_tx.send(Ok(tail)).await.is_ok()
+                    && let Some(ref t) = tracker
+                {
+                    t.record_bytes(tail_len);
+                }
+            }
+            Ok::<bool, DataFusionError>(input_ended)
+        };
+
+        let send = self.send_part_request(request);
+
+        // `join!` rather than `tokio::spawn`: the producer borrows `input` (and
+        // `schema`), so it cannot satisfy the `'static` bound a spawned task
+        // needs. Running both futures on this task lets them make progress
+        // concurrently without that constraint.
+        let (producer_result, send_result) = futures::join!(producer, send);
+        // Prefer the producer error (e.g. NaN rejection) over any HTTP error it
+        // induced.
+        let input_ended = producer_result?;
+        send_result?;
+
+        Ok(input_ended)
+    }
+}
+
 impl<S: HttpSend + 'static> DisplayAs for RemoteInsertExec<S> {
    fn fmt_as(&self, t: DisplayFormatType, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        match t {
@@ -278,6 +528,8 @@ impl<S: HttpSend + 'static> ExecutionPlan for RemoteInsertExec<S> {
            self.upload_id.clone(),
            self.tracker.clone(),
            self.branch.clone(),
+            self.max_bytes_per_request,
+            self.max_request_duration,
        )))
    }

@@ -310,8 +562,36 @@ impl<S: HttpSend + 'static> ExecutionPlan for RemoteInsertExec<S> {
        let upload_id = self.upload_id.clone();
        let tracker = self.tracker.clone();
        let branch = self.branch.clone();
+        let max_bytes_per_request = self.max_bytes_per_request;
+        let max_request_duration = self.max_request_duration;

        let stream = futures::stream::once(async move {
+            // Multipart writes with a byte budget split the partition into
+            // several bounded, still-streamed requests so no single request
+            // stays open long enough to hit the client read timeout.
+            if let (Some(upload_id), Some(max_bytes)) =
+                (upload_id.as_deref(), max_bytes_per_request)
+            {
+                let ctx = PartRequestCtx {
+                    client: &client,
+                    identifier: &identifier,
+                    table_name: &table_name,
+                    upload_id,
+                    branch: branch.as_deref(),
+                    overwrite,
+                };
+                ctx.send_multipart_chunked(max_bytes, max_request_duration, input_stream, tracker)
+                    .await?;
+                // Count 0 here as for the non-multipart path below: the parts are
+                // only staged, so the real row count is resolved when the caller
+                // completes the multipart write.
+                let count_array: ArrayRef = Arc::new(UInt64Array::from(vec![0u64]));
+                return Ok::<RecordBatch, DataFusionError>(RecordBatch::try_new(
+                    COUNT_SCHEMA.clone(),
+                    vec![count_array],
+                )?);
+            }
+
            let mut request = client
                .post(&format!("/v1/table/{}/insert/", identifier))
                .header(CONTENT_TYPE, ARROW_STREAM_CONTENT_TYPE);
@@ -423,9 +703,15 @@ mod tests {
    use arrow_schema::{DataType, Field, Schema as ArrowSchema};
    use datafusion::prelude::SessionContext;
    use datafusion_catalog::MemTable;
-    use std::sync::Arc;
+    use datafusion_common::{DataFusionError, Result as DataFusionResult};
+    use datafusion_execution::{SendableRecordBatchStream, TaskContext};
+    use datafusion_physical_expr::EquivalenceProperties;
+    use datafusion_physical_plan::stream::RecordBatchStreamAdapter;
+    use datafusion_physical_plan::{DisplayAs, DisplayFormatType, ExecutionPlan, PlanProperties};
    use std::sync::atomic::{AtomicUsize, Ordering};
+    use std::sync::{Arc, Mutex};

+    use super::RemoteInsertExec;
    use crate::Table;
    use crate::remote::ARROW_STREAM_CONTENT_TYPE;
    use crate::table::datafusion::BaseTableAdapter;
@@ -591,4 +877,489 @@ mod tests {
        // Verify: should have made exactly one HTTP request despite multiple input partitions
        assert_eq!(request_count.load(Ordering::SeqCst), 1);
    }
+
+    /// Build a single-partition input plan from the given batches.
+    async fn input_plan_from_batches(
+        schema: Arc<ArrowSchema>,
+        batches: Vec<arrow_array::RecordBatch>,
+    ) -> Arc<dyn ExecutionPlan> {
+        use datafusion_catalog::TableProvider;
+        let mem = MemTable::try_new(schema, vec![batches]).unwrap();
+        let ctx = SessionContext::new();
+        mem.scan(&ctx.state(), None, &[], None).await.unwrap()
+    }
+
+    /// Build a single-partition input plan from the batches spread across the
+    /// given partitions.
+    async fn input_plan_from_partitions(
+        schema: Arc<ArrowSchema>,
+        partitions: Vec<Vec<arrow_array::RecordBatch>>,
+    ) -> Arc<dyn ExecutionPlan> {
+        use datafusion_catalog::TableProvider;
+        let mem = MemTable::try_new(schema, partitions).unwrap();
+        let ctx = SessionContext::new();
+        mem.scan(&ctx.state(), None, &[], None).await.unwrap()
+    }
+
+    fn counting_insert_client(
+        counter: Arc<AtomicUsize>,
+    ) -> crate::remote::client::RestfulLanceDbClient<crate::remote::client::test_utils::MockSender>
+    {
+        crate::remote::client::test_utils::client_with_handler(move |request| {
+            let path = request.url().path();
+            assert_eq!(path, "/v1/table/my_table/insert/");
+            let query = request.url().query().unwrap_or("");
+            assert!(query.contains("upload_id=upload-1"), "query: {query}");
+            assert!(query.contains("upload_part_id="), "query: {query}");
+            counter.fetch_add(1, Ordering::SeqCst);
+            http::Response::builder()
+                .status(200)
+                .body(String::new())
+                .unwrap()
+        })
+    }
+
+    /// Insert handler that records the `upload_part_id` of every part request so
+    /// a test can assert the ids are distinct.
+    fn recording_insert_client(
+        part_ids: Arc<Mutex<Vec<String>>>,
+    ) -> crate::remote::client::RestfulLanceDbClient<crate::remote::client::test_utils::MockSender>
+    {
+        crate::remote::client::test_utils::client_with_handler(move |request| {
+            assert_eq!(request.url().path(), "/v1/table/my_table/insert/");
+            let part_id = request
+                .url()
+                .query_pairs()
+                .find(|(k, _)| k == "upload_part_id")
+                .map(|(_, v)| v.into_owned())
+                .expect("upload_part_id query param");
+            part_ids.lock().unwrap().push(part_id);
+            http::Response::builder()
+                .status(200)
+                .body(String::new())
+                .unwrap()
+        })
+    }
+
+    /// Single-partition input plan that yields one good batch and then an error,
+    /// for exercising the mid-part input-error abort path in `send_one_part`.
+    #[derive(Debug)]
+    struct ErroringExec {
+        schema: Arc<ArrowSchema>,
+        properties: Arc<PlanProperties>,
+    }
+
+    impl ErroringExec {
+        fn new() -> Self {
+            let schema = record_batch!(("id", Int32, [1, 2])).unwrap().schema();
+            let properties = PlanProperties::new(
+                EquivalenceProperties::new(schema.clone()),
+                datafusion_physical_plan::Partitioning::UnknownPartitioning(1),
+                datafusion_physical_plan::execution_plan::EmissionType::Incremental,
+                datafusion_physical_plan::execution_plan::Boundedness::Bounded,
+            );
+            Self {
+                schema,
+                properties: Arc::new(properties),
+            }
+        }
+    }
+
+    impl DisplayAs for ErroringExec {
+        fn fmt_as(
+            &self,
+            _t: DisplayFormatType,
+            f: &mut std::fmt::Formatter<'_>,
+        ) -> std::fmt::Result {
+            write!(f, "ErroringExec")
+        }
+    }
+
+    impl ExecutionPlan for ErroringExec {
+        fn name(&self) -> &str {
+            "ErroringExec"
+        }
+        fn properties(&self) -> &Arc<PlanProperties> {
+            &self.properties
+        }
+        fn children(&self) -> Vec<&Arc<dyn ExecutionPlan>> {
+            vec![]
+        }
+        fn with_new_children(
+            self: Arc<Self>,
+            _children: Vec<Arc<dyn ExecutionPlan>>,
+        ) -> DataFusionResult<Arc<dyn ExecutionPlan>> {
+            Ok(self)
+        }
+        fn execute(
+            &self,
+            _partition: usize,
+            _context: Arc<TaskContext>,
+        ) -> DataFusionResult<SendableRecordBatchStream> {
+            let batch = record_batch!(("id", Int32, [1, 2])).unwrap();
+            let stream = futures::stream::iter(vec![
+                Ok(batch),
+                Err(DataFusionError::Execution("boom".to_string())),
+            ]);
+            Ok(Box::pin(RecordBatchStreamAdapter::new(
+                self.schema.clone(),
+                stream,
+            )))
+        }
+    }
+
+    #[tokio::test]
+    async fn test_multipart_chunked_splits_into_parts() {
+        use futures::StreamExt;
+
+        let insert_count = Arc::new(AtomicUsize::new(0));
+        let client = counting_insert_client(insert_count.clone());
+
+        let schema = Arc::new(ArrowSchema::new(vec![Field::new(
+            "id",
+            DataType::Int32,
+            true,
+        )]));
+        let batches = vec![
+            record_batch!(("id", Int32, [1, 2])).unwrap(),
+            record_batch!(("id", Int32, [3, 4])).unwrap(),
+            record_batch!(("id", Int32, [5, 6])).unwrap(),
+        ];
+        let input = input_plan_from_batches(schema, batches).await;
+
+        // A 1-byte budget forces every batch into its own part.
+        let exec = RemoteInsertExec::new_multipart(
+            "my_table".to_string(),
+            "my_table".to_string(),
+            client,
+            input,
+            false,
+            "upload-1".to_string(),
+            None,
+            None,
+            Some(1),
+            None,
+        );
+
+        let mut stream = exec.execute(0, Arc::new(TaskContext::default())).unwrap();
+        while stream.next().await.transpose().unwrap().is_some() {}
+
+        assert_eq!(insert_count.load(Ordering::SeqCst), 3);
+    }
+
+    #[tokio::test]
+    async fn test_multipart_single_part_when_under_budget() {
+        use futures::StreamExt;
+
+        let insert_count = Arc::new(AtomicUsize::new(0));
+        let client = counting_insert_client(insert_count.clone());
+
+        let schema = Arc::new(ArrowSchema::new(vec![Field::new(
+            "id",
+            DataType::Int32,
+            true,
+        )]));
+        let batches = vec![
+            record_batch!(("id", Int32, [1, 2])).unwrap(),
+            record_batch!(("id", Int32, [3, 4])).unwrap(),
+            record_batch!(("id", Int32, [5, 6])).unwrap(),
+        ];
+        let input = input_plan_from_batches(schema, batches).await;
+
+        // A large byte budget and no time limit keep the whole partition in a
+        // single part.
+        let exec = RemoteInsertExec::new_multipart(
+            "my_table".to_string(),
+            "my_table".to_string(),
+            client,
+            input,
+            false,
+            "upload-1".to_string(),
+            None,
+            None,
+            Some(64 * 1024 * 1024),
+            None,
+        );
+
+        let mut stream = exec.execute(0, Arc::new(TaskContext::default())).unwrap();
+        while stream.next().await.transpose().unwrap().is_some() {}
+
+        assert_eq!(insert_count.load(Ordering::SeqCst), 1);
+    }
+
+    #[tokio::test]
+    async fn test_multipart_chunked_splits_by_duration() {
+        use futures::StreamExt;
+
+        let insert_count = Arc::new(AtomicUsize::new(0));
+        let client = counting_insert_client(insert_count.clone());
+
+        let schema = Arc::new(ArrowSchema::new(vec![Field::new(
+            "id",
+            DataType::Int32,
+            true,
+        )]));
+        let batches = vec![
+            record_batch!(("id", Int32, [1, 2])).unwrap(),
+            record_batch!(("id", Int32, [3, 4])).unwrap(),
+            record_batch!(("id", Int32, [5, 6])).unwrap(),
+        ];
+        let input = input_plan_from_batches(schema, batches).await;
+
+        // A large byte budget but a tiny duration budget: writing and sending
+        // one batch already takes longer than the limit, so each batch is cut
+        // into its own part on the time check rather than the byte check.
+        let exec = RemoteInsertExec::new_multipart(
+            "my_table".to_string(),
+            "my_table".to_string(),
+            client,
+            input,
+            false,
+            "upload-1".to_string(),
+            None,
+            None,
+            Some(64 * 1024 * 1024),
+            Some(std::time::Duration::from_nanos(1)),
+        );
+
+        let mut stream = exec.execute(0, Arc::new(TaskContext::default())).unwrap();
+        while stream.next().await.transpose().unwrap().is_some() {}
+
+        assert_eq!(insert_count.load(Ordering::SeqCst), 3);
+    }
+
+    #[tokio::test]
+    async fn test_multipart_empty_partition_stages_nothing() {
+        use futures::StreamExt;
+
+        let insert_count = Arc::new(AtomicUsize::new(0));
+        let client = counting_insert_client(insert_count.clone());
+
+        let schema = Arc::new(ArrowSchema::new(vec![Field::new(
+            "id",
+            DataType::Int32,
+            true,
+        )]));
+        // An empty partition should stage no parts; on the multipart path the
+        // write relies on another partition having data to commit.
+        let input = input_plan_from_batches(schema, vec![]).await;
+
+        let exec = RemoteInsertExec::new_multipart(
+            "my_table".to_string(),
+            "my_table".to_string(),
+            client,
+            input,
+            false,
+            "upload-1".to_string(),
+            None,
+            None,
+            Some(64 * 1024 * 1024),
+            None,
+        );
+
+        let mut stream = exec.execute(0, Arc::new(TaskContext::default())).unwrap();
+        while stream.next().await.transpose().unwrap().is_some() {}
+
+        assert_eq!(insert_count.load(Ordering::SeqCst), 0);
+    }
+
+    #[tokio::test]
+    async fn test_multipart_chunked_uses_distinct_part_ids() {
+        use futures::StreamExt;
+        use std::collections::HashSet;
+
+        let part_ids = Arc::new(Mutex::new(Vec::new()));
+        let client = recording_insert_client(part_ids.clone());
+
+        let schema = Arc::new(ArrowSchema::new(vec![Field::new(
+            "id",
+            DataType::Int32,
+            true,
+        )]));
+        let batches = vec![
+            record_batch!(("id", Int32, [1, 2])).unwrap(),
+            record_batch!(("id", Int32, [3, 4])).unwrap(),
+            record_batch!(("id", Int32, [5, 6])).unwrap(),
+        ];
+        let input = input_plan_from_batches(schema, batches).await;
+
+        // A 1-byte budget forces every batch into its own part.
+        let exec = RemoteInsertExec::new_multipart(
+            "my_table".to_string(),
+            "my_table".to_string(),
+            client,
+            input,
+            false,
+            "upload-1".to_string(),
+            None,
+            None,
+            Some(1),
+            None,
+        );
+
+        let mut stream = exec.execute(0, Arc::new(TaskContext::default())).unwrap();
+        while stream.next().await.transpose().unwrap().is_some() {}
+
+        let ids = part_ids.lock().unwrap().clone();
+        assert_eq!(ids.len(), 3, "expected one part id per part: {ids:?}");
+        assert!(
+            ids.iter().all(|id| !id.is_empty()),
+            "part ids must be non-empty: {ids:?}"
+        );
+        let unique: HashSet<&String> = ids.iter().collect();
+        assert_eq!(unique.len(), 3, "part ids must be distinct: {ids:?}");
+    }
+
+    #[tokio::test]
+    async fn test_multipart_chunks_each_partition_independently() {
+        use futures::StreamExt;
+
+        let insert_count = Arc::new(AtomicUsize::new(0));
+        let client = counting_insert_client(insert_count.clone());
+
+        let schema = Arc::new(ArrowSchema::new(vec![Field::new(
+            "id",
+            DataType::Int32,
+            true,
+        )]));
+        let partitions = vec![
+            // Partition 0: two batches, split into two parts by the 1-byte budget.
+            vec![
+                record_batch!(("id", Int32, [1, 2])).unwrap(),
+                record_batch!(("id", Int32, [3, 4])).unwrap(),
+            ],
+            // Partition 1: one batch, one part.
+            vec![record_batch!(("id", Int32, [5, 6])).unwrap()],
+        ];
+        let input = input_plan_from_partitions(schema, partitions).await;
+
+        let exec = RemoteInsertExec::new_multipart(
+            "my_table".to_string(),
+            "my_table".to_string(),
+            client,
+            input,
+            false,
+            "upload-1".to_string(),
+            None,
+            None,
+            Some(1),
+            None,
+        );
+
+        for partition in 0..2 {
+            let mut stream = exec
+                .execute(partition, Arc::new(TaskContext::default()))
+                .unwrap();
+            while stream.next().await.transpose().unwrap().is_some() {}
+        }
+
+        // 2 parts from partition 0 + 1 part from partition 1.
+        assert_eq!(insert_count.load(Ordering::SeqCst), 3);
+    }
+
+    #[tokio::test]
+    async fn test_multipart_input_error_surfaces_original() {
+        use futures::StreamExt;
+
+        let insert_count = Arc::new(AtomicUsize::new(0));
+        let client = counting_insert_client(insert_count.clone());
+
+        // A large byte budget keeps the good batch and the following error in
+        // the same part, exercising the mid-part abort path.
+        let input: Arc<dyn ExecutionPlan> = Arc::new(ErroringExec::new());
+        let exec = RemoteInsertExec::new_multipart(
+            "my_table".to_string(),
+            "my_table".to_string(),
+            client,
+            input,
+            false,
+            "upload-1".to_string(),
+            None,
+            None,
+            Some(64 * 1024 * 1024),
+            None,
+        );
+
+        let mut stream = exec.execute(0, Arc::new(TaskContext::default())).unwrap();
+        let mut err = None;
+        while let Some(item) = stream.next().await {
+            if let Err(e) = item {
+                err = Some(e);
+                break;
+            }
+        }
+
+        let err = err.expect("expected the input stream error to surface");
+        // The original DataFusion error must win over the HTTP error it induces.
+        assert!(
+            err.to_string().contains("boom"),
+            "expected original input error, got: {err}"
+        );
+    }
+
+    #[tokio::test]
+    async fn test_multipart_records_progress_within_a_part() {
+        use crate::table::write_progress::{ProgressCallback, WriteProgress, WriteProgressTracker};
+        use futures::StreamExt;
+
+        let insert_count = Arc::new(AtomicUsize::new(0));
+        let client = counting_insert_client(insert_count.clone());
+
+        let schema = Arc::new(ArrowSchema::new(vec![Field::new(
+            "id",
+            DataType::Int32,
+            true,
+        )]));
+        let batches = vec![
+            record_batch!(("id", Int32, [1, 2])).unwrap(),
+            record_batch!(("id", Int32, [3, 4])).unwrap(),
+            record_batch!(("id", Int32, [5, 6])).unwrap(),
+        ];
+        let input = input_plan_from_batches(schema, batches).await;
+
+        let observed = Arc::new(Mutex::new(Vec::<usize>::new()));
+        let observed_cb = observed.clone();
+        let callback: ProgressCallback = Arc::new(Mutex::new(move |p: &WriteProgress| {
+            observed_cb.lock().unwrap().push(p.output_bytes());
+        }));
+        let tracker = Arc::new(WriteProgressTracker::new(callback, None));
+
+        // A large byte budget keeps all three batches in one part; smooth
+        // progress therefore requires bytes to be reported per chunk rather than
+        // once when the part completes.
+        let exec = RemoteInsertExec::new_multipart(
+            "my_table".to_string(),
+            "my_table".to_string(),
+            client,
+            input,
+            false,
+            "upload-1".to_string(),
+            Some(tracker),
+            None,
+            Some(64 * 1024 * 1024),
+            None,
+        );
+
+        let mut stream = exec.execute(0, Arc::new(TaskContext::default())).unwrap();
+        while stream.next().await.transpose().unwrap().is_some() {}
+
+        assert_eq!(
+            insert_count.load(Ordering::SeqCst),
+            1,
+            "batches should all land in a single part"
+        );
+        let observed = observed.lock().unwrap();
+        assert!(
+            observed.len() > 1,
+            "expected multiple incremental progress updates within the part: {observed:?}"
+        );
+        assert!(
+            observed.windows(2).all(|w| w[1] >= w[0]),
+            "progress bytes should be monotonic: {observed:?}"
+        );
+        assert!(
+            *observed.last().unwrap() > 0,
+            "final progress should report bytes: {observed:?}"
+        );
+    }
 }
@@ -62,6 +62,7 @@ use self::dataset::DatasetConsistencyWrapper;
 use self::merge::MergeInsertBuilder;

 mod add_data;
+pub mod branch_merge;
 mod create_index;
 pub mod datafusion;
 pub(crate) mod dataset;
@@ -77,6 +78,10 @@ use crate::index::waiter::wait_for_index;
 #[cfg(feature = "remote")]
 pub(crate) use add_data::PreprocessingOutput;
 pub use add_data::{AddDataBuilder, AddDataMode, AddResult, NaNVectorBehavior};
+pub use branch_merge::{
+    BranchDiff, ColumnChange, ColumnSummary, IndexSummary, MergeBlocker, MergeBlockerCode,
+    MergeBranchResult, MergeBranchStatus, MergePreview, RowCountSummary,
+};
 pub use chrono::Duration;
 pub use delete::DeleteResult;
 use futures::future::join_all;
@@ -707,6 +712,19 @@ pub trait BaseTable: std::fmt::Display + std::fmt::Debug + Send + Sync {
    async fn list_branches(&self) -> Result<HashMap<String, BranchContents>>;
    /// Delete a branch.
    async fn delete_branch(&self, name: &str) -> Result<()>;
+    /// Diff a branch against main. Remote only.
+    async fn diff_branch(&self, _from_branch: &str) -> Result<BranchDiff> {
+        Err(Error::NotSupported {
+            message: "diff_branch is only supported on remote tables".into(),
+        })
+    }
+    /// Merge a branch into main, or dry-run. Remote only.
+    /// HTTP 409 still returns [`Ok`] with [`MergeBranchStatus::Rejected`].
+    async fn merge_branch(&self, _from_branch: &str, _dry_run: bool) -> Result<MergeBranchResult> {
+        Err(Error::NotSupported {
+            message: "merge_branch is only supported on remote tables".into(),
+        })
+    }
    /// The branch this handle is scoped to, or `None` for `main`.
    fn current_branch(&self) -> Option<String>;
    /// Get the table definition.
@@ -1953,6 +1971,21 @@ impl Table {
        self.inner.delete_branch(name).await
    }

+    /// Diff a branch against main. Remote only.
+    pub async fn diff_branch(&self, from_branch: &str) -> Result<BranchDiff> {
+        self.inner.diff_branch(from_branch).await
+    }
+
+    /// Merge a branch into main, or dry-run. Remote only.
+    /// HTTP 409 still returns [`Ok`] with [`MergeBranchStatus::Rejected`].
+    pub async fn merge_branch(
+        &self,
+        from_branch: &str,
+        dry_run: bool,
+    ) -> Result<MergeBranchResult> {
+        self.inner.merge_branch(from_branch, dry_run).await
+    }
+
    /// The branch this handle is scoped to, or `None` for `main`.
    pub fn current_branch(&self) -> Option<String> {
        self.inner.current_branch()
@@ -0,0 +1,114 @@
+// SPDX-License-Identifier: Apache-2.0
+// SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+//! Types for remote branch diff / merge against main.
+
+use serde::{Deserialize, Serialize};
+
+#[derive(Serialize, Deserialize, Clone, Debug, PartialEq, Eq)]
+#[serde(rename_all = "camelCase")]
+pub struct ColumnSummary {
+    pub name: String,
+    pub data_type: String,
+    pub nullable: bool,
+}
+
+#[derive(Serialize, Deserialize, Clone, Debug, PartialEq, Eq)]
+#[serde(rename_all = "camelCase")]
+pub struct ColumnChange {
+    pub name: String,
+    pub main: ColumnSummary,
+    pub branch: ColumnSummary,
+}
+
+#[derive(Serialize, Deserialize, Clone, Debug, PartialEq, Eq)]
+#[serde(rename_all = "camelCase")]
+pub struct IndexSummary {
+    pub index_name: String,
+    pub columns: Vec<String>,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub index_type: Option<String>,
+    pub status: String,
+}
+
+#[derive(Serialize, Deserialize, Clone, Debug, PartialEq, Eq)]
+#[serde(rename_all = "camelCase")]
+pub struct RowCountSummary {
+    pub unchanged: u64,
+    pub new_on_base: u64,
+    pub new_on_branch: u64,
+    pub stale_recompute: u64,
+    pub inputs_changed: u64,
+    pub delta_available: bool,
+}
+
+#[derive(Serialize, Deserialize, Clone, Debug, PartialEq, Eq)]
+#[serde(rename_all = "camelCase")]
+pub enum MergeBlockerCode {
+    BaseMoved,
+    RowCountMismatch,
+    RowsChanged,
+    ColumnRemoved,
+    ColumnChanged,
+    NoMergeableChanges,
+    NoColumnChanges,
+    InputColumnDependency,
+    ParentNotMain,
+    #[serde(other)]
+    Unknown,
+}
+
+#[derive(Serialize, Deserialize, Clone, Debug, PartialEq, Eq)]
+#[serde(rename_all = "camelCase")]
+pub struct MergeBlocker {
+    pub code: MergeBlockerCode,
+    pub message: String,
+}
+
+#[derive(Serialize, Deserialize, Clone, Debug, PartialEq, Eq)]
+#[serde(rename_all = "camelCase")]
+pub struct BranchDiff {
+    pub from_branch: String,
+    pub parent_version: u64,
+    pub main_version: u64,
+    pub branch_version: u64,
+    pub base_moved: bool,
+    pub row_count_main: u64,
+    pub row_count_branch: u64,
+    pub row_summary: RowCountSummary,
+    pub added_columns: Vec<ColumnSummary>,
+    pub removed_columns: Vec<ColumnSummary>,
+    pub changed_columns: Vec<ColumnChange>,
+    pub added_indexes: Vec<IndexSummary>,
+    pub removed_indexes: Vec<IndexSummary>,
+    pub mergeable: bool,
+    pub merge_blockers: Vec<MergeBlocker>,
+}
+
+#[derive(Serialize, Deserialize, Clone, Debug, PartialEq, Eq)]
+#[serde(rename_all = "camelCase")]
+pub struct MergePreview {
+    #[serde(default)]
+    pub promoted_columns: Vec<String>,
+}
+
+#[derive(Serialize, Deserialize, Clone, Debug, PartialEq, Eq)]
+#[serde(rename_all = "camelCase")]
+pub enum MergeBranchStatus {
+    Ready,
+    Rejected,
+    NotImplemented,
+    Merged,
+    #[serde(other)]
+    Unknown,
+}
+
+#[derive(Serialize, Deserialize, Clone, Debug, PartialEq, Eq)]
+#[serde(rename_all = "camelCase")]
+pub struct MergeBranchResult {
+    pub status: MergeBranchStatus,
+    pub diff: BranchDiff,
+    pub preview: MergePreview,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub main_version_after: Option<u64>,
+}
@@ -382,7 +382,9 @@ mod tests {
    use crate::connect;
    use crate::connection::ConnectBuilder;
    use crate::index::Index;
-    use crate::index::scalar::{BTreeIndexBuilder, BitmapIndexBuilder, FmIndexBuilder};
+    use crate::index::scalar::{
+        BTreeIndexBuilder, BitmapIndexBuilder, FmIndexBuilder, FtsIndexBuilder,
+    };
    use crate::index::vector::{
        IvfHnswFlatIndexBuilder, IvfHnswPqIndexBuilder, IvfHnswSqIndexBuilder,
    };
@@ -1362,7 +1364,10 @@ mod tests {
            .unwrap();

        table
-            .create_index(&["text"], Index::FTS(Default::default()))
+            .create_index(
+                &["text"],
+                Index::FTS(FtsIndexBuilder::default().block_size(256).unwrap()),
+            )
            .execute()
            .await
            .unwrap();
@@ -1372,6 +1377,10 @@ mod tests {
        assert_eq!(index.index_type, crate::index::IndexType::FTS);
        assert_eq!(index.columns, vec!["text".to_string()]);
        assert_eq!(index.name, "text_idx");
+        assert_eq!(index.index_version, Some(3));
+        let index_params: FtsIndexBuilder =
+            serde_json::from_str(index.index_details.as_deref().unwrap()).unwrap();
+        assert_eq!(index_params.posting_block_size(), 256);

        let num_rows = 120;
        let stats = table.index_stats("text_idx").await.unwrap().unwrap();
@@ -774,9 +774,6 @@ fn shard_writer_config_from_defaults(defaults: &HashMap<String, String>) -> Shar
    if let Some(v) = bool_of("durable_write") {
        config = config.with_durable_write(v);
    }
-    if let Some(v) = bool_of("sync_indexed_write") {
-        config = config.with_sync_indexed_write(v);
-    }
    if let Some(v) = usize_of("max_wal_buffer_size") {
        config = config.with_max_wal_buffer_size(v);
    }
@@ -798,12 +795,6 @@ fn shard_writer_config_from_defaults(defaults: &HashMap<String, String>) -> Shar
    if let Some(v) = millis_of("backpressure_log_interval_ms") {
        config = config.with_backpressure_log_interval(v);
    }
-    if let Some(v) = usize_of("async_index_buffer_rows") {
-        config = config.with_async_index_buffer_rows(v);
-    }
-    if let Some(v) = millis_of("async_index_interval_ms") {
-        config = config.with_async_index_interval(v);
-    }
    if let Some(v) = bool_of("enable_memtable") {
        config = config.with_enable_memtable(v);
    }
@@ -1008,13 +999,11 @@ mod tests {
        let defaults = HashMap::from([
            ("durable_write".to_string(), "false".to_string()),
            ("max_memtable_rows".to_string(), "4096".to_string()),
-            ("async_index_interval_ms".to_string(), "250".to_string()),
            ("unknown_key".to_string(), "ignored".to_string()),
        ]);
        let config = shard_writer_config_from_defaults(&defaults);
        assert!(!config.durable_write);
        assert_eq!(config.max_memtable_rows, 4096);
-        assert_eq!(config.async_index_interval, Duration::from_millis(250));
        assert_eq!(config.shard_spec_id, SHARDING_SPEC_ID);
    }

@@ -14,17 +14,15 @@ use lance::arrow::json::JsonDataType;
 use lance::dataset::{ReadParams, WriteParams};
 use lance::index::vector::utils::infer_vector_dim;
 use lance::io::{ObjectStoreParams, WrappingObjectStore};
-use lazy_static::lazy_static;
 use std::pin::Pin;

 use crate::error::{Error, Result};
 use datafusion_physical_plan::SendableRecordBatchStream;

-lazy_static! {
-    static ref TABLE_NAME_REGEX: regex::Regex = regex::Regex::new(r"^[a-zA-Z0-9_\-\.]+$").unwrap();
-    static ref NAMESPACE_NAME_REGEX: regex::Regex =
-        regex::Regex::new(r"^[a-zA-Z0-9_\-\.]+$").unwrap();
-}
+static TABLE_NAME_REGEX: std::sync::LazyLock<regex::Regex> =
+    std::sync::LazyLock::new(|| regex::Regex::new(r"^[a-zA-Z0-9_\-\.]+$").unwrap());
+static NAMESPACE_NAME_REGEX: std::sync::LazyLock<regex::Regex> =
+    std::sync::LazyLock::new(|| regex::Regex::new(r"^[a-zA-Z0-9_\-\.]+$").unwrap());

 pub trait PatchStoreParam {
    fn patch_with_store_wrapper(