implement a first version of the clear indexes

2025-12-08 21:55:42 +00:00 · 2023-05-04 20:25:22 +02:00
93 changed files with 1828 additions and 3172 deletions
--- a/.dockerignore
+++ b/.dockerignore
@@ -2,3 +2,4 @@ target
 Dockerfile
 .dockerignore
 .gitignore
+**/.git
--- a/.github/workflows/publish-apt-brew-pkg.yml
+++ b/.github/workflows/publish-apt-brew-pkg.yml
@@ -35,7 +35,7 @@ jobs:
    - name: Build deb package
      run: cargo deb -p meilisearch -o target/debian/meilisearch.deb
    - name: Upload debian pkg to release
-      uses: svenstaro/upload-release-action@2.6.1
+      uses: svenstaro/upload-release-action@2.5.0
      with:
        repo_token: ${{ secrets.MEILI_BOT_GH_PAT }}
        file: target/debian/meilisearch.deb
--- a/.github/workflows/publish-binaries.yml
+++ b/.github/workflows/publish-binaries.yml
@@ -54,7 +54,7 @@ jobs:
    # No need to upload binaries for dry run (cron)
    - name: Upload binaries to release
      if: github.event_name == 'release'
-      uses: svenstaro/upload-release-action@2.6.1
+      uses: svenstaro/upload-release-action@2.5.0
      with:
        repo_token: ${{ secrets.MEILI_BOT_GH_PAT }}
        file: target/release/meilisearch
@@ -87,7 +87,7 @@ jobs:
    # No need to upload binaries for dry run (cron)
    - name: Upload binaries to release
      if: github.event_name == 'release'
-      uses: svenstaro/upload-release-action@2.6.1
+      uses: svenstaro/upload-release-action@2.5.0
      with:
        repo_token: ${{ secrets.MEILI_BOT_GH_PAT }}
        file: target/release/${{ matrix.artifact_name }}
@@ -121,7 +121,7 @@ jobs:
      - name: Upload the binary to release
        # No need to upload binaries for dry run (cron)
        if: github.event_name == 'release'
-        uses: svenstaro/upload-release-action@2.6.1
+        uses: svenstaro/upload-release-action@2.5.0
        with:
          repo_token: ${{ secrets.MEILI_BOT_GH_PAT }}
          file: target/${{ matrix.target }}/release/meilisearch
@@ -183,7 +183,7 @@ jobs:
      - name: Upload the binary to release
        # No need to upload binaries for dry run (cron)
        if: github.event_name == 'release'
-        uses: svenstaro/upload-release-action@2.6.1
+        uses: svenstaro/upload-release-action@2.5.0
        with:
          repo_token: ${{ secrets.MEILI_BOT_GH_PAT }}
          file: target/${{ matrix.target }}/release/meilisearch
--- a/.github/workflows/publish-docker-images.yml
+++ b/.github/workflows/publish-docker-images.yml
@@ -58,9 +58,13 @@ jobs:

      - name: Set up QEMU
        uses: docker/setup-qemu-action@v2
+        with:
+          platforms: linux/amd64,linux/arm64

      - name: Set up Docker Buildx
        uses: docker/setup-buildx-action@v2
+        with:
+          platforms: linux/amd64,linux/arm64

      - name: Login to Docker Hub
        uses: docker/login-action@v2
@@ -88,10 +92,13 @@ jobs:
          push: true
          platforms: linux/amd64,linux/arm64
          tags: ${{ steps.meta.outputs.tags }}
+          builder: ${{ steps.buildx.outputs.name }}
          build-args: |
            COMMIT_SHA=${{ github.sha }}
            COMMIT_DATE=${{ steps.build-metadata.outputs.date }}
            GIT_TAG=${{ github.ref_name }}
+          cache-from: type=gha
+          cache-to: type=gha,mode=max

      # /!\ Don't touch this without checking with Cloud team
      - name: Send CI information to Cloud team
--- a/.github/workflows/sdks-tests.yml
+++ b/.github/workflows/sdks-tests.yml
@@ -3,11 +3,6 @@ name: SDKs tests

 on:
  workflow_dispatch:
-    inputs:
-      docker_image:
-        description: 'The Meilisearch Docker image used'
-        required: false
-        default: nightly
  schedule:
    - cron: "0 6 * * MON" # Every Monday at 6:00AM

@@ -16,28 +11,13 @@ env:
  MEILI_NO_ANALYTICS: 'true'

 jobs:
-  define-docker-image:
-    runs-on: ubuntu-latest
-    outputs:
-      docker-image: ${{ steps.define-image.outputs.docker-image }}
-    steps:
-      - uses: actions/checkout@v3
-      - name: Define the Docker image we need to use
-        id: define-image
-        run: |
-          event=${{ github.event.action }}
-          echo "docker-image=nightly" >> $GITHUB_OUTPUT
-          if [[ $event == 'workflow_dispatch' ]]; then
-            echo "docker-image=${{ github.event.inputs.docker_image }}" >> $GITHUB_OUTPUT
-          fi

  meilisearch-js-tests:
-    needs: define-docker-image
    name: JS SDK tests
    runs-on: ubuntu-latest
    services:
      meilisearch:
-        image: getmeili/meilisearch:${{ github.event.inputs.docker_image }}
+        image: getmeili/meilisearch:nightly
        env:
          MEILI_MASTER_KEY: ${{ env.MEILI_MASTER_KEY }}
          MEILI_NO_ANALYTICS: ${{ env.MEILI_NO_ANALYTICS }}
@@ -67,12 +47,11 @@ jobs:
        run: yarn test:env:browser

  instant-meilisearch-tests:
-    needs: define-docker-image
    name: instant-meilisearch tests
    runs-on: ubuntu-latest
    services:
      meilisearch:
-        image: getmeili/meilisearch:${{ github.event.inputs.docker_image }}
+        image: getmeili/meilisearch:nightly
        env:
          MEILI_MASTER_KEY: ${{ env.MEILI_MASTER_KEY }}
          MEILI_NO_ANALYTICS: ${{ env.MEILI_NO_ANALYTICS }}
@@ -94,12 +73,11 @@ jobs:
        run: yarn build

  meilisearch-php-tests:
-    needs: define-docker-image
    name: PHP SDK tests
    runs-on: ubuntu-latest
    services:
      meilisearch:
-        image: getmeili/meilisearch:${{ github.event.inputs.docker_image }}
+        image: getmeili/meilisearch:nightly
        env:
          MEILI_MASTER_KEY: ${{ env.MEILI_MASTER_KEY }}
          MEILI_NO_ANALYTICS: ${{ env.MEILI_NO_ANALYTICS }}
@@ -125,12 +103,11 @@ jobs:
          composer remove --dev guzzlehttp/guzzle http-interop/http-factory-guzzle

  meilisearch-python-tests:
-    needs: define-docker-image
    name: Python SDK tests
    runs-on: ubuntu-latest
    services:
      meilisearch:
-        image: getmeili/meilisearch:${{ github.event.inputs.docker_image }}
+        image: getmeili/meilisearch:nightly
        env:
          MEILI_MASTER_KEY: ${{ env.MEILI_MASTER_KEY }}
          MEILI_NO_ANALYTICS: ${{ env.MEILI_NO_ANALYTICS }}
@@ -150,12 +127,11 @@ jobs:
        run: pipenv run pytest

  meilisearch-go-tests:
-    needs: define-docker-image
    name: Go SDK tests
    runs-on: ubuntu-latest
    services:
      meilisearch:
-        image: getmeili/meilisearch:${{ github.event.inputs.docker_image }}
+        image: getmeili/meilisearch:nightly
        env:
          MEILI_MASTER_KEY: ${{ env.MEILI_MASTER_KEY }}
          MEILI_NO_ANALYTICS: ${{ env.MEILI_NO_ANALYTICS }}
@@ -163,7 +139,7 @@ jobs:
          - '7700:7700'
    steps:
      - name: Set up Go
-        uses: actions/setup-go@v4
+        uses: actions/setup-go@v3
        with:
          go-version: stable
      - uses: actions/checkout@v3
@@ -180,12 +156,11 @@ jobs:
        run: go test -v ./...

  meilisearch-ruby-tests:
-    needs: define-docker-image
    name: Ruby SDK tests
    runs-on: ubuntu-latest
    services:
      meilisearch:
-        image: getmeili/meilisearch:${{ github.event.inputs.docker_image }}
+        image: getmeili/meilisearch:nightly
        env:
          MEILI_MASTER_KEY: ${{ env.MEILI_MASTER_KEY }}
          MEILI_NO_ANALYTICS: ${{ env.MEILI_NO_ANALYTICS }}
@@ -205,12 +180,11 @@ jobs:
        run: bundle exec rspec

  meilisearch-rust-tests:
-    needs: define-docker-image
    name: Rust SDK tests
    runs-on: ubuntu-latest
    services:
      meilisearch:
-        image: getmeili/meilisearch:${{ github.event.inputs.docker_image }}
+        image: getmeili/meilisearch:nightly
        env:
          MEILI_MASTER_KEY: ${{ env.MEILI_MASTER_KEY }}
          MEILI_NO_ANALYTICS: ${{ env.MEILI_NO_ANALYTICS }}
--- a/.github/workflows/test-suite.yml
+++ b/.github/workflows/test-suite.yml
@@ -43,7 +43,7 @@ jobs:
          toolchain: nightly
          override: true
      - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.4.0
+        uses: Swatinem/rust-cache@v2.2.1
      - name: Run cargo check without any default features
        uses: actions-rs/cargo@v1
        with:
@@ -65,7 +65,7 @@ jobs:
    steps:
      - uses: actions/checkout@v3
      - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.4.0
+        uses: Swatinem/rust-cache@v2.2.1
      - name: Run cargo check without any default features
        uses: actions-rs/cargo@v1
        with:
@@ -105,29 +105,6 @@ jobs:
          command: test
          args: --workspace --locked --release --all-features

-  test-disabled-tokenization:
-    name: Test disabled tokenization
-    runs-on: ubuntu-latest
-    container:
-      image: ubuntu:18.04
-    if: github.event_name == 'schedule'
-    steps:
-      - uses: actions/checkout@v3
-      - name: Install needed dependencies
-        run: |
-          apt-get update
-          apt-get install --assume-yes build-essential curl
-      - uses: actions-rs/toolchain@v1 
-        with: 
-          toolchain: stable 
-          override: true 
-      - name: Run cargo tree without default features and check lindera is not present
-        run: |
-          cargo tree -f '{p} {f}' -e normal --no-default-features | grep lindera -vqz
-      - name: Run cargo tree with default features and check lindera is pressent
-        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
@@ -146,7 +123,7 @@ jobs:
          toolchain: stable
          override: true
      - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.4.0
+        uses: Swatinem/rust-cache@v2.2.1
      - name: Run tests in debug
        uses: actions-rs/cargo@v1
        with:
@@ -165,7 +142,7 @@ jobs:
          override: true
          components: clippy
      - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.4.0
+        uses: Swatinem/rust-cache@v2.2.1
      - name: Run cargo clippy
        uses: actions-rs/cargo@v1
        with:
@@ -184,7 +161,7 @@ jobs:
          override: true
          components: rustfmt
      - name: Cache dependencies
-        uses: Swatinem/rust-cache@v2.4.0
+        uses: Swatinem/rust-cache@v2.2.1
      - name: Run cargo fmt
        # Since we never ran the `build.rs` script in the benchmark directory we are missing one auto-generated import file.
        # Since we want to trigger (and fail) this action as fast as possible, instead of building the benchmark crate
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -463,7 +463,7 @@ checksum = "b645a089122eccb6111b4f81cbc1a49f5900ac4666bb93ac027feaecf15607bf"

 [[package]]
 name = "benchmarks"
-version = "1.2.0"
+version = "1.1.1"
 dependencies = [
 "anyhow",
 "bytes",
@@ -1207,15 +1207,9 @@ dependencies = [
 "winapi",
 ]

-[[package]]
-name = "doc-comment"
-version = "0.3.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fea41bba32d969b513997752735605054bc0dfa92b4c56bf1189f2e174be7a10"
-
 [[package]]
 name = "dump"
-version = "1.2.0"
+version = "1.1.1"
 dependencies = [
 "anyhow",
 "big_s",
@@ -1434,7 +1428,7 @@ dependencies = [

 [[package]]
 name = "file-store"
-version = "1.2.0"
+version = "1.1.1"
 dependencies = [
 "faux",
 "tempfile",
@@ -1456,7 +1450,7 @@ dependencies = [

 [[package]]
 name = "filter-parser"
-version = "1.2.0"
+version = "1.1.1"
 dependencies = [
 "insta",
 "nom",
@@ -1482,7 +1476,7 @@ dependencies = [

 [[package]]
 name = "flatten-serde-json"
-version = "1.2.0"
+version = "1.1.1"
 dependencies = [
 "criterion",
 "serde_json",
@@ -1769,15 +1763,6 @@ dependencies = [
 "byteorder",
 ]

-[[package]]
-name = "hashbrown"
-version = "0.11.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ab5ef0d4909ef3724cc8cce6ccc8572c5c817592e9285f5464f8e86f8bd3726e"
-dependencies = [
- "ahash 0.7.6",
-]
-
 [[package]]
 name = "hashbrown"
 version = "0.12.3"
@@ -1809,7 +1794,7 @@ checksum = "95505c38b4572b2d910cecb0281560f54b440a19336cbbcb27bf6ce6adc6f5a8"
 [[package]]
 name = "heed"
 version = "0.12.5"
-source = "git+https://github.com/meilisearch/heed?tag=v0.12.6#8c5b94225fc949c02bb7b900cc50ffaf6b584b1e"
+source = "git+https://github.com/meilisearch/heed?tag=v0.12.5#4158a6c484752afaaf9e2530a6ee0e7ab0f24ee8"
 dependencies = [
 "byteorder",
 "heed-traits",
@@ -1826,12 +1811,12 @@ dependencies = [
 [[package]]
 name = "heed-traits"
 version = "0.7.0"
-source = "git+https://github.com/meilisearch/heed?tag=v0.12.6#8c5b94225fc949c02bb7b900cc50ffaf6b584b1e"
+source = "git+https://github.com/meilisearch/heed?tag=v0.12.5#4158a6c484752afaaf9e2530a6ee0e7ab0f24ee8"

 [[package]]
 name = "heed-types"
 version = "0.7.2"
-source = "git+https://github.com/meilisearch/heed?tag=v0.12.6#8c5b94225fc949c02bb7b900cc50ffaf6b584b1e"
+source = "git+https://github.com/meilisearch/heed?tag=v0.12.5#4158a6c484752afaaf9e2530a6ee0e7ab0f24ee8"
 dependencies = [
 "bincode",
 "heed-traits",
@@ -1879,22 +1864,6 @@ dependencies = [
 "digest",
 ]

-[[package]]
-name = "hnsw"
-version = "0.11.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2b9740ebf8769ec4ad6762cc951ba18f39bba6dfbc2fbbe46285f7539af79752"
-dependencies = [
- "ahash 0.7.6",
- "hashbrown 0.11.2",
- "libm",
- "num-traits",
- "rand_core",
- "serde",
- "smallvec",
- "space",
-]
-
 [[package]]
 name = "http"
 version = "0.2.9"
@@ -1990,7 +1959,7 @@ dependencies = [

 [[package]]
 name = "index-scheduler"
-version = "1.2.0"
+version = "1.1.1"
 dependencies = [
 "anyhow",
 "big_s",
@@ -2025,7 +1994,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "bd070e393353796e801d209ad339e89596eb4c8d430d18ede6a1cced8fafbd99"
 dependencies = [
 "autocfg",
- "hashbrown 0.12.3",
+ "hashbrown",
 "serde",
 ]

@@ -2117,7 +2086,7 @@ checksum = "37228e06c75842d1097432d94d02f37fe3ebfca9791c2e8fef6e9db17ed128c1"
 dependencies = [
 "cedarwood",
 "fxhash",
- "hashbrown 0.12.3",
+ "hashbrown",
 "lazy_static",
 "phf",
 "phf_codegen",
@@ -2144,7 +2113,7 @@ dependencies = [

 [[package]]
 name = "json-depth-checker"
-version = "1.2.0"
+version = "1.1.1"
 dependencies = [
 "criterion",
 "serde_json",
@@ -2570,7 +2539,7 @@ checksum = "490cc448043f947bae3cbee9c203358d62dbee0db12107a74be5c30ccfd09771"

 [[package]]
 name = "meili-snap"
-version = "1.2.0"
+version = "1.1.1"
 dependencies = [
 "insta",
 "md5",
@@ -2579,7 +2548,7 @@ dependencies = [

 [[package]]
 name = "meilisearch"
-version = "1.2.0"
+version = "1.1.1"
 dependencies = [
 "actix-cors",
 "actix-http",
@@ -2627,7 +2596,6 @@ dependencies = [
 "num_cpus",
 "obkv",
 "once_cell",
- "ordered-float",
 "parking_lot",
 "permissive-json-pointer",
 "pin-project-lite",
@@ -2668,7 +2636,7 @@ dependencies = [

 [[package]]
 name = "meilisearch-auth"
-version = "1.2.0"
+version = "1.1.1"
 dependencies = [
 "base64 0.21.0",
 "enum-iterator",
@@ -2687,7 +2655,7 @@ dependencies = [

 [[package]]
 name = "meilisearch-types"
-version = "1.2.0"
+version = "1.1.1"
 dependencies = [
 "actix-web",
 "anyhow",
@@ -2741,13 +2709,12 @@ dependencies = [

 [[package]]
 name = "milli"
-version = "1.2.0"
+version = "1.1.1"
 dependencies = [
 "big_s",
 "bimap",
 "bincode",
 "bstr",
- "bytemuck",
 "byteorder",
 "charabia",
 "concat-arrays",
@@ -2763,7 +2730,6 @@ dependencies = [
 "geoutils",
 "grenad",
 "heed",
- "hnsw",
 "insta",
 "itertools",
 "json-depth-checker",
@@ -2778,7 +2744,6 @@ dependencies = [
 "once_cell",
 "ordered-float",
 "rand",
- "rand_pcg",
 "rayon",
 "roaring",
 "rstar",
@@ -2788,7 +2753,6 @@ dependencies = [
 "smallstr",
 "smallvec",
 "smartstring",
- "space",
 "tempfile",
 "thiserror",
 "time",
@@ -2977,9 +2941,9 @@ checksum = "624a8340c38c1b80fd549087862da4ba43e08858af025b236e509b6649fc13d5"

 [[package]]
 name = "ordered-float"
-version = "3.7.0"
+version = "3.6.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "2fc2dbde8f8a79f2102cc474ceb0ad68e3b80b85289ea62389b60e66777e4213"
+checksum = "13a384337e997e6860ffbaa83708b2ef329fd8c54cb67a5f64d421e0f943254f"
 dependencies = [
 "num-traits",
 ]
@@ -3100,7 +3064,7 @@ checksum = "478c572c3d73181ff3c2539045f6eb99e5491218eae919370993b890cdbdd98e"

 [[package]]
 name = "permissive-json-pointer"
-version = "1.2.0"
+version = "1.1.1"
 dependencies = [
 "big_s",
 "serde_json",
@@ -3363,16 +3327,6 @@ dependencies = [
 "getrandom",
 ]

-[[package]]
-name = "rand_pcg"
-version = "0.3.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "59cad018caf63deb318e5a4586d99a24424a364f40f1e5778c29aca23f4fc73e"
-dependencies = [
- "rand_core",
- "serde",
-]
-
 [[package]]
 name = "rayon"
 version = "1.7.0"
@@ -3810,9 +3764,6 @@ name = "smallvec"
 version = "1.10.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "a507befe795404456341dfab10cef66ead4c041f62b8b11bbb92bffe5d0953e0"
-dependencies = [
- "serde",
-]

 [[package]]
 name = "smartstring"
@@ -3835,16 +3786,6 @@ dependencies = [
 "winapi",
 ]

-[[package]]
-name = "space"
-version = "0.17.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c5ab9701ae895386d13db622abf411989deff7109b13b46b6173bb4ce5c1d123"
-dependencies = [
- "doc-comment",
- "num-traits",
-]
-
 [[package]]
 name = "spin"
 version = "0.5.2"
@@ -4492,7 +4433,7 @@ version = "0.16.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9c531a2dc4c462b833788be2c07eef4e621d0e9edbd55bf280cc164c1c1aa043"
 dependencies = [
- "hashbrown 0.12.3",
+ "hashbrown",
 "once_cell",
 ]

--- a/Cargo.toml
+++ b/Cargo.toml
@@ -17,7 +17,7 @@ members = [
 ]

 [workspace.package]
-version = "1.2.0"
+version = "1.1.1"
 authors = ["Quentin de Quelen <quentin@dequelen.me>", "Clément Renault <clement@meilisearch.com>"]
 description = "Meilisearch HTTP server"
 homepage = "https://meilisearch.com"
--- a/5
+++ b/5
@@ -1,3 +1,4 @@
+# syntax=docker/dockerfile:1.4
 # Compile
 FROM    rust:alpine3.16 AS compiler

@@ -11,7 +12,7 @@ ARG     GIT_TAG
 ENV     VERGEN_GIT_SHA=${COMMIT_SHA} VERGEN_GIT_COMMIT_TIMESTAMP=${COMMIT_DATE} VERGEN_GIT_SEMVER_LIGHTWEIGHT=${GIT_TAG}
 ENV     RUSTFLAGS="-C target-feature=-crt-static"

-COPY    . .
+COPY    --link . .
 RUN     set -eux; \
        apkArch="$(apk --print-arch)"; \
        if [ "$apkArch" = "aarch64" ]; then \
@@ -30,7 +31,7 @@ RUN     apk update --quiet \

 # add meilisearch to the `/bin` so you can run it from anywhere and it's easy
 # to find.
-COPY    --from=compiler /meilisearch/target/release/meilisearch /bin/meilisearch
+COPY    --from=compiler --link /meilisearch/target/release/meilisearch /bin/meilisearch
 # To stay compatible with the older version of the container (pre v0.27.0) we're
 # going to symlink the meilisearch binary in the path to `/meilisearch`
 RUN     ln -s /bin/meilisearch /meilisearch
--- a/assets/grafana-dashboard.json
+++ b/assets/grafana-dashboard.json
--- a/assets/prometheus-basic-scraper.yml
+++ b/assets/prometheus-basic-scraper.yml
@@ -1,19 +0,0 @@
-global:
-  scrape_interval:     15s # By default, scrape targets every 15 seconds.
-
-  # Attach these labels to any time series or alerts when communicating with
-  # external systems (federation, remote storage, Alertmanager).
-  external_labels:
-    monitor: 'codelab-monitor'
-
-# A scrape configuration containing exactly one endpoint to scrape:
-# Here it's Prometheus itself.
-scrape_configs:
-  # The job name is added as a label `job=<job_name>` to any timeseries scraped from this config.
-  - job_name: 'meilisearch'
-
-    # Override the global default and scrape targets from this job every 5 seconds.
-    scrape_interval: 5s
-
-    static_configs:
-      - targets: ['localhost:7700']
--- a/benchmarks/Cargo.toml
+++ b/benchmarks/Cargo.toml
@@ -13,7 +13,7 @@ license.workspace = true
 [dependencies]
 anyhow = "1.0.70"
 csv = "1.2.1"
-milli = { path = "../milli" }
+milli = { path = "../milli", default-features = false }
 mimalloc = { version = "0.1.36", default-features = false }
 serde_json = { version = "1.0.95", features = ["preserve_order"] }

@@ -31,7 +31,7 @@ flate2 = "1.0.25"
 reqwest = { version = "0.11.16", features = ["blocking", "rustls-tls"], default-features = false }

 [features]
-default = ["milli/all-tokenizations"]
+default = ["milli/default"]

 [[bench]]
 name = "search_songs"
--- a/config.toml
+++ b/config.toml
@@ -1,131 +1,128 @@
 # This file shows the default configuration of Meilisearch.
 # All variables are defined here: https://www.meilisearch.com/docs/learn/configuration/instance_options#environment-variables

+db_path = "./data.ms"
 # Designates the location where database files will be created and retrieved.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#database-path
-db_path = "./data.ms"

+env = "development"
 # Configures the instance's environment. Value must be either `production` or `development`.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#environment
-env = "development"

-# The address on which the HTTP server will listen.
 http_addr = "localhost:7700"
+# The address on which the HTTP server will listen.

+# master_key = "YOUR_MASTER_KEY_VALUE"
 # Sets the instance's master key, automatically protecting all routes except GET /health.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#master-key
-# master_key = "YOUR_MASTER_KEY_VALUE"

+# no_analytics = true
 # Deactivates Meilisearch's built-in telemetry when provided.
 # Meilisearch automatically collects data from all instances that do not opt out using this flag.
 # All gathered data is used solely for the purpose of improving Meilisearch, and can be deleted at any time.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#disable-analytics
-# no_analytics = true

+http_payload_size_limit = "100 MB"
 # Sets the maximum size of accepted payloads.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#payload-limit-size
-http_payload_size_limit = "100 MB"

+log_level = "INFO"
 # Defines how much detail should be present in Meilisearch's logs.
 # Meilisearch currently supports six log levels, listed in order of increasing verbosity:  `OFF`, `ERROR`, `WARN`, `INFO`, `DEBUG`, `TRACE`
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#log-level
-log_level = "INFO"

+# max_indexing_memory = "2 GiB"
 # Sets the maximum amount of RAM Meilisearch can use when indexing.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#max-indexing-memory
-# max_indexing_memory = "2 GiB"

+# max_indexing_threads = 4
 # Sets the maximum number of threads Meilisearch can use during indexing.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#max-indexing-threads
-# max_indexing_threads = 4

 #############
 ### DUMPS ###
 #############

+dump_dir = "dumps/"
 # Sets the directory where Meilisearch will create dump files.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#dump-directory
-dump_dir = "dumps/"

+# import_dump = "./path/to/my/file.dump"
 # Imports the dump file located at the specified path. Path must point to a .dump file.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#import-dump
-# import_dump = "./path/to/my/file.dump"

+ignore_missing_dump = false
 # Prevents Meilisearch from throwing an error when `import_dump` does not point to a valid dump file.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#ignore-missing-dump
-ignore_missing_dump = false

+ignore_dump_if_db_exists = false
 # Prevents a Meilisearch instance with an existing database from throwing an error when using `import_dump`.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#ignore-dump-if-db-exists
-ignore_dump_if_db_exists = false


 #################
 ### SNAPSHOTS ###
 #################

+schedule_snapshot = false
 # Enables scheduled snapshots when true, disable when false (the default).
 # If the value is given as an integer, then enables the scheduled snapshot with the passed value as the interval
 # between each snapshot, in seconds.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#schedule-snapshot-creation
-schedule_snapshot = false

+snapshot_dir = "snapshots/"
 # Sets the directory where Meilisearch will store snapshots.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#snapshot-destination
-snapshot_dir = "snapshots/"

+# import_snapshot = "./path/to/my/snapshot"
 # Launches Meilisearch after importing a previously-generated snapshot at the given filepath.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#import-snapshot
-# import_snapshot = "./path/to/my/snapshot"

+ignore_missing_snapshot = false
 # Prevents a Meilisearch instance from throwing an error when `import_snapshot` does not point to a valid snapshot file.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#ignore-missing-snapshot
-ignore_missing_snapshot = false

+ignore_snapshot_if_db_exists = false
 # Prevents a Meilisearch instance with an existing database from throwing an error when using `import_snapshot`.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#ignore-snapshot-if-db-exists
-ignore_snapshot_if_db_exists = false


 ###########
 ### SSL ###
 ###########

+# ssl_auth_path = "./path/to/root"
 # Enables client authentication in the specified path.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#ssl-authentication-path
-# ssl_auth_path = "./path/to/root"

+# ssl_cert_path = "./path/to/certfile"
 # Sets the server's SSL certificates.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#ssl-certificates-path
-# ssl_cert_path = "./path/to/certfile"

+# ssl_key_path = "./path/to/private-key"
 # Sets the server's SSL key files.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#ssl-key-path
-# ssl_key_path = "./path/to/private-key"

+# ssl_ocsp_path = "./path/to/ocsp-file"
 # Sets the server's OCSP file.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#ssl-ocsp-path
-# ssl_ocsp_path = "./path/to/ocsp-file"

+ssl_require_auth = false
 # Makes SSL authentication mandatory.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#ssl-require-auth
-ssl_require_auth = false

+ssl_resumption = false
 # Activates SSL session resumption.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#ssl-resumption
-ssl_resumption = false

+ssl_tickets = false
 # Activates SSL tickets.
 # https://www.meilisearch.com/docs/learn/configuration/instance_options#ssl-tickets
-ssl_tickets = false

 #############################
 ### Experimental features ###
 #############################

+experimental_enable_metrics = false
 # Experimental metrics feature. For more information, see: <https://github.com/meilisearch/meilisearch/discussions/3518>
 # Enables the Prometheus metrics on the `GET /metrics` endpoint.
-experimental_enable_metrics = false
-
-# Experimental RAM reduction during indexing, do not use in production, see: <https://github.com/meilisearch/product/discussions/652>
-experimental_reduce_indexing_memory_usage = false
--- a/dump/src/lib.rs
+++ b/dump/src/lib.rs
@@ -110,6 +110,9 @@ pub enum KindDump {
        allow_index_creation: bool,
    },
    IndexDeletion,
+    IndexClear {
+        index_uids: Vec<String>,
+    },
    IndexCreation {
        primary_key: Option<String>,
    },
@@ -180,6 +183,7 @@ impl From<KindWithContent> for KindDump {
                ..
            } => KindDump::Settings { settings: new_settings, is_deletion, allow_index_creation },
            KindWithContent::IndexDeletion { .. } => KindDump::IndexDeletion,
+            KindWithContent::IndexClear { index_uids } => KindDump::IndexClear { index_uids },
            KindWithContent::IndexCreation { primary_key, .. } => {
                KindDump::IndexCreation { primary_key }
            }
@@ -211,8 +215,8 @@ pub(crate) mod test {
    use maplit::btreeset;
    use meilisearch_types::index_uid_pattern::IndexUidPattern;
    use meilisearch_types::keys::{Action, Key};
+    use meilisearch_types::milli;
    use meilisearch_types::milli::update::Setting;
-    use meilisearch_types::milli::{self};
    use meilisearch_types::settings::{Checked, Settings};
    use meilisearch_types::tasks::{Details, Status};
    use serde_json::{json, Map, Value};
--- a/grafana-dashboards/dashboard.json
+++ b/grafana-dashboards/dashboard.json
--- a/index-scheduler/src/autobatcher.rs
+++ b/index-scheduler/src/autobatcher.rs
@@ -32,6 +32,7 @@ enum AutobatchKind {
    },
    IndexCreation,
    IndexDeletion,
+    IndexClear,
    IndexUpdate,
    IndexSwap,
 }
@@ -74,6 +75,7 @@ impl From<KindWithContent> for AutobatchKind {
                }
            }
            KindWithContent::IndexDeletion { .. } => AutobatchKind::IndexDeletion,
+            KindWithContent::IndexClear { .. } => AutobatchKind::IndexClear,
            KindWithContent::IndexCreation { .. } => AutobatchKind::IndexCreation,
            KindWithContent::IndexUpdate { .. } => AutobatchKind::IndexUpdate,
            KindWithContent::IndexSwap { .. } => AutobatchKind::IndexSwap,
@@ -123,6 +125,9 @@ pub enum BatchKind {
    IndexDeletion {
        ids: Vec<TaskId>,
    },
+    IndexClear {
+        id: TaskId,
+    },
    IndexCreation {
        id: TaskId,
    },
@@ -173,6 +178,7 @@ impl BatchKind {
        match AutobatchKind::from(kind) {
            K::IndexCreation => (Break(BatchKind::IndexCreation { id: task_id }), true),
            K::IndexDeletion => (Break(BatchKind::IndexDeletion { ids: vec![task_id] }), false),
+            K::IndexClear => (Break(BatchKind::IndexClear { id: task_id }), false),
            K::IndexUpdate => (Break(BatchKind::IndexUpdate { id: task_id }), false),
            K::IndexSwap => (Break(BatchKind::IndexSwap { id: task_id }), false),
            K::DocumentClear => (Continue(BatchKind::DocumentClear { ids: vec![task_id] }), false),
@@ -222,7 +228,7 @@ impl BatchKind {

        match (self, kind) {
            // We don't batch any of these operations
-            (this, K::IndexCreation | K::IndexUpdate | K::IndexSwap | K::DocumentDeletionByFilter) => Break(this),
+            (this, K::IndexCreation | K::IndexUpdate | K::IndexClear | K::IndexSwap | K::DocumentDeletionByFilter) => Break(this),
            // We must not batch tasks that don't have the same index creation rights if the index doesn't already exists.
            (this, kind) if !index_already_exists && this.allow_index_creation() == Some(false) && kind.allow_index_creation() == Some(true) => {
                Break(this)
@@ -480,6 +486,7 @@ impl BatchKind {
            (
                BatchKind::IndexCreation { .. }
                | BatchKind::IndexDeletion { .. }
+                | BatchKind::IndexClear { .. }
                | BatchKind::IndexUpdate { .. }
                | BatchKind::IndexSwap { .. }
                | BatchKind::DocumentDeletionByFilter { .. },
--- a/index-scheduler/src/batch.rs
+++ b/index-scheduler/src/batch.rs
@@ -24,7 +24,6 @@ use std::io::BufWriter;

 use dump::IndexMetadata;
 use log::{debug, error, info};
-use meilisearch_types::error::Code;
 use meilisearch_types::heed::{RoTxn, RwTxn};
 use meilisearch_types::milli::documents::{obkv_to_object, DocumentsBatchReader};
 use meilisearch_types::milli::heed::CompactionOption;
@@ -86,6 +85,10 @@ pub(crate) enum Batch {
        tasks: Vec<Task>,
        index_has_been_created: bool,
    },
+    IndexClear {
+        index_uids: Vec<String>,
+        task: Task,
+    },
    IndexSwap {
        task: Task,
    },
@@ -155,6 +158,7 @@ impl Batch {
            | Batch::TaskDeletion(task)
            | Batch::Dump(task)
            | Batch::IndexCreation { task, .. }
+            | Batch::IndexClear { task, .. }
            | Batch::IndexDocumentDeletionByFilter { task, .. }
            | Batch::IndexUpdate { task, .. } => vec![task.uid],
            Batch::SnapshotCreation(tasks) | Batch::IndexDeletion { tasks, .. } => {
@@ -190,6 +194,7 @@ impl Batch {
            | TaskDeletion(_)
            | SnapshotCreation(_)
            | Dump(_)
+            | IndexClear { .. }
            | IndexSwap { .. } => None,
            IndexOperation { op, .. } => Some(op.index_uid()),
            IndexCreation { index_uid, .. }
@@ -454,6 +459,14 @@ impl IndexScheduler {
                index_has_been_created: must_create_index,
                tasks: self.get_existing_tasks(rtxn, ids)?,
            })),
+            BatchKind::IndexClear { id } => {
+                let task = self.get_task(rtxn, id)?.ok_or(Error::CorruptedTaskQueue)?;
+                let index_uids = match &task.kind {
+                    KindWithContent::IndexClear { index_uids } => index_uids.clone(),
+                    _ => unreachable!(),
+                };
+                Ok(Some(Batch::IndexClear { index_uids, task }))
+            }
            BatchKind::IndexSwap { id } => {
                let task = self.get_task(rtxn, id)?.ok_or(Error::CorruptedTaskQueue)?;
                Ok(Some(Batch::IndexSwap { task }))
@@ -1018,6 +1031,13 @@ impl IndexScheduler {

                Ok(tasks)
            }
+            Batch::IndexClear { index_uids, mut task } => {
+                let wtxn = self.env.write_txn()?;
+                self.index_mapper.delete_indexes(wtxn, index_uids, false)?;
+                task.status = Status::Succeeded;
+
+                Ok(vec![task])
+            }
            Batch::IndexSwap { mut task } => {
                let mut wtxn = self.env.write_txn()?;
                let swaps = if let KindWithContent::IndexSwap { swaps } = &task.kind {
@@ -1492,12 +1512,7 @@ fn delete_document_by_filter(filter: &serde_json::Value, index: Index) -> Result
    Ok(if let Some(filter) = filter {
        let mut wtxn = index.write_txn()?;

-        let candidates = filter.evaluate(&wtxn, &index).map_err(|err| match err {
-            milli::Error::UserError(milli::UserError::InvalidFilter(_)) => {
-                Error::from(err).with_custom_error_code(Code::InvalidDocumentFilter)
-            }
-            e => e.into(),
-        })?;
+        let candidates = filter.evaluate(&wtxn, &index)?;
        let mut delete_operation = DeleteDocuments::new(&mut wtxn, &index)?;
        delete_operation.delete_documents(&candidates);
        let deleted_documents =
--- a/index-scheduler/src/error.rs
+++ b/index-scheduler/src/error.rs
@@ -46,8 +46,6 @@ impl From<DateField> for Code {
 #[allow(clippy::large_enum_variant)]
 #[derive(Error, Debug)]
 pub enum Error {
-    #[error("{1}")]
-    WithCustomErrorCode(Code, Box<Self>),
    #[error("Index `{0}` not found.")]
    IndexNotFound(String),
    #[error("Index `{0}` already exists.")]
@@ -146,7 +144,6 @@ impl Error {
    pub fn is_recoverable(&self) -> bool {
        match self {
            Error::IndexNotFound(_)
-            | Error::WithCustomErrorCode(_, _)
            | Error::IndexAlreadyExists(_)
            | Error::SwapDuplicateIndexFound(_)
            | Error::SwapDuplicateIndexesFound(_)
@@ -179,16 +176,11 @@ impl Error {
            Error::PlannedFailure => false,
        }
    }
-
-    pub fn with_custom_error_code(self, code: Code) -> Self {
-        Self::WithCustomErrorCode(code, Box::new(self))
-    }
 }

 impl ErrorCode for Error {
    fn error_code(&self) -> Code {
        match self {
-            Error::WithCustomErrorCode(code, _) => *code,
            Error::IndexNotFound(_) => Code::IndexNotFound,
            Error::IndexAlreadyExists(_) => Code::IndexAlreadyExists,
            Error::SwapDuplicateIndexesFound(_) => Code::InvalidSwapDuplicateIndexFound,
--- a/index-scheduler/src/index_mapper/index_map.rs
+++ b/index-scheduler/src/index_mapper/index_map.rs
@@ -5,7 +5,6 @@ use std::collections::BTreeMap;
 use std::path::Path;
 use std::time::Duration;

-use meilisearch_types::heed::flags::Flags;
 use meilisearch_types::heed::{EnvClosingEvent, EnvOpenOptions};
 use meilisearch_types::milli::Index;
 use time::OffsetDateTime;
@@ -54,7 +53,6 @@ pub struct IndexMap {
 pub struct ClosingIndex {
    uuid: Uuid,
    closing_event: EnvClosingEvent,
-    enable_mdb_writemap: bool,
    map_size: usize,
    generation: usize,
 }
@@ -70,7 +68,6 @@ impl ClosingIndex {
    pub fn wait_timeout(self, timeout: Duration) -> Option<ReopenableIndex> {
        self.closing_event.wait_timeout(timeout).then_some(ReopenableIndex {
            uuid: self.uuid,
-            enable_mdb_writemap: self.enable_mdb_writemap,
            map_size: self.map_size,
            generation: self.generation,
        })
@@ -79,7 +76,6 @@ impl ClosingIndex {

 pub struct ReopenableIndex {
    uuid: Uuid,
-    enable_mdb_writemap: bool,
    map_size: usize,
    generation: usize,
 }
@@ -107,7 +103,7 @@ impl ReopenableIndex {
                return Ok(());
            }
            map.unavailable.remove(&self.uuid);
-            map.create(&self.uuid, path, None, self.enable_mdb_writemap, self.map_size)?;
+            map.create(&self.uuid, path, None, self.map_size)?;
        }
        Ok(())
    }
@@ -174,17 +170,16 @@ impl IndexMap {
        uuid: &Uuid,
        path: &Path,
        date: Option<(OffsetDateTime, OffsetDateTime)>,
-        enable_mdb_writemap: bool,
        map_size: usize,
    ) -> Result<Index> {
        if !matches!(self.get_unavailable(uuid), Missing) {
            panic!("Attempt to open an index that was unavailable");
        }
-        let index = create_or_open_index(path, date, enable_mdb_writemap, map_size)?;
+        let index = create_or_open_index(path, date, map_size)?;
        match self.available.insert(*uuid, index.clone()) {
            InsertionOutcome::InsertedNew => (),
            InsertionOutcome::Evicted(evicted_uuid, evicted_index) => {
-                self.close(evicted_uuid, evicted_index, enable_mdb_writemap, 0);
+                self.close(evicted_uuid, evicted_index, 0);
            }
            InsertionOutcome::Replaced(_) => {
                panic!("Attempt to open an index that was already opened")
@@ -217,30 +212,17 @@ impl IndexMap {
    /// | Closing         | Closing       |
    /// | Available       | Closing       |
    ///
-    pub fn close_for_resize(
-        &mut self,
-        uuid: &Uuid,
-        enable_mdb_writemap: bool,
-        map_size_growth: usize,
-    ) {
+    pub fn close_for_resize(&mut self, uuid: &Uuid, map_size_growth: usize) {
        let Some(index) = self.available.remove(uuid) else { return; };
-        self.close(*uuid, index, enable_mdb_writemap, map_size_growth);
+        self.close(*uuid, index, map_size_growth);
    }

-    fn close(
-        &mut self,
-        uuid: Uuid,
-        index: Index,
-        enable_mdb_writemap: bool,
-        map_size_growth: usize,
-    ) {
+    fn close(&mut self, uuid: Uuid, index: Index, map_size_growth: usize) {
        let map_size = index.map_size().unwrap_or(DEFAULT_MAP_SIZE) + map_size_growth;
        let closing_event = index.prepare_for_closing();
        let generation = self.next_generation();
-        self.unavailable.insert(
-            uuid,
-            Some(ClosingIndex { uuid, closing_event, enable_mdb_writemap, map_size, generation }),
-        );
+        self.unavailable
+            .insert(uuid, Some(ClosingIndex { uuid, closing_event, map_size, generation }));
    }

    /// Attempts to delete and index.
@@ -300,15 +282,11 @@ impl IndexMap {
 fn create_or_open_index(
    path: &Path,
    date: Option<(OffsetDateTime, OffsetDateTime)>,
-    enable_mdb_writemap: bool,
    map_size: usize,
 ) -> Result<Index> {
    let mut options = EnvOpenOptions::new();
    options.map_size(clamp_to_page_size(map_size));
    options.max_readers(1024);
-    if enable_mdb_writemap {
-        unsafe { options.flag(Flags::MdbWriteMap) };
-    }

    if let Some((created, updated)) = date {
        Ok(Index::new_with_creation_dates(options, path, created, updated)?)
--- a/index-scheduler/src/index_mapper/mod.rs
+++ b/index-scheduler/src/index_mapper/mod.rs
@@ -66,8 +66,6 @@ pub struct IndexMapper {
    index_base_map_size: usize,
    /// The quantity by which the map size of an index is incremented upon reopening, in bytes.
    index_growth_amount: usize,
-    /// Whether we open a meilisearch index with the MDB_WRITEMAP option or not.
-    enable_mdb_writemap: bool,
    pub indexer_config: Arc<IndexerConfig>,
 }

@@ -90,17 +88,8 @@ pub enum IndexStatus {
 pub struct IndexStats {
    /// Number of documents in the index.
    pub number_of_documents: u64,
-    /// Size taken up by the index' DB, in bytes.
-    ///
-    /// This includes the size taken by both the used and free pages of the DB, and as the free pages
-    /// are not returned to the disk after a deletion, this number is typically larger than
-    /// `used_database_size` that only includes the size of the used pages.
+    /// Size of the index' DB, in bytes.
    pub database_size: u64,
-    /// Size taken by the used pages of the index' DB, in bytes.
-    ///
-    /// As the DB backend does not return to the disk the pages that are not currently used by the DB,
-    /// this value is typically smaller than `database_size`.
-    pub used_database_size: u64,
    /// Association of every field name with the number of times it occurs in the documents.
    pub field_distribution: FieldDistribution,
    /// Creation date of the index.
@@ -116,10 +105,10 @@ impl IndexStats {
    ///
    /// - rtxn: a RO transaction for the index, obtained from `Index::read_txn()`.
    pub fn new(index: &Index, rtxn: &RoTxn) -> Result<Self> {
+        let database_size = index.on_disk_size()?;
        Ok(IndexStats {
            number_of_documents: index.number_of_documents(rtxn)?,
-            database_size: index.on_disk_size()?,
-            used_database_size: index.used_size()?,
+            database_size,
            field_distribution: index.field_distribution(rtxn)?,
            created_at: index.created_at(rtxn)?,
            updated_at: index.updated_at(rtxn)?,
@@ -134,22 +123,15 @@ impl IndexMapper {
        index_base_map_size: usize,
        index_growth_amount: usize,
        index_count: usize,
-        enable_mdb_writemap: bool,
        indexer_config: IndexerConfig,
    ) -> Result<Self> {
-        let mut wtxn = env.write_txn()?;
-        let index_mapping = env.create_database(&mut wtxn, Some(INDEX_MAPPING))?;
-        let index_stats = env.create_database(&mut wtxn, Some(INDEX_STATS))?;
-        wtxn.commit()?;
-
        Ok(Self {
            index_map: Arc::new(RwLock::new(IndexMap::new(index_count))),
-            index_mapping,
-            index_stats,
+            index_mapping: env.create_database(Some(INDEX_MAPPING))?,
+            index_stats: env.create_database(Some(INDEX_STATS))?,
            base_path,
            index_base_map_size,
            index_growth_amount,
-            enable_mdb_writemap,
            indexer_config: Arc::new(indexer_config),
        })
    }
@@ -180,7 +162,6 @@ impl IndexMapper {
                    &uuid,
                    &index_path,
                    date,
-                    self.enable_mdb_writemap,
                    self.index_base_map_size,
                )?;

@@ -192,19 +173,37 @@ impl IndexMapper {
        }
    }

+    pub fn delete_index(&self, wtxn: RwTxn, name: &str) -> Result<()> {
+        self.delete_indexes(wtxn, Some(name), true)
+    }
+
    /// Removes the index from the mapping table and the in-memory index map
    /// but keeps the associated tasks.
-    pub fn delete_index(&self, mut wtxn: RwTxn, name: &str) -> Result<()> {
-        let uuid = self
-            .index_mapping
-            .get(&wtxn, name)?
-            .ok_or_else(|| Error::IndexNotFound(name.to_string()))?;
+    pub fn delete_indexes(
+        &self,
+        mut wtxn: RwTxn,
+        names: impl IntoIterator<Item = impl AsRef<str>>,
+        error_on_missing_index: bool,
+    ) -> Result<()> {
+        let indexes = names
+            .into_iter()
+            .map(|name| {
+                let name = name.as_ref().to_string();
+                let uuid = self
+                    .index_mapping
+                    .get(&wtxn, &name)?
+                    .ok_or_else(|| Error::IndexNotFound(name.to_string()))?;
+                Ok((name, uuid))
+            })
+            .filter(|res| error_on_missing_index || res.is_ok())
+            .collect::<Result<Vec<_>>>()?;

-        // Not an error if the index had no stats in cache.
-        self.index_stats.delete(&mut wtxn, &uuid)?;
-
-        // Once we retrieved the UUID of the index we remove it from the mapping table.
-        assert!(self.index_mapping.delete(&mut wtxn, name)?);
+        for (name, uuid) in indexes.iter() {
+            // Not an error if the index had no stats in cache.
+            self.index_stats.delete(&mut wtxn, uuid)?;
+            // Once we retrieved the UUID of the index we remove it from the mapping table.
+            assert!(self.index_mapping.delete(&mut wtxn, name)?);
+        }

        wtxn.commit()?;

@@ -222,51 +221,63 @@ impl IndexMapper {
        //    This can not be caused by indexation because deleting an index happens in the scheduler itself, so cannot be concurrent with indexation.
        //
        // In these situations, reporting the error through a panic is in order.
-        let closing_event = loop {
-            let mut lock = self.index_map.write().unwrap();
-            match lock.start_deletion(&uuid) {
-                Ok(env_closing) => break env_closing,
-                Err(Some(reopen)) => {
-                    // drop the lock here so that we don't synchronously wait for the index to close.
-                    drop(lock);
-                    tries += 1;
-                    if tries >= 100 {
-                        panic!("Too many attempts to close index {name} prior to deletion.")
+        let indexes = indexes
+            .into_iter()
+            .map(|(name, uuid)| {
+                let closing_event = loop {
+                    let mut lock = self.index_map.write().unwrap();
+                    match lock.start_deletion(&uuid) {
+                        Ok(env_closing) => break env_closing,
+                        Err(Some(reopen)) => {
+                            // drop the lock here so that we don't synchronously wait for the index to close.
+                            drop(lock);
+                            tries += 1;
+                            if tries >= 100 {
+                                panic!("Too many attempts to close index {name} prior to deletion.")
+                            }
+                            let reopen =
+                                if let Some(reopen) = reopen.wait_timeout(Duration::from_secs(6)) {
+                                    reopen
+                                } else {
+                                    continue;
+                                };
+                            reopen.close(&mut self.index_map.write().unwrap());
+                            continue;
+                        }
+                        // TODO: what is this case, what does that mean?
+                        Err(None) => return None,
                    }
-                    let reopen = if let Some(reopen) = reopen.wait_timeout(Duration::from_secs(6)) {
-                        reopen
-                    } else {
-                        continue;
-                    };
-                    reopen.close(&mut self.index_map.write().unwrap());
-                    continue;
-                }
-                Err(None) => return Ok(()),
-            }
-        };
+                };
+                Some((name, uuid, closing_event))
+            })
+            .filter_map(|thingy| thingy)
+            .map(|(name, uuid, closing)| {
+                (name.to_string(), uuid, self.base_path.join(uuid.to_string()), closing)
+            })
+            .collect::<Vec<_>>();

        let index_map = self.index_map.clone();
-        let index_path = self.base_path.join(uuid.to_string());
-        let index_name = name.to_string();
        thread::Builder::new()
            .name(String::from("index_deleter"))
            .spawn(move || {
-                // We first wait to be sure that the previously opened index is effectively closed.
-                // This can take a lot of time, this is why we do that in a separate thread.
-                if let Some(closing_event) = closing_event {
-                    closing_event.wait();
-                }
+                for (name, uuid, index_path, closing_event) in indexes {
+                    // We first wait to be sure that the previously opened index is effectively closed.
+                    // This can take a lot of time, this is why we do that in a separate thread.
+                    if let Some(closing_event) = closing_event {
+                        closing_event.wait();
+                    }

-                // Then we remove the content from disk.
-                if let Err(e) = fs::remove_dir_all(&index_path) {
-                    error!(
-                        "An error happened when deleting the index {} ({}): {}",
-                        index_name, uuid, e
-                    );
-                }
+                    // Then we remove the content from disk.
+                    if let Err(e) = fs::remove_dir_all(&index_path) {
+                        error!(
+                            "An error happened when deleting the index {} ({}): {}",
+                            name, uuid, e
+                        );
+                    }

-                // Finally we remove the entry from the index map.
-                index_map.write().unwrap().end_deletion(&uuid);
+                    // Finally we remove the entry from the index map.
+                    index_map.write().unwrap().end_deletion(&uuid);
+                }
            })
            .unwrap();

@@ -292,11 +303,7 @@ impl IndexMapper {
            .ok_or_else(|| Error::IndexNotFound(name.to_string()))?;

        // We remove the index from the in-memory index map.
-        self.index_map.write().unwrap().close_for_resize(
-            &uuid,
-            self.enable_mdb_writemap,
-            self.index_growth_amount,
-        );
+        self.index_map.write().unwrap().close_for_resize(&uuid, self.index_growth_amount);

        Ok(())
    }
@@ -361,7 +368,6 @@ impl IndexMapper {
                                &uuid,
                                &index_path,
                                None,
-                                self.enable_mdb_writemap,
                                self.index_base_map_size,
                            )?;
                        }
--- a/index-scheduler/src/lib.rs
+++ b/index-scheduler/src/lib.rs
@@ -31,7 +31,7 @@ mod uuid_codec;
 pub type Result<T> = std::result::Result<T, Error>;
 pub type TaskId = u32;

-use std::collections::{BTreeMap, HashMap};
+use std::collections::HashMap;
 use std::ops::{Bound, RangeBounds};
 use std::path::{Path, PathBuf};
 use std::sync::atomic::AtomicBool;
@@ -233,8 +233,6 @@ pub struct IndexSchedulerOptions {
    pub task_db_size: usize,
    /// The size, in bytes, with which a meilisearch index is opened the first time of each meilisearch index.
    pub index_base_map_size: usize,
-    /// Whether we open a meilisearch index with the MDB_WRITEMAP option or not.
-    pub enable_mdb_writemap: bool,
    /// The size, in bytes, by which the map size of an index is increased when it resized due to being full.
    pub index_growth_amount: usize,
    /// The number of indexes that can be concurrently opened in memory.
@@ -376,11 +374,6 @@ impl IndexScheduler {
        std::fs::create_dir_all(&options.indexes_path)?;
        std::fs::create_dir_all(&options.dumps_path)?;

-        if cfg!(windows) && options.enable_mdb_writemap {
-            // programmer error if this happens: in normal use passing the option on Windows is an error in main
-            panic!("Windows doesn't support the MDB_WRITEMAP LMDB option");
-        }
-
        let task_db_size = clamp_to_page_size(options.task_db_size);
        let budget = if options.indexer_config.skip_index_budget {
            IndexBudget {
@@ -403,37 +396,25 @@ impl IndexScheduler {
            .open(options.tasks_path)?;
        let file_store = FileStore::new(&options.update_file_path)?;

-        let mut wtxn = env.write_txn()?;
-        let all_tasks = env.create_database(&mut wtxn, Some(db_name::ALL_TASKS))?;
-        let status = env.create_database(&mut wtxn, Some(db_name::STATUS))?;
-        let kind = env.create_database(&mut wtxn, Some(db_name::KIND))?;
-        let index_tasks = env.create_database(&mut wtxn, Some(db_name::INDEX_TASKS))?;
-        let canceled_by = env.create_database(&mut wtxn, Some(db_name::CANCELED_BY))?;
-        let enqueued_at = env.create_database(&mut wtxn, Some(db_name::ENQUEUED_AT))?;
-        let started_at = env.create_database(&mut wtxn, Some(db_name::STARTED_AT))?;
-        let finished_at = env.create_database(&mut wtxn, Some(db_name::FINISHED_AT))?;
-        wtxn.commit()?;
-
        // allow unreachable_code to get rids of the warning in the case of a test build.
        let this = Self {
            must_stop_processing: MustStopProcessing::default(),
            processing_tasks: Arc::new(RwLock::new(ProcessingTasks::new())),
            file_store,
-            all_tasks,
-            status,
-            kind,
-            index_tasks,
-            canceled_by,
-            enqueued_at,
-            started_at,
-            finished_at,
+            all_tasks: env.create_database(Some(db_name::ALL_TASKS))?,
+            status: env.create_database(Some(db_name::STATUS))?,
+            kind: env.create_database(Some(db_name::KIND))?,
+            index_tasks: env.create_database(Some(db_name::INDEX_TASKS))?,
+            canceled_by: env.create_database(Some(db_name::CANCELED_BY))?,
+            enqueued_at: env.create_database(Some(db_name::ENQUEUED_AT))?,
+            started_at: env.create_database(Some(db_name::STARTED_AT))?,
+            finished_at: env.create_database(Some(db_name::FINISHED_AT))?,
            index_mapper: IndexMapper::new(
                &env,
                options.indexes_path,
                budget.map_size,
                options.index_growth_amount,
                budget.index_count,
-                options.enable_mdb_writemap,
                options.indexer_config,
            )?,
            env,
@@ -573,16 +554,10 @@ impl IndexScheduler {
        &self.index_mapper.indexer_config
    }

-    /// Return the real database size (i.e.: The size **with** the free pages)
    pub fn size(&self) -> Result<u64> {
        Ok(self.env.real_disk_size()?)
    }

-    /// Return the used database size (i.e.: The size **without** the free pages)
-    pub fn used_size(&self) -> Result<u64> {
-        Ok(self.env.non_free_pages_size()?)
-    }
-
    /// Return the index corresponding to the name.
    ///
    /// * If the index wasn't opened before, the index will be opened.
@@ -762,38 +737,6 @@ impl IndexScheduler {
        Ok(tasks)
    }

-    /// The returned structure contains:
-    /// 1. The name of the property being observed can be `statuses`, `types`, or `indexes`.
-    /// 2. The name of the specific data related to the property can be `enqueued` for the `statuses`, `settingsUpdate` for the `types`, or the name of the index for the `indexes`, for example.
-    /// 3. The number of times the properties appeared.
-    pub fn get_stats(&self) -> Result<BTreeMap<String, BTreeMap<String, u64>>> {
-        let rtxn = self.read_txn()?;
-
-        let mut res = BTreeMap::new();
-
-        res.insert(
-            "statuses".to_string(),
-            enum_iterator::all::<Status>()
-                .map(|s| Ok((s.to_string(), self.get_status(&rtxn, s)?.len())))
-                .collect::<Result<BTreeMap<String, u64>>>()?,
-        );
-        res.insert(
-            "types".to_string(),
-            enum_iterator::all::<Kind>()
-                .map(|s| Ok((s.to_string(), self.get_kind(&rtxn, s)?.len())))
-                .collect::<Result<BTreeMap<String, u64>>>()?,
-        );
-        res.insert(
-            "indexes".to_string(),
-            self.index_tasks
-                .iter(&rtxn)?
-                .map(|res| Ok(res.map(|(name, bitmap)| (name.to_string(), bitmap.len()))?))
-                .collect::<Result<BTreeMap<String, u64>>>()?,
-        );
-
-        Ok(res)
-    }
-
    /// Return true iff there is at least one task associated with this index
    /// that is processing.
    pub fn is_index_processing(&self, index: &str) -> Result<bool> {
@@ -1341,6 +1284,7 @@ impl<'a> Dump<'a> {
                KindDump::IndexDeletion => KindWithContent::IndexDeletion {
                    index_uid: task.index_uid.ok_or(Error::CorruptedDump)?,
                },
+                KindDump::IndexClear { index_uids } => KindWithContent::IndexClear { index_uids },
                KindDump::IndexCreation { primary_key } => KindWithContent::IndexCreation {
                    index_uid: task.index_uid.ok_or(Error::CorruptedDump)?,
                    primary_key,
@@ -1528,7 +1472,6 @@ mod tests {
                dumps_path: tempdir.path().join("dumps"),
                task_db_size: 1000 * 1000, // 1 MB, we don't use MiB on purpose.
                index_base_map_size: 1000 * 1000, // 1 MB, we don't use MiB on purpose.
-                enable_mdb_writemap: false,
                index_growth_amount: 1000 * 1000, // 1 MB
                index_count: 5,
                indexer_config,
--- a/index-scheduler/src/utils.rs
+++ b/index-scheduler/src/utils.rs
@@ -258,6 +258,7 @@ pub fn swap_index_uid_in_task(task: &mut Task, swap: (&str, &str)) {
        K::TaskCancelation { .. }
        | K::TaskDeletion { .. }
        | K::DumpCreation { .. }
+        | K::IndexClear { .. }
        | K::SnapshotCreation => (),
    };
    if let Some(Details::IndexSwap { swaps }) = &mut task.details {
@@ -466,7 +467,7 @@ impl IndexScheduler {
                        }
                    }
                    Details::DocumentDeletionByFilter { deleted_documents, original_filter: _ } => {
-                        assert_eq!(kind.as_kind(), Kind::DocumentDeletion);
+                        assert_eq!(kind.as_kind(), Kind::DocumentDeletionByFilter);
                        let (index_uid, _) = if let KindWithContent::DocumentDeletionByFilter {
                            ref index_uid,
                            ref filter_expr,
--- a/meilisearch-auth/src/lib.rs
+++ b/meilisearch-auth/src/lib.rs
@@ -45,11 +45,6 @@ impl AuthController {
        self.store.size()
    }

-    /// Return the used size of the `AuthController` database in bytes.
-    pub fn used_size(&self) -> Result<u64> {
-        self.store.used_size()
-    }
-
    pub fn create_key(&self, create_key: CreateApiKey) -> Result<Key> {
        match self.store.get_api_key(create_key.uid)? {
            Some(_) => Err(AuthControllerError::ApiKeyAlreadyExists(create_key.uid.to_string())),
--- a/meilisearch-auth/src/store.rs
+++ b/meilisearch-auth/src/store.rs
@@ -55,11 +55,9 @@ impl HeedAuthStore {
        let path = path.as_ref().join(AUTH_DB_PATH);
        create_dir_all(&path)?;
        let env = Arc::new(open_auth_store_env(path.as_ref())?);
-        let mut wtxn = env.write_txn()?;
-        let keys = env.create_database(&mut wtxn, Some(KEY_DB_NAME))?;
+        let keys = env.create_database(Some(KEY_DB_NAME))?;
        let action_keyid_index_expiration =
-            env.create_database(&mut wtxn, Some(KEY_ID_ACTION_INDEX_EXPIRATION_DB_NAME))?;
-        wtxn.commit()?;
+            env.create_database(Some(KEY_ID_ACTION_INDEX_EXPIRATION_DB_NAME))?;
        Ok(Self { env, keys, action_keyid_index_expiration, should_close_on_drop: true })
    }

@@ -75,11 +73,6 @@ impl HeedAuthStore {
        Ok(self.env.real_disk_size()?)
    }

-    /// Return the number of bytes actually used in the database
-    pub fn used_size(&self) -> Result<u64> {
-        Ok(self.env.non_free_pages_size()?)
-    }
-
    pub fn set_drop_on_close(&mut self, v: bool) {
        self.should_close_on_drop = v;
    }
--- a/meilisearch-types/Cargo.toml
+++ b/meilisearch-types/Cargo.toml
@@ -22,7 +22,7 @@ file-store = { path = "../file-store" }
 flate2 = "1.0.25"
 fst = "0.4.7"
 memmap2 = "0.5.10"
-milli = { path = "../milli" }
+milli = { path = "../milli", default-features = false }
 roaring = { version = "0.10.1", features = ["serde"] }
 serde = { version = "1.0.160", features = ["derive"] }
 serde-cs = "0.2.4"
@@ -40,7 +40,7 @@ meili-snap = { path = "../meili-snap" }

 [features]
 # all specialized tokenizations
-all-tokenizations = ["milli/all-tokenizations"]
+default = ["milli/default"]

 # chinese specialized tokenization
 chinese = ["milli/chinese"]
--- a/meilisearch-types/src/deserr/mod.rs
+++ b/meilisearch-types/src/deserr/mod.rs
@@ -150,7 +150,6 @@ make_missing_field_convenience_builder!(MissingApiKeyActions, missing_api_key_ac
 make_missing_field_convenience_builder!(MissingApiKeyExpiresAt, missing_api_key_expires_at);
 make_missing_field_convenience_builder!(MissingApiKeyIndexes, missing_api_key_indexes);
 make_missing_field_convenience_builder!(MissingSwapIndexes, missing_swap_indexes);
-make_missing_field_convenience_builder!(MissingDocumentFilter, missing_document_filter);

 // Integrate a sub-error into a [`DeserrError`] by taking its error message but using
 // the default error code (C) from `Self`
--- a/meilisearch-types/src/error.rs
+++ b/meilisearch-types/src/error.rs
@@ -214,14 +214,12 @@ InvalidApiKeyUid                      , InvalidRequest       , BAD_REQUEST ;
 InvalidContentType                    , InvalidRequest       , UNSUPPORTED_MEDIA_TYPE ;
 InvalidDocumentCsvDelimiter           , InvalidRequest       , BAD_REQUEST ;
 InvalidDocumentFields                 , InvalidRequest       , BAD_REQUEST ;
-MissingDocumentFilter                 , InvalidRequest       , BAD_REQUEST ;
 InvalidDocumentFilter                 , InvalidRequest       , BAD_REQUEST ;
 InvalidDocumentGeoField               , InvalidRequest       , BAD_REQUEST ;
-InvalidVectorDimensions               , InvalidRequest       , BAD_REQUEST ;
-InvalidVectorsType                    , InvalidRequest       , BAD_REQUEST ;
 InvalidDocumentId                     , InvalidRequest       , BAD_REQUEST ;
 InvalidDocumentLimit                  , InvalidRequest       , BAD_REQUEST ;
 InvalidDocumentOffset                 , InvalidRequest       , BAD_REQUEST ;
+InvalidDocumentDeleteFilter           , InvalidRequest       , BAD_REQUEST ;
 InvalidIndexLimit                     , InvalidRequest       , BAD_REQUEST ;
 InvalidIndexOffset                    , InvalidRequest       , BAD_REQUEST ;
 InvalidIndexPrimaryKey                , InvalidRequest       , BAD_REQUEST ;
@@ -334,8 +332,6 @@ impl ErrorCode for milli::Error {
                    UserError::InvalidSortableAttribute { .. } => Code::InvalidSearchSort,
                    UserError::CriterionError(_) => Code::InvalidSettingsRankingRules,
                    UserError::InvalidGeoField { .. } => Code::InvalidDocumentGeoField,
-                    UserError::InvalidVectorDimensions { .. } => Code::InvalidVectorDimensions,
-                    UserError::InvalidVectorsType { .. } => Code::InvalidVectorsType,
                    UserError::SortError(_) => Code::InvalidSearchSort,
                    UserError::InvalidMinTypoWordLenSetting(_, _) => {
                        Code::InvalidSettingsTypoTolerance
--- a/meilisearch-types/src/tasks.rs
+++ b/meilisearch-types/src/tasks.rs
@@ -46,6 +46,7 @@ impl Task {
            | SnapshotCreation
            | TaskCancelation { .. }
            | TaskDeletion { .. }
+            | IndexClear { .. }
            | IndexSwap { .. } => None,
            DocumentAdditionOrUpdate { index_uid, .. }
            | DocumentDeletion { index_uid, .. }
@@ -72,6 +73,7 @@ impl Task {
            | KindWithContent::DocumentClear { .. }
            | KindWithContent::SettingsUpdate { .. }
            | KindWithContent::IndexDeletion { .. }
+            | KindWithContent::IndexClear { .. }
            | KindWithContent::IndexCreation { .. }
            | KindWithContent::IndexUpdate { .. }
            | KindWithContent::IndexSwap { .. }
@@ -111,6 +113,9 @@ pub enum KindWithContent {
        is_deletion: bool,
        allow_index_creation: bool,
    },
+    IndexClear {
+        index_uids: Vec<String>,
+    },
    IndexDeletion {
        index_uid: String,
    },
@@ -156,6 +161,7 @@ impl KindWithContent {
            KindWithContent::SettingsUpdate { .. } => Kind::SettingsUpdate,
            KindWithContent::IndexCreation { .. } => Kind::IndexCreation,
            KindWithContent::IndexDeletion { .. } => Kind::IndexDeletion,
+            KindWithContent::IndexClear { .. } => Kind::IndexDeletion,
            KindWithContent::IndexUpdate { .. } => Kind::IndexUpdate,
            KindWithContent::IndexSwap { .. } => Kind::IndexSwap,
            KindWithContent::TaskCancelation { .. } => Kind::TaskCancelation,
@@ -181,6 +187,7 @@ impl KindWithContent {
            | IndexCreation { index_uid, .. }
            | IndexUpdate { index_uid, .. }
            | IndexDeletion { index_uid } => vec![index_uid],
+            IndexClear { index_uids } => index_uids.into_iter().map(|s| s.as_ref()).collect(),
            IndexSwap { swaps } => {
                let mut indexes = HashSet::<&str>::default();
                for swap in swaps {
@@ -214,7 +221,9 @@ impl KindWithContent {
                    deleted_documents: None,
                })
            }
-            KindWithContent::DocumentClear { .. } | KindWithContent::IndexDeletion { .. } => {
+            KindWithContent::DocumentClear { .. }
+            | KindWithContent::IndexDeletion { .. }
+            | KindWithContent::IndexClear { .. } => {
                Some(Details::ClearAll { deleted_documents: None })
            }
            KindWithContent::SettingsUpdate { new_settings, .. } => {
@@ -268,7 +277,7 @@ impl KindWithContent {
            KindWithContent::SettingsUpdate { new_settings, .. } => {
                Some(Details::SettingsUpdate { settings: new_settings.clone() })
            }
-            KindWithContent::IndexDeletion { .. } => None,
+            KindWithContent::IndexDeletion { .. } | KindWithContent::IndexClear { .. } => None,
            KindWithContent::IndexCreation { primary_key, .. }
            | KindWithContent::IndexUpdate { primary_key, .. } => {
                Some(Details::IndexInfo { primary_key: primary_key.clone() })
@@ -307,7 +316,7 @@ impl From<&KindWithContent> for Option<Details> {
            KindWithContent::SettingsUpdate { new_settings, .. } => {
                Some(Details::SettingsUpdate { settings: new_settings.clone() })
            }
-            KindWithContent::IndexDeletion { .. } => None,
+            KindWithContent::IndexDeletion { .. } | KindWithContent::IndexClear { .. } => None,
            KindWithContent::IndexCreation { primary_key, .. } => {
                Some(Details::IndexInfo { primary_key: primary_key.clone() })
            }
@@ -395,6 +404,7 @@ impl std::error::Error for ParseTaskStatusError {}
 pub enum Kind {
    DocumentAdditionOrUpdate,
    DocumentDeletion,
+    DocumentDeletionByFilter,
    SettingsUpdate,
    IndexCreation,
    IndexDeletion,
@@ -411,6 +421,7 @@ impl Kind {
        match self {
            Kind::DocumentAdditionOrUpdate
            | Kind::DocumentDeletion
+            | Kind::DocumentDeletionByFilter
            | Kind::SettingsUpdate
            | Kind::IndexCreation
            | Kind::IndexDeletion
@@ -428,6 +439,7 @@ impl Display for Kind {
        match self {
            Kind::DocumentAdditionOrUpdate => write!(f, "documentAdditionOrUpdate"),
            Kind::DocumentDeletion => write!(f, "documentDeletion"),
+            Kind::DocumentDeletionByFilter => write!(f, "documentDeletionByFilter"),
            Kind::SettingsUpdate => write!(f, "settingsUpdate"),
            Kind::IndexCreation => write!(f, "indexCreation"),
            Kind::IndexDeletion => write!(f, "indexDeletion"),
--- a/meilisearch/Cargo.toml
+++ b/meilisearch/Cargo.toml
@@ -48,7 +48,6 @@ mime = "0.3.17"
 num_cpus = "1.15.0"
 obkv = "0.2.0"
 once_cell = "1.17.1"
-ordered-float = "3.7.0"
 parking_lot = "0.12.1"
 permissive-json-pointer = { path = "../permissive-json-pointer" }
 pin-project-lite = "0.2.9"
@@ -107,7 +106,7 @@ vergen = { version = "7.5.1", default-features = false, features = ["git"] }
 zip = { version = "0.6.4", optional = true }

 [features]
-default = ["analytics", "meilisearch-types/all-tokenizations", "mini-dashboard"]
+default = ["analytics", "meilisearch-types/default", "mini-dashboard"]
 analytics = ["segment"]
 mini-dashboard = ["actix-web-static-files", "static-files", "anyhow", "cargo_toml", "hex", "reqwest", "sha-1", "tempfile", "zip"]
 chinese = ["meilisearch-types/chinese"]
--- a/meilisearch/src/analytics/mock_analytics.rs
+++ b/meilisearch/src/analytics/mock_analytics.rs
@@ -5,7 +5,7 @@ use actix_web::HttpRequest;
 use meilisearch_types::InstanceUid;
 use serde_json::Value;

-use super::{find_user_id, Analytics, DocumentDeletionKind, DocumentFetchKind};
+use super::{find_user_id, Analytics, DocumentDeletionKind};
 use crate::routes::indexes::documents::UpdateDocumentsQuery;
 use crate::routes::tasks::TasksFilterQuery;
 use crate::Opt;
@@ -71,8 +71,6 @@ impl Analytics for MockAnalytics {
        _request: &HttpRequest,
    ) {
    }
-    fn get_fetch_documents(&self, _documents_query: &DocumentFetchKind, _request: &HttpRequest) {}
-    fn post_fetch_documents(&self, _documents_query: &DocumentFetchKind, _request: &HttpRequest) {}
    fn get_tasks(&self, _query: &TasksFilterQuery, _request: &HttpRequest) {}
    fn health_seen(&self, _request: &HttpRequest) {}
 }
--- a/meilisearch/src/analytics/mod.rs
+++ b/meilisearch/src/analytics/mod.rs
@@ -67,12 +67,6 @@ pub enum DocumentDeletionKind {
    PerFilter,
 }

-#[derive(Copy, Clone, Debug, PartialEq, Eq)]
-pub enum DocumentFetchKind {
-    PerDocumentId,
-    Normal { with_filter: bool, limit: usize, offset: usize },
-}
-
 pub trait Analytics: Sync + Send {
    fn instance_uid(&self) -> Option<&InstanceUid>;

@@ -96,12 +90,6 @@ pub trait Analytics: Sync + Send {
        request: &HttpRequest,
    );

-    // this method should be called to aggregate a fetch documents request
-    fn get_fetch_documents(&self, documents_query: &DocumentFetchKind, request: &HttpRequest);
-
-    // this method should be called to aggregate a fetch documents request
-    fn post_fetch_documents(&self, documents_query: &DocumentFetchKind, request: &HttpRequest);
-
    // this method should be called to aggregate a add documents request
    fn delete_documents(&self, kind: DocumentDeletionKind, request: &HttpRequest);

--- a/meilisearch/src/analytics/segment_analytics.rs
+++ b/meilisearch/src/analytics/segment_analytics.rs
@@ -23,9 +23,7 @@ use tokio::select;
 use tokio::sync::mpsc::{self, Receiver, Sender};
 use uuid::Uuid;

-use super::{
-    config_user_id_path, DocumentDeletionKind, DocumentFetchKind, MEILISEARCH_CONFIG_PATH,
-};
+use super::{config_user_id_path, DocumentDeletionKind, MEILISEARCH_CONFIG_PATH};
 use crate::analytics::Analytics;
 use crate::option::{default_http_addr, IndexerOpts, MaxMemory, MaxThreads, ScheduleSnapshot};
 use crate::routes::indexes::documents::UpdateDocumentsQuery;
@@ -74,8 +72,6 @@ pub enum AnalyticsMsg {
    AggregateAddDocuments(DocumentsAggregator),
    AggregateDeleteDocuments(DocumentsDeletionAggregator),
    AggregateUpdateDocuments(DocumentsAggregator),
-    AggregateGetFetchDocuments(DocumentsFetchAggregator),
-    AggregatePostFetchDocuments(DocumentsFetchAggregator),
    AggregateTasks(TasksAggregator),
    AggregateHealth(HealthAggregator),
 }
@@ -143,8 +139,6 @@ impl SegmentAnalytics {
            add_documents_aggregator: DocumentsAggregator::default(),
            delete_documents_aggregator: DocumentsDeletionAggregator::default(),
            update_documents_aggregator: DocumentsAggregator::default(),
-            get_fetch_documents_aggregator: DocumentsFetchAggregator::default(),
-            post_fetch_documents_aggregator: DocumentsFetchAggregator::default(),
            get_tasks_aggregator: TasksAggregator::default(),
            health_aggregator: HealthAggregator::default(),
        });
@@ -211,16 +205,6 @@ impl super::Analytics for SegmentAnalytics {
        let _ = self.sender.try_send(AnalyticsMsg::AggregateUpdateDocuments(aggregate));
    }

-    fn get_fetch_documents(&self, documents_query: &DocumentFetchKind, request: &HttpRequest) {
-        let aggregate = DocumentsFetchAggregator::from_query(documents_query, request);
-        let _ = self.sender.try_send(AnalyticsMsg::AggregateGetFetchDocuments(aggregate));
-    }
-
-    fn post_fetch_documents(&self, documents_query: &DocumentFetchKind, request: &HttpRequest) {
-        let aggregate = DocumentsFetchAggregator::from_query(documents_query, request);
-        let _ = self.sender.try_send(AnalyticsMsg::AggregatePostFetchDocuments(aggregate));
-    }
-
    fn get_tasks(&self, query: &TasksFilterQuery, request: &HttpRequest) {
        let aggregate = TasksAggregator::from_query(query, request);
        let _ = self.sender.try_send(AnalyticsMsg::AggregateTasks(aggregate));
@@ -241,7 +225,6 @@ impl super::Analytics for SegmentAnalytics {
 struct Infos {
    env: String,
    experimental_enable_metrics: bool,
-    experimental_reduce_indexing_memory_usage: bool,
    db_path: bool,
    import_dump: bool,
    dump_dir: bool,
@@ -275,7 +258,6 @@ impl From<Opt> for Infos {
        let Opt {
            db_path,
            experimental_enable_metrics,
-            experimental_reduce_indexing_memory_usage,
            http_addr,
            master_key: _,
            env,
@@ -318,7 +300,6 @@ impl From<Opt> for Infos {
        Self {
            env,
            experimental_enable_metrics,
-            experimental_reduce_indexing_memory_usage,
            db_path: db_path != PathBuf::from("./data.ms"),
            import_dump: import_dump.is_some(),
            dump_dir: dump_dir != PathBuf::from("dumps/"),
@@ -357,8 +338,6 @@ pub struct Segment {
    add_documents_aggregator: DocumentsAggregator,
    delete_documents_aggregator: DocumentsDeletionAggregator,
    update_documents_aggregator: DocumentsAggregator,
-    get_fetch_documents_aggregator: DocumentsFetchAggregator,
-    post_fetch_documents_aggregator: DocumentsFetchAggregator,
    get_tasks_aggregator: TasksAggregator,
    health_aggregator: HealthAggregator,
 }
@@ -421,8 +400,6 @@ impl Segment {
                        Some(AnalyticsMsg::AggregateAddDocuments(agreg)) => self.add_documents_aggregator.aggregate(agreg),
                        Some(AnalyticsMsg::AggregateDeleteDocuments(agreg)) => self.delete_documents_aggregator.aggregate(agreg),
                        Some(AnalyticsMsg::AggregateUpdateDocuments(agreg)) => self.update_documents_aggregator.aggregate(agreg),
-                        Some(AnalyticsMsg::AggregateGetFetchDocuments(agreg)) => self.get_fetch_documents_aggregator.aggregate(agreg),
-                        Some(AnalyticsMsg::AggregatePostFetchDocuments(agreg)) => self.post_fetch_documents_aggregator.aggregate(agreg),
                        Some(AnalyticsMsg::AggregateTasks(agreg)) => self.get_tasks_aggregator.aggregate(agreg),
                        Some(AnalyticsMsg::AggregateHealth(agreg)) => self.health_aggregator.aggregate(agreg),
                        None => (),
@@ -473,10 +450,6 @@ impl Segment {
            .into_event(&self.user, "Documents Deleted");
        let update_documents = std::mem::take(&mut self.update_documents_aggregator)
            .into_event(&self.user, "Documents Updated");
-        let get_fetch_documents = std::mem::take(&mut self.get_fetch_documents_aggregator)
-            .into_event(&self.user, "Documents Fetched GET");
-        let post_fetch_documents = std::mem::take(&mut self.post_fetch_documents_aggregator)
-            .into_event(&self.user, "Documents Fetched POST");
        let get_tasks =
            std::mem::take(&mut self.get_tasks_aggregator).into_event(&self.user, "Tasks Seen");
        let health =
@@ -500,12 +473,6 @@ impl Segment {
        if let Some(update_documents) = update_documents {
            let _ = self.batcher.push(update_documents).await;
        }
-        if let Some(get_fetch_documents) = get_fetch_documents {
-            let _ = self.batcher.push(get_fetch_documents).await;
-        }
-        if let Some(post_fetch_documents) = post_fetch_documents {
-            let _ = self.batcher.push(post_fetch_documents).await;
-        }
        if let Some(get_tasks) = get_tasks {
            let _ = self.batcher.push(get_tasks).await;
        }
@@ -1168,76 +1135,3 @@ impl HealthAggregator {
        })
    }
 }
-
-#[derive(Default, Serialize)]
-pub struct DocumentsFetchAggregator {
-    #[serde(skip)]
-    timestamp: Option<OffsetDateTime>,
-
-    // context
-    #[serde(rename = "user-agent")]
-    user_agents: HashSet<String>,
-
-    #[serde(rename = "requests.max_limit")]
-    total_received: usize,
-
-    // a call on ../documents/:doc_id
-    per_document_id: bool,
-    // if a filter was used
-    per_filter: bool,
-
-    // pagination
-    #[serde(rename = "pagination.max_limit")]
-    max_limit: usize,
-    #[serde(rename = "pagination.max_offset")]
-    max_offset: usize,
-}
-
-impl DocumentsFetchAggregator {
-    pub fn from_query(query: &DocumentFetchKind, request: &HttpRequest) -> Self {
-        let (limit, offset) = match query {
-            DocumentFetchKind::PerDocumentId => (1, 0),
-            DocumentFetchKind::Normal { limit, offset, .. } => (*limit, *offset),
-        };
-        Self {
-            timestamp: Some(OffsetDateTime::now_utc()),
-            user_agents: extract_user_agents(request).into_iter().collect(),
-            total_received: 1,
-            per_document_id: matches!(query, DocumentFetchKind::PerDocumentId),
-            per_filter: matches!(query, DocumentFetchKind::Normal { with_filter, .. } if *with_filter),
-            max_limit: limit,
-            max_offset: offset,
-        }
-    }
-
-    /// Aggregate one [DocumentsFetchAggregator] into another.
-    pub fn aggregate(&mut self, other: Self) {
-        if self.timestamp.is_none() {
-            self.timestamp = other.timestamp;
-        }
-        for user_agent in other.user_agents {
-            self.user_agents.insert(user_agent);
-        }
-
-        self.total_received = self.total_received.saturating_add(other.total_received);
-        self.per_document_id |= other.per_document_id;
-        self.per_filter |= other.per_filter;
-
-        self.max_limit = self.max_limit.max(other.max_limit);
-        self.max_offset = self.max_offset.max(other.max_offset);
-    }
-
-    pub fn into_event(self, user: &User, event_name: &str) -> Option<Track> {
-        // if we had no timestamp it means we never encountered any events and
-        // thus we don't need to send this event.
-        let timestamp = self.timestamp?;
-
-        Some(Track {
-            timestamp: Some(timestamp),
-            user: user.clone(),
-            event: event_name.to_string(),
-            properties: serde_json::to_value(self).ok()?,
-            ..Default::default()
-        })
-    }
-}
--- a/meilisearch/src/error.rs
+++ b/meilisearch/src/error.rs
@@ -1,6 +1,5 @@
 use actix_web as aweb;
 use aweb::error::{JsonPayloadError, QueryPayloadError};
-use byte_unit::Byte;
 use meilisearch_types::document_formats::{DocumentFormatError, PayloadType};
 use meilisearch_types::error::{Code, ErrorCode, ResponseError};
 use meilisearch_types::index_uid::{IndexUid, IndexUidFormatError};
@@ -27,8 +26,8 @@ pub enum MeilisearchHttpError {
    InvalidExpression(&'static [&'static str], Value),
    #[error("A {0} payload is missing.")]
    MissingPayload(PayloadType),
-    #[error("The provided payload reached the size limit. The maximum accepted payload size is {}.",  Byte::from_bytes(*.0 as u64).get_appropriate_unit(true))]
-    PayloadTooLarge(usize),
+    #[error("The provided payload reached the size limit.")]
+    PayloadTooLarge,
    #[error("Two indexes must be given for each swap. The list `[{}]` contains {} indexes.",
        .0.iter().map(|uid| format!("\"{uid}\"")).collect::<Vec<_>>().join(", "), .0.len()
    )]
@@ -61,9 +60,9 @@ impl ErrorCode for MeilisearchHttpError {
            MeilisearchHttpError::MissingPayload(_) => Code::MissingPayload,
            MeilisearchHttpError::InvalidContentType(_, _) => Code::InvalidContentType,
            MeilisearchHttpError::DocumentNotFound(_) => Code::DocumentNotFound,
-            MeilisearchHttpError::EmptyFilter => Code::InvalidDocumentFilter,
+            MeilisearchHttpError::EmptyFilter => Code::InvalidDocumentDeleteFilter,
            MeilisearchHttpError::InvalidExpression(_, _) => Code::InvalidSearchFilter,
-            MeilisearchHttpError::PayloadTooLarge(_) => Code::PayloadTooLarge,
+            MeilisearchHttpError::PayloadTooLarge => Code::PayloadTooLarge,
            MeilisearchHttpError::SwapIndexPayloadWrongLength(_) => Code::InvalidSwapIndexes,
            MeilisearchHttpError::IndexUid(e) => e.error_code(),
            MeilisearchHttpError::SerdeJson(_) => Code::Internal,
--- a/meilisearch/src/extractors/payload.rs
+++ b/meilisearch/src/extractors/payload.rs
@@ -11,7 +11,6 @@ use crate::error::MeilisearchHttpError;
 pub struct Payload {
    payload: Decompress<dev::Payload>,
    limit: usize,
-    remaining: usize,
 }

 pub struct PayloadConfig {
@@ -44,7 +43,6 @@ impl FromRequest for Payload {
        ready(Ok(Payload {
            payload: Decompress::from_headers(payload.take(), req.headers()),
            limit,
-            remaining: limit,
        }))
    }
 }
@@ -56,14 +54,12 @@ impl Stream for Payload {
    fn poll_next(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Option<Self::Item>> {
        match Pin::new(&mut self.payload).poll_next(cx) {
            Poll::Ready(Some(result)) => match result {
-                Ok(bytes) => match self.remaining.checked_sub(bytes.len()) {
+                Ok(bytes) => match self.limit.checked_sub(bytes.len()) {
                    Some(new_limit) => {
-                        self.remaining = new_limit;
+                        self.limit = new_limit;
                        Poll::Ready(Some(Ok(bytes)))
                    }
-                    None => {
-                        Poll::Ready(Some(Err(MeilisearchHttpError::PayloadTooLarge(self.limit))))
-                    }
+                    None => Poll::Ready(Some(Err(MeilisearchHttpError::PayloadTooLarge))),
                },
                x => Poll::Ready(Some(x.map_err(MeilisearchHttpError::from))),
            },
--- a/meilisearch/src/lib.rs
+++ b/meilisearch/src/lib.rs
@@ -232,7 +232,6 @@ fn open_or_create_database_unchecked(
            dumps_path: opt.dump_dir.clone(),
            task_db_size: opt.max_task_db_size.get_bytes() as usize,
            index_base_map_size: opt.max_index_size.get_bytes() as usize,
-            enable_mdb_writemap: opt.experimental_reduce_indexing_memory_usage,
            indexer_config: (&opt.indexer_options).try_into()?,
            autobatching_enabled: true,
            max_number_of_tasks: 1_000_000,
--- a/meilisearch/src/main.rs
+++ b/meilisearch/src/main.rs
@@ -29,11 +29,6 @@ fn setup(opt: &Opt) -> anyhow::Result<()> {
 async fn main() -> anyhow::Result<()> {
    let (opt, config_read_from) = Opt::try_build()?;

-    anyhow::ensure!(
-        !(cfg!(windows) && opt.experimental_reduce_indexing_memory_usage),
-        "The `experimental-reduce-indexing-memory-usage` flag is not supported on Windows"
-    );
-
    setup(&opt)?;

    match (opt.env.as_ref(), &opt.master_key) {
--- a/meilisearch/src/metrics.rs
+++ b/meilisearch/src/metrics.rs
@@ -4,32 +4,20 @@ use prometheus::{
    register_int_gauge_vec, HistogramVec, IntCounterVec, IntGauge, IntGaugeVec,
 };

-/// Create evenly distributed buckets
-fn create_buckets() -> [f64; 29] {
-    (0..10)
-        .chain((10..100).step_by(10))
-        .chain((100..=1000).step_by(100))
-        .map(|i| i as f64 / 1000.)
-        .collect::<Vec<_>>()
-        .try_into()
-        .unwrap()
-}
+const HTTP_RESPONSE_TIME_CUSTOM_BUCKETS: &[f64; 14] = &[
+    0.0005, 0.0008, 0.00085, 0.0009, 0.00095, 0.001, 0.00105, 0.0011, 0.00115, 0.0012, 0.0015,
+    0.002, 0.003, 1.0,
+];

 lazy_static! {
-    pub static ref HTTP_RESPONSE_TIME_CUSTOM_BUCKETS: [f64; 29] = create_buckets();
-    pub static ref MEILISEARCH_HTTP_REQUESTS_TOTAL: IntCounterVec = register_int_counter_vec!(
-        opts!("meilisearch_http_requests_total", "Meilisearch HTTP requests total"),
+    pub static ref HTTP_REQUESTS_TOTAL: IntCounterVec = register_int_counter_vec!(
+        opts!("http_requests_total", "HTTP requests total"),
        &["method", "path"]
    )
    .expect("Can't create a metric");
    pub static ref MEILISEARCH_DB_SIZE_BYTES: IntGauge =
-        register_int_gauge!(opts!("meilisearch_db_size_bytes", "Meilisearch DB Size In Bytes"))
+        register_int_gauge!(opts!("meilisearch_db_size_bytes", "Meilisearch Db Size In Bytes"))
            .expect("Can't create a metric");
-    pub static ref MEILISEARCH_USED_DB_SIZE_BYTES: IntGauge = register_int_gauge!(opts!(
-        "meilisearch_used_db_size_bytes",
-        "Meilisearch Used DB Size In Bytes"
-    ))
-    .expect("Can't create a metric");
    pub static ref MEILISEARCH_INDEX_COUNT: IntGauge =
        register_int_gauge!(opts!("meilisearch_index_count", "Meilisearch Index Count"))
            .expect("Can't create a metric");
@@ -38,16 +26,11 @@ lazy_static! {
        &["index"]
    )
    .expect("Can't create a metric");
-    pub static ref MEILISEARCH_HTTP_RESPONSE_TIME_SECONDS: HistogramVec = register_histogram_vec!(
+    pub static ref HTTP_RESPONSE_TIME_SECONDS: HistogramVec = register_histogram_vec!(
        "http_response_time_seconds",
        "HTTP response times",
        &["method", "path"],
        HTTP_RESPONSE_TIME_CUSTOM_BUCKETS.to_vec()
    )
    .expect("Can't create a metric");
-    pub static ref MEILISEARCH_NB_TASKS: IntGaugeVec = register_int_gauge_vec!(
-        opts!("meilisearch_nb_tasks", "Meilisearch Number of tasks"),
-        &["kind", "value"]
-    )
-    .expect("Can't create a metric");
 }
--- a/meilisearch/src/middleware.rs
+++ b/meilisearch/src/middleware.rs
@@ -52,11 +52,11 @@ where
        if is_registered_resource {
            let request_method = req.method().to_string();
            histogram_timer = Some(
-                crate::metrics::MEILISEARCH_HTTP_RESPONSE_TIME_SECONDS
+                crate::metrics::HTTP_RESPONSE_TIME_SECONDS
                    .with_label_values(&[&request_method, request_path])
                    .start_timer(),
            );
-            crate::metrics::MEILISEARCH_HTTP_REQUESTS_TOTAL
+            crate::metrics::HTTP_REQUESTS_TOTAL
                .with_label_values(&[&request_method, request_path])
                .inc();
        }
--- a/meilisearch/src/option.rs
+++ b/meilisearch/src/option.rs
@@ -48,8 +48,6 @@ const MEILI_IGNORE_DUMP_IF_DB_EXISTS: &str = "MEILI_IGNORE_DUMP_IF_DB_EXISTS";
 const MEILI_DUMP_DIR: &str = "MEILI_DUMP_DIR";
 const MEILI_LOG_LEVEL: &str = "MEILI_LOG_LEVEL";
 const MEILI_EXPERIMENTAL_ENABLE_METRICS: &str = "MEILI_EXPERIMENTAL_ENABLE_METRICS";
-const MEILI_EXPERIMENTAL_REDUCE_INDEXING_MEMORY_USAGE: &str =
-    "MEILI_EXPERIMENTAL_REDUCE_INDEXING_MEMORY_USAGE";

 const DEFAULT_CONFIG_FILE_PATH: &str = "./config.toml";
 const DEFAULT_DB_PATH: &str = "./data.ms";
@@ -295,11 +293,6 @@ pub struct Opt {
    #[serde(default)]
    pub experimental_enable_metrics: bool,

-    /// Experimental RAM reduction during indexing, do not use in production, see: <https://github.com/meilisearch/product/discussions/652>
-    #[clap(long, env = MEILI_EXPERIMENTAL_REDUCE_INDEXING_MEMORY_USAGE)]
-    #[serde(default)]
-    pub experimental_reduce_indexing_memory_usage: bool,
-
    #[serde(flatten)]
    #[clap(flatten)]
    pub indexer_options: IndexerOpts,
@@ -392,7 +385,6 @@ impl Opt {
            #[cfg(all(not(debug_assertions), feature = "analytics"))]
            no_analytics,
            experimental_enable_metrics: enable_metrics_route,
-            experimental_reduce_indexing_memory_usage: reduce_indexing_memory_usage,
        } = self;
        export_to_env_if_not_present(MEILI_DB_PATH, db_path);
        export_to_env_if_not_present(MEILI_HTTP_ADDR, http_addr);
@@ -434,10 +426,6 @@ impl Opt {
            MEILI_EXPERIMENTAL_ENABLE_METRICS,
            enable_metrics_route.to_string(),
        );
-        export_to_env_if_not_present(
-            MEILI_EXPERIMENTAL_REDUCE_INDEXING_MEMORY_USAGE,
-            reduce_indexing_memory_usage.to_string(),
-        );
        indexer_options.export_to_env();
    }

--- a/meilisearch/src/routes/indexes/documents.rs
+++ b/meilisearch/src/routes/indexes/documents.rs
@@ -29,7 +29,7 @@ use tempfile::tempfile;
 use tokio::fs::File;
 use tokio::io::{AsyncSeekExt, AsyncWriteExt, BufWriter};

-use crate::analytics::{Analytics, DocumentDeletionKind, DocumentFetchKind};
+use crate::analytics::{Analytics, DocumentDeletionKind};
 use crate::error::MeilisearchHttpError;
 use crate::error::PayloadError::ReceivePayload;
 use crate::extractors::authentication::policies::*;
@@ -97,14 +97,10 @@ pub async fn get_document(
    index_scheduler: GuardedData<ActionPolicy<{ actions::DOCUMENTS_GET }>, Data<IndexScheduler>>,
    document_param: web::Path<DocumentParam>,
    params: AwebQueryParameter<GetDocument, DeserrQueryParamError>,
-    req: HttpRequest,
-    analytics: web::Data<dyn Analytics>,
 ) -> Result<HttpResponse, ResponseError> {
    let DocumentParam { index_uid, document_id } = document_param.into_inner();
    let index_uid = IndexUid::try_from(index_uid)?;

-    analytics.get_fetch_documents(&DocumentFetchKind::PerDocumentId, &req);
-
    let GetDocument { fields } = params.into_inner();
    let attributes_to_retrieve = fields.merge_star_and_none();

@@ -165,31 +161,16 @@ pub async fn documents_by_query_post(
    index_scheduler: GuardedData<ActionPolicy<{ actions::DOCUMENTS_GET }>, Data<IndexScheduler>>,
    index_uid: web::Path<String>,
    body: AwebJson<BrowseQuery, DeserrJsonError>,
-    req: HttpRequest,
-    analytics: web::Data<dyn Analytics>,
 ) -> Result<HttpResponse, ResponseError> {
    debug!("called with body: {:?}", body);

-    let body = body.into_inner();
-
-    analytics.post_fetch_documents(
-        &DocumentFetchKind::Normal {
-            with_filter: body.filter.is_some(),
-            limit: body.limit,
-            offset: body.offset,
-        },
-        &req,
-    );
-
-    documents_by_query(&index_scheduler, index_uid, body)
+    documents_by_query(&index_scheduler, index_uid, body.into_inner())
 }

 pub async fn get_documents(
    index_scheduler: GuardedData<ActionPolicy<{ actions::DOCUMENTS_GET }>, Data<IndexScheduler>>,
    index_uid: web::Path<String>,
    params: AwebQueryParameter<BrowseQueryGet, DeserrQueryParamError>,
-    req: HttpRequest,
-    analytics: web::Data<dyn Analytics>,
 ) -> Result<HttpResponse, ResponseError> {
    debug!("called with params: {:?}", params);

@@ -210,15 +191,6 @@ pub async fn get_documents(
        filter,
    };

-    analytics.get_fetch_documents(
-        &DocumentFetchKind::Normal {
-            with_filter: query.filter.is_some(),
-            limit: query.limit,
-            offset: query.offset,
-        },
-        &req,
-    );
-
    documents_by_query(&index_scheduler, index_uid, query)
 }

@@ -486,7 +458,7 @@ pub async fn delete_documents_batch(
 #[derive(Debug, Deserr)]
 #[deserr(error = DeserrJsonError, rename_all = camelCase, deny_unknown_fields)]
 pub struct DocumentDeletionByFilter {
-    #[deserr(error = DeserrJsonError<InvalidDocumentFilter>, missing_field_error = DeserrJsonError::missing_document_filter)]
+    #[deserr(error = DeserrJsonError<InvalidDocumentDeleteFilter>)]
    filter: Value,
 }

@@ -508,8 +480,8 @@ pub async fn delete_documents_by_filter(
    || -> Result<_, ResponseError> {
        Ok(crate::search::parse_filter(&filter)?.ok_or(MeilisearchHttpError::EmptyFilter)?)
    }()
-    // and whatever was the error, the error code should always be an InvalidDocumentFilter
-    .map_err(|err| ResponseError::from_msg(err.message, Code::InvalidDocumentFilter))?;
+    // and whatever was the error, the error code should always be an InvalidDocumentDeleteFilter
+    .map_err(|err| ResponseError::from_msg(err.message, Code::InvalidDocumentDeleteFilter))?;
    let task = KindWithContent::DocumentDeletionByFilter { index_uid, filter_expr: filter };

    let task: SummarizedTaskView =
@@ -568,12 +540,7 @@ fn retrieve_documents<S: AsRef<str>>(
    };

    let candidates = if let Some(filter) = filter {
-        filter.evaluate(&rtxn, index).map_err(|err| match err {
-            milli::Error::UserError(milli::UserError::InvalidFilter(_)) => {
-                ResponseError::from_msg(err.to_string(), Code::InvalidDocumentFilter)
-            }
-            e => e.into(),
-        })?
+        filter.evaluate(&rtxn, index)?
    } else {
        index.documents_ids(&rtxn)?
    };
--- a/meilisearch/src/routes/indexes/mod.rs
+++ b/meilisearch/src/routes/indexes/mod.rs
@@ -31,7 +31,8 @@ pub fn configure(cfg: &mut web::ServiceConfig) {
    cfg.service(
        web::resource("")
            .route(web::get().to(list_indexes))
-            .route(web::post().to(SeqHandler(create_index))),
+            .route(web::post().to(SeqHandler(create_index)))
+            .route(web::delete().to(SeqHandler(delete_all_indexes))),
    )
    .service(
        web::scope("/{index_uid}")
@@ -107,6 +108,22 @@ pub async fn list_indexes(
    Ok(HttpResponse::Ok().json(ret))
 }

+pub async fn delete_all_indexes(
+    index_scheduler: GuardedData<ActionPolicy<{ actions::INDEXES_DELETE }>, Data<IndexScheduler>>,
+    _req: HttpRequest,
+    _analytics: web::Data<dyn Analytics>,
+) -> Result<HttpResponse, ResponseError> {
+    let filters = index_scheduler.filters();
+    let indexes = index_scheduler.index_names()?;
+    let indexes = indexes.into_iter().filter(|uid| filters.is_index_authorized(uid)).collect();
+
+    let task = KindWithContent::IndexClear { index_uids: indexes };
+    let task: SummarizedTaskView =
+        tokio::task::spawn_blocking(move || index_scheduler.register(task)).await??.into();
+
+    Ok(HttpResponse::Accepted().json(task))
+}
+
 #[derive(Deserr, Debug)]
 #[deserr(error = DeserrJsonError, rename_all = camelCase, deny_unknown_fields)]
 pub struct IndexCreateRequest {
--- a/meilisearch/src/routes/indexes/search.rs
+++ b/meilisearch/src/routes/indexes/search.rs
@@ -34,8 +34,6 @@ pub fn configure(cfg: &mut web::ServiceConfig) {
 pub struct SearchQueryGet {
    #[deserr(default, error = DeserrQueryParamError<InvalidSearchQ>)]
    q: Option<String>,
-    #[deserr(default, error = DeserrQueryParamError<InvalidSearchQ>)]
-    vector: Option<Vec<f32>>,
    #[deserr(default = Param(DEFAULT_SEARCH_OFFSET()), error = DeserrQueryParamError<InvalidSearchOffset>)]
    offset: Param<usize>,
    #[deserr(default = Param(DEFAULT_SEARCH_LIMIT()), error = DeserrQueryParamError<InvalidSearchLimit>)]
@@ -82,7 +80,6 @@ impl From<SearchQueryGet> for SearchQuery {

        Self {
            q: other.q,
-            vector: other.vector,
            offset: other.offset.0,
            limit: other.limit.0,
            page: other.page.as_deref().copied(),
--- a/meilisearch/src/routes/metrics.rs
+++ b/meilisearch/src/routes/metrics.rs
@@ -17,7 +17,7 @@ pub fn configure(config: &mut web::ServiceConfig) {

 pub async fn get_metrics(
    index_scheduler: GuardedData<ActionPolicy<{ actions::METRICS_GET }>, Data<IndexScheduler>>,
-    auth_controller: Data<AuthController>,
+    auth_controller: GuardedData<ActionPolicy<{ actions::METRICS_GET }>, Data<AuthController>>,
 ) -> Result<HttpResponse, ResponseError> {
    let auth_filters = index_scheduler.filters();
    if !auth_filters.all_indexes_authorized() {
@@ -28,10 +28,10 @@ pub async fn get_metrics(
        return Err(error);
    }

-    let response = create_all_stats((*index_scheduler).clone(), auth_controller, auth_filters)?;
+    let response =
+        create_all_stats((*index_scheduler).clone(), (*auth_controller).clone(), auth_filters)?;

    crate::metrics::MEILISEARCH_DB_SIZE_BYTES.set(response.database_size as i64);
-    crate::metrics::MEILISEARCH_USED_DB_SIZE_BYTES.set(response.used_database_size as i64);
    crate::metrics::MEILISEARCH_INDEX_COUNT.set(response.indexes.len() as i64);

    for (index, value) in response.indexes.iter() {
@@ -40,14 +40,6 @@ pub async fn get_metrics(
            .set(value.number_of_documents as i64);
    }

-    for (kind, value) in index_scheduler.get_stats()? {
-        for (value, count) in value {
-            crate::metrics::MEILISEARCH_NB_TASKS
-                .with_label_values(&[&kind, &value])
-                .set(count as i64);
-        }
-    }
-
    let encoder = TextEncoder::new();
    let mut buffer = vec![];
    encoder.encode(&prometheus::gather(), &mut buffer).expect("Failed to encode metrics");
--- a/meilisearch/src/routes/mod.rs
+++ b/meilisearch/src/routes/mod.rs
@@ -231,8 +231,6 @@ pub async fn running() -> HttpResponse {
 #[serde(rename_all = "camelCase")]
 pub struct Stats {
    pub database_size: u64,
-    #[serde(skip)]
-    pub used_database_size: u64,
    #[serde(serialize_with = "time::serde::rfc3339::option::serialize")]
    pub last_update: Option<OffsetDateTime>,
    pub indexes: BTreeMap<String, indexes::IndexStats>,
@@ -261,7 +259,6 @@ pub fn create_all_stats(
    let mut last_task: Option<OffsetDateTime> = None;
    let mut indexes = BTreeMap::new();
    let mut database_size = 0;
-    let mut used_database_size = 0;

    for index_uid in index_scheduler.index_names()? {
        // Accumulate the size of all indexes, even unauthorized ones, so
@@ -269,7 +266,6 @@ pub fn create_all_stats(
        // See <https://github.com/meilisearch/meilisearch/pull/3541#discussion_r1126747643> for context.
        let stats = index_scheduler.index_stats(&index_uid)?;
        database_size += stats.inner_stats.database_size;
-        used_database_size += stats.inner_stats.used_database_size;

        if !filters.is_index_authorized(&index_uid) {
            continue;
@@ -282,14 +278,10 @@ pub fn create_all_stats(
    }

    database_size += index_scheduler.size()?;
-    used_database_size += index_scheduler.used_size()?;
    database_size += auth_controller.size()?;
-    used_database_size += auth_controller.used_size()?;
-    let update_file_size = index_scheduler.compute_update_file_size()?;
-    database_size += update_file_size;
-    used_database_size += update_file_size;
+    database_size += index_scheduler.compute_update_file_size()?;

-    let stats = Stats { database_size, used_database_size, last_update: last_task, indexes };
+    let stats = Stats { database_size, last_update: last_task, indexes };
    Ok(stats)
 }

--- a/meilisearch/src/routes/tasks.rs
+++ b/meilisearch/src/routes/tasks.rs
@@ -99,7 +99,7 @@ pub struct DetailsView {
    #[serde(skip_serializing_if = "Option::is_none")]
    pub deleted_tasks: Option<Option<u64>>,
    #[serde(skip_serializing_if = "Option::is_none")]
-    pub original_filter: Option<Option<String>>,
+    pub original_filter: Option<String>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub dump_uid: Option<Option<String>>,
    #[serde(skip_serializing_if = "Option::is_none")]
@@ -131,13 +131,12 @@ impl From<Details> for DetailsView {
            } => DetailsView {
                provided_ids: Some(received_document_ids),
                deleted_documents: Some(deleted_documents),
-                original_filter: Some(None),
                ..DetailsView::default()
            },
            Details::DocumentDeletionByFilter { original_filter, deleted_documents } => {
                DetailsView {
                    provided_ids: Some(0),
-                    original_filter: Some(Some(original_filter)),
+                    original_filter: Some(original_filter),
                    deleted_documents: Some(deleted_documents),
                    ..DetailsView::default()
                }
@@ -149,7 +148,7 @@ impl From<Details> for DetailsView {
                DetailsView {
                    matched_tasks: Some(matched_tasks),
                    canceled_tasks: Some(canceled_tasks),
-                    original_filter: Some(Some(original_filter)),
+                    original_filter: Some(original_filter),
                    ..DetailsView::default()
                }
            }
@@ -157,7 +156,7 @@ impl From<Details> for DetailsView {
                DetailsView {
                    matched_tasks: Some(matched_tasks),
                    deleted_tasks: Some(deleted_tasks),
-                    original_filter: Some(Some(original_filter)),
+                    original_filter: Some(original_filter),
                    ..DetailsView::default()
                }
            }
@@ -730,7 +729,7 @@ mod tests {
            let err = deserr_query_params::<TaskDeletionOrCancelationQuery>(params).unwrap_err();
            snapshot!(meili_snap::json_string!(err), @r###"
            {
-              "message": "Invalid value in parameter `types`: `createIndex` is not a valid task type. Available types are `documentAdditionOrUpdate`, `documentDeletion`, `settingsUpdate`, `indexCreation`, `indexDeletion`, `indexUpdate`, `indexSwap`, `taskCancelation`, `taskDeletion`, `dumpCreation`, `snapshotCreation`.",
+              "message": "Invalid value in parameter `types`: `createIndex` is not a valid task type. Available types are `documentAdditionOrUpdate`, `documentDeletion`, `documentDeletionByFilter`, `settingsUpdate`, `indexCreation`, `indexDeletion`, `indexUpdate`, `indexSwap`, `taskCancelation`, `taskDeletion`, `dumpCreation`, `snapshotCreation`.",
              "code": "invalid_task_types",
              "type": "invalid_request",
              "link": "https://docs.meilisearch.com/errors#invalid_task_types"
--- a/meilisearch/src/search.rs
+++ b/meilisearch/src/search.rs
@@ -9,15 +9,13 @@ use meilisearch_auth::IndexSearchRules;
 use meilisearch_types::deserr::DeserrJsonError;
 use meilisearch_types::error::deserr_codes::*;
 use meilisearch_types::index_uid::IndexUid;
-use meilisearch_types::milli::dot_product_similarity;
 use meilisearch_types::settings::DEFAULT_PAGINATION_MAX_TOTAL_HITS;
 use meilisearch_types::{milli, Document};
 use milli::tokenizer::TokenizerBuilder;
 use milli::{
    AscDesc, FieldId, FieldsIdsMap, Filter, FormatOptions, Index, MatchBounds, MatcherBuilder,
-    SortError, TermsMatchingStrategy, VectorOrArrayOfVectors, DEFAULT_VALUES_PER_FACET,
+    SortError, TermsMatchingStrategy, DEFAULT_VALUES_PER_FACET,
 };
-use ordered_float::OrderedFloat;
 use regex::Regex;
 use serde::Serialize;
 use serde_json::{json, Value};
@@ -33,13 +31,11 @@ pub const DEFAULT_CROP_MARKER: fn() -> String = || "…".to_string();
 pub const DEFAULT_HIGHLIGHT_PRE_TAG: fn() -> String = || "<em>".to_string();
 pub const DEFAULT_HIGHLIGHT_POST_TAG: fn() -> String = || "</em>".to_string();

-#[derive(Debug, Clone, Default, PartialEq, Deserr)]
+#[derive(Debug, Clone, Default, PartialEq, Eq, Deserr)]
 #[deserr(error = DeserrJsonError, rename_all = camelCase, deny_unknown_fields)]
 pub struct SearchQuery {
    #[deserr(default, error = DeserrJsonError<InvalidSearchQ>)]
    pub q: Option<String>,
-    #[deserr(default, error = DeserrJsonError<InvalidSearchQ>)]
-    pub vector: Option<Vec<f32>>,
    #[deserr(default = DEFAULT_SEARCH_OFFSET(), error = DeserrJsonError<InvalidSearchOffset>)]
    pub offset: usize,
    #[deserr(default = DEFAULT_SEARCH_LIMIT(), error = DeserrJsonError<InvalidSearchLimit>)]
@@ -84,15 +80,13 @@ impl SearchQuery {
 // This struct contains the fields of `SearchQuery` inline.
 // This is because neither deserr nor serde support `flatten` when using `deny_unknown_fields.
 // The `From<SearchQueryWithIndex>` implementation ensures both structs remain up to date.
-#[derive(Debug, Clone, PartialEq, Deserr)]
+#[derive(Debug, Clone, PartialEq, Eq, Deserr)]
 #[deserr(error = DeserrJsonError, rename_all = camelCase, deny_unknown_fields)]
 pub struct SearchQueryWithIndex {
    #[deserr(error = DeserrJsonError<InvalidIndexUid>, missing_field_error = DeserrJsonError::missing_index_uid)]
    pub index_uid: IndexUid,
    #[deserr(default, error = DeserrJsonError<InvalidSearchQ>)]
    pub q: Option<String>,
-    #[deserr(default, error = DeserrJsonError<InvalidSearchQ>)]
-    pub vector: Option<Vec<f32>>,
    #[deserr(default = DEFAULT_SEARCH_OFFSET(), error = DeserrJsonError<InvalidSearchOffset>)]
    pub offset: usize,
    #[deserr(default = DEFAULT_SEARCH_LIMIT(), error = DeserrJsonError<InvalidSearchLimit>)]
@@ -132,7 +126,6 @@ impl SearchQueryWithIndex {
        let SearchQueryWithIndex {
            index_uid,
            q,
-            vector,
            offset,
            limit,
            page,
@@ -154,7 +147,6 @@ impl SearchQueryWithIndex {
            index_uid,
            SearchQuery {
                q,
-                vector,
                offset,
                limit,
                page,
@@ -217,8 +209,6 @@ pub struct SearchHit {
 pub struct SearchResult {
    pub hits: Vec<SearchHit>,
    pub query: String,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub vector: Option<Vec<f32>>,
    pub processing_time_ms: u128,
    #[serde(flatten)]
    pub hits_info: HitsInfo,
@@ -280,10 +270,6 @@ pub fn perform_search(

    let mut search = index.search(&rtxn);

-    if let Some(ref vector) = query.vector {
-        search.vector(vector.clone());
-    }
-
    if let Some(ref query) = query.q {
        search.query(query);
    }
@@ -403,6 +389,7 @@ pub fn perform_search(
    formatter_builder.highlight_suffix(query.highlight_post_tag);

    let mut documents = Vec::new();
+
    let documents_iter = index.documents(&rtxn, documents_ids)?;

    for (_id, obkv) in documents_iter {
@@ -429,12 +416,6 @@ pub fn perform_search(
            insert_geo_distance(sort, &mut document);
        }

-        if let Some(vector) = query.vector.as_ref() {
-            if let Some(vectors) = extract_field("_vectors", &fields_ids_map, obkv)? {
-                insert_semantic_similarity(vector, vectors, &mut document);
-            }
-        }
-
        let hit = SearchHit { document, formatted, matches_position };
        documents.push(hit);
    }
@@ -484,8 +465,7 @@ pub fn perform_search(
    let result = SearchResult {
        hits: documents,
        hits_info,
-        query: query.q.unwrap_or_default(),
-        vector: query.vector,
+        query: query.q.clone().unwrap_or_default(),
        processing_time_ms: before_search.elapsed().as_millis(),
        facet_distribution,
        facet_stats,
@@ -509,20 +489,6 @@ fn insert_geo_distance(sorts: &[String], document: &mut Document) {
    }
 }

-fn insert_semantic_similarity(query: &[f32], vectors: Value, document: &mut Document) {
-    let vectors =
-        match serde_json::from_value(vectors).map(VectorOrArrayOfVectors::into_array_of_vectors) {
-            Ok(vectors) => vectors,
-            Err(_) => return,
-        };
-    let similarity = vectors
-        .into_iter()
-        .map(|v| OrderedFloat(dot_product_similarity(query, &v)))
-        .max()
-        .map(OrderedFloat::into_inner);
-    document.insert("_semanticSimilarity".to_string(), json!(similarity));
-}
-
 fn compute_formatted_options(
    attr_to_highlight: &HashSet<String>,
    attr_to_crop: &[String],
@@ -650,22 +616,6 @@ fn make_document(
    Ok(document)
 }

-/// Extract the JSON value under the field name specified
-/// but doesn't support nested objects.
-fn extract_field(
-    field_name: &str,
-    field_ids_map: &FieldsIdsMap,
-    obkv: obkv::KvReaderU16,
-) -> Result<Option<serde_json::Value>, MeilisearchHttpError> {
-    match field_ids_map.id(field_name) {
-        Some(fid) => match obkv.get(fid) {
-            Some(value) => Ok(serde_json::from_slice(value).map(Some)?),
-            None => Ok(None),
-        },
-        None => Ok(None),
-    }
-}
-
 fn format_fields<A: AsRef<[u8]>>(
    document: &Document,
    field_ids_map: &FieldsIdsMap,
--- a/meilisearch/tests/auth/authorization.rs
+++ b/meilisearch/tests/auth/authorization.rs
@@ -16,11 +16,8 @@ pub static AUTHORIZATIONS: Lazy<HashMap<(&'static str, &'static str), HashSet<&'
            ("GET",     "/indexes/products/search") =>                         hashset!{"search", "*"},
            ("POST",    "/indexes/products/documents") =>                      hashset!{"documents.add", "documents.*", "*"},
            ("GET",     "/indexes/products/documents") =>                      hashset!{"documents.get", "documents.*", "*"},
-            ("POST",    "/indexes/products/documents/fetch") =>                hashset!{"documents.get", "documents.*", "*"},
            ("GET",     "/indexes/products/documents/0") =>                    hashset!{"documents.get", "documents.*", "*"},
            ("DELETE",  "/indexes/products/documents/0") =>                    hashset!{"documents.delete", "documents.*", "*"},
-            ("POST",    "/indexes/products/documents/delete-batch") =>         hashset!{"documents.delete", "documents.*", "*"},
-            ("POST",    "/indexes/products/documents/delete") =>               hashset!{"documents.delete", "documents.*", "*"},
            ("GET",     "/tasks") =>                                           hashset!{"tasks.get", "tasks.*", "*"},
            ("DELETE",  "/tasks") =>                                           hashset!{"tasks.delete", "tasks.*", "*"},
            ("GET",     "/tasks?indexUid=products") =>                         hashset!{"tasks.get", "tasks.*", "*"},
--- a/meilisearch/tests/documents/add_documents.rs
+++ b/meilisearch/tests/documents/add_documents.rs
@@ -1781,7 +1781,7 @@ async fn error_add_documents_payload_size() {
    snapshot!(json_string!(response, { ".duration" => "[duration]", ".enqueuedAt" => "[date]", ".startedAt" => "[date]", ".finishedAt" => "[date]" }),
        @r###"
    {
-      "message": "The provided payload reached the size limit. The maximum accepted payload size is 10.00 MiB.",
+      "message": "The provided payload reached the size limit.",
      "code": "payload_too_large",
      "type": "invalid_request",
      "link": "https://docs.meilisearch.com/errors#payload_too_large"
--- a/meilisearch/tests/documents/errors.rs
+++ b/meilisearch/tests/documents/errors.rs
@@ -180,9 +180,9 @@ async fn get_all_documents_bad_filter() {
    snapshot!(json_string!(response), @r###"
    {
      "message": "Attribute `doggo` is not filterable. This index does not have configured filterable attributes.\n1:6 doggo=bernese",
-      "code": "invalid_document_filter",
+      "code": "invalid_search_filter",
      "type": "invalid_request",
-      "link": "https://docs.meilisearch.com/errors#invalid_document_filter"
+      "link": "https://docs.meilisearch.com/errors#invalid_search_filter"
    }
    "###);
 }
@@ -547,9 +547,9 @@ async fn delete_document_by_filter() {
    snapshot!(json_string!(response), @r###"
    {
      "message": "Invalid syntax for the filter parameter: `expected String, Array, found: true`.",
-      "code": "invalid_document_filter",
+      "code": "invalid_document_delete_filter",
      "type": "invalid_request",
-      "link": "https://docs.meilisearch.com/errors#invalid_document_filter"
+      "link": "https://docs.meilisearch.com/errors#invalid_document_delete_filter"
    }
    "###);

@@ -559,9 +559,9 @@ async fn delete_document_by_filter() {
    snapshot!(json_string!(response), @r###"
    {
      "message": "Was expecting an operation `=`, `!=`, `>=`, `>`, `<=`, `<`, `IN`, `NOT IN`, `TO`, `EXISTS`, `NOT EXISTS`, `IS NULL`, `IS NOT NULL`, `IS EMPTY`, `IS NOT EMPTY`, `_geoRadius`, or `_geoBoundingBox` at `hello`.\n1:6 hello",
-      "code": "invalid_document_filter",
+      "code": "invalid_document_delete_filter",
      "type": "invalid_request",
-      "link": "https://docs.meilisearch.com/errors#invalid_document_filter"
+      "link": "https://docs.meilisearch.com/errors#invalid_document_delete_filter"
    }
    "###);

@@ -571,21 +571,9 @@ async fn delete_document_by_filter() {
    snapshot!(json_string!(response), @r###"
    {
      "message": "Sending an empty filter is forbidden.",
-      "code": "invalid_document_filter",
+      "code": "invalid_document_delete_filter",
      "type": "invalid_request",
-      "link": "https://docs.meilisearch.com/errors#invalid_document_filter"
-    }
-    "###);
-
-    // do not send any filter
-    let (response, code) = index.delete_document_by_filter(json!({})).await;
-    snapshot!(code, @"400 Bad Request");
-    snapshot!(json_string!(response), @r###"
-    {
-      "message": "Missing field `filter`",
-      "code": "missing_document_filter",
-      "type": "invalid_request",
-      "link": "https://docs.meilisearch.com/errors#missing_document_filter"
+      "link": "https://docs.meilisearch.com/errors#invalid_document_delete_filter"
    }
    "###);

@@ -642,9 +630,9 @@ async fn delete_document_by_filter() {
      },
      "error": {
        "message": "Attribute `doggo` is not filterable. This index does not have configured filterable attributes.\n1:6 doggo = bernese",
-        "code": "invalid_document_filter",
+        "code": "invalid_search_filter",
        "type": "invalid_request",
-        "link": "https://docs.meilisearch.com/errors#invalid_document_filter"
+        "link": "https://docs.meilisearch.com/errors#invalid_search_filter"
      },
      "duration": "[duration]",
      "enqueuedAt": "[date]",
@@ -676,9 +664,9 @@ async fn delete_document_by_filter() {
      },
      "error": {
        "message": "Attribute `catto` is not filterable. Available filterable attributes are: `doggo`.\n1:6 catto = jorts",
-        "code": "invalid_document_filter",
+        "code": "invalid_search_filter",
        "type": "invalid_request",
-        "link": "https://docs.meilisearch.com/errors#invalid_document_filter"
+        "link": "https://docs.meilisearch.com/errors#invalid_search_filter"
      },
      "duration": "[duration]",
      "enqueuedAt": "[date]",
@@ -760,27 +748,4 @@ async fn fetch_document_by_filter() {
      "link": "https://docs.meilisearch.com/errors#invalid_document_filter"
    }
    "###);
-
-    let (response, code) = index.get_document_by_filter(json!({ "filter": "cool doggo" })).await;
-    snapshot!(code, @"400 Bad Request");
-    snapshot!(json_string!(response), @r###"
-    {
-      "message": "Was expecting an operation `=`, `!=`, `>=`, `>`, `<=`, `<`, `IN`, `NOT IN`, `TO`, `EXISTS`, `NOT EXISTS`, `IS NULL`, `IS NOT NULL`, `IS EMPTY`, `IS NOT EMPTY`, `_geoRadius`, or `_geoBoundingBox` at `cool doggo`.\n1:11 cool doggo",
-      "code": "invalid_document_filter",
-      "type": "invalid_request",
-      "link": "https://docs.meilisearch.com/errors#invalid_document_filter"
-    }
-    "###);
-
-    let (response, code) =
-        index.get_document_by_filter(json!({ "filter": "doggo = bernese" })).await;
-    snapshot!(code, @"400 Bad Request");
-    snapshot!(json_string!(response), @r###"
-    {
-      "message": "Attribute `doggo` is not filterable. Available filterable attributes are: `color`.\n1:6 doggo = bernese",
-      "code": "invalid_document_filter",
-      "type": "invalid_request",
-      "link": "https://docs.meilisearch.com/errors#invalid_document_filter"
-    }
-    "###);
 }
--- a/meilisearch/tests/search/errors.rs
+++ b/meilisearch/tests/search/errors.rs
@@ -946,7 +946,7 @@ async fn sort_unset_ranking_rule() {
    index.wait_task(1).await;

    let expected_response = json!({
-        "message": "You must specify where `sort` is listed in the rankingRules setting to use the sort parameter at search time.",
+        "message": "The sort ranking rule must be specified in the ranking rules settings to use the sort parameter at search time.",
        "code": "invalid_search_sort",
        "type": "invalid_request",
        "link": "https://docs.meilisearch.com/errors#invalid_search_sort"
--- a/meilisearch/tests/tasks/errors.rs
+++ b/meilisearch/tests/tasks/errors.rs
@@ -97,7 +97,7 @@ async fn task_bad_types() {
    snapshot!(code, @"400 Bad Request");
    snapshot!(json_string!(response), @r###"
    {
-      "message": "Invalid value in parameter `types`: `doggo` is not a valid task type. Available types are `documentAdditionOrUpdate`, `documentDeletion`, `settingsUpdate`, `indexCreation`, `indexDeletion`, `indexUpdate`, `indexSwap`, `taskCancelation`, `taskDeletion`, `dumpCreation`, `snapshotCreation`.",
+      "message": "Invalid value in parameter `types`: `doggo` is not a valid task type. Available types are `documentAdditionOrUpdate`, `documentDeletion`, `documentDeletionByFilter`, `settingsUpdate`, `indexCreation`, `indexDeletion`, `indexUpdate`, `indexSwap`, `taskCancelation`, `taskDeletion`, `dumpCreation`, `snapshotCreation`.",
      "code": "invalid_task_types",
      "type": "invalid_request",
      "link": "https://docs.meilisearch.com/errors#invalid_task_types"
@@ -108,7 +108,7 @@ async fn task_bad_types() {
    snapshot!(code, @"400 Bad Request");
    snapshot!(json_string!(response), @r###"
    {
-      "message": "Invalid value in parameter `types`: `doggo` is not a valid task type. Available types are `documentAdditionOrUpdate`, `documentDeletion`, `settingsUpdate`, `indexCreation`, `indexDeletion`, `indexUpdate`, `indexSwap`, `taskCancelation`, `taskDeletion`, `dumpCreation`, `snapshotCreation`.",
+      "message": "Invalid value in parameter `types`: `doggo` is not a valid task type. Available types are `documentAdditionOrUpdate`, `documentDeletion`, `documentDeletionByFilter`, `settingsUpdate`, `indexCreation`, `indexDeletion`, `indexUpdate`, `indexSwap`, `taskCancelation`, `taskDeletion`, `dumpCreation`, `snapshotCreation`.",
      "code": "invalid_task_types",
      "type": "invalid_request",
      "link": "https://docs.meilisearch.com/errors#invalid_task_types"
@@ -119,7 +119,7 @@ async fn task_bad_types() {
    snapshot!(code, @"400 Bad Request");
    snapshot!(json_string!(response), @r###"
    {
-      "message": "Invalid value in parameter `types`: `doggo` is not a valid task type. Available types are `documentAdditionOrUpdate`, `documentDeletion`, `settingsUpdate`, `indexCreation`, `indexDeletion`, `indexUpdate`, `indexSwap`, `taskCancelation`, `taskDeletion`, `dumpCreation`, `snapshotCreation`.",
+      "message": "Invalid value in parameter `types`: `doggo` is not a valid task type. Available types are `documentAdditionOrUpdate`, `documentDeletion`, `documentDeletionByFilter`, `settingsUpdate`, `indexCreation`, `indexDeletion`, `indexUpdate`, `indexSwap`, `taskCancelation`, `taskDeletion`, `dumpCreation`, `snapshotCreation`.",
      "code": "invalid_task_types",
      "type": "invalid_request",
      "link": "https://docs.meilisearch.com/errors#invalid_task_types"
--- a/meilisearch/tests/tasks/mod.rs
+++ b/meilisearch/tests/tasks/mod.rs
@@ -413,7 +413,7 @@ async fn test_summarized_document_addition_or_update() {
 }

 #[actix_web::test]
-async fn test_summarized_delete_documents_by_batch() {
+async fn test_summarized_delete_batch() {
    let server = Server::new().await;
    let index = server.index("test");
    index.delete_batch(vec![1, 2, 3]).await;
@@ -430,8 +430,7 @@ async fn test_summarized_delete_documents_by_batch() {
      "canceledBy": null,
      "details": {
        "providedIds": 3,
-        "deletedDocuments": 0,
-        "originalFilter": null
+        "deletedDocuments": 0
      },
      "error": {
        "message": "Index `test` not found.",
@@ -461,8 +460,7 @@ async fn test_summarized_delete_documents_by_batch() {
      "canceledBy": null,
      "details": {
        "providedIds": 1,
-        "deletedDocuments": 0,
-        "originalFilter": null
+        "deletedDocuments": 0
      },
      "error": null,
      "duration": "[duration]",
@@ -474,100 +472,7 @@ async fn test_summarized_delete_documents_by_batch() {
 }

 #[actix_web::test]
-async fn test_summarized_delete_documents_by_filter() {
-    let server = Server::new().await;
-    let index = server.index("test");
-
-    index.delete_document_by_filter(json!({ "filter": "doggo = bernese" })).await;
-    index.wait_task(0).await;
-    let (task, _) = index.get_task(0).await;
-    assert_json_snapshot!(task,
-        { ".duration" => "[duration]", ".enqueuedAt" => "[date]", ".startedAt" => "[date]", ".finishedAt" => "[date]" },
-        @r###"
-    {
-      "uid": 0,
-      "indexUid": "test",
-      "status": "failed",
-      "type": "documentDeletion",
-      "canceledBy": null,
-      "details": {
-        "providedIds": 0,
-        "deletedDocuments": 0,
-        "originalFilter": "\"doggo = bernese\""
-      },
-      "error": {
-        "message": "Index `test` not found.",
-        "code": "index_not_found",
-        "type": "invalid_request",
-        "link": "https://docs.meilisearch.com/errors#index_not_found"
-      },
-      "duration": "[duration]",
-      "enqueuedAt": "[date]",
-      "startedAt": "[date]",
-      "finishedAt": "[date]"
-    }
-    "###);
-
-    index.create(None).await;
-    index.delete_document_by_filter(json!({ "filter": "doggo = bernese" })).await;
-    index.wait_task(2).await;
-    let (task, _) = index.get_task(2).await;
-    assert_json_snapshot!(task,
-        { ".duration" => "[duration]", ".enqueuedAt" => "[date]", ".startedAt" => "[date]", ".finishedAt" => "[date]" },
-        @r###"
-    {
-      "uid": 2,
-      "indexUid": "test",
-      "status": "failed",
-      "type": "documentDeletion",
-      "canceledBy": null,
-      "details": {
-        "providedIds": 0,
-        "deletedDocuments": 0,
-        "originalFilter": "\"doggo = bernese\""
-      },
-      "error": {
-        "message": "Attribute `doggo` is not filterable. This index does not have configured filterable attributes.\n1:6 doggo = bernese",
-        "code": "invalid_document_filter",
-        "type": "invalid_request",
-        "link": "https://docs.meilisearch.com/errors#invalid_document_filter"
-      },
-      "duration": "[duration]",
-      "enqueuedAt": "[date]",
-      "startedAt": "[date]",
-      "finishedAt": "[date]"
-    }
-    "###);
-
-    index.update_settings(json!({ "filterableAttributes": ["doggo"] })).await;
-    index.delete_document_by_filter(json!({ "filter": "doggo = bernese" })).await;
-    index.wait_task(4).await;
-    let (task, _) = index.get_task(4).await;
-    assert_json_snapshot!(task,
-        { ".duration" => "[duration]", ".enqueuedAt" => "[date]", ".startedAt" => "[date]", ".finishedAt" => "[date]" },
-        @r###"
-    {
-      "uid": 4,
-      "indexUid": "test",
-      "status": "succeeded",
-      "type": "documentDeletion",
-      "canceledBy": null,
-      "details": {
-        "providedIds": 0,
-        "deletedDocuments": 0,
-        "originalFilter": "\"doggo = bernese\""
-      },
-      "error": null,
-      "duration": "[duration]",
-      "enqueuedAt": "[date]",
-      "startedAt": "[date]",
-      "finishedAt": "[date]"
-    }
-    "###);
-}
-
-#[actix_web::test]
-async fn test_summarized_delete_document_by_id() {
+async fn test_summarized_delete_document() {
    let server = Server::new().await;
    let index = server.index("test");
    index.delete_document(1).await;
@@ -584,8 +489,7 @@ async fn test_summarized_delete_document_by_id() {
      "canceledBy": null,
      "details": {
        "providedIds": 1,
-        "deletedDocuments": 0,
-        "originalFilter": null
+        "deletedDocuments": 0
      },
      "error": {
        "message": "Index `test` not found.",
@@ -615,8 +519,7 @@ async fn test_summarized_delete_document_by_id() {
      "canceledBy": null,
      "details": {
        "providedIds": 1,
-        "deletedDocuments": 0,
-        "originalFilter": null
+        "deletedDocuments": 0
      },
      "error": null,
      "duration": "[duration]",
--- a/milli/Cargo.toml
+++ b/milli/Cargo.toml
@@ -15,7 +15,6 @@ license.workspace = true
 bimap = { version = "0.6.3", features = ["serde"] }
 bincode = "1.3.3"
 bstr = "1.4.0"
-bytemuck = { version = "1.13.1", features = ["extern_crate_alloc"] }
 byteorder = "1.4.3"
 charabia = { version = "0.7.2", default-features = false }
 concat-arrays = "0.1.2"
@@ -26,39 +25,26 @@ flatten-serde-json = { path = "../flatten-serde-json" }
 fst = "0.4.7"
 fxhash = "0.2.1"
 geoutils = "0.5.1"
-grenad = { version = "0.4.4", default-features = false, features = [
-    "tempfile",
-] }
-heed = { git = "https://github.com/meilisearch/heed", tag = "v0.12.6", default-features = false, features = [
-    "lmdb",
-    "sync-read-txn",
-] }
-hnsw = { version = "0.11.0", features = ["serde1"] }
+grenad = { version = "0.4.4", default-features = false, features = ["tempfile"] }
+heed = { git = "https://github.com/meilisearch/heed", tag = "v0.12.5", default-features = false, features = ["lmdb", "sync-read-txn"] }
 json-depth-checker = { path = "../json-depth-checker" }
 levenshtein_automata = { version = "0.2.1", features = ["fst_automaton"] }
 memmap2 = "0.5.10"
 obkv = "0.2.0"
 once_cell = "1.17.1"
 ordered-float = "3.6.0"
-rand_pcg = { version = "0.3.1", features = ["serde1"] }
 rayon = "1.7.0"
 roaring = "0.10.1"
 rstar = { version = "0.10.0", features = ["serde"] }
 serde = { version = "1.0.160", features = ["derive"] }
 serde_json = { version = "1.0.95", features = ["preserve_order"] }
 slice-group-by = "0.3.0"
-space = "0.17.0"
-smallstr = { version = "0.3.0", features = ["serde"] }
+smallstr =  { version = "0.3.0", features = ["serde"] }
 smallvec = "1.10.0"
 smartstring = "1.0.1"
 tempfile = "3.5.0"
 thiserror = "1.0.40"
-time = { version = "0.3.20", features = [
-    "serde-well-known",
-    "formatting",
-    "parsing",
-    "macros",
-] }
+time = { version = "0.3.20", features = ["serde-well-known", "formatting", "parsing", "macros"] }
 uuid = { version = "1.3.1", features = ["v4"] }

 filter-parser = { path = "../filter-parser" }
@@ -77,13 +63,13 @@ big_s = "1.0.2"
 insta = "1.29.0"
 maplit = "1.0.2"
 md5 = "0.7.0"
-rand = { version = "0.8.5", features = ["small_rng"] }
+rand = {version = "0.8.5", features = ["small_rng"] }

 [target.'cfg(fuzzing)'.dev-dependencies]
 fuzzcheck = "0.12.1"

 [features]
-all-tokenizations = ["charabia/default"]
+default = [ "charabia/default" ]

 # Use POSIX semaphores instead of SysV semaphores in LMDB
 # For more information on this feature, see heed's Cargo.toml
--- a/milli/examples/search.rs
+++ b/milli/examples/search.rs
@@ -52,7 +52,6 @@ fn main() -> Result<(), Box<dyn Error>> {
            let docs = execute_search(
                &mut ctx,
                &(!query.trim().is_empty()).then(|| query.trim().to_owned()),
-                &None,
                TermsMatchingStrategy::Last,
                false,
                &None,
--- a/milli/src/distance.rs
+++ b/milli/src/distance.rs
@@ -1,44 +0,0 @@
-use serde::{Deserialize, Serialize};
-use space::Metric;
-
-#[derive(Debug, Default, Clone, Copy, Serialize, Deserialize)]
-pub struct DotProduct;
-
-impl Metric<Vec<f32>> for DotProduct {
-    type Unit = u32;
-
-    // TODO explain me this function, I don't understand why f32.to_bits is ordered.
-    // I tried to do this and it wasn't OK <https://stackoverflow.com/a/43305015/1941280>
-    //
-    // Following <https://docs.rs/space/0.17.0/space/trait.Metric.html>.
-    fn distance(&self, a: &Vec<f32>, b: &Vec<f32>) -> Self::Unit {
-        let dist = 1.0 - dot_product_similarity(a, b);
-        debug_assert!(!dist.is_nan());
-        dist.to_bits()
-    }
-}
-
-/// Returns the dot product similarity score that will between 0.0 and 1.0
-/// if both vectors are normalized. The higher the more similar the vectors are.
-pub fn dot_product_similarity(a: &[f32], b: &[f32]) -> f32 {
-    a.iter().zip(b).map(|(a, b)| a * b).sum()
-}
-
-#[derive(Debug, Default, Clone, Copy, Serialize, Deserialize)]
-pub struct Euclidean;
-
-impl Metric<Vec<f32>> for Euclidean {
-    type Unit = u32;
-
-    fn distance(&self, a: &Vec<f32>, b: &Vec<f32>) -> Self::Unit {
-        let dist = euclidean_squared_distance(a, b).sqrt();
-        debug_assert!(!dist.is_nan());
-        dist.to_bits()
-    }
-}
-
-/// Return the squared euclidean distance between both vectors that will
-/// between 0.0 and +inf. The smaller the nearer the vectors are.
-pub fn euclidean_squared_distance(a: &[f32], b: &[f32]) -> f32 {
-    a.iter().zip(b).map(|(a, b)| (a - b).powi(2)).sum()
-}
--- a/milli/src/error.rs
+++ b/milli/src/error.rs
@@ -110,13 +110,9 @@ only composed of alphanumeric characters (a-z A-Z 0-9), hyphens (-) and undersco
    },
    #[error(transparent)]
    InvalidGeoField(#[from] GeoError),
-    #[error("Invalid vector dimensions: expected: `{}`, found: `{}`.", .expected, .found)]
-    InvalidVectorDimensions { expected: usize, found: usize },
-    #[error("The `_vectors` field in the document with the id: `{document_id}` is not an array. Was expecting an array of floats or an array of arrays of floats but instead got `{value}`.")]
-    InvalidVectorsType { document_id: Value, value: Value },
    #[error("{0}")]
    InvalidFilter(String),
-    #[error("Invalid type for filter subexpression: expected: {}, found: {1}.", .0.join(", "))]
+    #[error("Invalid type for filter subexpression: `expected {}, found: {1}`.", .0.join(", "))]
    InvalidFilterExpression(&'static [&'static str], Value),
    #[error("Attribute `{}` is not sortable. {}",
        .field,
@@ -130,7 +126,7 @@ only composed of alphanumeric characters (a-z A-Z 0-9), hyphens (-) and undersco
    InvalidSortableAttribute { field: String, valid_fields: BTreeSet<String> },
    #[error("{}", HeedError::BadOpenOptions)]
    InvalidLmdbOpenOptions,
-    #[error("You must specify where `sort` is listed in the rankingRules setting to use the sort parameter at search time.")]
+    #[error("The sort ranking rule must be specified in the ranking rules settings to use the sort parameter at search time.")]
    SortRankingRuleMissing,
    #[error("The database file is in an invalid state.")]
    InvalidStoreFile,
--- a/milli/src/heed_codec/roaring_bitmap/cbo_roaring_bitmap_codec.rs
+++ b/milli/src/heed_codec/roaring_bitmap/cbo_roaring_bitmap_codec.rs
@@ -49,7 +49,7 @@ impl CboRoaringBitmapCodec {
        } else {
            // Otherwise, it means we used the classic RoaringBitmapCodec and
            // that the header takes threshold integers.
-            RoaringBitmap::deserialize_unchecked_from(bytes)
+            RoaringBitmap::deserialize_from(bytes)
        }
    }

@@ -69,7 +69,7 @@ impl CboRoaringBitmapCodec {
                    vec.push(integer);
                }
            } else {
-                roaring |= RoaringBitmap::deserialize_unchecked_from(bytes.as_ref())?;
+                roaring |= RoaringBitmap::deserialize_from(bytes.as_ref())?;
            }
        }

--- a/milli/src/heed_codec/roaring_bitmap/roaring_bitmap_codec.rs
+++ b/milli/src/heed_codec/roaring_bitmap/roaring_bitmap_codec.rs
@@ -8,7 +8,7 @@ impl heed::BytesDecode<'_> for RoaringBitmapCodec {
    type DItem = RoaringBitmap;

    fn bytes_decode(bytes: &[u8]) -> Option<Self::DItem> {
-        RoaringBitmap::deserialize_unchecked_from(bytes).ok()
+        RoaringBitmap::deserialize_from(bytes).ok()
    }
 }

--- a/milli/src/index.rs
+++ b/milli/src/index.rs
@@ -8,12 +8,10 @@ use charabia::{Language, Script};
 use heed::flags::Flags;
 use heed::types::*;
 use heed::{CompactionOption, Database, PolyDatabase, RoTxn, RwTxn};
-use rand_pcg::Pcg32;
 use roaring::RoaringBitmap;
 use rstar::RTree;
 use time::OffsetDateTime;

-use crate::distance::DotProduct;
 use crate::error::{InternalError, UserError};
 use crate::facet::FacetType;
 use crate::fields_ids_map::FieldsIdsMap;
@@ -23,14 +21,12 @@ use crate::heed_codec::facet::{
 };
 use crate::heed_codec::{ScriptLanguageCodec, StrBEU16Codec, StrRefCodec};
 use crate::{
-    default_criteria, CboRoaringBitmapCodec, Criterion, DocumentId, ExternalDocumentsIds,
-    FacetDistribution, FieldDistribution, FieldId, FieldIdWordCountCodec, GeoPoint, ObkvCodec,
-    Result, RoaringBitmapCodec, RoaringBitmapLenCodec, Search, U8StrStrCodec, BEU16, BEU32,
+    default_criteria, BEU32StrCodec, BoRoaringBitmapCodec, CboRoaringBitmapCodec, Criterion,
+    DocumentId, ExternalDocumentsIds, FacetDistribution, FieldDistribution, FieldId,
+    FieldIdWordCountCodec, GeoPoint, ObkvCodec, Result, RoaringBitmapCodec, RoaringBitmapLenCodec,
+    Search, U8StrStrCodec, BEU16, BEU32,
 };

-/// The HNSW data-structure that we serialize, fill and search in.
-pub type Hnsw = hnsw::Hnsw<DotProduct, Vec<f32>, Pcg32, 12, 24>;
-
 pub const DEFAULT_MIN_WORD_LEN_ONE_TYPO: u8 = 5;
 pub const DEFAULT_MIN_WORD_LEN_TWO_TYPOS: u8 = 9;

@@ -47,7 +43,6 @@ pub mod main_key {
    pub const FIELDS_IDS_MAP_KEY: &str = "fields-ids-map";
    pub const GEO_FACETED_DOCUMENTS_IDS_KEY: &str = "geo-faceted-documents-ids";
    pub const GEO_RTREE_KEY: &str = "geo-rtree";
-    pub const VECTOR_HNSW_KEY: &str = "vector-hnsw";
    pub const HARD_EXTERNAL_DOCUMENTS_IDS_KEY: &str = "hard-external-documents-ids";
    pub const NUMBER_FACETED_DOCUMENTS_IDS_PREFIX: &str = "number-faceted-documents-ids";
    pub const PRIMARY_KEY_KEY: &str = "primary-key";
@@ -92,7 +87,6 @@ pub mod db_name {
    pub const FACET_ID_STRING_DOCIDS: &str = "facet-id-string-docids";
    pub const FIELD_ID_DOCID_FACET_F64S: &str = "field-id-docid-facet-f64s";
    pub const FIELD_ID_DOCID_FACET_STRINGS: &str = "field-id-docid-facet-strings";
-    pub const VECTOR_ID_DOCID: &str = "vector-id-docids";
    pub const DOCUMENTS: &str = "documents";
    pub const SCRIPT_LANGUAGE_DOCIDS: &str = "script_language_docids";
 }
@@ -117,6 +111,9 @@ pub struct Index {
    /// A prefix of word and all the documents ids containing this prefix, from attributes for which typos are not allowed.
    pub exact_word_prefix_docids: Database<Str, RoaringBitmapCodec>,

+    /// Maps a word and a document id (u32) to all the positions where the given word appears.
+    pub docid_word_positions: Database<BEU32StrCodec, BoRoaringBitmapCodec>,
+
    /// Maps the proximity between a pair of words with all the docids where this relation appears.
    pub word_pair_proximity_docids: Database<U8StrStrCodec, CboRoaringBitmapCodec>,
    /// Maps the proximity between a pair of word and prefix with all the docids where this relation appears.
@@ -156,9 +153,6 @@ pub struct Index {
    /// Maps the document id, the facet field id and the strings.
    pub field_id_docid_facet_strings: Database<FieldDocIdFacetStringCodec, Str>,

-    /// Maps a vector id to the document id that have it.
-    pub vector_id_docid: Database<OwnedType<BEU32>, OwnedType<BEU32>>,
-
    /// Maps the document id to the document as an obkv store.
    pub(crate) documents: Database<OwnedType<BEU32>, ObkvCodec>,
 }
@@ -172,49 +166,37 @@ impl Index {
    ) -> Result<Index> {
        use db_name::*;

-        options.max_dbs(24);
+        options.max_dbs(23);
        unsafe { options.flag(Flags::MdbAlwaysFreePages) };

        let env = options.open(path)?;
-        let mut wtxn = env.write_txn()?;
-        let main = env.create_poly_database(&mut wtxn, Some(MAIN))?;
-        let word_docids = env.create_database(&mut wtxn, Some(WORD_DOCIDS))?;
-        let exact_word_docids = env.create_database(&mut wtxn, Some(EXACT_WORD_DOCIDS))?;
-        let word_prefix_docids = env.create_database(&mut wtxn, Some(WORD_PREFIX_DOCIDS))?;
-        let exact_word_prefix_docids =
-            env.create_database(&mut wtxn, Some(EXACT_WORD_PREFIX_DOCIDS))?;
-        let word_pair_proximity_docids =
-            env.create_database(&mut wtxn, Some(WORD_PAIR_PROXIMITY_DOCIDS))?;
-        let script_language_docids =
-            env.create_database(&mut wtxn, Some(SCRIPT_LANGUAGE_DOCIDS))?;
+        let main = env.create_poly_database(Some(MAIN))?;
+        let word_docids = env.create_database(Some(WORD_DOCIDS))?;
+        let exact_word_docids = env.create_database(Some(EXACT_WORD_DOCIDS))?;
+        let word_prefix_docids = env.create_database(Some(WORD_PREFIX_DOCIDS))?;
+        let exact_word_prefix_docids = env.create_database(Some(EXACT_WORD_PREFIX_DOCIDS))?;
+        let docid_word_positions = env.create_database(Some(DOCID_WORD_POSITIONS))?;
+        let word_pair_proximity_docids = env.create_database(Some(WORD_PAIR_PROXIMITY_DOCIDS))?;
+        let script_language_docids = env.create_database(Some(SCRIPT_LANGUAGE_DOCIDS))?;
        let word_prefix_pair_proximity_docids =
-            env.create_database(&mut wtxn, Some(WORD_PREFIX_PAIR_PROXIMITY_DOCIDS))?;
+            env.create_database(Some(WORD_PREFIX_PAIR_PROXIMITY_DOCIDS))?;
        let prefix_word_pair_proximity_docids =
-            env.create_database(&mut wtxn, Some(PREFIX_WORD_PAIR_PROXIMITY_DOCIDS))?;
-        let word_position_docids = env.create_database(&mut wtxn, Some(WORD_POSITION_DOCIDS))?;
-        let word_fid_docids = env.create_database(&mut wtxn, Some(WORD_FIELD_ID_DOCIDS))?;
-        let field_id_word_count_docids =
-            env.create_database(&mut wtxn, Some(FIELD_ID_WORD_COUNT_DOCIDS))?;
-        let word_prefix_position_docids =
-            env.create_database(&mut wtxn, Some(WORD_PREFIX_POSITION_DOCIDS))?;
-        let word_prefix_fid_docids =
-            env.create_database(&mut wtxn, Some(WORD_PREFIX_FIELD_ID_DOCIDS))?;
-        let facet_id_f64_docids = env.create_database(&mut wtxn, Some(FACET_ID_F64_DOCIDS))?;
-        let facet_id_string_docids =
-            env.create_database(&mut wtxn, Some(FACET_ID_STRING_DOCIDS))?;
-        let facet_id_exists_docids =
-            env.create_database(&mut wtxn, Some(FACET_ID_EXISTS_DOCIDS))?;
-        let facet_id_is_null_docids =
-            env.create_database(&mut wtxn, Some(FACET_ID_IS_NULL_DOCIDS))?;
-        let facet_id_is_empty_docids =
-            env.create_database(&mut wtxn, Some(FACET_ID_IS_EMPTY_DOCIDS))?;
-        let field_id_docid_facet_f64s =
-            env.create_database(&mut wtxn, Some(FIELD_ID_DOCID_FACET_F64S))?;
+            env.create_database(Some(PREFIX_WORD_PAIR_PROXIMITY_DOCIDS))?;
+        let word_position_docids = env.create_database(Some(WORD_POSITION_DOCIDS))?;
+        let word_fid_docids = env.create_database(Some(WORD_FIELD_ID_DOCIDS))?;
+        let field_id_word_count_docids = env.create_database(Some(FIELD_ID_WORD_COUNT_DOCIDS))?;
+        let word_prefix_position_docids = env.create_database(Some(WORD_PREFIX_POSITION_DOCIDS))?;
+        let word_prefix_fid_docids = env.create_database(Some(WORD_PREFIX_FIELD_ID_DOCIDS))?;
+        let facet_id_f64_docids = env.create_database(Some(FACET_ID_F64_DOCIDS))?;
+        let facet_id_string_docids = env.create_database(Some(FACET_ID_STRING_DOCIDS))?;
+        let facet_id_exists_docids = env.create_database(Some(FACET_ID_EXISTS_DOCIDS))?;
+        let facet_id_is_null_docids = env.create_database(Some(FACET_ID_IS_NULL_DOCIDS))?;
+        let facet_id_is_empty_docids = env.create_database(Some(FACET_ID_IS_EMPTY_DOCIDS))?;
+
+        let field_id_docid_facet_f64s = env.create_database(Some(FIELD_ID_DOCID_FACET_F64S))?;
        let field_id_docid_facet_strings =
-            env.create_database(&mut wtxn, Some(FIELD_ID_DOCID_FACET_STRINGS))?;
-        let vector_id_docid = env.create_database(&mut wtxn, Some(VECTOR_ID_DOCID))?;
-        let documents = env.create_database(&mut wtxn, Some(DOCUMENTS))?;
-        wtxn.commit()?;
+            env.create_database(Some(FIELD_ID_DOCID_FACET_STRINGS))?;
+        let documents = env.create_database(Some(DOCUMENTS))?;

        Index::set_creation_dates(&env, main, created_at, updated_at)?;

@@ -225,6 +207,7 @@ impl Index {
            exact_word_docids,
            word_prefix_docids,
            exact_word_prefix_docids,
+            docid_word_positions,
            word_pair_proximity_docids,
            script_language_docids,
            word_prefix_pair_proximity_docids,
@@ -241,7 +224,6 @@ impl Index {
            facet_id_is_empty_docids,
            field_id_docid_facet_f64s,
            field_id_docid_facet_strings,
-            vector_id_docid,
            documents,
        })
    }
@@ -513,26 +495,6 @@ impl Index {
        }
    }

-    /* vector HNSW */
-
-    /// Writes the provided `hnsw`.
-    pub(crate) fn put_vector_hnsw(&self, wtxn: &mut RwTxn, hnsw: &Hnsw) -> heed::Result<()> {
-        self.main.put::<_, Str, SerdeBincode<Hnsw>>(wtxn, main_key::VECTOR_HNSW_KEY, hnsw)
-    }
-
-    /// Delete the `hnsw`.
-    pub(crate) fn delete_vector_hnsw(&self, wtxn: &mut RwTxn) -> heed::Result<bool> {
-        self.main.delete::<_, Str>(wtxn, main_key::VECTOR_HNSW_KEY)
-    }
-
-    /// Returns the `hnsw`.
-    pub fn vector_hnsw(&self, rtxn: &RoTxn) -> Result<Option<Hnsw>> {
-        match self.main.get::<_, Str, SerdeBincode<Hnsw>>(rtxn, main_key::VECTOR_HNSW_KEY)? {
-            Some(hnsw) => Ok(Some(hnsw)),
-            None => Ok(None),
-        }
-    }
-
    /* field distribution */

    /// Writes the field distribution which associates every field name with
@@ -1497,9 +1459,9 @@ pub(crate) mod tests {

        db_snap!(index, field_distribution,
            @r###"
-        age              1      |
-        id               2      |
-        name             2      |
+        age              1     
+        id               2     
+        name             2     
        "###
        );

@@ -1517,9 +1479,9 @@ pub(crate) mod tests {

        db_snap!(index, field_distribution,
            @r###"
-        age              1      |
-        id               2      |
-        name             2      |
+        age              1     
+        id               2     
+        name             2     
        "###
        );

@@ -1533,9 +1495,9 @@ pub(crate) mod tests {

        db_snap!(index, field_distribution,
            @r###"
-        has_dog          1      |
-        id               2      |
-        name             2      |
+        has_dog          1     
+        id               2     
+        name             2     
        "###
        );
    }
--- a/milli/src/lib.rs
+++ b/milli/src/lib.rs
@@ -5,12 +5,57 @@
 #[global_allocator]
 pub static ALLOC: mimalloc::MiMalloc = mimalloc::MiMalloc;

+// #[cfg(test)]
+// pub mod allocator {
+//     use std::alloc::{GlobalAlloc, System};
+//     use std::sync::atomic::{self, AtomicI64};
+
+//     #[global_allocator]
+//     pub static ALLOC: CountingAlloc = CountingAlloc {
+//         max_resident: AtomicI64::new(0),
+//         resident: AtomicI64::new(0),
+//         allocated: AtomicI64::new(0),
+//     };
+
+//     pub struct CountingAlloc {
+//         pub max_resident: AtomicI64,
+//         pub resident: AtomicI64,
+//         pub allocated: AtomicI64,
+//     }
+//     unsafe impl GlobalAlloc for CountingAlloc {
+//         unsafe fn alloc(&self, layout: std::alloc::Layout) -> *mut u8 {
+//             self.allocated.fetch_add(layout.size() as i64, atomic::Ordering::SeqCst);
+//             let old_resident =
+//                 self.resident.fetch_add(layout.size() as i64, atomic::Ordering::SeqCst);
+
+//             let resident = old_resident + layout.size() as i64;
+//             self.max_resident.fetch_max(resident, atomic::Ordering::SeqCst);
+
+//             // if layout.size() > 1_000_000 {
+//             //     eprintln!(
+//             //         "allocating {} with new resident size: {resident}",
+//             //         layout.size() / 1_000_000
+//             //     );
+//             //     // let trace = std::backtrace::Backtrace::capture();
+//             //     // let t = trace.to_string();
+//             //     // eprintln!("{t}");
+//             // }
+
+//             System.alloc(layout)
+//         }
+
+//         unsafe fn dealloc(&self, ptr: *mut u8, layout: std::alloc::Layout) {
+//             self.resident.fetch_sub(layout.size() as i64, atomic::Ordering::Relaxed);
+//             System.dealloc(ptr, layout)
+//         }
+//     }
+// }
+
 #[macro_use]
 pub mod documents;

 mod asc_desc;
 mod criterion;
-pub mod distance;
 mod error;
 mod external_documents_ids;
 pub mod facet;
@@ -30,7 +75,6 @@ use std::convert::{TryFrom, TryInto};
 use std::hash::BuildHasherDefault;

 use charabia::normalizer::{CharNormalizer, CompatibilityDecompositionNormalizer};
-pub use distance::{dot_product_similarity, euclidean_squared_distance};
 pub use filter_parser::{Condition, FilterCondition, Span, Token};
 use fxhash::{FxHasher32, FxHasher64};
 pub use grenad::CompressionType;
@@ -285,35 +329,6 @@ pub fn normalize_facet(original: &str) -> String {
    CompatibilityDecompositionNormalizer.normalize_str(original.trim()).to_lowercase()
 }

-/// Represents either a vector or an array of multiple vectors.
-#[derive(serde::Serialize, serde::Deserialize, Debug)]
-#[serde(transparent)]
-pub struct VectorOrArrayOfVectors {
-    #[serde(with = "either::serde_untagged")]
-    inner: either::Either<Vec<f32>, Vec<Vec<f32>>>,
-}
-
-impl VectorOrArrayOfVectors {
-    pub fn into_array_of_vectors(self) -> Vec<Vec<f32>> {
-        match self.inner {
-            either::Either::Left(vector) => vec![vector],
-            either::Either::Right(vectors) => vectors,
-        }
-    }
-}
-
-/// Normalize a vector by dividing the dimensions by the lenght of it.
-pub fn normalize_vector(mut vector: Vec<f32>) -> Vec<f32> {
-    let squared: f32 = vector.iter().map(|x| x * x).sum();
-    let length = squared.sqrt();
-    if length <= f32::EPSILON {
-        vector
-    } else {
-        vector.iter_mut().for_each(|x| *x /= length);
-        vector
-    }
-}
-
 #[cfg(test)]
 mod tests {
    use serde_json::json;
--- a/milli/src/search/mod.rs
+++ b/milli/src/search/mod.rs
@@ -22,7 +22,6 @@ pub mod new;

 pub struct Search<'a> {
    query: Option<String>,
-    vector: Option<Vec<f32>>,
    // this should be linked to the String in the query
    filter: Option<Filter<'a>>,
    offset: usize,
@@ -40,7 +39,6 @@ impl<'a> Search<'a> {
    pub fn new(rtxn: &'a heed::RoTxn, index: &'a Index) -> Search<'a> {
        Search {
            query: None,
-            vector: None,
            filter: None,
            offset: 0,
            limit: 20,
@@ -59,11 +57,6 @@ impl<'a> Search<'a> {
        self
    }

-    pub fn vector(&mut self, vector: impl Into<Vec<f32>>) -> &mut Search<'a> {
-        self.vector = Some(vector.into());
-        self
-    }
-
    pub fn offset(&mut self, offset: usize) -> &mut Search<'a> {
        self.offset = offset;
        self
@@ -113,7 +106,6 @@ impl<'a> Search<'a> {
            execute_search(
                &mut ctx,
                &self.query,
-                &self.vector,
                self.terms_matching_strategy,
                self.exhaustive_number_hits,
                &self.filter,
@@ -140,7 +132,6 @@ impl fmt::Debug for Search<'_> {
    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
        let Search {
            query,
-            vector: _,
            filter,
            offset,
            limit,
@@ -154,7 +145,6 @@ impl fmt::Debug for Search<'_> {
        } = self;
        f.debug_struct("Search")
            .field("query", query)
-            .field("vector", &"[...]")
            .field("filter", filter)
            .field("offset", offset)
            .field("limit", limit)
@@ -211,14 +201,12 @@ pub fn build_dfa(word: &str, typos: u8, is_prefix: bool) -> DFA {

 #[cfg(test)]
 mod test {
-    #[allow(unused_imports)]
    use super::*;
+    use crate::index::tests::TempIndex;

-    #[cfg(feature = "japanese")]
+    #[cfg(feature = "default")]
    #[test]
    fn test_kanji_language_detection() {
-        use crate::index::tests::TempIndex;
-
        let index = TempIndex::new();

        index
--- a/milli/src/search/new/graph_based_ranking_rule.rs
+++ b/milli/src/search/new/graph_based_ranking_rule.rs
@@ -46,7 +46,7 @@ use super::logger::SearchLogger;
 use super::query_graph::QueryNode;
 use super::ranking_rule_graph::{
    ConditionDocIdsCache, DeadEndsCache, ExactnessGraph, FidGraph, PositionGraph, ProximityGraph,
-    RankingRuleGraph, RankingRuleGraphTrait, TypoGraph, WordsGraph,
+    RankingRuleGraph, RankingRuleGraphTrait, TypoGraph,
 };
 use super::small_bitmap::SmallBitmap;
 use super::{QueryGraph, RankingRule, RankingRuleOutput, SearchContext};
@@ -54,12 +54,6 @@ use crate::search::new::query_term::LocatedQueryTermSubset;
 use crate::search::new::ranking_rule_graph::PathVisitor;
 use crate::{Result, TermsMatchingStrategy};

-pub type Words = GraphBasedRankingRule<WordsGraph>;
-impl GraphBasedRankingRule<WordsGraph> {
-    pub fn new(terms_matching_strategy: TermsMatchingStrategy) -> Self {
-        Self::new_with_id("words".to_owned(), Some(terms_matching_strategy))
-    }
-}
 pub type Proximity = GraphBasedRankingRule<ProximityGraph>;
 impl GraphBasedRankingRule<ProximityGraph> {
    pub fn new(terms_matching_strategy: Option<TermsMatchingStrategy>) -> Self {
--- a/milli/src/search/new/logger/visual.rs
+++ b/milli/src/search/new/logger/visual.rs
@@ -4,6 +4,7 @@ use std::io::{BufWriter, Write};
 use std::path::{Path, PathBuf};
 use std::time::Instant;

+// use rand::random;
 use roaring::RoaringBitmap;

 use crate::search::new::interner::Interned;
@@ -12,7 +13,6 @@ use crate::search::new::query_term::LocatedQueryTermSubset;
 use crate::search::new::ranking_rule_graph::{
    Edge, FidCondition, FidGraph, PositionCondition, PositionGraph, ProximityCondition,
    ProximityGraph, RankingRuleGraph, RankingRuleGraphTrait, TypoCondition, TypoGraph,
-    WordsCondition, WordsGraph,
 };
 use crate::search::new::ranking_rules::BoxRankingRule;
 use crate::search::new::{QueryGraph, QueryNode, RankingRule, SearchContext, SearchLogger};
@@ -24,12 +24,11 @@ pub enum SearchEvents {
    RankingRuleSkipBucket { ranking_rule_idx: usize, bucket_len: u64 },
    RankingRuleEndIteration { ranking_rule_idx: usize, universe_len: u64 },
    ExtendResults { new: Vec<u32> },
+    WordsGraph { query_graph: QueryGraph },
    ProximityGraph { graph: RankingRuleGraph<ProximityGraph> },
    ProximityPaths { paths: Vec<Vec<Interned<ProximityCondition>>> },
    TypoGraph { graph: RankingRuleGraph<TypoGraph> },
    TypoPaths { paths: Vec<Vec<Interned<TypoCondition>>> },
-    WordsGraph { graph: RankingRuleGraph<WordsGraph> },
-    WordsPaths { paths: Vec<Vec<Interned<WordsCondition>>> },
    FidGraph { graph: RankingRuleGraph<FidGraph> },
    FidPaths { paths: Vec<Vec<Interned<FidCondition>>> },
    PositionGraph { graph: RankingRuleGraph<PositionGraph> },
@@ -140,11 +139,8 @@ impl SearchLogger<QueryGraph> for VisualSearchLogger {
        let Some(location) = self.location.last() else { return };
        match location {
            Location::Words => {
-                if let Some(graph) = state.downcast_ref::<RankingRuleGraph<WordsGraph>>() {
-                    self.events.push(SearchEvents::WordsGraph { graph: graph.clone() });
-                }
-                if let Some(paths) = state.downcast_ref::<Vec<Vec<Interned<WordsCondition>>>>() {
-                    self.events.push(SearchEvents::WordsPaths { paths: paths.clone() });
+                if let Some(query_graph) = state.downcast_ref::<QueryGraph>() {
+                    self.events.push(SearchEvents::WordsGraph { query_graph: query_graph.clone() });
                }
            }
            Location::Typo => {
@@ -333,6 +329,7 @@ impl<'ctx> DetailedLoggerFinish<'ctx> {
            SearchEvents::ExtendResults { new } => {
                self.write_extend_results(new)?;
            }
+            SearchEvents::WordsGraph { query_graph } => self.write_words_graph(query_graph)?,
            SearchEvents::ProximityGraph { graph } => self.write_rr_graph(&graph)?,
            SearchEvents::ProximityPaths { paths } => {
                self.write_rr_graph_paths::<ProximityGraph>(paths)?;
@@ -341,10 +338,6 @@ impl<'ctx> DetailedLoggerFinish<'ctx> {
            SearchEvents::TypoPaths { paths } => {
                self.write_rr_graph_paths::<TypoGraph>(paths)?;
            }
-            SearchEvents::WordsGraph { graph } => self.write_rr_graph(&graph)?,
-            SearchEvents::WordsPaths { paths } => {
-                self.write_rr_graph_paths::<WordsGraph>(paths)?;
-            }
            SearchEvents::FidGraph { graph } => self.write_rr_graph(&graph)?,
            SearchEvents::FidPaths { paths } => {
                self.write_rr_graph_paths::<FidGraph>(paths)?;
@@ -462,7 +455,7 @@ fill: \"#B6E2D3\"
                shape: class
                max_nbr_typo: {}",
                    term_subset.description(ctx),
-                    term_subset.max_typo_cost(ctx)
+                    term_subset.max_nbr_typos(ctx)
                )?;

                for w in term_subset.all_single_words_except_prefix_db(ctx)? {
@@ -489,6 +482,13 @@ fill: \"#B6E2D3\"
        }
        Ok(())
    }
+    fn write_words_graph(&mut self, qg: QueryGraph) -> Result<()> {
+        self.make_new_file_for_internal_state_if_needed()?;
+
+        self.write_query_graph(&qg)?;
+
+        Ok(())
+    }
    fn write_rr_graph<R: RankingRuleGraphTrait>(
        &mut self,
        graph: &RankingRuleGraph<R>,
--- a/milli/src/search/new/matches/matching_words.rs
+++ b/milli/src/search/new/matches/matching_words.rs
@@ -52,7 +52,7 @@ impl MatchingWords {
            words.push(LocatedMatchingWords {
                value: matching_words,
                positions: located_term.positions.clone(),
-                is_prefix: term.is_prefix(),
+                is_prefix: term.is_cached_prefix(),
                original_char_count: term.original_word(&ctx).chars().count(),
            });
        }
@@ -244,8 +244,6 @@ pub(crate) mod tests {
        temp_index
            .add_documents(documents!([
                { "id": 1, "name": "split this world westfali westfalia the Ŵôřlḑôle" },
-                { "id": 2, "name": "Westfália" },
-                { "id": 3, "name": "Ŵôřlḑôle" },
            ]))
            .unwrap();
        temp_index
@@ -307,7 +305,7 @@ pub(crate) mod tests {
                    ..Default::default()
                })
                .next(),
-            Some(MatchType::Full { char_len: 5, ids: &(2..=2) })
+            None
        );
        assert_eq!(
            matching_words
--- a/milli/src/search/new/matches/mod.rs
+++ b/milli/src/search/new/matches/mod.rs
@@ -499,37 +499,17 @@ mod tests {
    use charabia::TokenizerBuilder;
    use matching_words::tests::temp_index_with_documents;

+    use super::super::located_query_terms_from_tokens;
    use super::*;
-    use crate::index::tests::TempIndex;
-    use crate::{execute_search, SearchContext};
+    use crate::SearchContext;

    impl<'a> MatcherBuilder<'a, &[u8]> {
-        fn new_test(rtxn: &'a heed::RoTxn, index: &'a TempIndex, query: &str) -> Self {
-            let mut ctx = SearchContext::new(index, rtxn);
-            let crate::search::PartialSearchResult { located_query_terms, .. } = execute_search(
-                &mut ctx,
-                &Some(query.to_string()),
-                &None,
-                crate::TermsMatchingStrategy::default(),
-                false,
-                &None,
-                &None,
-                crate::search::new::GeoSortStrategy::default(),
-                0,
-                100,
-                Some(10),
-                &mut crate::DefaultSearchLogger,
-                &mut crate::DefaultSearchLogger,
-            )
-            .unwrap();
-
-            // consume context and located_query_terms to build MatchingWords.
-            let matching_words = match located_query_terms {
-                Some(located_query_terms) => MatchingWords::new(ctx, located_query_terms),
-                None => MatchingWords::default(),
-            };
-
-            MatcherBuilder::new(matching_words, TokenizerBuilder::new().build())
+        pub fn new_test(mut ctx: SearchContext, query: &'a str) -> Self {
+            let tokenizer = TokenizerBuilder::new().build();
+            let tokens = tokenizer.tokenize(query);
+            let query_terms = located_query_terms_from_tokens(&mut ctx, tokens, None).unwrap();
+            let matching_words = MatchingWords::new(ctx, query_terms);
+            Self::new(matching_words, TokenizerBuilder::new().build())
        }
    }

@@ -537,7 +517,8 @@ mod tests {
    fn format_identity() {
        let temp_index = temp_index_with_documents();
        let rtxn = temp_index.read_txn().unwrap();
-        let builder = MatcherBuilder::new_test(&rtxn, &temp_index, "split the world");
+        let ctx = SearchContext::new(&temp_index, &rtxn);
+        let builder = MatcherBuilder::new_test(ctx, "split the world");

        let format_options = FormatOptions { highlight: false, crop: None };

@@ -564,7 +545,8 @@ mod tests {
    fn format_highlight() {
        let temp_index = temp_index_with_documents();
        let rtxn = temp_index.read_txn().unwrap();
-        let builder = MatcherBuilder::new_test(&rtxn, &temp_index, "split the world");
+        let ctx = SearchContext::new(&temp_index, &rtxn);
+        let builder = MatcherBuilder::new_test(ctx, "split the world");

        let format_options = FormatOptions { highlight: true, crop: None };

@@ -607,7 +589,8 @@ mod tests {
    fn highlight_unicode() {
        let temp_index = temp_index_with_documents();
        let rtxn = temp_index.read_txn().unwrap();
-        let builder = MatcherBuilder::new_test(&rtxn, &temp_index, "world");
+        let ctx = SearchContext::new(&temp_index, &rtxn);
+        let builder = MatcherBuilder::new_test(ctx, "world");
        let format_options = FormatOptions { highlight: true, crop: None };

        // Text containing prefix match.
@@ -616,7 +599,7 @@ mod tests {
        // no crop should return complete text with highlighted matches.
        insta::assert_snapshot!(
            matcher.format(format_options),
-            @"<em>Ŵôřlḑ</em>ôle"
+            @"<em>Ŵôřlḑôle</em>"
        );

        // Text containing unicode match.
@@ -628,7 +611,8 @@ mod tests {
            @"<em>Ŵôřlḑ</em>"
        );

-        let builder = MatcherBuilder::new_test(&rtxn, &temp_index, "westfali");
+        let ctx = SearchContext::new(&temp_index, &rtxn);
+        let builder = MatcherBuilder::new_test(ctx, "westfali");
        let format_options = FormatOptions { highlight: true, crop: None };

        // Text containing unicode match.
@@ -637,7 +621,7 @@ mod tests {
        // no crop should return complete text with highlighted matches.
        insta::assert_snapshot!(
            matcher.format(format_options),
-            @"<em>Westfáli</em>a"
+            @"<em>Westfália</em>"
        );
    }

@@ -645,7 +629,8 @@ mod tests {
    fn format_crop() {
        let temp_index = temp_index_with_documents();
        let rtxn = temp_index.read_txn().unwrap();
-        let builder = MatcherBuilder::new_test(&rtxn, &temp_index, "split the world");
+        let ctx = SearchContext::new(&temp_index, &rtxn);
+        let builder = MatcherBuilder::new_test(ctx, "split the world");

        let format_options = FormatOptions { highlight: false, crop: Some(10) };

@@ -742,7 +727,8 @@ mod tests {
    fn format_highlight_crop() {
        let temp_index = temp_index_with_documents();
        let rtxn = temp_index.read_txn().unwrap();
-        let builder = MatcherBuilder::new_test(&rtxn, &temp_index, "split the world");
+        let ctx = SearchContext::new(&temp_index, &rtxn);
+        let builder = MatcherBuilder::new_test(ctx, "split the world");

        let format_options = FormatOptions { highlight: true, crop: Some(10) };

@@ -804,7 +790,8 @@ mod tests {
        //! testing: https://github.com/meilisearch/specifications/pull/120#discussion_r836536295
        let temp_index = temp_index_with_documents();
        let rtxn = temp_index.read_txn().unwrap();
-        let builder = MatcherBuilder::new_test(&rtxn, &temp_index, "split the world");
+        let ctx = SearchContext::new(&temp_index, &rtxn);
+        let builder = MatcherBuilder::new_test(ctx, "split the world");

        let text = "void void split the world void void.";

@@ -840,8 +827,8 @@ mod tests {
    fn partial_matches() {
        let temp_index = temp_index_with_documents();
        let rtxn = temp_index.read_txn().unwrap();
-        let mut builder =
-            MatcherBuilder::new_test(&rtxn, &temp_index, "the \"t he\" door \"do or\"");
+        let ctx = SearchContext::new(&temp_index, &rtxn);
+        let mut builder = MatcherBuilder::new_test(ctx, "the \"t he\" door \"do or\"");
        builder.highlight_prefix("_".to_string());
        builder.highlight_suffix("_".to_string());

--- a/milli/src/search/new/mod.rs
+++ b/milli/src/search/new/mod.rs
@@ -15,7 +15,11 @@ mod resolve_query_graph;
 mod small_bitmap;

 mod exact_attribute;
+// TODO: documentation + comments
+// implementation is currently an adaptation of the previous implementation to fit with the new model
 mod sort;
+// TODO: documentation + comments
+mod words;

 #[cfg(test)]
 mod tests;
@@ -28,7 +32,6 @@ use db_cache::DatabaseCache;
 use exact_attribute::ExactAttribute;
 use graph_based_ranking_rule::{Exactness, Fid, Position, Proximity, Typo};
 use heed::RoTxn;
-use hnsw::Searcher;
 use interner::{DedupInterner, Interner};
 pub use logger::visual::VisualSearchLogger;
 pub use logger::{DefaultSearchLogger, SearchLogger};
@@ -40,17 +43,13 @@ use ranking_rules::{
 use resolve_query_graph::{compute_query_graph_docids, PhraseDocIdsCache};
 use roaring::RoaringBitmap;
 use sort::Sort;
-use space::Neighbor;
+use words::Words;

 use self::geo_sort::GeoSort;
 pub use self::geo_sort::Strategy as GeoSortStrategy;
-use self::graph_based_ranking_rule::Words;
 use self::interner::Interned;
 use crate::search::new::distinct::apply_distinct_rule;
-use crate::{
-    normalize_vector, AscDesc, DocumentId, Filter, Index, Member, Result, TermsMatchingStrategy,
-    UserError, BEU32,
-};
+use crate::{AscDesc, DocumentId, Filter, Index, Member, Result, TermsMatchingStrategy, UserError};

 /// A structure used throughout the execution of a search query.
 pub struct SearchContext<'ctx> {
@@ -203,11 +202,6 @@ fn get_ranking_rules_for_query_graph_search<'ctx>(
    let mut sorted_fields = HashSet::new();
    let mut geo_sorted = false;

-    // Don't add the `words` ranking rule if the term matching strategy is `All`
-    if matches!(terms_matching_strategy, TermsMatchingStrategy::All) {
-        words = true;
-    }
-
    let mut ranking_rules: Vec<BoxRankingRule<QueryGraph>> = vec![];
    let settings_ranking_rules = ctx.index.criteria(ctx.txn)?;
    for rr in settings_ranking_rules {
@@ -354,7 +348,6 @@ fn resolve_sort_criteria<'ctx, Query: RankingRuleQueryTrait>(
 pub fn execute_search(
    ctx: &mut SearchContext,
    query: &Option<String>,
-    vector: &Option<Vec<f32>>,
    terms_matching_strategy: TermsMatchingStrategy,
    exhaustive_number_hits: bool,
    filters: &Option<Filter>,
@@ -404,8 +397,8 @@ pub fn execute_search(
        None
    };
    let bucket_sort_output = if let Some(query_terms) = query_terms {
-        let (graph, new_located_query_terms) = QueryGraph::from_query(ctx, &query_terms)?;
-        located_query_terms = Some(new_located_query_terms);
+        let graph = QueryGraph::from_query(ctx, &query_terms)?;
+        located_query_terms = Some(query_terms);

        let ranking_rules = get_ranking_rules_for_query_graph_search(
            ctx,
@@ -434,34 +427,6 @@ pub fn execute_search(

    let BucketSortOutput { docids, mut all_candidates } = bucket_sort_output;

-    let docids = match vector {
-        Some(vector) => {
-            // return the nearest documents that are also part of the candidates.
-            let mut searcher = Searcher::new();
-            let hnsw = ctx.index.vector_hnsw(ctx.txn)?.unwrap_or_default();
-            let ef = hnsw.len().min(100);
-            let mut dest = vec![Neighbor { index: 0, distance: 0 }; ef];
-            let vector = normalize_vector(vector.clone());
-            let neighbors = hnsw.nearest(&vector, ef, &mut searcher, &mut dest[..]);
-
-            let mut docids = Vec::new();
-            for Neighbor { index, distance: _ } in neighbors.iter() {
-                let index = BEU32::new(*index as u32);
-                let docid = ctx.index.vector_id_docid.get(ctx.txn, &index)?.unwrap().get();
-                if universe.contains(docid) {
-                    docids.push(docid);
-                    if docids.len() == (from + length) {
-                        break;
-                    }
-                }
-            }
-
-            docids.into_iter().skip(from).take(length).collect()
-        }
-        // return the search docids if the vector field is not specified
-        None => docids,
-    };
-
    // The candidates is the universe unless the exhaustive number of hits
    // is requested and a distinct attribute is set.
    if exhaustive_number_hits {
--- a/milli/src/search/new/query_graph.rs
+++ b/milli/src/search/new/query_graph.rs
@@ -88,15 +88,12 @@ pub struct QueryGraph {
 }

 impl QueryGraph {
-    /// Build the query graph from the parsed user search query, return an updated list of the located query terms
-    /// which contains ngrams.
+    /// Build the query graph from the parsed user search query.
    pub fn from_query(
        ctx: &mut SearchContext,
        // NOTE: the terms here must be consecutive
        terms: &[LocatedQueryTerm],
-    ) -> Result<(QueryGraph, Vec<LocatedQueryTerm>)> {
-        let mut new_located_query_terms = terms.to_vec();
-
+    ) -> Result<QueryGraph> {
        let nbr_typos = number_of_typos_allowed(ctx)?;

        let mut nodes_data: Vec<QueryNodeData> = vec![QueryNodeData::Start, QueryNodeData::End];
@@ -110,11 +107,10 @@ impl QueryGraph {
        let original_terms_len = terms.len();
        for term_idx in 0..original_terms_len {
            let mut new_nodes = vec![];
-
            let new_node_idx = add_node(
                &mut nodes_data,
                QueryNodeData::Term(LocatedQueryTermSubset {
-                    term_subset: QueryTermSubset::full(terms[term_idx].value),
+                    term_subset: QueryTermSubset::full(Interned::from_raw(term_idx as u16)),
                    positions: terms[term_idx].positions.clone(),
                    term_ids: term_idx as u8..=term_idx as u8,
                }),
@@ -125,7 +121,6 @@ impl QueryGraph {
                if let Some(ngram) =
                    query_term::make_ngram(ctx, &terms[term_idx - 1..=term_idx], &nbr_typos)?
                {
-                    new_located_query_terms.push(ngram.clone());
                    let ngram_idx = add_node(
                        &mut nodes_data,
                        QueryNodeData::Term(LocatedQueryTermSubset {
@@ -141,7 +136,6 @@ impl QueryGraph {
                if let Some(ngram) =
                    query_term::make_ngram(ctx, &terms[term_idx - 2..=term_idx], &nbr_typos)?
                {
-                    new_located_query_terms.push(ngram.clone());
                    let ngram_idx = add_node(
                        &mut nodes_data,
                        QueryNodeData::Term(LocatedQueryTermSubset {
@@ -173,7 +167,7 @@ impl QueryGraph {
        let mut graph = QueryGraph { root_node, end_node, nodes };
        graph.build_initial_edges();

-        Ok((graph, new_located_query_terms))
+        Ok(graph)
    }

    /// Remove the given nodes, connecting all their predecessors to all their successors.
--- a/milli/src/search/new/query_term/compute_derivations.rs
+++ b/milli/src/search/new/query_term/compute_derivations.rs
@@ -28,14 +28,16 @@ pub enum ZeroOrOneTypo {
 impl Interned<QueryTerm> {
    pub fn compute_fully_if_needed(self, ctx: &mut SearchContext) -> Result<()> {
        let s = ctx.term_interner.get_mut(self);
-        if s.max_levenshtein_distance <= 1 && s.one_typo.is_uninit() {
+        if s.max_nbr_typos == 0 {
+            s.one_typo = Lazy::Init(OneTypoTerm::default());
+            s.two_typo = Lazy::Init(TwoTypoTerm::default());
+        } else if s.max_nbr_typos == 1 && s.one_typo.is_uninit() {
            assert!(s.two_typo.is_uninit());
-            // Initialize one_typo subterm even if max_nbr_typo is 0 because of split words
            self.initialize_one_typo_subterm(ctx)?;
            let s = ctx.term_interner.get_mut(self);
            assert!(s.one_typo.is_init());
            s.two_typo = Lazy::Init(TwoTypoTerm::default());
-        } else if s.max_levenshtein_distance > 1 && s.two_typo.is_uninit() {
+        } else if s.max_nbr_typos > 1 && s.two_typo.is_uninit() {
            assert!(s.two_typo.is_uninit());
            self.initialize_one_and_two_typo_subterm(ctx)?;
            let s = ctx.term_interner.get_mut(self);
@@ -185,7 +187,7 @@ pub fn partially_initialized_term_from_word(
                original: ctx.word_interner.insert(word.to_owned()),
                ngram_words: None,
                is_prefix: false,
-                max_levenshtein_distance: 0,
+                max_nbr_typos: 0,
                zero_typo: <_>::default(),
                one_typo: Lazy::Init(<_>::default()),
                two_typo: Lazy::Init(<_>::default()),
@@ -256,7 +258,7 @@ pub fn partially_initialized_term_from_word(
    Ok(QueryTerm {
        original: word_interned,
        ngram_words: None,
-        max_levenshtein_distance: max_typo,
+        max_nbr_typos: max_typo,
        is_prefix,
        zero_typo,
        one_typo: Lazy::Uninit,
@@ -275,16 +277,7 @@ fn find_split_words(ctx: &mut SearchContext, word: &str) -> Result<Option<Intern
 impl Interned<QueryTerm> {
    fn initialize_one_typo_subterm(self, ctx: &mut SearchContext) -> Result<()> {
        let self_mut = ctx.term_interner.get_mut(self);
-
-        let allows_split_words = self_mut.allows_split_words();
-        let QueryTerm {
-            original,
-            is_prefix,
-            one_typo,
-            max_levenshtein_distance: max_nbr_typos,
-            ..
-        } = self_mut;
-
+        let QueryTerm { original, is_prefix, one_typo, .. } = self_mut;
        let original = *original;
        let is_prefix = *is_prefix;
        // let original_str = ctx.word_interner.get(*original).to_owned();
@@ -293,33 +286,26 @@ impl Interned<QueryTerm> {
        }
        let mut one_typo_words = BTreeSet::new();

-        if *max_nbr_typos > 0 {
-            find_zero_one_typo_derivations(ctx, original, is_prefix, |derived_word, nbr_typos| {
-                match nbr_typos {
-                    ZeroOrOneTypo::Zero => {}
-                    ZeroOrOneTypo::One => {
-                        if one_typo_words.len() < limits::MAX_ONE_TYPO_COUNT {
-                            one_typo_words.insert(derived_word);
-                        } else {
-                            return Ok(ControlFlow::Break(()));
-                        }
+        find_zero_one_typo_derivations(ctx, original, is_prefix, |derived_word, nbr_typos| {
+            match nbr_typos {
+                ZeroOrOneTypo::Zero => {}
+                ZeroOrOneTypo::One => {
+                    if one_typo_words.len() < limits::MAX_ONE_TYPO_COUNT {
+                        one_typo_words.insert(derived_word);
+                    } else {
+                        return Ok(ControlFlow::Break(()));
                    }
                }
-                Ok(ControlFlow::Continue(()))
-            })?;
-        }
-
-        let split_words = if allows_split_words {
-            let original_str = ctx.word_interner.get(original).to_owned();
-            find_split_words(ctx, original_str.as_str())?
-        } else {
-            None
-        };
+            }
+            Ok(ControlFlow::Continue(()))
+        })?;
+        let original_str = ctx.word_interner.get(original).to_owned();
+        let split_words = find_split_words(ctx, original_str.as_str())?;

        let self_mut = ctx.term_interner.get_mut(self);

        // Only add the split words to the derivations if:
-        // 1. the term is neither an ngram nor a phrase; OR
+        // 1. the term is not an ngram; OR
        // 2. the term is an ngram, but the split words are different from the ngram's component words
        let split_words = if let Some((ngram_words, split_words)) =
            self_mut.ngram_words.as_ref().zip(split_words.as_ref())
@@ -341,13 +327,7 @@ impl Interned<QueryTerm> {
    }
    fn initialize_one_and_two_typo_subterm(self, ctx: &mut SearchContext) -> Result<()> {
        let self_mut = ctx.term_interner.get_mut(self);
-        let QueryTerm {
-            original,
-            is_prefix,
-            two_typo,
-            max_levenshtein_distance: max_nbr_typos,
-            ..
-        } = self_mut;
+        let QueryTerm { original, is_prefix, two_typo, .. } = self_mut;
        let original_str = ctx.word_interner.get(*original).to_owned();
        if two_typo.is_init() {
            return Ok(());
@@ -355,37 +335,34 @@ impl Interned<QueryTerm> {
        let mut one_typo_words = BTreeSet::new();
        let mut two_typo_words = BTreeSet::new();

-        if *max_nbr_typos > 0 {
-            find_zero_one_two_typo_derivations(
-                *original,
-                *is_prefix,
-                ctx.index.words_fst(ctx.txn)?,
-                &mut ctx.word_interner,
-                |derived_word, nbr_typos| {
-                    if one_typo_words.len() >= limits::MAX_ONE_TYPO_COUNT
-                        && two_typo_words.len() >= limits::MAX_TWO_TYPOS_COUNT
-                    {
-                        // No chance we will add either one- or two-typo derivations anymore, stop iterating.
-                        return Ok(ControlFlow::Break(()));
-                    }
-                    match nbr_typos {
-                        NumberOfTypos::Zero => {}
-                        NumberOfTypos::One => {
-                            if one_typo_words.len() < limits::MAX_ONE_TYPO_COUNT {
-                                one_typo_words.insert(derived_word);
-                            }
-                        }
-                        NumberOfTypos::Two => {
-                            if two_typo_words.len() < limits::MAX_TWO_TYPOS_COUNT {
-                                two_typo_words.insert(derived_word);
-                            }
+        find_zero_one_two_typo_derivations(
+            *original,
+            *is_prefix,
+            ctx.index.words_fst(ctx.txn)?,
+            &mut ctx.word_interner,
+            |derived_word, nbr_typos| {
+                if one_typo_words.len() >= limits::MAX_ONE_TYPO_COUNT
+                    && two_typo_words.len() >= limits::MAX_TWO_TYPOS_COUNT
+                {
+                    // No chance we will add either one- or two-typo derivations anymore, stop iterating.
+                    return Ok(ControlFlow::Break(()));
+                }
+                match nbr_typos {
+                    NumberOfTypos::Zero => {}
+                    NumberOfTypos::One => {
+                        if one_typo_words.len() < limits::MAX_ONE_TYPO_COUNT {
+                            one_typo_words.insert(derived_word);
                        }
                    }
-                    Ok(ControlFlow::Continue(()))
-                },
-            )?;
-        }
-
+                    NumberOfTypos::Two => {
+                        if two_typo_words.len() < limits::MAX_TWO_TYPOS_COUNT {
+                            two_typo_words.insert(derived_word);
+                        }
+                    }
+                }
+                Ok(ControlFlow::Continue(()))
+            },
+        )?;
        let split_words = find_split_words(ctx, original_str.as_str())?;
        let self_mut = ctx.term_interner.get_mut(self);

--- a/milli/src/search/new/query_term/mod.rs
+++ b/milli/src/search/new/query_term/mod.rs
@@ -43,7 +43,7 @@ pub struct QueryTermSubset {
 pub struct QueryTerm {
    original: Interned<String>,
    ngram_words: Option<Vec<Interned<String>>>,
-    max_levenshtein_distance: u8,
+    max_nbr_typos: u8,
    is_prefix: bool,
    zero_typo: ZeroTypoTerm,
    // May not be computed yet
@@ -342,16 +342,10 @@ impl QueryTermSubset {
        }
        None
    }
-    pub fn max_typo_cost(&self, ctx: &SearchContext) -> u8 {
+    pub fn max_nbr_typos(&self, ctx: &SearchContext) -> u8 {
        let t = ctx.term_interner.get(self.original);
-        match t.max_levenshtein_distance {
-            0 => {
-                if t.allows_split_words() {
-                    1
-                } else {
-                    0
-                }
-            }
+        match t.max_nbr_typos {
+            0 => 0,
            1 => {
                if self.one_typo_subset.is_empty() {
                    0
@@ -444,9 +438,6 @@ impl QueryTerm {

        self.zero_typo.is_empty() && one_typo.is_empty() && two_typo.is_empty()
    }
-    fn allows_split_words(&self) -> bool {
-        self.zero_typo.phrase.is_none()
-    }
 }

 impl Interned<QueryTerm> {
@@ -479,9 +470,6 @@ impl QueryTerm {
    pub fn is_cached_prefix(&self) -> bool {
        self.zero_typo.use_prefix_db.is_some()
    }
-    pub fn is_prefix(&self) -> bool {
-        self.is_prefix
-    }
    pub fn original_word(&self, ctx: &SearchContext) -> String {
        ctx.word_interner.get(self.original).clone()
    }
--- a/milli/src/search/new/query_term/parse_query.rs
+++ b/milli/src/search/new/query_term/parse_query.rs
@@ -77,9 +77,13 @@ pub fn located_query_terms_from_tokens(
                }
            }
            TokenKind::Separator(separator_kind) => {
-                // add penalty for hard separators
-                if let SeparatorKind::Hard = separator_kind {
-                    position = position.wrapping_add(7);
+                match separator_kind {
+                    SeparatorKind::Hard => {
+                        position += 1;
+                    }
+                    SeparatorKind::Soft => {
+                        position += 0;
+                    }
                }

                phrase = 'phrase: {
@@ -213,7 +217,7 @@ pub fn make_ngram(
        original: ngram_str_interned,
        ngram_words: Some(words_interned),
        is_prefix,
-        max_levenshtein_distance: max_nbr_typos,
+        max_nbr_typos,
        zero_typo: term.zero_typo,
        one_typo: Lazy::Uninit,
        two_typo: Lazy::Uninit,
@@ -267,7 +271,7 @@ impl PhraseBuilder {
                QueryTerm {
                    original: ctx.word_interner.insert(phrase_desc),
                    ngram_words: None,
-                    max_levenshtein_distance: 0,
+                    max_nbr_typos: 0,
                    is_prefix: false,
                    zero_typo: ZeroTypoTerm {
                        phrase: Some(phrase),
@@ -284,36 +288,3 @@ impl PhraseBuilder {
        })
    }
 }
-
-#[cfg(test)]
-mod tests {
-    use charabia::TokenizerBuilder;
-
-    use super::*;
-    use crate::index::tests::TempIndex;
-
-    fn temp_index_with_documents() -> TempIndex {
-        let temp_index = TempIndex::new();
-        temp_index
-            .add_documents(documents!([
-                { "id": 1, "name": "split this world westfali westfalia the Ŵôřlḑôle" },
-                { "id": 2, "name": "Westfália" },
-                { "id": 3, "name": "Ŵôřlḑôle" },
-            ]))
-            .unwrap();
-        temp_index
-    }
-
-    #[test]
-    fn start_with_hard_separator() -> Result<()> {
-        let tokenizer = TokenizerBuilder::new().build();
-        let tokens = tokenizer.tokenize(".");
-        let index = temp_index_with_documents();
-        let rtxn = index.read_txn()?;
-        let mut ctx = SearchContext::new(&index, &rtxn);
-        // panics with `attempt to add with overflow` before <https://github.com/meilisearch/meilisearch/issues/3785>
-        let located_query_terms = located_query_terms_from_tokens(&mut ctx, tokens, None)?;
-        assert!(located_query_terms.is_empty());
-        Ok(())
-    }
-}
--- a/milli/src/search/new/ranking_rule_graph/cheapest_paths.rs
+++ b/milli/src/search/new/ranking_rule_graph/cheapest_paths.rs
@@ -205,12 +205,18 @@ impl<G: RankingRuleGraphTrait> VisitorState<G> {
 impl<G: RankingRuleGraphTrait> RankingRuleGraph<G> {
    pub fn find_all_costs_to_end(&self) -> MappedInterner<QueryNode, Vec<u64>> {
        let mut costs_to_end = self.query_graph.nodes.map(|_| vec![]);
+        let mut enqueued = SmallBitmap::new(self.query_graph.nodes.len());

-        self.traverse_breadth_first_backward(self.query_graph.end_node, |cur_node| {
-            if cur_node == self.query_graph.end_node {
-                *costs_to_end.get_mut(self.query_graph.end_node) = vec![0];
-                return;
-            }
+        let mut node_stack = VecDeque::new();
+
+        *costs_to_end.get_mut(self.query_graph.end_node) = vec![0];
+
+        for prev_node in self.query_graph.nodes.get(self.query_graph.end_node).predecessors.iter() {
+            node_stack.push_back(prev_node);
+            enqueued.insert(prev_node);
+        }
+
+        while let Some(cur_node) = node_stack.pop_front() {
            let mut self_costs = Vec::<u64>::new();

            let cur_node_edges = &self.edges_of_node.get(cur_node);
@@ -226,7 +232,13 @@ impl<G: RankingRuleGraphTrait> RankingRuleGraph<G> {
            self_costs.dedup();

            *costs_to_end.get_mut(cur_node) = self_costs;
-        });
+            for prev_node in self.query_graph.nodes.get(cur_node).predecessors.iter() {
+                if !enqueued.contains(prev_node) {
+                    node_stack.push_back(prev_node);
+                    enqueued.insert(prev_node);
+                }
+            }
+        }
        costs_to_end
    }

@@ -235,12 +247,17 @@ impl<G: RankingRuleGraphTrait> RankingRuleGraph<G> {
        node_with_removed_outgoing_conditions: Interned<QueryNode>,
        costs: &mut MappedInterner<QueryNode, Vec<u64>>,
    ) {
-        // Traverse the graph backward from the target node, recomputing the cost for each of its predecessors.
-        // We first check that no other node is contributing the same total cost to a predecessor before removing
-        // the cost from the predecessor.
-        self.traverse_breadth_first_backward(node_with_removed_outgoing_conditions, |cur_node| {
+        let mut enqueued = SmallBitmap::new(self.query_graph.nodes.len());
+        let mut node_stack = VecDeque::new();
+
+        enqueued.insert(node_with_removed_outgoing_conditions);
+        node_stack.push_back(node_with_removed_outgoing_conditions);
+
+        'main_loop: while let Some(cur_node) = node_stack.pop_front() {
            let mut costs_to_remove = FxHashSet::default();
-            costs_to_remove.extend(costs.get(cur_node).iter().copied());
+            for c in costs.get(cur_node) {
+                costs_to_remove.insert(*c);
+            }

            let cur_node_edges = &self.edges_of_node.get(cur_node);
            for edge_idx in cur_node_edges.iter() {
@@ -248,75 +265,22 @@ impl<G: RankingRuleGraphTrait> RankingRuleGraph<G> {
                for cost in costs.get(edge.dest_node).iter() {
                    costs_to_remove.remove(&(*cost + edge.cost as u64));
                    if costs_to_remove.is_empty() {
-                        return;
+                        continue 'main_loop;
                    }
                }
            }
            if costs_to_remove.is_empty() {
-                return;
+                continue 'main_loop;
            }
            let mut new_costs = BTreeSet::from_iter(costs.get(cur_node).iter().copied());
            for c in costs_to_remove {
                new_costs.remove(&c);
            }
            *costs.get_mut(cur_node) = new_costs.into_iter().collect();
-        });
-    }

-    /// Traverse the graph backwards from the given node such that every time
-    /// a node is visited, we are guaranteed that all its successors either:
-    /// 1. have already been visited; OR
-    /// 2. were not reachable from the given node
-    pub fn traverse_breadth_first_backward(
-        &self,
-        from: Interned<QueryNode>,
-        mut visit: impl FnMut(Interned<QueryNode>),
-    ) {
-        let mut reachable = SmallBitmap::for_interned_values_in(&self.query_graph.nodes);
-        {
-            // go backward to get the set of all reachable nodes from the given node
-            // the nodes that are not reachable will be set as `visited`
-            let mut stack = VecDeque::new();
-            let mut enqueued = SmallBitmap::for_interned_values_in(&self.query_graph.nodes);
-            enqueued.insert(from);
-            stack.push_back(from);
-            while let Some(n) = stack.pop_front() {
-                if reachable.contains(n) {
-                    continue;
-                }
-                reachable.insert(n);
-                for prev_node in self.query_graph.nodes.get(n).predecessors.iter() {
-                    if !enqueued.contains(prev_node) && !reachable.contains(prev_node) {
-                        stack.push_back(prev_node);
-                        enqueued.insert(prev_node);
-                    }
-                }
-            }
-        };
-        let mut unreachable_or_visited =
-            SmallBitmap::for_interned_values_in(&self.query_graph.nodes);
-        for (n, _) in self.query_graph.nodes.iter() {
-            if !reachable.contains(n) {
-                unreachable_or_visited.insert(n);
-            }
-        }
-
-        let mut enqueued = SmallBitmap::for_interned_values_in(&self.query_graph.nodes);
-        let mut stack = VecDeque::new();
-
-        enqueued.insert(from);
-        stack.push_back(from);
-
-        while let Some(cur_node) = stack.pop_front() {
-            if !self.query_graph.nodes.get(cur_node).successors.is_subset(&unreachable_or_visited) {
-                stack.push_back(cur_node);
-                continue;
-            }
-            unreachable_or_visited.insert(cur_node);
-            visit(cur_node);
            for prev_node in self.query_graph.nodes.get(cur_node).predecessors.iter() {
-                if !enqueued.contains(prev_node) && !unreachable_or_visited.contains(prev_node) {
-                    stack.push_back(prev_node);
+                if !enqueued.contains(prev_node) {
+                    node_stack.push_back(prev_node);
                    enqueued.insert(prev_node);
                }
            }
--- a/milli/src/search/new/ranking_rule_graph/mod.rs
+++ b/milli/src/search/new/ranking_rule_graph/mod.rs
@@ -20,8 +20,6 @@ mod position;
 mod proximity;
 /// Implementation of the `typo` ranking rule
 mod typo;
-/// Implementation of the `words` ranking rule
-mod words;

 use std::collections::BTreeSet;
 use std::hash::Hash;
@@ -35,7 +33,6 @@ pub use position::{PositionCondition, PositionGraph};
 pub use proximity::{ProximityCondition, ProximityGraph};
 use roaring::RoaringBitmap;
 pub use typo::{TypoCondition, TypoGraph};
-pub use words::{WordsCondition, WordsGraph};

 use super::interner::{DedupInterner, FixedSizeInterner, Interned, MappedInterner};
 use super::query_term::LocatedQueryTermSubset;
--- a/milli/src/search/new/ranking_rule_graph/position/mod.rs
+++ b/milli/src/search/new/ranking_rule_graph/position/mod.rs
@@ -111,16 +111,23 @@ impl RankingRuleGraphTrait for PositionGraph {

 fn cost_from_position(sum_positions: u32) -> u32 {
    match sum_positions {
-        0 => 0,
-        1 => 1,
-        2..=4 => 2,
-        5..=7 => 3,
-        8..=11 => 4,
-        12..=16 => 5,
-        17..=24 => 6,
-        25..=64 => 7,
-        65..=256 => 8,
-        257..=1024 => 9,
-        _ => 10,
+        0 | 1 | 2 | 3 => sum_positions,
+        4 | 5 => 4,
+        6 | 7 => 5,
+        8 | 9 => 6,
+        10 | 11 => 7,
+        12 | 13 => 8,
+        14 | 15 => 9,
+        16 | 17..=24 => 10,
+        25..=32 => 11,
+        33..=64 => 12,
+        65..=128 => 13,
+        129..=256 => 14,
+        257..=512 => 15,
+        513..=1024 => 16,
+        1025..=2048 => 17,
+        2049..=4096 => 18,
+        4097..=8192 => 19,
+        _ => 20,
    }
 }
--- a/milli/src/search/new/ranking_rule_graph/typo/mod.rs
+++ b/milli/src/search/new/ranking_rule_graph/typo/mod.rs
@@ -50,7 +50,7 @@ impl RankingRuleGraphTrait for TypoGraph {
        // 3-gram -> equivalent to 2 typos
        let base_cost = if term.term_ids.len() == 1 { 0 } else { term.term_ids.len() as u32 };

-        for nbr_typos in 0..=term.term_subset.max_typo_cost(ctx) {
+        for nbr_typos in 0..=term.term_subset.max_nbr_typos(ctx) {
            let mut term = term.clone();
            match nbr_typos {
                0 => {
--- a/milli/src/search/new/ranking_rule_graph/words/mod.rs
+++ b/milli/src/search/new/ranking_rule_graph/words/mod.rs
@@ -1,49 +0,0 @@
-use roaring::RoaringBitmap;
-
-use super::{ComputedCondition, RankingRuleGraphTrait};
-use crate::search::new::interner::{DedupInterner, Interned};
-use crate::search::new::query_term::LocatedQueryTermSubset;
-use crate::search::new::resolve_query_graph::compute_query_term_subset_docids;
-use crate::search::new::SearchContext;
-use crate::Result;
-
-#[derive(Clone, PartialEq, Eq, Hash)]
-pub struct WordsCondition {
-    term: LocatedQueryTermSubset,
-}
-
-pub enum WordsGraph {}
-
-impl RankingRuleGraphTrait for WordsGraph {
-    type Condition = WordsCondition;
-
-    fn resolve_condition(
-        ctx: &mut SearchContext,
-        condition: &Self::Condition,
-        universe: &RoaringBitmap,
-    ) -> Result<ComputedCondition> {
-        let WordsCondition { term, .. } = condition;
-        // maybe compute_query_term_subset_docids should accept a universe as argument
-        let mut docids = compute_query_term_subset_docids(ctx, &term.term_subset)?;
-        docids &= universe;
-
-        Ok(ComputedCondition {
-            docids,
-            universe_len: universe.len(),
-            start_term_subset: None,
-            end_term_subset: term.clone(),
-        })
-    }
-
-    fn build_edges(
-        _ctx: &mut SearchContext,
-        conditions_interner: &mut DedupInterner<Self::Condition>,
-        _from: Option<&LocatedQueryTermSubset>,
-        to_term: &LocatedQueryTermSubset,
-    ) -> Result<Vec<(u32, Interned<Self::Condition>)>> {
-        Ok(vec![(
-            to_term.term_ids.len() as u32,
-            conditions_interner.insert(WordsCondition { term: to_term.clone() }),
-        )])
-    }
-}
--- a/milli/src/search/new/tests/attribute_position.rs
+++ b/milli/src/search/new/tests/attribute_position.rs
@@ -138,7 +138,7 @@ fn test_attribute_position_simple() {
    s.terms_matching_strategy(TermsMatchingStrategy::All);
    s.query("quick brown");
    let SearchResult { documents_ids, .. } = s.execute().unwrap();
-    insta::assert_snapshot!(format!("{documents_ids:?}"), @"[10, 11, 12, 13, 2, 3, 4, 1, 0, 6, 8, 7, 9, 5]");
+    insta::assert_snapshot!(format!("{documents_ids:?}"), @"[10, 11, 12, 13, 3, 4, 2, 1, 0, 6, 8, 7, 9, 5]");
 }
 #[test]
 fn test_attribute_position_repeated() {
@@ -163,7 +163,7 @@ fn test_attribute_position_different_fields() {
    s.terms_matching_strategy(TermsMatchingStrategy::All);
    s.query("quick brown");
    let SearchResult { documents_ids, .. } = s.execute().unwrap();
-    insta::assert_snapshot!(format!("{documents_ids:?}"), @"[10, 11, 12, 13, 2, 3, 4, 1, 0, 6, 8, 7, 9, 5]");
+    insta::assert_snapshot!(format!("{documents_ids:?}"), @"[10, 11, 12, 13, 3, 4, 2, 1, 0, 6, 8, 7, 9, 5]");
 }

 #[test]
@@ -176,5 +176,5 @@ fn test_attribute_position_ngrams() {
    s.terms_matching_strategy(TermsMatchingStrategy::All);
    s.query("quick brown");
    let SearchResult { documents_ids, .. } = s.execute().unwrap();
-    insta::assert_snapshot!(format!("{documents_ids:?}"), @"[10, 11, 12, 13, 2, 3, 4, 1, 0, 6, 8, 7, 9, 5]");
+    insta::assert_snapshot!(format!("{documents_ids:?}"), @"[10, 11, 12, 13, 3, 4, 2, 1, 0, 6, 8, 7, 9, 5]");
 }
--- a/milli/src/search/new/tests/mod.rs
+++ b/milli/src/search/new/tests/mod.rs
@@ -4,7 +4,7 @@ pub mod distinct;
 pub mod exactness;
 pub mod geo_sort;
 pub mod integration;
-#[cfg(feature = "all-tokenizations")]
+#[cfg(feature = "default")]
 pub mod language;
 pub mod ngram_split_words;
 pub mod proximity;
--- a/milli/src/search/new/tests/ngram_split_words.rs
+++ b/milli/src/search/new/tests/ngram_split_words.rs
@@ -3,9 +3,9 @@ This module tests the following properties:

 1. Two consecutive words from a query can be combined into a "2gram"
 2. Three consecutive words from a query can be combined into a "3gram"
-3. A word from the query can be split into two consecutive words (split words), no matter how short it is
+3. A word from the query can be split into two consecutive words (split words)
 4. A 2gram can be split into two words
-5. A 3gram can be split into two words
+5. A 3gram cannot be split into two words
 6. 2grams can contain up to 1 typo
 7. 3grams cannot have typos
 8. 2grams and 3grams can be prefix tolerant
@@ -14,7 +14,6 @@ This module tests the following properties:
 11. Disabling typo tolerance does not disable ngram tolerance
 12. Prefix tolerance is disabled for the last word if a space follows it
 13. Ngrams cannot be formed by combining a phrase and a word or two phrases
-14. Split words are not disabled by the `disableOnAttribute` or `disableOnWords` typo settings
 */

 use crate::index::tests::TempIndex;
@@ -57,10 +56,6 @@ fn create_index() -> TempIndex {
            {
                "id": 5,
                "text": "sunflowering is not a verb"
-            },
-            {
-                "id": 6,
-                "text": "xy z"
            }
        ]))
        .unwrap();
@@ -268,11 +263,10 @@ fn test_disable_split_words() {
    s.query("sunflower ");
    let SearchResult { documents_ids, .. } = s.execute().unwrap();
    // no document containing `sun flower`
-    insta::assert_snapshot!(format!("{documents_ids:?}"), @"[1, 3]");
+    insta::assert_snapshot!(format!("{documents_ids:?}"), @"[3]");
    let texts = collect_field_values(&index, &txn, "text", &documents_ids);
    insta::assert_debug_snapshot!(texts, @r###"
    [
-        "\"the sun flower is tall\"",
        "\"the sunflower is tall\"",
    ]
    "###);
@@ -313,11 +307,10 @@ fn test_3gram_no_split_words() {
    let SearchResult { documents_ids, .. } = s.execute().unwrap();

    // no document with `sun flower`
-    insta::assert_snapshot!(format!("{documents_ids:?}"), @"[1, 2, 3, 5]");
+    insta::assert_snapshot!(format!("{documents_ids:?}"), @"[2, 3, 5]");
    let texts = collect_field_values(&index, &txn, "text", &documents_ids);
    insta::assert_debug_snapshot!(texts, @r###"
    [
-        "\"the sun flower is tall\"",
        "\"the sunflowers are pretty\"",
        "\"the sunflower is tall\"",
        "\"sunflowering is not a verb\"",
@@ -376,50 +369,3 @@ fn test_no_ngram_phrases() {
    ]
    "###);
 }
-
-#[test]
-fn test_short_split_words() {
-    let index = create_index();
-    let txn = index.read_txn().unwrap();
-
-    let mut s = Search::new(&txn, &index);
-    s.terms_matching_strategy(TermsMatchingStrategy::All);
-    s.query("xyz");
-    let SearchResult { documents_ids, .. } = s.execute().unwrap();
-
-    insta::assert_snapshot!(format!("{documents_ids:?}"), @"[6]");
-    let texts = collect_field_values(&index, &txn, "text", &documents_ids);
-    insta::assert_debug_snapshot!(texts, @r###"
-    [
-        "\"xy z\"",
-    ]
-    "###);
-}
-
-#[test]
-fn test_split_words_never_disabled() {
-    let index = create_index();
-
-    index
-        .update_settings(|s| {
-            s.set_exact_words(["sunflower"].iter().map(ToString::to_string).collect());
-            s.set_exact_attributes(["text"].iter().map(ToString::to_string).collect());
-        })
-        .unwrap();
-
-    let txn = index.read_txn().unwrap();
-
-    let mut s = Search::new(&txn, &index);
-    s.terms_matching_strategy(TermsMatchingStrategy::All);
-    s.query("the sunflower is tall");
-    let SearchResult { documents_ids, .. } = s.execute().unwrap();
-
-    insta::assert_snapshot!(format!("{documents_ids:?}"), @"[1, 3]");
-    let texts = collect_field_values(&index, &txn, "text", &documents_ids);
-    insta::assert_debug_snapshot!(texts, @r###"
-    [
-        "\"the sun flower is tall\"",
-        "\"the sunflower is tall\"",
-    ]
-    "###);
-}
--- a/milli/src/search/new/tests/typo.rs
+++ b/milli/src/search/new/tests/typo.rs
@@ -9,7 +9,7 @@ This module tests the following properties:
 6. A typo on the first letter of a word counts as two typos
 7. Phrases are not typo tolerant
 8. 2grams can have 1 typo if they are larger than `min_word_len_two_typos`
-9. 3grams are not typo tolerant (but they can be split into two words)
+9. 3grams are not typo tolerant
 10. The `typo` ranking rule assumes the role of the `words` ranking rule implicitly
 if `words` doesn't exist before it.
 11. The `typo` ranking rule places documents with the same number of typos in the same bucket
@@ -287,17 +287,16 @@ fn test_typo_exact_word() {
    ]
    "###);

-    // exact words do not disable prefix (sunflowering OK, but no sunflowar)
+    // exact words do not disable prefix (sunflowering OK, but no sunflowar or sun flower)
    let mut s = Search::new(&txn, &index);
    s.terms_matching_strategy(TermsMatchingStrategy::All);
    s.query("network interconnection sunflower");
    let SearchResult { documents_ids, .. } = s.execute().unwrap();
-    insta::assert_snapshot!(format!("{documents_ids:?}"), @"[16, 17, 18]");
+    insta::assert_snapshot!(format!("{documents_ids:?}"), @"[16, 18]");
    let texts = collect_field_values(&index, &txn, "text", &documents_ids);
    insta::assert_debug_snapshot!(texts, @r###"
    [
        "\"network interconnection sunflower\"",
-        "\"network interconnection sun flower\"",
        "\"network interconnection sunflowering\"",
    ]
    "###);
--- a/milli/src/search/new/words.rs
+++ b/milli/src/search/new/words.rs
@@ -0,0 +1,87 @@
+use roaring::RoaringBitmap;
+
+use super::logger::SearchLogger;
+use super::query_graph::QueryNode;
+use super::resolve_query_graph::compute_query_graph_docids;
+use super::small_bitmap::SmallBitmap;
+use super::{QueryGraph, RankingRule, RankingRuleOutput, SearchContext};
+use crate::{Result, TermsMatchingStrategy};
+
+pub struct Words {
+    exhausted: bool, // TODO: remove
+    query_graph: Option<QueryGraph>,
+    nodes_to_remove: Vec<SmallBitmap<QueryNode>>,
+    terms_matching_strategy: TermsMatchingStrategy,
+}
+impl Words {
+    pub fn new(terms_matching_strategy: TermsMatchingStrategy) -> Self {
+        Self {
+            exhausted: true,
+            query_graph: None,
+            nodes_to_remove: vec![],
+            terms_matching_strategy,
+        }
+    }
+}
+
+impl<'ctx> RankingRule<'ctx, QueryGraph> for Words {
+    fn id(&self) -> String {
+        "words".to_owned()
+    }
+    fn start_iteration(
+        &mut self,
+        ctx: &mut SearchContext<'ctx>,
+        _logger: &mut dyn SearchLogger<QueryGraph>,
+        _universe: &RoaringBitmap,
+        parent_query_graph: &QueryGraph,
+    ) -> Result<()> {
+        self.exhausted = false;
+        self.query_graph = Some(parent_query_graph.clone());
+        self.nodes_to_remove = match self.terms_matching_strategy {
+            TermsMatchingStrategy::Last => {
+                let mut ns = parent_query_graph.removal_order_for_terms_matching_strategy_last(ctx);
+                ns.reverse();
+                ns
+            }
+            TermsMatchingStrategy::All => {
+                vec![]
+            }
+        };
+        Ok(())
+    }
+
+    fn next_bucket(
+        &mut self,
+        ctx: &mut SearchContext<'ctx>,
+        logger: &mut dyn SearchLogger<QueryGraph>,
+        universe: &RoaringBitmap,
+    ) -> Result<Option<RankingRuleOutput<QueryGraph>>> {
+        if self.exhausted {
+            return Ok(None);
+        }
+        let Some(query_graph) = &mut self.query_graph else { panic!() };
+        logger.log_internal_state(query_graph);
+
+        let this_bucket = compute_query_graph_docids(ctx, query_graph, universe)?;
+
+        let child_query_graph = query_graph.clone();
+
+        if self.nodes_to_remove.is_empty() {
+            self.exhausted = true;
+        } else {
+            let nodes_to_remove = self.nodes_to_remove.pop().unwrap();
+            query_graph.remove_nodes_keep_edges(&nodes_to_remove.iter().collect::<Vec<_>>());
+        }
+        Ok(Some(RankingRuleOutput { query: child_query_graph, candidates: this_bucket }))
+    }
+
+    fn end_iteration(
+        &mut self,
+        _ctx: &mut SearchContext<'ctx>,
+        _logger: &mut dyn SearchLogger<QueryGraph>,
+    ) {
+        self.exhausted = true;
+        self.nodes_to_remove = vec![];
+        self.query_graph = None;
+    }
+}
--- a/milli/src/snapshot_tests.rs
+++ b/milli/src/snapshot_tests.rs
@@ -89,6 +89,7 @@ Create a snapshot test of the given database.
    - `exact_word_docids`
    - `word_prefix_docids`
    - `exact_word_prefix_docids`
+    - `docid_word_positions`
    - `word_pair_proximity_docids`
    - `word_prefix_pair_proximity_docids`
    - `word_position_docids`
@@ -216,6 +217,11 @@ pub fn snap_exact_word_prefix_docids(index: &Index) -> String {
        &format!("{s:<16} {}", display_bitmap(&b))
    })
 }
+pub fn snap_docid_word_positions(index: &Index) -> String {
+    make_db_snap_from_iter!(index, docid_word_positions, |((idx, s), b)| {
+        &format!("{idx:<6} {s:<16} {}", display_bitmap(&b))
+    })
+}
 pub fn snap_word_pair_proximity_docids(index: &Index) -> String {
    make_db_snap_from_iter!(index, word_pair_proximity_docids, |((proximity, word1, word2), b)| {
        &format!("{proximity:<2} {word1:<16} {word2:<16} {}", display_bitmap(&b))
@@ -318,7 +324,7 @@ pub fn snap_field_distributions(index: &Index) -> String {
    let rtxn = index.read_txn().unwrap();
    let mut snap = String::new();
    for (field, count) in index.field_distribution(&rtxn).unwrap() {
-        writeln!(&mut snap, "{field:<16} {count:<6} |").unwrap();
+        writeln!(&mut snap, "{field:<16} {count:<6}").unwrap();
    }
    snap
 }
@@ -328,7 +334,7 @@ pub fn snap_fields_ids_map(index: &Index) -> String {
    let mut snap = String::new();
    for field_id in fields_ids_map.ids() {
        let name = fields_ids_map.name(field_id).unwrap();
-        writeln!(&mut snap, "{field_id:<3} {name:<16} |").unwrap();
+        writeln!(&mut snap, "{field_id:<3} {name:<16}").unwrap();
    }
    snap
 }
@@ -471,6 +477,9 @@ macro_rules! full_snap_of_db {
    ($index:ident, exact_word_prefix_docids) => {{
        $crate::snapshot_tests::snap_exact_word_prefix_docids(&$index)
    }};
+    ($index:ident, docid_word_positions) => {{
+        $crate::snapshot_tests::snap_docid_word_positions(&$index)
+    }};
    ($index:ident, word_pair_proximity_docids) => {{
        $crate::snapshot_tests::snap_word_pair_proximity_docids(&$index)
    }};
--- a/milli/src/snapshots/index.rs/initial_field_distribution/1/field_distribution.snap
+++ b/milli/src/snapshots/index.rs/initial_field_distribution/1/field_distribution.snap
@@ -1,7 +1,7 @@
 ---
 source: milli/src/index.rs
 ---
-age              1      |
-id               2      |
-name             2      |
+age              1     
+id               2     
+name             2     

--- a/milli/src/snapshots/index.rs/initial_field_distribution/field_distribution.snap
+++ b/milli/src/snapshots/index.rs/initial_field_distribution/field_distribution.snap
@@ -1,7 +1,7 @@
 ---
 source: milli/src/index.rs
 ---
-age              1      |
-id               2      |
-name             2      |
+age              1     
+id               2     
+name             2     

--- a/milli/src/update/clear_documents.rs
+++ b/milli/src/update/clear_documents.rs
@@ -23,6 +23,7 @@ impl<'t, 'u, 'i> ClearDocuments<'t, 'u, 'i> {
            exact_word_docids,
            word_prefix_docids,
            exact_word_prefix_docids,
+            docid_word_positions,
            word_pair_proximity_docids,
            word_prefix_pair_proximity_docids,
            prefix_word_pair_proximity_docids,
@@ -39,7 +40,6 @@ impl<'t, 'u, 'i> ClearDocuments<'t, 'u, 'i> {
            facet_id_is_empty_docids,
            field_id_docid_facet_f64s,
            field_id_docid_facet_strings,
-            vector_id_docid,
            documents,
        } = self.index;

@@ -58,7 +58,6 @@ impl<'t, 'u, 'i> ClearDocuments<'t, 'u, 'i> {
        self.index.put_field_distribution(self.wtxn, &FieldDistribution::default())?;
        self.index.delete_geo_rtree(self.wtxn)?;
        self.index.delete_geo_faceted_documents_ids(self.wtxn)?;
-        self.index.delete_vector_hnsw(self.wtxn)?;

        // We clean all the faceted documents ids.
        for field_id in faceted_fields {
@@ -81,6 +80,7 @@ impl<'t, 'u, 'i> ClearDocuments<'t, 'u, 'i> {
        exact_word_docids.clear(self.wtxn)?;
        word_prefix_docids.clear(self.wtxn)?;
        exact_word_prefix_docids.clear(self.wtxn)?;
+        docid_word_positions.clear(self.wtxn)?;
        word_pair_proximity_docids.clear(self.wtxn)?;
        word_prefix_pair_proximity_docids.clear(self.wtxn)?;
        prefix_word_pair_proximity_docids.clear(self.wtxn)?;
@@ -97,7 +97,6 @@ impl<'t, 'u, 'i> ClearDocuments<'t, 'u, 'i> {
        facet_id_string_docids.clear(self.wtxn)?;
        field_id_docid_facet_f64s.clear(self.wtxn)?;
        field_id_docid_facet_strings.clear(self.wtxn)?;
-        vector_id_docid.clear(self.wtxn)?;
        documents.clear(self.wtxn)?;

        Ok(number_of_documents)
@@ -142,6 +141,7 @@ mod tests {

        assert!(index.word_docids.is_empty(&rtxn).unwrap());
        assert!(index.word_prefix_docids.is_empty(&rtxn).unwrap());
+        assert!(index.docid_word_positions.is_empty(&rtxn).unwrap());
        assert!(index.word_pair_proximity_docids.is_empty(&rtxn).unwrap());
        assert!(index.field_id_word_count_docids.is_empty(&rtxn).unwrap());
        assert!(index.word_prefix_pair_proximity_docids.is_empty(&rtxn).unwrap());
--- a/milli/src/update/delete_documents.rs
+++ b/milli/src/update/delete_documents.rs
@@ -1,13 +1,11 @@
 use std::collections::btree_map::Entry;
-use std::collections::{BTreeSet, HashMap, HashSet};
+use std::collections::{HashMap, HashSet};

 use fst::IntoStreamer;
 use heed::types::{ByteSlice, DecodeIgnore, Str, UnalignedSlice};
 use heed::{BytesDecode, BytesEncode, Database, RwIter};
-use hnsw::Searcher;
 use roaring::RoaringBitmap;
 use serde::{Deserialize, Serialize};
-use space::KnnPoints;
 use time::OffsetDateTime;

 use super::facet::delete::FacetsDelete;
@@ -16,9 +14,9 @@ use crate::error::InternalError;
 use crate::facet::FacetType;
 use crate::heed_codec::facet::FieldDocIdFacetCodec;
 use crate::heed_codec::CboRoaringBitmapCodec;
-use crate::index::Hnsw;
 use crate::{
-    ExternalDocumentsIds, FieldId, FieldIdMapMissingEntry, Index, Result, RoaringBitmapCodec, BEU32,
+    ExternalDocumentsIds, FieldId, FieldIdMapMissingEntry, Index, Result, RoaringBitmapCodec,
+    SmallString32, BEU32,
 };

 pub struct DeleteDocuments<'t, 'u, 'i> {
@@ -234,6 +232,7 @@ impl<'t, 'u, 'i> DeleteDocuments<'t, 'u, 'i> {
            exact_word_docids,
            word_prefix_docids,
            exact_word_prefix_docids,
+            docid_word_positions,
            word_pair_proximity_docids,
            field_id_word_count_docids,
            word_prefix_pair_proximity_docids,
@@ -250,12 +249,25 @@ impl<'t, 'u, 'i> DeleteDocuments<'t, 'u, 'i> {
            facet_id_exists_docids,
            facet_id_is_null_docids,
            facet_id_is_empty_docids,
-            vector_id_docid,
            documents,
        } = self.index;
-        // Remove from the documents database
+
+        // Retrieve the words contained in the documents.
+        let mut words = Vec::new();
        for docid in &self.to_delete_docids {
            documents.delete(self.wtxn, &BEU32::new(docid))?;
+
+            // We iterate through the words positions of the document id, retrieve the word and delete the positions.
+            // We create an iterator to be able to get the content and delete the key-value itself.
+            // It's faster to acquire a cursor to get and delete, as we avoid traversing the LMDB B-Tree two times but only once.
+            let mut iter = docid_word_positions.prefix_iter_mut(self.wtxn, &(docid, ""))?;
+            while let Some(result) = iter.next() {
+                let ((_docid, word), _positions) = result?;
+                // This boolean will indicate if we must remove this word from the words FST.
+                words.push((SmallString32::from(word), false));
+                // safety: we don't keep references from inside the LMDB database.
+                unsafe { iter.del_current()? };
+            }
        }
        // We acquire the current external documents ids map...
        // Note that its soft-deleted document ids field will be equal to the `to_delete_docids`
@@ -266,27 +278,42 @@ impl<'t, 'u, 'i> DeleteDocuments<'t, 'u, 'i> {
        let new_external_documents_ids = new_external_documents_ids.into_static();
        self.index.put_external_documents_ids(self.wtxn, &new_external_documents_ids)?;

-        let mut words_to_keep = BTreeSet::default();
-        let mut words_to_delete = BTreeSet::default();
+        // Maybe we can improve the get performance of the words
+        // if we sort the words first, keeping the LMDB pages in cache.
+        words.sort_unstable();
+
        // We iterate over the words and delete the documents ids
        // from the word docids database.
-        remove_from_word_docids(
-            self.wtxn,
-            word_docids,
-            &self.to_delete_docids,
-            &mut words_to_keep,
-            &mut words_to_delete,
-        )?;
-        remove_from_word_docids(
-            self.wtxn,
-            exact_word_docids,
-            &self.to_delete_docids,
-            &mut words_to_keep,
-            &mut words_to_delete,
-        )?;
+        for (word, must_remove) in &mut words {
+            remove_from_word_docids(
+                self.wtxn,
+                word_docids,
+                word.as_str(),
+                must_remove,
+                &self.to_delete_docids,
+            )?;
+
+            remove_from_word_docids(
+                self.wtxn,
+                exact_word_docids,
+                word.as_str(),
+                must_remove,
+                &self.to_delete_docids,
+            )?;
+        }

        // We construct an FST set that contains the words to delete from the words FST.
-        let words_to_delete = fst::Set::from_iter(words_to_delete.difference(&words_to_keep))?;
+        let words_to_delete =
+            words.iter().filter_map(
+                |(word, must_remove)| {
+                    if *must_remove {
+                        Some(word.as_str())
+                    } else {
+                        None
+                    }
+                },
+            );
+        let words_to_delete = fst::Set::from_iter(words_to_delete)?;

        let new_words_fst = {
            // We retrieve the current words FST from the database.
@@ -440,30 +467,6 @@ impl<'t, 'u, 'i> DeleteDocuments<'t, 'u, 'i> {
            &self.to_delete_docids,
        )?;

-        // An ugly and slow way to remove the vectors from the HNSW
-        // It basically reconstructs the HNSW from scratch without editing the current one.
-        let current_hnsw = self.index.vector_hnsw(self.wtxn)?.unwrap_or_default();
-        if !current_hnsw.is_empty() {
-            let mut new_hnsw = Hnsw::default();
-            let mut searcher = Searcher::new();
-            let mut new_vector_id_docids = Vec::new();
-
-            for result in vector_id_docid.iter(self.wtxn)? {
-                let (vector_id, docid) = result?;
-                if !self.to_delete_docids.contains(docid.get()) {
-                    let vector = current_hnsw.get_point(vector_id.get() as usize).clone();
-                    let vector_id = new_hnsw.insert(vector, &mut searcher);
-                    new_vector_id_docids.push((vector_id as u32, docid));
-                }
-            }
-
-            vector_id_docid.clear(self.wtxn)?;
-            for (vector_id, docid) in new_vector_id_docids {
-                vector_id_docid.put(self.wtxn, &BEU32::new(vector_id), &docid)?;
-            }
-            self.index.put_vector_hnsw(self.wtxn, &new_hnsw)?;
-        }
-
        self.index.put_soft_deleted_documents_ids(self.wtxn, &RoaringBitmap::new())?;

        Ok(DetailedDocumentDeletionResult {
@@ -529,24 +532,23 @@ fn remove_from_word_prefix_docids(
 fn remove_from_word_docids(
    txn: &mut heed::RwTxn,
    db: &heed::Database<Str, RoaringBitmapCodec>,
+    word: &str,
+    must_remove: &mut bool,
    to_remove: &RoaringBitmap,
-    words_to_keep: &mut BTreeSet<String>,
-    words_to_remove: &mut BTreeSet<String>,
 ) -> Result<()> {
    // We create an iterator to be able to get the content and delete the word docids.
    // It's faster to acquire a cursor to get and delete or put, as we avoid traversing
    // the LMDB B-Tree two times but only once.
-    let mut iter = db.iter_mut(txn)?;
-    while let Some((key, mut docids)) = iter.next().transpose()? {
-        let previous_len = docids.len();
-        docids -= to_remove;
-        if docids.is_empty() {
-            // safety: we don't keep references from inside the LMDB database.
-            unsafe { iter.del_current()? };
-            words_to_remove.insert(key.to_owned());
-        } else {
-            words_to_keep.insert(key.to_owned());
-            if docids.len() != previous_len {
+    let mut iter = db.prefix_iter_mut(txn, word)?;
+    if let Some((key, mut docids)) = iter.next().transpose()? {
+        if key == word {
+            let previous_len = docids.len();
+            docids -= to_remove;
+            if docids.is_empty() {
+                // safety: we don't keep references from inside the LMDB database.
+                unsafe { iter.del_current()? };
+                *must_remove = true;
+            } else if docids.len() != previous_len {
                let key = key.to_owned();
                // safety: we don't keep references from inside the LMDB database.
                unsafe { iter.put_current(&key, &docids)? };
@@ -625,7 +627,7 @@ mod tests {

    use super::*;
    use crate::index::tests::TempIndex;
-    use crate::{db_snap, Filter, Search};
+    use crate::{db_snap, Filter};

    fn delete_documents<'t>(
        wtxn: &mut RwTxn<'t, '_>,
@@ -1197,52 +1199,4 @@ mod tests {
            DeletionStrategy::AlwaysSoft,
        );
    }
-
-    #[test]
-    fn delete_words_exact_attributes() {
-        let index = TempIndex::new();
-
-        index
-            .update_settings(|settings| {
-                settings.set_primary_key(S("id"));
-                settings.set_searchable_fields(vec![S("text"), S("exact")]);
-                settings.set_exact_attributes(vec![S("exact")].into_iter().collect());
-            })
-            .unwrap();
-
-        index
-            .add_documents(documents!([
-                { "id": 0, "text": "hello" },
-                { "id": 1, "exact": "hello"}
-            ]))
-            .unwrap();
-        db_snap!(index, word_docids, 1, @r###"
-        hello            [0, ]
-        "###);
-        db_snap!(index, exact_word_docids, 1, @r###"
-        hello            [1, ]
-        "###);
-        db_snap!(index, words_fst, 1, @"300000000000000001084cfcfc2ce1000000016000000090ea47f");
-
-        let mut wtxn = index.write_txn().unwrap();
-        let deleted_internal_ids =
-            delete_documents(&mut wtxn, &index, &["1"], DeletionStrategy::AlwaysHard);
-        wtxn.commit().unwrap();
-
-        db_snap!(index, word_docids, 2, @r###"
-        hello            [0, ]
-        "###);
-        db_snap!(index, exact_word_docids, 2, @"");
-        db_snap!(index, words_fst, 2, @"300000000000000001084cfcfc2ce1000000016000000090ea47f");
-
-        insta::assert_snapshot!(format!("{deleted_internal_ids:?}"), @"[1]");
-        let txn = index.read_txn().unwrap();
-        let words = index.words_fst(&txn).unwrap().into_stream().into_strs().unwrap();
-        insta::assert_snapshot!(format!("{words:?}"), @r###"["hello"]"###);
-
-        let mut s = Search::new(&txn, &index);
-        s.query("hello");
-        let crate::SearchResult { documents_ids, .. } = s.execute().unwrap();
-        insta::assert_snapshot!(format!("{documents_ids:?}"), @"[0]");
-    }
 }
--- a/milli/src/update/facet/mod.rs
+++ b/milli/src/update/facet/mod.rs
@@ -261,9 +261,7 @@ pub(crate) mod test_helpers {
            let options = options.map_size(4096 * 4 * 1000 * 100);
            let tempdir = tempfile::TempDir::new().unwrap();
            let env = options.open(tempdir.path()).unwrap();
-            let mut wtxn = env.write_txn().unwrap();
-            let content = env.create_database(&mut wtxn, None).unwrap();
-            wtxn.commit().unwrap();
+            let content = env.create_database(None).unwrap();

            FacetIndex {
                content,
--- a/milli/src/update/index_documents/extract/extract_fid_word_count_docids.rs
+++ b/milli/src/update/index_documents/extract/extract_fid_word_count_docids.rs
@@ -1,6 +1,6 @@
 use std::collections::HashMap;
 use std::fs::File;
-use std::io;
+use std::{cmp, io};

 use grenad::Sorter;

@@ -54,10 +54,11 @@ pub fn extract_fid_word_count_docids<R: io::Read + io::Seek>(
        }

        for position in read_u32_ne_bytes(value) {
-            let (field_id, _) = relative_from_absolute_position(position);
+            let (field_id, position) = relative_from_absolute_position(position);
+            let word_count = position as u32 + 1;

            let value = document_fid_wordcount.entry(field_id as FieldId).or_insert(0);
-            *value += 1;
+            *value = cmp::max(*value, word_count);
        }
    }

@@ -82,7 +83,7 @@ fn drain_document_fid_wordcount_into_sorter(
    let mut key_buffer = Vec::new();

    for (fid, count) in document_fid_wordcount.drain() {
-        if count <= 30 {
+        if count <= 10 {
            key_buffer.clear();
            key_buffer.extend_from_slice(&fid.to_be_bytes());
            key_buffer.push(count as u8);
--- a/milli/src/update/index_documents/extract/extract_vector_points.rs
+++ b/milli/src/update/index_documents/extract/extract_vector_points.rs
@@ -1,69 +0,0 @@
-use std::convert::TryFrom;
-use std::fs::File;
-use std::io;
-
-use bytemuck::cast_slice;
-use serde_json::{from_slice, Value};
-
-use super::helpers::{create_writer, writer_into_reader, GrenadParameters};
-use crate::error::UserError;
-use crate::{FieldId, InternalError, Result, VectorOrArrayOfVectors};
-
-/// Extracts the embedding vector contained in each document under the `_vectors` field.
-///
-/// Returns the generated grenad reader containing the docid as key associated to the Vec<f32>
-#[logging_timer::time]
-pub fn extract_vector_points<R: io::Read + io::Seek>(
-    obkv_documents: grenad::Reader<R>,
-    indexer: GrenadParameters,
-    primary_key_id: FieldId,
-    vectors_fid: FieldId,
-) -> Result<grenad::Reader<File>> {
-    let mut writer = create_writer(
-        indexer.chunk_compression_type,
-        indexer.chunk_compression_level,
-        tempfile::tempfile()?,
-    );
-
-    let mut cursor = obkv_documents.into_cursor()?;
-    while let Some((docid_bytes, value)) = cursor.move_on_next()? {
-        let obkv = obkv::KvReader::new(value);
-
-        // since we only needs the primary key when we throw an error we create this getter to
-        // lazily get it when needed
-        let document_id = || -> Value {
-            let document_id = obkv.get(primary_key_id).unwrap();
-            serde_json::from_slice(document_id).unwrap()
-        };
-
-        // first we retrieve the _vectors field
-        if let Some(vectors) = obkv.get(vectors_fid) {
-            // extract the vectors
-            let vectors = match from_slice(vectors) {
-                Ok(vectors) => VectorOrArrayOfVectors::into_array_of_vectors(vectors),
-                Err(_) => {
-                    return Err(UserError::InvalidVectorsType {
-                        document_id: document_id(),
-                        value: from_slice(vectors).map_err(InternalError::SerdeJson)?,
-                    }
-                    .into())
-                }
-            };
-
-            for (i, vector) in vectors.into_iter().enumerate() {
-                match u16::try_from(i) {
-                    Ok(i) => {
-                        let mut key = docid_bytes.to_vec();
-                        key.extend_from_slice(&i.to_ne_bytes());
-                        let bytes = cast_slice(&vector);
-                        writer.insert(key, bytes)?;
-                    }
-                    Err(_) => continue,
-                }
-            }
-        }
-        // else => the `_vectors` object was `null`, there is nothing to do
-    }
-
-    writer_into_reader(writer)
-}
--- a/milli/src/update/index_documents/extract/mod.rs
+++ b/milli/src/update/index_documents/extract/mod.rs
@@ -4,7 +4,6 @@ mod extract_facet_string_docids;
 mod extract_fid_docid_facet_values;
 mod extract_fid_word_count_docids;
 mod extract_geo_points;
-mod extract_vector_points;
 mod extract_word_docids;
 mod extract_word_fid_docids;
 mod extract_word_pair_proximity_docids;
@@ -23,7 +22,6 @@ use self::extract_facet_string_docids::extract_facet_string_docids;
 use self::extract_fid_docid_facet_values::{extract_fid_docid_facet_values, ExtractedFacetValues};
 use self::extract_fid_word_count_docids::extract_fid_word_count_docids;
 use self::extract_geo_points::extract_geo_points;
-use self::extract_vector_points::extract_vector_points;
 use self::extract_word_docids::extract_word_docids;
 use self::extract_word_fid_docids::extract_word_fid_docids;
 use self::extract_word_pair_proximity_docids::extract_word_pair_proximity_docids;
@@ -47,7 +45,6 @@ pub(crate) fn data_from_obkv_documents(
    faceted_fields: HashSet<FieldId>,
    primary_key_id: FieldId,
    geo_fields_ids: Option<(FieldId, FieldId)>,
-    vectors_field_id: Option<FieldId>,
    stop_words: Option<fst::Set<&[u8]>>,
    max_positions_per_attributes: Option<u32>,
    exact_attributes: HashSet<FieldId>,
@@ -72,7 +69,6 @@ pub(crate) fn data_from_obkv_documents(
                    &faceted_fields,
                    primary_key_id,
                    geo_fields_ids,
-                    vectors_field_id,
                    &stop_words,
                    max_positions_per_attributes,
                )
@@ -283,7 +279,6 @@ fn send_and_extract_flattened_documents_data(
    faceted_fields: &HashSet<FieldId>,
    primary_key_id: FieldId,
    geo_fields_ids: Option<(FieldId, FieldId)>,
-    vectors_field_id: Option<FieldId>,
    stop_words: &Option<fst::Set<&[u8]>>,
    max_positions_per_attributes: Option<u32>,
 ) -> Result<(
@@ -312,25 +307,6 @@ fn send_and_extract_flattened_documents_data(
        });
    }

-    if let Some(vectors_field_id) = vectors_field_id {
-        let documents_chunk_cloned = flattened_documents_chunk.clone();
-        let lmdb_writer_sx_cloned = lmdb_writer_sx.clone();
-        rayon::spawn(move || {
-            let result = extract_vector_points(
-                documents_chunk_cloned,
-                indexer,
-                primary_key_id,
-                vectors_field_id,
-            );
-            let _ = match result {
-                Ok(vector_points) => {
-                    lmdb_writer_sx_cloned.send(Ok(TypedChunk::VectorPoints(vector_points)))
-                }
-                Err(error) => lmdb_writer_sx_cloned.send(Err(error)),
-            };
-        });
-    }
-
    let (docid_word_positions_chunk, docid_fid_facet_values_chunks): (Result<_>, Result<_>) =
        rayon::join(
            || {
@@ -349,6 +325,8 @@ fn send_and_extract_flattened_documents_data(
                // send docid_word_positions_chunk to DB writer
                let docid_word_positions_chunk =
                    unsafe { as_cloneable_grenad(&docid_word_positions_chunk)? };
+                let _ = lmdb_writer_sx
+                    .send(Ok(TypedChunk::DocidWordPositions(docid_word_positions_chunk.clone())));

                let _ =
                    lmdb_writer_sx.send(Ok(TypedChunk::ScriptLanguageDocids(script_language_pair)));
--- a/milli/src/update/index_documents/helpers/merge_functions.rs
+++ b/milli/src/update/index_documents/helpers/merge_functions.rs
@@ -4,6 +4,7 @@ use std::result::Result as StdResult;

 use roaring::RoaringBitmap;

+use super::read_u32_ne_bytes;
 use crate::heed_codec::CboRoaringBitmapCodec;
 use crate::update::index_documents::transform::Operation;
 use crate::Result;
@@ -21,6 +22,10 @@ pub fn concat_u32s_array<'a>(_key: &[u8], values: &[Cow<'a, [u8]>]) -> Result<Co
    }
 }

+pub fn roaring_bitmap_from_u32s_array(slice: &[u8]) -> RoaringBitmap {
+    read_u32_ne_bytes(slice).collect()
+}
+
 pub fn serialize_roaring_bitmap(bitmap: &RoaringBitmap, buffer: &mut Vec<u8>) -> io::Result<()> {
    buffer.clear();
    buffer.reserve(bitmap.serialized_size());
--- a/milli/src/update/index_documents/helpers/mod.rs
+++ b/milli/src/update/index_documents/helpers/mod.rs
@@ -14,8 +14,8 @@ pub use grenad_helpers::{
 };
 pub use merge_functions::{
    concat_u32s_array, keep_first, keep_latest_obkv, merge_cbo_roaring_bitmaps,
-    merge_obkvs_and_operations, merge_roaring_bitmaps, merge_two_obkvs, serialize_roaring_bitmap,
-    MergeFn,
+    merge_obkvs_and_operations, merge_roaring_bitmaps, merge_two_obkvs,
+    roaring_bitmap_from_u32s_array, serialize_roaring_bitmap, MergeFn,
 };

 use crate::MAX_WORD_LENGTH;
--- a/milli/src/update/index_documents/mod.rs
+++ b/milli/src/update/index_documents/mod.rs
@@ -304,8 +304,6 @@ where
            }
            None => None,
        };
-        // get the fid of the `_vectors` field.
-        let vectors_field_id = self.index.fields_ids_map(self.wtxn)?.id("_vectors");

        let stop_words = self.index.stop_words(self.wtxn)?;
        let exact_attributes = self.index.exact_attributes_ids(self.wtxn)?;
@@ -342,7 +340,6 @@ where
                    faceted_fields,
                    primary_key_id,
                    geo_fields_ids,
-                    vectors_field_id,
                    stop_words,
                    max_positions_per_attributes,
                    exact_attributes,
@@ -1584,7 +1581,7 @@ mod tests {
        assert_eq!(count, 4);
    }

-    #[cfg(feature = "chinese")]
+    #[cfg(feature = "default")]
    #[test]
    fn test_meilisearch_1714() {
        let index = TempIndex::new();
@@ -2474,11 +2471,11 @@ mod tests {
              {
                "id": 3,
                "text": "a a a a a a a a a a a a a a a a a
-                a a a a a a a a a a a a a a a a a a a a a a a a a a
-                a a a a a a a a a a a a a a a a a a a a a a a a a a
-                a a a a a a a a a a a a a a a a a a a a a a a a a a
-                a a a a a a a a a a a a a a a a a a a a a a a a a a
-                a a a a a a a a a a a a a a a a a a a a a a a a a a
+                a a a a a a a a a a a a a a a a a a a a a a a a a a 
+                a a a a a a a a a a a a a a a a a a a a a a a a a a 
+                a a a a a a a a a a a a a a a a a a a a a a a a a a 
+                a a a a a a a a a a a a a a a a a a a a a a a a a a 
+                a a a a a a a a a a a a a a a a a a a a a a a a a a 
                a a a a a a a a a a a a a a a a a a a a a "
             }
            ]))
@@ -2516,5 +2513,6 @@ mod tests {

        db_snap!(index, word_fid_docids, 3, @"4c2e2a1832e5802796edc1638136d933");
        db_snap!(index, word_position_docids, 3, @"74f556b91d161d997a89468b4da1cb8f");
+        db_snap!(index, docid_word_positions, 3, @"5287245332627675740b28bd46e1cde1");
    }
 }
--- a/milli/src/update/index_documents/typed_chunk.rs
+++ b/milli/src/update/index_documents/typed_chunk.rs
@@ -4,29 +4,27 @@ use std::convert::TryInto;
 use std::fs::File;
 use std::io;

-use bytemuck::allocation::pod_collect_to_vec;
 use charabia::{Language, Script};
 use grenad::MergerBuilder;
 use heed::types::ByteSlice;
-use heed::RwTxn;
-use hnsw::Searcher;
+use heed::{BytesDecode, RwTxn};
 use roaring::RoaringBitmap;
-use space::KnnPoints;

 use super::helpers::{
-    self, merge_ignore_values, serialize_roaring_bitmap, valid_lmdb_key, CursorClonableMmap,
+    self, merge_ignore_values, roaring_bitmap_from_u32s_array, serialize_roaring_bitmap,
+    valid_lmdb_key, CursorClonableMmap,
 };
 use super::{ClonableMmap, MergeFn};
-use crate::error::UserError;
 use crate::facet::FacetType;
 use crate::update::facet::FacetsUpdate;
-use crate::update::index_documents::helpers::{as_cloneable_grenad, try_split_array_at};
+use crate::update::index_documents::helpers::as_cloneable_grenad;
 use crate::{
-    lat_lng_to_xyz, normalize_vector, CboRoaringBitmapCodec, DocumentId, GeoPoint, Index, Result,
-    BEU32,
+    lat_lng_to_xyz, BoRoaringBitmapCodec, CboRoaringBitmapCodec, DocumentId, GeoPoint, Index,
+    Result,
 };

 pub(crate) enum TypedChunk {
+    DocidWordPositions(grenad::Reader<CursorClonableMmap>),
    FieldIdDocidFacetStrings(grenad::Reader<CursorClonableMmap>),
    FieldIdDocidFacetNumbers(grenad::Reader<CursorClonableMmap>),
    Documents(grenad::Reader<CursorClonableMmap>),
@@ -45,7 +43,6 @@ pub(crate) enum TypedChunk {
    FieldIdFacetIsNullDocids(grenad::Reader<File>),
    FieldIdFacetIsEmptyDocids(grenad::Reader<File>),
    GeoPoints(grenad::Reader<File>),
-    VectorPoints(grenad::Reader<File>),
    ScriptLanguageDocids(HashMap<(Script, Language), RoaringBitmap>),
 }

@@ -59,6 +56,29 @@ pub(crate) fn write_typed_chunk_into_index(
 ) -> Result<(RoaringBitmap, bool)> {
    let mut is_merged_database = false;
    match typed_chunk {
+        TypedChunk::DocidWordPositions(docid_word_positions_iter) => {
+            write_entries_into_database(
+                docid_word_positions_iter,
+                &index.docid_word_positions,
+                wtxn,
+                index_is_empty,
+                |value, buffer| {
+                    // ensure that values are unique and ordered
+                    let positions = roaring_bitmap_from_u32s_array(value);
+                    BoRoaringBitmapCodec::serialize_into(&positions, buffer);
+                    Ok(buffer)
+                },
+                |new_values, db_values, buffer| {
+                    let new_values = roaring_bitmap_from_u32s_array(new_values);
+                    let positions = match BoRoaringBitmapCodec::bytes_decode(db_values) {
+                        Some(db_values) => new_values | db_values,
+                        None => new_values, // should not happen
+                    };
+                    BoRoaringBitmapCodec::serialize_into(&positions, buffer);
+                    Ok(())
+                },
+            )?;
+        }
        TypedChunk::Documents(obkv_documents_iter) => {
            let mut cursor = obkv_documents_iter.into_cursor()?;
            while let Some((key, value)) = cursor.move_on_next()? {
@@ -229,40 +249,6 @@ pub(crate) fn write_typed_chunk_into_index(
            index.put_geo_rtree(wtxn, &rtree)?;
            index.put_geo_faceted_documents_ids(wtxn, &geo_faceted_docids)?;
        }
-        TypedChunk::VectorPoints(vector_points) => {
-            let mut hnsw = index.vector_hnsw(wtxn)?.unwrap_or_default();
-            let mut searcher = Searcher::new();
-
-            let mut expected_dimensions = match index.vector_id_docid.iter(wtxn)?.next() {
-                Some(result) => {
-                    let (vector_id, _) = result?;
-                    Some(hnsw.get_point(vector_id.get() as usize).len())
-                }
-                None => None,
-            };
-
-            let mut cursor = vector_points.into_cursor()?;
-            while let Some((key, value)) = cursor.move_on_next()? {
-                // convert the key back to a u32 (4 bytes)
-                let (left, _index) = try_split_array_at(key).unwrap();
-                let docid = DocumentId::from_be_bytes(left);
-                // convert the vector back to a Vec<f32>
-                let vector: Vec<f32> = pod_collect_to_vec(value);
-
-                // TODO Move this error in the vector extractor
-                let found = vector.len();
-                let expected = *expected_dimensions.get_or_insert(found);
-                if expected != found {
-                    return Err(UserError::InvalidVectorDimensions { expected, found })?;
-                }
-
-                let vector = normalize_vector(vector);
-                let vector_id = hnsw.insert(vector, &mut searcher) as u32;
-                index.vector_id_docid.put(wtxn, &BEU32::new(vector_id), &BEU32::new(docid))?;
-            }
-            log::debug!("There are {} entries in the HNSW so far", hnsw.len());
-            index.put_vector_hnsw(wtxn, &hnsw)?;
-        }
        TypedChunk::ScriptLanguageDocids(hash_pair) => {
            let mut buffer = Vec::new();
            for (key, value) in hash_pair {