Check LMDB-patch with mutexes works well

Remove risk of command injection

.github/ISSUE_TEMPLATE/new_feature_issue.md

@@ -24,6 +24,11 @@ TBD
 - [ ] If not, add the `no db change` label to your PR, and you're good to merge.
 - [ ] If yes, add the `db change` label to your PR. You'll receive a message explaining you what to do.
 
+### Reminders when adding features
+
+- [ ] Write unit tests using insta
+- [ ] Write declarative integration tests in [workloads/tests](https://github.com/meilisearch/meilisearch/tree/main/workloads/test). Specify the routes to call and then call `cargo xtask test workloads/tests/YOUR_TEST.json --update-responses` so that responses are automatically filled.
+
 ### Reminders when modifying the API
 
 - [ ] Update the openAPI file with utoipa:

.github/workflows/bench-manual.yml

@@ -18,7 +18,7 @@ jobs:
     timeout-minutes: 180 # 3h
     steps:
       - uses: actions/checkout@v5
-      - uses: dtolnay/rust-toolchain@1.89
+      - uses: dtolnay/rust-toolchain@1.91.1
         with:
           profile: minimal
 

.github/workflows/bench-pr.yml

@@ -66,9 +66,7 @@ jobs:
           fetch-depth: 0 # fetch full history to be able to get main commit sha
           ref: ${{ steps.comment-branch.outputs.head_ref }}
 
-      - uses: dtolnay/rust-toolchain@1.89
-        with:
-          profile: minimal
+      - uses: dtolnay/rust-toolchain@1.91.1
 
       - name: Run benchmarks on PR ${{ github.event.issue.id }}
         run: |

.github/workflows/bench-push-indexing.yml

@@ -12,9 +12,7 @@ jobs:
     timeout-minutes: 180 # 3h
     steps:
       - uses: actions/checkout@v5
-      - uses: dtolnay/rust-toolchain@1.89
-        with:
-          profile: minimal
+      - uses: dtolnay/rust-toolchain@1.91.1
 
       # Run benchmarks
       - name: Run benchmarks - Dataset ${BENCH_NAME} - Branch main - Commit ${{ github.sha }}

.github/workflows/benchmarks-manual.yml

@@ -18,7 +18,7 @@ jobs:
     timeout-minutes: 4320 # 72h
     steps:
       - uses: actions/checkout@v5
-      - uses: dtolnay/rust-toolchain@1.89
+      - uses: dtolnay/rust-toolchain@1.91.1
         with:
           profile: minimal
 

.github/workflows/benchmarks-pr.yml

@@ -44,7 +44,7 @@ jobs:
             exit 1
           fi
 
-      - uses: dtolnay/rust-toolchain@1.89
+      - uses: dtolnay/rust-toolchain@1.91.1
         with:
           profile: minimal
 

.github/workflows/benchmarks-push-indexing.yml

@@ -16,7 +16,7 @@ jobs:
     timeout-minutes: 4320 # 72h
     steps:
       - uses: actions/checkout@v5
-      - uses: dtolnay/rust-toolchain@1.89
+      - uses: dtolnay/rust-toolchain@1.91.1
         with:
           profile: minimal
 

.github/workflows/benchmarks-push-search-geo.yml

@@ -15,7 +15,7 @@ jobs:
     runs-on: benchmarks
     steps:
       - uses: actions/checkout@v5
-      - uses: dtolnay/rust-toolchain@1.89
+      - uses: dtolnay/rust-toolchain@1.91.1
         with:
           profile: minimal
 

.github/workflows/benchmarks-push-search-songs.yml

@@ -15,7 +15,7 @@ jobs:
     runs-on: benchmarks
     steps:
       - uses: actions/checkout@v5
-      - uses: dtolnay/rust-toolchain@1.89
+      - uses: dtolnay/rust-toolchain@1.91.1
         with:
           profile: minimal
 

.github/workflows/benchmarks-push-search-wiki.yml

@@ -15,7 +15,7 @@ jobs:
     runs-on: benchmarks
     steps:
       - uses: actions/checkout@v5
-      - uses: dtolnay/rust-toolchain@1.89
+      - uses: dtolnay/rust-toolchain@1.91.1
         with:
           profile: minimal
 

.github/workflows/db-change-comments.yml

@@ -6,7 +6,7 @@ on:
 
 env:
   MESSAGE: |
-    ### Hello, I'm a bot 🤖 
+    ### Hello, I'm a bot 🤖
 
     You are receiving this message because you declared that this PR make changes to the Meilisearch database.
     Depending on the nature of the change, additional actions might be required on your part. The following sections detail the additional actions depending on the nature of the change, please copy the relevant section in the description of your PR, and make sure to perform the required actions.
@@ -19,6 +19,7 @@ env:
 
     - [ ] Detail the change to the DB format and why they are forward compatible
     - [ ] Forward-compatibility: A database created before this PR and using the features touched by this PR was able to be opened by a Meilisearch produced by the code of this PR.
+    - [ ] Declarative test: add a [declarative test containing a dumpless upgrade](https://github.com/meilisearch/meilisearch/blob/main/TESTING.md#typical-usage)
 
 
     ## This PR makes breaking changes
@@ -35,8 +36,7 @@ env:
       - [ ] Write the code to go from the old database to the new one
         - If the change happened in milli, the upgrade function should be written and called [here](https://github.com/meilisearch/meilisearch/blob/3fd86e8d76d7d468b0095d679adb09211ca3b6c0/crates/milli/src/update/upgrade/mod.rs#L24-L47)
         - If the change happened in the index-scheduler, we've never done it yet, but the right place to do it should be [here](https://github.com/meilisearch/meilisearch/blob/3fd86e8d76d7d468b0095d679adb09211ca3b6c0/crates/index-scheduler/src/scheduler/process_upgrade/mod.rs#L13)
-      - [ ] Write an integration test [here](https://github.com/meilisearch/meilisearch/blob/main/crates/meilisearch/tests/upgrade/mod.rs) ensuring you can read the old database, upgrade to the new database, and read the new database as expected
-    
+      - [ ] Declarative test: add a [declarative test containing a dumpless upgrade](https://github.com/meilisearch/meilisearch/blob/main/TESTING.md#typical-usage)
 
 jobs:
   add-comment:

.github/workflows/flaky-tests.yml

@@ -3,7 +3,7 @@ name: Look for flaky tests
 on:
   workflow_dispatch:
   schedule:
-    - cron: '0 4 * * *' # Every day at 4:00AM
+    - cron: "0 4 * * *" # Every day at 4:00AM
 
 jobs:
   flaky:
@@ -13,11 +13,17 @@ jobs:
       image: ubuntu:22.04
     steps:
       - uses: actions/checkout@v5
+      - name: Clean space as per https://github.com/actions/virtual-environments/issues/709
+        run: |
+          sudo rm -rf "/opt/ghc" || true
+          sudo rm -rf "/usr/share/dotnet" || true
+          sudo rm -rf "/usr/local/lib/android" || true
+          sudo rm -rf "/usr/local/share/boost" || true
       - name: Install needed dependencies
         run: |
           apt-get update && apt-get install -y curl
           apt-get install build-essential -y
-      - uses: dtolnay/rust-toolchain@1.89
+      - uses: dtolnay/rust-toolchain@1.91.1
       - name: Install cargo-flaky
         run: cargo install cargo-flaky
       - name: Run cargo flaky in the dumps

.github/workflows/fuzzer-indexing.yml

@@ -12,9 +12,7 @@ jobs:
     timeout-minutes: 4320 # 72h
     steps:
       - uses: actions/checkout@v5
-      - uses: dtolnay/rust-toolchain@1.89
-        with:
-          profile: minimal
+      - uses: dtolnay/rust-toolchain@1.91.1
 
       # Run benchmarks
       - name: Run the fuzzer

.github/workflows/publish-apt-brew-pkg.yml

@@ -25,7 +25,13 @@ jobs:
         run: |
           apt-get update && apt-get install -y curl
           apt-get install build-essential -y
-      - uses: dtolnay/rust-toolchain@1.89
+      - name: Clean space as per https://github.com/actions/virtual-environments/issues/709
+        run: |
+          sudo rm -rf "/opt/ghc" || true
+          sudo rm -rf "/usr/share/dotnet" || true
+          sudo rm -rf "/usr/local/lib/android" || true
+          sudo rm -rf "/usr/local/share/boost" || true
+      - uses: dtolnay/rust-toolchain@1.91.1
       - name: Install cargo-deb
         run: cargo install cargo-deb
       - uses: actions/checkout@v5

.github/workflows/publish-docker-images.yml

@@ -208,8 +208,8 @@ jobs:
           done
           cosign sign --yes ${images}
 
-      # /!\ Don't touch this without checking with Cloud team
-      - name: Send CI information to Cloud team
+      # /!\ Don't touch this without checking with engineers working on the Cloud code base on #discussion-engineering Slack channel
+      - name: Notify meilisearch-cloud 
         # Do not send if nightly build (i.e. 'schedule' or 'workflow_dispatch' event)
         if: ${{ (github.event_name == 'push') && (matrix.edition == 'enterprise') }}
         uses: peter-evans/repository-dispatch@v3
@@ -218,3 +218,14 @@ jobs:
           repository: meilisearch/meilisearch-cloud
           event-type: cloud-docker-build
           client-payload: '{ "meilisearch_version": "${{ github.ref_name }}", "stable": "${{ steps.check-tag-format.outputs.stable }}" }'
+
+      # /!\ Don't touch this without checking with integration team members on #discussion-integrations Slack channel 
+      - name: Notify meilisearch-kubernetes
+        # Do not send if nightly build (i.e. 'schedule' or 'workflow_dispatch' event), or if not stable
+        if: ${{ github.event_name == 'push' && matrix.edition == 'community' && steps.check-tag-format.outputs.stable == 'true' }}
+        uses: peter-evans/repository-dispatch@v3
+        with:
+          token: ${{ secrets.MEILI_BOT_GH_PAT }}
+          repository: meilisearch/meilisearch-kubernetes
+          event-type: meilisearch-release
+          client-payload: '{ "version": "${{ github.ref_name }}" }'

.github/workflows/publish-release-assets.yml

@@ -76,7 +76,7 @@ jobs:
     needs: check-version
     steps:
       - uses: actions/checkout@v5
-      - uses: dtolnay/rust-toolchain@1.89
+      - uses: dtolnay/rust-toolchain@1.91.1
       - name: Build
         run: cargo build --release --locked ${{ matrix.feature-flag }} ${{ matrix.extra-args }}
       # No need to upload binaries for dry run (cron or workflow_dispatch)

.github/workflows/sdks-tests.yml

@@ -25,14 +25,18 @@ jobs:
       - uses: actions/checkout@v5
       - name: Define the Docker image we need to use
         id: define-image
+        env:
+          EVENT_NAME: ${{ github.event_name }}
+          DOCKER_IMAGE_INPUT: ${{ github.event.inputs.docker_image }}
         run: |
-          event=${{ github.event_name }}
           echo "docker-image=nightly" >> $GITHUB_OUTPUT
-          if [[ $event == 'workflow_dispatch' ]]; then
-            echo "docker-image=${{ github.event.inputs.docker_image }}" >> $GITHUB_OUTPUT
+          if [[ "$EVENT_NAME" == 'workflow_dispatch' ]]; then
+            echo "docker-image=$DOCKER_IMAGE_INPUT" >> $GITHUB_OUTPUT
           fi
       - name: Docker image is ${{ steps.define-image.outputs.docker-image }}
-        run: echo "Docker image is ${{ steps.define-image.outputs.docker-image }}"
+        env:
+          DOCKER_IMAGE: ${{ steps.define-image.outputs.docker-image }}
+        run: echo "Docker image is $DOCKER_IMAGE"
 
 ##########
 ## SDKs ##

.github/workflows/test-suite.yml

@@ -19,31 +19,36 @@ jobs:
     runs-on: ${{ matrix.runner }}
     strategy:
       matrix:
-        runner: [ubuntu-24.04, ubuntu-24.04-arm]
+        runner: [ubuntu-22.04, ubuntu-22.04-arm]
         features: ["", "--features enterprise"]
-    container:
-      # Use ubuntu-22.04 to compile with glibc 2.35
-      image: ubuntu:22.04
     steps:
       - uses: actions/checkout@v5
-      - name: Install needed dependencies
+      - name: check free space before
+        run: df -h
+      - name: Clean space as per https://github.com/actions/virtual-environments/issues/709
         run: |
-          apt-get update && apt-get install -y curl
-          apt-get install build-essential -y
+          sudo rm -rf "/opt/ghc" || true
+          sudo rm -rf "/usr/share/dotnet" || true
+          sudo rm -rf "/usr/local/lib/android" || true
+          sudo rm -rf "/usr/local/share/boost" || true
+      - name: check free space after
+        run: df -h
       - name: Setup test with Rust stable
-        uses: dtolnay/rust-toolchain@1.89
+        uses: dtolnay/rust-toolchain@1.91.1
       - name: Cache dependencies
         uses: Swatinem/rust-cache@v2.8.0
-      - name: Run cargo check without any default features
+        with:
+          key: ${{ matrix.features }}
+      - name: Run cargo build without any default features
         uses: actions-rs/cargo@v1
         with:
           command: build
-          args: --locked --release --no-default-features --all
+          args: --locked --no-default-features --all
       - name: Run cargo test
         uses: actions-rs/cargo@v1
         with:
           command: test
-          args: --locked --release --all ${{ matrix.features }}
+          args: --locked --all ${{ matrix.features }}
 
   test-others:
     name: Tests on ${{ matrix.os }}
@@ -53,53 +58,56 @@ jobs:
       matrix:
         os: [macos-14, windows-2022]
         features: ["", "--features enterprise"]
+    if: github.event_name == 'schedule' || github.event_name == 'workflow_dispatch'
     steps:
       - uses: actions/checkout@v5
       - name: Cache dependencies
         uses: Swatinem/rust-cache@v2.8.0
-      - uses: dtolnay/rust-toolchain@1.89
-      - name: Run cargo check without any default features
+      - uses: dtolnay/rust-toolchain@1.91.1
+      - name: Run cargo build without any default features
         uses: actions-rs/cargo@v1
         with:
           command: build
-          args: --locked --release --no-default-features --all
+          args: --locked --no-default-features --all
       - name: Run cargo test
         uses: actions-rs/cargo@v1
         with:
           command: test
-          args: --locked --release --all ${{ matrix.features }}
+          args: --locked --all ${{ matrix.features }}
 
   test-all-features:
     name: Tests almost all features
-    runs-on: ubuntu-latest
-    container:
-      # Use ubuntu-22.04 to compile with glibc 2.35
-      image: ubuntu:22.04
+    runs-on: ubuntu-22.04
     if: github.event_name == 'schedule' || github.event_name == 'workflow_dispatch'
     steps:
       - uses: actions/checkout@v5
-      - name: Install needed dependencies
+      - name: Clean space as per https://github.com/actions/virtual-environments/issues/709
         run: |
-          apt-get update
-          apt-get install --assume-yes build-essential curl
-      - uses: dtolnay/rust-toolchain@1.89
+          sudo rm -rf "/opt/ghc" || true
+          sudo rm -rf "/usr/share/dotnet" || true
+          sudo rm -rf "/usr/local/lib/android" || true
+          sudo rm -rf "/usr/local/share/boost" || true
+      - uses: dtolnay/rust-toolchain@1.91.1
       - name: Run cargo build with almost all features
         run: |
-          cargo build --workspace --locked --release --features "$(cargo xtask list-features --exclude-feature cuda,test-ollama)"
+          cargo build --workspace --locked --features "$(cargo xtask list-features --exclude-feature cuda,test-ollama)"
       - name: Run cargo test with almost all features
         run: |
-          cargo test --workspace --locked --release --features "$(cargo xtask list-features --exclude-feature cuda,test-ollama)"
+          cargo test --workspace --locked --features "$(cargo xtask list-features --exclude-feature cuda,test-ollama)"
 
   ollama-ubuntu:
     name: Test with Ollama
-    runs-on: ubuntu-latest
-    strategy:
-      matrix:
-        features: ["", "--features enterprise"]
+    runs-on: ubuntu-22.04
     env:
       MEILI_TEST_OLLAMA_SERVER: "http://localhost:11434"
     steps:
       - uses: actions/checkout@v5
+      - name: Clean space as per https://github.com/actions/virtual-environments/issues/709
+        run: |
+          sudo rm -rf "/opt/ghc" || true
+          sudo rm -rf "/usr/share/dotnet" || true
+          sudo rm -rf "/usr/local/lib/android" || true
+          sudo rm -rf "/usr/local/share/boost" || true
       - name: Install Ollama
         run: |
           curl -fsSL https://ollama.com/install.sh | sudo -E sh
@@ -123,21 +131,21 @@ jobs:
         uses: actions-rs/cargo@v1
         with:
           command: test
-          args: --locked --release --all --features test-ollama ollama ${{ matrix.features }}
+          args: --locked -p meilisearch --features test-ollama ollama
 
   test-disabled-tokenization:
     name: Test disabled tokenization
-    runs-on: ubuntu-latest
-    container:
-      image: ubuntu:22.04
+    runs-on: ubuntu-22.04
     if: github.event_name == 'schedule' || github.event_name == 'workflow_dispatch'
     steps:
       - uses: actions/checkout@v5
-      - name: Install needed dependencies
+      - name: Clean space as per https://github.com/actions/virtual-environments/issues/709
         run: |
-          apt-get update
-          apt-get install --assume-yes build-essential curl
-      - uses: dtolnay/rust-toolchain@1.89
+          sudo rm -rf "/opt/ghc" || true
+          sudo rm -rf "/usr/share/dotnet" || true
+          sudo rm -rf "/usr/local/lib/android" || true
+          sudo rm -rf "/usr/local/share/boost" || true
+      - uses: dtolnay/rust-toolchain@1.91.1
       - name: Run cargo tree without default features and check lindera is not present
         run: |
           if cargo tree -f '{p} {f}' -e normal --no-default-features | grep -qz lindera; then
@@ -148,35 +156,39 @@ jobs:
         run: |
           cargo tree -f '{p} {f}' -e normal | grep lindera -qz
 
-  # We run tests in debug also, to make sure that the debug_assertions are hit
-  test-debug:
-    name: Run tests in debug
+  build:
+    name: Build in release
+    runs-on: ubuntu-22.04
+    steps:
+      - uses: actions/checkout@v5
+      - name: Clean space as per https://github.com/actions/virtual-environments/issues/709
+        run: |
+          sudo rm -rf "/opt/ghc" || true
+          sudo rm -rf "/usr/share/dotnet" || true
+          sudo rm -rf "/usr/local/lib/android" || true
+          sudo rm -rf "/usr/local/share/boost" || true
+      - uses: dtolnay/rust-toolchain@1.91.1
+      - name: Cache dependencies
+        uses: Swatinem/rust-cache@v2.8.0
+      - name: Build
+        run: cargo build --release --locked --target x86_64-unknown-linux-gnu
+
+  clippy:
+    name: Run Clippy
     runs-on: ubuntu-22.04
     strategy:
       matrix:
         features: ["", "--features enterprise"]
     steps:
       - uses: actions/checkout@v5
-      - uses: dtolnay/rust-toolchain@1.89
-      - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.8.0
-      - name: Run tests in debug
-        uses: actions-rs/cargo@v1
+      - name: Clean space as per https://github.com/actions/virtual-environments/issues/709
+        run: |
+          sudo rm -rf "/opt/ghc" || true
+          sudo rm -rf "/usr/share/dotnet" || true
+          sudo rm -rf "/usr/local/lib/android" || true
+          sudo rm -rf "/usr/local/share/boost" || true
+      - uses: dtolnay/rust-toolchain@1.91.1
         with:
-          command: test
-          args: --locked --all ${{ matrix.features }}
-
-  clippy:
-    name: Run Clippy
-    runs-on: ubuntu-latest
-    strategy:
-      matrix:
-        features: ["", "--features enterprise"]
-    steps:
-      - uses: actions/checkout@v5
-      - uses: dtolnay/rust-toolchain@1.89
-        with:
-          profile: minimal
           components: clippy
       - name: Cache dependencies
         uses: Swatinem/rust-cache@v2.8.0
@@ -188,14 +200,17 @@ jobs:
 
   fmt:
     name: Run Rustfmt
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-22.04
     steps:
       - uses: actions/checkout@v5
-      - uses: dtolnay/rust-toolchain@1.89
+      - name: Clean space as per https://github.com/actions/virtual-environments/issues/709
+        run: |
+          sudo rm -rf "/opt/ghc" || true
+          sudo rm -rf "/usr/share/dotnet" || true
+          sudo rm -rf "/usr/local/lib/android" || true
+          sudo rm -rf "/usr/local/share/boost" || true
+      - uses: dtolnay/rust-toolchain@1.91.1
         with:
-          profile: minimal
-          toolchain: nightly-2024-07-09
-          override: true
           components: rustfmt
       - name: Cache dependencies
         uses: Swatinem/rust-cache@v2.8.0
@@ -206,3 +221,23 @@ jobs:
         run: |
           echo -ne "\n" > crates/benchmarks/benches/datasets_paths.rs
           cargo fmt --all -- --check
+
+  declarative-tests:
+    name: Run declarative tests
+    runs-on: ubuntu-22.04-arm
+    permissions:
+      contents: read
+    steps:
+      - uses: actions/checkout@v5
+      - name: Clean space as per https://github.com/actions/virtual-environments/issues/709
+        run: |
+          sudo rm -rf "/opt/ghc" || true
+          sudo rm -rf "/usr/share/dotnet" || true
+          sudo rm -rf "/usr/local/lib/android" || true
+          sudo rm -rf "/usr/local/share/boost" || true
+      - uses: dtolnay/rust-toolchain@1.91.1
+      - name: Cache dependencies
+        uses: Swatinem/rust-cache@v2.8.0
+      - name: Run declarative tests
+        run: |
+          cargo xtask test workloads/tests/*.json

.github/workflows/update-cargo-toml-version.yml

@@ -18,9 +18,13 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v5
-      - uses: dtolnay/rust-toolchain@1.89
-        with:
-          profile: minimal
+      - name: Clean space as per https://github.com/actions/virtual-environments/issues/709
+        run: |
+          sudo rm -rf "/opt/ghc" || true
+          sudo rm -rf "/usr/share/dotnet" || true
+          sudo rm -rf "/usr/local/lib/android" || true
+          sudo rm -rf "/usr/local/share/boost" || true
+      - uses: dtolnay/rust-toolchain@1.91.1
       - name: Install sd
         run: cargo install sd
       - name: Update Cargo.toml file

BENCHMARKS.md

@@ -124,6 +124,7 @@ They are JSON files with the following structure (comments are not actually supp
 {
   // Name of the workload. Must be unique to the workload, as it will be used to group results on the dashboard.
   "name": "hackernews.ndjson_1M,no-threads",
+  "type": "bench",
   // Number of consecutive runs of the commands that should be performed.
   // Each run uses a fresh instance of Meilisearch and a fresh database.
   // Each run produces its own report file.

Cargo.lock

@@ -580,7 +580,7 @@ source = "git+https://github.com/meilisearch/bbqueue#e8af4a4bccc8eb36b2b0442c4a9
 
 [[package]]
 name = "benchmarks"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "anyhow",
  "bumpalo",
@@ -790,11 +790,11 @@ dependencies = [
 
 [[package]]
 name = "build-info"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "anyhow",
  "time",
- "vergen-git2",
+ "vergen-gitcl",
 ]
 
 [[package]]
@@ -1786,7 +1786,7 @@ dependencies = [
 
 [[package]]
 name = "dump"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "anyhow",
  "big_s",
@@ -2018,7 +2018,7 @@ checksum = "37909eebbb50d72f9059c3b6d82c0463f2ff062c9e95845c43a6c9c0355411be"
 
 [[package]]
 name = "file-store"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "tempfile",
  "thiserror 2.0.17",
@@ -2040,7 +2040,7 @@ dependencies = [
 
 [[package]]
 name = "filter-parser"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "insta",
  "levenshtein_automata",
@@ -2068,7 +2068,7 @@ dependencies = [
 
 [[package]]
 name = "flatten-serde-json"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "criterion",
  "serde_json",
@@ -2231,7 +2231,7 @@ dependencies = [
 
 [[package]]
 name = "fuzzers"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "arbitrary",
  "bumpalo",
@@ -2604,19 +2604,6 @@ version = "0.32.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "e629b9b98ef3dd8afe6ca2bd0f89306cec16d43d907889945bc5d6687f2f13c7"
 
-[[package]]
-name = "git2"
-version = "0.20.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2deb07a133b1520dc1a5690e9bd08950108873d7ed5de38dcc74d3b5ebffa110"
-dependencies = [
- "bitflags 2.10.0",
- "libc",
- "libgit2-sys",
- "log",
- "url",
-]
-
 [[package]]
 name = "glob"
 version = "0.3.3"
@@ -2711,9 +2698,9 @@ dependencies = [
 
 [[package]]
 name = "hannoy"
-version = "0.0.9-nested-rtxns-2"
+version = "0.1.0-nested-rtxns"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "06eda090938d9dcd568c8c2a5de383047ed9191578ebf4a342d2975d16e621f2"
+checksum = "be82bf3f2108ddc8885e3d306fcd7f4692066bfe26065ca8b42ba417f3c26dd1"
 dependencies = [
  "bytemuck",
  "byteorder",
@@ -2803,8 +2790,7 @@ checksum = "2304e00983f87ffb38b55b444b5e3b60a884b5d30c0fca7d82fe33449bbe55ea"
 [[package]]
 name = "heed"
 version = "0.22.1-nested-rtxns-6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c69e07cd539834bedcfa938f3d7d8520cce1ad2b0776c122b5ccdf8fd5bafe12"
+source = "git+https://github.com/meilisearch/heed?branch=allow-nested-rtxn-from-wtxn#464a9c33f76e0679115eef7379c9bdde9d5e1eda"
 dependencies = [
  "bitflags 2.10.0",
  "byteorder",
@@ -2821,14 +2807,12 @@ dependencies = [
 [[package]]
 name = "heed-traits"
 version = "0.20.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "eb3130048d404c57ce5a1ac61a903696e8fcde7e8c2991e9fcfc1f27c3ef74ff"
+source = "git+https://github.com/meilisearch/heed?branch=allow-nested-rtxn-from-wtxn#464a9c33f76e0679115eef7379c9bdde9d5e1eda"
 
 [[package]]
 name = "heed-types"
 version = "0.21.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "13c255bdf46e07fb840d120a36dcc81f385140d7191c76a7391672675c01a55d"
+source = "git+https://github.com/meilisearch/heed?branch=allow-nested-rtxn-from-wtxn#464a9c33f76e0679115eef7379c9bdde9d5e1eda"
 dependencies = [
  "bincode 1.3.3",
  "byteorder",
@@ -3198,7 +3182,7 @@ dependencies = [
 
 [[package]]
 name = "index-scheduler"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "anyhow",
  "backoff",
@@ -3460,7 +3444,7 @@ dependencies = [
 
 [[package]]
 name = "json-depth-checker"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "criterion",
  "serde_json",
@@ -3557,18 +3541,6 @@ dependencies = [
  "rle-decode-fast",
 ]
 
-[[package]]
-name = "libgit2-sys"
-version = "0.18.2+1.9.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1c42fe03df2bd3c53a3a9c7317ad91d80c81cd1fb0caec8d7cc4cd2bfa10c222"
-dependencies = [
- "cc",
- "libc",
- "libz-sys",
- "pkg-config",
-]
-
 [[package]]
 name = "libloading"
 version = "0.8.9"
@@ -3626,18 +3598,6 @@ dependencies = [
  "zlib-rs",
 ]
 
-[[package]]
-name = "libz-sys"
-version = "1.1.22"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "8b70e7a7df205e92a1a4cd9aaae7898dac0aa555503cc0a649494d0d60e7651d"
-dependencies = [
- "cc",
- "libc",
- "pkg-config",
- "vcpkg",
-]
-
 [[package]]
 name = "lindera"
 version = "0.43.3"
@@ -3842,8 +3802,7 @@ checksum = "6373607a59f0be73a39b6fe456b8192fcc3585f602af20751600e974dd455e77"
 [[package]]
 name = "lmdb-master-sys"
 version = "0.2.6-nested-rtxns-6"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e113d9bf240f974fbe7fd516cbfd8c422e925c0655495501c7237548425493d0"
+source = "git+https://github.com/meilisearch/heed?branch=allow-nested-rtxn-from-wtxn#464a9c33f76e0679115eef7379c9bdde9d5e1eda"
 dependencies = [
  "cc",
  "doxygen-rs",
@@ -3974,7 +3933,7 @@ checksum = "ae960838283323069879657ca3de837e9f7bbb4c7bf6ea7f1b290d5e9476d2e0"
 
 [[package]]
 name = "meili-snap"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "insta",
  "md5 0.8.0",
@@ -3985,7 +3944,7 @@ dependencies = [
 
 [[package]]
 name = "meilisearch"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "actix-cors",
  "actix-http",
@@ -4083,7 +4042,7 @@ dependencies = [
 
 [[package]]
 name = "meilisearch-auth"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "base64 0.22.1",
  "enum-iterator",
@@ -4102,7 +4061,7 @@ dependencies = [
 
 [[package]]
 name = "meilisearch-types"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "actix-web",
  "anyhow",
@@ -4137,7 +4096,7 @@ dependencies = [
 
 [[package]]
 name = "meilitool"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "anyhow",
  "clap",
@@ -4171,7 +4130,7 @@ dependencies = [
 
 [[package]]
 name = "milli"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "arroy",
  "bbqueue",
@@ -4750,7 +4709,7 @@ checksum = "9b4f627cb1b25917193a259e49bdad08f671f8d9708acfd5fe0a8c1455d87220"
 
 [[package]]
 name = "permissive-json-pointer"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "big_s",
  "serde_json",
@@ -6072,6 +6031,20 @@ name = "similar"
 version = "2.7.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "bbbb5d9659141646ae647b42fe094daf6c6192d1620870b449d9557f748b2daa"
+dependencies = [
+ "bstr",
+ "unicode-segmentation",
+]
+
+[[package]]
+name = "similar-asserts"
+version = "1.7.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b5b441962c817e33508847a22bd82f03a30cff43642dc2fae8b050566121eb9a"
+dependencies = [
+ "console",
+ "similar",
+]
 
 [[package]]
 name = "simple_asn1"
@@ -7105,12 +7078,6 @@ version = "0.1.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ba73ea9cf16a25df0c8caa16c51acb937d5712a8429db78a3ee29d5dcacd3a65"
 
-[[package]]
-name = "vcpkg"
-version = "0.2.15"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "accd4ea62f7bb7a82fe23066fb0957d48ef677f6eeb8215f372f52e48bb32426"
-
 [[package]]
 name = "vergen"
 version = "9.0.6"
@@ -7124,14 +7091,13 @@ dependencies = [
 ]
 
 [[package]]
-name = "vergen-git2"
-version = "1.0.7"
+name = "vergen-gitcl"
+version = "1.0.8"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "4f6ee511ec45098eabade8a0750e76eec671e7fb2d9360c563911336bea9cac1"
+checksum = "b9dfc1de6eb2e08a4ddf152f1b179529638bedc0ea95e6d667c014506377aefe"
 dependencies = [
  "anyhow",
  "derive_builder",
- "git2",
  "rustversion",
  "time",
  "vergen",
@@ -7783,7 +7749,7 @@ dependencies = [
 
 [[package]]
 name = "xtask"
-version = "1.28.1"
+version = "1.29.0"
 dependencies = [
  "anyhow",
  "build-info",
@@ -7792,9 +7758,11 @@ dependencies = [
  "futures-core",
  "futures-util",
  "reqwest",
+ "semver",
  "serde",
  "serde_json",
  "sha2",
+ "similar-asserts",
  "sysinfo",
  "time",
  "tokio",

Cargo.toml

@@ -23,7 +23,7 @@ members = [
 ]
 
 [workspace.package]
-version = "1.28.1"
+version = "1.29.0"
 authors = [
     "Quentin de Quelen <quentin@dequelen.me>",
     "Clément Renault <clement@meilisearch.com>",
@@ -52,3 +52,6 @@ opt-level = 3
 opt-level = 3
 [profile.dev.package.gemm-f16]
 opt-level = 3
+
+[patch.crates-io]
+heed = { git = "https://github.com/meilisearch/heed", branch = "allow-nested-rtxn-from-wtxn" }

TESTING.md

@@ -0,0 +1,326 @@
+# Declarative tests
+
+Declarative tests ensure that Meilisearch features remain stable across versions.
+
+While we already have unit tests, those are run against **temporary databases** that are created fresh each time and therefore never risk corruption.
+
+Declarative tests instead **simulate the lifetime of a database**: they chain together commands and requests to change the binary, verifying that database state and API responses remain consistent.
+
+## Basic example
+
+```jsonc
+{
+  "type": "test",
+  "name": "api-keys",
+  "binary": { // the first command will run on the binary following this specification.
+    "source": "release", // get the binary as a release from GitHub
+    "version": "1.19.0", // version to fetch
+    "edition": "community" // edition to fetch
+  },
+  "commands": []
+}
+```
+
+This example defines a no-op test (it does nothing).
+
+If the file is saved at `workloads/tests/example.json`, you can run it with:
+
+```bash
+cargo xtask test workloads/tests/example.json
+```
+
+## Commands
+
+Commands represent API requests sent to Meilisearch endpoints during a test.
+
+They are executed sequentially, and their responses can be validated to ensure consistent behavior across upgrades.
+
+```jsonc
+
+{
+  "route": "keys",
+  "method": "POST",
+  "body": {
+    "inline": {
+      "actions": [
+        "search",
+        "documents.add"
+      ],
+      "description": "Test API Key",
+      "expiresAt": null,
+      "indexes": [ "movies" ]
+    }
+  }
+}
+```
+
+This command issues a `POST /keys` request, creating an API key with permissions to search and add documents in the `movies` index.
+
+### Using assets in commands
+
+To keep tests concise and reusable, you can define **assets** at the root of the workload file.
+
+Assets are external data sources (such as datasets) that are cached between runs, making tests faster and easier to read.
+
+```jsonc
+{
+  "type": "test",
+  "name": "movies",
+  "binary": {
+    "source": "release",
+    "version": "1.19.0",
+    "edition": "community"
+  },
+  "assets": {
+    "movies.json": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/movies.json",
+      "sha256": "5b6e4cb660bc20327776e8a33ea197b43d9ec84856710ead1cc87ab24df77de1"
+    }
+  },
+  "commands": [
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "movies.json"
+      }
+    }
+  ]
+}
+```
+
+In this example:
+- The `movies.json` dataset is defined as an asset, pointing to a remote URL.
+- The SHA-256 checksum ensures integrity.
+- The `POST /indexes/movies/documents` command uses this asset as the request body.
+
+This makes the test much cleaner than inlining a large dataset directly into the command.
+
+For asset handling, please refer to the [declarative benchmarks documentation](/BENCHMARKS.md#adding-new-assets).
+
+### Asserting responses
+
+Commands can specify both the **expected status code** and the **expected response body**.
+
+```jsonc
+{
+  "route": "indexes/movies/documents",
+  "method": "POST",
+  "body": {
+    "asset": "movies.json"
+  },
+  "expectedStatus": 202,
+  "expectedResponse": {
+    "enqueuedAt": "[timestamp]", // Set to a bracketed string to ignore the value
+    "indexUid": "movies",
+    "status": "enqueued",
+    "taskUid": 1,
+    "type": "documentAdditionOrUpdate"
+  },
+  "synchronous": "WaitForTask"
+}
+```
+
+Manually writing `expectedResponse` fields can be tedious.
+
+Instead, you can let the test runner populate them automatically:
+
+```bash
+# Run the workload to populate expected fields. Only adds the missing ones, doesn't change existing data
+cargo xtask test workloads/tests/example.json --add-missing-responses
+
+# OR
+
+# Run the workload to populate expected fields. Updates all fields including existing ones
+cargo xtask test workloads/tests/example.json --update-responses
+```
+
+This workflow is recommended:
+
+1. Write the test without expected fields.
+2. Run it with `--add-missing-responses` to capture the actual responses.
+3. Review and commit the generated expectations.
+
+## Changing binary
+
+It is possible to insert an instruction to change the current Meilisearch instance from one binary specification to another during a test.
+
+When executed, such an instruction will:
+1. Stop the current Meilisearch instance.
+2. Fetch the binary specified by the instruction.
+3. Restart the server with the specified binary on the same database.
+
+```jsonc
+{
+  "type": "test",
+  "name": "movies",
+  "binary": {
+    "source": "release",
+    "version": "1.19.0", // start with version v1.19.0
+    "edition": "community"
+  },
+  "assets": {
+    "movies.json": {
+      "local_location": null,
+      "remote_location": "https://milli-benchmarks.fra1.digitaloceanspaces.com/bench/datasets/movies.json",
+      "sha256": "5b6e4cb660bc20327776e8a33ea197b43d9ec84856710ead1cc87ab24df77de1"
+    }
+  },
+  "commands": [
+    // setup some data
+    {
+      "route": "indexes/movies/documents",
+      "method": "POST",
+      "body": {
+        "asset": "movies.json"
+      }
+    },
+    // switch binary to v1.24.0
+    {
+      "binary": {
+        "source": "release",
+        "version": "1.24.0",
+        "edition": "community"
+      }
+    }
+  ]
+}
+```
+
+### Typical Usage
+
+In most cases, the change binary instruction will be used to update a database.
+
+- **Set up** some data using commands on an older version.
+- **Upgrade** to the latest version.
+- **Assert** that the data and API behavior remain correct after the upgrade.
+
+To properly test the dumpless upgrade, one should typically:
+
+1. Open the database without processing the update task: Use a `binary` instruction to switch to the desired version, passing `--experimental-dumpless-upgrade` and `--experimental-max-number-of-batched-tasks=0` as extra CLI arguments
+2. Check that the search, stats and task queue still work.
+3. Open the database and process the update task: Use a `binary` instruction to switch to the desired version, passing `--experimental-dumpless-upgrade` as the extra CLI argument. Use a `health` command to wait for the upgrade task to finish.
+4. Check that the indexing, search, stats, and task queue still work.
+
+```jsonc
+{
+  "type": "test",
+  "name": "movies",
+  "binary": {
+    "source": "release",
+    "version": "1.12.0",
+    "edition": "community"
+  },
+  "commands": [
+    // 0. Run commands to populate the database
+    {
+      // ..
+    },
+    // 1. Open the database with new MS without processing the update task
+    {
+      "binary": {
+        "source": "build", // build the binary from the sources in the current git repository
+        "edition": "community",
+        "extraCliArgs": [
+          "--experimental-dumpless-upgrade", // allows to open with a newer MS
+          "--experimental-max-number-of-batched-tasks=0" // prevent processing of the update task
+        ]
+      }
+    },
+    // 2. Check the search etc.
+    {
+      // ..
+    },
+    // 3. Open the database with new MS and processing the update task
+    {
+      "binary": {
+        "source": "build", // build the binary from the sources in the current git repository
+        "edition": "community",
+        "extraCliArgs": [
+          "--experimental-dumpless-upgrade" // allows to open with a newer MS
+          // no `--experimental-max-number-of-batched-tasks=0`
+        ]
+      }
+    },
+    // 4. Check the indexing, search, etc.
+    {
+      // ..
+    }
+  ]
+}
+```
+
+This ensures backward compatibility: databases created with older Meilisearch versions should remain functional and consistent after an upgrade.
+
+## Variables
+
+Sometimes a command needs to use a value returned by a **previous response**.
+These values can be captured and reused using the register field.
+
+```jsonc
+{
+  "route": "keys",
+  "method": "POST",
+  "body": {
+    "inline": {
+        "actions": [
+        "search",
+        "documents.add"
+        ],
+        "description": "Test API Key",
+        "expiresAt": null,
+        "indexes": [ "movies" ]
+    }
+  },
+  "expectedResponse": {
+      "key": "c6f64630bad2996b1f675007c8800168e14adf5d6a7bb1a400a6d2b158050eaf",
+      // ...
+  },
+  "register": {
+    "key": "/key"
+  },
+  "synchronous": "WaitForResponse"
+}
+```
+
+The `register` field captures the value at the JSON path `/key` from the response.
+Paths follow the **JavaScript Object Notation Pointer (RFC 6901)** format.
+Registered variables are available for all subsequent commands.
+
+Registered variables can be referenced by wrapping their name in double curly braces:
+
+In the route/path:
+
+```jsonc
+{
+  "route": "tasks/{{ task_id }}",
+  "method": "GET"
+}
+```
+
+In the request body:
+
+```jsonc
+{
+  "route": "indexes/movies/documents",
+  "method": "PATCH",
+  "body": {
+    "inline": {
+      "id": "{{ document_id }}",
+      "overview": "Shazam turns evil and the world is in danger.",
+    }
+  }
+}
+```
+
+Or they can be referenced by their name (**without curly braces**) as an API key:
+
+```jsonc
+{
+  "route": "indexes/movies/documents",
+  "method": "POST",
+  "body": { /* ... */ },
+  "apiKeyVariable": "key" // The **content** of the key variable will be used as an API key
+}
+```

crates/benchmarks/benches/indexing.rs

@@ -21,6 +21,10 @@ use roaring::RoaringBitmap;
 #[global_allocator]
 static ALLOC: mimalloc::MiMalloc = mimalloc::MiMalloc;
 
+fn no_cancel() -> bool {
+    false
+}
+
 const BENCHMARK_ITERATION: usize = 10;
 
 fn setup_dir(path: impl AsRef<Path>) {
@@ -65,7 +69,7 @@ fn setup_settings<'t>(
     let sortable_fields = sortable_fields.iter().map(|s| s.to_string()).collect();
     builder.set_sortable_fields(sortable_fields);
 
-    builder.execute(&|| false, &Progress::default(), Default::default()).unwrap();
+    builder.execute(&no_cancel, &Progress::default(), Default::default()).unwrap();
 }
 
 fn setup_index_with_settings(
@@ -152,7 +156,7 @@ fn indexing_songs_default(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -168,7 +172,7 @@ fn indexing_songs_default(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -220,7 +224,7 @@ fn reindexing_songs_default(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -236,7 +240,7 @@ fn reindexing_songs_default(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -266,7 +270,7 @@ fn reindexing_songs_default(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -282,7 +286,7 @@ fn reindexing_songs_default(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -336,7 +340,7 @@ fn deleting_songs_in_batches_default(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -352,7 +356,7 @@ fn deleting_songs_in_batches_default(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -414,7 +418,7 @@ fn indexing_songs_in_three_batches_default(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -430,7 +434,7 @@ fn indexing_songs_in_three_batches_default(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -460,7 +464,7 @@ fn indexing_songs_in_three_batches_default(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -476,7 +480,7 @@ fn indexing_songs_in_three_batches_default(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -502,7 +506,7 @@ fn indexing_songs_in_three_batches_default(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -518,7 +522,7 @@ fn indexing_songs_in_three_batches_default(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -571,7 +575,7 @@ fn indexing_songs_without_faceted_numbers(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -587,7 +591,7 @@ fn indexing_songs_without_faceted_numbers(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -639,7 +643,7 @@ fn indexing_songs_without_faceted_fields(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -655,7 +659,7 @@ fn indexing_songs_without_faceted_fields(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -707,7 +711,7 @@ fn indexing_wiki(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -723,7 +727,7 @@ fn indexing_wiki(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -774,7 +778,7 @@ fn reindexing_wiki(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -790,7 +794,7 @@ fn reindexing_wiki(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -820,7 +824,7 @@ fn reindexing_wiki(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -836,7 +840,7 @@ fn reindexing_wiki(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -889,7 +893,7 @@ fn deleting_wiki_in_batches_default(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -905,7 +909,7 @@ fn deleting_wiki_in_batches_default(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -967,7 +971,7 @@ fn indexing_wiki_in_three_batches(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -983,7 +987,7 @@ fn indexing_wiki_in_three_batches(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -1014,7 +1018,7 @@ fn indexing_wiki_in_three_batches(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -1030,7 +1034,7 @@ fn indexing_wiki_in_three_batches(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -1057,7 +1061,7 @@ fn indexing_wiki_in_three_batches(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -1073,7 +1077,7 @@ fn indexing_wiki_in_three_batches(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -1125,7 +1129,7 @@ fn indexing_movies_default(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -1141,7 +1145,7 @@ fn indexing_movies_default(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -1192,7 +1196,7 @@ fn reindexing_movies_default(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -1208,7 +1212,7 @@ fn reindexing_movies_default(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -1238,7 +1242,7 @@ fn reindexing_movies_default(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -1254,7 +1258,7 @@ fn reindexing_movies_default(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -1307,7 +1311,7 @@ fn deleting_movies_in_batches_default(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -1323,7 +1327,7 @@ fn deleting_movies_in_batches_default(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -1372,7 +1376,7 @@ fn delete_documents_from_ids(index: Index, document_ids_to_delete: Vec<RoaringBi
             Some(primary_key),
             &document_changes,
             RuntimeEmbedders::default(),
-            &|| false,
+            &no_cancel,
             &Progress::default(),
             &Default::default(),
         )
@@ -1422,7 +1426,7 @@ fn indexing_movies_in_three_batches(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -1438,7 +1442,7 @@ fn indexing_movies_in_three_batches(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -1468,7 +1472,7 @@ fn indexing_movies_in_three_batches(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -1484,7 +1488,7 @@ fn indexing_movies_in_three_batches(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -1510,7 +1514,7 @@ fn indexing_movies_in_three_batches(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -1526,7 +1530,7 @@ fn indexing_movies_in_three_batches(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -1601,7 +1605,7 @@ fn indexing_nested_movies_default(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -1617,7 +1621,7 @@ fn indexing_nested_movies_default(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -1693,7 +1697,7 @@ fn deleting_nested_movies_in_batches_default(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -1709,7 +1713,7 @@ fn deleting_nested_movies_in_batches_default(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -1777,7 +1781,7 @@ fn indexing_nested_movies_without_faceted_fields(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -1793,7 +1797,7 @@ fn indexing_nested_movies_without_faceted_fields(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -1845,7 +1849,7 @@ fn indexing_geo(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -1861,7 +1865,7 @@ fn indexing_geo(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -1912,7 +1916,7 @@ fn reindexing_geo(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -1928,7 +1932,7 @@ fn reindexing_geo(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -1958,7 +1962,7 @@ fn reindexing_geo(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -1974,7 +1978,7 @@ fn reindexing_geo(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )
@@ -2027,7 +2031,7 @@ fn deleting_geo_in_batches_default(c: &mut Criterion) {
                         &rtxn,
                         None,
                         &mut new_fields_ids_map,
-                        &|| false,
+                        &no_cancel,
                         Progress::default(),
                         None,
                     )
@@ -2043,7 +2047,7 @@ fn deleting_geo_in_batches_default(c: &mut Criterion) {
                     primary_key,
                     &document_changes,
                     RuntimeEmbedders::default(),
-                    &|| false,
+                    &no_cancel,
                     &Progress::default(),
                     &Default::default(),
                 )

crates/build-info/Cargo.toml

@@ -15,4 +15,4 @@ time = { version = "0.3.44", features = ["parsing"] }
 
 [build-dependencies]
 anyhow = "1.0.100"
-vergen-git2 = "1.0.7"
+vergen-gitcl = "1.0.8"

crates/build-info/build.rs

@@ -15,7 +15,7 @@ fn emit_git_variables() -> anyhow::Result<()> {
     // Note: any code that needs VERGEN_ environment variables should take care to define them manually in the Dockerfile and pass them
     // in the corresponding GitHub workflow (publish_docker.yml).
     // This is due to the Dockerfile building the binary outside of the git directory.
-    let mut builder = vergen_git2::Git2Builder::default();
+    let mut builder = vergen_gitcl::GitclBuilder::default();
 
     builder.branch(true);
     builder.commit_timestamp(true);
@@ -25,5 +25,5 @@ fn emit_git_variables() -> anyhow::Result<()> {
 
     let git2 = builder.build()?;
 
-    vergen_git2::Emitter::default().fail_on_error().add_instructions(&git2)?.emit()
+    vergen_gitcl::Emitter::default().fail_on_error().add_instructions(&git2)?.emit()
 }

crates/build-info/src/main.rs

@@ -0,0 +1,6 @@
+use build_info::BuildInfo;
+
+fn main() {
+    let info = BuildInfo::from_build();
+    dbg!(info);
+}

crates/dump/src/reader/v2/settings.rs

@@ -107,19 +107,14 @@ impl Settings<Unchecked> {
     }
 }
 
-#[derive(Debug, Clone, PartialEq)]
+#[derive(Default, Debug, Clone, PartialEq)]
 pub enum Setting<T> {
     Set(T),
     Reset,
+    #[default]
     NotSet,
 }
 
-impl<T> Default for Setting<T> {
-    fn default() -> Self {
-        Self::NotSet
-    }
-}
-
 impl<T> Setting<T> {
     pub const fn is_not_set(&self) -> bool {
         matches!(self, Self::NotSet)

crates/dump/src/reader/v3/settings.rs

@@ -161,19 +161,14 @@ pub struct Facets {
     pub min_level_size: Option<NonZeroUsize>,
 }
 
-#[derive(Debug, Clone, PartialEq, Eq)]
+#[derive(Default, Debug, Clone, PartialEq, Eq)]
 pub enum Setting<T> {
     Set(T),
     Reset,
+    #[default]
     NotSet,
 }
 
-impl<T> Default for Setting<T> {
-    fn default() -> Self {
-        Self::NotSet
-    }
-}
-
 impl<T> Setting<T> {
     pub fn map<U, F>(self, f: F) -> Setting<U>
     where

crates/dump/src/reader/v4/meta.rs

@@ -1,9 +1,7 @@
 use std::fmt::{self, Display, Formatter};
-use std::marker::PhantomData;
 use std::str::FromStr;
 
-use serde::de::Visitor;
-use serde::{Deserialize, Deserializer};
+use serde::Deserialize;
 use uuid::Uuid;
 
 use super::settings::{Settings, Unchecked};
@@ -82,59 +80,3 @@ impl Display for IndexUidFormatError {
 }
 
 impl std::error::Error for IndexUidFormatError {}
-
-/// A type that tries to match either a star (*) or
-/// any other thing that implements `FromStr`.
-#[derive(Debug)]
-#[cfg_attr(test, derive(serde::Serialize))]
-pub enum StarOr<T> {
-    Star,
-    Other(T),
-}
-
-impl<'de, T, E> Deserialize<'de> for StarOr<T>
-where
-    T: FromStr<Err = E>,
-    E: Display,
-{
-    fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
-    where
-        D: Deserializer<'de>,
-    {
-        /// Serde can't differentiate between `StarOr::Star` and `StarOr::Other` without a tag.
-        /// Simply using `#[serde(untagged)]` + `#[serde(rename="*")]` will lead to attempting to
-        /// deserialize everything as a `StarOr::Other`, including "*".
-        /// [`#[serde(other)]`](https://serde.rs/variant-attrs.html#other) might have helped but is
-        /// not supported on untagged enums.
-        struct StarOrVisitor<T>(PhantomData<T>);
-
-        impl<T, FE> Visitor<'_> for StarOrVisitor<T>
-        where
-            T: FromStr<Err = FE>,
-            FE: Display,
-        {
-            type Value = StarOr<T>;
-
-            fn expecting(&self, formatter: &mut Formatter) -> std::fmt::Result {
-                formatter.write_str("a string")
-            }
-
-            fn visit_str<SE>(self, v: &str) -> Result<Self::Value, SE>
-            where
-                SE: serde::de::Error,
-            {
-                match v {
-                    "*" => Ok(StarOr::Star),
-                    v => {
-                        let other = FromStr::from_str(v).map_err(|e: T::Err| {
-                            SE::custom(format!("Invalid `other` value: {}", e))
-                        })?;
-                        Ok(StarOr::Other(other))
-                    }
-                }
-            }
-        }
-
-        deserializer.deserialize_str(StarOrVisitor(PhantomData))
-    }
-}

crates/dump/src/reader/v4/settings.rs

@@ -192,19 +192,14 @@ pub struct Facets {
     pub min_level_size: Option<NonZeroUsize>,
 }
 
-#[derive(Debug, Clone, PartialEq, Eq, Copy)]
+#[derive(Default, Debug, Clone, PartialEq, Eq, Copy)]
 pub enum Setting<T> {
     Set(T),
     Reset,
+    #[default]
     NotSet,
 }
 
-impl<T> Default for Setting<T> {
-    fn default() -> Self {
-        Self::NotSet
-    }
-}
-
 impl<T> Setting<T> {
     pub fn set(self) -> Option<T> {
         match self {

crates/dump/src/reader/v5/settings.rs

@@ -47,20 +47,15 @@ pub struct Settings<T> {
     pub _kind: PhantomData<T>,
 }
 
-#[derive(Debug, Clone, PartialEq, Eq, Copy)]
+#[derive(Default, Debug, Clone, PartialEq, Eq, Copy)]
 #[cfg_attr(test, derive(serde::Serialize))]
 pub enum Setting<T> {
     Set(T),
     Reset,
+    #[default]
     NotSet,
 }
 
-impl<T> Default for Setting<T> {
-    fn default() -> Self {
-        Self::NotSet
-    }
-}
-
 impl<T> Setting<T> {
     pub fn set(self) -> Option<T> {
         match self {

crates/dump/src/reader/v5/tasks.rs

@@ -322,7 +322,7 @@ impl From<Task> for TaskView {
             _ => None,
         });
 
-        let duration = finished_at.zip(started_at).map(|(tf, ts)| (tf - ts));
+        let duration = finished_at.zip(started_at).map(|(tf, ts)| tf - ts);
 
         Self {
             uid: id,

crates/index-scheduler/src/insta_snapshot.rs

@@ -6,7 +6,7 @@ use meilisearch_types::heed::types::{SerdeBincode, SerdeJson, Str};
 use meilisearch_types::heed::{Database, RoTxn};
 use meilisearch_types::milli::{CboRoaringBitmapCodec, RoaringBitmapCodec, BEU32};
 use meilisearch_types::tasks::{Details, Kind, Status, Task};
-use meilisearch_types::versioning;
+use meilisearch_types::versioning::{self, VERSION_MAJOR, VERSION_MINOR, VERSION_PATCH};
 use roaring::RoaringBitmap;
 
 use crate::index_mapper::IndexMapper;
@@ -320,7 +320,11 @@ fn snapshot_details(d: &Details) -> String {
             format!("{{ url: {url:?}, api_key: {api_key:?}, payload_size: {payload_size:?}, indexes: {indexes:?} }}")
         }
         Details::UpgradeDatabase { from, to } => {
-            format!("{{ from: {from:?}, to: {to:?} }}")
+            if to == &(VERSION_MAJOR, VERSION_MINOR, VERSION_PATCH) {
+                format!("{{ from: {from:?}, to: [current version] }}")
+            } else {
+                format!("{{ from: {from:?}, to: {to:?} }}")
+            }
         }
         Details::IndexCompaction { index_uid, pre_compaction_size, post_compaction_size } => {
             format!("{{ index_uid: {index_uid:?}, pre_compaction_size: {pre_compaction_size:?}, post_compaction_size: {post_compaction_size:?} }}")
@@ -400,7 +404,21 @@ pub fn snapshot_batch(batch: &Batch) -> String {
 
     snap.push('{');
     snap.push_str(&format!("uid: {uid}, "));
-    snap.push_str(&format!("details: {}, ", serde_json::to_string(details).unwrap()));
+    let details = if let Some(upgrade_to) = &details.upgrade_to {
+        if upgrade_to.as_str()
+            == format!("v{VERSION_MAJOR}.{VERSION_MINOR}.{VERSION_PATCH}").as_str()
+        {
+            let mut details = details.clone();
+
+            details.upgrade_to = Some("[current version]".into());
+            serde_json::to_string(&details).unwrap()
+        } else {
+            serde_json::to_string(details).unwrap()
+        }
+    } else {
+        serde_json::to_string(details).unwrap()
+    };
+    snap.push_str(&format!("details: {details}, "));
     snap.push_str(&format!("stats: {}, ", serde_json::to_string(&stats).unwrap()));
     if !embedder_stats.skip_serializing() {
         snap.push_str(&format!(

crates/index-scheduler/src/scheduler/snapshots/test_failure.rs/upgrade_failure/after_processing_everything.snap

@@ -6,7 +6,7 @@ source: crates/index-scheduler/src/scheduler/test_failure.rs
 []
 ----------------------------------------------------------------------
 ### All Tasks:
-0 {uid: 0, batch_uid: 0, status: succeeded, details: { from: (1, 12, 0), to: (1, 28, 1) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
+0 {uid: 0, batch_uid: 0, status: succeeded, details: { from: (1, 12, 0), to: [current version] }, kind: UpgradeDatabase { from: (1, 12, 0) }}
 1 {uid: 1, batch_uid: 1, status: succeeded, details: { primary_key: Some("mouse"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "catto", primary_key: Some("mouse") }}
 2 {uid: 2, batch_uid: 2, status: succeeded, details: { primary_key: Some("bone"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "doggo", primary_key: Some("bone") }}
 3 {uid: 3, batch_uid: 3, status: failed, error: ResponseError { code: 200, message: "Index `doggo` already exists.", error_code: "index_already_exists", error_type: "invalid_request", error_link: "https://docs.meilisearch.com/errors#index_already_exists" }, details: { primary_key: Some("bone"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "doggo", primary_key: Some("bone") }}
@@ -57,7 +57,7 @@ girafo: { number_of_documents: 0, field_distribution: {} }
 [timestamp] [4,]
 ----------------------------------------------------------------------
 ### All Batches:
-0 {uid: 0, details: {"upgradeFrom":"v1.12.0","upgradeTo":"v1.28.1"}, stats: {"totalNbTasks":1,"status":{"succeeded":1},"types":{"upgradeDatabase":1},"indexUids":{}}, stop reason: "stopped after the last task of type `upgradeDatabase` because they cannot be batched with tasks of any other type.", }
+0 {uid: 0, details: {"upgradeFrom":"v1.12.0","upgradeTo":"[current version]"}, stats: {"totalNbTasks":1,"status":{"succeeded":1},"types":{"upgradeDatabase":1},"indexUids":{}}, stop reason: "stopped after the last task of type `upgradeDatabase` because they cannot be batched with tasks of any other type.", }
 1 {uid: 1, details: {"primaryKey":"mouse"}, stats: {"totalNbTasks":1,"status":{"succeeded":1},"types":{"indexCreation":1},"indexUids":{"catto":1}}, stop reason: "created batch containing only task with id 1 of type `indexCreation` that cannot be batched with any other task.", }
 2 {uid: 2, details: {"primaryKey":"bone"}, stats: {"totalNbTasks":1,"status":{"succeeded":1},"types":{"indexCreation":1},"indexUids":{"doggo":1}}, stop reason: "created batch containing only task with id 2 of type `indexCreation` that cannot be batched with any other task.", }
 3 {uid: 3, details: {"primaryKey":"bone"}, stats: {"totalNbTasks":1,"status":{"failed":1},"types":{"indexCreation":1},"indexUids":{"doggo":1}}, stop reason: "created batch containing only task with id 3 of type `indexCreation` that cannot be batched with any other task.", }

crates/index-scheduler/src/scheduler/snapshots/test_failure.rs/upgrade_failure/register_automatic_upgrade_task.snap

@@ -6,7 +6,7 @@ source: crates/index-scheduler/src/scheduler/test_failure.rs
 []
 ----------------------------------------------------------------------
 ### All Tasks:
-0 {uid: 0, status: enqueued, details: { from: (1, 12, 0), to: (1, 28, 1) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
+0 {uid: 0, status: enqueued, details: { from: (1, 12, 0), to: [current version] }, kind: UpgradeDatabase { from: (1, 12, 0) }}
 ----------------------------------------------------------------------
 ### Status:
 enqueued [0,]

crates/index-scheduler/src/scheduler/snapshots/test_failure.rs/upgrade_failure/registered_a_task_while_the_upgrade_task_is_enqueued.snap

@@ -6,7 +6,7 @@ source: crates/index-scheduler/src/scheduler/test_failure.rs
 []
 ----------------------------------------------------------------------
 ### All Tasks:
-0 {uid: 0, status: enqueued, details: { from: (1, 12, 0), to: (1, 28, 1) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
+0 {uid: 0, status: enqueued, details: { from: (1, 12, 0), to: [current version] }, kind: UpgradeDatabase { from: (1, 12, 0) }}
 1 {uid: 1, status: enqueued, details: { primary_key: Some("mouse"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "catto", primary_key: Some("mouse") }}
 ----------------------------------------------------------------------
 ### Status:

crates/index-scheduler/src/scheduler/snapshots/test_failure.rs/upgrade_failure/upgrade_task_failed.snap

@@ -6,7 +6,7 @@ source: crates/index-scheduler/src/scheduler/test_failure.rs
 []
 ----------------------------------------------------------------------
 ### All Tasks:
-0 {uid: 0, batch_uid: 0, status: failed, error: ResponseError { code: 200, message: "Planned failure for tests.", error_code: "internal", error_type: "internal", error_link: "https://docs.meilisearch.com/errors#internal" }, details: { from: (1, 12, 0), to: (1, 28, 1) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
+0 {uid: 0, batch_uid: 0, status: failed, error: ResponseError { code: 200, message: "Planned failure for tests.", error_code: "internal", error_type: "internal", error_link: "https://docs.meilisearch.com/errors#internal" }, details: { from: (1, 12, 0), to: [current version] }, kind: UpgradeDatabase { from: (1, 12, 0) }}
 1 {uid: 1, status: enqueued, details: { primary_key: Some("mouse"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "catto", primary_key: Some("mouse") }}
 ----------------------------------------------------------------------
 ### Status:
@@ -37,7 +37,7 @@ catto [1,]
 [timestamp] [0,]
 ----------------------------------------------------------------------
 ### All Batches:
-0 {uid: 0, details: {"upgradeFrom":"v1.12.0","upgradeTo":"v1.28.1"}, stats: {"totalNbTasks":1,"status":{"failed":1},"types":{"upgradeDatabase":1},"indexUids":{}}, stop reason: "stopped after the last task of type `upgradeDatabase` because they cannot be batched with tasks of any other type.", }
+0 {uid: 0, details: {"upgradeFrom":"v1.12.0","upgradeTo":"[current version]"}, stats: {"totalNbTasks":1,"status":{"failed":1},"types":{"upgradeDatabase":1},"indexUids":{}}, stop reason: "stopped after the last task of type `upgradeDatabase` because they cannot be batched with tasks of any other type.", }
 ----------------------------------------------------------------------
 ### Batch to tasks mapping:
 0 [0,]

crates/index-scheduler/src/scheduler/snapshots/test_failure.rs/upgrade_failure/upgrade_task_failed_again.snap

@@ -6,7 +6,7 @@ source: crates/index-scheduler/src/scheduler/test_failure.rs
 []
 ----------------------------------------------------------------------
 ### All Tasks:
-0 {uid: 0, batch_uid: 0, status: failed, error: ResponseError { code: 200, message: "Planned failure for tests.", error_code: "internal", error_type: "internal", error_link: "https://docs.meilisearch.com/errors#internal" }, details: { from: (1, 12, 0), to: (1, 28, 1) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
+0 {uid: 0, batch_uid: 0, status: failed, error: ResponseError { code: 200, message: "Planned failure for tests.", error_code: "internal", error_type: "internal", error_link: "https://docs.meilisearch.com/errors#internal" }, details: { from: (1, 12, 0), to: [current version] }, kind: UpgradeDatabase { from: (1, 12, 0) }}
 1 {uid: 1, status: enqueued, details: { primary_key: Some("mouse"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "catto", primary_key: Some("mouse") }}
 2 {uid: 2, status: enqueued, details: { primary_key: Some("bone"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "doggo", primary_key: Some("bone") }}
 ----------------------------------------------------------------------
@@ -40,7 +40,7 @@ doggo [2,]
 [timestamp] [0,]
 ----------------------------------------------------------------------
 ### All Batches:
-0 {uid: 0, details: {"upgradeFrom":"v1.12.0","upgradeTo":"v1.28.1"}, stats: {"totalNbTasks":1,"status":{"failed":1},"types":{"upgradeDatabase":1},"indexUids":{}}, stop reason: "stopped after the last task of type `upgradeDatabase` because they cannot be batched with tasks of any other type.", }
+0 {uid: 0, details: {"upgradeFrom":"v1.12.0","upgradeTo":"[current version]"}, stats: {"totalNbTasks":1,"status":{"failed":1},"types":{"upgradeDatabase":1},"indexUids":{}}, stop reason: "stopped after the last task of type `upgradeDatabase` because they cannot be batched with tasks of any other type.", }
 ----------------------------------------------------------------------
 ### Batch to tasks mapping:
 0 [0,]

crates/index-scheduler/src/scheduler/snapshots/test_failure.rs/upgrade_failure/upgrade_task_succeeded.snap

@@ -6,7 +6,7 @@ source: crates/index-scheduler/src/scheduler/test_failure.rs
 []
 ----------------------------------------------------------------------
 ### All Tasks:
-0 {uid: 0, batch_uid: 0, status: succeeded, details: { from: (1, 12, 0), to: (1, 28, 1) }, kind: UpgradeDatabase { from: (1, 12, 0) }}
+0 {uid: 0, batch_uid: 0, status: succeeded, details: { from: (1, 12, 0), to: [current version] }, kind: UpgradeDatabase { from: (1, 12, 0) }}
 1 {uid: 1, status: enqueued, details: { primary_key: Some("mouse"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "catto", primary_key: Some("mouse") }}
 2 {uid: 2, status: enqueued, details: { primary_key: Some("bone"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "doggo", primary_key: Some("bone") }}
 3 {uid: 3, status: enqueued, details: { primary_key: Some("bone"), old_new_uid: None, new_index_uid: None }, kind: IndexCreation { index_uid: "doggo", primary_key: Some("bone") }}
@@ -43,7 +43,7 @@ doggo [2,3,]
 [timestamp] [0,]
 ----------------------------------------------------------------------
 ### All Batches:
-0 {uid: 0, details: {"upgradeFrom":"v1.12.0","upgradeTo":"v1.28.1"}, stats: {"totalNbTasks":1,"status":{"succeeded":1},"types":{"upgradeDatabase":1},"indexUids":{}}, stop reason: "stopped after the last task of type `upgradeDatabase` because they cannot be batched with tasks of any other type.", }
+0 {uid: 0, details: {"upgradeFrom":"v1.12.0","upgradeTo":"[current version]"}, stats: {"totalNbTasks":1,"status":{"succeeded":1},"types":{"upgradeDatabase":1},"indexUids":{}}, stop reason: "stopped after the last task of type `upgradeDatabase` because they cannot be batched with tasks of any other type.", }
 ----------------------------------------------------------------------
 ### Batch to tasks mapping:
 0 [0,]

crates/index-scheduler/src/upgrade/mod.rs

@@ -1,7 +1,7 @@
 use anyhow::bail;
 use meilisearch_types::heed::{Env, RwTxn, WithoutTls};
 use meilisearch_types::tasks::{Details, KindWithContent, Status, Task};
-use meilisearch_types::versioning::{VERSION_MAJOR, VERSION_MINOR, VERSION_PATCH};
+use meilisearch_types::versioning;
 use time::OffsetDateTime;
 use tracing::info;
 
@@ -9,83 +9,82 @@ use crate::queue::TaskQueue;
 use crate::versioning::Versioning;
 
 trait UpgradeIndexScheduler {
-    fn upgrade(
-        &self,
-        env: &Env<WithoutTls>,
-        wtxn: &mut RwTxn,
-        original: (u32, u32, u32),
-    ) -> anyhow::Result<()>;
-    fn target_version(&self) -> (u32, u32, u32);
+    fn upgrade(&self, env: &Env<WithoutTls>, wtxn: &mut RwTxn) -> anyhow::Result<()>;
+    /// Whether the migration should be applied, depending on the initial version of the index scheduler before
+    /// any migration was applied
+    fn must_upgrade(&self, initial_version: (u32, u32, u32)) -> bool;
+    /// A progress-centric description of the migration
+    fn description(&self) -> &'static str;
 }
 
+/// Upgrade the index scheduler to the binary version.
+///
+/// # Warning
+///
+/// The current implementation uses a single wtxn to the index scheduler for the whole duration of the upgrade.
+/// If migrations start taking take a long time, it might prevent tasks from being registered.
+/// If this issue manifests, then it can be mitigated by adding a `fn target_version` to `UpgradeIndexScheduler`,
+/// to be able to write intermediate versions and drop the wtxn between applying migrations.
 pub fn upgrade_index_scheduler(
     env: &Env<WithoutTls>,
     versioning: &Versioning,
-    from: (u32, u32, u32),
-    to: (u32, u32, u32),
+    initial_version: (u32, u32, u32),
 ) -> anyhow::Result<()> {
-    let current_major = to.0;
-    let current_minor = to.1;
-    let current_patch = to.2;
+    let target_major: u32 = versioning::VERSION_MAJOR;
+    let target_minor: u32 = versioning::VERSION_MINOR;
+    let target_patch: u32 = versioning::VERSION_PATCH;
+    let target_version = (target_major, target_minor, target_patch);
 
-    let upgrade_functions: &[&dyn UpgradeIndexScheduler] = &[
-        // This is the last upgrade function, it will be called when the index is up to date.
-        // any other upgrade function should be added before this one.
-        &ToCurrentNoOp {},
-    ];
-
-    let start = match from {
-        (1, 12, _) => 0,
-        (1, 13, _) => 0,
-        (1, 14, _) => 0,
-        (1, 15, _) => 0,
-        (1, 16, _) => 0,
-        (1, 17, _) => 0,
-        (1, 18, _) => 0,
-        (1, 19, _) => 0,
-        (1, 20, _) => 0,
-        (1, 21, _) => 0,
-        (1, 22, _) => 0,
-        (1, 23, _) => 0,
-        (1, 24, _) => 0,
-        (1, 25, _) => 0,
-        (1, 26, _) => 0,
-        (1, 27, _) => 0,
-        (1, 28, _) => 0,
-        (major, minor, patch) => {
-            if major > current_major
-                || (major == current_major && minor > current_minor)
-                || (major == current_major && minor == current_minor && patch > current_patch)
-            {
-                bail!(
-                "Database version {major}.{minor}.{patch} is higher than the Meilisearch version {current_major}.{current_minor}.{current_patch}. Downgrade is not supported",
-                );
-            } else if major < 1 || (major == current_major && minor < 12) {
-                bail!(
-                "Database version {major}.{minor}.{patch} is too old for the experimental dumpless upgrade feature. Please generate a dump using the v{major}.{minor}.{patch} and import it in the v{current_major}.{current_minor}.{current_patch}",
-            );
-            } else {
-                bail!("Unknown database version: v{major}.{minor}.{patch}");
-            }
-        }
-    };
-
-    info!("Upgrading the task queue");
-    let mut local_from = from;
-    for upgrade in upgrade_functions[start..].iter() {
-        let target = upgrade.target_version();
-        info!(
-            "Upgrading from v{}.{}.{} to v{}.{}.{}",
-            local_from.0, local_from.1, local_from.2, target.0, target.1, target.2
-        );
-        let mut wtxn = env.write_txn()?;
-        upgrade.upgrade(env, &mut wtxn, local_from)?;
-        versioning.set_version(&mut wtxn, target)?;
-        wtxn.commit()?;
-        local_from = target;
+    if initial_version == target_version {
+        return Ok(());
     }
 
+    let upgrade_functions: &[&dyn UpgradeIndexScheduler] = &[
+        // List all upgrade functions to apply in order here.
+    ];
+
+    let (initial_major, initial_minor, initial_patch) = initial_version;
+
+    if initial_version > target_version {
+        bail!(
+                "Database version {initial_major}.{initial_minor}.{initial_patch} is higher than the Meilisearch version {target_major}.{target_minor}.{target_patch}. Downgrade is not supported",
+            );
+    }
+
+    if initial_version < (1, 12, 0) {
+        bail!(
+                "Database version {initial_major}.{initial_minor}.{initial_patch} is too old for the experimental dumpless upgrade feature. Please generate a dump using the v{initial_major}.{initial_minor}.{initial_patch} and import it in the v{target_major}.{target_minor}.{target_patch}",
+            );
+    }
+
+    info!("Upgrading the task queue");
     let mut wtxn = env.write_txn()?;
+    let migration_count = upgrade_functions.len();
+    for (migration_index, upgrade) in upgrade_functions.iter().enumerate() {
+        if upgrade.must_upgrade(initial_version) {
+            info!(
+                "[{migration_index}/{migration_count}]Applying migration: {}",
+                upgrade.description()
+            );
+
+            upgrade.upgrade(env, &mut wtxn)?;
+
+            info!(
+                "[{}/{migration_count}]Migration applied: {}",
+                migration_index + 1,
+                upgrade.description()
+            )
+        } else {
+            info!(
+                "[{migration_index}/{migration_count}]Skipping unnecessary migration: {}",
+                upgrade.description()
+            )
+        }
+    }
+
+    versioning.set_version(&mut wtxn, target_version)?;
+    info!("Task queue upgraded, spawning the upgrade database task");
+
     let queue = TaskQueue::new(env, &mut wtxn)?;
     let uid = queue.next_task_id(&wtxn)?;
     queue.register(
@@ -98,9 +97,9 @@ pub fn upgrade_index_scheduler(
             finished_at: None,
             error: None,
             canceled_by: None,
-            details: Some(Details::UpgradeDatabase { from, to }),
+            details: Some(Details::UpgradeDatabase { from: initial_version, to: target_version }),
             status: Status::Enqueued,
-            kind: KindWithContent::UpgradeDatabase { from },
+            kind: KindWithContent::UpgradeDatabase { from: initial_version },
             network: None,
             custom_metadata: None,
         },
@@ -109,21 +108,3 @@ pub fn upgrade_index_scheduler(
 
     Ok(())
 }
-
-#[allow(non_camel_case_types)]
-struct ToCurrentNoOp {}
-
-impl UpgradeIndexScheduler for ToCurrentNoOp {
-    fn upgrade(
-        &self,
-        _env: &Env<WithoutTls>,
-        _wtxn: &mut RwTxn,
-        _original: (u32, u32, u32),
-    ) -> anyhow::Result<()> {
-        Ok(())
-    }
-
-    fn target_version(&self) -> (u32, u32, u32) {
-        (VERSION_MAJOR, VERSION_MINOR, VERSION_PATCH)
-    }
-}

crates/index-scheduler/src/versioning.rs

@@ -64,14 +64,7 @@ impl Versioning {
         };
         wtxn.commit()?;
 
-        let bin_major: u32 = versioning::VERSION_MAJOR;
-        let bin_minor: u32 = versioning::VERSION_MINOR;
-        let bin_patch: u32 = versioning::VERSION_PATCH;
-        let to = (bin_major, bin_minor, bin_patch);
-
-        if from != to {
-            upgrade_index_scheduler(env, &this, from, to)?;
-        }
+        upgrade_index_scheduler(env, &this, from)?;
 
         // Once we reach this point it means the upgrade process, if there was one is entirely finished
         // we can safely say we reached the latest version of the index scheduler

crates/meilisearch-types/src/network.rs

@@ -1,6 +1,7 @@
-use serde::{Deserialize, Serialize};
 use std::collections::BTreeMap;
 
+use serde::{Deserialize, Serialize};
+
 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq, Default)]
 #[serde(rename_all = "camelCase")]
 pub struct Network {

crates/meilisearch/src/analytics/segment_analytics.rs

@@ -1,7 +1,7 @@
 use std::any::TypeId;
 use std::collections::{HashMap, HashSet};
 use std::fs;
-use std::path::{Path, PathBuf};
+use std::path::Path;
 use std::sync::Arc;
 use std::time::{Duration, Instant};
 
@@ -344,14 +344,14 @@ impl Infos {
             experimental_no_edition_2024_for_dumps,
             experimental_vector_store_setting: vector_store_setting,
             gpu_enabled: meilisearch_types::milli::vector::is_cuda_enabled(),
-            db_path: db_path != PathBuf::from("./data.ms"),
+            db_path: db_path != Path::new("./data.ms"),
             import_dump: import_dump.is_some(),
-            dump_dir: dump_dir != PathBuf::from("dumps/"),
+            dump_dir: dump_dir != Path::new("dumps/"),
             ignore_missing_dump,
             ignore_dump_if_db_exists,
             import_snapshot: import_snapshot.is_some(),
             schedule_snapshot,
-            snapshot_dir: snapshot_dir != PathBuf::from("snapshots/"),
+            snapshot_dir: snapshot_dir != Path::new("snapshots/"),
             uses_s3_snapshots: s3_snapshot_options.is_some(),
             ignore_missing_snapshot,
             ignore_snapshot_if_db_exists,

crates/meilisearch/src/routes/metrics.rs

@@ -183,7 +183,11 @@ pub async fn get_metrics(
     crate::metrics::MEILISEARCH_LAST_FINISHED_BATCHES_PROGRESS_TRACE_MS.reset();
     let (batches, _total) = index_scheduler.get_batches_from_authorized_indexes(
         // Fetch the finished batches...
-        &Query { statuses: Some(vec![Status::Succeeded, Status::Failed]), ..Query::default() },
+        &Query {
+            statuses: Some(vec![Status::Succeeded, Status::Failed]),
+            limit: Some(1),
+            ..Query::default()
+        },
         auth_filters,
     )?;
     // ...and get the last batch only.

crates/meilisearch/src/search/mod.rs

@@ -789,11 +789,12 @@ impl TryFrom<Value> for ExternalDocumentId {
     }
 }
 
-#[derive(Debug, Copy, Clone, PartialEq, Eq, Deserr, ToSchema, Serialize)]
+#[derive(Default, Debug, Copy, Clone, PartialEq, Eq, Deserr, ToSchema, Serialize)]
 #[deserr(rename_all = camelCase)]
 #[serde(rename_all = "camelCase")]
 pub enum MatchingStrategy {
     /// Remove query words from last to first
+    #[default]
     Last,
     /// All query words are mandatory
     All,
@@ -801,12 +802,6 @@ pub enum MatchingStrategy {
     Frequency,
 }
 
-impl Default for MatchingStrategy {
-    fn default() -> Self {
-        Self::Last
-    }
-}
-
 impl From<MatchingStrategy> for TermsMatchingStrategy {
     fn from(other: MatchingStrategy) -> Self {
         match other {

crates/meilisearch/tests/auth/tenant_token.rs

@@ -187,7 +187,7 @@ macro_rules! compute_forbidden_search {
 
 #[actix_rt::test]
 async fn search_authorized_simple_token() {
-    let tenant_tokens = vec![
+    let tenant_tokens = [
         hashmap! {
             "searchRules" => json!({"*": {}}),
             "exp" => json!((OffsetDateTime::now_utc() + Duration::hours(1)).unix_timestamp())
@@ -239,7 +239,7 @@ async fn search_authorized_simple_token() {
 
 #[actix_rt::test]
 async fn search_authorized_filter_token() {
-    let tenant_tokens = vec![
+    let tenant_tokens = [
         hashmap! {
             "searchRules" => json!({"*": {"filter": "color = blue"}}),
             "exp" => json!((OffsetDateTime::now_utc() + Duration::hours(1)).unix_timestamp())
@@ -292,7 +292,7 @@ async fn search_authorized_filter_token() {
 
 #[actix_rt::test]
 async fn filter_search_authorized_filter_token() {
-    let tenant_tokens = vec![
+    let tenant_tokens = [
         hashmap! {
             "searchRules" => json!({"*": {"filter": "color = blue"}}),
             "exp" => json!((OffsetDateTime::now_utc() + Duration::hours(1)).unix_timestamp())
@@ -353,7 +353,7 @@ async fn filter_search_authorized_filter_token() {
 /// Tests that those Tenant Token are incompatible with the REFUSED_KEYS defined above.
 #[actix_rt::test]
 async fn error_search_token_forbidden_parent_key() {
-    let tenant_tokens = vec![
+    let tenant_tokens = [
         hashmap! {
             "searchRules" => json!({"*": {}}),
             "exp" => json!((OffsetDateTime::now_utc() + Duration::hours(1)).unix_timestamp())
@@ -389,7 +389,7 @@ async fn error_search_token_forbidden_parent_key() {
 
 #[actix_rt::test]
 async fn error_search_forbidden_token() {
-    let tenant_tokens = vec![
+    let tenant_tokens = [
         // bad index
         hashmap! {
             "searchRules" => json!({"products": {}}),

crates/meilisearch/tests/auth/tenant_token_multi_search.rs

@@ -680,7 +680,7 @@ async fn multi_search_authorized_simple_token() {
 
 #[actix_rt::test]
 async fn single_search_authorized_filter_token() {
-    let tenant_tokens = vec![
+    let tenant_tokens = [
         hashmap! {
             "searchRules" => json!({"*": {"filter": "color = blue"}}),
             "exp" => json!((OffsetDateTime::now_utc() + Duration::hours(1)).unix_timestamp())
@@ -733,7 +733,7 @@ async fn single_search_authorized_filter_token() {
 
 #[actix_rt::test]
 async fn multi_search_authorized_filter_token() {
-    let both_tenant_tokens = vec![
+    let both_tenant_tokens = [
         hashmap! {
             "searchRules" => json!({"sales": {"filter": "color = blue"}, "products": {"filter": "doggos.age <= 5"}}),
             "exp" => json!((OffsetDateTime::now_utc() + Duration::hours(1)).unix_timestamp())
@@ -842,7 +842,7 @@ async fn filter_single_search_authorized_filter_token() {
 
 #[actix_rt::test]
 async fn filter_multi_search_authorized_filter_token() {
-    let tenant_tokens = vec![
+    let tenant_tokens = [
         hashmap! {
             "searchRules" => json!({"sales": {"filter": "color = blue"}, "products": {"filter": "doggos.age <= 5"}}),
             "exp" => json!((OffsetDateTime::now_utc() + Duration::hours(1)).unix_timestamp())
@@ -900,7 +900,7 @@ async fn filter_multi_search_authorized_filter_token() {
 /// Tests that those Tenant Token are incompatible with the REFUSED_KEYS defined above.
 #[actix_rt::test]
 async fn error_single_search_token_forbidden_parent_key() {
-    let tenant_tokens = vec![
+    let tenant_tokens = [
         hashmap! {
             "searchRules" => json!({"*": {}}),
             "exp" => json!((OffsetDateTime::now_utc() + Duration::hours(1)).unix_timestamp())
@@ -941,7 +941,7 @@ async fn error_single_search_token_forbidden_parent_key() {
 /// Tests that those Tenant Token are incompatible with the REFUSED_KEYS defined above.
 #[actix_rt::test]
 async fn error_multi_search_token_forbidden_parent_key() {
-    let tenant_tokens = vec![
+    let tenant_tokens = [
         hashmap! {
             "searchRules" => json!({"*": {}}),
             "exp" => json!((OffsetDateTime::now_utc() + Duration::hours(1)).unix_timestamp())

crates/meilisearch/tests/settings/get_settings.rs

@@ -197,7 +197,7 @@ test_setting_routes!(
     {
       setting: vector_store,
       update_verb: patch,
-      default_value: null
+      default_value: "experimental"
     },
 );
 

crates/meilisearch/tests/settings/mod.rs

@@ -2,6 +2,7 @@ mod chat;
 mod distinct;
 mod errors;
 mod get_settings;
+mod parent_seachable_fields;
 mod prefix_search_settings;
 mod proximity_settings;
 mod tokenizer_customization;

crates/meilisearch/tests/settings/parent_seachable_fields.rs

@@ -0,0 +1,114 @@
+use meili_snap::{json_string, snapshot};
+use once_cell::sync::Lazy;
+
+use crate::common::Server;
+use crate::json;
+
+static DOCUMENTS: Lazy<crate::common::Value> = Lazy::new(|| {
+    json!([
+        {
+            "id": 1,
+            "meta": {
+                "title": "Soup of the day",
+                "description": "many the fish",
+            }
+        },
+        {
+            "id": 2,
+            "meta": {
+                "title": "Soup of day",
+                "description": "many the lazy fish",
+            }
+        },
+        {
+            "id": 3,
+            "meta": {
+                "title": "the Soup of day",
+                "description": "many the fish",
+            }
+        },
+    ])
+});
+
+#[actix_rt::test]
+async fn nested_field_becomes_searchable() {
+    let server = Server::new_shared();
+    let index = server.unique_index();
+
+    let (task, _status_code) = index.add_documents(DOCUMENTS.clone(), None).await;
+    server.wait_task(task.uid()).await.succeeded();
+
+    let (response, code) = index
+        .update_settings(json!({
+            "searchableAttributes": ["meta.title"]
+        }))
+        .await;
+    assert_eq!("202", code.as_str(), "{response:?}");
+    server.wait_task(response.uid()).await.succeeded();
+
+    // We expect no documents when searching for
+    // a nested non-searchable field
+    index
+        .search(json!({"q": "many fish"}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"[]"###);
+        })
+        .await;
+
+    let (response, code) = index
+        .update_settings(json!({
+            "searchableAttributes": ["meta.title", "meta.description"]
+        }))
+        .await;
+    assert_eq!("202", code.as_str(), "{response:?}");
+    server.wait_task(response.uid()).await.succeeded();
+
+    // We expect all the documents when the nested field becomes searchable
+    index
+        .search(json!({"q": "many fish"}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"
+            [
+              {
+                "id": 1,
+                "meta": {
+                  "title": "Soup of the day",
+                  "description": "many the fish"
+                }
+              },
+              {
+                "id": 3,
+                "meta": {
+                  "title": "the Soup of day",
+                  "description": "many the fish"
+                }
+              },
+              {
+                "id": 2,
+                "meta": {
+                  "title": "Soup of day",
+                  "description": "many the lazy fish"
+                }
+              }
+            ]
+            "###);
+        })
+        .await;
+
+    let (response, code) = index
+        .update_settings(json!({
+            "searchableAttributes": ["meta.title"]
+        }))
+        .await;
+    assert_eq!("202", code.as_str(), "{response:?}");
+    server.wait_task(response.uid()).await.succeeded();
+
+    // We expect no documents when searching for
+    // a nested non-searchable field
+    index
+        .search(json!({"q": "many fish"}), |response, code| {
+            snapshot!(code, @"200 OK");
+            snapshot!(json_string!(response["hits"]), @r###"[]"###);
+        })
+        .await;
+}

crates/meilisearch/tests/upgrade/mod.rs

@@ -42,8 +42,16 @@ async fn version_too_old() {
     std::fs::create_dir_all(&db_path).unwrap();
     std::fs::write(db_path.join("VERSION"), "1.11.9999").unwrap();
     let options = Opt { experimental_dumpless_upgrade: true, ..default_settings };
-    let err = Server::new_with_options(options).await.map(|_| ()).unwrap_err();
-    snapshot!(err, @"Database version 1.11.9999 is too old for the experimental dumpless upgrade feature. Please generate a dump using the v1.11.9999 and import it in the v1.28.1");
+    let err = Server::new_with_options(options).await.map(|_| ()).unwrap_err().to_string();
+
+    let major = meilisearch_types::versioning::VERSION_MAJOR;
+    let minor = meilisearch_types::versioning::VERSION_MINOR;
+    let patch = meilisearch_types::versioning::VERSION_PATCH;
+
+    let current_version = format!("{major}.{minor}.{patch}");
+    let err = err.replace(&current_version, "[current version]");
+
+    snapshot!(err, @"Database version 1.11.9999 is too old for the experimental dumpless upgrade feature. Please generate a dump using the v1.11.9999 and import it in the v[current version]");
 }
 
 #[actix_rt::test]
@@ -54,11 +62,21 @@ async fn version_requires_downgrade() {
     std::fs::create_dir_all(&db_path).unwrap();
     let major = meilisearch_types::versioning::VERSION_MAJOR;
     let minor = meilisearch_types::versioning::VERSION_MINOR;
-    let patch = meilisearch_types::versioning::VERSION_PATCH + 1;
-    std::fs::write(db_path.join("VERSION"), format!("{major}.{minor}.{patch}")).unwrap();
+    let mut patch = meilisearch_types::versioning::VERSION_PATCH;
+
+    let current_version = format!("{major}.{minor}.{patch}");
+    patch += 1;
+    let future_version = format!("{major}.{minor}.{patch}");
+
+    std::fs::write(db_path.join("VERSION"), &future_version).unwrap();
     let options = Opt { experimental_dumpless_upgrade: true, ..default_settings };
     let err = Server::new_with_options(options).await.map(|_| ()).unwrap_err();
-    snapshot!(err, @"Database version 1.28.2 is higher than the Meilisearch version 1.28.1. Downgrade is not supported");
+
+    let err = err.to_string();
+    let err = err.replace(&current_version, "[current version]");
+    let err = err.replace(&future_version, "[future version]");
+
+    snapshot!(err, @"Database version [future version] is higher than the Meilisearch version [current version]. Downgrade is not supported");
 }
 
 #[actix_rt::test]

crates/meilisearch/tests/upgrade/v1_12/snapshots/v1_12_0.rs/check_the_index_scheduler/batches_filter_afterEnqueuedAt_equal_2025-01-16T16_47_41.snap

@@ -8,7 +8,7 @@ source: crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs
       "progress": null,
       "details": {
         "upgradeFrom": "v1.12.0",
-        "upgradeTo": "v1.28.1"
+        "upgradeTo": "[current version]"
       },
       "stats": {
         "totalNbTasks": 1,

crates/meilisearch/tests/upgrade/v1_12/snapshots/v1_12_0.rs/check_the_index_scheduler/batches_filter_afterFinishedAt_equal_2025-01-16T16_47_41.snap

@@ -8,7 +8,7 @@ source: crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs
       "progress": null,
       "details": {
         "upgradeFrom": "v1.12.0",
-        "upgradeTo": "v1.28.1"
+        "upgradeTo": "[current version]"
       },
       "stats": {
         "totalNbTasks": 1,

crates/meilisearch/tests/upgrade/v1_12/snapshots/v1_12_0.rs/check_the_index_scheduler/batches_filter_afterStartedAt_equal_2025-01-16T16_47_41.snap

@@ -8,7 +8,7 @@ source: crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs
       "progress": null,
       "details": {
         "upgradeFrom": "v1.12.0",
-        "upgradeTo": "v1.28.1"
+        "upgradeTo": "[current version]"
       },
       "stats": {
         "totalNbTasks": 1,

crates/meilisearch/tests/upgrade/v1_12/snapshots/v1_12_0.rs/check_the_index_scheduler/tasks_filter_afterEnqueuedAt_equal_2025-01-16T16_47_41.snap

@@ -12,7 +12,7 @@ source: crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs
       "canceledBy": null,
       "details": {
         "upgradeFrom": "v1.12.0",
-        "upgradeTo": "v1.28.1"
+        "upgradeTo": "[current version]"
       },
       "error": null,
       "duration": "[duration]",

crates/meilisearch/tests/upgrade/v1_12/snapshots/v1_12_0.rs/check_the_index_scheduler/tasks_filter_afterFinishedAt_equal_2025-01-16T16_47_41.snap

@@ -12,7 +12,7 @@ source: crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs
       "canceledBy": null,
       "details": {
         "upgradeFrom": "v1.12.0",
-        "upgradeTo": "v1.28.1"
+        "upgradeTo": "[current version]"
       },
       "error": null,
       "duration": "[duration]",

crates/meilisearch/tests/upgrade/v1_12/snapshots/v1_12_0.rs/check_the_index_scheduler/tasks_filter_afterStartedAt_equal_2025-01-16T16_47_41.snap

@@ -12,7 +12,7 @@ source: crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs
       "canceledBy": null,
       "details": {
         "upgradeFrom": "v1.12.0",
-        "upgradeTo": "v1.28.1"
+        "upgradeTo": "[current version]"
       },
       "error": null,
       "duration": "[duration]",

crates/meilisearch/tests/upgrade/v1_12/snapshots/v1_12_0.rs/check_the_index_scheduler/the_whole_batch_queue_once_everything_has_been_processed.snap

@@ -8,7 +8,7 @@ source: crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs
       "progress": null,
       "details": {
         "upgradeFrom": "v1.12.0",
-        "upgradeTo": "v1.28.1"
+        "upgradeTo": "[current version]"
       },
       "stats": {
         "totalNbTasks": 1,

crates/meilisearch/tests/upgrade/v1_12/snapshots/v1_12_0.rs/check_the_index_scheduler/the_whole_task_queue_once_everything_has_been_processed.snap

@@ -12,7 +12,7 @@ source: crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs
       "canceledBy": null,
       "details": {
         "upgradeFrom": "v1.12.0",
-        "upgradeTo": "v1.28.1"
+        "upgradeTo": "[current version]"
       },
       "error": null,
       "duration": "[duration]",

crates/meilisearch/tests/upgrade/v1_12/v1_12_0.rs

@@ -166,55 +166,55 @@ async fn check_the_index_scheduler(server: &Server) {
     // We rewrite the first task for all calls because it may be the upgrade database with unknown dates and duration.
     // The other tasks should NOT change
     let (tasks, _) = server.tasks_filter("limit=1000").await;
-    snapshot!(json_string!(tasks, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "the_whole_task_queue_once_everything_has_been_processed");
+    snapshot!(json_string!(tasks, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "the_whole_task_queue_once_everything_has_been_processed");
     let (batches, _) = server.batches_filter("limit=1000").await;
-    snapshot!(json_string!(batches, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "the_whole_batch_queue_once_everything_has_been_processed");
+    snapshot!(json_string!(batches, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "the_whole_batch_queue_once_everything_has_been_processed");
 
     // Tests all the tasks query parameters
     let (tasks, _) = server.tasks_filter("uids=10").await;
-    snapshot!(json_string!(tasks, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_uids_equal_10");
+    snapshot!(json_string!(tasks, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_uids_equal_10");
     let (tasks, _) = server.tasks_filter("batchUids=10").await;
-    snapshot!(json_string!(tasks, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_batchUids_equal_10");
+    snapshot!(json_string!(tasks, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_batchUids_equal_10");
     let (tasks, _) = server.tasks_filter("statuses=canceled").await;
-    snapshot!(json_string!(tasks, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_statuses_equal_canceled");
+    snapshot!(json_string!(tasks, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_statuses_equal_canceled");
     // types has already been tested above to retrieve the upgrade database
     let (tasks, _) = server.tasks_filter("canceledBy=19").await;
-    snapshot!(json_string!(tasks, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_canceledBy_equal_19");
+    snapshot!(json_string!(tasks, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_canceledBy_equal_19");
     let (tasks, _) = server.tasks_filter("beforeEnqueuedAt=2025-01-16T16:47:41Z").await;
-    snapshot!(json_string!(tasks, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_beforeEnqueuedAt_equal_2025-01-16T16_47_41");
+    snapshot!(json_string!(tasks, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_beforeEnqueuedAt_equal_2025-01-16T16_47_41");
     let (tasks, _) = server.tasks_filter("afterEnqueuedAt=2025-01-16T16:47:41Z").await;
-    snapshot!(json_string!(tasks, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_afterEnqueuedAt_equal_2025-01-16T16_47_41");
+    snapshot!(json_string!(tasks, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_afterEnqueuedAt_equal_2025-01-16T16_47_41");
     let (tasks, _) = server.tasks_filter("beforeStartedAt=2025-01-16T16:47:41Z").await;
-    snapshot!(json_string!(tasks, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_beforeStartedAt_equal_2025-01-16T16_47_41");
+    snapshot!(json_string!(tasks, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_beforeStartedAt_equal_2025-01-16T16_47_41");
     let (tasks, _) = server.tasks_filter("afterStartedAt=2025-01-16T16:47:41Z").await;
-    snapshot!(json_string!(tasks, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_afterStartedAt_equal_2025-01-16T16_47_41");
+    snapshot!(json_string!(tasks, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_afterStartedAt_equal_2025-01-16T16_47_41");
     let (tasks, _) = server.tasks_filter("beforeFinishedAt=2025-01-16T16:47:41Z").await;
-    snapshot!(json_string!(tasks, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_beforeFinishedAt_equal_2025-01-16T16_47_41");
+    snapshot!(json_string!(tasks, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_beforeFinishedAt_equal_2025-01-16T16_47_41");
     let (tasks, _) = server.tasks_filter("afterFinishedAt=2025-01-16T16:47:41Z").await;
-    snapshot!(json_string!(tasks, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_afterFinishedAt_equal_2025-01-16T16_47_41");
+    snapshot!(json_string!(tasks, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]" }), name: "tasks_filter_afterFinishedAt_equal_2025-01-16T16_47_41");
 
     // Tests all the batches query parameters
     let (batches, _) = server.batches_filter("uids=10").await;
-    snapshot!(json_string!(batches, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_uids_equal_10");
+    snapshot!(json_string!(batches, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_uids_equal_10");
     let (batches, _) = server.batches_filter("batchUids=10").await;
-    snapshot!(json_string!(batches, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_batchUids_equal_10");
+    snapshot!(json_string!(batches, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_batchUids_equal_10");
     let (batches, _) = server.batches_filter("statuses=canceled").await;
-    snapshot!(json_string!(batches, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_statuses_equal_canceled");
+    snapshot!(json_string!(batches, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_statuses_equal_canceled");
     // types has already been tested above to retrieve the upgrade database
     let (batches, _) = server.batches_filter("canceledBy=19").await;
-    snapshot!(json_string!(batches, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_canceledBy_equal_19");
+    snapshot!(json_string!(batches, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_canceledBy_equal_19");
     let (batches, _) = server.batches_filter("beforeEnqueuedAt=2025-01-16T16:47:41Z").await;
-    snapshot!(json_string!(batches, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_beforeEnqueuedAt_equal_2025-01-16T16_47_41");
+    snapshot!(json_string!(batches, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_beforeEnqueuedAt_equal_2025-01-16T16_47_41");
     let (batches, _) = server.batches_filter("afterEnqueuedAt=2025-01-16T16:47:41Z").await;
-    snapshot!(json_string!(batches, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_afterEnqueuedAt_equal_2025-01-16T16_47_41");
+    snapshot!(json_string!(batches, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_afterEnqueuedAt_equal_2025-01-16T16_47_41");
     let (batches, _) = server.batches_filter("beforeStartedAt=2025-01-16T16:47:41Z").await;
-    snapshot!(json_string!(batches, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_beforeStartedAt_equal_2025-01-16T16_47_41");
+    snapshot!(json_string!(batches, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_beforeStartedAt_equal_2025-01-16T16_47_41");
     let (batches, _) = server.batches_filter("afterStartedAt=2025-01-16T16:47:41Z").await;
-    snapshot!(json_string!(batches, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_afterStartedAt_equal_2025-01-16T16_47_41");
+    snapshot!(json_string!(batches, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_afterStartedAt_equal_2025-01-16T16_47_41");
     let (batches, _) = server.batches_filter("beforeFinishedAt=2025-01-16T16:47:41Z").await;
-    snapshot!(json_string!(batches, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_beforeFinishedAt_equal_2025-01-16T16_47_41");
+    snapshot!(json_string!(batches, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_beforeFinishedAt_equal_2025-01-16T16_47_41");
     let (batches, _) = server.batches_filter("afterFinishedAt=2025-01-16T16:47:41Z").await;
-    snapshot!(json_string!(batches, { ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_afterFinishedAt_equal_2025-01-16T16_47_41");
+    snapshot!(json_string!(batches, { ".results[0].details.upgradeTo" => "[current version]", ".results[0].duration" => "[duration]", ".results[0].enqueuedAt" => "[date]", ".results[0].startedAt" => "[date]", ".results[0].finishedAt" => "[date]", ".results[0].stats.progressTrace" => "[progressTrace]", ".results[0].stats.internalDatabaseSizes" => "[internalDatabaseSizes]", ".results[0].stats.writeChannelCongestion" => "[writeChannelCongestion]" }), name: "batches_filter_afterFinishedAt_equal_2025-01-16T16_47_41");
 
     let (stats, _) = server.stats().await;
     assert_json_snapshot!(stats, {

crates/meilisearch/tests/vector/binary_quantized.rs

@@ -104,8 +104,8 @@ async fn binary_quantize_before_sending_documents() {
             "manual": {
               "embeddings": [
                 [
-                  -1.0,
-                  -1.0,
+                  0.0,
+                  0.0,
                   1.0
                 ]
               ],
@@ -122,7 +122,7 @@ async fn binary_quantize_before_sending_documents() {
                 [
                   1.0,
                   1.0,
-                  -1.0
+                  0.0
                 ]
               ],
               "regenerate": false
@@ -191,8 +191,8 @@ async fn binary_quantize_after_sending_documents() {
             "manual": {
               "embeddings": [
                 [
-                  -1.0,
-                  -1.0,
+                  0.0,
+                  0.0,
                   1.0
                 ]
               ],
@@ -209,7 +209,7 @@ async fn binary_quantize_after_sending_documents() {
                 [
                   1.0,
                   1.0,
-                  -1.0
+                  0.0
                 ]
               ],
               "regenerate": false

crates/meilisearch/tests/vector/huggingface.rs

@@ -0,0 +1,43 @@
+use meili_snap::snapshot;
+
+use crate::common::{GetAllDocumentsOptions, Server};
+use crate::json;
+
+#[actix_rt::test]
+async fn hf_bge_m3_force_cls_settings() {
+    let server = Server::new_shared();
+    let index = server.unique_index();
+
+    let (response, code) = index
+        .update_settings(json!({
+          "embedders": {
+              "default": {
+                  "source": "huggingFace",
+                  "model": "baai/bge-m3",
+                  "revision": "5617a9f61b028005a4858fdac845db406aefb181",
+                  "pooling": "forceCls",
+                  // minimal template to allow potential document embedding if used later
+                  "documentTemplate": "{{doc.title}}"
+              }
+          }
+        }))
+        .await;
+    snapshot!(code, @"202 Accepted");
+    server.wait_task(response.uid()).await.succeeded();
+
+    // Try to embed one simple document
+    let (task, code) =
+        index.add_documents(json!([{ "id": 1, "title": "Hello world" }]), None).await;
+    snapshot!(code, @"202 Accepted");
+    server.wait_task(task.uid()).await.succeeded();
+
+    // Retrieve the document with vectors and assert embeddings were produced
+    let (documents, _code) = index
+        .get_all_documents(GetAllDocumentsOptions { retrieve_vectors: true, ..Default::default() })
+        .await;
+    let has_vectors = documents["results"][0]["_vectors"]["default"]["embeddings"]
+        .as_array()
+        .map(|a| !a.is_empty())
+        .unwrap_or(false);
+    snapshot!(has_vectors, @"true");
+}

crates/meilisearch/tests/vector/mod.rs

@@ -1,5 +1,6 @@
 mod binary_quantized;
 mod fragments;
+mod huggingface;
 #[cfg(feature = "test-ollama")]
 mod ollama;
 mod openai;

crates/meilisearch/tests/vector/ollama.rs

@@ -500,13 +500,6 @@ async fn test_both_apis() {
     snapshot!(code, @"200 OK");
     snapshot!(json_string!(response["hits"]), @r###"
     [
-      {
-        "id": 0,
-        "name": "kefir",
-        "gender": "M",
-        "birthyear": 2023,
-        "breed": "Patou"
-      },
       {
         "id": 2,
         "name": "Vénus",
@@ -527,6 +520,13 @@ async fn test_both_apis() {
         "gender": "M",
         "birthyear": 1995,
         "breed": "Labrador Retriever"
+      },
+      {
+        "id": 0,
+        "name": "kefir",
+        "gender": "M",
+        "birthyear": 2023,
+        "breed": "Patou"
       }
     ]
     "###);
@@ -540,13 +540,6 @@ async fn test_both_apis() {
     snapshot!(code, @"200 OK");
     snapshot!(json_string!(response["hits"]), @r###"
     [
-      {
-        "id": 0,
-        "name": "kefir",
-        "gender": "M",
-        "birthyear": 2023,
-        "breed": "Patou"
-      },
       {
         "id": 2,
         "name": "Vénus",
@@ -567,6 +560,13 @@ async fn test_both_apis() {
         "gender": "M",
         "birthyear": 1995,
         "breed": "Labrador Retriever"
+      },
+      {
+        "id": 0,
+        "name": "kefir",
+        "gender": "M",
+        "birthyear": 2023,
+        "breed": "Patou"
       }
     ]
     "###);
@@ -581,18 +581,11 @@ async fn test_both_apis() {
     snapshot!(json_string!(response["hits"]), @r###"
     [
       {
-        "id": 0,
-        "name": "kefir",
+        "id": 1,
+        "name": "Intel",
         "gender": "M",
-        "birthyear": 2023,
-        "breed": "Patou"
-      },
-      {
-        "id": 3,
-        "name": "Max",
-        "gender": "M",
-        "birthyear": 1995,
-        "breed": "Labrador Retriever"
+        "birthyear": 2011,
+        "breed": "Beagle"
       },
       {
         "id": 2,
@@ -602,11 +595,18 @@ async fn test_both_apis() {
         "breed": "Jack Russel Terrier"
       },
       {
-        "id": 1,
-        "name": "Intel",
+        "id": 3,
+        "name": "Max",
         "gender": "M",
-        "birthyear": 2011,
-        "breed": "Beagle"
+        "birthyear": 1995,
+        "breed": "Labrador Retriever"
+      },
+      {
+        "id": 0,
+        "name": "kefir",
+        "gender": "M",
+        "birthyear": 2023,
+        "breed": "Patou"
       }
     ]
     "###);
@@ -621,18 +621,11 @@ async fn test_both_apis() {
     snapshot!(json_string!(response["hits"]), @r###"
     [
       {
-        "id": 0,
-        "name": "kefir",
+        "id": 1,
+        "name": "Intel",
         "gender": "M",
-        "birthyear": 2023,
-        "breed": "Patou"
-      },
-      {
-        "id": 3,
-        "name": "Max",
-        "gender": "M",
-        "birthyear": 1995,
-        "breed": "Labrador Retriever"
+        "birthyear": 2011,
+        "breed": "Beagle"
       },
       {
         "id": 2,
@@ -642,11 +635,18 @@ async fn test_both_apis() {
         "breed": "Jack Russel Terrier"
       },
       {
-        "id": 1,
-        "name": "Intel",
+        "id": 3,
+        "name": "Max",
         "gender": "M",
-        "birthyear": 2011,
-        "breed": "Beagle"
+        "birthyear": 1995,
+        "breed": "Labrador Retriever"
+      },
+      {
+        "id": 0,
+        "name": "kefir",
+        "gender": "M",
+        "birthyear": 2023,
+        "breed": "Patou"
       }
     ]
     "###);
@@ -661,18 +661,11 @@ async fn test_both_apis() {
     snapshot!(json_string!(response["hits"]), @r###"
     [
       {
-        "id": 0,
-        "name": "kefir",
+        "id": 1,
+        "name": "Intel",
         "gender": "M",
-        "birthyear": 2023,
-        "breed": "Patou"
-      },
-      {
-        "id": 3,
-        "name": "Max",
-        "gender": "M",
-        "birthyear": 1995,
-        "breed": "Labrador Retriever"
+        "birthyear": 2011,
+        "breed": "Beagle"
       },
       {
         "id": 2,
@@ -682,11 +675,18 @@ async fn test_both_apis() {
         "breed": "Jack Russel Terrier"
       },
       {
-        "id": 1,
-        "name": "Intel",
+        "id": 3,
+        "name": "Max",
         "gender": "M",
-        "birthyear": 2011,
-        "breed": "Beagle"
+        "birthyear": 1995,
+        "breed": "Labrador Retriever"
+      },
+      {
+        "id": 0,
+        "name": "kefir",
+        "gender": "M",
+        "birthyear": 2023,
+        "breed": "Patou"
       }
     ]
     "###);
@@ -701,18 +701,11 @@ async fn test_both_apis() {
     snapshot!(json_string!(response["hits"]), @r###"
     [
       {
-        "id": 0,
-        "name": "kefir",
+        "id": 1,
+        "name": "Intel",
         "gender": "M",
-        "birthyear": 2023,
-        "breed": "Patou"
-      },
-      {
-        "id": 3,
-        "name": "Max",
-        "gender": "M",
-        "birthyear": 1995,
-        "breed": "Labrador Retriever"
+        "birthyear": 2011,
+        "breed": "Beagle"
       },
       {
         "id": 2,
@@ -722,11 +715,18 @@ async fn test_both_apis() {
         "breed": "Jack Russel Terrier"
       },
       {
-        "id": 1,
-        "name": "Intel",
+        "id": 3,
+        "name": "Max",
         "gender": "M",
-        "birthyear": 2011,
-        "breed": "Beagle"
+        "birthyear": 1995,
+        "breed": "Labrador Retriever"
+      },
+      {
+        "id": 0,
+        "name": "kefir",
+        "gender": "M",
+        "birthyear": 2023,
+        "breed": "Patou"
       }
     ]
     "###);

crates/milli/Cargo.toml

@@ -91,7 +91,7 @@ rhai = { version = "1.23.6", features = [
     "sync",
 ] }
 arroy = "0.6.4-nested-rtxns"
-hannoy = { version = "0.0.9-nested-rtxns-2", features = ["arroy"] }
+hannoy = { version = "0.1.0-nested-rtxns", features = ["arroy"] }
 rand = "0.8.5"
 tracing = "0.1.41"
 ureq = { version = "2.12.1", features = ["json"] }

crates/milli/src/fields_ids_map/metadata.rs

@@ -18,6 +18,8 @@ use crate::{
 pub struct Metadata {
     /// The weight as defined in the FieldidsWeightsMap of the searchable attribute if it is searchable.
     pub searchable: Option<Weight>,
+    /// The field is part of the exact attributes.
+    pub exact: bool,
     /// The field is part of the sortable attributes.
     pub sortable: bool,
     /// The field is defined as the distinct attribute.
@@ -209,6 +211,7 @@ impl Metadata {
 #[derive(Debug, Clone)]
 pub struct MetadataBuilder {
     searchable_attributes: Option<Vec<String>>,
+    exact_searchable_attributes: Vec<String>,
     filterable_attributes: Vec<FilterableAttributesRule>,
     sortable_attributes: HashSet<String>,
     localized_attributes: Option<Vec<LocalizedAttributesRule>>,
@@ -220,15 +223,18 @@ impl MetadataBuilder {
     pub fn from_index(index: &Index, rtxn: &RoTxn) -> Result<Self> {
         let searchable_attributes = index
             .user_defined_searchable_fields(rtxn)?
-            .map(|fields| fields.into_iter().map(|s| s.to_string()).collect());
+            .map(|fields| fields.into_iter().map(String::from).collect());
+        let exact_searchable_attributes =
+            index.exact_attributes(rtxn)?.into_iter().map(String::from).collect();
         let filterable_attributes = index.filterable_attributes_rules(rtxn)?;
         let sortable_attributes = index.sortable_fields(rtxn)?;
         let localized_attributes = index.localized_attributes_rules(rtxn)?;
-        let distinct_attribute = index.distinct_field(rtxn)?.map(|s| s.to_string());
+        let distinct_attribute = index.distinct_field(rtxn)?.map(String::from);
         let asc_desc_attributes = index.asc_desc_fields(rtxn)?;
 
         Ok(Self::new(
             searchable_attributes,
+            exact_searchable_attributes,
             filterable_attributes,
             sortable_attributes,
             localized_attributes,
@@ -242,6 +248,7 @@ impl MetadataBuilder {
     /// This is used for testing, prefer using `MetadataBuilder::from_index` instead.
     pub fn new(
         searchable_attributes: Option<Vec<String>>,
+        exact_searchable_attributes: Vec<String>,
         filterable_attributes: Vec<FilterableAttributesRule>,
         sortable_attributes: HashSet<String>,
         localized_attributes: Option<Vec<LocalizedAttributesRule>>,
@@ -256,6 +263,7 @@ impl MetadataBuilder {
 
         Self {
             searchable_attributes,
+            exact_searchable_attributes,
             filterable_attributes,
             sortable_attributes,
             localized_attributes,
@@ -269,6 +277,7 @@ impl MetadataBuilder {
             // Vectors fields are not searchable, filterable, distinct or asc_desc
             return Metadata {
                 searchable: None,
+                exact: false,
                 sortable: false,
                 distinct: false,
                 asc_desc: false,
@@ -296,6 +305,7 @@ impl MetadataBuilder {
             // Geo fields are not searchable, distinct or asc_desc
             return Metadata {
                 searchable: None,
+                exact: false,
                 sortable,
                 distinct: false,
                 asc_desc: false,
@@ -309,6 +319,7 @@ impl MetadataBuilder {
             debug_assert!(!sortable, "geojson fields should not be sortable");
             return Metadata {
                 searchable: None,
+                exact: false,
                 sortable,
                 distinct: false,
                 asc_desc: false,
@@ -329,6 +340,8 @@ impl MetadataBuilder {
             None => Some(0),
         };
 
+        let exact = self.exact_searchable_attributes.iter().any(|attr| is_faceted_by(field, attr));
+
         let distinct =
             self.distinct_attribute.as_ref().is_some_and(|distinct_field| field == distinct_field);
         let asc_desc = self.asc_desc_attributes.contains(field);
@@ -343,6 +356,7 @@ impl MetadataBuilder {
 
         Metadata {
             searchable,
+            exact,
             sortable,
             distinct,
             asc_desc,

crates/milli/src/index.rs

@@ -281,6 +281,9 @@ impl Index {
                 &mut wtxn,
                 (constants::VERSION_MAJOR, constants::VERSION_MINOR, constants::VERSION_PATCH),
             )?;
+            // The database before v1.29 defaulted to using arroy, so we
+            // need to set it explicitly because the new default is hannoy.
+            this.put_vector_store(&mut wtxn, VectorStoreBackend::Hannoy)?;
         }
         wtxn.commit()?;
 

crates/milli/src/search/mod.rs

@@ -385,9 +385,10 @@ pub struct SearchResult {
     pub query_vector: Option<Embedding>,
 }
 
-#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+#[derive(Default, Debug, Clone, Copy, PartialEq, Eq)]
 pub enum TermsMatchingStrategy {
     // remove last word first
+    #[default]
     Last,
     // all words are mandatory
     All,
@@ -395,12 +396,6 @@ pub enum TermsMatchingStrategy {
     Frequency,
 }
 
-impl Default for TermsMatchingStrategy {
-    fn default() -> Self {
-        Self::Last
-    }
-}
-
 impl From<MatchingStrategy> for TermsMatchingStrategy {
     fn from(other: MatchingStrategy) -> Self {
         match other {

crates/milli/src/update/index_documents/helpers/grenad_helpers.rs

@@ -124,7 +124,7 @@ impl GrenadParameters {
     /// This should be called inside of a rayon thread pool,
     /// otherwise, it will take the global number of threads.
     pub fn max_memory_by_thread(&self) -> Option<usize> {
-        self.max_memory.map(|max_memory| (max_memory / rayon::current_num_threads()))
+        self.max_memory.map(|max_memory| max_memory / rayon::current_num_threads())
     }
 }
 

crates/milli/src/update/index_documents/mod.rs

@@ -54,11 +54,12 @@ pub struct DocumentAdditionResult {
     pub number_of_documents: u64,
 }
 
-#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, Serialize, Deserialize)]
+#[derive(Default, Debug, Clone, Copy, PartialEq, Eq, Hash, Serialize, Deserialize)]
 #[non_exhaustive]
 pub enum IndexDocumentsMethod {
     /// Replace the previous document with the new one,
     /// removing all the already known attributes.
+    #[default]
     ReplaceDocuments,
 
     /// Merge the previous version of the document with the new version,
@@ -66,12 +67,6 @@ pub enum IndexDocumentsMethod {
     UpdateDocuments,
 }
 
-impl Default for IndexDocumentsMethod {
-    fn default() -> Self {
-        Self::ReplaceDocuments
-    }
-}
-
 pub struct IndexDocuments<'t, 'i, 'a, FP, FA> {
     wtxn: &'t mut heed::RwTxn<'i>,
     index: &'i Index,
@@ -806,6 +801,10 @@ mod tests {
     use crate::vector::db::IndexEmbeddingConfig;
     use crate::{all_obkv_to_json, db_snap, Filter, FilterableAttributesRule, Search, UserError};
 
+    fn no_cancel() -> bool {
+        false
+    }
+
     #[test]
     fn simple_document_replacement() {
         let index = TempIndex::new();
@@ -1985,7 +1984,7 @@ mod tests {
                 &rtxn,
                 None,
                 &mut new_fields_ids_map,
-                &|| false,
+                &no_cancel,
                 Progress::default(),
                 None,
             )
@@ -2038,7 +2037,7 @@ mod tests {
                 &rtxn,
                 None,
                 &mut new_fields_ids_map,
-                &|| false,
+                &no_cancel,
                 Progress::default(),
                 None,
             )
@@ -2057,7 +2056,7 @@ mod tests {
             primary_key,
             &document_changes,
             RuntimeEmbedders::default(),
-            &|| false,
+            &no_cancel,
             &Progress::default(),
             &Default::default(),
         )
@@ -2127,7 +2126,7 @@ mod tests {
                 &rtxn,
                 None,
                 &mut new_fields_ids_map,
-                &|| false,
+                &no_cancel,
                 Progress::default(),
                 None,
             )
@@ -2146,7 +2145,7 @@ mod tests {
             primary_key,
             &document_changes,
             RuntimeEmbedders::default(),
-            &|| false,
+            &no_cancel,
             &Progress::default(),
             &Default::default(),
         )
@@ -2317,7 +2316,7 @@ mod tests {
                 &rtxn,
                 None,
                 &mut new_fields_ids_map,
-                &|| false,
+                &no_cancel,
                 Progress::default(),
                 None,
             )
@@ -2333,7 +2332,7 @@ mod tests {
             primary_key,
             &document_changes,
             embedders,
-            &|| false,
+            &no_cancel,
             &Progress::default(),
             &Default::default(),
         )
@@ -2381,7 +2380,7 @@ mod tests {
                 &rtxn,
                 None,
                 &mut new_fields_ids_map,
-                &|| false,
+                &no_cancel,
                 Progress::default(),
                 None,
             )
@@ -2397,7 +2396,7 @@ mod tests {
             primary_key,
             &document_changes,
             embedders,
-            &|| false,
+            &no_cancel,
             &Progress::default(),
             &Default::default(),
         )
@@ -2436,7 +2435,7 @@ mod tests {
                 &rtxn,
                 None,
                 &mut new_fields_ids_map,
-                &|| false,
+                &no_cancel,
                 Progress::default(),
                 None,
             )
@@ -2452,7 +2451,7 @@ mod tests {
             primary_key,
             &document_changes,
             embedders,
-            &|| false,
+            &no_cancel,
             &Progress::default(),
             &Default::default(),
         )
@@ -2490,7 +2489,7 @@ mod tests {
                 &rtxn,
                 None,
                 &mut new_fields_ids_map,
-                &|| false,
+                &no_cancel,
                 Progress::default(),
                 None,
             )
@@ -2506,7 +2505,7 @@ mod tests {
             primary_key,
             &document_changes,
             embedders,
-            &|| false,
+            &no_cancel,
             &Progress::default(),
             &Default::default(),
         )
@@ -2546,7 +2545,7 @@ mod tests {
                 &rtxn,
                 None,
                 &mut new_fields_ids_map,
-                &|| false,
+                &no_cancel,
                 Progress::default(),
                 None,
             )
@@ -2562,7 +2561,7 @@ mod tests {
             primary_key,
             &document_changes,
             embedders,
-            &|| false,
+            &no_cancel,
             &Progress::default(),
             &Default::default(),
         )
@@ -2607,7 +2606,7 @@ mod tests {
                 &rtxn,
                 None,
                 &mut new_fields_ids_map,
-                &|| false,
+                &no_cancel,
                 Progress::default(),
                 None,
             )
@@ -2623,7 +2622,7 @@ mod tests {
             primary_key,
             &document_changes,
             embedders,
-            &|| false,
+            &no_cancel,
             &Progress::default(),
             &Default::default(),
         )
@@ -2661,7 +2660,7 @@ mod tests {
                 &rtxn,
                 None,
                 &mut new_fields_ids_map,
-                &|| false,
+                &no_cancel,
                 Progress::default(),
                 None,
             )
@@ -2677,7 +2676,7 @@ mod tests {
             primary_key,
             &document_changes,
             embedders,
-            &|| false,
+            &no_cancel,
             &Progress::default(),
             &Default::default(),
         )
@@ -2715,7 +2714,7 @@ mod tests {
                 &rtxn,
                 None,
                 &mut new_fields_ids_map,
-                &|| false,
+                &no_cancel,
                 Progress::default(),
                 None,
             )
@@ -2731,7 +2730,7 @@ mod tests {
             primary_key,
             &document_changes,
             embedders,
-            &|| false,
+            &no_cancel,
             &Progress::default(),
             &Default::default(),
         )
@@ -2927,7 +2926,7 @@ mod tests {
                 &rtxn,
                 None,
                 &mut new_fields_ids_map,
-                &|| false,
+                &no_cancel,
                 Progress::default(),
                 None,
             )
@@ -2943,7 +2942,7 @@ mod tests {
             primary_key,
             &document_changes,
             embedders,
-            &|| false,
+            &no_cancel,
             &Progress::default(),
             &Default::default(),
         )
@@ -2988,7 +2987,7 @@ mod tests {
                 &rtxn,
                 None,
                 &mut new_fields_ids_map,
-                &|| false,
+                &no_cancel,
                 Progress::default(),
                 None,
             )
@@ -3004,7 +3003,7 @@ mod tests {
             primary_key,
             &document_changes,
             embedders,
-            &|| false,
+            &no_cancel,
             &Progress::default(),
             &Default::default(),
         )
@@ -3046,7 +3045,7 @@ mod tests {
                 &rtxn,
                 None,
                 &mut new_fields_ids_map,
-                &|| false,
+                &no_cancel,
                 Progress::default(),
                 None,
             )
@@ -3062,7 +3061,7 @@ mod tests {
             primary_key,
             &document_changes,
             embedders,
-            &|| false,
+            &no_cancel,
             &Progress::default(),
             &Default::default(),
         )

crates/milli/src/update/new/extract/searchable/extract_word_docids.rs

@@ -8,17 +8,26 @@ use bumpalo::Bump;
 
 use super::match_searchable_field;
 use super::tokenize_document::{tokenizer_builder, DocumentTokenizer};
+use crate::fields_ids_map::metadata::Metadata;
 use crate::update::new::document::DocumentContext;
 use crate::update::new::extract::cache::BalancedCaches;
 use crate::update::new::extract::perm_json_p::contained_in;
+use crate::update::new::extract::searchable::has_searchable_children;
 use crate::update::new::indexer::document_changes::{
     extract, DocumentChanges, Extractor, IndexingContext,
 };
+use crate::update::new::indexer::settings_changes::{
+    settings_change_extract, DocumentsIndentifiers, SettingsChangeExtractor,
+};
 use crate::update::new::ref_cell_ext::RefCellExt as _;
 use crate::update::new::steps::IndexingStep;
 use crate::update::new::thread_local::{FullySend, MostlySend, ThreadLocal};
-use crate::update::new::DocumentChange;
-use crate::{bucketed_position, DocumentId, FieldId, Result, MAX_POSITION_PER_ATTRIBUTE};
+use crate::update::new::{DocumentChange, DocumentIdentifiers};
+use crate::update::settings::SettingsDelta;
+use crate::{
+    bucketed_position, DocumentId, FieldId, PatternMatch, Result, UserError,
+    MAX_POSITION_PER_ATTRIBUTE,
+};
 
 const MAX_COUNTED_WORDS: usize = 30;
 
@@ -34,6 +43,15 @@ pub struct WordDocidsBalancedCaches<'extractor> {
 
 unsafe impl MostlySend for WordDocidsBalancedCaches<'_> {}
 
+/// Whether to extract or skip fields during word extraction.
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+enum FieldDbExtraction {
+    /// Extract the word and put it in to the fid-based databases.
+    Extract,
+    /// Do not store the word in the fid-based databases.
+    Skip,
+}
+
 impl<'extractor> WordDocidsBalancedCaches<'extractor> {
     pub fn new_in(buckets: usize, max_memory: Option<usize>, alloc: &'extractor Bump) -> Self {
         Self {
@@ -47,12 +65,14 @@ impl<'extractor> WordDocidsBalancedCaches<'extractor> {
         }
     }
 
+    #[allow(clippy::too_many_arguments)]
     fn insert_add_u32(
         &mut self,
         field_id: FieldId,
         position: u16,
         word: &str,
         exact: bool,
+        field_db_extraction: FieldDbExtraction,
         docid: u32,
         bump: &Bump,
     ) -> Result<()> {
@@ -66,11 +86,13 @@ impl<'extractor> WordDocidsBalancedCaches<'extractor> {
         let buffer_size = word_bytes.len() + 1 + size_of::<FieldId>();
         let mut buffer = BumpVec::with_capacity_in(buffer_size, bump);
 
-        buffer.clear();
-        buffer.extend_from_slice(word_bytes);
-        buffer.push(0);
-        buffer.extend_from_slice(&field_id.to_be_bytes());
-        self.word_fid_docids.insert_add_u32(&buffer, docid)?;
+        if field_db_extraction == FieldDbExtraction::Extract {
+            buffer.clear();
+            buffer.extend_from_slice(word_bytes);
+            buffer.push(0);
+            buffer.extend_from_slice(&field_id.to_be_bytes());
+            self.word_fid_docids.insert_add_u32(&buffer, docid)?;
+        }
 
         let position = bucketed_position(position);
         buffer.clear();
@@ -83,21 +105,26 @@ impl<'extractor> WordDocidsBalancedCaches<'extractor> {
             self.flush_fid_word_count(&mut buffer)?;
         }
 
-        self.fid_word_count
-            .entry(field_id)
-            .and_modify(|(_current_count, new_count)| *new_count.get_or_insert(0) += 1)
-            .or_insert((None, Some(1)));
+        if field_db_extraction == FieldDbExtraction::Extract {
+            self.fid_word_count
+                .entry(field_id)
+                .and_modify(|(_current_count, new_count)| *new_count.get_or_insert(0) += 1)
+                .or_insert((None, Some(1)));
+        }
+
         self.current_docid = Some(docid);
 
         Ok(())
     }
 
+    #[allow(clippy::too_many_arguments)]
     fn insert_del_u32(
         &mut self,
         field_id: FieldId,
         position: u16,
         word: &str,
         exact: bool,
+        field_db_extraction: FieldDbExtraction,
         docid: u32,
         bump: &Bump,
     ) -> Result<()> {
@@ -111,11 +138,13 @@ impl<'extractor> WordDocidsBalancedCaches<'extractor> {
         let buffer_size = word_bytes.len() + 1 + size_of::<FieldId>();
         let mut buffer = BumpVec::with_capacity_in(buffer_size, bump);
 
-        buffer.clear();
-        buffer.extend_from_slice(word_bytes);
-        buffer.push(0);
-        buffer.extend_from_slice(&field_id.to_be_bytes());
-        self.word_fid_docids.insert_del_u32(&buffer, docid)?;
+        if field_db_extraction == FieldDbExtraction::Extract {
+            buffer.clear();
+            buffer.extend_from_slice(word_bytes);
+            buffer.push(0);
+            buffer.extend_from_slice(&field_id.to_be_bytes());
+            self.word_fid_docids.insert_del_u32(&buffer, docid)?;
+        }
 
         let position = bucketed_position(position);
         buffer.clear();
@@ -128,10 +157,12 @@ impl<'extractor> WordDocidsBalancedCaches<'extractor> {
             self.flush_fid_word_count(&mut buffer)?;
         }
 
-        self.fid_word_count
-            .entry(field_id)
-            .and_modify(|(current_count, _new_count)| *current_count.get_or_insert(0) += 1)
-            .or_insert((Some(1), None));
+        if field_db_extraction == FieldDbExtraction::Extract {
+            self.fid_word_count
+                .entry(field_id)
+                .and_modify(|(current_count, _new_count)| *current_count.get_or_insert(0) += 1)
+                .or_insert((Some(1), None));
+        }
 
         self.current_docid = Some(docid);
 
@@ -325,6 +356,24 @@ impl WordDocidsExtractors {
             exact_attributes.iter().any(|attr| contained_in(fname, attr))
                 || disabled_typos_terms.is_exact(word)
         };
+
+        let mut should_tokenize = |field_name: &str| {
+            let Some((field_id, meta)) = new_fields_ids_map.id_with_metadata_or_insert(field_name)
+            else {
+                return Err(UserError::AttributeLimitReached.into());
+            };
+
+            let pattern_match = if meta.is_searchable() {
+                PatternMatch::Match
+            } else {
+                // TODO: should be a match on the field_name using `match_field_legacy` function,
+                //       but for legacy reasons we iterate over all the fields to fill the field_id_map.
+                PatternMatch::Parent
+            };
+
+            Ok((field_id, pattern_match))
+        };
+
         match document_change {
             DocumentChange::Deletion(inner) => {
                 let mut token_fn = |fname: &str, fid, pos, word: &str| {
@@ -333,13 +382,14 @@ impl WordDocidsExtractors {
                         pos,
                         word,
                         is_exact(fname, word),
+                        FieldDbExtraction::Extract,
                         inner.docid(),
                         doc_alloc,
                     )
                 };
                 document_tokenizer.tokenize_document(
                     inner.current(rtxn, index, context.db_fields_ids_map)?,
-                    new_fields_ids_map,
+                    &mut should_tokenize,
                     &mut token_fn,
                 )?;
             }
@@ -361,13 +411,14 @@ impl WordDocidsExtractors {
                         pos,
                         word,
                         is_exact(fname, word),
+                        FieldDbExtraction::Extract,
                         inner.docid(),
                         doc_alloc,
                     )
                 };
                 document_tokenizer.tokenize_document(
                     inner.current(rtxn, index, context.db_fields_ids_map)?,
-                    new_fields_ids_map,
+                    &mut should_tokenize,
                     &mut token_fn,
                 )?;
 
@@ -377,13 +428,14 @@ impl WordDocidsExtractors {
                         pos,
                         word,
                         is_exact(fname, word),
+                        FieldDbExtraction::Extract,
                         inner.docid(),
                         doc_alloc,
                     )
                 };
                 document_tokenizer.tokenize_document(
                     inner.merged(rtxn, index, context.db_fields_ids_map)?,
-                    new_fields_ids_map,
+                    &mut should_tokenize,
                     &mut token_fn,
                 )?;
             }
@@ -394,13 +446,14 @@ impl WordDocidsExtractors {
                         pos,
                         word,
                         is_exact(fname, word),
+                        FieldDbExtraction::Extract,
                         inner.docid(),
                         doc_alloc,
                     )
                 };
                 document_tokenizer.tokenize_document(
                     inner.inserted(),
-                    new_fields_ids_map,
+                    &mut should_tokenize,
                     &mut token_fn,
                 )?;
             }
@@ -411,3 +464,292 @@ impl WordDocidsExtractors {
         cached_sorter.flush_fid_word_count(&mut buffer)
     }
 }
+
+pub struct WordDocidsSettingsExtractorsData<'a, SD> {
+    tokenizer: DocumentTokenizer<'a>,
+    max_memory_by_thread: Option<usize>,
+    buckets: usize,
+    settings_delta: &'a SD,
+}
+
+impl<'extractor, SD: SettingsDelta + Sync> SettingsChangeExtractor<'extractor>
+    for WordDocidsSettingsExtractorsData<'_, SD>
+{
+    type Data = RefCell<Option<WordDocidsBalancedCaches<'extractor>>>;
+
+    fn init_data<'doc>(&'doc self, extractor_alloc: &'extractor Bump) -> crate::Result<Self::Data> {
+        Ok(RefCell::new(Some(WordDocidsBalancedCaches::new_in(
+            self.buckets,
+            self.max_memory_by_thread,
+            extractor_alloc,
+        ))))
+    }
+
+    fn process<'doc>(
+        &'doc self,
+        documents: impl Iterator<Item = crate::Result<DocumentIdentifiers<'doc>>>,
+        context: &'doc DocumentContext<Self::Data>,
+    ) -> crate::Result<()> {
+        for document in documents {
+            let document = document?;
+            SettingsChangeWordDocidsExtractors::extract_document_from_settings_change(
+                document,
+                context,
+                &self.tokenizer,
+                self.settings_delta,
+            )?;
+        }
+        Ok(())
+    }
+}
+
+pub struct SettingsChangeWordDocidsExtractors;
+
+impl SettingsChangeWordDocidsExtractors {
+    pub fn run_extraction<'fid, 'indexer, 'index, 'extractor, SD, MSP>(
+        settings_delta: &SD,
+        documents: &'indexer DocumentsIndentifiers<'indexer>,
+        indexing_context: IndexingContext<'fid, 'indexer, 'index, MSP>,
+        extractor_allocs: &'extractor mut ThreadLocal<FullySend<Bump>>,
+        step: IndexingStep,
+    ) -> Result<WordDocidsCaches<'extractor>>
+    where
+        SD: SettingsDelta + Sync,
+        MSP: Fn() -> bool + Sync,
+    {
+        // Warning: this is duplicated code from extract_word_pair_proximity_docids.rs
+        // TODO we need to read the new AND old settings to support changing global parameters
+        let rtxn = indexing_context.index.read_txn()?;
+        let stop_words = indexing_context.index.stop_words(&rtxn)?;
+        let allowed_separators = indexing_context.index.allowed_separators(&rtxn)?;
+        let allowed_separators: Option<Vec<_>> =
+            allowed_separators.as_ref().map(|s| s.iter().map(String::as_str).collect());
+        let dictionary = indexing_context.index.dictionary(&rtxn)?;
+        let dictionary: Option<Vec<_>> =
+            dictionary.as_ref().map(|s| s.iter().map(String::as_str).collect());
+        let mut builder = tokenizer_builder(
+            stop_words.as_ref(),
+            allowed_separators.as_deref(),
+            dictionary.as_deref(),
+        );
+        let tokenizer = builder.build();
+        let localized_attributes_rules =
+            indexing_context.index.localized_attributes_rules(&rtxn)?.unwrap_or_default();
+        let document_tokenizer = DocumentTokenizer {
+            tokenizer: &tokenizer,
+            localized_attributes_rules: &localized_attributes_rules,
+            max_positions_per_attributes: MAX_POSITION_PER_ATTRIBUTE,
+        };
+        let extractor_data = WordDocidsSettingsExtractorsData {
+            tokenizer: document_tokenizer,
+            max_memory_by_thread: indexing_context.grenad_parameters.max_memory_by_thread(),
+            buckets: rayon::current_num_threads(),
+            settings_delta,
+        };
+        let datastore = ThreadLocal::new();
+        {
+            let span = tracing::debug_span!(target: "indexing::documents::extract", "vectors");
+            let _entered = span.enter();
+
+            settings_change_extract(
+                documents,
+                &extractor_data,
+                indexing_context,
+                extractor_allocs,
+                &datastore,
+                step,
+            )?;
+        }
+
+        let mut merger = WordDocidsCaches::new();
+        for cache in datastore.into_iter().flat_map(RefCell::into_inner) {
+            merger.push(cache)?;
+        }
+
+        Ok(merger)
+    }
+
+    /// Extracts document words from a settings change.
+    fn extract_document_from_settings_change<SD: SettingsDelta>(
+        document: DocumentIdentifiers<'_>,
+        context: &DocumentContext<RefCell<Option<WordDocidsBalancedCaches>>>,
+        document_tokenizer: &DocumentTokenizer,
+        settings_delta: &SD,
+    ) -> Result<()> {
+        let mut cached_sorter_ref = context.data.borrow_mut_or_yield();
+        let cached_sorter = cached_sorter_ref.as_mut().unwrap();
+        let doc_alloc = &context.doc_alloc;
+
+        let new_fields_ids_map = settings_delta.new_fields_ids_map();
+        let old_fields_ids_map = context.index.fields_ids_map_with_metadata(&context.rtxn)?;
+        let old_searchable = settings_delta.old_searchable_attributes().as_ref();
+        let new_searchable = settings_delta.new_searchable_attributes().as_ref();
+
+        let current_document = document.current(
+            &context.rtxn,
+            context.index,
+            old_fields_ids_map.as_fields_ids_map(),
+        )?;
+
+        #[derive(Debug, Clone, Copy, PartialEq)]
+        enum ActionToOperate {
+            ReindexAllFields,
+            // TODO improve by listing field prefixes
+            IndexAddedFields,
+            SkipDocument,
+        }
+
+        let mut action = ActionToOperate::SkipDocument;
+        // Here we do a preliminary check to determine the action to take.
+        // This check doesn't trigger the tokenizer as we never return
+        // PatternMatch::Match.
+        document_tokenizer.tokenize_document(
+            current_document,
+            &mut |field_name| {
+                let fid = new_fields_ids_map.id(field_name).expect("All fields IDs must exist");
+
+                // If the document must be reindexed, early return NoMatch to stop the scanning process.
+                if action == ActionToOperate::ReindexAllFields {
+                    return Ok((fid, PatternMatch::NoMatch));
+                }
+
+                let old_field_metadata = old_fields_ids_map.metadata(fid).unwrap();
+                let new_field_metadata = new_fields_ids_map.metadata(fid).unwrap();
+
+                action = match (old_field_metadata, new_field_metadata) {
+                    // At least one field is added or removed from the exact fields => ReindexAllFields
+                    (Metadata { exact: old_exact, .. }, Metadata { exact: new_exact, .. })
+                        if old_exact != new_exact =>
+                    {
+                        ActionToOperate::ReindexAllFields
+                    }
+                    // At least one field is removed from the searchable fields => ReindexAllFields
+                    (Metadata { searchable: Some(_), .. }, Metadata { searchable: None, .. }) => {
+                        ActionToOperate::ReindexAllFields
+                    }
+                    // At least one field is added in the searchable fields => IndexAddedFields
+                    (Metadata { searchable: None, .. }, Metadata { searchable: Some(_), .. }) => {
+                        // We can safely overwrite the action, because we early return when action is ReindexAllFields.
+                        ActionToOperate::IndexAddedFields
+                    }
+                    _ => action,
+                };
+
+                Ok((fid, PatternMatch::Parent))
+            },
+            &mut |_, _, _, _| Ok(()),
+        )?;
+
+        // Early return when we don't need to index the document
+        if action == ActionToOperate::SkipDocument {
+            return Ok(());
+        }
+
+        let mut should_tokenize = |field_name: &str| {
+            let field_id = new_fields_ids_map.id(field_name).expect("All fields IDs must exist");
+            let old_field_metadata = old_fields_ids_map.metadata(field_id).unwrap();
+            let new_field_metadata = new_fields_ids_map.metadata(field_id).unwrap();
+
+            let pattern_match = match action {
+                ActionToOperate::ReindexAllFields => {
+                    if old_field_metadata.is_searchable() || new_field_metadata.is_searchable() {
+                        PatternMatch::Match
+                    // If any old or new field is searchable then we need to iterate over all fields
+                    // else if any field matches we need to iterate over all fields
+                    } else if has_searchable_children(
+                        field_name,
+                        old_searchable.zip(new_searchable).map(|(old, new)| old.iter().chain(new)),
+                    ) {
+                        PatternMatch::Parent
+                    } else {
+                        PatternMatch::NoMatch
+                    }
+                }
+                ActionToOperate::IndexAddedFields => {
+                    // Was not searchable but now is
+                    if !old_field_metadata.is_searchable() && new_field_metadata.is_searchable() {
+                        PatternMatch::Match
+                    // If the field is now a parent of a searchable field
+                    } else if has_searchable_children(field_name, new_searchable) {
+                        PatternMatch::Parent
+                    } else {
+                        PatternMatch::NoMatch
+                    }
+                }
+                ActionToOperate::SkipDocument => unreachable!(),
+            };
+
+            Ok((field_id, pattern_match))
+        };
+
+        let old_disabled_typos_terms = settings_delta.old_disabled_typos_terms();
+        let new_disabled_typos_terms = settings_delta.new_disabled_typos_terms();
+        let mut token_fn = |_field_name: &str, field_id, pos, word: &str| {
+            let old_field_metadata = old_fields_ids_map.metadata(field_id).unwrap();
+            let new_field_metadata = new_fields_ids_map.metadata(field_id).unwrap();
+
+            match (old_field_metadata, new_field_metadata) {
+                (
+                    Metadata { searchable: Some(_), exact: old_exact, .. },
+                    Metadata { searchable: None, .. },
+                ) => cached_sorter.insert_del_u32(
+                    field_id,
+                    pos,
+                    word,
+                    old_exact || old_disabled_typos_terms.is_exact(word),
+                    // We deleted the field globally
+                    FieldDbExtraction::Skip,
+                    document.docid(),
+                    doc_alloc,
+                ),
+                (
+                    Metadata { searchable: None, .. },
+                    Metadata { searchable: Some(_), exact: new_exact, .. },
+                ) => cached_sorter.insert_add_u32(
+                    field_id,
+                    pos,
+                    word,
+                    new_exact || new_disabled_typos_terms.is_exact(word),
+                    FieldDbExtraction::Extract,
+                    document.docid(),
+                    doc_alloc,
+                ),
+                (Metadata { searchable: None, .. }, Metadata { searchable: None, .. }) => {
+                    unreachable!()
+                }
+                (Metadata { exact: old_exact, .. }, Metadata { exact: new_exact, .. }) => {
+                    cached_sorter.insert_del_u32(
+                        field_id,
+                        pos,
+                        word,
+                        old_exact || old_disabled_typos_terms.is_exact(word),
+                        // The field has already been extracted
+                        FieldDbExtraction::Skip,
+                        document.docid(),
+                        doc_alloc,
+                    )?;
+                    cached_sorter.insert_add_u32(
+                        field_id,
+                        pos,
+                        word,
+                        new_exact || new_disabled_typos_terms.is_exact(word),
+                        // The field has already been extracted
+                        FieldDbExtraction::Skip,
+                        document.docid(),
+                        doc_alloc,
+                    )
+                }
+            }
+        };
+
+        // TODO we must tokenize twice when we change global parameters like stop words,
+        //      the language settings, dictionary, separators, non-separators...
+        document_tokenizer.tokenize_document(
+            current_document,
+            &mut should_tokenize,
+            &mut token_fn,
+        )?;
+
+        Ok(())
+    }
+}

crates/milli/src/update/new/extract/searchable/extract_word_pair_proximity_docids.rs

@@ -6,17 +6,24 @@ use bumpalo::Bump;
 
 use super::match_searchable_field;
 use super::tokenize_document::{tokenizer_builder, DocumentTokenizer};
+use crate::fields_ids_map::metadata::Metadata;
+use crate::proximity::ProximityPrecision::*;
 use crate::proximity::{index_proximity, MAX_DISTANCE};
 use crate::update::new::document::{Document, DocumentContext};
 use crate::update::new::extract::cache::BalancedCaches;
 use crate::update::new::indexer::document_changes::{
     extract, DocumentChanges, Extractor, IndexingContext,
 };
+use crate::update::new::indexer::settings_change_extract;
+use crate::update::new::indexer::settings_changes::{
+    DocumentsIndentifiers, SettingsChangeExtractor,
+};
 use crate::update::new::ref_cell_ext::RefCellExt as _;
 use crate::update::new::steps::IndexingStep;
 use crate::update::new::thread_local::{FullySend, ThreadLocal};
-use crate::update::new::DocumentChange;
-use crate::{FieldId, GlobalFieldsIdsMap, Result, MAX_POSITION_PER_ATTRIBUTE};
+use crate::update::new::{DocumentChange, DocumentIdentifiers};
+use crate::update::settings::SettingsDelta;
+use crate::{FieldId, PatternMatch, Result, UserError, MAX_POSITION_PER_ATTRIBUTE};
 
 pub struct WordPairProximityDocidsExtractorData<'a> {
     tokenizer: DocumentTokenizer<'a>,
@@ -116,7 +123,7 @@ impl WordPairProximityDocidsExtractor {
     // and to store the docids of the documents that have a number of words in a given field
     // equal to or under than MAX_COUNTED_WORDS.
     fn extract_document_change(
-        context: &DocumentContext<RefCell<BalancedCaches>>,
+        context: &DocumentContext<RefCell<BalancedCaches<'_>>>,
         document_tokenizer: &DocumentTokenizer,
         searchable_attributes: Option<&[&str]>,
         document_change: DocumentChange,
@@ -147,8 +154,12 @@ impl WordPairProximityDocidsExtractor {
                 process_document_tokens(
                     document,
                     document_tokenizer,
-                    new_fields_ids_map,
                     &mut word_positions,
+                    &mut |field_name| {
+                        new_fields_ids_map
+                            .id_with_metadata_or_insert(field_name)
+                            .ok_or(UserError::AttributeLimitReached.into())
+                    },
                     &mut |(w1, w2), prox| {
                         del_word_pair_proximity.push(((w1, w2), prox));
                     },
@@ -170,8 +181,12 @@ impl WordPairProximityDocidsExtractor {
                 process_document_tokens(
                     document,
                     document_tokenizer,
-                    new_fields_ids_map,
                     &mut word_positions,
+                    &mut |field_name| {
+                        new_fields_ids_map
+                            .id_with_metadata_or_insert(field_name)
+                            .ok_or(UserError::AttributeLimitReached.into())
+                    },
                     &mut |(w1, w2), prox| {
                         del_word_pair_proximity.push(((w1, w2), prox));
                     },
@@ -180,8 +195,12 @@ impl WordPairProximityDocidsExtractor {
                 process_document_tokens(
                     document,
                     document_tokenizer,
-                    new_fields_ids_map,
                     &mut word_positions,
+                    &mut |field_name| {
+                        new_fields_ids_map
+                            .id_with_metadata_or_insert(field_name)
+                            .ok_or(UserError::AttributeLimitReached.into())
+                    },
                     &mut |(w1, w2), prox| {
                         add_word_pair_proximity.push(((w1, w2), prox));
                     },
@@ -192,8 +211,12 @@ impl WordPairProximityDocidsExtractor {
                 process_document_tokens(
                     document,
                     document_tokenizer,
-                    new_fields_ids_map,
                     &mut word_positions,
+                    &mut |field_name| {
+                        new_fields_ids_map
+                            .id_with_metadata_or_insert(field_name)
+                            .ok_or(UserError::AttributeLimitReached.into())
+                    },
                     &mut |(w1, w2), prox| {
                         add_word_pair_proximity.push(((w1, w2), prox));
                     },
@@ -257,8 +280,8 @@ fn drain_word_positions(
 fn process_document_tokens<'doc>(
     document: impl Document<'doc>,
     document_tokenizer: &DocumentTokenizer,
-    fields_ids_map: &mut GlobalFieldsIdsMap,
     word_positions: &mut VecDeque<(Rc<str>, u16)>,
+    field_id_and_metadata: &mut impl FnMut(&str) -> Result<(FieldId, Metadata)>,
     word_pair_proximity: &mut impl FnMut((Rc<str>, Rc<str>), u8),
 ) -> Result<()> {
     let mut field_id = None;
@@ -279,8 +302,248 @@ fn process_document_tokens<'doc>(
         word_positions.push_back((Rc::from(word), pos));
         Ok(())
     };
-    document_tokenizer.tokenize_document(document, fields_ids_map, &mut token_fn)?;
+
+    let mut should_tokenize = |field_name: &str| {
+        let (field_id, meta) = field_id_and_metadata(field_name)?;
+
+        let pattern_match = if meta.is_searchable() {
+            PatternMatch::Match
+        } else {
+            // TODO: should be a match on the field_name using `match_field_legacy` function,
+            //       but for legacy reasons we iterate over all the fields to fill the field_id_map.
+            PatternMatch::Parent
+        };
+
+        Ok((field_id, pattern_match))
+    };
+
+    document_tokenizer.tokenize_document(document, &mut should_tokenize, &mut token_fn)?;
 
     drain_word_positions(word_positions, word_pair_proximity);
     Ok(())
 }
+
+pub struct WordPairProximityDocidsSettingsExtractorsData<'a, SD> {
+    tokenizer: DocumentTokenizer<'a>,
+    max_memory_by_thread: Option<usize>,
+    buckets: usize,
+    settings_delta: &'a SD,
+}
+
+impl<'extractor, SD: SettingsDelta + Sync> SettingsChangeExtractor<'extractor>
+    for WordPairProximityDocidsSettingsExtractorsData<'_, SD>
+{
+    type Data = RefCell<BalancedCaches<'extractor>>;
+
+    fn init_data<'doc>(&'doc self, extractor_alloc: &'extractor Bump) -> crate::Result<Self::Data> {
+        Ok(RefCell::new(BalancedCaches::new_in(
+            self.buckets,
+            self.max_memory_by_thread,
+            extractor_alloc,
+        )))
+    }
+
+    fn process<'doc>(
+        &'doc self,
+        documents: impl Iterator<Item = crate::Result<DocumentIdentifiers<'doc>>>,
+        context: &'doc DocumentContext<Self::Data>,
+    ) -> crate::Result<()> {
+        for document in documents {
+            let document = document?;
+            SettingsChangeWordPairProximityDocidsExtractors::extract_document_from_settings_change(
+                document,
+                context,
+                &self.tokenizer,
+                self.settings_delta,
+            )?;
+        }
+        Ok(())
+    }
+}
+
+pub struct SettingsChangeWordPairProximityDocidsExtractors;
+
+impl SettingsChangeWordPairProximityDocidsExtractors {
+    pub fn run_extraction<'fid, 'indexer, 'index, 'extractor, SD, MSP>(
+        settings_delta: &SD,
+        documents: &'indexer DocumentsIndentifiers<'indexer>,
+        indexing_context: IndexingContext<'fid, 'indexer, 'index, MSP>,
+        extractor_allocs: &'extractor mut ThreadLocal<FullySend<Bump>>,
+        step: IndexingStep,
+    ) -> Result<Vec<BalancedCaches<'extractor>>>
+    where
+        SD: SettingsDelta + Sync,
+        MSP: Fn() -> bool + Sync,
+    {
+        // Warning: this is duplicated code from extract_word_docids.rs
+        let rtxn = indexing_context.index.read_txn()?;
+        let stop_words = indexing_context.index.stop_words(&rtxn)?;
+        let allowed_separators = indexing_context.index.allowed_separators(&rtxn)?;
+        let allowed_separators: Option<Vec<_>> =
+            allowed_separators.as_ref().map(|s| s.iter().map(String::as_str).collect());
+        let dictionary = indexing_context.index.dictionary(&rtxn)?;
+        let dictionary: Option<Vec<_>> =
+            dictionary.as_ref().map(|s| s.iter().map(String::as_str).collect());
+        let mut builder = tokenizer_builder(
+            stop_words.as_ref(),
+            allowed_separators.as_deref(),
+            dictionary.as_deref(),
+        );
+        let tokenizer = builder.build();
+        let localized_attributes_rules =
+            indexing_context.index.localized_attributes_rules(&rtxn)?.unwrap_or_default();
+        let document_tokenizer = DocumentTokenizer {
+            tokenizer: &tokenizer,
+            localized_attributes_rules: &localized_attributes_rules,
+            max_positions_per_attributes: MAX_POSITION_PER_ATTRIBUTE,
+        };
+        let extractor_data = WordPairProximityDocidsSettingsExtractorsData {
+            tokenizer: document_tokenizer,
+            max_memory_by_thread: indexing_context.grenad_parameters.max_memory_by_thread(),
+            buckets: rayon::current_num_threads(),
+            settings_delta,
+        };
+        let datastore = ThreadLocal::new();
+        {
+            let span = tracing::trace_span!(target: "indexing::documents::extract", "word_pair_proximity_docids_extraction");
+            let _entered = span.enter();
+
+            settings_change_extract(
+                documents,
+                &extractor_data,
+                indexing_context,
+                extractor_allocs,
+                &datastore,
+                step,
+            )?;
+        }
+
+        Ok(datastore.into_iter().map(RefCell::into_inner).collect())
+    }
+
+    /// Extracts document words from a settings change.
+    fn extract_document_from_settings_change<SD: SettingsDelta>(
+        document: DocumentIdentifiers<'_>,
+        context: &DocumentContext<RefCell<BalancedCaches<'_>>>,
+        document_tokenizer: &DocumentTokenizer,
+        settings_delta: &SD,
+    ) -> Result<()> {
+        let mut cached_sorter = context.data.borrow_mut_or_yield();
+        let doc_alloc = &context.doc_alloc;
+
+        let new_fields_ids_map = settings_delta.new_fields_ids_map();
+        let old_fields_ids_map = settings_delta.old_fields_ids_map();
+        let old_proximity_precision = *settings_delta.old_proximity_precision();
+        let new_proximity_precision = *settings_delta.new_proximity_precision();
+
+        let current_document = document.current(
+            &context.rtxn,
+            context.index,
+            old_fields_ids_map.as_fields_ids_map(),
+        )?;
+
+        #[derive(Debug, Clone, Copy, PartialEq)]
+        enum ActionToOperate {
+            ReindexAllFields,
+            SkipDocument,
+        }
+
+        // TODO prefix_fid delete_old_fid_based_databases
+        let mut action = match (old_proximity_precision, new_proximity_precision) {
+            (ByAttribute, ByWord) => ActionToOperate::ReindexAllFields,
+            (_, _) => ActionToOperate::SkipDocument,
+        };
+
+        // Here we do a preliminary check to determine the action to take.
+        // This check doesn't trigger the tokenizer as we never return
+        // PatternMatch::Match.
+        if action != ActionToOperate::ReindexAllFields {
+            document_tokenizer.tokenize_document(
+                current_document,
+                &mut |field_name| {
+                    let fid = new_fields_ids_map.id(field_name).expect("All fields IDs must exist");
+
+                    // If the document must be reindexed, early return NoMatch to stop the scanning process.
+                    if action == ActionToOperate::ReindexAllFields {
+                        return Ok((fid, PatternMatch::NoMatch));
+                    }
+
+                    let old_field_metadata = old_fields_ids_map.metadata(fid).unwrap();
+                    let new_field_metadata = new_fields_ids_map.metadata(fid).unwrap();
+
+                    action = match (old_field_metadata, new_field_metadata) {
+                        // At least one field is removed or added from the searchable fields
+                        (
+                            Metadata { searchable: Some(_), .. },
+                            Metadata { searchable: None, .. },
+                        )
+                        | (
+                            Metadata { searchable: None, .. },
+                            Metadata { searchable: Some(_), .. },
+                        ) => ActionToOperate::ReindexAllFields,
+                        _ => action,
+                    };
+
+                    Ok((fid, PatternMatch::Parent))
+                },
+                &mut |_, _, _, _| Ok(()),
+            )?;
+        }
+
+        // Early return when we don't need to index the document
+        if action == ActionToOperate::SkipDocument {
+            return Ok(());
+        }
+
+        let mut del_word_pair_proximity = bumpalo::collections::Vec::new_in(doc_alloc);
+        let mut add_word_pair_proximity = bumpalo::collections::Vec::new_in(doc_alloc);
+
+        // is a vecdequeue, and will be smol, so can stay on the heap for now
+        let mut word_positions: VecDeque<(Rc<str>, u16)> =
+            VecDeque::with_capacity(MAX_DISTANCE as usize);
+
+        process_document_tokens(
+            current_document,
+            // TODO Tokenize must be based on old settings
+            document_tokenizer,
+            &mut word_positions,
+            &mut |field_name| {
+                Ok(old_fields_ids_map.id_with_metadata(field_name).expect("All fields must exist"))
+            },
+            &mut |(w1, w2), prox| {
+                del_word_pair_proximity.push(((w1, w2), prox));
+            },
+        )?;
+
+        process_document_tokens(
+            current_document,
+            // TODO Tokenize must be based on new settings
+            document_tokenizer,
+            &mut word_positions,
+            &mut |field_name| {
+                Ok(new_fields_ids_map.id_with_metadata(field_name).expect("All fields must exist"))
+            },
+            &mut |(w1, w2), prox| {
+                add_word_pair_proximity.push(((w1, w2), prox));
+            },
+        )?;
+
+        let mut key_buffer = bumpalo::collections::Vec::new_in(doc_alloc);
+
+        del_word_pair_proximity.sort_unstable();
+        del_word_pair_proximity.dedup_by(|(k1, _), (k2, _)| k1 == k2);
+        for ((w1, w2), prox) in del_word_pair_proximity.iter() {
+            let key = build_key(*prox, w1, w2, &mut key_buffer);
+            cached_sorter.insert_del_u32(key, document.docid())?;
+        }
+
+        add_word_pair_proximity.sort_unstable();
+        add_word_pair_proximity.dedup_by(|(k1, _), (k2, _)| k1 == k2);
+        for ((w1, w2), prox) in add_word_pair_proximity.iter() {
+            let key = build_key(*prox, w1, w2, &mut key_buffer);
+            cached_sorter.insert_add_u32(key, document.docid())?;
+        }
+
+        Ok(())
+    }
+}

crates/milli/src/update/new/extract/searchable/mod.rs

@@ -2,8 +2,12 @@ mod extract_word_docids;
 mod extract_word_pair_proximity_docids;
 mod tokenize_document;
 
-pub use extract_word_docids::{WordDocidsCaches, WordDocidsExtractors};
-pub use extract_word_pair_proximity_docids::WordPairProximityDocidsExtractor;
+pub use extract_word_docids::{
+    SettingsChangeWordDocidsExtractors, WordDocidsCaches, WordDocidsExtractors,
+};
+pub use extract_word_pair_proximity_docids::{
+    SettingsChangeWordPairProximityDocidsExtractors, WordPairProximityDocidsExtractor,
+};
 
 use crate::attribute_patterns::{match_field_legacy, PatternMatch};
 
@@ -27,3 +31,17 @@ pub fn match_searchable_field(
 
     selection
 }
+
+/// return `true` if the provided `field_name` is a parent of at least one of the fields contained in `searchable`,
+/// or if `searchable` is `None`.
+fn has_searchable_children<I, A>(field_name: &str, searchable: Option<I>) -> bool
+where
+    I: IntoIterator<Item = A>,
+    A: AsRef<str>,
+{
+    searchable.is_none_or(|fields| {
+        fields
+            .into_iter()
+            .any(|attr| match_field_legacy(attr.as_ref(), field_name) == PatternMatch::Parent)
+    })
+}

crates/milli/src/update/new/extract/searchable/tokenize_document.rs

@@ -8,10 +8,7 @@ use crate::update::new::document::Document;
 use crate::update::new::extract::perm_json_p::{
     seek_leaf_values_in_array, seek_leaf_values_in_object, Depth,
 };
-use crate::{
-    FieldId, GlobalFieldsIdsMap, InternalError, LocalizedAttributesRule, Result, UserError,
-    MAX_WORD_LENGTH,
-};
+use crate::{FieldId, InternalError, LocalizedAttributesRule, Result, MAX_WORD_LENGTH};
 
 // todo: should be crate::proximity::MAX_DISTANCE but it has been forgotten
 const MAX_DISTANCE: u32 = 8;
@@ -26,26 +23,25 @@ impl DocumentTokenizer<'_> {
     pub fn tokenize_document<'doc>(
         &self,
         document: impl Document<'doc>,
-        field_id_map: &mut GlobalFieldsIdsMap,
+        should_tokenize: &mut impl FnMut(&str) -> Result<(FieldId, PatternMatch)>,
         token_fn: &mut impl FnMut(&str, FieldId, u16, &str) -> Result<()>,
     ) -> Result<()> {
         let mut field_position = HashMap::new();
-        let mut tokenize_field = |field_name: &str, _depth, value: &Value| {
-            let Some((field_id, meta)) = field_id_map.id_with_metadata_or_insert(field_name) else {
-                return Err(UserError::AttributeLimitReached.into());
-            };
-
-            if meta.is_searchable() {
-                self.tokenize_field(field_id, field_name, value, token_fn, &mut field_position)?;
-            }
-
-            // todo: should be a match on the field_name using `match_field_legacy` function,
-            // but for legacy reasons we iterate over all the fields to fill the field_id_map.
-            Ok(PatternMatch::Match)
-        };
-
         for entry in document.iter_top_level_fields() {
             let (field_name, value) = entry?;
+
+            if let (_, PatternMatch::NoMatch) = should_tokenize(field_name)? {
+                continue;
+            }
+
+            let mut tokenize_field = |field_name: &str, _depth, value: &Value| {
+                let (fid, pattern_match) = should_tokenize(field_name)?;
+                if pattern_match == PatternMatch::Match {
+                    self.tokenize_field(fid, field_name, value, token_fn, &mut field_position)?;
+                }
+                Ok(pattern_match)
+            };
+
             // parse json.
             match serde_json::to_value(value).map_err(InternalError::SerdeJson)? {
                 Value::Object(object) => seek_leaf_values_in_object(
@@ -192,7 +188,7 @@ mod test {
     use super::*;
     use crate::fields_ids_map::metadata::{FieldIdMapWithMetadata, MetadataBuilder};
     use crate::update::new::document::{DocumentFromVersions, Versions};
-    use crate::FieldsIdsMap;
+    use crate::{FieldsIdsMap, GlobalFieldsIdsMap, UserError};
 
     #[test]
     fn test_tokenize_document() {
@@ -231,6 +227,7 @@ mod test {
                 Default::default(),
                 Default::default(),
                 Default::default(),
+                Default::default(),
                 None,
                 None,
                 Default::default(),
@@ -251,15 +248,19 @@ mod test {
         let document = Versions::single(document);
         let document = DocumentFromVersions::new(&document);
 
+        let mut should_tokenize = |field_name: &str| {
+            let Some(field_id) = global_fields_ids_map.id_or_insert(field_name) else {
+                return Err(UserError::AttributeLimitReached.into());
+            };
+
+            Ok((field_id, PatternMatch::Match))
+        };
+
         document_tokenizer
-            .tokenize_document(
-                document,
-                &mut global_fields_ids_map,
-                &mut |_fname, fid, pos, word| {
-                    words.insert([fid, pos], word.to_string());
-                    Ok(())
-                },
-            )
+            .tokenize_document(document, &mut should_tokenize, &mut |_fname, fid, pos, word| {
+                words.insert([fid, pos], word.to_string());
+                Ok(())
+            })
             .unwrap();
 
         snapshot!(format!("{:#?}", words), @r###"

crates/milli/src/update/new/extract/vectors/mod.rs

@@ -1,5 +1,6 @@
 use std::cell::RefCell;
 use std::fmt::Debug;
+use std::sync::RwLock;
 
 use bumpalo::collections::Vec as BVec;
 use bumpalo::Bump;
@@ -27,7 +28,10 @@ use crate::vector::extractor::{
 use crate::vector::session::{EmbedSession, Input, Metadata, OnEmbed};
 use crate::vector::settings::ReindexAction;
 use crate::vector::{Embedding, RuntimeEmbedder, RuntimeEmbedders, RuntimeFragment};
-use crate::{DocumentId, FieldDistribution, InternalError, Result, ThreadPoolNoAbort, UserError};
+use crate::{
+    DocumentId, FieldDistribution, GlobalFieldsIdsMap, InternalError, Result, ThreadPoolNoAbort,
+    UserError,
+};
 
 pub struct EmbeddingExtractor<'a, 'b> {
     embedders: &'a RuntimeEmbedders,
@@ -321,6 +325,15 @@ impl<'extractor, SD: SettingsDelta + Sync> SettingsChangeExtractor<'extractor>
         let old_embedders = self.settings_delta.old_embedders();
         let unused_vectors_distribution = UnusedVectorsDistributionBump::new_in(&context.doc_alloc);
 
+        // We get a reference to the new and old fields ids maps but
+        // note that those are local versions where updates to them
+        // will not be reflected in the database. It's not an issue
+        // because new settings do not generate new fields.
+        let new_fields_ids_map = RwLock::new(self.settings_delta.new_fields_ids_map().clone());
+        let new_fields_ids_map = RefCell::new(GlobalFieldsIdsMap::new(&new_fields_ids_map));
+        let old_fields_ids_map = RwLock::new(self.settings_delta.old_fields_ids_map().clone());
+        let old_fields_ids_map = RefCell::new(GlobalFieldsIdsMap::new(&old_fields_ids_map));
+
         let mut all_chunks = BVec::with_capacity_in(embedders.len(), &context.doc_alloc);
         let embedder_configs = context.index.embedding_configs();
         for (embedder_name, action) in self.settings_delta.embedder_actions().iter() {
@@ -396,6 +409,7 @@ impl<'extractor, SD: SettingsDelta + Sync> SettingsChangeExtractor<'extractor>
                         if !must_regenerate {
                             continue;
                         }
+
                         // we need to regenerate the prompts for the document
                         chunks.settings_change_autogenerated(
                             document.docid(),
@@ -406,7 +420,8 @@ impl<'extractor, SD: SettingsDelta + Sync> SettingsChangeExtractor<'extractor>
                                 context.db_fields_ids_map,
                             )?,
                             self.settings_delta,
-                            context.new_fields_ids_map,
+                            &old_fields_ids_map,
+                            &new_fields_ids_map,
                             &unused_vectors_distribution,
                             old_is_user_provided,
                             fragments_changed,
@@ -442,7 +457,8 @@ impl<'extractor, SD: SettingsDelta + Sync> SettingsChangeExtractor<'extractor>
                                     context.db_fields_ids_map,
                                 )?,
                                 self.settings_delta,
-                                context.new_fields_ids_map,
+                                &old_fields_ids_map,
+                                &new_fields_ids_map,
                                 &unused_vectors_distribution,
                                 old_is_user_provided,
                                 true,
@@ -638,7 +654,8 @@ impl<'a, 'b, 'extractor> Chunks<'a, 'b, 'extractor> {
         external_docid: &'a str,
         document: D,
         settings_delta: &SD,
-        fields_ids_map: &'a RefCell<crate::GlobalFieldsIdsMap>,
+        old_fields_ids_map: &'a RefCell<GlobalFieldsIdsMap<'a>>,
+        new_fields_ids_map: &'a RefCell<GlobalFieldsIdsMap<'a>>,
         unused_vectors_distribution: &UnusedVectorsDistributionBump<'a>,
         old_is_user_provided: bool,
         full_reindex: bool,
@@ -733,10 +750,17 @@ impl<'a, 'b, 'extractor> Chunks<'a, 'b, 'extractor> {
                     old_embedder.as_ref().map(|old_embedder| &old_embedder.document_template)
                 };
 
-                let extractor =
-                    DocumentTemplateExtractor::new(document_template, doc_alloc, fields_ids_map);
+                let extractor = DocumentTemplateExtractor::new(
+                    document_template,
+                    doc_alloc,
+                    new_fields_ids_map,
+                );
                 let old_extractor = old_document_template.map(|old_document_template| {
-                    DocumentTemplateExtractor::new(old_document_template, doc_alloc, fields_ids_map)
+                    DocumentTemplateExtractor::new(
+                        old_document_template,
+                        doc_alloc,
+                        old_fields_ids_map,
+                    )
                 });
                 let metadata =
                     Metadata { docid, external_docid, extractor_id: extractor.extractor_id() };

crates/milli/src/update/new/indexer/extract.rs

@@ -372,11 +372,10 @@ where
     SD: SettingsDelta + Sync,
 {
     // Create the list of document ids to extract
-    let rtxn = indexing_context.index.read_txn()?;
-    let all_document_ids =
-        indexing_context.index.documents_ids(&rtxn)?.into_iter().collect::<Vec<_>>();
-    let primary_key =
-        primary_key_from_db(indexing_context.index, &rtxn, &indexing_context.db_fields_ids_map)?;
+    let index = indexing_context.index;
+    let rtxn = index.read_txn()?;
+    let all_document_ids = index.documents_ids(&rtxn)?.into_iter().collect::<Vec<_>>();
+    let primary_key = primary_key_from_db(index, &rtxn, &indexing_context.db_fields_ids_map)?;
     let documents = DocumentsIndentifiers::new(&all_document_ids, primary_key);
 
     let span =
@@ -391,6 +390,133 @@ where
         extractor_allocs,
     )?;
 
+    {
+        let WordDocidsCaches {
+            word_docids,
+            word_fid_docids,
+            exact_word_docids,
+            word_position_docids,
+            fid_word_count_docids,
+        } = {
+            let span = tracing::trace_span!(target: "indexing::documents::extract", "word_docids");
+            let _entered = span.enter();
+            SettingsChangeWordDocidsExtractors::run_extraction(
+                settings_delta,
+                &documents,
+                indexing_context,
+                extractor_allocs,
+                IndexingStep::ExtractingWords,
+            )?
+        };
+
+        indexing_context.progress.update_progress(IndexingStep::MergingWordCaches);
+
+        {
+            let span = tracing::trace_span!(target: "indexing::documents::merge", "word_docids");
+            let _entered = span.enter();
+            indexing_context.progress.update_progress(MergingWordCache::WordDocids);
+
+            merge_and_send_docids(
+                word_docids,
+                index.word_docids.remap_types(),
+                index,
+                extractor_sender.docids::<WordDocids>(),
+                &indexing_context.must_stop_processing,
+            )?;
+        }
+
+        {
+            let span =
+                tracing::trace_span!(target: "indexing::documents::merge", "word_fid_docids");
+            let _entered = span.enter();
+            indexing_context.progress.update_progress(MergingWordCache::WordFieldIdDocids);
+
+            merge_and_send_docids(
+                word_fid_docids,
+                index.word_fid_docids.remap_types(),
+                index,
+                extractor_sender.docids::<WordFidDocids>(),
+                &indexing_context.must_stop_processing,
+            )?;
+        }
+
+        {
+            let span =
+                tracing::trace_span!(target: "indexing::documents::merge", "exact_word_docids");
+            let _entered = span.enter();
+            indexing_context.progress.update_progress(MergingWordCache::ExactWordDocids);
+
+            merge_and_send_docids(
+                exact_word_docids,
+                index.exact_word_docids.remap_types(),
+                index,
+                extractor_sender.docids::<ExactWordDocids>(),
+                &indexing_context.must_stop_processing,
+            )?;
+        }
+
+        {
+            let span =
+                tracing::trace_span!(target: "indexing::documents::merge", "word_position_docids");
+            let _entered = span.enter();
+            indexing_context.progress.update_progress(MergingWordCache::WordPositionDocids);
+
+            merge_and_send_docids(
+                word_position_docids,
+                index.word_position_docids.remap_types(),
+                index,
+                extractor_sender.docids::<WordPositionDocids>(),
+                &indexing_context.must_stop_processing,
+            )?;
+        }
+
+        {
+            let span =
+                tracing::trace_span!(target: "indexing::documents::merge", "fid_word_count_docids");
+            let _entered = span.enter();
+            indexing_context.progress.update_progress(MergingWordCache::FieldIdWordCountDocids);
+
+            merge_and_send_docids(
+                fid_word_count_docids,
+                index.field_id_word_count_docids.remap_types(),
+                index,
+                extractor_sender.docids::<FidWordCountDocids>(),
+                &indexing_context.must_stop_processing,
+            )?;
+        }
+    }
+
+    // Run the proximity extraction only if the precision is ByWord.
+    let new_proximity_precision = settings_delta.new_proximity_precision();
+    if *new_proximity_precision == ProximityPrecision::ByWord {
+        let caches = {
+            let span = tracing::trace_span!(target: "indexing::documents::extract", "word_pair_proximity_docids");
+            let _entered = span.enter();
+
+            SettingsChangeWordPairProximityDocidsExtractors::run_extraction(
+                settings_delta,
+                &documents,
+                indexing_context,
+                extractor_allocs,
+                IndexingStep::ExtractingWordProximity,
+            )?
+        };
+
+        {
+            let span = tracing::trace_span!(target: "indexing::documents::merge", "word_pair_proximity_docids");
+            let _entered = span.enter();
+            indexing_context.progress.update_progress(IndexingStep::MergingWordProximity);
+
+            merge_and_send_docids(
+                caches,
+                index.word_pair_proximity_docids.remap_types(),
+                index,
+                extractor_sender.docids::<WordPairProximityDocids>(),
+                &indexing_context.must_stop_processing,
+            )?;
+        }
+    }
+
     'vectors: {
         if settings_delta.embedder_actions().is_empty() {
             break 'vectors;

crates/milli/src/update/new/indexer/mod.rs

@@ -1,4 +1,4 @@
-use std::collections::BTreeMap;
+use std::collections::{BTreeMap, BTreeSet};
 use std::sync::atomic::AtomicBool;
 use std::sync::{Arc, Once, RwLock};
 use std::thread::{self, Builder};
@@ -8,9 +8,11 @@ use document_changes::{DocumentChanges, IndexingContext};
 pub use document_deletion::DocumentDeletion;
 pub use document_operation::{DocumentOperation, PayloadStats};
 use hashbrown::HashMap;
-use heed::{RoTxn, RwTxn};
+use heed::types::DecodeIgnore;
+use heed::{BytesDecode, Database, RoTxn, RwTxn};
 pub use partial_dump::PartialDump;
 pub use post_processing::recompute_word_fst_from_word_docids_database;
+pub use settings_changes::settings_change_extract;
 pub use update_by_function::UpdateByFunction;
 pub use write::ChannelCongestion;
 use write::{build_vectors, update_index, write_to_db};
@@ -20,12 +22,18 @@ use super::steps::IndexingStep;
 use super::thread_local::ThreadLocal;
 use crate::documents::PrimaryKey;
 use crate::fields_ids_map::metadata::{FieldIdMapWithMetadata, MetadataBuilder};
+use crate::heed_codec::StrBEU16Codec;
 use crate::progress::{EmbedderStats, Progress};
+use crate::proximity::ProximityPrecision;
+use crate::update::new::steps::SettingsIndexerStep;
+use crate::update::new::FacetFieldIdsDelta;
 use crate::update::settings::SettingsDelta;
 use crate::update::GrenadParameters;
 use crate::vector::settings::{EmbedderAction, RemoveFragments, WriteBackToDocuments};
 use crate::vector::{Embedder, RuntimeEmbedders, VectorStore};
-use crate::{FieldsIdsMap, GlobalFieldsIdsMap, Index, InternalError, Result, ThreadPoolNoAbort};
+use crate::{
+    Error, FieldsIdsMap, GlobalFieldsIdsMap, Index, InternalError, Result, ThreadPoolNoAbort,
+};
 
 #[cfg(not(feature = "enterprise"))]
 pub mod community_edition;
@@ -242,6 +250,20 @@ where
     SD: SettingsDelta + Sync,
 {
     delete_old_embedders_and_fragments(wtxn, index, settings_delta)?;
+    delete_old_fid_based_databases(wtxn, index, settings_delta, must_stop_processing, progress)?;
+
+    // Clear word_pair_proximity if byWord to byAttribute
+    let old_proximity_precision = settings_delta.old_proximity_precision();
+    let new_proximity_precision = settings_delta.new_proximity_precision();
+    if *old_proximity_precision == ProximityPrecision::ByWord
+        && *new_proximity_precision == ProximityPrecision::ByAttribute
+    {
+        index.word_pair_proximity_docids.clear(wtxn)?;
+    }
+
+    // TODO delete useless searchable databases
+    //      - Clear fid_prefix_* in the post processing
+    //      - clear the prefix + fid_prefix if setting `PrefixSearch` is enabled
 
     let mut bbbuffers = Vec::new();
     let finished_extraction = AtomicBool::new(false);
@@ -300,6 +322,8 @@ where
                 .unwrap()
             })?;
 
+        let global_fields_ids_map = GlobalFieldsIdsMap::new(&new_fields_ids_map);
+
         let new_embedders = settings_delta.new_embedders();
         let embedder_actions = settings_delta.embedder_actions();
         let index_embedder_category_ids = settings_delta.new_embedder_category_id();
@@ -334,6 +358,18 @@ where
         })
         .unwrap()?;
 
+        pool.install(|| {
+            // WARN When implementing the facets don't forget this
+            let facet_field_ids_delta = FacetFieldIdsDelta::new(0, 0);
+            post_processing::post_process(
+                indexing_context,
+                wtxn,
+                global_fields_ids_map,
+                facet_field_ids_delta,
+            )
+        })
+        .unwrap()?;
+
         indexing_context.progress.update_progress(IndexingStep::BuildingGeoJson);
         index.cellulite.build(
             wtxn,
@@ -463,6 +499,106 @@ where
     Ok(())
 }
 
+/// Deletes entries refering the provided
+/// fids from the fid-based databases.
+fn delete_old_fid_based_databases<SD, MSP>(
+    wtxn: &mut RwTxn<'_>,
+    index: &Index,
+    settings_delta: &SD,
+    must_stop_processing: &MSP,
+    progress: &Progress,
+) -> Result<()>
+where
+    SD: SettingsDelta + Sync,
+    MSP: Fn() -> bool + Sync,
+{
+    let fids_to_delete: Option<BTreeSet<_>> = {
+        let rtxn = index.read_txn()?;
+        let fields_ids_map = index.fields_ids_map(&rtxn)?;
+        let old_searchable_attributes = settings_delta.old_searchable_attributes().as_ref();
+        let new_searchable_attributes = settings_delta.new_searchable_attributes().as_ref();
+        old_searchable_attributes.zip(new_searchable_attributes).map(|(old, new)| {
+            old.iter()
+                // Ignore the field if it is not searchable anymore
+                // or if it was never referenced in any document
+                .filter_map(|name| if new.contains(name) { None } else { fields_ids_map.id(name) })
+                .collect()
+        })
+    };
+
+    let Some(fids_to_delete) = fids_to_delete else {
+        return Ok(());
+    };
+
+    progress.update_progress(SettingsIndexerStep::DeletingOldWordFidDocids);
+    delete_old_word_fid_docids(wtxn, index.word_fid_docids, must_stop_processing, &fids_to_delete)?;
+
+    progress.update_progress(SettingsIndexerStep::DeletingOldFidWordCountDocids);
+    delete_old_fid_word_count_docids(wtxn, index, must_stop_processing, &fids_to_delete)?;
+
+    progress.update_progress(SettingsIndexerStep::DeletingOldWordPrefixFidDocids);
+    delete_old_word_fid_docids(
+        wtxn,
+        index.word_prefix_fid_docids,
+        must_stop_processing,
+        &fids_to_delete,
+    )?;
+
+    Ok(())
+}
+
+fn delete_old_word_fid_docids<'txn, MSP, DC>(
+    wtxn: &mut RwTxn<'txn>,
+    database: Database<StrBEU16Codec, DC>,
+    must_stop_processing: &MSP,
+    fids_to_delete: &BTreeSet<u16>,
+) -> Result<(), Error>
+where
+    MSP: Fn() -> bool + Sync,
+    DC: BytesDecode<'txn>,
+{
+    let mut iter = database.iter_mut(wtxn)?.remap_data_type::<DecodeIgnore>();
+    while let Some(((_word, fid), ())) = iter.next().transpose()? {
+        // TODO should I call it that often?
+        if must_stop_processing() {
+            return Err(Error::InternalError(InternalError::AbortedIndexation));
+        }
+
+        if fids_to_delete.contains(&fid) {
+            // safety: We don't keep any references to the data.
+            unsafe { iter.del_current()? };
+        }
+    }
+
+    Ok(())
+}
+
+fn delete_old_fid_word_count_docids<MSP>(
+    wtxn: &mut RwTxn<'_>,
+    index: &Index,
+    must_stop_processing: &MSP,
+    fids_to_delete: &BTreeSet<u16>,
+) -> Result<(), Error>
+where
+    MSP: Fn() -> bool + Sync,
+{
+    let db = index.field_id_word_count_docids.remap_data_type::<DecodeIgnore>();
+    for &fid_to_delete in fids_to_delete {
+        if must_stop_processing() {
+            return Err(Error::InternalError(InternalError::AbortedIndexation));
+        }
+
+        let mut iter = db.prefix_iter_mut(wtxn, &(fid_to_delete, 0))?;
+        while let Some(((fid, _word_count), ())) = iter.next().transpose()? {
+            debug_assert_eq!(fid, fid_to_delete);
+            // safety: We don't keep any references to the data.
+            unsafe { iter.del_current()? };
+        }
+    }
+
+    Ok(())
+}
+
 fn indexer_memory_settings(
     current_num_threads: usize,
     grenad_parameters: GrenadParameters,

crates/milli/src/update/new/steps.rs

@@ -28,6 +28,9 @@ make_enum_progress! {
         ChangingVectorStore,
         UsingStableIndexer,
         UsingExperimentalIndexer,
+        DeletingOldWordFidDocids,
+        DeletingOldFidWordCountDocids,
+        DeletingOldWordPrefixFidDocids,
     }
 }
 

crates/milli/src/update/settings.rs

@@ -48,10 +48,11 @@ use crate::{
     ChannelCongestion, FieldId, FilterableAttributesRule, Index, LocalizedAttributesRule, Result,
 };
 
-#[derive(Debug, Clone, PartialEq, Eq, Copy)]
+#[derive(Default, Debug, Clone, PartialEq, Eq, Copy)]
 pub enum Setting<T> {
     Set(T),
     Reset,
+    #[default]
     NotSet,
 }
 
@@ -71,12 +72,6 @@ where
     }
 }
 
-impl<T> Default for Setting<T> {
-    fn default() -> Self {
-        Self::NotSet
-    }
-}
-
 impl<T> Setting<T> {
     pub fn set(self) -> Option<T> {
         match self {
@@ -1589,33 +1584,33 @@ impl<'a, 't, 'i> Settings<'a, 't, 'i> {
 
         // only use the new indexer when only the embedder possibly changed
         if let Self {
-            searchable_fields: Setting::NotSet,
+            searchable_fields: _,
             displayed_fields: Setting::NotSet,
             filterable_fields: Setting::NotSet,
             sortable_fields: Setting::NotSet,
             criteria: Setting::NotSet,
-            stop_words: Setting::NotSet,
-            non_separator_tokens: Setting::NotSet,
-            separator_tokens: Setting::NotSet,
-            dictionary: Setting::NotSet,
+            stop_words: Setting::NotSet, // TODO (require force reindexing of searchables)
+            non_separator_tokens: Setting::NotSet, // TODO (require force reindexing of searchables)
+            separator_tokens: Setting::NotSet, // TODO (require force reindexing of searchables)
+            dictionary: Setting::NotSet, // TODO (require force reindexing of searchables)
             distinct_field: Setting::NotSet,
             synonyms: Setting::NotSet,
             primary_key: Setting::NotSet,
             authorize_typos: Setting::NotSet,
             min_word_len_two_typos: Setting::NotSet,
             min_word_len_one_typo: Setting::NotSet,
-            exact_words: Setting::NotSet,
-            exact_attributes: Setting::NotSet,
+            exact_words: Setting::NotSet, // TODO (require force reindexing of searchables)
+            exact_attributes: _,
             max_values_per_facet: Setting::NotSet,
             sort_facet_values_by: Setting::NotSet,
             pagination_max_total_hits: Setting::NotSet,
-            proximity_precision: Setting::NotSet,
+            proximity_precision: _,
             embedder_settings: _,
             search_cutoff: Setting::NotSet,
-            localized_attributes_rules: Setting::NotSet,
-            prefix_search: Setting::NotSet,
+            localized_attributes_rules: Setting::NotSet, // TODO to start with
+            prefix_search: Setting::NotSet,              // TODO continue with this
             facet_search: Setting::NotSet,
-            disable_on_numbers: Setting::NotSet,
+            disable_on_numbers: Setting::NotSet, // TODO (require force reindexing of searchables)
             chat: Setting::NotSet,
             vector_store: Setting::NotSet,
             wtxn: _,
@@ -1632,10 +1627,12 @@ impl<'a, 't, 'i> Settings<'a, 't, 'i> {
             // Update index settings
             let embedding_config_updates = self.update_embedding_configs()?;
             self.update_user_defined_searchable_attributes()?;
+            self.update_exact_attributes()?;
+            self.update_proximity_precision()?;
 
-            let mut new_inner_settings =
-                InnerIndexSettings::from_index(self.index, self.wtxn, None)?;
-            new_inner_settings.recompute_searchables(self.wtxn, self.index)?;
+            // Note that we don't need to update the searchables here,
+            // as it will be done after the settings update.
+            let new_inner_settings = InnerIndexSettings::from_index(self.index, self.wtxn, None)?;
 
             let primary_key_id = self
                 .index
@@ -2062,9 +2059,12 @@ impl InnerIndexSettings {
         let sortable_fields = index.sortable_fields(rtxn)?;
         let asc_desc_fields = index.asc_desc_fields(rtxn)?;
         let distinct_field = index.distinct_field(rtxn)?.map(|f| f.to_string());
-        let user_defined_searchable_attributes = index
-            .user_defined_searchable_fields(rtxn)?
-            .map(|fields| fields.into_iter().map(|f| f.to_string()).collect());
+        let user_defined_searchable_attributes = match index.user_defined_searchable_fields(rtxn)? {
+            Some(fields) if fields.contains(&"*") => None,
+            Some(fields) => Some(fields.into_iter().map(|f| f.to_string()).collect()),
+            None => None,
+        };
+
         let builder = MetadataBuilder::from_index(index, rtxn)?;
         let fields_ids_map = FieldIdMapWithMetadata::new(fields_ids_map, builder);
         let disabled_typos_terms = index.disabled_typos_terms(rtxn)?;
@@ -2578,8 +2578,20 @@ fn deserialize_sub_embedder(
 /// Implement this trait for the settings delta type.
 /// This is used in the new settings update flow and will allow to easily replace the old settings delta type: `InnerIndexSettingsDiff`.
 pub trait SettingsDelta {
-    fn new_embedders(&self) -> &RuntimeEmbedders;
+    fn old_fields_ids_map(&self) -> &FieldIdMapWithMetadata;
+    fn new_fields_ids_map(&self) -> &FieldIdMapWithMetadata;
+
+    fn old_searchable_attributes(&self) -> &Option<Vec<String>>;
+    fn new_searchable_attributes(&self) -> &Option<Vec<String>>;
+
+    fn old_disabled_typos_terms(&self) -> &DisabledTyposTerms;
+    fn new_disabled_typos_terms(&self) -> &DisabledTyposTerms;
+
+    fn old_proximity_precision(&self) -> &ProximityPrecision;
+    fn new_proximity_precision(&self) -> &ProximityPrecision;
+
     fn old_embedders(&self) -> &RuntimeEmbedders;
+    fn new_embedders(&self) -> &RuntimeEmbedders;
     fn new_embedder_category_id(&self) -> &HashMap<String, u8>;
     fn embedder_actions(&self) -> &BTreeMap<String, EmbedderAction>;
     fn try_for_each_fragment_diff<F, E>(
@@ -2589,7 +2601,6 @@ pub trait SettingsDelta {
     ) -> std::result::Result<(), E>
     where
         F: FnMut(FragmentDiff) -> std::result::Result<(), E>;
-    fn new_fields_ids_map(&self) -> &FieldIdMapWithMetadata;
 }
 
 pub struct FragmentDiff<'a> {
@@ -2598,26 +2609,47 @@ pub struct FragmentDiff<'a> {
 }
 
 impl SettingsDelta for InnerIndexSettingsDiff {
-    fn new_embedders(&self) -> &RuntimeEmbedders {
-        &self.new.runtime_embedders
+    fn old_fields_ids_map(&self) -> &FieldIdMapWithMetadata {
+        &self.old.fields_ids_map
+    }
+    fn new_fields_ids_map(&self) -> &FieldIdMapWithMetadata {
+        &self.new.fields_ids_map
+    }
+
+    fn old_searchable_attributes(&self) -> &Option<Vec<String>> {
+        &self.old.user_defined_searchable_attributes
+    }
+    fn new_searchable_attributes(&self) -> &Option<Vec<String>> {
+        &self.new.user_defined_searchable_attributes
+    }
+
+    fn old_disabled_typos_terms(&self) -> &DisabledTyposTerms {
+        &self.old.disabled_typos_terms
+    }
+    fn new_disabled_typos_terms(&self) -> &DisabledTyposTerms {
+        &self.new.disabled_typos_terms
+    }
+
+    fn old_proximity_precision(&self) -> &ProximityPrecision {
+        &self.old.proximity_precision
+    }
+    fn new_proximity_precision(&self) -> &ProximityPrecision {
+        &self.new.proximity_precision
     }
 
     fn old_embedders(&self) -> &RuntimeEmbedders {
         &self.old.runtime_embedders
     }
+    fn new_embedders(&self) -> &RuntimeEmbedders {
+        &self.new.runtime_embedders
+    }
 
     fn new_embedder_category_id(&self) -> &HashMap<String, u8> {
         &self.new.embedder_category_id
     }
-
     fn embedder_actions(&self) -> &BTreeMap<String, EmbedderAction> {
         &self.embedding_config_updates
     }
-
-    fn new_fields_ids_map(&self) -> &FieldIdMapWithMetadata {
-        &self.new.fields_ids_map
-    }
-
     fn try_for_each_fragment_diff<F, E>(
         &self,
         embedder_name: &str,

crates/milli/src/update/test_settings.rs

@@ -14,28 +14,21 @@ fn set_and_reset_searchable_fields() {
     let index = TempIndex::new();
 
     // First we send 3 documents with ids from 1 to 3.
-    let mut wtxn = index.write_txn().unwrap();
-
     index
-        .add_documents_using_wtxn(
-            &mut wtxn,
-            documents!([
-                { "id": 1, "name": "kevin", "age": 23 },
-                { "id": 2, "name": "kevina", "age": 21},
-                { "id": 3, "name": "benoit", "age": 34 }
-            ]),
-        )
+        .add_documents(documents!([
+            { "id": 1, "name": "kevin", "age": 23 },
+            { "id": 2, "name": "kevina", "age": 21},
+            { "id": 3, "name": "benoit", "age": 34 }
+        ]))
         .unwrap();
 
     // We change the searchable fields to be the "name" field only.
     index
-        .update_settings_using_wtxn(&mut wtxn, |settings| {
+        .update_settings(|settings| {
             settings.set_searchable_fields(vec!["name".into()]);
         })
         .unwrap();
 
-    wtxn.commit().unwrap();
-
     db_snap!(index, fields_ids_map, @r###"
     0   id               |
     1   name             |

crates/milli/src/update/upgrade/mod.rs

@@ -5,103 +5,36 @@ mod v1_15;
 mod v1_16;
 
 use heed::RwTxn;
-use v1_12::{V1_12_3_To_V1_13_0, V1_12_To_V1_12_3};
-use v1_13::{V1_13_0_To_V1_13_1, V1_13_1_To_Latest_V1_13};
-use v1_14::Latest_V1_13_To_Latest_V1_14;
-use v1_15::Latest_V1_14_To_Latest_V1_15;
-use v1_16::Latest_V1_15_To_V1_16_0;
+use v1_12::{FixFieldDistribution, RecomputeStats};
+use v1_13::AddNewStats;
+use v1_14::UpgradeArroyVersion;
+use v1_15::RecomputeWordFst;
+use v1_16::SwitchToMultimodal;
 
 use crate::constants::{VERSION_MAJOR, VERSION_MINOR, VERSION_PATCH};
 use crate::progress::{Progress, VariableNameStep};
 use crate::{Index, InternalError, Result};
 
 trait UpgradeIndex {
+    /// Returns `true` if `upgrade` should be called when the index started with version `initial_version`.
+    fn must_upgrade(&self, initial_version: (u32, u32, u32)) -> bool;
+
     /// Returns `true` if the index scheduler must regenerate its cached stats.
-    fn upgrade(
-        &self,
-        wtxn: &mut RwTxn,
-        index: &Index,
-        original: (u32, u32, u32),
-        progress: Progress,
-    ) -> Result<bool>;
-    fn target_version(&self) -> (u32, u32, u32);
+    fn upgrade(&self, wtxn: &mut RwTxn, index: &Index, progress: Progress) -> Result<bool>;
+
+    /// Description of the upgrade for progress display purposes.
+    fn description(&self) -> &'static str;
 }
 
 const UPGRADE_FUNCTIONS: &[&dyn UpgradeIndex] = &[
-    &V1_12_To_V1_12_3 {},
-    &V1_12_3_To_V1_13_0 {},
-    &V1_13_0_To_V1_13_1 {},
-    &V1_13_1_To_Latest_V1_13 {},
-    &Latest_V1_13_To_Latest_V1_14 {},
-    &Latest_V1_14_To_Latest_V1_15 {},
-    &Latest_V1_15_To_V1_16_0 {},
-    &ToTargetNoOp { target: (1, 18, 0) },
-    &ToTargetNoOp { target: (1, 19, 0) },
-    &ToTargetNoOp { target: (1, 20, 0) },
-    &ToTargetNoOp { target: (1, 21, 0) },
-    &ToTargetNoOp { target: (1, 22, 0) },
-    &ToTargetNoOp { target: (1, 23, 0) },
-    &ToTargetNoOp { target: (1, 24, 0) },
-    &ToTargetNoOp { target: (1, 25, 0) },
-    &ToTargetNoOp { target: (1, 26, 0) },
-    &ToTargetNoOp { target: (1, 27, 0) },
-    &ToTargetNoOp { target: (1, 28, 0) },
-    // This is the last upgrade function, it will be called when the index is up to date.
-    // any other upgrade function should be added before this one.
-    &ToCurrentNoOp {},
+    &FixFieldDistribution {},
+    &RecomputeStats {},
+    &AddNewStats {},
+    &UpgradeArroyVersion {},
+    &RecomputeWordFst {},
+    &SwitchToMultimodal {},
 ];
 
-/// Causes a compile-time error if the argument is not in range of `0..UPGRADE_FUNCTIONS.len()`
-macro_rules! function_index {
-    ($start:expr) => {{
-        const _CHECK_INDEX: () = {
-            if $start >= $crate::update::upgrade::UPGRADE_FUNCTIONS.len() {
-                panic!("upgrade functions out of range")
-            }
-        };
-
-        $start
-    }};
-}
-
-const fn start(from: (u32, u32, u32)) -> Option<usize> {
-    let start = match from {
-        (1, 12, 0..=2) => function_index!(0),
-        (1, 12, 3..) => function_index!(1),
-        (1, 13, 0) => function_index!(2),
-        (1, 13, _) => function_index!(4),
-        (1, 14, _) => function_index!(5),
-        // We must handle the current version in the match because in case of a failure some index may have been upgraded but not other.
-        (1, 15, _) => function_index!(6),
-        (1, 16, _) | (1, 17, _) => function_index!(7),
-        (1, 18, _) => function_index!(8),
-        (1, 19, _) => function_index!(9),
-        (1, 20, _) => function_index!(10),
-        (1, 21, _) => function_index!(11),
-        (1, 22, _) => function_index!(12),
-        (1, 23, _) => function_index!(13),
-        (1, 24, _) => function_index!(14),
-        (1, 25, _) => function_index!(15),
-        (1, 26, _) => function_index!(16),
-        (1, 27, _) => function_index!(17),
-        (1, 28, _) => function_index!(18),
-        // We deliberately don't add a placeholder with (VERSION_MAJOR, VERSION_MINOR, VERSION_PATCH) here to force manually
-        // considering dumpless upgrade.
-        (_major, _minor, _patch) => return None,
-    };
-
-    Some(start)
-}
-
-/// Causes a compile-time error if the latest package cannot be upgraded.
-///
-/// This serves as a reminder to consider the proper dumpless upgrade implementation when changing the package version.
-const _CHECK_PACKAGE_CAN_UPGRADE: () = {
-    if start((VERSION_MAJOR, VERSION_MINOR, VERSION_PATCH)).is_none() {
-        panic!("cannot upgrade from latest package version")
-    }
-};
-
 /// Return true if the cached stats of the index must be regenerated
 pub fn upgrade<MSP>(
     wtxn: &mut RwTxn,
@@ -113,79 +46,34 @@ pub fn upgrade<MSP>(
 where
     MSP: Fn() -> bool + Sync,
 {
-    let from = index.get_version(wtxn)?.unwrap_or(db_version);
+    let upgrade_functions = UPGRADE_FUNCTIONS;
 
-    let start =
-        start(from).ok_or_else(|| InternalError::CannotUpgradeToVersion(from.0, from.1, from.2))?;
+    let initial_version = index.get_version(wtxn)?.unwrap_or(db_version);
 
     enum UpgradeVersion {}
-    let upgrade_path = &UPGRADE_FUNCTIONS[start..];
 
-    let mut current_version = from;
     let mut regenerate_stats = false;
-    for (i, upgrade) in upgrade_path.iter().enumerate() {
+    for (i, upgrade) in upgrade_functions.iter().enumerate() {
         if (must_stop_processing)() {
             return Err(crate::Error::InternalError(InternalError::AbortedIndexation));
         }
-        let target = upgrade.target_version();
-        progress.update_progress(VariableNameStep::<UpgradeVersion>::new(
-            format!(
-                "Upgrading from v{}.{}.{} to v{}.{}.{}",
-                current_version.0,
-                current_version.1,
-                current_version.2,
-                target.0,
-                target.1,
-                target.2
-            ),
-            i as u32,
-            upgrade_path.len() as u32,
-        ));
-        regenerate_stats |= upgrade.upgrade(wtxn, index, from, progress.clone())?;
-        index.put_version(wtxn, target)?;
-        current_version = target;
+        if upgrade.must_upgrade(initial_version) {
+            regenerate_stats |= upgrade.upgrade(wtxn, index, progress.clone())?;
+            progress.update_progress(VariableNameStep::<UpgradeVersion>::new(
+                upgrade.description(),
+                i as u32,
+                upgrade_functions.len() as u32,
+            ));
+        } else {
+            progress.update_progress(VariableNameStep::<UpgradeVersion>::new(
+                "Skipping migration that must not be applied",
+                i as u32,
+                upgrade_functions.len() as u32,
+            ));
+        }
     }
 
+    index.put_version(wtxn, (VERSION_MAJOR, VERSION_MINOR, VERSION_PATCH))?;
+
     Ok(regenerate_stats)
 }
-
-#[allow(non_camel_case_types)]
-struct ToCurrentNoOp {}
-
-impl UpgradeIndex for ToCurrentNoOp {
-    fn upgrade(
-        &self,
-        _wtxn: &mut RwTxn,
-        _index: &Index,
-        _original: (u32, u32, u32),
-        _progress: Progress,
-    ) -> Result<bool> {
-        Ok(false)
-    }
-
-    fn target_version(&self) -> (u32, u32, u32) {
-        (VERSION_MAJOR, VERSION_MINOR, VERSION_PATCH)
-    }
-}
-
-/// Perform no operation during the upgrade except changing to the specified target version.
-#[allow(non_camel_case_types)]
-struct ToTargetNoOp {
-    pub target: (u32, u32, u32),
-}
-
-impl UpgradeIndex for ToTargetNoOp {
-    fn upgrade(
-        &self,
-        _wtxn: &mut RwTxn,
-        _index: &Index,
-        _original: (u32, u32, u32),
-        _progress: Progress,
-    ) -> Result<bool> {
-        Ok(false)
-    }
-
-    fn target_version(&self) -> (u32, u32, u32) {
-        self.target
-    }
-}

crates/milli/src/update/upgrade/v1_12.rs

@@ -4,17 +4,10 @@ use super::UpgradeIndex;
 use crate::progress::Progress;
 use crate::{make_enum_progress, Index, Result};
 
-#[allow(non_camel_case_types)]
-pub(super) struct V1_12_To_V1_12_3 {}
+pub(super) struct FixFieldDistribution {}
 
-impl UpgradeIndex for V1_12_To_V1_12_3 {
-    fn upgrade(
-        &self,
-        wtxn: &mut RwTxn,
-        index: &Index,
-        _original: (u32, u32, u32),
-        progress: Progress,
-    ) -> Result<bool> {
+impl UpgradeIndex for FixFieldDistribution {
+    fn upgrade(&self, wtxn: &mut RwTxn, index: &Index, progress: Progress) -> Result<bool> {
         make_enum_progress! {
             enum FieldDistribution {
                 RebuildingFieldDistribution,
@@ -25,27 +18,28 @@ impl UpgradeIndex for V1_12_To_V1_12_3 {
         Ok(true)
     }
 
-    fn target_version(&self) -> (u32, u32, u32) {
-        (1, 12, 3)
+    fn must_upgrade(&self, initial_version: (u32, u32, u32)) -> bool {
+        initial_version < (1, 12, 3)
+    }
+
+    fn description(&self) -> &'static str {
+        "Recomputing field distribution which was wrong before v1.12.3"
     }
 }
 
-#[allow(non_camel_case_types)]
-pub(super) struct V1_12_3_To_V1_13_0 {}
+pub(super) struct RecomputeStats {}
 
-impl UpgradeIndex for V1_12_3_To_V1_13_0 {
-    fn upgrade(
-        &self,
-        _wtxn: &mut RwTxn,
-        _index: &Index,
-        _original: (u32, u32, u32),
-        _progress: Progress,
-    ) -> Result<bool> {
+impl UpgradeIndex for RecomputeStats {
+    fn upgrade(&self, _wtxn: &mut RwTxn, _index: &Index, _progress: Progress) -> Result<bool> {
         // recompute the indexes stats
         Ok(true)
     }
 
-    fn target_version(&self) -> (u32, u32, u32) {
-        (1, 13, 0)
+    fn must_upgrade(&self, initial_version: (u32, u32, u32)) -> bool {
+        initial_version < (1, 13, 0)
+    }
+
+    fn description(&self) -> &'static str {
+        "Recomputing stats"
     }
 }

crates/milli/src/update/upgrade/v1_13.rs

@@ -5,17 +5,10 @@ use crate::database_stats::DatabaseStats;
 use crate::progress::Progress;
 use crate::{make_enum_progress, Index, Result};
 
-#[allow(non_camel_case_types)]
-pub(super) struct V1_13_0_To_V1_13_1();
+pub(super) struct AddNewStats();
 
-impl UpgradeIndex for V1_13_0_To_V1_13_1 {
-    fn upgrade(
-        &self,
-        wtxn: &mut RwTxn,
-        index: &Index,
-        _original: (u32, u32, u32),
-        progress: Progress,
-    ) -> Result<bool> {
+impl UpgradeIndex for AddNewStats {
+    fn upgrade(&self, wtxn: &mut RwTxn, index: &Index, progress: Progress) -> Result<bool> {
         make_enum_progress! {
             enum DocumentsStats {
                 CreatingDocumentsStats,
@@ -30,26 +23,11 @@ impl UpgradeIndex for V1_13_0_To_V1_13_1 {
         Ok(true)
     }
 
-    fn target_version(&self) -> (u32, u32, u32) {
-        (1, 13, 1)
-    }
-}
-
-#[allow(non_camel_case_types)]
-pub(super) struct V1_13_1_To_Latest_V1_13();
-
-impl UpgradeIndex for V1_13_1_To_Latest_V1_13 {
-    fn upgrade(
-        &self,
-        _wtxn: &mut RwTxn,
-        _index: &Index,
-        _original: (u32, u32, u32),
-        _progress: Progress,
-    ) -> Result<bool> {
-        Ok(false)
-    }
-
-    fn target_version(&self) -> (u32, u32, u32) {
-        (1, 13, 3)
+    fn must_upgrade(&self, initial_version: (u32, u32, u32)) -> bool {
+        initial_version < (1, 13, 1)
+    }
+
+    fn description(&self) -> &'static str {
+        "Computing newly introduced document stats"
     }
 }

crates/milli/src/update/upgrade/v1_14.rs

@@ -5,17 +5,10 @@ use super::UpgradeIndex;
 use crate::progress::Progress;
 use crate::{make_enum_progress, Index, Result};
 
-#[allow(non_camel_case_types)]
-pub(super) struct Latest_V1_13_To_Latest_V1_14();
+pub(super) struct UpgradeArroyVersion();
 
-impl UpgradeIndex for Latest_V1_13_To_Latest_V1_14 {
-    fn upgrade(
-        &self,
-        wtxn: &mut RwTxn,
-        index: &Index,
-        _original: (u32, u32, u32),
-        progress: Progress,
-    ) -> Result<bool> {
+impl UpgradeIndex for UpgradeArroyVersion {
+    fn upgrade(&self, wtxn: &mut RwTxn, index: &Index, progress: Progress) -> Result<bool> {
         make_enum_progress! {
             enum VectorStore {
                 UpdateInternalVersions,
@@ -35,7 +28,11 @@ impl UpgradeIndex for Latest_V1_13_To_Latest_V1_14 {
         Ok(false)
     }
 
-    fn target_version(&self) -> (u32, u32, u32) {
-        (1, 14, 0)
+    fn must_upgrade(&self, initial_version: (u32, u32, u32)) -> bool {
+        initial_version < (1, 14, 0)
+    }
+
+    fn description(&self) -> &'static str {
+        "Updating vector store with an internal version"
     }
 }

crates/milli/src/update/upgrade/v1_15.rs

@@ -7,25 +7,21 @@ use crate::progress::Progress;
 use crate::update::new::indexer::recompute_word_fst_from_word_docids_database;
 use crate::{Index, Result};
 
-#[allow(non_camel_case_types)]
-pub(super) struct Latest_V1_14_To_Latest_V1_15();
+pub(super) struct RecomputeWordFst();
 
-impl UpgradeIndex for Latest_V1_14_To_Latest_V1_15 {
-    fn upgrade(
-        &self,
-        wtxn: &mut RwTxn,
-        index: &Index,
-        _original: (u32, u32, u32),
-        progress: Progress,
-    ) -> Result<bool> {
+impl UpgradeIndex for RecomputeWordFst {
+    fn upgrade(&self, wtxn: &mut RwTxn, index: &Index, progress: Progress) -> Result<bool> {
         // Recompute the word FST from the word docids database.
         recompute_word_fst_from_word_docids_database(index, wtxn, &progress)?;
 
         Ok(false)
     }
+    fn must_upgrade(&self, initial_version: (u32, u32, u32)) -> bool {
+        initial_version < (1, 15, 0)
+    }
 
-    fn target_version(&self) -> (u32, u32, u32) {
-        (1, 15, 0)
+    fn description(&self) -> &'static str {
+        "Recomputing word FST from word docids database as it was wrong before v1.15.0"
     }
 }
 

crates/milli/src/update/upgrade/v1_16.rs

@@ -6,17 +6,10 @@ use crate::progress::Progress;
 use crate::vector::db::{EmbedderInfo, EmbeddingStatus};
 use crate::{Index, InternalError, Result};
 
-#[allow(non_camel_case_types)]
-pub(super) struct Latest_V1_15_To_V1_16_0();
+pub(super) struct SwitchToMultimodal();
 
-impl UpgradeIndex for Latest_V1_15_To_V1_16_0 {
-    fn upgrade(
-        &self,
-        wtxn: &mut RwTxn,
-        index: &Index,
-        _original: (u32, u32, u32),
-        _progress: Progress,
-    ) -> Result<bool> {
+impl UpgradeIndex for SwitchToMultimodal {
+    fn upgrade(&self, wtxn: &mut RwTxn, index: &Index, _progress: Progress) -> Result<bool> {
         let v1_15_indexing_configs = index
             .main
             .remap_types::<Str, SerdeJson<Vec<super::v1_15::IndexEmbeddingConfig>>>()
@@ -41,8 +34,11 @@ impl UpgradeIndex for Latest_V1_15_To_V1_16_0 {
 
         Ok(false)
     }
+    fn must_upgrade(&self, initial_version: (u32, u32, u32)) -> bool {
+        initial_version < (1, 16, 0)
+    }
 
-    fn target_version(&self) -> (u32, u32, u32) {
-        (1, 16, 0)
+    fn description(&self) -> &'static str {
+        "Migrating the database for multimodal support"
     }
 }

crates/milli/src/vector/embedder/hf.rs

@@ -2,6 +2,7 @@ use candle_core::Tensor;
 use candle_nn::VarBuilder;
 use candle_transformers::models::bert::{BertModel, Config as BertConfig, DTYPE};
 use candle_transformers::models::modernbert::{Config as ModernConfig, ModernBert};
+use candle_transformers::models::xlm_roberta::{Config as XlmRobertaConfig, XLMRobertaModel};
 // FIXME: currently we'll be using the hub to retrieve model, in the future we might want to embed it into Meilisearch itself
 use hf_hub::api::sync::Api;
 use hf_hub::{Repo, RepoType};
@@ -89,6 +90,7 @@ impl Default for EmbedderOptions {
 enum ModelKind {
     Bert(BertModel),
     Modern(ModernBert),
+    XlmRoberta(XLMRobertaModel),
 }
 
 /// Perform embedding of documents and queries
@@ -304,7 +306,8 @@ impl Embedder {
         };
 
         let is_modern = has_arch("modernbert");
-        tracing::debug!(is_modern, model_type, "detected HF architecture");
+        let is_xlm_roberta = has_arch("xlm-roberta") || has_arch("xlm_roberta");
+        tracing::debug!(is_modern, is_xlm_roberta, model_type, "detected HF architecture");
 
         let mut tokenizer = Tokenizer::from_file(&tokenizer_filename)
             .map_err(|inner| NewEmbedderError::open_tokenizer(tokenizer_filename, inner))?;
@@ -340,6 +343,18 @@ impl Embedder {
                 )
             })?;
             ModelKind::Modern(ModernBert::load(vb, &config).map_err(NewEmbedderError::load_model)?)
+        } else if is_xlm_roberta {
+            let config: XlmRobertaConfig = serde_json::from_str(&config_str).map_err(|inner| {
+                NewEmbedderError::deserialize_config(
+                    options.model.clone(),
+                    config_str.clone(),
+                    config_filename.clone(),
+                    inner,
+                )
+            })?;
+            ModelKind::XlmRoberta(
+                XLMRobertaModel::new(&config, vb).map_err(NewEmbedderError::load_model)?,
+            )
         } else {
             let config: BertConfig = serde_json::from_str(&config_str).map_err(|inner| {
                 NewEmbedderError::deserialize_config(
@@ -451,6 +466,19 @@ impl Embedder {
                 let mask = Tensor::stack(&[mask], 0).map_err(EmbedError::tensor_shape)?;
                 model.forward(&token_ids, &mask).map_err(EmbedError::model_forward)?
             }
+            ModelKind::XlmRoberta(model) => {
+                let mut mask_vec = tokens.get_attention_mask().to_vec();
+                if mask_vec.len() > self.max_len {
+                    mask_vec.truncate(self.max_len);
+                }
+                let mask = Tensor::new(mask_vec.as_slice(), &self.device)
+                    .map_err(EmbedError::tensor_shape)?;
+                let mask = Tensor::stack(&[mask], 0).map_err(EmbedError::tensor_shape)?;
+                let token_type_ids = token_ids.zeros_like().map_err(EmbedError::tensor_shape)?;
+                model
+                    .forward(&token_ids, &mask, &token_type_ids, None, None, None)
+                    .map_err(EmbedError::model_forward)?
+            }
         };
 
         let embedding = Self::pooling(embeddings, self.pooling)?;

crates/milli/src/vector/embeddings.rs

@@ -67,7 +67,7 @@ impl<F> Embeddings<F> {
     ///
     /// If `embeddings.len() % self.dimension != 0`, then the append operation fails.
     pub fn append(&mut self, mut embeddings: Vec<F>) -> Result<(), Vec<F>> {
-        if embeddings.len() % self.dimension != 0 {
+        if !embeddings.len().is_multiple_of(self.dimension) {
             return Err(embeddings);
         }
         self.data.append(&mut embeddings);

crates/milli/src/vector/store.rs

@@ -1,5 +1,5 @@
 use hannoy::distances::{Cosine, Hamming};
-use hannoy::ItemId;
+use hannoy::{ItemId, Searched};
 use heed::{RoTxn, RwTxn, Unspecified};
 use ordered_float::OrderedFloat;
 use rand::SeedableRng as _;
@@ -974,7 +974,7 @@ impl VectorStore {
             }
 
             if let Some(mut ret) = searcher.by_item(rtxn, item)? {
-                results.append(&mut ret);
+                results.append(&mut ret.nns);
             }
         }
         results.sort_unstable_by_key(|(_, distance)| OrderedFloat(*distance));
@@ -1028,10 +1028,9 @@ impl VectorStore {
                 searcher.candidates(filter);
             }
 
-            let (res, _degraded) =
-                &mut searcher
-                    .by_vector_with_cancellation(rtxn, vector, || time_budget.exceeded())?;
-            results.append(res);
+            let Searched { mut nns, did_cancel: _ } =
+                searcher.by_vector_with_cancellation(rtxn, vector, || time_budget.exceeded())?;
+            results.append(&mut nns);
         }
 
         results.sort_unstable_by_key(|(_, distance)| OrderedFloat(*distance));

crates/xtask/Cargo.toml

@@ -22,6 +22,7 @@ reqwest = { version = "0.12.24", features = [
     "json",
     "rustls-tls",
 ], default-features = false }
+semver = "1.0.27"
 serde = { version = "1.0.228", features = ["derive"] }
 serde_json = "1.0.145"
 sha2 = "0.10.9"
@@ -42,3 +43,4 @@ tracing = "0.1.41"
 tracing-subscriber = "0.3.20"
 tracing-trace = { version = "0.1.0", path = "../tracing-trace" }
 uuid = { version = "1.18.1", features = ["v7", "serde"] }
+similar-asserts = "1.7.0"

crates/xtask/src/bench/command.rs

@@ -1,194 +0,0 @@
-use std::collections::BTreeMap;
-use std::fmt::Display;
-use std::io::Read as _;
-
-use anyhow::{bail, Context as _};
-use serde::Deserialize;
-
-use super::assets::{fetch_asset, Asset};
-use super::client::{Client, Method};
-
-#[derive(Clone, Deserialize)]
-pub struct Command {
-    pub route: String,
-    pub method: Method,
-    #[serde(default)]
-    pub body: Body,
-    #[serde(default)]
-    pub synchronous: SyncMode,
-}
-
-#[derive(Default, Clone, Deserialize)]
-#[serde(untagged)]
-pub enum Body {
-    Inline {
-        inline: serde_json::Value,
-    },
-    Asset {
-        asset: String,
-    },
-    #[default]
-    Empty,
-}
-
-impl Body {
-    pub fn get(
-        self,
-        assets: &BTreeMap<String, Asset>,
-        asset_folder: &str,
-    ) -> anyhow::Result<Option<(Vec<u8>, &'static str)>> {
-        Ok(match self {
-            Body::Inline { inline: body } => Some((
-                serde_json::to_vec(&body)
-                    .context("serializing to bytes")
-                    .context("while getting inline body")?,
-                "application/json",
-            )),
-            Body::Asset { asset: name } => Some({
-                let context = || format!("while getting body from asset '{name}'");
-                let (mut file, format) =
-                    fetch_asset(&name, assets, asset_folder).with_context(context)?;
-                let mut buf = Vec::new();
-                file.read_to_end(&mut buf).with_context(context)?;
-                (buf, format.to_content_type(&name))
-            }),
-            Body::Empty => None,
-        })
-    }
-}
-
-impl Display for Command {
-    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        write!(f, "{:?} {} ({:?})", self.method, self.route, self.synchronous)
-    }
-}
-
-#[derive(Default, Debug, Clone, Copy, Deserialize)]
-pub enum SyncMode {
-    DontWait,
-    #[default]
-    WaitForResponse,
-    WaitForTask,
-}
-
-pub async fn run_batch(
-    client: &Client,
-    batch: &[Command],
-    assets: &BTreeMap<String, Asset>,
-    asset_folder: &str,
-) -> anyhow::Result<()> {
-    let [.., last] = batch else { return Ok(()) };
-    let sync = last.synchronous;
-
-    let mut tasks = tokio::task::JoinSet::new();
-
-    for command in batch {
-        // FIXME: you probably don't want to copy assets everytime here
-        tasks.spawn({
-            let client = client.clone();
-            let command = command.clone();
-            let assets = assets.clone();
-            let asset_folder = asset_folder.to_owned();
-
-            async move { run(client, command, &assets, &asset_folder).await }
-        });
-    }
-
-    while let Some(result) = tasks.join_next().await {
-        result
-            .context("panicked while executing command")?
-            .context("error while executing command")?;
-    }
-
-    match sync {
-        SyncMode::DontWait => {}
-        SyncMode::WaitForResponse => {}
-        SyncMode::WaitForTask => wait_for_tasks(client).await?,
-    }
-
-    Ok(())
-}
-
-async fn wait_for_tasks(client: &Client) -> anyhow::Result<()> {
-    loop {
-        let response = client
-            .get("tasks?statuses=enqueued,processing")
-            .send()
-            .await
-            .context("could not wait for tasks")?;
-        let response: serde_json::Value = response
-            .json()
-            .await
-            .context("could not deserialize response to JSON")
-            .context("could not wait for tasks")?;
-        match response.get("total") {
-            Some(serde_json::Value::Number(number)) => {
-                let number = number.as_u64().with_context(|| {
-                    format!("waiting for tasks: could not parse 'total' as integer, got {}", number)
-                })?;
-                if number == 0 {
-                    break;
-                } else {
-                    tokio::time::sleep(std::time::Duration::from_secs(1)).await;
-                    continue;
-                }
-            }
-            Some(thing_else) => {
-                bail!(format!(
-                    "waiting for tasks: could not parse 'total' as a number, got '{thing_else}'"
-                ))
-            }
-            None => {
-                bail!(format!(
-                    "waiting for tasks: expected response to contain 'total', got '{response}'"
-                ))
-            }
-        }
-    }
-    Ok(())
-}
-
-#[tracing::instrument(skip(client, command, assets, asset_folder), fields(command = %command))]
-pub async fn run(
-    client: Client,
-    mut command: Command,
-    assets: &BTreeMap<String, Asset>,
-    asset_folder: &str,
-) -> anyhow::Result<()> {
-    // memtake the body here to leave an empty body in its place, so that command is not partially moved-out
-    let body = std::mem::take(&mut command.body)
-        .get(assets, asset_folder)
-        .with_context(|| format!("while getting body for command {command}"))?;
-
-    let request = client.request(command.method.into(), &command.route);
-
-    let request = if let Some((body, content_type)) = body {
-        request.body(body).header(reqwest::header::CONTENT_TYPE, content_type)
-    } else {
-        request
-    };
-
-    let response =
-        request.send().await.with_context(|| format!("error sending command: {}", command))?;
-
-    let code = response.status();
-    if code.is_client_error() {
-        tracing::error!(%command, %code, "error in workload file");
-        let response: serde_json::Value = response
-            .json()
-            .await
-            .context("could not deserialize response as JSON")
-            .context("parsing error in workload file when sending command")?;
-        bail!("error in workload file: server responded with error code {code} and '{response}'")
-    } else if code.is_server_error() {
-        tracing::error!(%command, %code, "server error");
-        let response: serde_json::Value = response
-            .json()
-            .await
-            .context("could not deserialize response as JSON")
-            .context("parsing server error when sending command")?;
-        bail!("server error: server responded with error code {code} and '{response}'")
-    }
-
-    Ok(())
-}

crates/xtask/src/bench/dashboard.rs

@@ -7,9 +7,9 @@ use tokio::task::AbortHandle;
 use tracing_trace::processor::span_stats::CallStats;
 use uuid::Uuid;
 
-use super::client::Client;
 use super::env_info;
-use super::workload::Workload;
+use super::workload::BenchWorkload;
+use crate::common::client::Client;
 
 #[derive(Debug, Clone)]
 pub enum DashboardClient {
@@ -89,7 +89,7 @@ impl DashboardClient {
     pub async fn create_workload(
         &self,
         invocation_uuid: Uuid,
-        workload: &Workload,
+        workload: &BenchWorkload,
     ) -> anyhow::Result<Uuid> {
         let Self::Client(dashboard_client) = self else { return Ok(Uuid::now_v7()) };
 

crates/xtask/src/bench/mod.rs

@@ -1,51 +1,36 @@
-mod assets;
-mod client;
-mod command;
 mod dashboard;
 mod env_info;
-mod meili_process;
 mod workload;
 
-use std::io::LineWriter;
-use std::path::PathBuf;
+use crate::common::args::CommonArgs;
+use crate::common::logs::setup_logs;
+use crate::common::workload::Workload;
+use std::{path::PathBuf, sync::Arc};
 
-use anyhow::Context;
+use anyhow::{bail, Context};
 use clap::Parser;
-use tracing_subscriber::fmt::format::FmtSpan;
-use tracing_subscriber::layer::SubscriberExt;
-use tracing_subscriber::Layer;
 
-use self::client::Client;
-use self::workload::Workload;
+use crate::common::client::Client;
+pub use workload::BenchWorkload;
 
-pub fn default_http_addr() -> String {
-    "127.0.0.1:7700".to_string()
-}
 pub fn default_report_folder() -> String {
     "./bench/reports/".into()
 }
 
-pub fn default_asset_folder() -> String {
-    "./bench/assets/".into()
-}
-
-pub fn default_log_filter() -> String {
-    "info".into()
-}
-
 pub fn default_dashboard_url() -> String {
     "http://localhost:9001".into()
 }
 
 /// Run benchmarks from a workload
 #[derive(Parser, Debug)]
-pub struct BenchDeriveArgs {
-    /// Filename of the workload file, pass multiple filenames
-    /// to run multiple workloads in the specified order.
-    ///
-    /// Each workload run will get its own report file.
-    #[arg(value_name = "WORKLOAD_FILE", last = false)]
-    workload_file: Vec<PathBuf>,
+pub struct BenchArgs {
+    /// Common arguments shared with other commands
+    #[command(flatten)]
+    common: CommonArgs,
+
+    /// Meilisearch master keys
+    #[arg(long)]
+    pub master_key: Option<String>,
 
     /// URL of the dashboard.
     #[arg(long, default_value_t = default_dashboard_url())]
@@ -59,34 +44,14 @@ pub struct BenchDeriveArgs {
     #[arg(long, default_value_t = default_report_folder())]
     report_folder: String,
 
-    /// Directory to store the remote assets.
-    #[arg(long, default_value_t = default_asset_folder())]
-    asset_folder: String,
-
-    /// Log directives
-    #[arg(short, long, default_value_t = default_log_filter())]
-    log_filter: String,
-
     /// Benchmark dashboard API key
     #[arg(long)]
     api_key: Option<String>,
 
-    /// Meilisearch master keys
-    #[arg(long)]
-    master_key: Option<String>,
-
-    /// Authentication bearer for fetching assets
-    #[arg(long)]
-    assets_key: Option<String>,
-
     /// Reason for the benchmark invocation
     #[arg(short, long)]
     reason: Option<String>,
 
-    /// The maximum time in seconds we allow for fetching the task queue before timing out.
-    #[arg(long, default_value_t = 60)]
-    tasks_queue_timeout_secs: u64,
-
     /// The path to the binary to run.
     ///
     /// If unspecified, runs `cargo run` after building Meilisearch with `cargo build`.
@@ -94,18 +59,8 @@ pub struct BenchDeriveArgs {
     binary_path: Option<PathBuf>,
 }
 
-pub fn run(args: BenchDeriveArgs) -> anyhow::Result<()> {
-    // setup logs
-    let filter: tracing_subscriber::filter::Targets =
-        args.log_filter.parse().context("invalid --log-filter")?;
-
-    let subscriber = tracing_subscriber::registry().with(
-        tracing_subscriber::fmt::layer()
-            .with_writer(|| LineWriter::new(std::io::stderr()))
-            .with_span_events(FmtSpan::NEW | FmtSpan::CLOSE)
-            .with_filter(filter),
-    );
-    tracing::subscriber::set_global_default(subscriber).context("could not setup logging")?;
+pub fn run(args: BenchArgs) -> anyhow::Result<()> {
+    setup_logs(&args.common.log_filter)?;
 
     // fetch environment and build info
     let env = env_info::Environment::generate_from_current_config();
@@ -116,8 +71,11 @@ pub fn run(args: BenchDeriveArgs) -> anyhow::Result<()> {
     let _scope = rt.enter();
 
     // setup clients
-    let assets_client =
-        Client::new(None, args.assets_key.as_deref(), Some(std::time::Duration::from_secs(3600)))?; // 1h
+    let assets_client = Client::new(
+        None,
+        args.common.assets_key.as_deref(),
+        Some(std::time::Duration::from_secs(3600)), // 1h
+    )?;
 
     let dashboard_client = if args.no_dashboard {
         dashboard::DashboardClient::new_dry()
@@ -134,11 +92,11 @@ pub fn run(args: BenchDeriveArgs) -> anyhow::Result<()> {
         None,
     )?;
 
-    let meili_client = Client::new(
+    let meili_client = Arc::new(Client::new(
         Some("http://127.0.0.1:7700".into()),
         args.master_key.as_deref(),
-        Some(std::time::Duration::from_secs(args.tasks_queue_timeout_secs)),
-    )?;
+        Some(std::time::Duration::from_secs(args.common.tasks_queue_timeout_secs)),
+    )?);
 
     // enter runtime
 
@@ -146,11 +104,11 @@ pub fn run(args: BenchDeriveArgs) -> anyhow::Result<()> {
         dashboard_client.send_machine_info(&env).await?;
 
         let commit_message = build_info.commit_msg.unwrap_or_default().split('\n').next().unwrap();
-        let max_workloads = args.workload_file.len();
+        let max_workloads = args.common.workload_file.len();
         let reason: Option<&str> = args.reason.as_deref();
         let invocation_uuid = dashboard_client.create_invocation(build_info.clone(), commit_message, env, max_workloads, reason).await?;
 
-        tracing::info!(workload_count = args.workload_file.len(), "handling workload files");
+        tracing::info!(workload_count = args.common.workload_file.len(), "handling workload files");
 
         // main task
         let workload_runs = tokio::spawn(
@@ -158,13 +116,17 @@ pub fn run(args: BenchDeriveArgs) -> anyhow::Result<()> {
                 let dashboard_client = dashboard_client.clone();
                 let mut dashboard_urls = Vec::new();
                 async move {
-            for workload_file in args.workload_file.iter() {
+            for workload_file in args.common.workload_file.iter() {
                 let workload: Workload = serde_json::from_reader(
                     std::fs::File::open(workload_file)
                         .with_context(|| format!("error opening {}", workload_file.display()))?,
                 )
                 .with_context(|| format!("error parsing {} as JSON", workload_file.display()))?;
 
+                let Workload::Bench(workload) = workload else {
+                    bail!("workload file {} is not a bench workload", workload_file.display());
+                };
+
                 let workload_name = workload.name.clone();
 
                 workload::execute(

crates/xtask/src/bench/workload.rs

@@ -1,24 +1,28 @@
-use std::collections::BTreeMap;
+use std::collections::{BTreeMap, HashMap};
 use std::fs::File;
 use std::io::{Seek as _, Write as _};
 use std::path::Path;
+use std::sync::Arc;
 
 use anyhow::{bail, Context as _};
 use futures_util::TryStreamExt as _;
-use serde::Deserialize;
+use serde::{Deserialize, Serialize};
 use serde_json::json;
 use tokio::task::JoinHandle;
 use uuid::Uuid;
 
-use super::assets::Asset;
-use super::client::Client;
-use super::command::SyncMode;
 use super::dashboard::DashboardClient;
-use super::BenchDeriveArgs;
-use crate::bench::{assets, meili_process};
+use super::BenchArgs;
+use crate::common::assets::{self, Asset};
+use crate::common::client::Client;
+use crate::common::command::{run_commands, Command};
+use crate::common::instance::Binary;
+use crate::common::process::{self, delete_db, start_meili};
 
-#[derive(Deserialize)]
-pub struct Workload {
+/// A bench workload.
+/// Not to be confused with [a test workload](crate::test::workload::Workload).
+#[derive(Serialize, Deserialize, Debug)]
+pub struct BenchWorkload {
     pub name: String,
     pub run_count: u16,
     pub extra_cli_args: Vec<String>,
@@ -26,30 +30,34 @@ pub struct Workload {
     #[serde(default)]
     pub target: String,
     #[serde(default)]
-    pub precommands: Vec<super::command::Command>,
-    pub commands: Vec<super::command::Command>,
+    pub precommands: Vec<Command>,
+    pub commands: Vec<Command>,
 }
 
-async fn run_commands(
+async fn run_workload_commands(
     dashboard_client: &DashboardClient,
     logs_client: &Client,
-    meili_client: &Client,
+    meili_client: &Arc<Client>,
     workload_uuid: Uuid,
-    workload: &Workload,
-    args: &BenchDeriveArgs,
+    workload: &BenchWorkload,
+    args: &BenchArgs,
     run_number: u16,
 ) -> anyhow::Result<JoinHandle<anyhow::Result<File>>> {
     let report_folder = &args.report_folder;
     let workload_name = &workload.name;
+    let assets = Arc::new(workload.assets.clone());
+    let asset_folder = args.common.asset_folder.clone().leak();
 
-    for batch in workload
-        .precommands
-        .as_slice()
-        .split_inclusive(|command| !matches!(command.synchronous, SyncMode::DontWait))
-    {
-        super::command::run_batch(meili_client, batch, &workload.assets, &args.asset_folder)
-            .await?;
-    }
+    run_commands(
+        meili_client,
+        &workload.precommands,
+        0,
+        &assets,
+        asset_folder,
+        &mut HashMap::new(),
+        false,
+    )
+    .await?;
 
     std::fs::create_dir_all(report_folder)
         .with_context(|| format!("could not create report directory at {report_folder}"))?;
@@ -59,14 +67,16 @@ async fn run_commands(
 
     let report_handle = start_report(logs_client, trace_filename, &workload.target).await?;
 
-    for batch in workload
-        .commands
-        .as_slice()
-        .split_inclusive(|command| !matches!(command.synchronous, SyncMode::DontWait))
-    {
-        super::command::run_batch(meili_client, batch, &workload.assets, &args.asset_folder)
-            .await?;
-    }
+    run_commands(
+        meili_client,
+        &workload.commands,
+        0,
+        &assets,
+        asset_folder,
+        &mut HashMap::new(),
+        false,
+    )
+    .await?;
 
     let processor =
         stop_report(dashboard_client, logs_client, workload_uuid, report_filename, report_handle)
@@ -81,14 +91,14 @@ pub async fn execute(
     assets_client: &Client,
     dashboard_client: &DashboardClient,
     logs_client: &Client,
-    meili_client: &Client,
+    meili_client: &Arc<Client>,
     invocation_uuid: Uuid,
     master_key: Option<&str>,
-    workload: Workload,
-    args: &BenchDeriveArgs,
+    workload: BenchWorkload,
+    args: &BenchArgs,
     binary_path: Option<&Path>,
 ) -> anyhow::Result<()> {
-    assets::fetch_assets(assets_client, &workload.assets, &args.asset_folder).await?;
+    assets::fetch_assets(assets_client, &workload.assets, &args.common.asset_folder).await?;
 
     let workload_uuid = dashboard_client.create_workload(invocation_uuid, &workload).await?;
 
@@ -129,38 +139,33 @@ pub async fn execute(
 async fn execute_run(
     dashboard_client: &DashboardClient,
     logs_client: &Client,
-    meili_client: &Client,
+    meili_client: &Arc<Client>,
     workload_uuid: Uuid,
     master_key: Option<&str>,
-    workload: &Workload,
-    args: &BenchDeriveArgs,
+    workload: &BenchWorkload,
+    args: &BenchArgs,
     binary_path: Option<&Path>,
     run_number: u16,
 ) -> anyhow::Result<tokio::task::JoinHandle<anyhow::Result<std::fs::File>>> {
-    meili_process::delete_db();
+    delete_db().await;
 
-    let run_command = match binary_path {
-        Some(binary_path) => tokio::process::Command::new(binary_path),
-        None => {
-            meili_process::build().await?;
-            let mut command = tokio::process::Command::new("cargo");
-            command
-                .arg("run")
-                .arg("--release")
-                .arg("-p")
-                .arg("meilisearch")
-                .arg("--bin")
-                .arg("meilisearch")
-                .arg("--");
-            command
-        }
+    let binary = match binary_path {
+        Some(binary_path) => Binary {
+            source: crate::common::instance::BinarySource::Path(binary_path.to_owned()),
+            extra_cli_args: workload.extra_cli_args.clone(),
+        },
+        None => Binary {
+            source: crate::common::instance::BinarySource::Build {
+                edition: crate::common::instance::Edition::Community,
+            },
+            extra_cli_args: workload.extra_cli_args.clone(),
+        },
     };
 
     let meilisearch =
-        meili_process::start(meili_client, master_key, workload, &args.asset_folder, run_command)
-            .await?;
+        start_meili(meili_client, master_key, &binary, &args.common.asset_folder).await?;
 
-    let processor = run_commands(
+    let processor = run_workload_commands(
         dashboard_client,
         logs_client,
         meili_client,
@@ -171,7 +176,7 @@ async fn execute_run(
     )
     .await?;
 
-    meili_process::kill(meilisearch).await;
+    process::kill_meili(meilisearch).await;
 
     tracing::info!(run_number, "Successful run");
 

crates/xtask/src/common/args.rs

@@ -0,0 +1,36 @@
+use clap::Parser;
+use std::path::PathBuf;
+
+pub fn default_asset_folder() -> String {
+    "./bench/assets/".into()
+}
+
+pub fn default_log_filter() -> String {
+    "info".into()
+}
+
+#[derive(Parser, Debug, Clone)]
+pub struct CommonArgs {
+    /// Filename of the workload file, pass multiple filenames
+    /// to run multiple workloads in the specified order.
+    ///
+    /// For benches, each workload run will get its own report file.
+    #[arg(value_name = "WORKLOAD_FILE", last = false)]
+    pub workload_file: Vec<PathBuf>,
+
+    /// Directory to store the remote assets.
+    #[arg(long, default_value_t = default_asset_folder())]
+    pub asset_folder: String,
+
+    /// Log directives
+    #[arg(short, long, default_value_t = default_log_filter())]
+    pub log_filter: String,
+
+    /// Authentication bearer for fetching assets
+    #[arg(long)]
+    pub assets_key: Option<String>,
+
+    /// The maximum time in seconds we allow for fetching the task queue before timing out.
+    #[arg(long, default_value_t = 60)]
+    pub tasks_queue_timeout_secs: u64,
+}

crates/xtask/src/common/assets.rs

@@ -3,21 +3,22 @@ use std::io::{Read as _, Seek as _, Write as _};
 
 use anyhow::{bail, Context};
 use futures_util::TryStreamExt as _;
-use serde::Deserialize;
+use serde::{Deserialize, Serialize};
 use sha2::Digest;
 
 use super::client::Client;
 
-#[derive(Deserialize, Clone)]
+#[derive(Serialize, Deserialize, Clone, Debug)]
 pub struct Asset {
     pub local_location: Option<String>,
     pub remote_location: Option<String>,
-    #[serde(default)]
+    #[serde(default, skip_serializing_if = "AssetFormat::is_default")]
     pub format: AssetFormat,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
     pub sha256: Option<String>,
 }
 
-#[derive(Deserialize, Default, Copy, Clone)]
+#[derive(Serialize, Deserialize, Default, Copy, Clone, Debug)]
 pub enum AssetFormat {
     #[default]
     Auto,
@@ -27,6 +28,10 @@ pub enum AssetFormat {
 }
 
 impl AssetFormat {
+    fn is_default(&self) -> bool {
+        matches!(self, AssetFormat::Auto)
+    }
+
     pub fn to_content_type(self, filename: &str) -> &'static str {
         match self {
             AssetFormat::Auto => Self::auto_detect(filename).to_content_type(filename),
@@ -166,7 +171,14 @@ fn check_sha256(name: &str, asset: &Asset, mut file: std::fs::File) -> anyhow::R
             }
         }
         None => {
-            tracing::warn!(sha256 = file_hash, "Skipping hash for asset {name} that doesn't have one. Please add it to workload file");
+            let msg = match name.starts_with("meilisearch-") {
+                true => "Please add it to crates/xtask/src/common/instance/release.rs",
+                false => "Please add it to workload file",
+            };
+            tracing::warn!(
+                sha256 = file_hash,
+                "Skipping hash for asset {name} that doesn't have one. {msg}"
+            );
             true
         }
     })

crates/xtask/src/common/client.rs

@@ -1,5 +1,5 @@
 use anyhow::Context;
-use serde::Deserialize;
+use serde::{Deserialize, Serialize};
 
 #[derive(Debug, Clone)]
 pub struct Client {
@@ -61,7 +61,7 @@ impl Client {
     }
 }
 
-#[derive(Debug, Clone, Copy, Deserialize)]
+#[derive(Debug, Clone, Copy, Serialize, Deserialize)]
 #[serde(rename_all = "SCREAMING_SNAKE_CASE")]
 pub enum Method {
     Get,

crates/xtask/src/common/command.rs

@@ -0,0 +1,430 @@
+use std::collections::{BTreeMap, HashMap};
+use std::fmt::Display;
+use std::io::Read as _;
+use std::sync::Arc;
+
+use anyhow::{bail, Context as _};
+use reqwest::StatusCode;
+use serde::{Deserialize, Serialize};
+use serde_json::Value;
+use similar_asserts::SimpleDiff;
+
+use crate::common::assets::{fetch_asset, Asset};
+use crate::common::client::{Client, Method};
+
+#[derive(Serialize, Deserialize, Clone, Debug)]
+#[serde(rename_all = "camelCase", deny_unknown_fields)]
+pub struct Command {
+    pub route: String,
+    pub method: Method,
+    #[serde(default)]
+    pub body: Body,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub expected_status: Option<u16>,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub expected_response: Option<serde_json::Value>,
+    #[serde(default, skip_serializing_if = "HashMap::is_empty")]
+    pub register: HashMap<String, String>,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub api_key_variable: Option<String>,
+    #[serde(default)]
+    pub synchronous: SyncMode,
+}
+
+#[derive(Default, Clone, Serialize, Deserialize, Debug)]
+#[serde(untagged)]
+pub enum Body {
+    Inline {
+        inline: serde_json::Value,
+    },
+    Asset {
+        asset: String,
+    },
+    #[default]
+    Empty,
+}
+
+impl Body {
+    pub fn get(
+        self,
+        assets: &BTreeMap<String, Asset>,
+        registered: &HashMap<String, Value>,
+        asset_folder: &str,
+    ) -> anyhow::Result<Option<(Vec<u8>, &'static str)>> {
+        Ok(match self {
+            Body::Inline { inline: mut body } => {
+                if !registered.is_empty() {
+                    insert_variables(&mut body, registered);
+                }
+
+                Some((
+                    serde_json::to_vec(&body)
+                        .context("serializing to bytes")
+                        .context("while getting inline body")?,
+                    "application/json",
+                ))
+            }
+            Body::Asset { asset: name } => Some({
+                let context = || format!("while getting body from asset '{name}'");
+                let (mut file, format) =
+                    fetch_asset(&name, assets, asset_folder).with_context(context)?;
+                let mut buf = Vec::new();
+                file.read_to_end(&mut buf).with_context(context)?;
+                (buf, format.to_content_type(&name))
+            }),
+            Body::Empty => None,
+        })
+    }
+}
+
+impl Display for Command {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "{:?} {} ({:?})", self.method, self.route, self.synchronous)
+    }
+}
+
+#[derive(Default, Debug, Clone, Copy, Serialize, Deserialize, PartialEq)]
+pub enum SyncMode {
+    DontWait,
+    #[default]
+    WaitForResponse,
+    WaitForTask,
+}
+
+async fn run_batch(
+    client: &Arc<Client>,
+    batch: &[Command],
+    first_command_index: usize,
+    assets: &Arc<BTreeMap<String, Asset>>,
+    asset_folder: &'static str,
+    registered: &mut HashMap<String, Value>,
+    return_response: bool,
+) -> anyhow::Result<Vec<(Value, StatusCode)>> {
+    let [.., last] = batch else { return Ok(Vec::new()) };
+    let sync = last.synchronous;
+    let batch_len = batch.len();
+
+    let mut tasks = Vec::with_capacity(batch.len());
+    for (index, command) in batch.iter().cloned().enumerate() {
+        let client2 = Arc::clone(client);
+        let assets2 = Arc::clone(assets);
+        let needs_response = return_response || !command.register.is_empty();
+        let registered2 = registered.clone(); // FIXME: cloning the whole map for each command is inefficient
+        tasks.push(tokio::spawn(async move {
+            run(
+                &client2,
+                &command,
+                first_command_index + index,
+                &assets2,
+                registered2,
+                asset_folder,
+                needs_response,
+            )
+            .await
+        }));
+    }
+
+    let mut outputs = Vec::with_capacity(if return_response { batch_len } else { 0 });
+    for (task, command) in tasks.into_iter().zip(batch.iter()) {
+        let output = task.await.context("task panicked")??;
+        if let Some(output) = output {
+            for (name, path) in &command.register {
+                let value = output
+                    .0
+                    .pointer(path)
+                    .with_context(|| format!("could not find path '{path}' in response (required to register '{name}')"))?
+                    .clone();
+                registered.insert(name.clone(), value);
+            }
+
+            if return_response {
+                outputs.push(output);
+            }
+        }
+    }
+
+    match sync {
+        SyncMode::DontWait => {}
+        SyncMode::WaitForResponse => {}
+        SyncMode::WaitForTask => wait_for_tasks(client).await?,
+    }
+
+    Ok(outputs)
+}
+
+async fn wait_for_tasks(client: &Client) -> anyhow::Result<()> {
+    loop {
+        let response = client
+            .get("tasks?statuses=enqueued,processing")
+            .send()
+            .await
+            .context("could not wait for tasks")?;
+        let response: serde_json::Value = response
+            .json()
+            .await
+            .context("could not deserialize response to JSON")
+            .context("could not wait for tasks")?;
+        match response.get("total") {
+            Some(serde_json::Value::Number(number)) => {
+                let number = number.as_u64().with_context(|| {
+                    format!("waiting for tasks: could not parse 'total' as integer, got {}", number)
+                })?;
+                if number == 0 {
+                    break;
+                } else {
+                    tokio::time::sleep(std::time::Duration::from_secs(1)).await;
+                    continue;
+                }
+            }
+            Some(thing_else) => {
+                bail!(format!(
+                    "waiting for tasks: could not parse 'total' as a number, got '{thing_else}'"
+                ))
+            }
+            None => {
+                bail!(format!(
+                    "waiting for tasks: expected response to contain 'total', got '{response}'"
+                ))
+            }
+        }
+    }
+    Ok(())
+}
+
+fn json_eq_ignore(reference: &Value, value: &Value) -> bool {
+    match reference {
+        Value::Null | Value::Bool(_) | Value::Number(_) => reference == value,
+        Value::String(s) => (s.starts_with('[') && s.ends_with(']')) || reference == value,
+        Value::Array(values) => match value {
+            Value::Array(other_values) => {
+                if values.len() != other_values.len() {
+                    return false;
+                }
+                for (value, other_value) in values.iter().zip(other_values.iter()) {
+                    if !json_eq_ignore(value, other_value) {
+                        return false;
+                    }
+                }
+                true
+            }
+            _ => false,
+        },
+        Value::Object(map) => match value {
+            Value::Object(other_map) => {
+                if map.len() != other_map.len() {
+                    return false;
+                }
+                for (key, value) in map.iter() {
+                    match other_map.get(key) {
+                        Some(other_value) => {
+                            if !json_eq_ignore(value, other_value) {
+                                return false;
+                            }
+                        }
+                        None => return false,
+                    }
+                }
+                true
+            }
+            _ => false,
+        },
+    }
+}
+
+#[tracing::instrument(skip(client, command, assets, registered, asset_folder), fields(command = %command))]
+pub async fn run(
+    client: &Client,
+    command: &Command,
+    command_index: usize,
+    assets: &BTreeMap<String, Asset>,
+    registered: HashMap<String, Value>,
+    asset_folder: &str,
+    return_value: bool,
+) -> anyhow::Result<Option<(Value, StatusCode)>> {
+    // Try to replace variables in the route
+    let mut route = &command.route;
+    let mut owned_route;
+    if !registered.is_empty() {
+        while let (Some(pos1), Some(pos2)) = (route.find("{{"), route.rfind("}}")) {
+            if pos2 > pos1 {
+                let name = route[pos1 + 2..pos2].trim();
+                if let Some(replacement) = registered.get(name).and_then(|r| r.as_str()) {
+                    let mut new_route = String::new();
+                    new_route.push_str(&route[..pos1]);
+                    new_route.push_str(replacement);
+                    new_route.push_str(&route[pos2 + 2..]);
+                    owned_route = new_route;
+                    route = &owned_route;
+                    continue;
+                }
+            }
+            break;
+        }
+    }
+
+    // memtake the body here to leave an empty body in its place, so that command is not partially moved-out
+    let body = command
+        .body
+        .clone()
+        .get(assets, &registered, asset_folder)
+        .with_context(|| format!("while getting body for command {command}"))?;
+
+    let mut request = client.request(command.method.into(), route);
+
+    // Replace the api key
+    if let Some(var_name) = &command.api_key_variable {
+        if let Some(api_key) = registered.get(var_name).and_then(|v| v.as_str()) {
+            request = request.header("Authorization", format!("Bearer {api_key}"));
+        } else {
+            bail!("could not find API key variable '{var_name}' in registered values");
+        }
+    }
+
+    let request = if let Some((body, content_type)) = body {
+        request.body(body).header(reqwest::header::CONTENT_TYPE, content_type)
+    } else {
+        request
+    };
+
+    let response =
+        request.send().await.with_context(|| format!("error sending command: {}", command))?;
+
+    let code = response.status();
+
+    if !return_value {
+        if let Some(expected_status) = command.expected_status {
+            if code.as_u16() != expected_status {
+                let response = response
+                    .text()
+                    .await
+                    .context("could not read response body as text")
+                    .context("reading response body when checking expected status")?;
+                bail!("unexpected status code: got {}, expected {expected_status}, response body: '{response}'", code.as_u16());
+            }
+        } else if code.is_client_error() {
+            tracing::error!(%command, %code, "error in workload file");
+            let response: serde_json::Value = response
+                .json()
+                .await
+                .context("could not deserialize response as JSON")
+                .context("parsing error in workload file when sending command")?;
+            bail!(
+                "error in workload file: server responded with error code {code} and '{response}'"
+            )
+        } else if code.is_server_error() {
+            tracing::error!(%command, %code, "server error");
+            let response: serde_json::Value = response
+                .json()
+                .await
+                .context("could not deserialize response as JSON")
+                .context("parsing server error when sending command")?;
+            bail!("server error: server responded with error code {code} and '{response}'")
+        }
+    }
+
+    if let Some(expected_response) = &command.expected_response {
+        let mut evaluated_expected_response;
+
+        let expected_response = if !registered.is_empty() {
+            evaluated_expected_response = expected_response.clone();
+            insert_variables(&mut evaluated_expected_response, &registered);
+            &evaluated_expected_response
+        } else {
+            expected_response
+        };
+
+        let response: serde_json::Value = response
+            .json()
+            .await
+            .context("could not deserialize response as JSON")
+            .context("parsing response when checking expected response")?;
+        if return_value {
+            return Ok(Some((response, code)));
+        }
+        if !json_eq_ignore(expected_response, &response) {
+            let expected_pretty = serde_json::to_string_pretty(expected_response)
+                .context("serializing expected response as pretty JSON")?;
+            let response_pretty = serde_json::to_string_pretty(&response)
+                .context("serializing response as pretty JSON")?;
+            let diff = SimpleDiff::from_str(&expected_pretty, &response_pretty, "expected", "got");
+            bail!("command #{command_index} unexpected response:\n{diff}");
+        }
+    } else if return_value {
+        let response: serde_json::Value = response
+            .json()
+            .await
+            .context("could not deserialize response as JSON")
+            .context("parsing response when recording expected response")?;
+        return Ok(Some((response, code)));
+    }
+
+    Ok(None)
+}
+
+pub async fn run_commands(
+    client: &Arc<Client>,
+    commands: &[Command],
+    mut first_command_index: usize,
+    assets: &Arc<BTreeMap<String, Asset>>,
+    asset_folder: &'static str,
+    registered: &mut HashMap<String, Value>,
+    return_response: bool,
+) -> anyhow::Result<Vec<(Value, StatusCode)>> {
+    let mut responses = Vec::new();
+    for batch in
+        commands.split_inclusive(|command| !matches!(command.synchronous, SyncMode::DontWait))
+    {
+        let mut new_responses = run_batch(
+            client,
+            batch,
+            first_command_index,
+            assets,
+            asset_folder,
+            registered,
+            return_response,
+        )
+        .await?;
+        responses.append(&mut new_responses);
+
+        first_command_index += batch.len();
+    }
+
+    Ok(responses)
+}
+
+pub fn health_command() -> Command {
+    Command {
+        route: "/health".into(),
+        method: crate::common::client::Method::Get,
+        body: Default::default(),
+        register: HashMap::new(),
+        synchronous: SyncMode::WaitForResponse,
+        expected_status: None,
+        expected_response: None,
+        api_key_variable: None,
+    }
+}
+
+pub fn insert_variables(value: &mut Value, registered: &HashMap<String, Value>) {
+    match value {
+        Value::Null | Value::Bool(_) | Value::Number(_) => (),
+        Value::String(s) => {
+            if s.starts_with("{{") && s.ends_with("}}") {
+                let name = s[2..s.len() - 2].trim();
+                if let Some(replacement) = registered.get(name) {
+                    *value = replacement.clone();
+                }
+            }
+        }
+        Value::Array(values) => {
+            for value in values {
+                insert_variables(value, registered);
+            }
+        }
+        Value::Object(map) => {
+            for (_key, value) in map.iter_mut() {
+                insert_variables(value, registered);
+            }
+        }
+    }
+}

crates/xtask/src/common/instance/mod.rs

@@ -0,0 +1,113 @@
+use std::fmt::Display;
+use std::path::PathBuf;
+
+use serde::{Deserialize, Serialize};
+
+mod release;
+
+pub use release::{add_releases_to_assets, Release};
+
+/// A binary to execute on a temporary DB.
+///
+/// - The URL of the binary will be in the form <http://localhost:PORT>, where `PORT`
+///   is selected by the runner.
+/// - The database will be temporary, cleaned before use, and will be selected by the runner.
+#[derive(Clone, Debug, Serialize, Deserialize)]
+#[serde(rename_all = "camelCase")]
+pub struct Binary {
+    /// Describes how this binary should be instantiated
+    #[serde(flatten)]
+    pub source: BinarySource,
+    /// Extra CLI arguments to pass to the binary.
+    ///
+    /// Should be Meilisearch CLI options.
+    #[serde(default, skip_serializing_if = "Vec::is_empty")]
+    pub extra_cli_args: Vec<String>,
+}
+
+impl Display for Binary {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "{}", self.source)?;
+        if !self.extra_cli_args.is_empty() {
+            write!(f, "with arguments: {:?}", self.extra_cli_args)?;
+        }
+        Ok(())
+    }
+}
+
+impl Binary {
+    pub fn as_release(&self) -> Option<&Release> {
+        if let BinarySource::Release(release) = &self.source {
+            Some(release)
+        } else {
+            None
+        }
+    }
+
+    pub fn binary_path(&self, asset_folder: &str) -> anyhow::Result<Option<PathBuf>> {
+        self.source.binary_path(asset_folder)
+    }
+}
+
+#[derive(Clone, Debug, Serialize, Deserialize)]
+#[serde(rename_all = "camelCase", deny_unknown_fields, tag = "source")]
+/// Description of how to get a binary to instantiate.
+pub enum BinarySource {
+    /// Compile and run the binary from the current repository.=
+    Build {
+        #[serde(default)]
+        edition: Edition,
+    },
+    /// Get a release from GitHub
+    Release(Release),
+    /// Run the binary from the specified local path.
+    Path(PathBuf),
+}
+
+impl Display for BinarySource {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            BinarySource::Build { edition: Edition::Community } => {
+                f.write_str("git with community edition")
+            }
+            BinarySource::Build { edition: Edition::Enterprise } => {
+                f.write_str("git with enterprise edition")
+            }
+            BinarySource::Release(release) => write!(f, "{release}"),
+            BinarySource::Path(path) => write!(f, "binary at `{}`", path.display()),
+        }
+    }
+}
+
+impl Default for BinarySource {
+    fn default() -> Self {
+        Self::Build { edition: Default::default() }
+    }
+}
+
+impl BinarySource {
+    fn binary_path(&self, asset_folder: &str) -> anyhow::Result<Option<PathBuf>> {
+        Ok(match self {
+            Self::Release(release) => Some(release.binary_path(asset_folder)?),
+            Self::Build { .. } => None,
+            Self::Path(path) => Some(path.clone()),
+        })
+    }
+}
+
+#[derive(Debug, Clone, Copy, Default, Serialize, Deserialize)]
+#[serde(rename_all = "camelCase", deny_unknown_fields)]
+pub enum Edition {
+    #[default]
+    Community,
+    Enterprise,
+}
+
+impl Edition {
+    fn binary_base(&self) -> &'static str {
+        match self {
+            Edition::Community => "meilisearch",
+            Edition::Enterprise => "meilisearch-enterprise",
+        }
+    }
+}