Merge remote-tracking branch 'origin/main' into yang/udf-secret-api

.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.38.0-beta.13"
+current_version = "0.38.0-beta.11"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

.github/dependabot.yml

@@ -44,27 +44,3 @@ updates:
       python-deps:
         patterns:
           - "*"
-
-  # The npm ecosystem covers pnpm lockfiles. There are two separate installs:
-  # the bindings themselves and the examples, which have their own lockfile.
-  # As with cargo and pip above, only bump the lockfile — the version ranges
-  # in package.json are our consumers' constraints, not ours.
-  - package-ecosystem: npm
-    directory: /nodejs
-    schedule:
-      interval: weekly
-    versioning-strategy: lockfile-only
-    groups:
-      nodejs-deps:
-        patterns:
-          - "*"
-
-  - package-ecosystem: npm
-    directory: /nodejs/examples
-    schedule:
-      interval: weekly
-    versioning-strategy: lockfile-only
-    groups:
-      nodejs-examples-deps:
-        patterns:
-          - "*"

.github/workflows/dev.yml

@@ -29,14 +29,12 @@ jobs:
     steps:
       - uses: actions/setup-node@v6
         with:
-          node-version: "24"
-      - uses: pnpm/action-setup@v6
-        with:
-          version: 11.1.1
+          node-version: "18"
       # These rules are disabled because Github will always ensure there
       # is a blank line between the title and the body and Github will
       # word wrap the description field to ensure a reasonable max line
       # length.
+      - run: npm install @commitlint/config-conventional
       - run: >
           echo 'module.exports = {
             "rules": {
@@ -45,11 +43,7 @@ jobs:
               "body-leading-blank": [0, "always"]
             }
           }' > .commitlintrc.js
-      - run: >
-          pnpm dlx
-          --package @commitlint/cli@21.2.2
-          --package @commitlint/config-conventional@21.2.2
-          commitlint --extends @commitlint/config-conventional --verbose <<< $COMMIT_MSG
+      - run: npx commitlint --extends @commitlint/config-conventional --verbose <<< $COMMIT_MSG
         env:
           COMMIT_MSG: >
             ${{ github.event.pull_request.title }}
@@ -60,7 +54,7 @@ jobs:
         with:
           script: |
             const message = `**ACTION NEEDED**
-
+              
               Lance follows the [Conventional Commits specification](https://www.conventionalcommits.org/en/v1.0.0/) for release automation.
 
               The PR title and description are used as the merge commit message.\

.github/workflows/docs-link-check.yml

@@ -56,7 +56,7 @@ jobs:
         uses: lycheeverse/lychee-action@e7477775783ea5526144ba13e8db5eec57747ce8 # v2.9.0
         with:
           # Restricted to http(s) on purpose. Much of docs/src is generated
-          # API reference (the js/ tree comes from `pnpm run docs` in nodejs)
+          # API reference (the js/ tree comes from `npm run docs` in nodejs)
           # and the hand-written pages use mkdocstrings cross-references and
           # nav-relative paths that only resolve in the site mkdocs builds,
           # not in this checkout, so relative links would be reported as

.github/workflows/docs.yml

@@ -55,7 +55,9 @@ jobs:
       - name: Set up node
         uses: actions/setup-node@v6
         with:
-          node-version: 24
+          node-version: 20
+          cache: 'npm'
+          cache-dependency-path: docs/package-lock.json
       - name: Install node dependencies
         working-directory: nodejs
         run: |

.github/workflows/nodejs.yml

@@ -47,8 +47,9 @@ jobs:
         version: 11.1.1
     - uses: actions/setup-node@v6
       with:
-        # Build on a supported LTS; the matrix job below covers every
-        # Node version the library claims to support.
+        # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
+        # in October. The library itself still supports Node >= 18
+        # (see test matrix below).
         node-version: 24
         cache: 'pnpm'
         cache-dependency-path: nodejs/pnpm-lock.yaml
@@ -83,7 +84,7 @@ jobs:
     timeout-minutes: 30
     strategy:
       matrix:
-        node-version: [ "22", "24", "26" ]
+        node-version: [ "18", "20" ]
     runs-on: "ubuntu-22.04"
     defaults:
       run:
@@ -100,9 +101,9 @@ jobs:
     - uses: actions/setup-node@v6
       name: Setup Node.js 24 for build
       with:
-        # Build and install once on a fixed version so the generated docs
-        # are identical across matrix legs; the tests below then run on each
-        # supported Node version.
+        # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
+        # in October. Build/install runs on Node 24; tests run on the
+        # matrix version below using direct jest invocation.
         node-version: 24
         cache: 'pnpm'
         cache-dependency-path: nodejs/pnpm-lock.yaml
@@ -151,9 +152,9 @@ jobs:
         S3_TEST: "1"
         # Newer @smithy/core uses dynamic ESM imports.
         NODE_OPTIONS: "--experimental-vm-modules"
-      # Invoke the installed jest binary directly; the pnpm shim is set up
-      # against the build-phase Node, not the version selected above.
-      run: node_modules/.bin/jest --verbose
+      # Invoke jest directly because pnpm 11 itself requires Node 22+
+      # while the matrix tests on older Node versions.
+      run: npx jest --verbose
     - name: Test examples
       working-directory: ./
       env:
@@ -163,7 +164,7 @@ jobs:
       run: |
         python ci/mock_openai.py &
         cd nodejs/examples
-        node_modules/.bin/jest --testEnvironment jest-environment-node-single-context --verbose
+        npx jest --testEnvironment jest-environment-node-single-context --verbose
   macos:
     timeout-minutes: 30
     # macos-15 ships a newer linker; the older macos-14 linker fails to insert
@@ -184,7 +185,8 @@ jobs:
         version: 11.1.1
     - uses: actions/setup-node@v6
       with:
-        # pnpm 11 requires Node >= 22.13.
+        # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
+        # in October.
         node-version: 24
         cache: 'pnpm'
         cache-dependency-path: nodejs/pnpm-lock.yaml

.github/workflows/npm-publish.yml

@@ -40,31 +40,40 @@ jobs:
           - target: aarch64-apple-darwin
             host: macos-latest
             features: fp16kernels
-            # Fat LTO was ~111 of this job's ~113 minutes.
-            lto: thin
-            codegen_units: 16
             pre_build: |-
               brew install protobuf
+              # Fat LTO (the workspace default in .cargo/config.toml) is
+              # single-threaded and is the peak-memory step of the build. On
+              # this runner it accounted for ~111 of the job's ~113 minutes,
+              # making it the critical path of the entire publish pipeline.
+              # ThinLTO parallelizes it across the runner's cores, for a few
+              # percent of runtime performance.
+              export CARGO_PROFILE_RELEASE_LTO=thin
+              export CARGO_PROFILE_RELEASE_CODEGEN_UNITS=16
           - target: x86_64-pc-windows-msvc
             host: windows-2025
             features: ","
-            # The lower peak also keeps this on the standard 4-core runner.
-            lto: thin
-            codegen_units: 16
             pre_build: |-
               choco install --no-progress protoc ninja nasm
               tail -n 1000 /c/ProgramData/chocolatey/logs/chocolatey.log
               # There is an issue where choco doesn't add nasm to the path
               export PATH="$PATH:/c/Program Files/NASM"
               nasm -v
+              # See the ThinLTO note on aarch64-apple-darwin above. Keeping
+              # peak memory down is also what lets this run on the standard
+              # 4-core runner: the 8-core larger runner was only needed to
+              # stop fat LTO from OOMing rustc-LLVM.
+              export CARGO_PROFILE_RELEASE_LTO=thin
+              export CARGO_PROFILE_RELEASE_CODEGEN_UNITS=16
           - target: aarch64-pc-windows-msvc
             host: windows-2025
             features: ","
-            lto: thin
-            codegen_units: 16
             pre_build: |-
                 choco install --no-progress protoc
                 rustup target add aarch64-pc-windows-msvc
+                # See the ThinLTO note on aarch64-apple-darwin above.
+                export CARGO_PROFILE_RELEASE_LTO=thin
+                export CARGO_PROFILE_RELEASE_CODEGEN_UNITS=16
           - target: x86_64-unknown-linux-gnu
             host: ubuntu-latest
             features: fp16kernels
@@ -94,14 +103,6 @@ jobs:
             # https://github.com/napi-rs/napi-rs/blob/main/debian-aarch64.Dockerfile
             docker: ghcr.io/napi-rs/napi-rs/nodejs-rust:lts-debian-aarch64
             features: "fp16kernels"
-            # Fat LTO OOM-killed rustc every nightly; even with lld it peaked
-            # at 31391 MiB of the runner's 32 GiB.
-            lto: thin
-            codegen_units: 16
-            # arm64 Linux links through GNU `ld` where x86_64 defaults to
-            # `rust-lld`, which is why only arm64 OOM'd. lld cut the largest
-            # linker process 7.0 -> 4.0 GiB (lancedb/sophon#7313).
-            linker: /tmp/aarch64-lld-clang
             pre_build: |-
               set -e &&
               apt-get update &&
@@ -111,30 +112,9 @@ jobs:
               # AT_HWCAP2 (added in Linux 3.17). Define it for aws-lc-sys.
               export CFLAGS="$CFLAGS -DAT_HWCAP2=26" &&
               rustup target add aarch64-unknown-linux-gnu
-              # Not `&&`-chained: in dash, errexit does not fire for a
-              # non-final command in an `&&` list, so failures were ignored.
-              #
-              # A wrapper rather than `-C link-arg` because the per-target
-              # rustflags variable does not reach every unit that links, while
-              # the linker variable does. `clang` because GCC silently ignores
-              # `-fuse-ld=lld` unless built with lld support. Two echoes
-              # because printf's newline escape gets rewritten to `;` between
-              # here and the container.
-              echo '#!/bin/sh' > /tmp/aarch64-lld-clang
-              echo 'exec clang --target=aarch64-unknown-linux-gnu --sysroot=/usr/aarch64-unknown-linux-gnu/aarch64-unknown-linux-gnu/sysroot --gcc-toolchain=/usr/aarch64-unknown-linux-gnu -fuse-ld=lld "$@"' >> /tmp/aarch64-lld-clang
-              chmod 0755 /tmp/aarch64-lld-clang
-              # Fail now, not at the cdylib link ~30 minutes later. Linking at
-              # all also proves lld resolved; clang errors out when it cannot.
-              echo 'int main(void){return 0;}' > /tmp/probe.c
-              /tmp/aarch64-lld-clang /tmp/probe.c -o /tmp/probe
-              readelf -h /tmp/probe | grep AArch64
           - target: aarch64-unknown-linux-musl
             host: ubuntu-2404-8x-x64
             features: ","
-            # Fat LTO took the whole runner down. lld cannot help: it died
-            # inside rustc's LLVM, before any linker was spawned.
-            lto: thin
-            codegen_units: 16
             pre_build: |-
               set -e &&
               sudo apt-get update &&
@@ -143,19 +123,6 @@ jobs:
               export EXTRA_ARGS="-x"
     name: build - ${{ matrix.settings.target }}
     runs-on: ${{ matrix.settings.host }}
-    # On the job, not exported from `pre_build`: `Swatinem/rust-cache` hashes
-    # `CARGO_*` into its cache key before any step runs, so a step-local export
-    # leaves the key unchanged while cargo still rebuilds cold. The ThinLTO
-    # legs had been doing that every run.
-    #
-    # Not `RUSTFLAGS`: setting it, even to "", discards every config-file
-    # rustflag, silently dropping .cargo/config.toml's `target-cpu` and
-    # `target-feature` from the published binaries.
-    env:
-      CARGO_PROFILE_RELEASE_LTO: ${{ matrix.settings.lto || 'fat' }}
-      CARGO_PROFILE_RELEASE_CODEGEN_UNITS: ${{ matrix.settings.codegen_units || '1' }}
-      # Empty elsewhere: a per-target variable is only read for that triple.
-      CARGO_TARGET_AARCH64_UNKNOWN_LINUX_GNU_LINKER: ${{ matrix.settings.linker }}
     defaults:
       run:
         working-directory: nodejs
@@ -168,7 +135,8 @@ jobs:
       - name: Setup node
         uses: actions/setup-node@v6
         with:
-          # pnpm 11 requires Node >= 22.13.
+          # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
+          # in October.
           node-version: 24
           cache: pnpm
           cache-dependency-path: nodejs/pnpm-lock.yaml
@@ -201,15 +169,19 @@ jobs:
           # creating ref). The nightly cadence also keeps entries inside
           # GitHub's 7-day eviction window, which a tag-only trigger would not.
           save-if: ${{ github.ref == 'refs/heads/main' }}
-      # Docker builds can use rust-cache too: the workspace is bind-mounted, so
-      # `target/` lives on the host and rust-cache's prune keeps the entry
-      # small.
+      # Docker builds can use rust-cache too. `target/` already lives on the
+      # host because the whole workspace is bind-mounted into the container, and
+      # rust-cache's prune and save run host-side, so they can manage it -- which
+      # is what keeps the entry to dependency artifacts rather than a multi-GB
+      # copy of everything.
       #
       # Two differences from the native builds. The container's CARGO_HOME is
-      # bind-mounted from `.cargo-cache` rather than ~/.cargo, so that is cached
-      # explicitly. And the key uses the *host* rustc version, not the compiler
-      # that built these artifacts -- safe, since cargo fingerprints the real
-      # one; a base-image bump just costs one cold build.
+      # bind-mounted from `.cargo-cache` rather than the host's ~/.cargo, so that
+      # has to be cached explicitly. And the key is derived from the *host* rustc
+      # version, which is not the compiler that produced these artifacts; that is
+      # safe because cargo fingerprints the real compiler and rebuilds on a
+      # mismatch, it just means a base-image toolchain bump costs one cold build
+      # instead of invalidating the key.
       - name: Cache cargo (docker builds)
         uses: Swatinem/rust-cache@v2
         if: ${{ matrix.settings.docker }}
@@ -238,19 +210,14 @@ jobs:
           # cache step above saves. Previously the registry mounts pointed at
           # `.cargo/...`, a path nothing cached, so the container re-downloaded
           # the whole crate registry on every run.
-          #
-          # `docker run` inherits nothing; `-e NAME` carries the job's `env:` in.
           options: "--user 0:0 -v ${{ github.workspace }}/.cargo-cache/git/db:/usr/local/cargo/git/db \
                     -v ${{ github.workspace }}/.cargo-cache/registry/cache:/usr/local/cargo/registry/cache \
                     -v ${{ github.workspace }}/.cargo-cache/registry/index:/usr/local/cargo/registry/index \
-                    -e CARGO_PROFILE_RELEASE_LTO \
-                    -e CARGO_PROFILE_RELEASE_CODEGEN_UNITS \
-                    -e CARGO_TARGET_AARCH64_UNKNOWN_LINUX_GNU_LINKER \
                     -v ${{ github.workspace }}:/build -w /build/nodejs"
           run: |
             set -e
             ${{ matrix.settings.pre_build }}
-            node_modules/.bin/napi build --platform --release \
+            npx napi build --platform --release \
               --features ${{ matrix.settings.features }} \
               --target ${{ matrix.settings.target }} \
               --dts ../lancedb/native.d.ts \
@@ -270,7 +237,7 @@ jobs:
       - name: Build
         run: |
           ${{ matrix.settings.pre_build }}
-          node_modules/.bin/napi build --platform --release \
+          npx napi build --platform --release \
               --features ${{ matrix.settings.features }} \
               --target ${{ matrix.settings.target }} \
               --dts ../lancedb/native.d.ts \
@@ -289,18 +256,6 @@ jobs:
         if: always()
         run: df -h
         shell: bash
-      - name: Report peak memory
-        if: always() && runner.os == 'Linux'
-        shell: bash
-        run: |
-          peak=$(find /sys/fs/cgroup -name memory.peak -readable \
-                 -exec cat {} + 2>/dev/null | sort -n | tail -1)
-          if [ -n "$peak" ]; then
-            echo "peak memory: $((peak / 1024 / 1024)) MiB"
-          else
-            echo "peak memory: unavailable (no readable cgroup v2 memory.peak)"
-          fi
-          free -g || true
       - name: Upload artifact
         uses: actions/upload-artifact@v7
         with:
@@ -338,7 +293,7 @@ jobs:
           - target: aarch64-unknown-linux-gnu
             host: ubuntu-2404-8x-arm64
         node:
-          - '22'
+          - '20'
     runs-on: ${{ matrix.settings.host }}
     defaults:
       run:
@@ -384,9 +339,9 @@ jobs:
       - name: Move built files
         run: cp dist/native.d.ts dist/native.js dist/*.node lancedb/
       - name: Test bindings
-        # Invoke the installed jest binary directly; the pnpm shim is set up
-        # against the install-phase Node, not the version selected above.
-        run: node_modules/.bin/jest --verbose
+        # Invoke jest directly because pnpm 11 itself requires Node 22+
+        # while the matrix tests on older Node versions.
+        run: npx jest --verbose
   publish:
     name: Publish
     runs-on: ubuntu-latest

.github/workflows/update_package_lock_run.yml

@@ -0,0 +1,22 @@
+name: Update package-lock.json
+
+on:
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v6
+        with:
+          ref: main
+          persist-credentials: false
+          fetch-depth: 0
+          lfs: true
+      - uses: ./.github/workflows/update_package_lock
+        with:
+          github_token: ${{ secrets.LANCEDB_RELEASE_TOKEN }}

.github/workflows/update_package_lock_run_nodejs.yml

@@ -0,0 +1,22 @@
+name: Update NodeJs package-lock.json
+
+on:
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v6
+        with:
+          ref: main
+          persist-credentials: false
+          fetch-depth: 0
+          lfs: true
+      - uses: ./.github/workflows/update_package_lock_nodejs
+        with:
+          github_token: ${{ secrets.LANCEDB_RELEASE_TOKEN }}

.pre-commit-config.yaml

@@ -20,10 +20,7 @@ repos:
     hooks:
       - id: local-biome-check
         name: biome check
-        # Use the biome from nodejs/package.json rather than a separately
-        # pinned one: the two drifted apart and disagreed on formatting, so
-        # this hook rejected code that `pnpm lint` accepted.
-        entry: nodejs/node_modules/.bin/biome check --config-path nodejs/biome.json nodejs/
+        entry: npx @biomejs/biome@1.8.3 check --config-path nodejs/biome.json nodejs/
         language: system
         types: [text]
         files: "nodejs/.*"

AGENTS.md

@@ -38,7 +38,7 @@ Before committing changes, run formatting for every language you touched. At min
 * Rust changes: run `cargo fmt --all`.
 * Python changes: run `ruff format .` and `ruff check .` from the repository root,
   and run targeted tests through `cd python && uv run ...`.
-* TypeScript changes: run the relevant `pnpm` lint, format, build, and docs commands in `nodejs`.
+* TypeScript changes: run the relevant `npm`/`pnpm` lint, format, build, and docs commands in `nodejs`.
 
 Before creating a PR, the exact value passed to `gh pr create --title` must follow
 Conventional Commits, such as `fix: support nested field paths in native index creation`
@@ -101,12 +101,12 @@ Python bindings changes:
 TypeScript bindings changes:
 
 1. Add napi-rs method binding on `Table` in `nodejs/src/table.rs`.
-2. Run `pnpm build` to generate TypeScript definitions.
+2. Run `npm run build` to generate TypeScript definitions.
 3. Add typescript method on abstract class `Table` in `nodejs/src/table.ts`.
 4. Add concrete method on `LocalTable` class in `nodejs/src/native_table.ts`.
     * Note: despite the name, this class is also used for remote tables.
 5. Add test in `nodejs/__test__/table.test.ts`.
-6. Run `pnpm run docs` to generate TypeScript documentation.
+6. Run `npm run docs` to generate TypeScript documentation.
 
 ## Python API reference
 

Cargo.lock

@@ -1597,9 +1597,9 @@ checksum = "613afe47fcd5fac7ccf1db93babcb082c5994d996f20b8b159f2ad1658eb5724"
 
 [[package]]
 name = "chacha20"
-version = "0.10.2"
+version = "0.10.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "65c35e4b699c7e15ccbe7ee35c005e4fc0a278d22238a2857e6ce2dadeda1b06"
+checksum = "6f8d983286843e49675a4b7a2d174efe136dc93a18d69130dd18198a6c167601"
 dependencies = [
  "cfg-if 1.0.4",
  "cpufeatures 0.3.0",
@@ -3455,8 +3455,8 @@ checksum = "42703706b716c37f96a77aea830392ad231f44c9e9a67872fa5548707e11b11c"
 
 [[package]]
 name = "fsst"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "rand 0.9.5",
@@ -4815,8 +4815,8 @@ checksum = "e037a2e1d8d5fdbd49b16a4ea09d5d6401c1f29eca5ff29d03d3824dba16256a"
 
 [[package]]
 name = "lance"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arc-swap",
  "arrow",
@@ -4888,8 +4888,8 @@ dependencies = [
 
 [[package]]
 name = "lance-arrow"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -4911,7 +4911,7 @@ dependencies = [
 [[package]]
 name = "lance-arrow-scalar"
 version = "58.0.0"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -4925,7 +4925,7 @@ dependencies = [
 [[package]]
 name = "lance-arrow-stats"
 version = "58.0.0"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "arrow-schema",
@@ -4934,8 +4934,8 @@ dependencies = [
 
 [[package]]
 name = "lance-bitpacking"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrayref",
  "crunchy",
@@ -4945,8 +4945,8 @@ dependencies = [
 
 [[package]]
 name = "lance-core"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -4983,8 +4983,8 @@ dependencies = [
 
 [[package]]
 name = "lance-datafusion"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow",
  "arrow-array",
@@ -5013,8 +5013,8 @@ dependencies = [
 
 [[package]]
 name = "lance-datagen"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow",
  "arrow-array",
@@ -5031,8 +5031,8 @@ dependencies = [
 
 [[package]]
 name = "lance-derive"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -5041,8 +5041,8 @@ dependencies = [
 
 [[package]]
 name = "lance-encoding"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-arith",
  "arrow-array",
@@ -5075,8 +5075,8 @@ dependencies = [
 
 [[package]]
 name = "lance-file"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-arith",
  "arrow-array",
@@ -5107,8 +5107,8 @@ dependencies = [
 
 [[package]]
 name = "lance-index"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arc-swap",
  "arrow",
@@ -5172,8 +5172,8 @@ dependencies = [
 
 [[package]]
 name = "lance-index-core"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "arrow-schema",
@@ -5195,8 +5195,8 @@ dependencies = [
 
 [[package]]
 name = "lance-io"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow",
  "arrow-array",
@@ -5236,8 +5236,8 @@ dependencies = [
 
 [[package]]
 name = "lance-linalg"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "arrow-schema",
@@ -5251,8 +5251,8 @@ dependencies = [
 
 [[package]]
 name = "lance-namespace"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow",
  "async-trait",
@@ -5264,8 +5264,8 @@ dependencies = [
 
 [[package]]
 name = "lance-namespace-impls"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow",
  "arrow-ipc",
@@ -5318,8 +5318,8 @@ dependencies = [
 
 [[package]]
 name = "lance-select"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -5333,8 +5333,8 @@ dependencies = [
 
 [[package]]
 name = "lance-table"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow",
  "arrow-array",
@@ -5374,8 +5374,8 @@ dependencies = [
 
 [[package]]
 name = "lance-testing"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "arrow-schema",
@@ -5388,8 +5388,8 @@ dependencies = [
 
 [[package]]
 name = "lance-tokenizer"
-version = "12.0.0-beta.5"
-source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.5#556637791d0048c2b4f1342dd84b67c8bbd65259"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "frostem",
  "icu_segmenter",
@@ -5402,7 +5402,7 @@ dependencies = [
 
 [[package]]
 name = "lancedb"
-version = "0.38.0-beta.12"
+version = "0.38.0-beta.11"
 dependencies = [
  "ahash",
  "anyhow",
@@ -5490,7 +5490,7 @@ dependencies = [
 
 [[package]]
 name = "lancedb-nodejs"
-version = "0.38.0-beta.12"
+version = "0.38.0-beta.11"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -5515,7 +5515,7 @@ dependencies = [
 
 [[package]]
 name = "lancedb-python"
-version = "0.38.0-beta.12"
+version = "0.38.0-beta.11"
 dependencies = [
  "arrow",
  "async-trait",

Cargo.toml

@@ -13,20 +13,20 @@ categories = ["database-implementations"]
 rust-version = "1.91.0"
 
 [workspace.dependencies]
-lance = { "version" = "=12.0.0-beta.5", default-features = false, "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-core = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-datagen = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-file = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-io = { "version" = "=12.0.0-beta.5", default-features = false, "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-index = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-linalg = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-namespace = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-namespace-impls = { "version" = "=12.0.0-beta.5", default-features = false, "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-table = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-testing = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-datafusion = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-encoding = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
-lance-arrow = { "version" = "=12.0.0-beta.5", "tag" = "v12.0.0-beta.5", "git" = "https://github.com/lance-format/lance.git" }
+lance = { "version" = "=12.0.0-beta.2", default-features = false, "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-core = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-datagen = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-file = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-io = { "version" = "=12.0.0-beta.2", default-features = false, "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-index = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-linalg = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-namespace = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-namespace-impls = { "version" = "=12.0.0-beta.2", default-features = false, "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-table = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-testing = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-datafusion = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-encoding = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-arrow = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
 lancedb = { path = "rust/lancedb", default-features = false }
 ahash = "0.8"
 # Note that this one does not include pyarrow

Makefile

@@ -5,5 +5,5 @@ licenses:
 	cd python && cargo about generate ../about.hbs -o RUST_THIRD_PARTY_LICENSES.html -c ../about.toml
 	cd python && uv sync --all-extras && uv tool run pip-licenses --python .venv/bin/python --format=markdown --with-urls --output-file=PYTHON_THIRD_PARTY_LICENSES.md
 	cd nodejs && cargo about generate ../about.hbs -o RUST_THIRD_PARTY_LICENSES.html -c ../about.toml
-	cd nodejs && pnpm dlx license-checker@25 --markdown --out NODEJS_THIRD_PARTY_LICENSES.md
+	cd nodejs && npx license-checker --markdown --out NODEJS_THIRD_PARTY_LICENSES.md
 	cd java && ./mvnw license:aggregate-add-third-party -q

ci/update_lockfiles.sh

@@ -12,12 +12,16 @@ done
 # This updates the lockfile without building
 cargo metadata --quiet > /dev/null
 
+pushd nodejs || exit 1
+npm install --package-lock-only --silent
+popd
+
 if git diff --quiet --exit-code; then
   echo "No lockfile changes to commit; skipping amend."
 elif $AMEND; then
-  git add Cargo.lock
+  git add Cargo.lock nodejs/package-lock.json
   git commit --amend --no-edit
 else
-  git add Cargo.lock
+  git add Cargo.lock nodejs/package-lock.json
   git commit -m "Update lockfiles"
 fi

deny.toml

@@ -131,13 +131,18 @@ allow = [
     "BSD-3-Clause",
     "ISC",
     "Unicode-3.0",
+    "Unicode-DFS-2016",
     "Zlib",
     "CC0-1.0",
     "MPL-2.0",
     "BSL-1.0",
+    "OpenSSL",
     # 0BSD ("BSD Zero Clause") is effectively public domain — no attribution
     # required. Pulled in by `mock_instant`.
     "0BSD",
+    # bzip2-1.0.6 is the permissive upstream bzip2 license (BSD-like). Pulled
+    # in by `libbz2-rs-sys`, the pure-Rust bzip2 implementation.
+    "bzip2-1.0.6",
     # CDLA-Permissive-2.0 is a permissive data license used by `webpki-roots`
     # for the Mozilla CA root bundle. Data-only, distribution-compatible.
     "CDLA-Permissive-2.0",
@@ -145,7 +150,12 @@ allow = [
 confidence-threshold = 0.8
 # Per-crate license exceptions: allow a license for a specific crate only,
 # rather than globally via the `allow` list above.
-exceptions = []
+exceptions = [
+    # CDDL-1.0 (copyleft) is pulled in only as a dev/profiling dependency via
+    # `inferno` -> `pprof` -> `lance-testing`; it is a test dependency that we
+    # do not distribute, so scope the allowance to `inferno` alone.
+    { allow = ["CDDL-1.0"], crate = "inferno" },
+]
 # Crates whose license cannot be determined from Cargo metadata but whose
 # license we've manually confirmed from upstream. Keep this list minimal.
 [[licenses.clarify]]

docs/README.md

@@ -47,24 +47,22 @@ pytest -vv python/tests/docs
 
 ### Checking typescript examples
 
-The examples depend on `@lancedb/lancedb` at `file:../dist`, so the package must be
-built before running the tests. This uses pnpm; see the
-[Typescript contributing guide](../nodejs/CONTRIBUTING.md) for the toolchain setup.
+The `@lancedb/lancedb` package must be built before running the tests:
 
 ```shell
 pushd nodejs
-pnpm install
-pnpm build
+npm ci
+npm run build
 popd
 ```
 
-Then you can run the examples by going to the `nodejs/examples` directory, which is a
-separate pnpm package with its own lockfile:
+Then you can run the examples by going to the `nodejs/examples` directory and
+running the tests like a normal npm package:
 
 ```shell
 pushd nodejs/examples
-pnpm install
-pnpm test
+npm ci
+npm test
 popd
 ```
 
@@ -86,7 +84,6 @@ The new files should be checked into the repository.
 
 ```shell
 pushd nodejs
-# `pnpm docs` would invoke pnpm's built-in `docs` command, not the script.
-pnpm run docs
+npm run docs
 popd
 ```

docs/package-lock.json

@@ -0,0 +1,135 @@
+{
+  "name": "lancedb-docs-test",
+  "version": "1.0.0",
+  "lockfileVersion": 3,
+  "requires": true,
+  "packages": {
+    "": {
+      "name": "lancedb-docs-test",
+      "version": "1.0.0",
+      "license": "Apache 2",
+      "dependencies": {
+        "apache-arrow": "file:../node/node_modules/apache-arrow",
+        "vectordb": "file:../node"
+      },
+      "devDependencies": {
+        "@types/node": "^20.11.8",
+        "typescript": "^5.3.3"
+      }
+    },
+    "../node": {
+      "name": "vectordb",
+      "version": "0.21.2-beta.0",
+      "cpu": [
+        "x64",
+        "arm64"
+      ],
+      "license": "Apache-2.0",
+      "os": [
+        "darwin",
+        "linux",
+        "win32"
+      ],
+      "dependencies": {
+        "@neon-rs/load": "^0.0.74",
+        "axios": "^1.4.0"
+      },
+      "devDependencies": {
+        "@neon-rs/cli": "^0.0.160",
+        "@types/chai": "^4.3.4",
+        "@types/chai-as-promised": "^7.1.5",
+        "@types/mocha": "^10.0.1",
+        "@types/node": "^18.16.2",
+        "@types/sinon": "^10.0.15",
+        "@types/temp": "^0.9.1",
+        "@types/uuid": "^9.0.3",
+        "@typescript-eslint/eslint-plugin": "^5.59.1",
+        "apache-arrow-old": "npm:apache-arrow@13.0.0",
+        "cargo-cp-artifact": "^0.1",
+        "chai": "^4.3.7",
+        "chai-as-promised": "^7.1.1",
+        "eslint": "^8.39.0",
+        "eslint-config-standard-with-typescript": "^34.0.1",
+        "eslint-plugin-import": "^2.26.0",
+        "eslint-plugin-n": "^15.7.0",
+        "eslint-plugin-promise": "^6.1.1",
+        "mocha": "^10.2.0",
+        "openai": "^4.24.1",
+        "sinon": "^15.1.0",
+        "temp": "^0.9.4",
+        "ts-node": "^10.9.1",
+        "ts-node-dev": "^2.0.0",
+        "typedoc": "^0.24.7",
+        "typedoc-plugin-markdown": "^3.15.3",
+        "typescript": "^5.1.0",
+        "uuid": "^9.0.0"
+      },
+      "optionalDependencies": {
+        "@lancedb/vectordb-darwin-arm64": "0.21.2-beta.0",
+        "@lancedb/vectordb-darwin-x64": "0.21.2-beta.0",
+        "@lancedb/vectordb-linux-arm64-gnu": "0.21.2-beta.0",
+        "@lancedb/vectordb-linux-x64-gnu": "0.21.2-beta.0",
+        "@lancedb/vectordb-win32-x64-msvc": "0.21.2-beta.0"
+      },
+      "peerDependencies": {
+        "@apache-arrow/ts": "^14.0.2",
+        "apache-arrow": "^14.0.2"
+      }
+    },
+    "../node/node_modules/apache-arrow": {
+      "version": "14.0.2",
+      "license": "Apache-2.0",
+      "dependencies": {
+        "@types/command-line-args": "5.2.0",
+        "@types/command-line-usage": "5.0.2",
+        "@types/node": "20.3.0",
+        "@types/pad-left": "2.1.1",
+        "command-line-args": "5.2.1",
+        "command-line-usage": "7.0.1",
+        "flatbuffers": "23.5.26",
+        "json-bignum": "^0.0.3",
+        "pad-left": "^2.1.0",
+        "tslib": "^2.5.3"
+      },
+      "bin": {
+        "arrow2csv": "bin/arrow2csv.js"
+      }
+    },
+    "node_modules/@types/node": {
+      "version": "20.11.8",
+      "resolved": "https://registry.npmjs.org/@types/node/-/node-20.11.8.tgz",
+      "integrity": "sha512-i7omyekpPTNdv4Jb/Rgqg0RU8YqLcNsI12quKSDkRXNfx7Wxdm6HhK1awT3xTgEkgxPn3bvnSpiEAc7a7Lpyow==",
+      "dev": true,
+      "dependencies": {
+        "undici-types": "~5.26.4"
+      }
+    },
+    "node_modules/apache-arrow": {
+      "resolved": "../node/node_modules/apache-arrow",
+      "link": true
+    },
+    "node_modules/typescript": {
+      "version": "5.3.3",
+      "resolved": "https://registry.npmjs.org/typescript/-/typescript-5.3.3.tgz",
+      "integrity": "sha512-pXWcraxM0uxAS+tN0AG/BF2TyqmHO014Z070UsJ+pFvYuRSq8KH8DmWpnbXe0pEPDHXZV3FcAbJkijJ5oNEnWw==",
+      "dev": true,
+      "bin": {
+        "tsc": "bin/tsc",
+        "tsserver": "bin/tsserver"
+      },
+      "engines": {
+        "node": ">=14.17"
+      }
+    },
+    "node_modules/undici-types": {
+      "version": "5.26.5",
+      "resolved": "https://registry.npmjs.org/undici-types/-/undici-types-5.26.5.tgz",
+      "integrity": "sha512-JlCMO+ehdEIKqlFxk6IfVoAUVmgz7cU7zD/h9XZ0qzeosSHmUJVOzSQvvYSYWXkFXC+IfLKSIffhv0sVZup6pA==",
+      "dev": true
+    },
+    "node_modules/vectordb": {
+      "resolved": "../node",
+      "link": true
+    }
+  }
+}

docs/package.json

@@ -0,0 +1,20 @@
+{
+  "name": "lancedb-docs-test",
+  "version": "1.0.0",
+  "description": "auto-generated tests from doc",
+  "author": "dev@lancedb.com",
+  "license": "Apache 2",
+  "dependencies": {
+    "apache-arrow": "file:../node/node_modules/apache-arrow",
+    "vectordb": "file:../node"
+  },
+  "scripts": {
+    "build": "tsc -b && cd ../node && npm run build-release",
+    "example": "npm run build && node",
+    "test": "npm run build && ls dist/*.js | xargs -n 1 node"
+  },
+  "devDependencies": {
+    "@types/node": "^20.11.8",
+    "typescript": "^5.3.3"
+  }
+}

docs/src/java/java.md

@@ -14,7 +14,7 @@ Add the following dependency to your `pom.xml`:
 <dependency>
     <groupId>com.lancedb</groupId>
     <artifactId>lancedb-core</artifactId>
-    <version>0.38.0-beta.13</version>
+    <version>0.38.0-beta.11</version>
 </dependency>
 ```
 

docs/tsconfig.json

@@ -0,0 +1,17 @@
+{
+  "include": [
+    "src/*.ts",
+  ],
+  "compilerOptions": {
+    "target": "es2022",
+    "module": "nodenext",
+    "declaration": true,
+    "outDir": "./dist",
+    "strict": true,
+    "allowJs": true,
+    "resolveJsonModule": true,
+  },
+  "exclude": [
+    "./dist/*",
+  ]
+}

java/lancedb-core/pom.xml

@@ -8,7 +8,7 @@
     <parent>
       <groupId>com.lancedb</groupId>
       <artifactId>lancedb-parent</artifactId>
-      <version>0.38.0-beta.13</version>
+      <version>0.38.0-beta.11</version>
       <relativePath>../pom.xml</relativePath>
     </parent>
 

java/pom.xml

@@ -6,7 +6,7 @@
 
     <groupId>com.lancedb</groupId>
     <artifactId>lancedb-parent</artifactId>
-    <version>0.38.0-beta.13</version>
+    <version>0.38.0-beta.11</version>
     <packaging>pom</packaging>
     <name>${project.artifactId}</name>
     <description>LanceDB Java SDK Parent POM</description>
@@ -28,7 +28,7 @@
     <properties>
         <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
         <arrow.version>15.0.0</arrow.version>
-        <lance-core.version>12.0.0-beta.5</lance-core.version>
+        <lance-core.version>12.0.0-beta.2</lance-core.version>
         <spotless.skip>false</spotless.skip>
         <spotless.version>2.30.0</spotless.version>
         <spotless.java.googlejavaformat.version>1.7</spotless.java.googlejavaformat.version>

nodejs/Cargo.toml

@@ -1,7 +1,7 @@
 [package]
 name = "lancedb-nodejs"
 edition.workspace = true
-version = "0.38.0-beta.13"
+version = "0.38.0-beta.11"
 publish = false
 license.workspace = true
 description.workspace = true

nodejs/__test__/package.test.ts

@@ -5,8 +5,8 @@ import packageJson = require("../package.json");
 
 describe("package metadata", () => {
   it("requires Node.js type declarations compatible with the runtime", () => {
-    expect(packageJson.engines.node).toBe(">= 22");
-    expect(packageJson.peerDependencies["@types/node"]).toBe(">=22");
+    expect(packageJson.engines.node).toBe(">= 18");
+    expect(packageJson.peerDependencies["@types/node"]).toBe(">=18");
     expect(packageJson.peerDependenciesMeta["@types/node"]).toEqual({
       optional: true,
     });

nodejs/__test__/remote.test.ts

@@ -3,7 +3,6 @@
 
 import * as http from "http";
 import { RequestListener } from "http";
-import packageJson = require("../package.json");
 import {
   ClientConfig,
   Connection,
@@ -71,13 +70,7 @@ async function withMockDatabase(
   try {
     await callback(db);
   } finally {
-    // `close()` alone leaves the port bound until keep-alive sockets drain, so
-    // a single failing test would cascade into EADDRINUSE for every test after
-    // it. Destroy the connections and wait for the port to actually be free.
-    await new Promise<void>((resolve) => {
-      server.closeAllConnections();
-      server.close(() => resolve());
-    });
+    server.close();
   }
 }
 
@@ -138,7 +131,7 @@ describe("remote connection", () => {
       (req, res) => {
         expect(req.headers["x-api-key"]).toEqual("fake");
         expect(req.headers["user-agent"]).toEqual(
-          `LanceDB-Node-Client/${packageJson.version}`,
+          `LanceDB-Node-Client/${process.env.npm_package_version}`,
         );
 
         const body = JSON.stringify({ tables: [] });

nodejs/examples/package.json

@@ -8,8 +8,7 @@
     "//1": "--experimental-vm-modules is needed to run jest with sentence-transformers",
     "//2": "--testEnvironment is needed to run jest with sentence-transformers",
     "//3": "See: https://github.com/huggingface/transformers.js/issues/57",
-    "//4": "jest is invoked by its JS entry, not node_modules/.bin/jest: under pnpm that path is a shell shim, which `node` cannot execute",
-    "test": "node --experimental-vm-modules node_modules/jest/bin/jest.js --testEnvironment jest-environment-node-single-context --verbose",
+    "test": "node --experimental-vm-modules node_modules/.bin/jest --testEnvironment jest-environment-node-single-context --verbose",
     "lint": "biome check *.ts && biome format *.ts",
     "lint-ci": "biome ci .",
     "lint-fix": "biome check --write *.ts && pnpm format",

nodejs/npm/darwin-arm64/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-darwin-arm64",
-	"version": "0.38.0-beta.13",
+	"version": "0.38.0-beta.11",
 	"os": ["darwin"],
 	"cpu": ["arm64"],
 	"main": "lancedb.darwin-arm64.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-gnu",
-	"version": "0.38.0-beta.13",
+	"version": "0.38.0-beta.11",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-gnu.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-musl",
-	"version": "0.38.0-beta.13",
+	"version": "0.38.0-beta.11",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-musl.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-gnu",
-	"version": "0.38.0-beta.13",
+	"version": "0.38.0-beta.11",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-gnu.node",

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-musl",
-	"version": "0.38.0-beta.13",
+	"version": "0.38.0-beta.11",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-musl.node",

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

@@ -1,6 +1,6 @@
 {
   "name": "@lancedb/lancedb-win32-arm64-msvc",
-  "version": "0.38.0-beta.13",
+  "version": "0.38.0-beta.11",
   "os": [
     "win32"
   ],

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-win32-x64-msvc",
-	"version": "0.38.0-beta.13",
+	"version": "0.38.0-beta.11",
 	"os": ["win32"],
 	"cpu": ["x64"],
 	"main": "lancedb.win32-x64-msvc.node",

nodejs/package.json

@@ -11,7 +11,7 @@
     "ann"
   ],
   "private": false,
-  "version": "0.38.0-beta.13",
+  "version": "0.38.0-beta.11",
   "main": "dist/index.js",
   "exports": {
     ".": "./dist/index.js",
@@ -67,7 +67,7 @@
     "timeout": "3m"
   },
   "engines": {
-    "node": ">= 22"
+    "node": ">= 18"
   },
   "packageManager": "pnpm@11.1.1",
   "cpu": ["x64", "arm64"],
@@ -101,7 +101,7 @@
     "openai": "4.29.2"
   },
   "peerDependencies": {
-    "@types/node": ">=22",
+    "@types/node": ">=18",
     "apache-arrow": ">=15.0.0 <=18.1.0"
   },
   "peerDependenciesMeta": {

python/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb-python"
-version = "0.38.0-beta.13"
+version = "0.38.0-beta.11"
 publish = false
 edition.workspace = true
 description = "Python bindings for LanceDB"

python/python/lancedb/db.py

@@ -16,6 +16,7 @@ from typing import (
     Iterable,
     List,
     Literal,
+    Mapping,
     Optional,
     Union,
 )
@@ -687,17 +688,35 @@ class DBConnection(EnforceOverrides):
         """
         raise NotImplementedError("serialize is not supported for this connection type")
 
-    def create_function(self, definition: UdfDefinition) -> FunctionVersion:
+    def create_function(
+        self,
+        definition: UdfDefinition,
+        *,
+        secrets: Optional[Mapping[str, str]] = None,
+    ) -> FunctionVersion:
         """Register a scalar Python UDF and wait for its immutable version.
 
+        ``secrets`` must contain exactly the names declared by
+        ``@udf(secrets=[...])``. Values are sent in the create request and
+        stored server-side in the private execution artifact; returned
+        Function and Job metadata contain only the declared names.
         This is the blocking counterpart of :meth:`create_function_async`.
         Local connections raise ``NotImplementedError``.
         """
-        return self.create_function_async(definition).wait()
+        return self.create_function_async(definition, secrets=secrets).wait()
 
-    def create_function_async(self, definition: UdfDefinition) -> Job[FunctionVersion]:
+    def create_function_async(
+        self,
+        definition: UdfDefinition,
+        *,
+        secrets: Optional[Mapping[str, str]] = None,
+    ) -> Job[FunctionVersion]:
         """Register a scalar Python UDF through the remote Function catalog.
 
+        ``secrets`` must contain exactly the names declared by
+        ``@udf(secrets=[...])``. Values are sent in the create request and
+        stored server-side in the private execution artifact; returned
+        Function and Job metadata contain only the declared names.
         Submission returns a typed job. The immutable Function version becomes
         available only when :meth:`Job.wait` succeeds. Local connections raise
         ``NotImplementedError``.
@@ -1405,8 +1424,13 @@ class LanceDBConnection(DBConnection):
         return Job(self._conn.job(job_id))
 
     @override
-    def create_function_async(self, definition: UdfDefinition) -> Job[FunctionVersion]:
-        job = LOOP.run(self._conn.create_function_async(definition))
+    def create_function_async(
+        self,
+        definition: UdfDefinition,
+        *,
+        secrets: Optional[Mapping[str, str]] = None,
+    ) -> Job[FunctionVersion]:
+        job = LOOP.run(self._conn.create_function_async(definition, secrets=secrets))
         return Job(job)
 
     @override
@@ -2225,17 +2249,24 @@ class AsyncConnection(object):
         return AsyncJob(self._inner.job(job_id))
 
     async def create_function_async(
-        self, definition: UdfDefinition
+        self,
+        definition: UdfDefinition,
+        *,
+        secrets: Optional[Mapping[str, str]] = None,
     ) -> AsyncJob[FunctionVersion]:
         """Register a scalar Python UDF through the remote Function catalog.
 
+        ``secrets`` must contain exactly the names declared by
+        ``@udf(secrets=[...])``. Values are sent in the create request and
+        stored server-side in the private execution artifact; returned
+        Function and Job metadata contain only the declared names.
         The returned typed job resolves to the immutable Function version.
         Local connections raise ``NotImplementedError``.
         """
         if not isinstance(definition, UdfDefinition):
             raise TypeError("create_function_async requires a @udf definition")
         inner = await self._inner.create_function_async(
-            definition.registration_request.to_canonical_json()
+            definition._submission_json(secrets)
         )
         return _typed_job(inner, FunctionVersion.from_json)
 

python/python/lancedb/functions.py

@@ -4,7 +4,7 @@
 """Canonical Function values exchanged with LanceDB Enterprise services.
 
 These immutable models contain client/wire state only. Catalog persistence,
-environment bake, and execution are owned by Sophon.
+environment bake, secret resolution, and execution are owned by Sophon.
 ``RefreshColumnResult`` is also the backend-neutral result of a local
 expression-backed refresh job.
 """
@@ -229,7 +229,7 @@ class PythonEnvironmentSpec(_RemoteValue):
 
 
 class PythonRuntimeSpec(_RemoteValue):
-    """Remote runtime definition with environment values.
+    """Remote runtime definition with non-secret environment values.
 
     V1 supports ``kind="python"``. Newer runtime kinds remain readable, while
     their unknown payload fields are intentionally not retained by the client.
@@ -268,6 +268,7 @@ class FunctionVersion(_RemoteValue):
     runtime: PythonRuntimeSpec
     runtime_digest: str
     environment_digest: str
+    required_secrets: tuple[str, ...] = ()
     created_at: str
 
     def __call__(self, **inputs: Any) -> FunctionApplication:
@@ -329,12 +330,17 @@ class FunctionVersion(_RemoteValue):
 
 
 class FunctionRegistrationRequest(_RemoteValue):
-    """Stable remote registration envelope produced by :func:`udf`."""
+    """Stable remote registration envelope produced by :func:`udf`.
+
+    Only secret names are represented. Secret values are supplied separately
+    when the definition is submitted and are not part of this durable value.
+    """
 
     name: str
     artifact: FunctionArtifactRequest
     signature: FunctionSignature
     runtime: PythonRuntimeSpec
+    required_secrets: tuple[str, ...] = ()
 
 
 class FunctionVersionRef(_OpenRemoteValue):
@@ -479,6 +485,27 @@ class RefreshColumnResult(_RemoteValue):
 
 
 _FUNCTION_NAME = re.compile(r"^[A-Za-z_][A-Za-z0-9_.-]*$")
+_SECRET_NAME = re.compile(r"^[A-Za-z_][A-Za-z0-9_]*$")
+# Keep this byte limit aligned with Sophon's MAX_FUNCTION_SECRET_VALUE_BYTES.
+_MAX_FUNCTION_SECRET_VALUE_BYTES = 64 * 1024
+_MAX_FUNCTION_SECRET_VALUES_BYTES = 512 * 1024
+
+
+def _validate_secret_value(name: str, value: Any) -> str:
+    """Validate one secret value before building the create request."""
+    if not isinstance(value, str):
+        raise TypeError(f"Function secret {name!r} value must be a string")
+    if not value:
+        raise ValueError(f"Function secret {name!r} value must be non-empty")
+    if "\0" in value:
+        raise ValueError(f"Function secret {name!r} value must not contain NUL")
+    value_bytes = len(value.encode("utf-8"))
+    if value_bytes > _MAX_FUNCTION_SECRET_VALUE_BYTES:
+        raise ValueError(
+            f"Function secret {name!r} value exceeds the "
+            f"{_MAX_FUNCTION_SECRET_VALUE_BYTES}-byte limit"
+        )
+    return value
 
 
 _GRAMMAR_PRIMITIVES = (
@@ -502,107 +529,31 @@ _GRAMMAR_PRIMITIVES = (
 
 
 def _canonical_arrow_type(data_type: pa.DataType) -> str:
-    """The compact Function grammar, or canonical exact JSON for nested types."""
-    grammar = _grammar_arrow_type(data_type)
-    if grammar is not None:
-        return grammar
-    exact = _exact_arrow_type(data_type)
-    return json.dumps(exact, ensure_ascii=False, sort_keys=True, separators=(",", ":"))
-
-
-def _grammar_arrow_type(data_type: pa.DataType) -> Optional[str]:
+    """The server's V1 Function type grammar. Anything outside it is rejected
+    here rather than at registration."""
     for candidate, name in _GRAMMAR_PRIMITIVES:
         if data_type == candidate:
             return name
     if pa.types.is_list(data_type) or pa.types.is_large_list(data_type):
-        item = _grammar_list_item(data_type)
-        if item is None:
-            return None
         prefix = "list" if pa.types.is_list(data_type) else "large_list"
-        return f"{prefix}<{item}>"
+        return f"{prefix}<{_canonical_list_item(data_type)}>"
     if pa.types.is_fixed_size_list(data_type) and data_type.list_size > 0:
-        item = _grammar_list_item(data_type)
-        if item is not None:
-            return f"fixed_size_list<{item}, {data_type.list_size}>"
-    return None
+        return (
+            f"fixed_size_list<{_canonical_list_item(data_type)}, {data_type.list_size}>"
+        )
+    raise TypeError(f"unsupported Arrow type for Function signature: {data_type}")
 
 
-def _grammar_list_item(data_type: pa.DataType) -> Optional[str]:
+def _canonical_list_item(data_type: pa.DataType) -> str:
     """The grammar names only the item type; it always means a non-nullable
-    child called `item`, so other child properties require exact JSON."""
+    child called `item`, so any other child metadata cannot be represented."""
     child = data_type.value_field
     if child.name != "item" or child.nullable or child.metadata:
-        return None
-    return _grammar_arrow_type(child.type)
-
-
-def _validate_exact_arrow_field(field: pa.Field) -> None:
-    if not field.name:
         raise TypeError(
-            "unsupported Arrow type for Function signature: field names "
-            "must not be empty"
+            "unsupported Arrow type for Function signature: list items must be a "
+            f"non-nullable field named 'item', got {child}"
         )
-    if field.metadata:
-        raise TypeError(
-            "unsupported Arrow type for Function signature: field metadata "
-            f"is not supported, got {field}"
-        )
-
-
-def _exact_arrow_field(field: pa.Field) -> dict[str, Any]:
-    _validate_exact_arrow_field(field)
-    return {
-        "name": field.name,
-        "nullable": field.nullable,
-        "type": _exact_arrow_type(field.type),
-    }
-
-
-def _exact_arrow_type(data_type: pa.DataType) -> dict[str, Any]:
-    for candidate, name in _GRAMMAR_PRIMITIVES:
-        if data_type == candidate:
-            return {"type": name}
-    if pa.types.is_struct(data_type):
-        fields = list(data_type)
-        names = [field.name for field in fields]
-        if not fields or len(set(names)) != len(names):
-            raise TypeError(
-                "unsupported Arrow type for Function signature: structs must have "
-                "non-empty, uniquely named fields"
-            )
-        return {
-            "type": "struct",
-            "fields": [_exact_arrow_field(field) for field in fields],
-        }
-    if (
-        pa.types.is_list(data_type)
-        or pa.types.is_large_list(data_type)
-        or pa.types.is_fixed_size_list(data_type)
-    ):
-        if pa.types.is_fixed_size_list(data_type):
-            if data_type.value_field.name != "item":
-                raise TypeError(
-                    "unsupported Arrow type for Function signature: fixed-size list "
-                    "items must be named 'item'"
-                )
-            if data_type.list_size <= 0:
-                raise TypeError(
-                    f"unsupported Arrow type for Function signature: {data_type}"
-                )
-        value: dict[str, Any] = {
-            "type": (
-                "list"
-                if pa.types.is_list(data_type)
-                else "large_list"
-                if pa.types.is_large_list(data_type)
-                else "fixed_size_list"
-            ),
-            "fields": [_exact_arrow_field(data_type.value_field)],
-        }
-        if pa.types.is_fixed_size_list(data_type):
-            value["length"] = data_type.list_size
-        return value
-    raise TypeError(f"unsupported Arrow type for Function signature: {data_type}")
+    return _canonical_arrow_type(child.type)
 
 
 def _list_of(item: pa.DataType) -> pa.DataType:
@@ -676,11 +627,8 @@ def _callable_parameters(function: Callable[..., Any]) -> tuple[inspect.Paramete
 
 def _function_output(output: pa.DataType | pa.Field | pa.Schema) -> FunctionOutput:
     if isinstance(output, pa.Schema):
-        if output.metadata:
-            raise TypeError("Function output schema metadata is not supported")
         fields = tuple(output)
     elif isinstance(output, pa.Field) and pa.types.is_struct(output.type):
-        _validate_exact_arrow_field(output)
         if output.nullable:
             raise ValueError("Function output must be non-nullable")
         fields = tuple(output.type)
@@ -696,7 +644,6 @@ def _function_output(output: pa.DataType | pa.Field | pa.Schema) -> FunctionOutp
             raise TypeError(
                 "output_schema must be a PyArrow DataType, Field, or Schema"
             )
-        _validate_exact_arrow_field(field)
         if field.nullable:
             raise ValueError("Function output must be non-nullable")
         return FunctionOutput(
@@ -709,8 +656,6 @@ def _function_output(output: pa.DataType | pa.Field | pa.Schema) -> FunctionOutp
         raise ValueError("named-struct Function output must contain at least one field")
     if any(field.nullable for field in fields):
         raise ValueError("Function output fields must be non-nullable")
-    for field in fields:
-        _validate_exact_arrow_field(field)
     names = [field.name for field in fields]
     if len(set(names)) != len(names):
         raise ValueError("Function output field names must be unique")
@@ -739,10 +684,6 @@ def _infer_signature(
     if input_schema is not None:
         if not isinstance(input_schema, pa.Schema):
             raise TypeError("input_schema must be a PyArrow Schema")
-        if input_schema.metadata:
-            raise TypeError("Function input schema metadata is not supported")
-        for field in input_schema:
-            _validate_exact_arrow_field(field)
         expected = tuple(parameter.name for parameter in parameters)
         actual = tuple(input_schema.names)
         if actual != expected:
@@ -995,6 +936,7 @@ class UdfDefinition:
         output_schema: Optional[pa.DataType | pa.Field | pa.Schema],
         pip: tuple[str, ...],
         env: Mapping[str, str],
+        secrets: tuple[str, ...],
         python_version: Optional[str],
         conda: tuple[str, ...] = (),
         conda_channels: tuple[str, ...] = (),
@@ -1021,6 +963,17 @@ class UdfDefinition:
             for key, value in environment.items()
         ):
             raise TypeError("Function env keys and values must be strings")
+        required_secrets = tuple(sorted(set(secrets)))
+        invalid_secrets = [
+            secret for secret in required_secrets if not _SECRET_NAME.fullmatch(secret)
+        ]
+        if invalid_secrets:
+            raise ValueError(f"invalid Function secret names: {invalid_secrets!r}")
+        overlap = set(environment) & set(required_secrets)
+        if overlap:
+            raise ValueError(
+                f"Function env and secret names must be disjoint: {sorted(overlap)!r}"
+            )
         signature = _infer_signature(function, input_schema, output_schema)
         source = _package_source(function)
         digest = f"sha256:{hashlib.sha256(source).hexdigest()}"
@@ -1049,14 +1002,65 @@ class UdfDefinition:
             ),
             signature=signature,
             runtime=runtime,
+            required_secrets=required_secrets,
         )
         functools.update_wrapper(self, function)
 
     @property
     def registration_request(self) -> FunctionRegistrationRequest:
-        """The immutable request sent by ``create_function_async``."""
+        """The immutable, value-free client model for a Function submission."""
         return self._request
 
+    def _submission_json(self, secrets: Optional[Mapping[str, str]]) -> str:
+        """Build one registration submission without retaining values on self."""
+        if secrets is None:
+            secret_values: Mapping[str, str] = {}
+        elif not isinstance(secrets, Mapping):
+            raise TypeError("Function secrets must be a mapping of names to strings")
+        else:
+            secret_values = secrets
+
+        if any(not isinstance(name, str) for name in secret_values):
+            raise TypeError("Function secret names must be strings")
+        expected = set(self._request.required_secrets)
+        provided = set(secret_values)
+        if provided != expected:
+            missing = sorted(expected - provided)
+            unexpected = sorted(provided - expected)
+            details = []
+            if missing:
+                details.append(f"missing: {missing!r}")
+            if unexpected:
+                details.append(f"unexpected: {unexpected!r}")
+            raise ValueError(
+                "Function secret values must exactly match the declared secrets ("
+                + "; ".join(details)
+                + ")"
+            )
+
+        canonical_values = {}
+        total_bytes = 0
+        for name in sorted(secret_values):
+            value = _validate_secret_value(name, secret_values[name])
+            total_bytes += len(value.encode("utf-8"))
+            if total_bytes > _MAX_FUNCTION_SECRET_VALUES_BYTES:
+                raise ValueError(
+                    "Function secret values exceed the "
+                    f"{_MAX_FUNCTION_SECRET_VALUES_BYTES}-byte request limit"
+                )
+            canonical_values[name] = value
+
+        submission = self._request._known_dict()
+        if canonical_values:
+            submission["secret_values"] = canonical_values
+        return json.dumps(
+            submission,
+            ensure_ascii=False,
+            allow_nan=False,
+            sort_keys=True,
+            separators=(",", ":"),
+        )
+
     def __call__(self, *args, **kwargs):
         return self._function(*args, **kwargs)
 
@@ -1074,6 +1078,7 @@ def udf(
     output_schema: Optional[pa.DataType | pa.Field | pa.Schema] = None,
     pip: tuple[str, ...] | list[str] = (),
     env: Optional[Mapping[str, str]] = None,
+    secrets: tuple[str, ...] | list[str] = (),
     python_version: Optional[str] = None,
     conda: tuple[str, ...] | list[str] = (),
     conda_channels: tuple[str, ...] | list[str] = (),
@@ -1088,6 +1093,7 @@ def udf(
     output_schema: Optional[pa.DataType | pa.Field | pa.Schema] = None,
     pip: tuple[str, ...] | list[str] = (),
     env: Optional[Mapping[str, str]] = None,
+    secrets: tuple[str, ...] | list[str] = (),
     python_version: Optional[str] = None,
     conda: tuple[str, ...] | list[str] = (),
     conda_channels: tuple[str, ...] | list[str] = (),
@@ -1118,7 +1124,10 @@ def udf(
     conda_channels : sequence of str, optional
         Conda channels in priority order; requires ``conda``.
     env : mapping of str to str, optional
-        Environment variables included in the Function definition.
+        Non-secret environment variables. Use ``secrets`` for credentials.
+    secrets : sequence of str, optional
+        Names of secrets required by the callable. Supply their values separately
+        to ``create_function`` or ``create_function_async``.
     python_version : str, optional
         Remote Python major/minor version. Defaults to the client version.
 
@@ -1140,11 +1149,15 @@ def udf(
     Examples
     --------
     >>> from lancedb import udf
-    >>> @udf(pip=["numpy==2.2.0"])
+    >>> @udf(pip=["numpy==2.2.0"], secrets=["MODEL_TOKEN"])
     ... def score(value: float) -> float:
     ...     return value * 2
     >>> score(1.5)
     3.0
+    >>> db.create_function(  # doctest: +SKIP
+    ...     score, secrets={"MODEL_TOKEN": "user-secret-value"}
+    ... )
+
     """
 
     def decorate(target: Callable[..., Any]) -> UdfDefinition:
@@ -1155,6 +1168,7 @@ def udf(
             output_schema=output_schema,
             pip=tuple(pip),
             env={} if env is None else env,
+            secrets=tuple(secrets),
             python_version=python_version,
             conda=tuple(conda),
             conda_channels=tuple(conda_channels),

python/python/lancedb/remote/db.py

@@ -7,7 +7,7 @@ import json
 import logging
 from concurrent.futures import ThreadPoolExecutor
 import sys
-from typing import TYPE_CHECKING, Any, Dict, Iterable, List, Optional, Union
+from typing import TYPE_CHECKING, Any, Dict, Iterable, List, Mapping, Optional, Union
 from urllib.parse import urlparse
 import warnings
 
@@ -742,8 +742,15 @@ class RemoteDBConnection(DBConnection):
         return Job(self._conn.job(job_id))
 
     @override
-    def create_function_async(self, definition: UdfDefinition) -> Job[FunctionVersion]:
-        return Job(LOOP.run(self._conn.create_function_async(definition)))
+    def create_function_async(
+        self,
+        definition: UdfDefinition,
+        *,
+        secrets: Optional[Mapping[str, str]] = None,
+    ) -> Job[FunctionVersion]:
+        return Job(
+            LOOP.run(self._conn.create_function_async(definition, secrets=secrets))
+        )
 
     @override
     def get_function(self, name: str, *, version: str) -> FunctionVersion:

python/python/lancedb/remote/table.py

@@ -67,15 +67,7 @@ from ..query import (
     LanceTakeQueryBuilder,
     LanceVectorQueryBuilder,
 )
-from ..table import (
-    AsyncTable,
-    BlobMode,
-    Branches,
-    IndexStatistics,
-    Query,
-    Table,
-    Tags,
-)
+from ..table import AsyncTable, BlobMode, Branches, IndexStatistics, Query, Table, Tags
 from ..types import BaseTokenizerType
 
 

python/python/lancedb/table.py

@@ -2165,11 +2165,9 @@ class Table(ABC):
             Function columns are supported only on LanceDB Cloud and
             Enterprise.
         computed: Dict[str, str], optional
-            A mapping from output column names to SQL expressions derives each
-            output field from its expression. A direct projection of a Blob v2
-            field inherits Blob v2 semantics; other expressions derive their
-            ordinary Arrow type. Mapping order is declaration and dependency
-            order.
+            A map of column name to a SQL expression defining the column. The
+            column's type and inputs are derived from the expression, so no
+            data type is supplied.
 
             Unlike ``transforms``, the expression is stored rather than
             evaluated now: the column is committed with no values, and rows get
@@ -6270,11 +6268,8 @@ class AsyncTable:
             Function columns are supported only on LanceDB Cloud and
             Enterprise.
         computed: Dict[str, str], optional
-            A mapping from output column names to SQL expressions derives each
-            output field from its expression. A direct projection of a Blob v2
-            field inherits Blob v2 semantics; other expressions derive their
-            ordinary Arrow type. Mapping order is declaration and dependency
-            order.
+            A map of column name to a SQL expression defining the column. The
+            column's type and inputs are derived from the expression.
 
             Unlike ``transforms``, the expression is stored rather than
             evaluated now: the column is committed with no values, and rows get

python/python/tests/test_first_class_function_slice1.py

@@ -37,6 +37,21 @@ def job_result(name: str) -> dict:
     return json.loads(fixture(name))["result"]
 
 
+def assert_no_secret_values(value):
+    if isinstance(value, dict):
+        for key, child in value.items():
+            assert key not in {
+                "secret_value",
+                "secret_values",
+                "resolved_secret",
+                "resolved_secrets",
+            }
+            assert_no_secret_values(child)
+    elif isinstance(value, list):
+        for child in value:
+            assert_no_secret_values(child)
+
+
 def test_public_function_values_are_in_api_reference():
     docs = Path(__file__).parents[3] / "docs" / "src" / "python" / "python.md"
     rendered = docs.read_text()
@@ -94,6 +109,7 @@ def test_function_version_identity_is_immutable_and_exact():
     version = FunctionVersion.from_json(json.dumps(value))
     assert version.name == "embed"
     assert version.version == "fv_01K3EXACT"
+    assert version.required_secrets == ("HF_TOKEN",)
 
     with pytest.raises((TypeError, ValueError)):
         version.version = "fv_changed"
@@ -276,6 +292,15 @@ def test_refresh_result_rejects_non_u64_values(field):
         RefreshColumnResult.from_json(json.dumps(value))
 
 
+def test_canonical_client_values_contain_secret_names_only():
+    version = FunctionVersion.from_json(
+        json.dumps(job_result("remote_function_job.json"))
+    )
+    canonical = json.loads(version.to_canonical_json())
+    assert canonical["required_secrets"] == ["HF_TOKEN"]
+    assert_no_secret_values(canonical)
+
+
 class _FunctionDeclarationInner:
     def __init__(self):
         self.calls = []

python/python/tests/test_first_class_function_slice2.py

@@ -19,7 +19,13 @@ import pyarrow as pa
 import pytest
 
 import lancedb
-from lancedb.functions import UdfDefinition, udf
+from lancedb.functions import (
+    _MAX_FUNCTION_SECRET_VALUE_BYTES,
+    _MAX_FUNCTION_SECRET_VALUES_BYTES,
+    FunctionRegistrationRequest,
+    UdfDefinition,
+    udf,
+)
 
 THRESHOLD = 20
 _CACHE = None
@@ -39,12 +45,28 @@ FIXTURES = (
 @udf(
     pip=["numpy>=2"],
     env={"MODE": "test"},
+    secrets=["API_TOKEN"],
     python_version="3.12",
 )
 def normalize_score(value: float) -> float:
     return value / 100.0
 
 
+def _assert_no_secret_values(value):
+    if isinstance(value, dict):
+        for key, child in value.items():
+            assert key not in {
+                "secret_value",
+                "secret_values",
+                "resolved_secret",
+                "resolved_secrets",
+            }
+            _assert_no_secret_values(child)
+    elif isinstance(value, list):
+        for child in value:
+            _assert_no_secret_values(child)
+
+
 def test_scalar_udf_matches_shared_registration_golden_and_remains_callable():
     assert isinstance(normalize_score, UdfDefinition)
     assert normalize_score(25.0) == 0.25
@@ -59,6 +81,8 @@ def test_scalar_udf_matches_shared_registration_golden_and_remains_callable():
         "kind": "scalar_to_arrow_batch",
         "version": 1,
     }
+    assert request["required_secrets"] == ["API_TOKEN"]
+    _assert_no_secret_values(request)
 
 
 def _run_packaged(definition, *args):
@@ -168,7 +192,7 @@ def test_udf_resolves_module_globals_before_builtins(tmp_path):
         udf(module.uses_callable_shadow)
 
 
-def test_canonical_arrow_type_prefers_the_compact_grammar():
+def test_canonical_arrow_type_is_exactly_the_grammar():
     from lancedb.functions import _GRAMMAR_PRIMITIVES, _canonical_arrow_type
 
     golden = json.loads(
@@ -181,13 +205,6 @@ def test_canonical_arrow_type_prefers_the_compact_grammar():
         case["arrow_type"] for case in golden["valid"] if "<" not in case["arrow_type"]
     ]
     assert [name for _, name in _GRAMMAR_PRIMITIVES] == primitives
-    assert _canonical_arrow_type(pa.list_(pa.field("item", pa.float32(), False))) == (
-        "list<float32>"
-    )
-    assert (
-        _canonical_arrow_type(pa.large_list(pa.field("item", pa.float32(), False)))
-        == "large_list<float32>"
-    )
     for outside in [
         pa.timestamp("us"),
         pa.decimal128(10, 2),
@@ -195,6 +212,7 @@ def test_canonical_arrow_type_prefers_the_compact_grammar():
         pa.large_binary(),
         pa.binary(4),
         pa.duration("s"),
+        pa.struct([pa.field("a", pa.int32())]),
         pa.list_(pa.float32(), 0),
         pa.list_(pa.timestamp("us")),
     ]:
@@ -378,35 +396,21 @@ def test_udf_recursion_versus_a_rebound_module_name(tmp_path):
         output_schema=None,
         pip=(),
         env={},
+        secrets=(),
         python_version=None,
     )
     with pytest.raises(ValueError, match="binds that name to another value"):
         udf(raw_fact)
 
 
-def test_canonical_arrow_type_uses_exact_json_for_list_child_properties():
+def test_canonical_arrow_type_rejects_unrepresentable_list_children():
     from lancedb.functions import _canonical_arrow_type
 
-    nullable = pa.list_(pa.float32())
-    assert json.loads(_canonical_arrow_type(nullable)) == {
-        "type": "list",
-        "fields": [
-            {
-                "name": "item",
-                "nullable": True,
-                "type": {"type": "float32"},
-            }
-        ],
-    }
-    named = pa.list_(pa.field("custom", pa.float32(), nullable=False))
-    assert json.loads(_canonical_arrow_type(named))["fields"][0]["name"] == "custom"
     for outside in [
+        pa.list_(pa.float32()),  # pyarrow default: nullable child
+        pa.list_(pa.field("custom", pa.float32(), nullable=False)),
         pa.list_(pa.field("item", pa.float32(), nullable=False, metadata={"k": "v"})),
         pa.list_(pa.field("item", pa.float32(), nullable=False), 0),
-        pa.list_(
-            pa.field("item", pa.float32(), nullable=False, metadata={"k": "v"}), 3
-        ),
-        pa.list_(pa.field("custom", pa.float32(), nullable=False), 3),
     ]:
         with pytest.raises(TypeError, match="unsupported Arrow type"):
             _canonical_arrow_type(outside)
@@ -416,29 +420,6 @@ def test_canonical_arrow_type_uses_exact_json_for_list_child_properties():
         )
         == "fixed_size_list<float32, 3>"
     )
-    fixed = json.loads(_canonical_arrow_type(pa.list_(pa.float32(), 3)))
-    assert fixed == {
-        "type": "fixed_size_list",
-        "fields": [
-            {
-                "name": "item",
-                "nullable": True,
-                "type": {"type": "float32"},
-            }
-        ],
-        "length": 3,
-    }
-    large = json.loads(_canonical_arrow_type(pa.large_list(pa.float32())))
-    assert large["type"] == "large_list"
-    assert large["fields"][0]["nullable"] is True
-
-    for invalid_struct in [
-        pa.struct([]),
-        pa.struct([pa.field("a", pa.int32()), pa.field("a", pa.int64())]),
-        pa.struct([pa.field("", pa.int32())]),
-    ]:
-        with pytest.raises(TypeError, match="unsupported Arrow type"):
-            _canonical_arrow_type(invalid_struct)
 
 
 def _calls_missing(value: int) -> int:
@@ -526,105 +507,6 @@ def test_explicit_arrow_schema_is_deterministic():
     assert signature.output.nullable is False
 
 
-def test_nested_struct_output_uses_canonical_exact_json():
-    token = pa.struct(
-        [
-            pa.field("position", pa.int32(), nullable=False),
-            pa.field("value", pa.string(), nullable=False),
-            pa.field("length", pa.int32(), nullable=False),
-        ]
-    )
-    analysis = pa.struct(
-        [
-            pa.field("normalized_text", pa.string(), nullable=False),
-            pa.field("has_content", pa.bool_(), nullable=False),
-            pa.field(
-                "metrics",
-                pa.struct(
-                    [
-                        pa.field("character_count", pa.int64(), nullable=False),
-                        pa.field("word_count", pa.int32(), nullable=False),
-                        pa.field("average_word_length", pa.float64(), nullable=False),
-                    ]
-                ),
-                nullable=False,
-            ),
-            pa.field(
-                "diagnostics",
-                pa.struct(
-                    [
-                        pa.field("status", pa.string(), nullable=False),
-                        pa.field(
-                            "normalization",
-                            pa.struct(
-                                [
-                                    pa.field("changed", pa.bool_(), nullable=False),
-                                    pa.field(
-                                        "original_length", pa.int64(), nullable=False
-                                    ),
-                                ]
-                            ),
-                            nullable=False,
-                        ),
-                    ]
-                ),
-                nullable=False,
-            ),
-            pa.field(
-                "token_preview",
-                pa.list_(pa.field("item", token, nullable=False)),
-                nullable=False,
-            ),
-        ]
-    )
-
-    @udf(
-        input_schema=pa.schema([pa.field("text", pa.string(), nullable=False)]),
-        output_schema=pa.field("analysis", analysis, nullable=False),
-    )
-    def analyze(text):
-        return {"normalized_text": text}
-
-    output = analyze.registration_request.signature.output
-    assert output.kind == "named_struct"
-    assert [field.name for field in output.fields] == [
-        "normalized_text",
-        "has_content",
-        "metrics",
-        "diagnostics",
-        "token_preview",
-    ]
-    metrics = json.loads(output.fields[2].arrow_type)
-    assert metrics == {
-        "type": "struct",
-        "fields": [
-            {
-                "name": "character_count",
-                "nullable": False,
-                "type": {"type": "int64"},
-            },
-            {
-                "name": "word_count",
-                "nullable": False,
-                "type": {"type": "int32"},
-            },
-            {
-                "name": "average_word_length",
-                "nullable": False,
-                "type": {"type": "float64"},
-            },
-        ],
-    }
-    preview = json.loads(output.fields[4].arrow_type)
-    assert preview["type"] == "list"
-    assert preview["fields"][0]["type"]["type"] == "struct"
-    assert [field["name"] for field in preview["fields"][0]["type"]["fields"]] == [
-        "position",
-        "value",
-        "length",
-    ]
-
-
 def test_annotation_and_explicit_schema_validation_fail_closed():
     with pytest.raises(TypeError, match="missing Function annotations"):
 
@@ -668,80 +550,55 @@ def test_annotation_and_explicit_schema_validation_fail_closed():
         def nullable_explicit(value):
             return value
 
-    for invalid_field in [
-        pa.field("", pa.int32(), nullable=False),
-        pa.field("result", pa.int32(), nullable=False, metadata={"k": "v"}),
-    ]:
-        with pytest.raises(TypeError, match="unsupported Arrow type"):
 
-            @udf(
-                input_schema=pa.schema([pa.field("value", pa.int64())]),
-                output_schema=pa.schema([invalid_field]),
-            )
-            def invalid_explicit_field(value):
-                return value
+def test_secret_names_are_canonical_and_disjoint_from_environment():
+    @udf(secrets=["Z_TOKEN", "A_TOKEN", "Z_TOKEN"])
+    def canonical_secrets(value: int) -> int:
+        return value
 
-    with pytest.raises(TypeError, match="unsupported Arrow type"):
+    assert canonical_secrets.registration_request.required_secrets == (
+        "A_TOKEN",
+        "Z_TOKEN",
+    )
 
-        @udf(
-            input_schema=pa.schema(
-                [pa.field("value", pa.int64(), metadata={"k": "v"})]
-            ),
-            output_schema=pa.int64(),
-        )
-        def input_field_metadata(value):
+    with pytest.raises(ValueError, match="must be disjoint"):
+
+        @udf(env={"TOKEN": "plaintext"}, secrets=["TOKEN"])
+        def overlapping(value: int) -> int:
             return value
 
-    with pytest.raises(TypeError, match="unsupported Arrow type"):
 
-        @udf(
-            input_schema=pa.schema([pa.field("value", pa.int64())]),
-            output_schema=pa.field(
-                "result", pa.int64(), nullable=False, metadata={"k": "v"}
-            ),
-        )
-        def scalar_output_field_metadata(value):
-            return value
+def test_declared_secret_api_still_requires_explicit_create_values():
+    @udf(secrets=["API_TOKEN"])
+    def declared_secret(value: int) -> int:
+        return value
 
-    struct_type = pa.struct([pa.field("value", pa.int64(), nullable=False)])
-    with pytest.raises(TypeError, match="unsupported Arrow type"):
+    with pytest.raises(ValueError, match="missing"):
+        declared_secret._submission_json(None)
+    submission = json.loads(
+        declared_secret._submission_json({"API_TOKEN": "explicit-secret"})
+    )
+    assert submission["required_secrets"] == ["API_TOKEN"]
+    assert submission["secret_values"] == {"API_TOKEN": "explicit-secret"}
 
-        @udf(
-            input_schema=pa.schema([pa.field("value", pa.int64())]),
-            output_schema=pa.field(
-                "result", struct_type, nullable=False, metadata={"k": "v"}
-            ),
-        )
-        def struct_output_field_metadata(value):
-            return {"value": value}
 
-    for input_schema, output_schema in [
-        (
-            pa.schema([pa.field("value", pa.int64())], metadata={"k": "v"}),
-            pa.int64(),
-        ),
-        (
-            pa.schema([pa.field("value", pa.int64())]),
-            pa.schema(
-                [pa.field("result", pa.int64(), nullable=False)],
-                metadata={"k": "v"},
-            ),
-        ),
-    ]:
-        with pytest.raises(TypeError, match="schema metadata"):
+def test_no_secrets_preserve_canonical_registration_shape():
+    @udf
+    def no_secrets(value: int) -> int:
+        return value
 
-            @udf(input_schema=input_schema, output_schema=output_schema)
-            def schema_metadata(value):
-                return value
+    canonical = json.loads(no_secrets.registration_request.to_canonical_json())
+    assert "required_secrets" not in canonical
+    assert json.loads(no_secrets._submission_json(None)) == canonical
 
 
 def test_local_function_catalog_operations_are_not_supported(tmp_path):
     db = lancedb.connect(tmp_path)
     message = "Function catalog operations are not supported by this database"
     with pytest.raises(NotImplementedError, match=message):
-        db.create_function(normalize_score)
+        db.create_function(normalize_score, secrets={"API_TOKEN": "value"})
     with pytest.raises(NotImplementedError, match=message):
-        db.create_function_async(normalize_score)
+        db.create_function_async(normalize_score, secrets={"API_TOKEN": "value"})
     with pytest.raises(NotImplementedError, match=message):
         db.get_function("normalize_score", version="fv_exact")
 
@@ -771,6 +628,7 @@ def _mock_remote_function_catalog():
                     "runtime": body["runtime"],
                     "runtime_digest": "sha256:runtime",
                     "environment_digest": "sha256:environment",
+                    "required_secrets": body.get("required_secrets", []),
                     "created_at": "2026-08-21T00:00:00Z",
                 }
                 response = {"job_id": "job-register"}
@@ -817,7 +675,9 @@ def test_remote_registration_job_and_exact_version_reopen_round_trip():
             host_override=host,
             client_config={"retry_config": {"retries": 0}},
         )
-        registration = db.create_function_async(normalize_score)
+        registration = db.create_function_async(
+            normalize_score, secrets={"API_TOKEN": "secret-value"}
+        )
         assert registration.id == "job-register"
         created = registration.wait()
         reopened = db.get_function("normalize_score", version=created.version)
@@ -826,9 +686,18 @@ def test_remote_registration_job_and_exact_version_reopen_round_trip():
     assert reopened.name == "normalize_score"
     assert reopened.version == "fv_exact"
     create_request = state["requests"][0][1]
-    assert create_request == json.loads(
+    expected = json.loads(normalize_score.registration_request.to_canonical_json())
+    expected["secret_values"] = {"API_TOKEN": "secret-value"}
+    assert create_request == expected
+    durable_request = FunctionRegistrationRequest.from_json(json.dumps(create_request))
+    assert not hasattr(durable_request, "secret_values")
+    assert "secret_values" not in json.loads(durable_request.to_canonical_json())
+    assert "secret_values" not in json.loads(
         normalize_score.registration_request.to_canonical_json()
     )
+    assert "secret-value" not in repr(normalize_score)
+    assert "secret-value" not in repr(normalize_score.registration_request)
+    assert not hasattr(created, "secret_values")
 
 
 def test_blocking_remote_registration_returns_function_version():
@@ -839,7 +708,9 @@ def test_blocking_remote_registration_returns_function_version():
             host_override=host,
             client_config={"retry_config": {"retries": 0}},
         )
-        created = db.create_function(normalize_score)
+        created = db.create_function(
+            normalize_score, secrets={"API_TOKEN": "blocking-secret"}
+        )
 
     assert created.name == "normalize_score"
     assert created.version == "fv_exact"
@@ -847,3 +718,121 @@ def test_blocking_remote_registration_returns_function_version():
         "/v1/functions/create",
         "/v1/jobs/describe",
     ]
+    assert state["requests"][0][1]["secret_values"] == {"API_TOKEN": "blocking-secret"}
+
+
+@pytest.mark.parametrize(
+    ("secret_values", "error_type", "message"),
+    [
+        (None, ValueError, "missing"),
+        ({}, ValueError, "missing"),
+        ({"OTHER": "value"}, ValueError, "missing.*unexpected"),
+        ({"API_TOKEN": ""}, ValueError, "non-empty"),
+        ({"API_TOKEN": "bad\0value"}, ValueError, "NUL"),
+        ({"API_TOKEN": 123}, TypeError, "must be a string"),
+        ([("API_TOKEN", "value")], TypeError, "must be a mapping"),
+    ],
+)
+def test_secret_values_are_validated_before_remote_request(
+    secret_values, error_type, message
+):
+    with _mock_remote_function_catalog() as (host, state):
+        db = lancedb.connect(
+            "db://dev",
+            api_key="fake",
+            host_override=host,
+            client_config={"retry_config": {"retries": 0}},
+        )
+        with pytest.raises(error_type, match=message):
+            db.create_function_async(normalize_score, secrets=secret_values)
+        assert state["requests"] == []
+
+
+@pytest.mark.parametrize(
+    "value",
+    [
+        "x" * _MAX_FUNCTION_SECRET_VALUE_BYTES,
+        "é" * (_MAX_FUNCTION_SECRET_VALUE_BYTES // len("é".encode("utf-8"))),
+    ],
+    ids=["ascii", "multibyte"],
+)
+def test_secret_value_accepts_exact_utf8_byte_limit(value):
+    submission = json.loads(normalize_score._submission_json({"API_TOKEN": value}))
+    assert submission["secret_values"]["API_TOKEN"] == value
+    assert len(value.encode("utf-8")) == _MAX_FUNCTION_SECRET_VALUE_BYTES
+
+
+@pytest.mark.parametrize(
+    "value",
+    [
+        "x" * (_MAX_FUNCTION_SECRET_VALUE_BYTES + 1),
+        "é" * (_MAX_FUNCTION_SECRET_VALUE_BYTES // len("é".encode("utf-8")) + 1),
+    ],
+    ids=["ascii", "multibyte"],
+)
+def test_secret_value_rejects_over_utf8_byte_limit_before_json_construction(
+    monkeypatch, value
+):
+    def fail_if_json_construction_starts(self):
+        pytest.fail("oversized secret reached JSON construction")
+
+    monkeypatch.setattr(
+        FunctionRegistrationRequest, "_known_dict", fail_if_json_construction_starts
+    )
+    with pytest.raises(ValueError, match=r"exceeds the 65536-byte limit"):
+        normalize_score._submission_json({"API_TOKEN": value})
+
+
+def test_secret_values_accept_exact_aggregate_utf8_byte_limit(monkeypatch):
+    names = tuple(f"SECRET_{index}" for index in range(8))
+    value = "é" * (_MAX_FUNCTION_SECRET_VALUE_BYTES // len("é".encode("utf-8")))
+    values = {name: value for name in names}
+    monkeypatch.setattr(
+        normalize_score,
+        "_request",
+        normalize_score._request._copy(update={"required_secrets": names}),
+    )
+
+    submission = json.loads(normalize_score._submission_json(values))
+
+    assert submission["secret_values"] == values
+    assert sum(len(item.encode("utf-8")) for item in values.values()) == (
+        _MAX_FUNCTION_SECRET_VALUES_BYTES
+    )
+
+
+def test_secret_values_reject_aggregate_over_limit_before_construction(monkeypatch):
+    names = tuple(f"SECRET_{index}" for index in range(9))
+    values = {name: "x" * _MAX_FUNCTION_SECRET_VALUE_BYTES for name in names}
+    monkeypatch.setattr(
+        normalize_score,
+        "_request",
+        normalize_score._request._copy(update={"required_secrets": names}),
+    )
+
+    def fail_if_json_construction_starts(self):
+        pytest.fail("oversized aggregate reached JSON construction")
+
+    monkeypatch.setattr(
+        FunctionRegistrationRequest, "_known_dict", fail_if_json_construction_starts
+    )
+    with pytest.raises(ValueError, match=r"exceed.*524288-byte request limit"):
+        normalize_score._submission_json(values)
+
+
+@pytest.mark.asyncio
+async def test_async_remote_registration_submits_secret_values_only_once():
+    with _mock_remote_function_catalog() as (host, state):
+        db = await lancedb.connect_async(
+            "db://dev",
+            api_key="fake",
+            host_override=host,
+            client_config={"retry_config": {"retries": 0}},
+        )
+        registration = await db.create_function_async(
+            normalize_score, secrets={"API_TOKEN": "async-secret"}
+        )
+        created = await registration.wait()
+
+    assert state["requests"][0][1]["secret_values"] == {"API_TOKEN": "async-secret"}
+    assert not hasattr(created, "secret_values")

python/python/tests/test_table.py

@@ -4087,29 +4087,6 @@ def test_computed_column_rejects_transforms_and_computed_together(tmp_path):
         table.add_columns({"a": "x + 1"}, computed={"b": "x * 2"})
 
 
-def test_computed_column_blob_projection_inherits_semantics(tmp_path):
-    schema = pa.schema([pa.field("id", pa.int64()), lancedb.blob("image")])
-    db = lancedb.connect(tmp_path)
-    table = db.create_table("computed_column_blob", schema=schema)
-    table.add(
-        [
-            {"id": 1, "image": b"hello"},
-            {"id": 2, "image": b""},
-            {"id": 3, "image": None},
-        ]
-    )
-
-    table.add_columns(computed={"image_copy": "image", "second_copy": "image_copy"})
-    assert table.refresh_column("image_copy").rows_filled == 2
-    assert table.refresh_column("second_copy").rows_filled == 2
-    assert table.blob_columns() == ["image", "image_copy", "second_copy"]
-
-    hits = table.search().with_row_id(True).limit(10).to_arrow()
-    rows = sorted(zip(hits["id"].to_pylist(), hits["_rowid"].to_pylist()))
-    copied = table.fetch_blobs("second_copy", [row_id for _, row_id in rows])
-    assert copied.to_pylist() == [b"hello", b"", None]
-
-
 @pytest.mark.asyncio
 async def test_computed_column_async(tmp_path):
     db = await lancedb.connect_async(tmp_path)

rust/lancedb/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb"
-version = "0.38.0-beta.13"
+version = "0.38.0-beta.11"
 edition.workspace = true
 description = "LanceDB: A serverless, low-latency vector database for AI applications"
 license.workspace = true

rust/lancedb/src/function.rs

@@ -5,9 +5,9 @@
 //! backend-neutral terminal result of a computed-column refresh.
 //!
 //! This module contains client/wire values only. Catalog persistence,
-//! environment bake, and execution are owned by Sophon.
+//! environment bake, secret resolution, and execution are owned by Sophon.
 
-use std::collections::BTreeMap;
+use std::collections::{BTreeMap, BTreeSet};
 
 use serde::de::{self, DeserializeOwned};
 use serde::{Deserialize, Deserializer, Serialize, Serializer};
@@ -15,6 +15,16 @@ use serde_json::Value;
 
 use crate::{Error, Result};
 
+// Keep these byte limits aligned with Sophon's Function submission validation.
+pub(crate) const MAX_FUNCTION_SECRET_VALUE_BYTES: usize = 64 * 1024;
+const MAX_FUNCTION_SECRET_VALUES_BYTES: usize = 512 * 1024;
+
+fn is_portable_environment_name(name: &str) -> bool {
+    let mut bytes = name.bytes();
+    matches!(bytes.next(), Some(b'A'..=b'Z' | b'a'..=b'z' | b'_'))
+        && bytes.all(|byte| matches!(byte, b'A'..=b'Z' | b'a'..=b'z' | b'0'..=b'9' | b'_'))
+}
+
 fn invalid_json(error: impl std::fmt::Display) -> Error {
     Error::InvalidInput {
         message: format!("invalid remote Function JSON: {error}"),
@@ -198,6 +208,11 @@ pub struct PythonEnvironmentSpec {
 }
 
 /// Reproducible Python runtime definition understood by Sophon.
+///
+/// `env` contains non-secret values. Secret values are submission-only in the
+/// client model and do not become part of this public runtime identity;
+/// [`FunctionVersion::required_secrets`] contains names only. Sophon persists
+/// submitted values separately in the private execution artifact.
 #[derive(Debug, Clone, PartialEq, Eq)]
 #[non_exhaustive]
 pub enum PythonRuntimeSpec {
@@ -239,7 +254,7 @@ impl PythonRuntimeSpec {
         }
     }
 
-    /// Environment variables, or `None` for an unknown kind.
+    /// Non-secret environment variables, or `None` for an unknown kind.
     pub fn env(&self) -> Option<&BTreeMap<String, String>> {
         match self {
             Self::Python { env, .. } => Some(env),
@@ -324,6 +339,8 @@ pub struct FunctionVersion {
     runtime: PythonRuntimeSpec,
     runtime_digest: String,
     environment_digest: String,
+    #[serde(default, skip_serializing_if = "Vec::is_empty")]
+    required_secrets: Vec<String>,
     created_at: String,
 }
 
@@ -356,6 +373,12 @@ impl FunctionVersion {
         &self.environment_digest
     }
 
+    /// Required secret names. Resolved values exist only in Sophon's private
+    /// execution artifact and worker launch path.
+    pub fn required_secrets(&self) -> &[String] {
+        &self.required_secrets
+    }
+
     pub fn created_at(&self) -> &str {
         &self.created_at
     }
@@ -397,12 +420,115 @@ pub struct FunctionArtifactRequest {
 }
 
 /// Stable request envelope for remote immutable Function registration.
-#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+///
+/// Secret values are submission-only in the client model. Sophon persists them
+/// in the database-scoped private execution artifact; returned
+/// [`FunctionVersion`] and Job metadata contain only
+/// [`Self::required_secrets`] names. Debug formatting always redacts values.
+#[derive(Clone, PartialEq, Eq, Serialize, Deserialize)]
 pub struct FunctionRegistrationRequest {
     pub name: String,
     pub artifact: FunctionArtifactRequest,
     pub signature: FunctionSignature,
     pub runtime: PythonRuntimeSpec,
+    #[serde(default, skip_serializing_if = "Vec::is_empty")]
+    pub required_secrets: Vec<String>,
+    #[serde(default, skip_serializing_if = "BTreeMap::is_empty")]
+    pub secret_values: BTreeMap<String, String>,
+}
+
+impl FunctionRegistrationRequest {
+    pub(crate) fn validate_secret_values(&self) -> Result<()> {
+        let mut required = BTreeSet::new();
+        for name in &self.required_secrets {
+            if !is_portable_environment_name(name) {
+                return Err(Error::InvalidInput {
+                    message: format!(
+                        "Function secret name {name:?} must be a portable environment variable name"
+                    ),
+                });
+            }
+            if !required.insert(name) {
+                return Err(Error::InvalidInput {
+                    message: format!("Function required_secrets contains duplicate name {name:?}"),
+                });
+            }
+        }
+
+        if let PythonRuntimeSpec::Python { env, .. } = &self.runtime
+            && let Some(name) = required.iter().find(|name| env.contains_key(**name))
+        {
+            return Err(Error::InvalidInput {
+                message: format!(
+                    "Function runtime env and secret names must be disjoint: {name:?}"
+                ),
+            });
+        }
+
+        let provided = self.secret_values.keys().collect::<BTreeSet<_>>();
+        if required != provided {
+            return Err(Error::InvalidInput {
+                message: "Function secret_values keys must exactly match required_secrets"
+                    .to_string(),
+            });
+        }
+
+        let mut total_bytes = 0usize;
+        for (name, value) in &self.secret_values {
+            if value.is_empty() {
+                return Err(Error::InvalidInput {
+                    message: format!("Function secret {name:?} value must be non-empty"),
+                });
+            }
+            if value.contains('\0') {
+                return Err(Error::InvalidInput {
+                    message: format!("Function secret {name:?} value must not contain NUL"),
+                });
+            }
+            if value.len() > MAX_FUNCTION_SECRET_VALUE_BYTES {
+                return Err(Error::InvalidInput {
+                    message: format!(
+                        "Function secret {name:?} value exceeds the \
+                         {MAX_FUNCTION_SECRET_VALUE_BYTES}-byte limit"
+                    ),
+                });
+            }
+            total_bytes =
+                total_bytes
+                    .checked_add(value.len())
+                    .ok_or_else(|| Error::InvalidInput {
+                        message: "Function secret values exceed the request byte limit".to_string(),
+                    })?;
+        }
+        if total_bytes > MAX_FUNCTION_SECRET_VALUES_BYTES {
+            return Err(Error::InvalidInput {
+                message: format!(
+                    "Function secret values exceed the \
+                     {MAX_FUNCTION_SECRET_VALUES_BYTES}-byte request limit"
+                ),
+            });
+        }
+        Ok(())
+    }
+}
+
+impl std::fmt::Debug for FunctionRegistrationRequest {
+    fn fmt(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        let secret_values = self
+            .secret_values
+            .keys()
+            .map(|name| (name, "[REDACTED]"))
+            .collect::<BTreeMap<_, _>>();
+        formatter
+            .debug_struct("FunctionRegistrationRequest")
+            .field("name", &self.name)
+            .field("artifact", &self.artifact)
+            .field("signature", &self.signature)
+            .field("runtime", &self.runtime)
+            .field("required_secrets", &self.required_secrets)
+            .field("secret_values", &secret_values)
+            .finish()
+    }
 }
 
 impl_json!(FunctionRegistrationRequest);
@@ -587,6 +713,185 @@ impl RefreshColumnResult {
 
 impl_json!(RefreshColumnResult);
 
+#[cfg(test)]
+mod secret_value_tests {
+    use super::{
+        FunctionRegistrationRequest, MAX_FUNCTION_SECRET_VALUE_BYTES,
+        MAX_FUNCTION_SECRET_VALUES_BYTES, PythonRuntimeSpec,
+    };
+    use crate::Error;
+
+    fn request() -> FunctionRegistrationRequest {
+        FunctionRegistrationRequest::from_json(include_str!(
+            "../tests/fixtures/first_class_functions/v1/remote_function_registration_request.json"
+        ))
+        .unwrap()
+    }
+
+    #[test]
+    fn validates_secret_name_and_value_invariants() {
+        let missing = request();
+        assert!(matches!(
+            missing.validate_secret_values(),
+            Err(Error::InvalidInput { message }) if message.contains("exactly match")
+        ));
+
+        let mut empty = request();
+        empty
+            .secret_values
+            .insert("API_TOKEN".to_string(), String::new());
+        assert!(matches!(
+            empty.validate_secret_values(),
+            Err(Error::InvalidInput { message }) if message.contains("non-empty")
+        ));
+
+        let mut nul = request();
+        nul.secret_values
+            .insert("API_TOKEN".to_string(), "before\0after".to_string());
+        assert!(matches!(
+            nul.validate_secret_values(),
+            Err(Error::InvalidInput { message }) if message.contains("NUL")
+        ));
+
+        let mut unexpected = request();
+        unexpected
+            .secret_values
+            .insert("OTHER".to_string(), "value".to_string());
+        assert!(matches!(
+            unexpected.validate_secret_values(),
+            Err(Error::InvalidInput { message }) if message.contains("exactly match")
+        ));
+    }
+
+    #[test]
+    fn rejects_invalid_duplicate_and_overlapping_secret_declarations() {
+        let mut invalid_name = request();
+        invalid_name.required_secrets = vec!["BAD=NAME".to_string()];
+        invalid_name
+            .secret_values
+            .insert("BAD=NAME".to_string(), "secret".to_string());
+
+        let mut duplicate = request();
+        duplicate.required_secrets = vec!["API_TOKEN".to_string(), "API_TOKEN".to_string()];
+        duplicate
+            .secret_values
+            .insert("API_TOKEN".to_string(), "secret".to_string());
+
+        let mut overlap = request();
+        overlap
+            .secret_values
+            .insert("API_TOKEN".to_string(), "secret".to_string());
+        if let PythonRuntimeSpec::Python { env, .. } = &mut overlap.runtime {
+            env.insert("API_TOKEN".to_string(), "public".to_string());
+        }
+
+        assert!(matches!(
+            invalid_name.validate_secret_values(),
+            Err(Error::InvalidInput { message }) if message.contains("portable environment variable")
+        ));
+        assert!(matches!(
+            duplicate.validate_secret_values(),
+            Err(Error::InvalidInput { message }) if message.contains("duplicate")
+        ));
+        assert!(matches!(
+            overlap.validate_secret_values(),
+            Err(Error::InvalidInput { message }) if message.contains("must be disjoint")
+        ));
+    }
+
+    #[test]
+    fn enforces_portable_secret_name_boundaries() {
+        for name in ["A", "_", "A0_"] {
+            let mut request = request();
+            request.required_secrets = vec![name.to_string()];
+            request
+                .secret_values
+                .insert(name.to_string(), "secret".to_string());
+            request.validate_secret_values().unwrap();
+        }
+
+        for name in ["", "0TOKEN", "BAD-NAME", "TÖKEN"] {
+            let mut request = request();
+            request.required_secrets = vec![name.to_string()];
+            request
+                .secret_values
+                .insert(name.to_string(), "secret".to_string());
+            assert!(matches!(
+                request.validate_secret_values(),
+                Err(Error::InvalidInput { message })
+                    if message.contains("portable environment variable")
+            ));
+        }
+    }
+
+    #[test]
+    fn accepts_exact_secret_value_utf8_byte_limit() {
+        for value in [
+            "x".repeat(MAX_FUNCTION_SECRET_VALUE_BYTES),
+            "é".repeat(MAX_FUNCTION_SECRET_VALUE_BYTES / "é".len()),
+        ] {
+            assert_eq!(value.len(), MAX_FUNCTION_SECRET_VALUE_BYTES);
+            let mut request = request();
+            request.secret_values.insert("API_TOKEN".to_string(), value);
+            request.validate_secret_values().unwrap();
+        }
+    }
+
+    #[test]
+    fn rejects_secret_value_over_utf8_byte_limit() {
+        for value in [
+            "x".repeat(MAX_FUNCTION_SECRET_VALUE_BYTES + 1),
+            "é".repeat(MAX_FUNCTION_SECRET_VALUE_BYTES / "é".len() + 1),
+        ] {
+            assert!(value.len() > MAX_FUNCTION_SECRET_VALUE_BYTES);
+            let mut request = request();
+            request.secret_values.insert("API_TOKEN".to_string(), value);
+            assert!(matches!(
+                request.validate_secret_values(),
+                Err(Error::InvalidInput { message }) if message.contains("65536-byte limit")
+            ));
+        }
+    }
+
+    #[test]
+    fn rejects_aggregate_secret_value_bytes_over_server_limit() {
+        let mut request = request();
+        request.required_secrets = (0..9).map(|index| format!("SECRET_{index}")).collect();
+        request.secret_values = request
+            .required_secrets
+            .iter()
+            .map(|name| (name.clone(), "x".repeat(MAX_FUNCTION_SECRET_VALUE_BYTES)))
+            .collect();
+
+        assert!(matches!(
+            request.validate_secret_values(),
+            Err(Error::InvalidInput { message })
+                if message.contains(&format!("{MAX_FUNCTION_SECRET_VALUES_BYTES}-byte request limit"))
+        ));
+    }
+
+    #[test]
+    fn accepts_exact_aggregate_secret_value_byte_limit() {
+        let mut request = request();
+        request.required_secrets = (0..8).map(|index| format!("SECRET_{index}")).collect();
+        request.secret_values = request
+            .required_secrets
+            .iter()
+            .map(|name| (name.clone(), "x".repeat(MAX_FUNCTION_SECRET_VALUE_BYTES)))
+            .collect();
+
+        assert_eq!(
+            request
+                .secret_values
+                .values()
+                .map(String::len)
+                .sum::<usize>(),
+            MAX_FUNCTION_SECRET_VALUES_BYTES
+        );
+        request.validate_secret_values().unwrap();
+    }
+}
+
 #[cfg(test)]
 mod conda_environment_tests {
     use super::PythonEnvironmentSpec;

rust/lancedb/src/remote/client.rs

@@ -7,6 +7,7 @@ use reqwest::{
     Body, Request, RequestBuilder, Response,
     header::{HeaderMap, HeaderValue},
 };
+use serde_json::Value;
 use std::{collections::HashMap, future::Future, str::FromStr, sync::Arc, time::Duration};
 
 use crate::error::{Error, Result};
@@ -14,6 +15,60 @@ use crate::remote::db::RemoteOptions;
 use crate::remote::retry::{ResolvedRetryConfig, RetryCounter};
 
 const REQUEST_ID_HEADER: HeaderName = HeaderName::from_static("x-request-id");
+const REDACTED_JSON_VALUE: &str = "[REDACTED]";
+const SUPPRESSED_JSON_BODY: &str = "[JSON BODY SUPPRESSED]";
+
+fn is_sensitive_json_field(name: &str) -> bool {
+    name.to_ascii_lowercase().contains("secret")
+}
+
+fn redact_sensitive_json_fields(value: &mut Value) {
+    match value {
+        Value::Object(fields) => {
+            for (name, child) in fields {
+                if is_sensitive_json_field(name) {
+                    *child = Value::String(REDACTED_JSON_VALUE.to_string());
+                } else {
+                    redact_sensitive_json_fields(child);
+                }
+            }
+        }
+        Value::Array(values) => values.iter_mut().for_each(redact_sensitive_json_fields),
+        _ => {}
+    }
+}
+
+fn redacted_json_body(request: &Request) -> Option<String> {
+    let body = request.body()?.as_bytes()?;
+    let mut value = serde_json::from_slice(body).ok()?;
+    redact_sensitive_json_fields(&mut value);
+    serde_json::to_string(&value).ok()
+}
+
+fn request_log_message(request: &Request, request_id: &str) -> String {
+    let prefix = format!(
+        "Sending request_id={}: {} {}",
+        request_id,
+        request.method(),
+        request.url()
+    );
+    let content_type = request
+        .headers()
+        .get("content-type")
+        .and_then(|value| value.to_str().ok())
+        .and_then(|value| value.split(';').next());
+    if content_type.is_some_and(|value| value.eq_ignore_ascii_case("application/json")) {
+        // Never format the raw Request here: its Debug representation is not a
+        // redaction boundary and may include the original body. If the JSON body
+        // cannot be structurally parsed, suppress it instead of logging raw bytes.
+        let body = redacted_json_body(request).unwrap_or_else(|| SUPPRESSED_JSON_BODY.to_string());
+        format!("{prefix} with body {body}")
+    } else {
+        // Method and URL are sufficient request context. Raw Request formatting
+        // may expose headers or a non-JSON body, so it is never a logging fallback.
+        prefix
+    }
+}
 
 /// Configuration for TLS/mTLS settings.
 #[derive(Clone, Debug)]
@@ -839,22 +894,9 @@ impl<S: HttpSend> RestfulLanceDbClient<S> {
         }
     }
 
-    pub(crate) fn log_request(&self, request: &Request, request_id: &String) {
+    pub(crate) fn log_request(&self, request: &Request, request_id: &str) {
         if log::log_enabled!(log::Level::Debug) {
-            let content_type = request
-                .headers()
-                .get("content-type")
-                .map(|v| v.to_str().unwrap());
-            if content_type == Some("application/json") {
-                let body = request.body().as_ref().unwrap().as_bytes().unwrap();
-                let body = String::from_utf8_lossy(body);
-                debug!(
-                    "Sending request_id={}: {:?} with body {}",
-                    request_id, request, body
-                );
-            } else {
-                debug!("Sending request_id={}: {:?}", request_id, request);
-            }
+            debug!("{}", request_log_message(request, request_id));
         }
     }
 
@@ -1077,6 +1119,49 @@ mod tests {
         ENV_MUTEX.lock().unwrap_or_else(|e| e.into_inner())
     }
 
+    #[test]
+    fn test_request_log_message_redacts_secrets_and_never_formats_raw_requests() {
+        const SECRET_SENTINEL: &str = "udf-secret-log-sentinel-7e4e";
+        const MALFORMED_SENTINEL: &str = "malformed-secret-log-sentinel-b652";
+        const NON_JSON_SENTINEL: &str = "non-json-secret-log-sentinel-7fd1";
+
+        let request = reqwest::Client::new()
+            .post("https://example.com/v1/functions/create")
+            .json(&serde_json::json!({
+                "name": "uses_secret",
+                "nested": {
+                    "secret_values": {"OPENAI_API_KEY": SECRET_SENTINEL},
+                    "safe": "visible-value"
+                }
+            }))
+            .build()
+            .unwrap();
+        let log_message = request_log_message(&request, "valid-json");
+
+        let malformed_request = reqwest::Client::new()
+            .post("https://example.com/v1/functions/create")
+            .header("content-type", "application/json; charset=utf-8")
+            .body(format!(r#"{{"secret_values":"{MALFORMED_SENTINEL}""#))
+            .build()
+            .unwrap();
+        let malformed_log_message = request_log_message(&malformed_request, "malformed-json");
+
+        let non_json_request = reqwest::Client::new()
+            .post("https://example.com/v1/functions/create")
+            .header("content-type", "text/plain")
+            .body(NON_JSON_SENTINEL)
+            .build()
+            .unwrap();
+        let non_json_log_message = request_log_message(&non_json_request, "non-json");
+
+        assert!(log_message.contains("visible-value"));
+        assert!(log_message.contains(REDACTED_JSON_VALUE));
+        assert!(!log_message.contains(SECRET_SENTINEL));
+        assert!(malformed_log_message.contains(SUPPRESSED_JSON_BODY));
+        assert!(!malformed_log_message.contains(MALFORMED_SENTINEL));
+        assert!(!non_json_log_message.contains(NON_JSON_SENTINEL));
+    }
+
     #[test]
     fn test_timeout_config_default() {
         let config = TimeoutConfig::default();

rust/lancedb/src/remote/db.rs

@@ -554,6 +554,7 @@ impl<S: HttpSend> Database for RemoteDatabase<S> {
         &self,
         request: FunctionRegistrationRequest,
     ) -> Result<Job<FunctionVersion>> {
+        request.validate_secret_values()?;
         let req = self.client.post("/v1/functions/create").json(&request);
         let (request_id, response) = self.client.send(req).await?;
         let response = self.client.check_response(&request_id, response).await?;
@@ -2642,7 +2643,8 @@ mod tests {
         );
         const FUNCTION_JOB: &str =
             include_str!("../../tests/fixtures/first_class_functions/v1/remote_function_job.json");
-        let expected: serde_json::Value = serde_json::from_str(REQUEST).unwrap();
+        let mut expected: serde_json::Value = serde_json::from_str(REQUEST).unwrap();
+        expected["secret_values"] = serde_json::json!({"API_TOKEN": "secret-value"});
         let conn = Connection::new_with_handler(move |request| match request.url().path() {
             "/v1/functions/create" => {
                 assert_eq!(request.method(), &reqwest::Method::POST);
@@ -2660,7 +2662,10 @@ mod tests {
                 .unwrap(),
             path => panic!("unexpected path: {path}"),
         });
-        let request = crate::function::FunctionRegistrationRequest::from_json(REQUEST).unwrap();
+        let mut request = crate::function::FunctionRegistrationRequest::from_json(REQUEST).unwrap();
+        request
+            .secret_values
+            .insert("API_TOKEN".to_string(), "secret-value".to_string());
         let job = conn.create_function_async(request).await.unwrap();
         assert_eq!(job.id(), Some("job-function-1"));
         let version = job.wait().await.unwrap();
@@ -2668,6 +2673,32 @@ mod tests {
         assert_eq!(version.version(), "fv_01K3EXACT");
     }
 
+    #[tokio::test]
+    async fn test_create_function_async_validates_secrets_before_serialization_and_send() {
+        const REQUEST: &str = include_str!(
+            "../../tests/fixtures/first_class_functions/v1/remote_function_registration_request.json"
+        );
+        let sends = Arc::new(AtomicUsize::new(0));
+        let sends_ref = sends.clone();
+        let conn = Connection::new_with_handler(move |_| {
+            sends_ref.fetch_add(1, Ordering::SeqCst);
+            http::Response::builder().status(500).body("").unwrap()
+        });
+        let mut request = crate::function::FunctionRegistrationRequest::from_json(REQUEST).unwrap();
+        request.secret_values.insert(
+            "API_TOKEN".to_string(),
+            "x".repeat(crate::function::MAX_FUNCTION_SECRET_VALUE_BYTES + 1),
+        );
+
+        let error = conn.create_function_async(request).await.unwrap_err();
+
+        assert!(matches!(
+            error,
+            Error::InvalidInput { message } if message.contains("65536-byte limit")
+        ));
+        assert_eq!(sends.load(Ordering::SeqCst), 0);
+    }
+
     #[tokio::test]
     async fn test_get_function_requires_and_sends_exact_version() {
         const VERSION: &str = include_str!(

rust/lancedb/src/remote/table.rs

@@ -3180,8 +3180,8 @@ impl<S: HttpSend> BaseTable for RemoteTable<S> {
             self.schema().await?.as_ref(),
             "schema evolution",
         )?;
-        // The server plans the declaration against its table schema, including
-        // Blob v2 semantics inherited by a direct field projection.
+        // The server plans the declaration: expression validation, type
+        // inference and the persisted binding all happen there.
         let entries = columns
             .iter()
             .map(
@@ -5734,8 +5734,9 @@ mod tests {
             ))
             .execute()
             .await;
-        let Err(err) = result else {
-            panic!("legacy remote query unexpectedly succeeded")
+        let err = match result {
+            Ok(_) => panic!("legacy remote query unexpectedly succeeded"),
+            Err(err) => err,
         };
 
         assert!(
@@ -7387,8 +7388,8 @@ mod tests {
         assert_eq!(result.version, if old_server { 0 } else { 43 });
     }
 
-    /// A declaration is sent as `{name, computed}` for the server to plan; the
-    /// client never types the expression itself.
+    /// A declaration is sent as `{name, computed}` entries for the server to
+    /// plan; the client never types the expression itself.
     #[tokio::test]
     async fn test_add_computed_columns_sends_the_expression() {
         let table = Table::new_with_handler("my_table", |request| match request.url().path() {

rust/lancedb/src/table.rs

@@ -750,8 +750,8 @@ pub trait BaseTable: std::fmt::Display + std::fmt::Debug + Send + Sync {
     /// Declare computed columns, each defined by a SQL expression.
     ///
     /// Where the declaration is planned depends on the backend: a local table
-    /// validates and types the expression itself, while a remote one sends the
-    /// expression for the server to plan.
+    /// validates and types the expression itself, a remote one sends the text
+    /// for the server to plan.
     async fn add_computed_columns(
         &self,
         _columns: &[(String, String)],
@@ -5763,13 +5763,6 @@ mod tests {
         assert!(index_bytes > 0);
         assert_eq!(with_index, data_only + index_bytes);
 
-        // Release builds reject unstable overlay datasets unless explicitly opted in.
-        if !lance_table::feature_flags::can_read_dataset(
-            lance_table::feature_flags::FLAG_UNSTABLE_DATA_OVERLAY_FILES,
-        ) {
-            return;
-        }
-
         // Commit an overlay file supplying new `foo` values for the first three
         // rows of fragment 0. There is no high-level API that writes overlays
         // yet, so write the overlay's data file and commit the `DataOverlay`

rust/lancedb/src/table/computed_columns.rs

@@ -9,35 +9,29 @@
 //! refresh fills the rows.
 //!
 //! The rule is tagged by kind ([`ComputedColumnKind`]) because kinds differ in
-//! where the column's type and inputs come from. A SQL expression determines
-//! its inputs and physical result type. A direct projection of a Blob v2 field
-//! also inherits that field's semantic type while execution continues to use
-//! `LargeBinary`. A kind resolved through a registry cannot be typed without
-//! consulting it.
-//! Registered Functions use an exact remote version plus a schema-level
-//! Function binding; unknown newer kinds remain readable and fail closed
-//! before mutation.
+//! where the column's type and inputs come from. A SQL expression is
+//! self-describing -- both are derived from the expression, so a caller writes
+//! neither -- while a kind resolved through a registry cannot be typed without
+//! consulting it. Registered Functions use an exact remote version plus a
+//! schema-level Function binding; unknown newer kinds remain readable and fail
+//! closed before mutation.
 //!
 //! [`computed_columns`] and [`computed_column_from_field`] read declarations
 //! back off a schema.
 
-use std::collections::{BTreeSet, HashMap, HashSet};
+use std::collections::{BTreeSet, HashMap};
 use std::sync::Arc;
 
-use arrow_schema::{DataType, Field as ArrowField, Fields, Schema as ArrowSchema, SchemaRef};
-use datafusion_common::{ScalarValue, tree_node::TreeNode};
-use datafusion_expr::Expr;
+use arrow_schema::{DataType, Field as ArrowField, Schema as ArrowSchema, SchemaRef};
+use datafusion_common::tree_node::TreeNode;
 use datafusion_physical_plan::PhysicalExpr;
 use lance::dataset::NewColumnTransform;
-use lance_arrow::FieldExt;
-use lance_core::datatypes::{BLOB_V2_DESC_FIELD, format_field_path_minimal, parse_field_path};
 use lance_datafusion::planner::Planner;
 use lance_namespace::models::{JsonArrowDataType, JsonArrowField, JsonArrowSchema};
 use serde::{Deserialize, Serialize};
 use serde_json::Value;
 
 use crate::function::{FunctionApplication, FunctionBinding};
-use crate::utils::resolve_arrow_field_path;
 use crate::{Error, Result};
 
 /// Field metadata key marking a column as computed. The value is `"true"`.
@@ -1112,20 +1106,15 @@ pub(crate) fn ensure_no_foreign_declarations<'a>(
     fields: impl IntoIterator<Item = &'a Arc<ArrowField>>,
 ) -> Result<()> {
     for field in fields {
-        ensure_no_foreign_declaration(field)?;
-    }
-    Ok(())
-}
-
-fn ensure_no_foreign_declaration(field: &ArrowField) -> Result<()> {
-    if field.metadata().keys().any(|k| is_declaration_key(k)) {
-        return Err(Error::InvalidInput {
-            message: format!(
-                "field '{}' carries computed-column metadata; declare computed columns \
-                 with add_columns().computed()",
-                field.name()
-            ),
-        });
+        if field.metadata().keys().any(|k| is_declaration_key(k)) {
+            return Err(Error::InvalidInput {
+                message: format!(
+                    "field '{}' carries computed-column metadata; declare computed columns \
+                     with add_columns().computed()",
+                    field.name()
+                ),
+            });
+        }
     }
     Ok(())
 }
@@ -1173,154 +1162,15 @@ pub(crate) struct BoundExpression {
     /// The columns the expression names, as written; nested inputs keep
     /// their dotted path.
     pub inputs: Vec<String>,
-    /// The top-level columns evaluation reads, in physical-expression order.
-    /// A nested input appears through its root.
+    /// The top-level columns evaluation reads, in [`Self::read_schema`]
+    /// order. A nested input appears through its root.
     pub roots: Vec<String>,
+    /// The projected schema evaluation runs against.
+    pub read_schema: SchemaRef,
     /// The compiled expression.
     pub physical: Arc<dyn PhysicalExpr>,
     /// The type the expression yields.
     pub data_type: DataType,
-    /// Blob v2 leaves the scan must materialize as `LargeBinary`.
-    pub blob_paths: Vec<String>,
-    /// A directly projected Blob v2 field whose semantics the output inherits.
-    projected_blob_field: Option<ArrowField>,
-}
-
-fn is_direct_field_projection(expr: &Expr) -> bool {
-    match expr {
-        Expr::Column(_) => true,
-        Expr::ScalarFunction(function)
-            if function.name() == "get_field" && function.args.len() == 2 =>
-        {
-            is_direct_field_projection(&function.args[0])
-                && matches!(
-                    &function.args[1],
-                    Expr::Literal(ScalarValue::Utf8(Some(_)), _)
-                )
-        }
-        _ => false,
-    }
-}
-
-fn projected_blob_field(schema: &ArrowSchema, expr: &Expr) -> Result<Option<ArrowField>> {
-    if !is_direct_field_projection(expr) {
-        return Ok(None);
-    }
-    let paths = Planner::column_names_in_expr(expr);
-    let [path] = paths.as_slice() else {
-        return Ok(None);
-    };
-    let (_, field) = resolve_arrow_field_path(schema, path)?;
-    Ok(field.is_blob_v2().then_some(field))
-}
-
-fn collect_blob_paths(field: &ArrowField, parent: &[String], paths: &mut Vec<Vec<String>>) {
-    let mut path = parent.to_vec();
-    path.push(field.name().clone());
-    if field.is_blob_v2() {
-        paths.push(path);
-        return;
-    }
-    match field.data_type() {
-        DataType::Struct(children) => {
-            for child in children {
-                collect_blob_paths(child, &path, paths);
-            }
-        }
-        DataType::List(child)
-        | DataType::LargeList(child)
-        | DataType::FixedSizeList(child, _)
-        | DataType::Map(child, _) => collect_blob_paths(child, &path, paths),
-        _ => {}
-    }
-}
-
-fn schema_blob_paths(schema: &ArrowSchema) -> Vec<Vec<String>> {
-    let mut paths = Vec::new();
-    for field in schema.fields() {
-        collect_blob_paths(field, &[], &mut paths);
-    }
-    paths
-}
-
-fn transform_blob_field(
-    field: &ArrowField,
-    parent: &[String],
-    materialized: &HashSet<Vec<String>>,
-) -> ArrowField {
-    let mut path = parent.to_vec();
-    path.push(field.name().clone());
-    if field.is_blob_v2() {
-        if materialized.contains(&path) {
-            return ArrowField::new(field.name(), DataType::LargeBinary, field.is_nullable());
-        }
-        return ArrowField::new(
-            field.name(),
-            BLOB_V2_DESC_FIELD.data_type().clone(),
-            field.is_nullable(),
-        )
-        .with_metadata(BLOB_V2_DESC_FIELD.metadata().clone());
-    }
-
-    let data_type = match field.data_type() {
-        DataType::Struct(children) => DataType::Struct(
-            children
-                .iter()
-                .map(|child| Arc::new(transform_blob_field(child, &path, materialized)))
-                .collect(),
-        ),
-        DataType::List(child) => {
-            DataType::List(Arc::new(transform_blob_field(child, &path, materialized)))
-        }
-        DataType::LargeList(child) => {
-            DataType::LargeList(Arc::new(transform_blob_field(child, &path, materialized)))
-        }
-        DataType::FixedSizeList(child, size) => DataType::FixedSizeList(
-            Arc::new(transform_blob_field(child, &path, materialized)),
-            *size,
-        ),
-        DataType::Map(child, sorted) => DataType::Map(
-            Arc::new(transform_blob_field(child, &path, materialized)),
-            *sorted,
-        ),
-        _ => return field.clone(),
-    };
-    ArrowField::new(field.name(), data_type, field.is_nullable())
-        .with_metadata(field.metadata().clone())
-}
-
-fn blob_runtime_schema(schema: &ArrowSchema, materialized: &HashSet<Vec<String>>) -> SchemaRef {
-    Arc::new(ArrowSchema::new_with_metadata(
-        schema
-            .fields()
-            .iter()
-            .map(|field| Arc::new(transform_blob_field(field, &[], materialized)))
-            .collect::<Fields>(),
-        schema.metadata().clone(),
-    ))
-}
-
-fn referenced_blob_paths(schema: &ArrowSchema, inputs: &[String]) -> Result<Vec<Vec<String>>> {
-    let input_paths = inputs
-        .iter()
-        .map(|input| {
-            parse_field_path(input).map_err(|error| Error::InvalidInput {
-                message: format!("invalid computed-column input path '{input}': {error}"),
-            })
-        })
-        .collect::<Result<Vec<_>>>()?;
-    Ok(schema_blob_paths(schema)
-        .into_iter()
-        .filter(|blob_path| {
-            input_paths.iter().any(|input_path| {
-                input_path.len() <= blob_path.len()
-                    && input_path
-                        .iter()
-                        .zip(blob_path)
-                        .all(|(input, blob)| input == blob)
-            })
-        })
-        .collect())
 }
 
 /// Parse, resolve and compile `expression` against `schema`.
@@ -1335,18 +1185,10 @@ pub(crate) fn bind(schema: SchemaRef, column: &str, expression: &str) -> Result<
         message,
     };
 
-    // Blob v2 is a semantic type whose runtime expression ABI is
-    // `LargeBinary`. Parse against that ABI first so a direct Blob reference
-    // is not mistaken for its storage descriptor struct.
-    let all_blob_paths = schema_blob_paths(schema.as_ref())
-        .into_iter()
-        .collect::<HashSet<_>>();
-    let parsing_schema = blob_runtime_schema(schema.as_ref(), &all_blob_paths);
-    let planner = Planner::new(parsing_schema);
+    let planner = Planner::new(schema.clone());
     let parsed = planner
         .parse_expr(expression)
         .map_err(|e| invalid(e.to_string()))?;
-    let projected_blob_field = projected_blob_field(schema.as_ref(), &parsed)?;
 
     // A declaration is evaluated more than once -- staging and writing are
     // separate passes, and a refresh years later replays the same text -- so
@@ -1376,19 +1218,13 @@ pub(crate) fn bind(schema: SchemaRef, column: &str, expression: &str) -> Result<
     inputs.sort();
     inputs.dedup();
 
-    let blob_paths = referenced_blob_paths(schema.as_ref(), &inputs)?;
-    let runtime_schema = blob_runtime_schema(
-        schema.as_ref(),
-        &blob_paths.iter().cloned().collect::<HashSet<_>>(),
-    );
-
     // A nested input is recorded by its path but read through its root
     // column; Schema::index_of resolves top-level names only. Resolved here
     // rather than left to the planner so an unknown column names itself in
     // the error instead of surfacing as a plan failure.
     let mut indices = Vec::with_capacity(inputs.len());
     for input in &inputs {
-        let index = runtime_schema
+        let index = schema
             .index_of(root(input))
             .map_err(|_| invalid(format!("unknown column '{input}'")))?;
         if !indices.contains(&index) {
@@ -1401,7 +1237,7 @@ pub(crate) fn bind(schema: SchemaRef, column: &str, expression: &str) -> Result<
     // compiles the expression has to be built on the projected schema
     // evaluation will actually read.
     let read_schema = Arc::new(
-        runtime_schema
+        schema
             .project(&indices)
             .map_err(|e| invalid(e.to_string()))?,
     );
@@ -1411,8 +1247,7 @@ pub(crate) fn bind(schema: SchemaRef, column: &str, expression: &str) -> Result<
         .map(|field| field.name().clone())
         .collect();
 
-    let runtime_planner = Planner::new(runtime_schema);
-    let optimized = runtime_planner
+    let optimized = planner
         .optimize_expr(parsed)
         .map_err(|e| invalid(e.to_string()))?;
     let physical = Planner::new(read_schema.clone())
@@ -1425,16 +1260,9 @@ pub(crate) fn bind(schema: SchemaRef, column: &str, expression: &str) -> Result<
     Ok(BoundExpression {
         inputs,
         roots,
+        read_schema,
         physical,
         data_type,
-        blob_paths: blob_paths
-            .iter()
-            .map(|path| {
-                let segments = path.iter().map(String::as_str).collect::<Vec<_>>();
-                format_field_path_minimal(&segments)
-            })
-            .collect(),
-        projected_blob_field,
     })
 }
 
@@ -1445,23 +1273,18 @@ pub(crate) fn bind(schema: SchemaRef, column: &str, expression: &str) -> Result<
 /// refresh time: that the expression parses, that every column it reads
 /// exists, and that the target name is free. A declaration that survives this
 /// is one a refresh can always act on.
-///
-/// Each accepted column joins the schema the next one resolves against, so a
-/// batch may declare `a` and then `b = a + 1` in one commit. Refresh order
-/// then matters, and refresh enforces it: `b` is refused while `a` still has
-/// unfilled rows.
-fn plan_declarations(schema: SchemaRef, columns: &[(String, String)]) -> Result<Vec<ArrowField>> {
+pub(crate) fn plan(schema: SchemaRef, columns: &[(String, String)]) -> Result<Vec<ArrowField>> {
     if columns.is_empty() {
         return Err(Error::InvalidInput {
             message: "at least one computed column is required".into(),
         });
     }
 
-    let mut schema = schema;
     let mut fields = Vec::with_capacity(columns.len());
+    let mut declared: Vec<&str> = Vec::with_capacity(columns.len());
 
     for (name, expression) in columns {
-        if schema.field_with_name(name).is_ok() {
+        if schema.field_with_name(name).is_ok() || declared.contains(&name.as_str()) {
             return Err(Error::ColumnAlreadyExists { name: name.clone() });
         }
 
@@ -1469,65 +1292,16 @@ fn plan_declarations(schema: SchemaRef, columns: &[(String, String)]) -> Result<
 
         // Declared columns start entirely null, so nullability is a property
         // of the declaration rather than of what the expression yields.
-        let computed_metadata = computed_column_metadata(expression, &bound.inputs);
-        let field = match bound.projected_blob_field {
-            Some(source) => {
-                let mut metadata = source.metadata().clone();
-                metadata.retain(|key, _| !is_declaration_key(key));
-                metadata.extend(computed_metadata);
-                source
-                    .with_name(name)
-                    .with_nullable(true)
-                    .with_metadata(metadata)
-            }
-            None => ArrowField::new(name, bound.data_type, true).with_metadata(computed_metadata),
-        };
-        schema = Arc::new(ArrowSchema::new_with_metadata(
-            schema
-                .fields()
-                .iter()
-                .cloned()
-                .chain(std::iter::once(Arc::new(field.clone())))
-                .collect::<Fields>(),
-            schema.metadata().clone(),
-        ));
-        fields.push(field);
+        fields.push(
+            ArrowField::new(name, bound.data_type, true)
+                .with_metadata(computed_column_metadata(expression, &bound.inputs)),
+        );
+        declared.push(name);
     }
 
     Ok(fields)
 }
 
-pub(crate) fn plan(schema: SchemaRef, columns: &[(String, String)]) -> Result<Vec<ArrowField>> {
-    plan_declarations(schema, columns)
-}
-
-/// Run the schema-level checks of
-/// [`AddColumnsBuilder::computed`](super::AddColumnsBuilder::computed) against
-/// `schema` without committing: the Function-binding guard and the planning of
-/// every declaration. For callers that stage declarations behind other work
-/// and need those rejections before any of it lands.
-///
-/// Only the schema is consulted. Declaring also refuses a table with an LSM
-/// write spec or retained SSTables; that is table state, checked at commit.
-///
-/// ```
-/// # use std::sync::Arc;
-/// # use arrow_schema::{DataType, Field, Schema};
-/// use lancedb::table::computed_columns::validate_declarations;
-///
-/// let schema = Arc::new(Schema::new(vec![Field::new("x", DataType::Int32, false)]));
-/// let declarations = vec![
-///     ("a".to_string(), "x + 1".to_string()),
-///     ("b".to_string(), "a * 2".to_string()),
-/// ];
-/// assert!(validate_declarations(schema.clone(), &declarations).is_ok());
-/// assert!(validate_declarations(schema, &[("c".into(), "random()".into())]).is_err());
-/// ```
-pub fn validate_declarations(schema: SchemaRef, columns: &[(String, String)]) -> Result<()> {
-    ensure_no_function_bindings_for_mutation(schema.as_ref(), "schema evolution")?;
-    plan(schema, columns).map(drop)
-}
-
 /// Build the transform that declares `columns` against `schema`.
 ///
 /// An all-null column is how a binding with no values yet is carried into a
@@ -1539,7 +1313,7 @@ pub(crate) fn declare(
     schema: SchemaRef,
     columns: &[(String, String)],
 ) -> Result<NewColumnTransform> {
-    let fields = plan_declarations(schema, columns)?;
+    let fields = plan(schema, columns)?;
     Ok(NewColumnTransform::AllNulls(Arc::new(ArrowSchema::new(
         fields,
     ))))
@@ -1566,22 +1340,6 @@ pub(super) async fn add_foreign_kind(table: &crate::Table, name: &str, kind: &st
 
 #[cfg(test)]
 mod tests {
-    /// The gate's reproducer: the validator applies the same schema-level
-    /// guard declaring does, so a staging caller is refused before it commits
-    /// anything else.
-    #[test]
-    fn test_validate_declarations_matches_schema_admission_barriers() {
-        let schema = Arc::new(ArrowSchema::new_with_metadata(
-            vec![ArrowField::new("x", DataType::Int32, true)],
-            HashMap::from([(
-                FUNCTION_BINDINGS_META_KEY.to_string(),
-                "not valid binding metadata".to_string(),
-            )]),
-        ));
-        let declarations = vec![("a".to_string(), "x + 1".to_string())];
-        assert!(super::validate_declarations(schema, &declarations).is_err());
-    }
-
     #[test]
     fn output_arrow_type_grammar_matches_the_shared_golden() {
         let golden: serde_json::Value = serde_json::from_str(include_str!(
@@ -1665,44 +1423,6 @@ mod tests {
         );
     }
 
-    #[test]
-    fn test_direct_blob_projection_inherits_semantics() {
-        let schema = Arc::new(ArrowSchema::new(vec![crate::blob("image", false)]));
-        let fields = plan(
-            schema,
-            &[
-                ("first".to_string(), "image".to_string()),
-                ("second".to_string(), "first".to_string()),
-            ],
-        )
-        .unwrap();
-
-        for field in &fields {
-            assert!(field.is_blob_v2());
-            assert!(field.is_nullable());
-        }
-        assert_eq!(
-            fields[1]
-                .metadata()
-                .get(EXPRESSION_META_KEY)
-                .map(String::as_str),
-            Some("first")
-        );
-    }
-
-    #[test]
-    fn test_blob_expression_transformation_does_not_inherit_semantics() {
-        let schema = Arc::new(ArrowSchema::new(vec![crate::blob("image", true)]));
-        let fields = plan(
-            schema,
-            &[("payload".to_string(), "coalesce(image, image)".to_string())],
-        )
-        .unwrap();
-
-        assert!(!fields[0].is_blob_v2());
-        assert_eq!(fields[0].data_type(), &DataType::LargeBinary);
-    }
-
     /// The binding reaches the schema only if `AllNulls` carries per-field
     /// metadata through the commit. The whole representation rests on it.
     #[tokio::test]
@@ -1862,40 +1582,6 @@ mod tests {
         assert!(declared(&table).await.is_empty());
     }
 
-    /// A batch may build on itself: one commit, and the later entry's inputs
-    /// name the earlier one.
-    #[tokio::test]
-    async fn test_a_declaration_may_read_one_declared_before_it() {
-        let table = table_with_ints("chain").await;
-        let before = table.version().await.unwrap();
-        add_computed(
-            &table,
-            &[("a".into(), "x + 1".into()), ("b".into(), "a * 2".into())],
-        )
-        .await
-        .unwrap();
-        assert_eq!(table.version().await.unwrap(), before + 1);
-        let declared = declared(&table).await;
-        assert_eq!(declared[1].name, "b");
-        assert_eq!(declared[1].inputs, vec!["a".to_string()]);
-
-        // Order is the dependency order; reading ahead is still unknown.
-        let err = add_computed(
-            &table,
-            &[("c".into(), "d + 1".into()), ("d".into(), "x + 1".into())],
-        )
-        .await
-        .unwrap_err();
-        assert!(matches!(err, Error::InvalidExpression { column, .. } if column == "c"));
-        assert!(
-            validate_declarations(
-                table.schema().await.unwrap(),
-                &[("e".into(), "random()".into())]
-            )
-            .is_err()
-        );
-    }
-
     /// A column added by an ordinary transform is materialized, not bound, so
     /// it carries no declaration to report.
     #[tokio::test]

rust/lancedb/src/table/refresh.rs

@@ -7,16 +7,6 @@
 //! therefore idempotent and does not observe input mutation -- once a row is
 //! filled, changing what the expression reads leaves the stored result alone.
 //!
-//! A column's computed inputs are filled first -- the dependency graph is
-//! walked once, each reachable column filled once in dependency order, each
-//! fill its own commit. Every fill in the pass, the requested column's
-//! included, covers only the fragments of the snapshot the pass started
-//! from: a commit may rebase over a concurrent append, and the fragment that
-//! admits carries placeholder nulls no earlier fill covered, so it waits for
-//! a later refresh rather than being read as values. Two concurrent fills of
-//! one input collide on its field in lance's conflict check, so a dependent
-//! fill can only commit over inputs that were durable when it read them.
-//!
 //! Two passes per fragment. The first scans only the unfilled live rows and
 //! evaluates the expression over them, which yields the exact fill count and
 //! decides whether the fragment is staged at all -- a fragment where nothing
@@ -29,14 +19,10 @@
 //! inputs masked to null first, so a poison value in a row nobody is filling
 //! cannot fail the refresh.
 
-use std::collections::HashSet;
 use std::sync::Arc;
 
-use arrow_array::{
-    Array, ArrayRef, BooleanArray, LargeBinaryArray, RecordBatch, RecordBatchOptions, StructArray,
-    new_null_array,
-};
-use arrow_schema::{DataType, Field as ArrowField, Schema as ArrowSchema};
+use arrow_array::{ArrayRef, BooleanArray, RecordBatch, RecordBatchOptions};
+use arrow_schema::Schema as ArrowSchema;
 use datafusion_expr::ColumnarValue;
 use futures::{Stream, StreamExt, TryStreamExt};
 use lance::Dataset;
@@ -44,7 +30,7 @@ use lance::dataset::WriteDestination;
 use lance::dataset::fragment::FileFragment;
 use lance::dataset::transaction::Operation;
 use lance_core::ROW_ID;
-use lance_core::datatypes::{BlobHandling, Schema as LanceSchema};
+use lance_core::datatypes::Schema as LanceSchema;
 use serde::{Deserialize, Serialize};
 
 use super::computed_columns::{BoundExpression, ComputedColumnKind, computed_column_from_field};
@@ -55,8 +41,7 @@ use crate::{Error, Result};
 /// The result of refreshing a computed column.
 #[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, Default)]
 pub struct RefreshColumnResult {
-    /// Rows that had a value computed, in the requested column only; inputs
-    /// filled on its behalf are not counted.
+    /// Rows that had a value computed.
     #[serde(default)]
     pub rows_filled: u64,
     /// The commit version associated with the operation.
@@ -67,7 +52,6 @@ pub struct RefreshColumnResult {
 struct RefreshExecution {
     result: RefreshColumnResult,
     source_version: u64,
-    published_version: Option<u64>,
 }
 
 /// Internal implementation of the refresh logic.
@@ -90,12 +74,7 @@ async fn execute_refresh_column_with_source(
 
     let expression = declared_expression(&dataset, column)?;
     let schema = Arc::new(ArrowSchema::from(dataset.schema()));
-    let bound = Arc::new(super::computed_columns::bind(
-        schema.clone(),
-        column,
-        &expression,
-    )?);
-    ensure_inputs_filled(&dataset, &schema, column, &bound).await?;
+    let bound = Arc::new(super::computed_columns::bind(schema, column, &expression)?);
     let field = dataset
         .schema()
         .field(column)
@@ -108,7 +87,6 @@ async fn execute_refresh_column_with_source(
         fields: vec![field.clone()],
         metadata: Default::default(),
     };
-    let output_is_blob = field.is_blob_v2();
 
     let mut rows_filled = 0u64;
     let mut replacements = Vec::new();
@@ -118,30 +96,29 @@ async fn execute_refresh_column_with_source(
             continue;
         }
         rows_filled += gained;
-        let values =
-            fill_stream(&dataset, &fragment, bound.clone(), column, output_is_blob).await?;
+        let values = fill_stream(&dataset, &fragment, bound.clone(), column).await?;
         replacements.push(fragment.write_columns(values, &column_schema).await?);
     }
 
-    let source_version = dataset.version().version;
     if replacements.is_empty() {
+        let source_version = dataset.version().version;
         return Ok(RefreshExecution {
             result: RefreshColumnResult {
                 rows_filled: 0,
                 version: source_version,
             },
             source_version,
-            published_version: None,
         });
     }
 
+    let read_version = dataset.version().version;
     // The dataset's own session, so registrations and caches survive the
     // commit being installed on the handle.
     let session = dataset.session();
     let new_dataset = Dataset::commit(
         WriteDestination::Dataset(dataset.clone()),
         Operation::DataReplacement { replacements },
-        Some(source_version),
+        Some(read_version),
         None,
         None,
         session,
@@ -156,52 +133,10 @@ async fn execute_refresh_column_with_source(
             rows_filled,
             version,
         },
-        source_version,
-        published_version: Some(version),
+        source_version: read_version,
     })
 }
 
-/// Refuse while a computed input still has rows a refresh of it would fill:
-/// read now, its placeholder null would be evaluated as a value and kept.
-async fn ensure_inputs_filled(
-    dataset: &Dataset,
-    schema: &Arc<ArrowSchema>,
-    column: &str,
-    bound: &BoundExpression,
-) -> Result<()> {
-    for input in &bound.roots {
-        let Some(declaration) = schema
-            .field_with_name(input)
-            .ok()
-            .and_then(computed_column_from_field)
-        else {
-            continue;
-        };
-        let ComputedColumnKind::Sql { expression } = &declaration.kind else {
-            return Err(Error::NotSupported {
-                message: format!(
-                    "computed column '{column}' reads '{input}', whose fill state this \
-                     refresh cannot check; refresh '{input}' first"
-                ),
-            });
-        };
-        let input_bound = super::computed_columns::bind(schema.clone(), input, expression)?;
-        let mut unfilled = 0u64;
-        for fragment in dataset.get_fragments() {
-            unfilled += count_fragment_gains(dataset, &fragment, &input_bound, input).await?;
-        }
-        if unfilled > 0 {
-            return Err(Error::InvalidInput {
-                message: format!(
-                    "computed column '{column}' reads '{input}', which has {unfilled} unfilled \
-                     rows; refresh '{input}' first"
-                ),
-            });
-        }
-    }
-    Ok(())
-}
-
 /// Run the refresh as a [`Job`] in this process.
 pub(crate) async fn execute_refresh_column_async(
     table: &NativeTable,
@@ -225,7 +160,8 @@ pub(crate) async fn execute_refresh_column_async(
             rows_failed: 0,
             rows_remaining: 0,
             source_version: execution.source_version,
-            published_version: execution.published_version,
+            published_version: (execution.result.rows_filled > 0)
+                .then_some(execution.result.version),
         })
     })))
 }
@@ -300,15 +236,12 @@ fn evaluation_batch(
     mask_out: Option<&BooleanArray>,
 ) -> lance_core::Result<RecordBatch> {
     let mut columns = Vec::with_capacity(bound.roots.len());
-    let mut fields = Vec::with_capacity(bound.roots.len());
     for name in &bound.roots {
-        let index = batch.schema_ref().index_of(name).map_err(|_| {
+        let column = batch.column_by_name(name).ok_or_else(|| {
             lance_core::Error::invalid_input(format!(
                 "refreshing a computed column read no {name} column"
             ))
         })?;
-        let column = batch.column(index);
-        fields.push(batch.schema_ref().field(index).clone());
         // Rows outside the mask must not reach the expression: a value in a
         // deleted or already-filled row can be one it would choke on.
         columns.push(match mask_out {
@@ -317,7 +250,7 @@ fn evaluation_batch(
         });
     }
     Ok(RecordBatch::try_new_with_options(
-        Arc::new(ArrowSchema::new(fields)),
+        bound.read_schema.clone(),
         columns,
         &RecordBatchOptions::new().with_row_count(Some(batch.num_rows())),
     )?)
@@ -338,99 +271,6 @@ fn evaluate(bound: &BoundExpression, batch: &RecordBatch) -> lance_core::Result<
     }
 }
 
-fn materialized_blob_ids(schema: &LanceSchema, paths: &[String]) -> Result<HashSet<u32>> {
-    paths
-        .iter()
-        .map(|path| {
-            let field = schema
-                .resolve(path)
-                .and_then(|fields| fields.last().copied())
-                .ok_or_else(|| Error::InvalidInput {
-                    message: format!("computed Blob input '{path}' no longer exists"),
-                })?;
-            if !field.is_blob_v2() {
-                return Err(Error::InvalidInput {
-                    message: format!("computed Blob input '{path}' is no longer Blob v2"),
-                });
-            }
-            u32::try_from(field.id).map_err(|_| Error::InvalidInput {
-                message: format!(
-                    "computed Blob input '{path}' has invalid field id {}",
-                    field.id
-                ),
-            })
-        })
-        .collect()
-}
-
-fn configure_blob_inputs(
-    scanner: &mut lance::dataset::scanner::Scanner,
-    schema: &LanceSchema,
-    bound: &BoundExpression,
-    extra_blob_id: Option<u32>,
-) -> Result<()> {
-    let mut ids = materialized_blob_ids(schema, &bound.blob_paths)?;
-    ids.extend(extra_blob_id);
-    scanner.blob_handling(BlobHandling::SomeBlobsBinary(ids));
-    Ok(())
-}
-
-fn blob_array_from_binary(
-    array: &ArrayRef,
-    target_field: &ArrowField,
-) -> lance_core::Result<ArrayRef> {
-    let values = array
-        .as_any()
-        .downcast_ref::<LargeBinaryArray>()
-        .ok_or_else(|| {
-            lance_core::Error::invalid_input(format!(
-                "a Blob v2 computed output produced {}, expected LargeBinary",
-                array.data_type()
-            ))
-        })?;
-    let mut builder = lance::blob::BlobArrayBuilder::new(values.len());
-    for index in 0..values.len() {
-        if values.is_null(index) {
-            builder.push_null()?;
-        } else {
-            builder.push_bytes(values.value(index))?;
-        }
-    }
-    let minimal = builder.finish()?;
-    let minimal = minimal
-        .as_any()
-        .downcast_ref::<StructArray>()
-        .ok_or_else(|| lance_core::Error::internal("Blob builder returned a non-struct array"))?;
-    let DataType::Struct(target_fields) = target_field.data_type() else {
-        return Err(lance_core::Error::invalid_input(format!(
-            "Blob v2 output field '{}' has non-struct type {}",
-            target_field.name(),
-            target_field.data_type()
-        )));
-    };
-    let columns = target_fields
-        .iter()
-        .map(|field| match field.name().as_str() {
-            "data" | "uri" => minimal
-                .column_by_name(field.name())
-                .cloned()
-                .ok_or_else(|| {
-                    lance_core::Error::internal(format!("Blob builder omitted '{}'", field.name()))
-                }),
-            "position" | "size" => Ok(new_null_array(field.data_type(), minimal.len())),
-            name => Err(lance_core::Error::invalid_input(format!(
-                "Blob v2 output field '{}' has unsupported logical child '{name}'",
-                target_field.name()
-            ))),
-        })
-        .collect::<lance_core::Result<Vec<_>>>()?;
-    Ok(Arc::new(StructArray::try_new(
-        target_fields.clone(),
-        columns,
-        minimal.nulls().cloned(),
-    )?))
-}
-
 /// How many rows of one fragment would gain a value.
 ///
 /// Scans only the unfilled live rows -- deleted rows never reach the
@@ -449,7 +289,6 @@ async fn count_fragment_gains(
         .with_row_id()
         .filter(&format!("{} IS NULL", quote_identifier(column)))?
         .project(&bound.roots)?;
-    configure_blob_inputs(&mut scanner, dataset.schema(), bound, None)?;
 
     let mut gained = 0u64;
     let mut batches = scanner.try_into_stream().await?;
@@ -471,7 +310,6 @@ async fn fill_stream(
     fragment: &FileFragment,
     bound: Arc<BoundExpression>,
     column: &str,
-    output_is_blob: bool,
 ) -> Result<impl Stream<Item = lance_core::Result<RecordBatch>> + Send + use<>> {
     let mut projection: Vec<String> = bound.roots.clone();
     projection.push(column.to_string());
@@ -481,20 +319,6 @@ async fn fill_stream(
         .with_row_id()
         .include_deleted_rows()
         .project(&projection)?;
-    let output_blob_id = output_is_blob
-        .then(|| {
-            dataset
-                .schema()
-                .field(column)
-                .and_then(|field| u32::try_from(field.id).ok())
-        })
-        .flatten();
-    configure_blob_inputs(
-        &mut scanner,
-        dataset.schema(),
-        bound.as_ref(),
-        output_blob_id,
-    )?;
 
     let projected = Arc::new(ArrowSchema::new(vec![
         ArrowSchema::from(dataset.schema())
@@ -530,11 +354,6 @@ async fn fill_stream(
 
         let computed = evaluate(&bound, &evaluation_batch(&batch, &bound, Some(&keep))?)?;
         let merged = arrow_select::zip::zip(&fill, &computed, existing)?;
-        let merged = if output_is_blob {
-            blob_array_from_binary(&merged, projected.field(0))?
-        } else {
-            merged
-        };
         Ok(RecordBatch::try_new(projected.clone(), vec![merged])?)
     }))
 }
@@ -543,12 +362,8 @@ async fn fill_stream(
 mod tests {
     use std::sync::Arc;
 
-    use arrow_array::{
-        Array, ArrayRef, Int32Array, LargeBinaryArray, RecordBatch, StructArray, record_batch,
-    };
-    use arrow_schema::Field as ArrowField;
+    use arrow_array::{Int32Array, record_batch};
     use futures::TryStreamExt;
-    use lance_core::ROW_ID;
 
     use crate::connect;
     use crate::query::{ExecutableQuery, QueryBase, Select};
@@ -569,8 +384,7 @@ mod tests {
             .version)
     }
 
-    async fn read(table: &Table, column: &str) -> Vec<Option<i64>> {
-        use arrow_array::{Array, Int64Array};
+    async fn read(table: &Table, column: &str) -> Vec<Option<i32>> {
         let batches = table
             .query()
             .select(Select::columns(&[column]))
@@ -580,19 +394,15 @@ mod tests {
             .try_collect::<Vec<_>>()
             .await
             .unwrap();
-        let mut values: Vec<Option<i64>> = batches
+        let mut values: Vec<Option<i32>> = batches
             .iter()
             .flat_map(|batch| {
-                let array = &batch[column];
-                match array.as_any().downcast_ref::<Int32Array>() {
-                    Some(ints) => ints.iter().map(|v| v.map(i64::from)).collect::<Vec<_>>(),
-                    None => array
-                        .as_any()
-                        .downcast_ref::<Int64Array>()
-                        .unwrap()
-                        .iter()
-                        .collect::<Vec<_>>(),
-                }
+                batch[column]
+                    .as_any()
+                    .downcast_ref::<Int32Array>()
+                    .unwrap()
+                    .iter()
+                    .collect::<Vec<_>>()
             })
             .collect();
         values.sort();
@@ -604,117 +414,6 @@ mod tests {
         table.add(batch).execute().await.unwrap();
     }
 
-    #[test]
-    fn test_blob_output_matches_complete_logical_field() {
-        let values: ArrayRef = Arc::new(LargeBinaryArray::from(vec![
-            Some(b"hello".as_slice()),
-            None,
-        ]));
-        let field = ArrowField::new(
-            "image",
-            lance_core::datatypes::BLOB_V2_LOGICAL_TYPE.clone(),
-            true,
-        );
-
-        let output = super::blob_array_from_binary(&values, &field).unwrap();
-        assert_eq!(output.data_type(), field.data_type());
-        let output = output.as_any().downcast_ref::<StructArray>().unwrap();
-        assert_eq!(output.column_by_name("position").unwrap().null_count(), 2);
-        assert_eq!(output.column_by_name("size").unwrap().null_count(), 2);
-    }
-
-    /// The gate's reproducer: `b = coalesce(a, 0)` refreshed before `a`
-    /// must not bake zeros from `a`'s placeholder null. It is refused, and
-    /// names the input, until `a` is filled -- after every append too.
-    #[tokio::test]
-    async fn test_dependent_refresh_refuses_an_unfilled_input() {
-        let table = table_with("dependent_refresh_order", vec![1, 2, 3]).await;
-        table
-            .add_columns()
-            .computed("a", "x + 1")
-            .computed("b", "coalesce(a, 0)")
-            .execute()
-            .await
-            .unwrap();
-
-        let err = table.refresh_column("b").await.unwrap_err();
-        assert!(
-            matches!(&err, Error::InvalidInput { message } if message.contains("refresh 'a' first")),
-            "{err}"
-        );
-        assert_eq!(read(&table, "b").await, vec![None, None, None]);
-
-        assert_eq!(table.refresh_column("a").await.unwrap().rows_filled, 3);
-        assert_eq!(table.refresh_column("b").await.unwrap().rows_filled, 3);
-        assert_eq!(read(&table, "b").await, vec![Some(2), Some(3), Some(4)]);
-
-        append(&table, vec![10]).await;
-        assert!(table.refresh_column("b").await.is_err());
-        table.refresh_column("a").await.unwrap();
-        assert_eq!(table.refresh_column("b").await.unwrap().rows_filled, 1);
-        assert_eq!(
-            table.count_rows(Some("b = 0".to_string())).await.unwrap(),
-            0
-        );
-    }
-
-    /// Names that need quoting, and a nested input, survive the trip through
-    /// declaration metadata and the dependency check: the recorded inputs
-    /// are matched by name, never re-parsed as SQL.
-    #[tokio::test]
-    async fn test_dependent_refresh_handles_awkward_column_names() {
-        use arrow_array::{Int32Array, StructArray};
-        use arrow_schema::{DataType, Field, Fields};
-
-        let conn = connect("memory://").execute().await.unwrap();
-        let age_fields = Fields::from(vec![Field::new("age", DataType::Int32, true)]);
-        let meta = StructArray::new(
-            age_fields.clone(),
-            vec![Arc::new(Int32Array::from(vec![10, 20])) as _],
-            None,
-        );
-        let schema = Arc::new(arrow_schema::Schema::new(vec![
-            Field::new("camelCase", DataType::Int32, true),
-            Field::new("with-hyphen", DataType::Int32, true),
-            Field::new("meta", DataType::Struct(age_fields), true),
-        ]));
-        let batch = arrow_array::RecordBatch::try_new(
-            schema,
-            vec![
-                Arc::new(Int32Array::from(vec![1, 2])) as _,
-                Arc::new(Int32Array::from(vec![100, 200])) as _,
-                Arc::new(meta) as _,
-            ],
-        )
-        .unwrap();
-        let table = conn
-            .create_table("awkward_names", batch)
-            .execute()
-            .await
-            .unwrap();
-
-        table
-            .add_columns()
-            .computed("y", "`camelCase` * 2")
-            .computed("z", "coalesce(y, 0) + `with-hyphen` + meta.age")
-            .execute()
-            .await
-            .unwrap();
-        let z = crate::table::computed_columns::computed_columns(
-            table.schema().await.unwrap().as_ref(),
-        )
-        .into_iter()
-        .find(|c| c.name == "z")
-        .unwrap();
-        assert_eq!(z.inputs, vec!["meta.age", "with-hyphen", "y"]);
-
-        let err = table.refresh_column("z").await.unwrap_err();
-        assert!(err.to_string().contains("refresh 'y' first"), "{err}");
-        assert_eq!(table.refresh_column("y").await.unwrap().rows_filled, 2);
-        assert_eq!(table.refresh_column("z").await.unwrap().rows_filled, 2);
-        assert_eq!(read(&table, "z").await, vec![Some(112), Some(224)]);
-    }
-
     #[tokio::test]
     async fn test_refresh_fills_a_declared_column() {
         let table = table_with("refresh_fills", vec![1, 2, 3]).await;
@@ -952,8 +651,7 @@ mod tests {
 
         let read_back = read(&table, "doubled").await;
         assert_eq!(read_back.len(), 20_000);
-        let mut expected: Vec<Option<i64>> =
-            values.iter().map(|v| Some(i64::from(v * 2))).collect();
+        let mut expected: Vec<Option<i32>> = values.iter().map(|v| Some(v * 2)).collect();
         expected.sort();
         assert_eq!(read_back, expected);
     }
@@ -1310,366 +1008,4 @@ mod tests {
         let err = table.refresh_column("embedding").await.unwrap_err();
         assert!(matches!(err, Error::NotSupported { message } if message.contains("udf")));
     }
-
-    fn blob_batch(ids: Vec<i32>, payloads: Vec<Option<&[u8]>>) -> RecordBatch {
-        use arrow_array::Int32Array;
-        use arrow_schema::{Field, Schema};
-
-        let mut builder = lance::blob::BlobArrayBuilder::new(payloads.len());
-        for payload in payloads {
-            match payload {
-                Some(payload) => builder.push_bytes(payload).unwrap(),
-                None => builder.push_null().unwrap(),
-            }
-        }
-        RecordBatch::try_new(
-            Arc::new(Schema::new(vec![
-                Field::new("id", arrow_schema::DataType::Int32, false),
-                crate::blob("image", true),
-            ])),
-            vec![Arc::new(Int32Array::from(ids)), builder.finish().unwrap()],
-        )
-        .unwrap()
-    }
-
-    async fn create_blob_table(path: &std::path::Path, batch: RecordBatch) -> Table {
-        let conn = connect(path.to_str().unwrap()).execute().await.unwrap();
-        conn.create_table("blobs", batch).execute().await.unwrap()
-    }
-
-    #[tokio::test]
-    async fn test_refresh_inherits_and_publishes_blob_output() {
-        use arrow_array::UInt64Array;
-        use lance_arrow::{
-            BLOB_DEDICATED_SIZE_THRESHOLD_META_KEY, BLOB_INLINE_SIZE_THRESHOLD_META_KEY,
-        };
-        use lance_core::datatypes::BlobKind;
-
-        use crate::table::schema_evolution::FieldMetadataUpdate;
-
-        let tmp = tempfile::tempdir().unwrap();
-        let table = create_blob_table(
-            tmp.path(),
-            blob_batch(
-                vec![1, 2, 3, 4],
-                vec![Some(b"hello"), Some(b"ab"), Some(b""), None],
-            ),
-        )
-        .await;
-        table
-            .add_columns()
-            .computed("image_copy", "image")
-            .execute()
-            .await
-            .unwrap();
-        table
-            .update_field_metadata(&[FieldMetadataUpdate::new("image_copy")
-                .set(BLOB_INLINE_SIZE_THRESHOLD_META_KEY, "1")
-                .set(BLOB_DEDICATED_SIZE_THRESHOLD_META_KEY, "4")])
-            .await
-            .unwrap();
-
-        let first_refresh = table.refresh_column("image_copy").await.unwrap();
-        assert_eq!(first_refresh.rows_filled, 3);
-        assert_eq!(
-            table.blob_columns().await.unwrap(),
-            vec!["image".to_string(), "image_copy".to_string()]
-        );
-
-        let batches = table
-            .query()
-            .with_row_id()
-            .execute()
-            .await
-            .unwrap()
-            .try_collect::<Vec<_>>()
-            .await
-            .unwrap();
-        let batch = arrow_select::concat::concat_batches(&batches[0].schema(), &batches).unwrap();
-        assert!(
-            batch
-                .column_by_name("image_copy")
-                .unwrap()
-                .as_any()
-                .is::<arrow_array::StructArray>()
-        );
-        let row_ids = batch
-            .column_by_name(ROW_ID)
-            .unwrap()
-            .as_any()
-            .downcast_ref::<UInt64Array>()
-            .unwrap()
-            .values()
-            .to_vec();
-        let original = table.fetch_blobs("image", &row_ids).await.unwrap();
-        let copied = table.fetch_blobs("image_copy", &row_ids).await.unwrap();
-        assert_eq!(original, copied);
-        let ids = batch
-            .column_by_name("id")
-            .unwrap()
-            .as_any()
-            .downcast_ref::<Int32Array>()
-            .unwrap();
-        let files = table
-            .fetch_blob_files("image_copy", &row_ids)
-            .await
-            .unwrap();
-        let mut layouts = ids
-            .values()
-            .iter()
-            .copied()
-            .zip(files)
-            .map(|(id, file)| (id, file.and_then(|file| file.kind())))
-            .collect::<Vec<_>>();
-        layouts.sort_by_key(|(id, _)| *id);
-        assert_eq!(
-            layouts,
-            vec![
-                (1, Some(BlobKind::Dedicated)),
-                (2, Some(BlobKind::Packed)),
-                (3, Some(BlobKind::Inline)),
-                (4, None),
-            ]
-        );
-
-        table
-            .add(blob_batch(vec![5], vec![Some(b"appended")]))
-            .execute()
-            .await
-            .unwrap();
-        table
-            .optimize(crate::table::OptimizeAction::Compact {
-                options: crate::table::CompactionOptions::default(),
-                remap_options: None,
-            })
-            .await
-            .unwrap();
-        assert_eq!(
-            table
-                .refresh_column("image_copy")
-                .await
-                .unwrap()
-                .rows_filled,
-            1
-        );
-        assert_eq!(
-            table
-                .refresh_column("image_copy")
-                .await
-                .unwrap()
-                .rows_filled,
-            0
-        );
-
-        table.checkout(first_refresh.version).await.unwrap();
-        assert_eq!(table.count_rows(None).await.unwrap(), 4);
-        assert_eq!(
-            table.blob_columns().await.unwrap(),
-            vec!["image".to_string(), "image_copy".to_string()]
-        );
-        table.checkout_latest().await.unwrap();
-    }
-
-    #[tokio::test]
-    async fn test_refresh_inherits_nested_struct_blob_input() {
-        use arrow_array::{Int32Array, StructArray, UInt64Array};
-        use arrow_schema::{DataType, Field, Fields, Schema};
-
-        let tmp = tempfile::tempdir().unwrap();
-        let mut blob_builder = lance::blob::BlobArrayBuilder::new(2);
-        blob_builder.push_bytes(b"nested").unwrap();
-        blob_builder.push_null().unwrap();
-        let blob_field = crate::blob("image", true);
-        let metadata_fields = Fields::from(vec![blob_field.clone()]);
-        let metadata = StructArray::new(
-            metadata_fields.clone(),
-            vec![blob_builder.finish().unwrap()],
-            None,
-        );
-        let batch = RecordBatch::try_new(
-            Arc::new(Schema::new(vec![
-                Field::new("id", DataType::Int32, false),
-                Field::new("metadata", DataType::Struct(metadata_fields), true),
-            ])),
-            vec![Arc::new(Int32Array::from(vec![1, 2])), Arc::new(metadata)],
-        )
-        .unwrap();
-        let table = create_blob_table(tmp.path(), batch).await;
-        table
-            .add_columns()
-            .computed("payload_copy", "metadata.image")
-            .execute()
-            .await
-            .unwrap();
-
-        assert_eq!(
-            table
-                .refresh_column("payload_copy")
-                .await
-                .unwrap()
-                .rows_filled,
-            1
-        );
-        assert_eq!(
-            table.blob_columns().await.unwrap(),
-            vec!["metadata.image".to_string(), "payload_copy".to_string()]
-        );
-        let batches = table
-            .query()
-            .with_row_id()
-            .execute()
-            .await
-            .unwrap()
-            .try_collect::<Vec<_>>()
-            .await
-            .unwrap();
-        let row_ids = batches[0]
-            .column_by_name(ROW_ID)
-            .unwrap()
-            .as_any()
-            .downcast_ref::<UInt64Array>()
-            .unwrap()
-            .values();
-        let payloads = table.fetch_blobs("payload_copy", row_ids).await.unwrap();
-        assert_eq!(payloads.value(0), b"nested");
-        assert!(payloads.is_null(1));
-    }
-
-    #[tokio::test]
-    async fn test_refresh_preserves_list_shape_when_materializing_blob_input() {
-        use arrow_array::{Int32Array, ListArray};
-        use arrow_buffer::{OffsetBuffer, ScalarBuffer};
-        use arrow_schema::{DataType, Field, Schema};
-
-        let tmp = tempfile::tempdir().unwrap();
-        let mut blob_builder = lance::blob::BlobArrayBuilder::new(3);
-        blob_builder.push_bytes(b"a").unwrap();
-        blob_builder.push_bytes(b"bb").unwrap();
-        blob_builder.push_null().unwrap();
-        let item = Arc::new(crate::blob("item", true));
-        let images = ListArray::new(
-            item.clone(),
-            OffsetBuffer::new(ScalarBuffer::from(vec![0, 2, 3])),
-            blob_builder.finish().unwrap(),
-            None,
-        );
-        let batch = RecordBatch::try_new(
-            Arc::new(Schema::new(vec![
-                Field::new("id", DataType::Int32, false),
-                Field::new("images", DataType::List(item), true),
-            ])),
-            vec![Arc::new(Int32Array::from(vec![1, 2])), Arc::new(images)],
-        )
-        .unwrap();
-        let table = create_blob_table(tmp.path(), batch).await;
-        table
-            .add_columns()
-            .computed("image_payloads", "images")
-            .execute()
-            .await
-            .unwrap();
-
-        assert_eq!(
-            table
-                .refresh_column("image_payloads")
-                .await
-                .unwrap()
-                .rows_filled,
-            2
-        );
-        let batches = table
-            .query()
-            .select(Select::columns(&["image_payloads"]))
-            .execute()
-            .await
-            .unwrap()
-            .try_collect::<Vec<_>>()
-            .await
-            .unwrap();
-        let output = batches[0]
-            .column_by_name("image_payloads")
-            .unwrap()
-            .as_any()
-            .downcast_ref::<ListArray>()
-            .unwrap();
-        assert_eq!(output.value_offsets(), &[0, 2, 3]);
-        assert!(output.values().as_any().is::<LargeBinaryArray>());
-    }
-
-    #[tokio::test]
-    async fn test_refresh_inherits_external_blob_input() {
-        use arrow_array::{Int32Array, StringArray, UInt64Array};
-        use arrow_schema::{DataType, Field, Schema};
-
-        let tmp = tempfile::tempdir().unwrap();
-        let payload = b"external-payload";
-        let path = tmp.path().join("payload.bin");
-        std::fs::write(&path, payload).unwrap();
-        let uri = url::Url::from_file_path(path).unwrap().to_string();
-        let conn = connect(tmp.path().join("db").to_str().unwrap())
-            .execute()
-            .await
-            .unwrap();
-        let table = conn
-            .create_empty_table(
-                "external",
-                Arc::new(Schema::new(vec![
-                    Field::new("id", DataType::Int32, false),
-                    crate::blob("image", true),
-                ])),
-            )
-            .execute()
-            .await
-            .unwrap();
-        let batch = RecordBatch::try_new(
-            Arc::new(Schema::new(vec![
-                Field::new("id", DataType::Int32, false),
-                Field::new("image", DataType::Utf8, true),
-            ])),
-            vec![
-                Arc::new(Int32Array::from(vec![1])),
-                Arc::new(StringArray::from(vec![Some(uri)])),
-            ],
-        )
-        .unwrap();
-        table
-            .add(batch)
-            .allow_external_blob_outside_bases(true)
-            .execute()
-            .await
-            .unwrap();
-        table
-            .add_columns()
-            .computed("payload_copy", "image")
-            .execute()
-            .await
-            .unwrap();
-
-        assert_eq!(
-            table
-                .refresh_column("payload_copy")
-                .await
-                .unwrap()
-                .rows_filled,
-            1
-        );
-        let batches = table
-            .query()
-            .with_row_id()
-            .execute()
-            .await
-            .unwrap()
-            .try_collect::<Vec<_>>()
-            .await
-            .unwrap();
-        let row_ids = batches[0]
-            .column_by_name(ROW_ID)
-            .unwrap()
-            .as_any()
-            .downcast_ref::<UInt64Array>()
-            .unwrap()
-            .values();
-        let payloads = table.fetch_blobs("payload_copy", row_ids).await.unwrap();
-        assert_eq!(payloads.value(0), payload);
-    }
 }

rust/lancedb/tests/first_class_function_slice1.rs

@@ -20,6 +20,25 @@ fn job_result(name: &str) -> Value {
     serde_json::from_str::<Value>(&fixture(name)).expect("remote Job fixture")["result"].clone()
 }
 
+fn assert_no_secret_values(value: &Value) {
+    match value {
+        Value::Object(values) => {
+            for (key, value) in values {
+                assert!(
+                    !matches!(
+                        key.as_str(),
+                        "secret_value" | "secret_values" | "resolved_secret" | "resolved_secrets"
+                    ),
+                    "client canonical value must not model resolved secret material"
+                );
+                assert_no_secret_values(value);
+            }
+        }
+        Value::Array(values) => values.iter().for_each(assert_no_secret_values),
+        _ => {}
+    }
+}
+
 #[test]
 fn function_version_job_result_matches_shared_canonical_golden() {
     let result = job_result("remote_function_job.json");
@@ -28,6 +47,7 @@ fn function_version_job_result_matches_shared_canonical_golden() {
     assert_eq!(version.name(), "embed");
     assert_eq!(version.version(), "fv_01K3EXACT");
     assert_eq!(version.runtime_digest(), "sha256:runtime");
+    assert_eq!(version.required_secrets(), &["HF_TOKEN"]);
     assert_eq!(
         version.to_canonical_json().expect("canonical JSON"),
         fixture("remote_function_version.canonical.json").trim()
@@ -142,3 +162,21 @@ fn floating_point_application_literals_are_rejected_consistently() {
             .contains("floating-point Function literals")
     );
 }
+
+#[test]
+fn canonical_client_values_contain_secret_names_only() {
+    let result = job_result("remote_function_job.json");
+    let version = FunctionVersion::from_json(&result.to_string()).expect("FunctionVersion result");
+    let canonical: Value = serde_json::from_str(
+        &version
+            .to_canonical_json()
+            .expect("canonical FunctionVersion"),
+    )
+    .expect("canonical JSON");
+
+    assert_eq!(
+        canonical["required_secrets"],
+        serde_json::json!(["HF_TOKEN"])
+    );
+    assert_no_secret_values(&canonical);
+}

rust/lancedb/tests/first_class_function_slice2.rs

@@ -6,6 +6,7 @@ use std::path::PathBuf;
 
 use lancedb::Error;
 use lancedb::function::FunctionRegistrationRequest;
+use serde_json::Value;
 
 fn fixture(name: &str) -> String {
     let path = PathBuf::from(env!("CARGO_MANIFEST_DIR"))
@@ -14,6 +15,25 @@ fn fixture(name: &str) -> String {
     fs::read_to_string(path).expect("fixture must be readable")
 }
 
+fn assert_no_secret_values(value: &Value) {
+    match value {
+        Value::Object(values) => {
+            for (key, value) in values {
+                assert!(
+                    !matches!(
+                        key.as_str(),
+                        "secret_value" | "secret_values" | "resolved_secret" | "resolved_secrets"
+                    ),
+                    "registration requests must not model resolved secret material"
+                );
+                assert_no_secret_values(value);
+            }
+        }
+        Value::Array(values) => values.iter().for_each(assert_no_secret_values),
+        _ => {}
+    }
+}
+
 #[test]
 fn registration_request_matches_shared_canonical_golden() {
     let request = FunctionRegistrationRequest::from_json(&fixture(
@@ -22,10 +42,33 @@ fn registration_request_matches_shared_canonical_golden() {
     .expect("registration request");
     assert_eq!(request.name, "normalize_score");
     assert_eq!(request.artifact.adapter.kind, "scalar_to_arrow_batch");
+    assert_eq!(request.required_secrets, ["API_TOKEN"]);
+    assert!(request.secret_values.is_empty());
     assert_eq!(
         request.to_canonical_json().expect("canonical request"),
         fixture("remote_function_registration_request.canonical.json").trim()
     );
+
+    let value: Value =
+        serde_json::from_str(&request.to_canonical_json().expect("canonical request"))
+            .expect("request JSON");
+    assert_no_secret_values(&value);
+}
+
+#[test]
+fn registration_request_serializes_secret_values_but_redacts_debug_output() {
+    let mut value: Value =
+        serde_json::from_str(&fixture("remote_function_registration_request.json")).unwrap();
+    value["secret_values"] = serde_json::json!({"API_TOKEN": "secret-plaintext"});
+    let request = FunctionRegistrationRequest::from_json(&value.to_string()).unwrap();
+
+    assert_eq!(request.secret_values["API_TOKEN"], "secret-plaintext");
+    let canonical = request.to_canonical_json().unwrap();
+    assert!(canonical.contains("secret-plaintext"));
+    let debug = format!("{request:?}");
+    assert!(debug.contains("API_TOKEN"));
+    assert!(debug.contains("[REDACTED]"));
+    assert!(!debug.contains("secret-plaintext"));
 }
 
 #[tokio::test]

rust/lancedb/tests/fixtures/first_class_functions/v1/remote_function_job.json

@@ -24,6 +24,7 @@
     },
     "runtime_digest": "sha256:runtime",
     "environment_digest": "sha256:environment",
+    "required_secrets": ["HF_TOKEN"],
     "created_at": "2026-08-21T00:00:00Z"
   },
   "future_job": {"trace_id": "trace-1"}

rust/lancedb/tests/fixtures/first_class_functions/v1/remote_function_registration_request.canonical.json

@@ -1 +1 @@
-{"artifact":{"adapter":{"kind":"scalar_to_arrow_batch","version":1},"content":{"data":"ZnJvbSBfX2Z1dHVyZV9fIGltcG9ydCBhbm5vdGF0aW9ucwoKZGVmIG5vcm1hbGl6ZV9zY29yZSh2YWx1ZTogZmxvYXQpIC0+IGZsb2F0OgogICAgcmV0dXJuIHZhbHVlIC8gMTAwLjAK","encoding":"base64"},"digest":"sha256:760784bdcef57b802f389b97804cc0b618aae39e86044733451bcd0b13089a7f","entrypoint":"normalize_score","kind":"python_callable"},"name":"normalize_score","runtime":{"env":{"MODE":"test"},"environment":{"kind":"pip","packages":["numpy>=2"]},"kind":"python","python_version":"3.12"},"signature":{"inputs":[{"arrow_type":"float64","name":"value","nullable":false}],"output":{"arrow_type":"float64","kind":"scalar","nullable":false}}}
+{"artifact":{"adapter":{"kind":"scalar_to_arrow_batch","version":1},"content":{"data":"ZnJvbSBfX2Z1dHVyZV9fIGltcG9ydCBhbm5vdGF0aW9ucwoKZGVmIG5vcm1hbGl6ZV9zY29yZSh2YWx1ZTogZmxvYXQpIC0+IGZsb2F0OgogICAgcmV0dXJuIHZhbHVlIC8gMTAwLjAK","encoding":"base64"},"digest":"sha256:760784bdcef57b802f389b97804cc0b618aae39e86044733451bcd0b13089a7f","entrypoint":"normalize_score","kind":"python_callable"},"name":"normalize_score","required_secrets":["API_TOKEN"],"runtime":{"env":{"MODE":"test"},"environment":{"kind":"pip","packages":["numpy>=2"]},"kind":"python","python_version":"3.12"},"signature":{"inputs":[{"arrow_type":"float64","name":"value","nullable":false}],"output":{"arrow_type":"float64","kind":"scalar","nullable":false}}}

rust/lancedb/tests/fixtures/first_class_functions/v1/remote_function_registration_request.json

@@ -39,5 +39,8 @@
     "env": {
       "MODE": "test"
     }
-  }
+  },
+  "required_secrets": [
+    "API_TOKEN"
+  ]
 }

rust/lancedb/tests/fixtures/first_class_functions/v1/remote_function_version.canonical.json

@@ -1 +1 @@
-{"artifact":{"digest":"sha256:code","entrypoint":"embed","kind":"python_callable"},"created_at":"2026-08-21T00:00:00Z","environment_digest":"sha256:environment","name":"embed","runtime":{"env":{"TOKENIZERS_PARALLELISM":"false"},"environment":{"kind":"pip","packages":["sentence-transformers>=3"]},"kind":"python","python_version":"3.12"},"runtime_digest":"sha256:runtime","signature":{"inputs":[{"arrow_type":"utf8","name":"text","nullable":true}],"output":{"arrow_type":"list<float32>","kind":"scalar","nullable":false}},"version":"fv_01K3EXACT"}
+{"artifact":{"digest":"sha256:code","entrypoint":"embed","kind":"python_callable"},"created_at":"2026-08-21T00:00:00Z","environment_digest":"sha256:environment","name":"embed","required_secrets":["HF_TOKEN"],"runtime":{"env":{"TOKENIZERS_PARALLELISM":"false"},"environment":{"kind":"pip","packages":["sentence-transformers>=3"]},"kind":"python","python_version":"3.12"},"runtime_digest":"sha256:runtime","signature":{"inputs":[{"arrow_type":"utf8","name":"text","nullable":true}],"output":{"arrow_type":"list<float32>","kind":"scalar","nullable":false}},"version":"fv_01K3EXACT"}