chore: feature gate vector_index (#7428 )

Signed-off-by: discord9 <discord9@163.com>
fix: using anonymous s3 access when ak and sk is not provided (#7425 )
2025-12-22 22:20:02 +00:00 · 2025-12-17 07:14:25 +00:00 · 2025-12-17 06:34:29 +00:00 · 2025-12-17 04:18:23 +00:00 · 2025-12-17 03:14:35 +00:00 · 2025-12-17 01:29:36 +00:00
23 changed files with 1228 additions and 657 deletions
--- a/.github/actions/setup-greptimedb-cluster/action.yml
+++ b/.github/actions/setup-greptimedb-cluster/action.yml
@@ -51,7 +51,7 @@ runs:
    run: |
      helm upgrade \
        --install my-greptimedb \
-        --set meta.backendStorage.etcd.endpoints=${{ inputs.etcd-endpoints }} \
+        --set 'meta.backendStorage.etcd.endpoints[0]=${{ inputs.etcd-endpoints }}' \
        --set meta.enableRegionFailover=${{ inputs.enable-region-failover }} \
        --set image.registry=${{ inputs.image-registry }} \
        --set image.repository=${{ inputs.image-repository }}  \
--- a/.github/scripts/deploy-greptimedb.sh
+++ b/.github/scripts/deploy-greptimedb.sh
@@ -81,7 +81,7 @@ function deploy_greptimedb_cluster() {
    --create-namespace \
    --set image.tag="$GREPTIMEDB_IMAGE_TAG" \
    --set initializer.tag="$GREPTIMEDB_INITIALIZER_IMAGE_TAG" \
-    --set meta.backendStorage.etcd.endpoints="etcd.$install_namespace:2379" \
+    --set "meta.backendStorage.etcd.endpoints[0]=etcd.$install_namespace.svc.cluster.local:2379" \
    --set meta.backendStorage.etcd.storeKeyPrefix="$cluster_name" \
    -n "$install_namespace"

@@ -119,7 +119,7 @@ function deploy_greptimedb_cluster_with_s3_storage() {
    --create-namespace \
    --set image.tag="$GREPTIMEDB_IMAGE_TAG" \
    --set initializer.tag="$GREPTIMEDB_INITIALIZER_IMAGE_TAG" \
-    --set meta.backendStorage.etcd.endpoints="etcd.$install_namespace:2379" \
+    --set "meta.backendStorage.etcd.endpoints[0]=etcd.$install_namespace.svc.cluster.local:2379" \
    --set meta.backendStorage.etcd.storeKeyPrefix="$cluster_name" \
    --set objectStorage.s3.bucket="$AWS_CI_TEST_BUCKET" \
    --set objectStorage.s3.region="$AWS_REGION" \
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -212,7 +212,7 @@ checksum = "d301b3b94cb4b2f23d7917810addbbaff90738e0ca2be692bd027e70d7e0330c"

 [[package]]
 name = "api"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "arrow-schema",
 "common-base",
@@ -733,7 +733,7 @@ dependencies = [

 [[package]]
 name = "auth"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -1383,7 +1383,7 @@ dependencies = [

 [[package]]
 name = "cache"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "catalog",
 "common-error",
@@ -1418,7 +1418,7 @@ checksum = "37b2a672a2cb129a2e41c10b1224bb368f9f37a2b16b612598138befd7b37eb5"

 [[package]]
 name = "catalog"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "arrow",
@@ -1763,7 +1763,7 @@ checksum = "b94f61472cee1439c0b966b47e3aca9ae07e45d070759512cd390ea2bebc6675"

 [[package]]
 name = "cli"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-stream",
 "async-trait",
@@ -1816,7 +1816,7 @@ dependencies = [

 [[package]]
 name = "client"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "arc-swap",
@@ -1849,7 +1849,7 @@ dependencies = [
 "snafu 0.8.6",
 "store-api",
 "substrait 0.37.3",
- "substrait 1.0.0-beta.2",
+ "substrait 1.0.0-beta.3",
 "tokio",
 "tokio-stream",
 "tonic 0.13.1",
@@ -1889,7 +1889,7 @@ dependencies = [

 [[package]]
 name = "cmd"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-trait",
 "auth",
@@ -1977,6 +1977,17 @@ dependencies = [
 "unicode-width 0.2.1",
 ]

+[[package]]
+name = "codespan-reporting"
+version = "0.13.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "af491d569909a7e4dee0ad7db7f5341fef5c614d5b8ec8cf765732aba3cff681"
+dependencies = [
+ "serde",
+ "termcolor",
+ "unicode-width 0.2.1",
+]
+
 [[package]]
 name = "colorchoice"
 version = "1.0.4"
@@ -2012,7 +2023,7 @@ checksum = "55b672471b4e9f9e95499ea597ff64941a309b2cdbffcc46f2cc5e2d971fd335"

 [[package]]
 name = "common-base"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "anymap2",
 "async-trait",
@@ -2036,14 +2047,14 @@ dependencies = [

 [[package]]
 name = "common-catalog"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "const_format",
 ]

 [[package]]
 name = "common-config"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-base",
 "common-error",
@@ -2068,7 +2079,7 @@ dependencies = [

 [[package]]
 name = "common-datasource"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "arrow",
 "arrow-schema",
@@ -2103,7 +2114,7 @@ dependencies = [

 [[package]]
 name = "common-decimal"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "bigdecimal 0.4.8",
 "common-error",
@@ -2116,7 +2127,7 @@ dependencies = [

 [[package]]
 name = "common-error"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-macro",
 "http 1.3.1",
@@ -2127,7 +2138,7 @@ dependencies = [

 [[package]]
 name = "common-event-recorder"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -2149,7 +2160,7 @@ dependencies = [

 [[package]]
 name = "common-frontend"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -2171,7 +2182,7 @@ dependencies = [

 [[package]]
 name = "common-function"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "ahash 0.8.12",
 "api",
@@ -2231,7 +2242,7 @@ dependencies = [

 [[package]]
 name = "common-greptimedb-telemetry"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-trait",
 "common-runtime",
@@ -2248,7 +2259,7 @@ dependencies = [

 [[package]]
 name = "common-grpc"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "arrow-flight",
@@ -2283,7 +2294,7 @@ dependencies = [

 [[package]]
 name = "common-grpc-expr"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "common-base",
@@ -2303,7 +2314,7 @@ dependencies = [

 [[package]]
 name = "common-macro"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "greptime-proto",
 "once_cell",
@@ -2314,7 +2325,7 @@ dependencies = [

 [[package]]
 name = "common-mem-prof"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "anyhow",
 "common-error",
@@ -2330,7 +2341,7 @@ dependencies = [

 [[package]]
 name = "common-memory-manager"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-error",
 "common-macro",
@@ -2343,7 +2354,7 @@ dependencies = [

 [[package]]
 name = "common-meta"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "anymap2",
 "api",
@@ -2415,7 +2426,7 @@ dependencies = [

 [[package]]
 name = "common-options"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-grpc",
 "humantime-serde",
@@ -2424,11 +2435,11 @@ dependencies = [

 [[package]]
 name = "common-plugins"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"

 [[package]]
 name = "common-pprof"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-error",
 "common-macro",
@@ -2440,7 +2451,7 @@ dependencies = [

 [[package]]
 name = "common-procedure"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-stream",
@@ -2469,7 +2480,7 @@ dependencies = [

 [[package]]
 name = "common-procedure-test"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-trait",
 "common-procedure",
@@ -2479,7 +2490,7 @@ dependencies = [

 [[package]]
 name = "common-query"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -2505,7 +2516,7 @@ dependencies = [

 [[package]]
 name = "common-recordbatch"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "arc-swap",
 "common-base",
@@ -2529,7 +2540,7 @@ dependencies = [

 [[package]]
 name = "common-runtime"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-trait",
 "clap 4.5.40",
@@ -2558,7 +2569,7 @@ dependencies = [

 [[package]]
 name = "common-session"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "serde",
 "strum 0.27.1",
@@ -2566,7 +2577,7 @@ dependencies = [

 [[package]]
 name = "common-sql"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-base",
 "common-decimal",
@@ -2584,7 +2595,7 @@ dependencies = [

 [[package]]
 name = "common-stat"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-base",
 "common-runtime",
@@ -2599,7 +2610,7 @@ dependencies = [

 [[package]]
 name = "common-telemetry"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "backtrace",
 "common-base",
@@ -2628,7 +2639,7 @@ dependencies = [

 [[package]]
 name = "common-test-util"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "client",
 "common-grpc",
@@ -2641,7 +2652,7 @@ dependencies = [

 [[package]]
 name = "common-time"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "arrow",
 "chrono",
@@ -2659,7 +2670,7 @@ dependencies = [

 [[package]]
 name = "common-version"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "build-data",
 "cargo-manifest",
@@ -2670,7 +2681,7 @@ dependencies = [

 [[package]]
 name = "common-wal"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-base",
 "common-error",
@@ -2693,7 +2704,7 @@ dependencies = [

 [[package]]
 name = "common-workload"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "common-telemetry",
 "serde",
@@ -3169,6 +3180,68 @@ dependencies = [
 "cipher",
 ]

+[[package]]
+name = "cxx"
+version = "1.0.190"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a7620f6cfc4dcca21f2b085b7a890e16c60fd66f560cd69ee60594908dc72ab1"
+dependencies = [
+ "cc",
+ "cxx-build",
+ "cxxbridge-cmd",
+ "cxxbridge-flags",
+ "cxxbridge-macro",
+ "foldhash 0.2.0",
+ "link-cplusplus",
+]
+
+[[package]]
+name = "cxx-build"
+version = "1.0.190"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7a9bc1a22964ff6a355fbec24cf68266a0ed28f8b84c0864c386474ea3d0e479"
+dependencies = [
+ "cc",
+ "codespan-reporting 0.13.1",
+ "indexmap 2.11.4",
+ "proc-macro2",
+ "quote",
+ "scratch",
+ "syn 2.0.106",
+]
+
+[[package]]
+name = "cxxbridge-cmd"
+version = "1.0.190"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b1f29a879d35f7906e3c9b77d7a1005a6a0787d330c09dfe4ffb5f617728cb44"
+dependencies = [
+ "clap 4.5.40",
+ "codespan-reporting 0.13.1",
+ "indexmap 2.11.4",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.106",
+]
+
+[[package]]
+name = "cxxbridge-flags"
+version = "1.0.190"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d67109015f93f683e364085aa6489a5b2118b4a40058482101d699936a7836d6"
+
+[[package]]
+name = "cxxbridge-macro"
+version = "1.0.190"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d187e019e7b05a1f3e69a8396b70800ee867aa9fc2ab972761173ccee03742df"
+dependencies = [
+ "indexmap 2.11.4",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.106",
+]
+
 [[package]]
 name = "darling"
 version = "0.14.4"
@@ -3939,7 +4012,7 @@ dependencies = [

 [[package]]
 name = "datanode"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "arrow-flight",
@@ -4003,7 +4076,7 @@ dependencies = [

 [[package]]
 name = "datatypes"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "arrow",
 "arrow-array",
@@ -4677,7 +4750,7 @@ checksum = "37909eebbb50d72f9059c3b6d82c0463f2ff062c9e95845c43a6c9c0355411be"

 [[package]]
 name = "file-engine"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -4809,7 +4882,7 @@ checksum = "8bf7cc16383c4b8d58b9905a8509f02926ce3058053c056376248d958c9df1e8"

 [[package]]
 name = "flow"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "arrow",
@@ -4878,7 +4951,7 @@ dependencies = [
 "sql",
 "store-api",
 "strum 0.27.1",
- "substrait 1.0.0-beta.2",
+ "substrait 1.0.0-beta.3",
 "table",
 "tokio",
 "tonic 0.13.1",
@@ -4916,6 +4989,12 @@ version = "0.1.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d9c4f5dac5e15c24eb999c26181a6ca40b39fe946cbe4c263c7209467bc83af2"

+[[package]]
+name = "foldhash"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "77ce24cb58228fbb8aa041425bb1050850ac19177686ea6e0f41a70416f56fdb"
+
 [[package]]
 name = "form_urlencoded"
 version = "1.2.2"
@@ -4933,7 +5012,7 @@ checksum = "28dd6caf6059519a65843af8fe2a3ae298b14b80179855aeb4adc2c1934ee619"

 [[package]]
 name = "frontend"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "arc-swap",
@@ -5516,7 +5595,7 @@ checksum = "5971ac85611da7067dbfcabef3c70ebb5606018acd9e2a3903a0da507521e0d5"
 dependencies = [
 "allocator-api2",
 "equivalent",
- "foldhash",
+ "foldhash 0.1.5",
 ]

 [[package]]
@@ -6148,7 +6227,7 @@ dependencies = [

 [[package]]
 name = "index"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-trait",
 "asynchronous-codec",
@@ -6161,6 +6240,7 @@ dependencies = [
 "common-telemetry",
 "common-test-util",
 "criterion 0.4.0",
+ "datatypes",
 "fastbloom",
 "fst",
 "futures",
@@ -6169,6 +6249,7 @@ dependencies = [
 "jieba-rs",
 "lazy_static",
 "mockall",
+ "nalgebra",
 "pin-project",
 "prost 0.13.5",
 "puffin",
@@ -6186,6 +6267,7 @@ dependencies = [
 "tempfile",
 "tokio",
 "tokio-util",
+ "usearch",
 "uuid",
 ]

@@ -7017,6 +7099,15 @@ dependencies = [
 "vcpkg",
 ]

+[[package]]
+name = "link-cplusplus"
+version = "1.0.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7f78c730aaa7d0b9336a299029ea49f9ee53b0ed06e9202e8cb7db9bae7b8c82"
+dependencies = [
+ "cc",
+]
+
 [[package]]
 name = "linked-hash-map"
 version = "0.5.6"
@@ -7077,7 +7168,7 @@ checksum = "13dc2df351e3202783a1fe0d44375f7295ffb4049267b0f3018346dc122a1d94"

 [[package]]
 name = "log-query"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "chrono",
 "common-error",
@@ -7089,7 +7180,7 @@ dependencies = [

 [[package]]
 name = "log-store"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-stream",
 "async-trait",
@@ -7390,7 +7481,7 @@ dependencies = [

 [[package]]
 name = "meta-client"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -7418,7 +7509,7 @@ dependencies = [

 [[package]]
 name = "meta-srv"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -7518,7 +7609,7 @@ dependencies = [

 [[package]]
 name = "metric-engine"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "aquamarine",
@@ -7615,7 +7706,7 @@ dependencies = [

 [[package]]
 name = "mito-codec"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "bytes",
@@ -7640,7 +7731,7 @@ dependencies = [

 [[package]]
 name = "mito2"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "aquamarine",
@@ -8380,7 +8471,7 @@ dependencies = [

 [[package]]
 name = "object-store"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "anyhow",
 "bytes",
@@ -8665,7 +8756,7 @@ dependencies = [

 [[package]]
 name = "operator"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "ahash 0.8.12",
 "api",
@@ -8725,7 +8816,7 @@ dependencies = [
 "sql",
 "sqlparser",
 "store-api",
- "substrait 1.0.0-beta.2",
+ "substrait 1.0.0-beta.3",
 "table",
 "tokio",
 "tokio-util",
@@ -9011,7 +9102,7 @@ dependencies = [

 [[package]]
 name = "partition"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -9368,7 +9459,7 @@ checksum = "8b870d8c151b6f2fb93e84a13146138f05d02ed11c7e7c54f8826aaaf7c9f184"

 [[package]]
 name = "pipeline"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "ahash 0.8.12",
 "api",
@@ -9524,7 +9615,7 @@ dependencies = [

 [[package]]
 name = "plugins"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "auth",
 "catalog",
@@ -9826,7 +9917,7 @@ dependencies = [

 [[package]]
 name = "promql"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "ahash 0.8.12",
 "async-trait",
@@ -10109,7 +10200,7 @@ dependencies = [

 [[package]]
 name = "puffin"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-compression 0.4.19",
 "async-trait",
@@ -10151,7 +10242,7 @@ dependencies = [

 [[package]]
 name = "query"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "ahash 0.8.12",
 "api",
@@ -10218,7 +10309,7 @@ dependencies = [
 "sql",
 "sqlparser",
 "store-api",
- "substrait 1.0.0-beta.2",
+ "substrait 1.0.0-beta.3",
 "table",
 "tokio",
 "tokio-stream",
@@ -11290,6 +11381,12 @@ version = "1.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "94143f37725109f92c262ed2cf5e59bce7498c01bcc1502d7b9afe439a4e9f49"

+[[package]]
+name = "scratch"
+version = "1.0.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d68f2ec51b097e4c1a75b681a8bec621909b5e91f15bb7b840c4f2f7b01148b2"
+
 [[package]]
 name = "scrypt"
 version = "0.11.0"
@@ -11554,7 +11651,7 @@ dependencies = [

 [[package]]
 name = "servers"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "ahash 0.8.12",
 "api",
@@ -11682,7 +11779,7 @@ dependencies = [

 [[package]]
 name = "session"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "ahash 0.8.12",
 "api",
@@ -12016,7 +12113,7 @@ dependencies = [

 [[package]]
 name = "sql"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "arrow-buffer",
@@ -12076,7 +12173,7 @@ dependencies = [

 [[package]]
 name = "sqlness-runner"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-trait",
 "clap 4.5.40",
@@ -12353,7 +12450,7 @@ dependencies = [

 [[package]]
 name = "standalone"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-trait",
 "catalog",
@@ -12394,7 +12491,7 @@ checksum = "a2eb9349b6444b326872e140eb1cf5e7c522154d69e7a0ffb0fb81c06b37543f"

 [[package]]
 name = "store-api"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "aquamarine",
@@ -12607,7 +12704,7 @@ dependencies = [

 [[package]]
 name = "substrait"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "async-trait",
 "bytes",
@@ -12730,7 +12827,7 @@ dependencies = [

 [[package]]
 name = "table"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "async-trait",
@@ -12999,7 +13096,7 @@ checksum = "8f50febec83f5ee1df3015341d8bd429f2d1cc62bcba7ea2076759d315084683"

 [[package]]
 name = "tests-fuzz"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "arbitrary",
 "async-trait",
@@ -13043,7 +13140,7 @@ dependencies = [

 [[package]]
 name = "tests-integration"
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 dependencies = [
 "api",
 "arrow-flight",
@@ -13118,7 +13215,7 @@ dependencies = [
 "sqlx",
 "standalone",
 "store-api",
- "substrait 1.0.0-beta.2",
+ "substrait 1.0.0-beta.3",
 "table",
 "tempfile",
 "time",
@@ -14143,6 +14240,16 @@ version = "2.1.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "daf8dba3b7eb870caf1ddeed7bc9d2a049f3cfdfae7cb521b087cc33ae4c49da"

+[[package]]
+name = "usearch"
+version = "2.21.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2cc9fc5f872a3a4f9081d5f42624d788231b763e1846c829b9968a3755ac884d"
+dependencies = [
+ "cxx",
+ "cxx-build",
+]
+
 [[package]]
 name = "utf8-ranges"
 version = "1.0.5"
@@ -14282,7 +14389,7 @@ dependencies = [
 "ciborium",
 "cidr",
 "clap 4.5.40",
- "codespan-reporting",
+ "codespan-reporting 0.12.0",
 "community-id",
 "convert_case 0.7.1",
 "crc",
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -75,7 +75,7 @@ members = [
 resolver = "2"

 [workspace.package]
-version = "1.0.0-beta.2"
+version = "1.0.0-beta.3"
 edition = "2024"
 license = "Apache-2.0"

--- a/src/catalog/src/system_schema/information_schema.rs
+++ b/src/catalog/src/system_schema/information_schema.rs
@@ -428,7 +428,7 @@ pub trait InformationExtension {
 }

 /// The request to inspect the datanode.
-#[derive(Debug, Clone, PartialEq, Eq)]
+#[derive(Debug, Clone, PartialEq)]
 pub struct DatanodeInspectRequest {
    /// Kind to fetch from datanode.
    pub kind: DatanodeInspectKind,
--- a/src/datatypes/src/schema.rs
+++ b/src/datatypes/src/schema.rs
@@ -33,7 +33,8 @@ pub use crate::schema::column_schema::{
    COLUMN_SKIPPING_INDEX_OPT_KEY_FALSE_POSITIVE_RATE, COLUMN_SKIPPING_INDEX_OPT_KEY_GRANULARITY,
    COLUMN_SKIPPING_INDEX_OPT_KEY_TYPE, COMMENT_KEY, ColumnExtType, ColumnSchema, FULLTEXT_KEY,
    FulltextAnalyzer, FulltextBackend, FulltextOptions, INVERTED_INDEX_KEY, Metadata,
-    SKIPPING_INDEX_KEY, SkippingIndexOptions, SkippingIndexType, TIME_INDEX_KEY,
+    SKIPPING_INDEX_KEY, SkippingIndexOptions, SkippingIndexType, TIME_INDEX_KEY, VECTOR_INDEX_KEY,
+    VectorDistanceMetric, VectorIndexEngineType, VectorIndexOptions,
 };
 pub use crate::schema::constraint::ColumnDefaultConstraint;
 pub use crate::schema::raw::RawSchema;
--- a/src/datatypes/src/schema/column_schema.rs
+++ b/src/datatypes/src/schema/column_schema.rs
@@ -46,6 +46,8 @@ pub const FULLTEXT_KEY: &str = "greptime:fulltext";
 pub const INVERTED_INDEX_KEY: &str = "greptime:inverted_index";
 /// Key used to store skip options in arrow field's metadata.
 pub const SKIPPING_INDEX_KEY: &str = "greptime:skipping_index";
+/// Key used to store vector index options in arrow field's metadata.
+pub const VECTOR_INDEX_KEY: &str = "greptime:vector_index";

 /// Keys used in fulltext options
 pub const COLUMN_FULLTEXT_CHANGE_OPT_KEY_ENABLE: &str = "enable";
@@ -216,6 +218,53 @@ impl ColumnSchema {
        self.metadata.contains_key(INVERTED_INDEX_KEY)
    }

+    /// Checks if this column has a vector index.
+    pub fn is_vector_indexed(&self) -> bool {
+        match self.vector_index_options() {
+            Ok(opts) => opts.is_some(),
+            Err(e) => {
+                common_telemetry::warn!(
+                    "Failed to deserialize vector_index_options for column '{}': {}",
+                    self.name,
+                    e
+                );
+                false
+            }
+        }
+    }
+
+    /// Gets the vector index options.
+    pub fn vector_index_options(&self) -> Result<Option<VectorIndexOptions>> {
+        match self.metadata.get(VECTOR_INDEX_KEY) {
+            None => Ok(None),
+            Some(json) => {
+                let options =
+                    serde_json::from_str(json).context(error::DeserializeSnafu { json })?;
+                Ok(Some(options))
+            }
+        }
+    }
+
+    /// Sets the vector index options.
+    pub fn set_vector_index_options(&mut self, options: &VectorIndexOptions) -> Result<()> {
+        self.metadata.insert(
+            VECTOR_INDEX_KEY.to_string(),
+            serde_json::to_string(options).context(error::SerializeSnafu)?,
+        );
+        Ok(())
+    }
+
+    /// Removes the vector index options.
+    pub fn unset_vector_index_options(&mut self) {
+        self.metadata.remove(VECTOR_INDEX_KEY);
+    }
+
+    /// Sets vector index options and returns self for chaining.
+    pub fn with_vector_index_options(mut self, options: &VectorIndexOptions) -> Result<Self> {
+        self.set_vector_index_options(options)?;
+        Ok(self)
+    }
+
    /// Set default constraint.
    ///
    /// If a default constraint exists for the column, this method will
@@ -964,6 +1013,181 @@ impl TryFrom<HashMap<String, String>> for SkippingIndexOptions {
    }
 }

+/// Distance metric for vector similarity search.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize, Default, Visit, VisitMut)]
+#[serde(rename_all = "lowercase")]
+pub enum VectorDistanceMetric {
+    /// Squared Euclidean distance (L2^2).
+    #[default]
+    L2sq,
+    /// Cosine distance (1 - cosine similarity).
+    Cosine,
+    /// Inner product (negative, for maximum inner product search).
+    #[serde(alias = "ip")]
+    InnerProduct,
+}
+
+impl fmt::Display for VectorDistanceMetric {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        match self {
+            VectorDistanceMetric::L2sq => write!(f, "l2sq"),
+            VectorDistanceMetric::Cosine => write!(f, "cosine"),
+            VectorDistanceMetric::InnerProduct => write!(f, "ip"),
+        }
+    }
+}
+
+impl std::str::FromStr for VectorDistanceMetric {
+    type Err = String;
+
+    fn from_str(s: &str) -> std::result::Result<Self, Self::Err> {
+        match s.to_lowercase().as_str() {
+            "l2sq" | "l2" | "euclidean" => Ok(VectorDistanceMetric::L2sq),
+            "cosine" | "cos" => Ok(VectorDistanceMetric::Cosine),
+            "inner_product" | "ip" | "dot" => Ok(VectorDistanceMetric::InnerProduct),
+            _ => Err(format!(
+                "Unknown distance metric: {}. Expected: l2sq, cosine, or ip",
+                s
+            )),
+        }
+    }
+}
+
+impl VectorDistanceMetric {
+    /// Returns the metric as u8 for blob serialization.
+    pub fn as_u8(&self) -> u8 {
+        match self {
+            Self::L2sq => 0,
+            Self::Cosine => 1,
+            Self::InnerProduct => 2,
+        }
+    }
+
+    /// Parses metric from u8 (used when reading blob).
+    pub fn try_from_u8(v: u8) -> Option<Self> {
+        match v {
+            0 => Some(Self::L2sq),
+            1 => Some(Self::Cosine),
+            2 => Some(Self::InnerProduct),
+            _ => None,
+        }
+    }
+}
+
+/// Default HNSW connectivity parameter.
+const DEFAULT_VECTOR_INDEX_CONNECTIVITY: u32 = 16;
+/// Default expansion factor during index construction.
+const DEFAULT_VECTOR_INDEX_EXPANSION_ADD: u32 = 128;
+/// Default expansion factor during search.
+const DEFAULT_VECTOR_INDEX_EXPANSION_SEARCH: u32 = 64;
+
+fn default_vector_index_connectivity() -> u32 {
+    DEFAULT_VECTOR_INDEX_CONNECTIVITY
+}
+
+fn default_vector_index_expansion_add() -> u32 {
+    DEFAULT_VECTOR_INDEX_EXPANSION_ADD
+}
+
+fn default_vector_index_expansion_search() -> u32 {
+    DEFAULT_VECTOR_INDEX_EXPANSION_SEARCH
+}
+
+/// Supported vector index engine types.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Default, Serialize, Deserialize, Visit, VisitMut)]
+#[serde(rename_all = "lowercase")]
+pub enum VectorIndexEngineType {
+    /// USearch HNSW implementation.
+    #[default]
+    Usearch,
+    // Future: Vsag,
+}
+
+impl VectorIndexEngineType {
+    /// Returns the engine type as u8 for blob serialization.
+    pub fn as_u8(&self) -> u8 {
+        match self {
+            Self::Usearch => 0,
+        }
+    }
+
+    /// Parses engine type from u8 (used when reading blob).
+    pub fn try_from_u8(v: u8) -> Option<Self> {
+        match v {
+            0 => Some(Self::Usearch),
+            _ => None,
+        }
+    }
+}
+
+impl fmt::Display for VectorIndexEngineType {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        match self {
+            Self::Usearch => write!(f, "usearch"),
+        }
+    }
+}
+
+impl std::str::FromStr for VectorIndexEngineType {
+    type Err = String;
+
+    fn from_str(s: &str) -> std::result::Result<Self, Self::Err> {
+        match s.to_lowercase().as_str() {
+            "usearch" => Ok(Self::Usearch),
+            _ => Err(format!(
+                "Unknown vector index engine: {}. Expected: usearch",
+                s
+            )),
+        }
+    }
+}
+
+/// Options for vector index (HNSW).
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, Visit, VisitMut)]
+#[serde(rename_all = "kebab-case")]
+pub struct VectorIndexOptions {
+    /// Vector index engine type (default: usearch).
+    #[serde(default)]
+    pub engine: VectorIndexEngineType,
+    /// Distance metric for similarity search.
+    #[serde(default)]
+    pub metric: VectorDistanceMetric,
+    /// HNSW connectivity parameter (M in the paper).
+    /// Higher values improve recall but increase memory usage.
+    #[serde(default = "default_vector_index_connectivity")]
+    pub connectivity: u32,
+    /// Expansion factor during index construction (ef_construction).
+    /// Higher values improve index quality but slow down construction.
+    #[serde(default = "default_vector_index_expansion_add")]
+    pub expansion_add: u32,
+    /// Expansion factor during search (ef_search).
+    /// Higher values improve recall but slow down search.
+    #[serde(default = "default_vector_index_expansion_search")]
+    pub expansion_search: u32,
+}
+
+impl Default for VectorIndexOptions {
+    fn default() -> Self {
+        Self {
+            engine: VectorIndexEngineType::default(),
+            metric: VectorDistanceMetric::default(),
+            connectivity: DEFAULT_VECTOR_INDEX_CONNECTIVITY,
+            expansion_add: DEFAULT_VECTOR_INDEX_EXPANSION_ADD,
+            expansion_search: DEFAULT_VECTOR_INDEX_EXPANSION_SEARCH,
+        }
+    }
+}
+
+impl fmt::Display for VectorIndexOptions {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        write!(
+            f,
+            "engine={}, metric={}, connectivity={}, expansion_add={}, expansion_search={}",
+            self.engine, self.metric, self.connectivity, self.expansion_add, self.expansion_search
+        )
+    }
+}
+
 #[cfg(test)]
 mod tests {
    use std::sync::Arc;
--- a/src/index/Cargo.toml
+++ b/src/index/Cargo.toml
@@ -7,6 +7,9 @@ license.workspace = true
 [lints]
 workspace = true

+[features]
+vector_index = ["dep:usearch"]
+
 [dependencies]
 async-trait.workspace = true
 asynchronous-codec = "0.7.0"
@@ -17,6 +20,7 @@ common-error.workspace = true
 common-macro.workspace = true
 common-runtime.workspace = true
 common-telemetry.workspace = true
+datatypes.workspace = true
 fastbloom = "0.8"
 fst.workspace = true
 futures.workspace = true
@@ -25,6 +29,7 @@ itertools.workspace = true
 jieba-rs = "0.8"
 lazy_static.workspace = true
 mockall.workspace = true
+nalgebra.workspace = true
 pin-project.workspace = true
 prost.workspace = true
 puffin.workspace = true
@@ -39,6 +44,7 @@ tantivy = { version = "0.24", features = ["zstd-compression"] }
 tantivy-jieba = "0.16"
 tokio.workspace = true
 tokio-util.workspace = true
+usearch = { version = "2.21", default-features = false, features = ["fp16lib"], optional = true }
 uuid.workspace = true

 [dev-dependencies]
--- a/src/index/src/lib.rs
+++ b/src/index/src/lib.rs
@@ -22,6 +22,8 @@ pub mod external_provider;
 pub mod fulltext_index;
 pub mod inverted_index;
 pub mod target;
+#[cfg(feature = "vector_index")]
+pub mod vector;

 pub type Bytes = Vec<u8>;
 pub type BytesRef<'a> = &'a [u8];
--- a/src/index/src/vector.rs
+++ b/src/index/src/vector.rs
@@ -0,0 +1,163 @@
+// Copyright 2023 Greptime Team
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+//! Vector index types and options.
+//!
+//! This module re-exports types from `datatypes` and provides conversions
+//! to USearch types, as well as distance computation functions.
+
+pub use datatypes::schema::{VectorDistanceMetric, VectorIndexOptions};
+use nalgebra::DVectorView;
+pub use usearch::MetricKind;
+
+/// Converts a VectorDistanceMetric to a USearch MetricKind.
+pub fn distance_metric_to_usearch(metric: VectorDistanceMetric) -> MetricKind {
+    match metric {
+        VectorDistanceMetric::L2sq => MetricKind::L2sq,
+        VectorDistanceMetric::Cosine => MetricKind::Cos,
+        VectorDistanceMetric::InnerProduct => MetricKind::IP,
+    }
+}
+
+/// Computes distance between two vectors using the specified metric.
+///
+/// Uses SIMD-optimized implementations via nalgebra.
+///
+/// **Note:** The caller must ensure that the two vectors have the same length
+/// and are non-empty. Empty vectors return 0.0 for all metrics.
+pub fn compute_distance(v1: &[f32], v2: &[f32], metric: VectorDistanceMetric) -> f32 {
+    // Empty vectors are degenerate; return 0.0 uniformly across all metrics.
+    if v1.is_empty() || v2.is_empty() {
+        return 0.0;
+    }
+
+    match metric {
+        VectorDistanceMetric::L2sq => l2sq(v1, v2),
+        VectorDistanceMetric::Cosine => cosine(v1, v2),
+        VectorDistanceMetric::InnerProduct => -dot(v1, v2),
+    }
+}
+
+/// Calculates the squared L2 distance between two vectors.
+fn l2sq(lhs: &[f32], rhs: &[f32]) -> f32 {
+    let lhs = DVectorView::from_slice(lhs, lhs.len());
+    let rhs = DVectorView::from_slice(rhs, rhs.len());
+    (lhs - rhs).norm_squared()
+}
+
+/// Calculates the cosine distance between two vectors.
+///
+/// Returns a value in `[0.0, 2.0]` where 0.0 means identical direction and 2.0 means
+/// opposite direction. For degenerate cases (zero or near-zero magnitude vectors),
+/// returns 1.0 (maximum uncertainty) to avoid NaN and ensure safe index operations.
+fn cosine(lhs: &[f32], rhs: &[f32]) -> f32 {
+    let lhs_vec = DVectorView::from_slice(lhs, lhs.len());
+    let rhs_vec = DVectorView::from_slice(rhs, rhs.len());
+
+    let dot_product = lhs_vec.dot(&rhs_vec);
+    let lhs_norm = lhs_vec.norm();
+    let rhs_norm = rhs_vec.norm();
+
+    // Zero-magnitude vectors have undefined direction; return max distance as safe fallback.
+    if dot_product.abs() < f32::EPSILON
+        || lhs_norm.abs() < f32::EPSILON
+        || rhs_norm.abs() < f32::EPSILON
+    {
+        return 1.0;
+    }
+
+    let cos_similar = dot_product / (lhs_norm * rhs_norm);
+    let res = 1.0 - cos_similar;
+    // Clamp near-zero results to exactly 0.0 to avoid floating-point artifacts.
+    if res.abs() < f32::EPSILON { 0.0 } else { res }
+}
+
+/// Calculates the dot product between two vectors.
+fn dot(lhs: &[f32], rhs: &[f32]) -> f32 {
+    let lhs = DVectorView::from_slice(lhs, lhs.len());
+    let rhs = DVectorView::from_slice(rhs, rhs.len());
+    lhs.dot(&rhs)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_distance_metric_to_usearch() {
+        assert_eq!(
+            distance_metric_to_usearch(VectorDistanceMetric::L2sq),
+            MetricKind::L2sq
+        );
+        assert_eq!(
+            distance_metric_to_usearch(VectorDistanceMetric::Cosine),
+            MetricKind::Cos
+        );
+        assert_eq!(
+            distance_metric_to_usearch(VectorDistanceMetric::InnerProduct),
+            MetricKind::IP
+        );
+    }
+
+    #[test]
+    fn test_vector_index_options_default() {
+        let options = VectorIndexOptions::default();
+        assert_eq!(options.metric, VectorDistanceMetric::L2sq);
+        assert_eq!(options.connectivity, 16);
+        assert_eq!(options.expansion_add, 128);
+        assert_eq!(options.expansion_search, 64);
+    }
+
+    #[test]
+    fn test_compute_distance_l2sq() {
+        let v1 = vec![1.0, 2.0, 3.0];
+        let v2 = vec![4.0, 5.0, 6.0];
+        // L2sq = (4-1)^2 + (5-2)^2 + (6-3)^2 = 9 + 9 + 9 = 27
+        let dist = compute_distance(&v1, &v2, VectorDistanceMetric::L2sq);
+        assert!((dist - 27.0).abs() < 1e-6);
+    }
+
+    #[test]
+    fn test_compute_distance_cosine() {
+        let v1 = vec![1.0, 0.0, 0.0];
+        let v2 = vec![0.0, 1.0, 0.0];
+        // Orthogonal vectors have cosine similarity of 0, distance of 1
+        let dist = compute_distance(&v1, &v2, VectorDistanceMetric::Cosine);
+        assert!((dist - 1.0).abs() < 1e-6);
+    }
+
+    #[test]
+    fn test_compute_distance_inner_product() {
+        let v1 = vec![1.0, 2.0, 3.0];
+        let v2 = vec![4.0, 5.0, 6.0];
+        // Inner product = 1*4 + 2*5 + 3*6 = 4 + 10 + 18 = 32
+        // Distance is negated: -32
+        let dist = compute_distance(&v1, &v2, VectorDistanceMetric::InnerProduct);
+        assert!((dist - (-32.0)).abs() < 1e-6);
+    }
+
+    #[test]
+    fn test_compute_distance_empty_vectors() {
+        // Empty vectors should return 0.0 uniformly for all metrics
+        assert_eq!(compute_distance(&[], &[], VectorDistanceMetric::L2sq), 0.0);
+        assert_eq!(
+            compute_distance(&[], &[], VectorDistanceMetric::Cosine),
+            0.0
+        );
+        assert_eq!(
+            compute_distance(&[], &[], VectorDistanceMetric::InnerProduct),
+            0.0
+        );
+    }
+}
--- a/src/mito2/src/flush.rs
+++ b/src/mito2/src/flush.rs
@@ -774,7 +774,12 @@ fn memtable_flat_sources(
            let iter = only_range.build_record_batch_iter(None)?;
            // Dedup according to append mode and merge mode.
            // Even single range may have duplicate rows.
-            let iter = maybe_dedup_one(options, field_column_start, iter);
+            let iter = maybe_dedup_one(
+                options.append_mode,
+                options.merge_mode(),
+                field_column_start,
+                iter,
+            );
            flat_sources.sources.push(FlatSource::Iter(iter));
        };
    } else {
@@ -842,17 +847,18 @@ fn merge_and_dedup(
    Ok(maybe_dedup)
 }

-fn maybe_dedup_one(
-    options: &RegionOptions,
+pub fn maybe_dedup_one(
+    append_mode: bool,
+    merge_mode: MergeMode,
    field_column_start: usize,
    input_iter: BoxedRecordBatchIterator,
 ) -> BoxedRecordBatchIterator {
-    if options.append_mode {
+    if append_mode {
        // No dedup in append mode
        input_iter
    } else {
        // Dedup according to merge mode.
-        match options.merge_mode() {
+        match merge_mode {
            MergeMode::LastRow => {
                Box::new(FlatDedupIterator::new(input_iter, FlatLastRow::new(false)))
            }
--- a/src/mito2/src/memtable.rs
+++ b/src/mito2/src/memtable.rs
@@ -55,10 +55,8 @@ pub mod time_partition;
 pub mod time_series;
 pub(crate) mod version;

-#[cfg(any(test, feature = "test"))]
-pub use bulk::part::BulkPart;
 pub use bulk::part::{
-    BulkPartEncoder, BulkPartMeta, UnorderedPart, record_batch_estimated_size,
+    BulkPart, BulkPartEncoder, BulkPartMeta, UnorderedPart, record_batch_estimated_size,
    sort_primary_key_record_batch,
 };
 #[cfg(any(test, feature = "test"))]
--- a/src/mito2/src/memtable/bulk.rs
+++ b/src/mito2/src/memtable/bulk.rs
@@ -668,10 +668,10 @@ impl BulkMemtable {
 }

 /// Iterator builder for bulk range
-struct BulkRangeIterBuilder {
-    part: BulkPart,
-    context: Arc<BulkIterContext>,
-    sequence: Option<SequenceRange>,
+pub struct BulkRangeIterBuilder {
+    pub part: BulkPart,
+    pub context: Arc<BulkIterContext>,
+    pub sequence: Option<SequenceRange>,
 }

 impl IterBuilder for BulkRangeIterBuilder {
@@ -1188,7 +1188,6 @@ impl MemtableBuilder for BulkMemtableBuilder {

 #[cfg(test)]
 mod tests {
-
    use mito_codec::row_converter::build_primary_key_codec;

    use super::*;
--- a/src/mito2/src/memtable/bulk/part.rs
+++ b/src/mito2/src/memtable/bulk/part.rs
@@ -1211,343 +1211,24 @@ impl BulkPartEncoder {
    }
 }

-/// Converts mutations to record batches.
-fn mutations_to_record_batch(
-    mutations: &[Mutation],
-    metadata: &RegionMetadataRef,
-    pk_encoder: &DensePrimaryKeyCodec,
-    dedup: bool,
-) -> Result<Option<(RecordBatch, i64, i64)>> {
-    let total_rows: usize = mutations
-        .iter()
-        .map(|m| m.rows.as_ref().map(|r| r.rows.len()).unwrap_or(0))
-        .sum();
-
-    if total_rows == 0 {
-        return Ok(None);
-    }
-
-    let mut pk_builder = BinaryBuilder::with_capacity(total_rows, 0);
-
-    let mut ts_vector: Box<dyn MutableVector> = metadata
-        .time_index_column()
-        .column_schema
-        .data_type
-        .create_mutable_vector(total_rows);
-    let mut sequence_builder = UInt64Builder::with_capacity(total_rows);
-    let mut op_type_builder = UInt8Builder::with_capacity(total_rows);
-
-    let mut field_builders: Vec<Box<dyn MutableVector>> = metadata
-        .field_columns()
-        .map(|f| f.column_schema.data_type.create_mutable_vector(total_rows))
-        .collect();
-
-    let mut pk_buffer = vec![];
-    for m in mutations {
-        let Some(key_values) = KeyValuesRef::new(metadata, m) else {
-            continue;
-        };
-
-        for row in key_values.iter() {
-            pk_buffer.clear();
-            pk_encoder
-                .encode_to_vec(row.primary_keys(), &mut pk_buffer)
-                .context(EncodeSnafu)?;
-            pk_builder.append_value(pk_buffer.as_bytes());
-            ts_vector.push_value_ref(&row.timestamp());
-            sequence_builder.append_value(row.sequence());
-            op_type_builder.append_value(row.op_type() as u8);
-            for (builder, field) in field_builders.iter_mut().zip(row.fields()) {
-                builder.push_value_ref(&field);
-            }
-        }
-    }
-
-    let arrow_schema = to_sst_arrow_schema(metadata);
-    // safety: timestamp column must be valid, and values must not be None.
-    let timestamp_unit = metadata
-        .time_index_column()
-        .column_schema
-        .data_type
-        .as_timestamp()
-        .unwrap()
-        .unit();
-    let sorter = ArraysSorter {
-        encoded_primary_keys: pk_builder.finish(),
-        timestamp_unit,
-        timestamp: ts_vector.to_vector().to_arrow_array(),
-        sequence: sequence_builder.finish(),
-        op_type: op_type_builder.finish(),
-        fields: field_builders
-            .iter_mut()
-            .map(|f| f.to_vector().to_arrow_array()),
-        dedup,
-        arrow_schema,
-    };
-
-    sorter.sort().map(Some)
-}
-
-struct ArraysSorter<I> {
-    encoded_primary_keys: BinaryArray,
-    timestamp_unit: TimeUnit,
-    timestamp: ArrayRef,
-    sequence: UInt64Array,
-    op_type: UInt8Array,
-    fields: I,
-    dedup: bool,
-    arrow_schema: SchemaRef,
-}
-
-impl<I> ArraysSorter<I>
-where
-    I: Iterator<Item = ArrayRef>,
-{
-    /// Converts arrays to record batch.
-    fn sort(self) -> Result<(RecordBatch, i64, i64)> {
-        debug_assert!(!self.timestamp.is_empty());
-        debug_assert!(self.timestamp.len() == self.sequence.len());
-        debug_assert!(self.timestamp.len() == self.op_type.len());
-        debug_assert!(self.timestamp.len() == self.encoded_primary_keys.len());
-
-        let timestamp_iter = timestamp_array_to_iter(self.timestamp_unit, &self.timestamp);
-        let (mut min_timestamp, mut max_timestamp) = (i64::MAX, i64::MIN);
-        let mut to_sort = self
-            .encoded_primary_keys
-            .iter()
-            .zip(timestamp_iter)
-            .zip(self.sequence.iter())
-            .map(|((pk, timestamp), sequence)| {
-                max_timestamp = max_timestamp.max(*timestamp);
-                min_timestamp = min_timestamp.min(*timestamp);
-                (pk, timestamp, sequence)
-            })
-            .enumerate()
-            .collect::<Vec<_>>();
-
-        to_sort.sort_unstable_by(|(_, (l_pk, l_ts, l_seq)), (_, (r_pk, r_ts, r_seq))| {
-            l_pk.cmp(r_pk)
-                .then(l_ts.cmp(r_ts))
-                .then(l_seq.cmp(r_seq).reverse())
-        });
-
-        if self.dedup {
-            // Dedup by timestamps while ignore sequence.
-            to_sort.dedup_by(|(_, (l_pk, l_ts, _)), (_, (r_pk, r_ts, _))| {
-                l_pk == r_pk && l_ts == r_ts
-            });
-        }
-
-        let indices = UInt32Array::from_iter_values(to_sort.iter().map(|v| v.0 as u32));
-
-        let pk_dictionary = Arc::new(binary_array_to_dictionary(
-            // safety: pk must be BinaryArray
-            arrow::compute::take(
-                &self.encoded_primary_keys,
-                &indices,
-                Some(TakeOptions {
-                    check_bounds: false,
-                }),
-            )
-            .context(ComputeArrowSnafu)?
-            .as_any()
-            .downcast_ref::<BinaryArray>()
-            .unwrap(),
-        )?) as ArrayRef;
-
-        let mut arrays = Vec::with_capacity(self.arrow_schema.fields.len());
-        for arr in self.fields {
-            arrays.push(
-                arrow::compute::take(
-                    &arr,
-                    &indices,
-                    Some(TakeOptions {
-                        check_bounds: false,
-                    }),
-                )
-                .context(ComputeArrowSnafu)?,
-            );
-        }
-
-        let timestamp = arrow::compute::take(
-            &self.timestamp,
-            &indices,
-            Some(TakeOptions {
-                check_bounds: false,
-            }),
-        )
-        .context(ComputeArrowSnafu)?;
-
-        arrays.push(timestamp);
-        arrays.push(pk_dictionary);
-        arrays.push(
-            arrow::compute::take(
-                &self.sequence,
-                &indices,
-                Some(TakeOptions {
-                    check_bounds: false,
-                }),
-            )
-            .context(ComputeArrowSnafu)?,
-        );
-
-        arrays.push(
-            arrow::compute::take(
-                &self.op_type,
-                &indices,
-                Some(TakeOptions {
-                    check_bounds: false,
-                }),
-            )
-            .context(ComputeArrowSnafu)?,
-        );
-
-        let batch = RecordBatch::try_new(self.arrow_schema, arrays).context(NewRecordBatchSnafu)?;
-        Ok((batch, min_timestamp, max_timestamp))
-    }
-}
-
-/// Converts timestamp array to an iter of i64 values.
-fn timestamp_array_to_iter(
-    timestamp_unit: TimeUnit,
-    timestamp: &ArrayRef,
-) -> impl Iterator<Item = &i64> {
-    match timestamp_unit {
-        // safety: timestamp column must be valid.
-        TimeUnit::Second => timestamp
-            .as_any()
-            .downcast_ref::<TimestampSecondArray>()
-            .unwrap()
-            .values()
-            .iter(),
-        TimeUnit::Millisecond => timestamp
-            .as_any()
-            .downcast_ref::<TimestampMillisecondArray>()
-            .unwrap()
-            .values()
-            .iter(),
-        TimeUnit::Microsecond => timestamp
-            .as_any()
-            .downcast_ref::<TimestampMicrosecondArray>()
-            .unwrap()
-            .values()
-            .iter(),
-        TimeUnit::Nanosecond => timestamp
-            .as_any()
-            .downcast_ref::<TimestampNanosecondArray>()
-            .unwrap()
-            .values()
-            .iter(),
-    }
-}
-
-/// Converts a **sorted** [BinaryArray] to [DictionaryArray].
-fn binary_array_to_dictionary(input: &BinaryArray) -> Result<PrimaryKeyArray> {
-    if input.is_empty() {
-        return Ok(DictionaryArray::new(
-            UInt32Array::from(Vec::<u32>::new()),
-            Arc::new(BinaryArray::from_vec(vec![])) as ArrayRef,
-        ));
-    }
-    let mut keys = Vec::with_capacity(16);
-    let mut values = BinaryBuilder::new();
-    let mut prev: usize = 0;
-    keys.push(prev as u32);
-    values.append_value(input.value(prev));
-
-    for current_bytes in input.iter().skip(1) {
-        // safety: encoded pk must present.
-        let current_bytes = current_bytes.unwrap();
-        let prev_bytes = input.value(prev);
-        if current_bytes != prev_bytes {
-            values.append_value(current_bytes);
-            prev += 1;
-        }
-        keys.push(prev as u32);
-    }
-
-    Ok(DictionaryArray::new(
-        UInt32Array::from(keys),
-        Arc::new(values.finish()) as ArrayRef,
-    ))
-}
-
 #[cfg(test)]
 mod tests {
-    use std::collections::VecDeque;
-
    use api::v1::{Row, SemanticType, WriteHint};
    use datafusion_common::ScalarValue;
    use datatypes::arrow::array::Float64Array;
-    use datatypes::prelude::{ConcreteDataType, ScalarVector, Value};
+    use datatypes::prelude::{ConcreteDataType, Value};
    use datatypes::schema::ColumnSchema;
-    use datatypes::vectors::{Float64Vector, TimestampMillisecondVector};
    use store_api::metadata::{ColumnMetadata, RegionMetadataBuilder};
    use store_api::storage::RegionId;
    use store_api::storage::consts::ReservedColumnId;

    use super::*;
    use crate::memtable::bulk::context::BulkIterContext;
-    use crate::sst::parquet::format::{PrimaryKeyReadFormat, ReadFormat};
    use crate::sst::{FlatSchemaOptions, to_flat_sst_arrow_schema};
    use crate::test_util::memtable_util::{
        build_key_values_with_ts_seq_values, metadata_for_test, region_metadata_to_row_schema,
    };

-    fn check_binary_array_to_dictionary(
-        input: &[&[u8]],
-        expected_keys: &[u32],
-        expected_values: &[&[u8]],
-    ) {
-        let input = BinaryArray::from_iter_values(input.iter());
-        let array = binary_array_to_dictionary(&input).unwrap();
-        assert_eq!(
-            &expected_keys,
-            &array.keys().iter().map(|v| v.unwrap()).collect::<Vec<_>>()
-        );
-        assert_eq!(
-            expected_values,
-            &array
-                .values()
-                .as_any()
-                .downcast_ref::<BinaryArray>()
-                .unwrap()
-                .iter()
-                .map(|v| v.unwrap())
-                .collect::<Vec<_>>()
-        );
-    }
-
-    #[test]
-    fn test_binary_array_to_dictionary() {
-        check_binary_array_to_dictionary(&[], &[], &[]);
-
-        check_binary_array_to_dictionary(&["a".as_bytes()], &[0], &["a".as_bytes()]);
-
-        check_binary_array_to_dictionary(
-            &["a".as_bytes(), "a".as_bytes()],
-            &[0, 0],
-            &["a".as_bytes()],
-        );
-
-        check_binary_array_to_dictionary(
-            &["a".as_bytes(), "a".as_bytes(), "b".as_bytes()],
-            &[0, 0, 1],
-            &["a".as_bytes(), "b".as_bytes()],
-        );
-
-        check_binary_array_to_dictionary(
-            &[
-                "a".as_bytes(),
-                "a".as_bytes(),
-                "b".as_bytes(),
-                "c".as_bytes(),
-            ],
-            &[0, 0, 1, 2],
-            &["a".as_bytes(), "b".as_bytes(), "c".as_bytes()],
-        );
-    }
-
    struct MutationInput<'a> {
        k0: &'a str,
        k1: u32,
@@ -1563,232 +1244,6 @@ mod tests {
        v1: &'a [Option<f64>],
    }

-    fn check_mutations_to_record_batches(
-        input: &[MutationInput],
-        expected: &[BatchOutput],
-        expected_timestamp: (i64, i64),
-        dedup: bool,
-    ) {
-        let metadata = metadata_for_test();
-        let mutations = input
-            .iter()
-            .map(|m| {
-                build_key_values_with_ts_seq_values(
-                    &metadata,
-                    m.k0.to_string(),
-                    m.k1,
-                    m.timestamps.iter().copied(),
-                    m.v1.iter().copied(),
-                    m.sequence,
-                )
-                .mutation
-            })
-            .collect::<Vec<_>>();
-        let total_rows: usize = mutations
-            .iter()
-            .flat_map(|m| m.rows.iter())
-            .map(|r| r.rows.len())
-            .sum();
-
-        let pk_encoder = DensePrimaryKeyCodec::new(&metadata);
-
-        let (batch, _, _) = mutations_to_record_batch(&mutations, &metadata, &pk_encoder, dedup)
-            .unwrap()
-            .unwrap();
-        let read_format = PrimaryKeyReadFormat::new_with_all_columns(metadata.clone());
-        let mut batches = VecDeque::new();
-        read_format
-            .convert_record_batch(&batch, None, &mut batches)
-            .unwrap();
-        if !dedup {
-            assert_eq!(
-                total_rows,
-                batches.iter().map(|b| { b.num_rows() }).sum::<usize>()
-            );
-        }
-        let batch_values = batches
-            .into_iter()
-            .map(|b| {
-                let pk_values = pk_encoder.decode(b.primary_key()).unwrap().into_dense();
-                let timestamps = b
-                    .timestamps()
-                    .as_any()
-                    .downcast_ref::<TimestampMillisecondVector>()
-                    .unwrap()
-                    .iter_data()
-                    .map(|v| v.unwrap().0.value())
-                    .collect::<Vec<_>>();
-                let float_values = b.fields()[1]
-                    .data
-                    .as_any()
-                    .downcast_ref::<Float64Vector>()
-                    .unwrap()
-                    .iter_data()
-                    .collect::<Vec<_>>();
-
-                (pk_values, timestamps, float_values)
-            })
-            .collect::<Vec<_>>();
-        assert_eq!(expected.len(), batch_values.len());
-
-        for idx in 0..expected.len() {
-            assert_eq!(expected[idx].pk_values, &batch_values[idx].0);
-            assert_eq!(expected[idx].timestamps, &batch_values[idx].1);
-            assert_eq!(expected[idx].v1, &batch_values[idx].2);
-        }
-    }
-
-    #[test]
-    fn test_mutations_to_record_batch() {
-        check_mutations_to_record_batches(
-            &[MutationInput {
-                k0: "a",
-                k1: 0,
-                timestamps: &[0],
-                v1: &[Some(0.1)],
-                sequence: 0,
-            }],
-            &[BatchOutput {
-                pk_values: &[Value::String("a".into()), Value::UInt32(0)],
-                timestamps: &[0],
-                v1: &[Some(0.1)],
-            }],
-            (0, 0),
-            true,
-        );
-
-        check_mutations_to_record_batches(
-            &[
-                MutationInput {
-                    k0: "a",
-                    k1: 0,
-                    timestamps: &[0],
-                    v1: &[Some(0.1)],
-                    sequence: 0,
-                },
-                MutationInput {
-                    k0: "b",
-                    k1: 0,
-                    timestamps: &[0],
-                    v1: &[Some(0.0)],
-                    sequence: 0,
-                },
-                MutationInput {
-                    k0: "a",
-                    k1: 0,
-                    timestamps: &[1],
-                    v1: &[Some(0.2)],
-                    sequence: 1,
-                },
-                MutationInput {
-                    k0: "a",
-                    k1: 1,
-                    timestamps: &[1],
-                    v1: &[Some(0.3)],
-                    sequence: 2,
-                },
-            ],
-            &[
-                BatchOutput {
-                    pk_values: &[Value::String("a".into()), Value::UInt32(0)],
-                    timestamps: &[0, 1],
-                    v1: &[Some(0.1), Some(0.2)],
-                },
-                BatchOutput {
-                    pk_values: &[Value::String("a".into()), Value::UInt32(1)],
-                    timestamps: &[1],
-                    v1: &[Some(0.3)],
-                },
-                BatchOutput {
-                    pk_values: &[Value::String("b".into()), Value::UInt32(0)],
-                    timestamps: &[0],
-                    v1: &[Some(0.0)],
-                },
-            ],
-            (0, 1),
-            true,
-        );
-
-        check_mutations_to_record_batches(
-            &[
-                MutationInput {
-                    k0: "a",
-                    k1: 0,
-                    timestamps: &[0],
-                    v1: &[Some(0.1)],
-                    sequence: 0,
-                },
-                MutationInput {
-                    k0: "b",
-                    k1: 0,
-                    timestamps: &[0],
-                    v1: &[Some(0.0)],
-                    sequence: 0,
-                },
-                MutationInput {
-                    k0: "a",
-                    k1: 0,
-                    timestamps: &[0],
-                    v1: &[Some(0.2)],
-                    sequence: 1,
-                },
-            ],
-            &[
-                BatchOutput {
-                    pk_values: &[Value::String("a".into()), Value::UInt32(0)],
-                    timestamps: &[0],
-                    v1: &[Some(0.2)],
-                },
-                BatchOutput {
-                    pk_values: &[Value::String("b".into()), Value::UInt32(0)],
-                    timestamps: &[0],
-                    v1: &[Some(0.0)],
-                },
-            ],
-            (0, 0),
-            true,
-        );
-        check_mutations_to_record_batches(
-            &[
-                MutationInput {
-                    k0: "a",
-                    k1: 0,
-                    timestamps: &[0],
-                    v1: &[Some(0.1)],
-                    sequence: 0,
-                },
-                MutationInput {
-                    k0: "b",
-                    k1: 0,
-                    timestamps: &[0],
-                    v1: &[Some(0.0)],
-                    sequence: 0,
-                },
-                MutationInput {
-                    k0: "a",
-                    k1: 0,
-                    timestamps: &[0],
-                    v1: &[Some(0.2)],
-                    sequence: 1,
-                },
-            ],
-            &[
-                BatchOutput {
-                    pk_values: &[Value::String("a".into()), Value::UInt32(0)],
-                    timestamps: &[0, 0],
-                    v1: &[Some(0.2), Some(0.1)],
-                },
-                BatchOutput {
-                    pk_values: &[Value::String("b".into()), Value::UInt32(0)],
-                    timestamps: &[0],
-                    v1: &[Some(0.0)],
-                },
-            ],
-            (0, 0),
-            false,
-        );
-    }
-
    fn encode(input: &[MutationInput]) -> EncodedBulkPart {
        let metadata = metadata_for_test();
        let kvs = input
--- a/src/mito2/src/sst.rs
+++ b/src/mito2/src/sst.rs
@@ -121,7 +121,7 @@ impl FlatSchemaOptions {
 ///
 /// The schema is:
 /// ```text
-/// primary key columns, field columns, time index, __prmary_key, __sequence, __op_type
+/// primary key columns, field columns, time index, __primary_key, __sequence, __op_type
 /// ```
 ///
 /// # Panics
--- a/src/object-store/src/config.rs
+++ b/src/object-store/src/config.rs
@@ -16,6 +16,7 @@ use std::time::Duration;

 use common_base::readable_size::ReadableSize;
 use common_base::secrets::{ExposeSecret, SecretString};
+use common_telemetry::tracing::warn;
 use opendal::services::{Azblob, Gcs, Oss, S3};
 use serde::{Deserialize, Serialize};

@@ -123,11 +124,18 @@ impl From<&S3Connection> for S3 {
    fn from(connection: &S3Connection) -> Self {
        let root = util::normalize_dir(&connection.root);

-        let mut builder = S3::default()
-            .root(&root)
-            .bucket(&connection.bucket)
-            .access_key_id(connection.access_key_id.expose_secret())
-            .secret_access_key(connection.secret_access_key.expose_secret());
+        let mut builder = S3::default().root(&root).bucket(&connection.bucket);
+
+        if !connection.access_key_id.expose_secret().is_empty()
+            && !connection.secret_access_key.expose_secret().is_empty()
+        {
+            builder = builder
+                .access_key_id(connection.access_key_id.expose_secret())
+                .secret_access_key(connection.secret_access_key.expose_secret());
+        } else {
+            warn!("No access key id or secret access key provided, using anonymous access");
+            builder = builder.allow_anonymous().disable_ec2_metadata();
+        }

        if let Some(endpoint) = &connection.endpoint {
            builder = builder.endpoint(endpoint);
--- a/src/sql/src/error.rs
+++ b/src/sql/src/error.rs
@@ -285,6 +285,13 @@ pub enum Error {
        location: Location,
    },

+    #[snafu(display("Failed to set VECTOR index option"))]
+    SetVectorIndexOption {
+        source: datatypes::error::Error,
+        #[snafu(implicit)]
+        location: Location,
+    },
+
    #[snafu(display(
        "Invalid partition number: {}, should be in range [2, 65536]",
        partition_num
@@ -394,7 +401,9 @@ impl ErrorExt for Error {
            ConvertValue { .. } => StatusCode::Unsupported,

            PermissionDenied { .. } => StatusCode::PermissionDenied,
-            SetFulltextOption { .. } | SetSkippingIndexOption { .. } => StatusCode::Unexpected,
+            SetFulltextOption { .. }
+            | SetSkippingIndexOption { .. }
+            | SetVectorIndexOption { .. } => StatusCode::Unexpected,
        }
    }

--- a/src/sql/src/parsers/create_parser.rs
+++ b/src/sql/src/parsers/create_parser.rs
@@ -43,6 +43,7 @@ use crate::parser::{FLOW, ParserContext};
 use crate::parsers::tql_parser;
 use crate::parsers::utils::{
    self, validate_column_fulltext_create_option, validate_column_skipping_index_create_option,
+    validate_column_vector_index_create_option,
 };
 use crate::statements::create::{
    Column, ColumnExtensions, CreateDatabase, CreateExternalTable, CreateFlow, CreateTable,
@@ -60,6 +61,7 @@ pub const EXPIRE: &str = "EXPIRE";
 pub const AFTER: &str = "AFTER";
 pub const INVERTED: &str = "INVERTED";
 pub const SKIPPING: &str = "SKIPPING";
+pub const VECTOR: &str = "VECTOR";

 pub type RawIntervalExpr = String;

@@ -928,6 +930,61 @@ impl<'a> ParserContext<'a> {
            is_index_declared |= true;
        }

+        // vector index
+        if let Token::Word(word) = parser.peek_token().token
+            && word.value.eq_ignore_ascii_case(VECTOR)
+        {
+            parser.next_token();
+            // Consume `INDEX` keyword
+            ensure!(
+                parser.parse_keyword(Keyword::INDEX),
+                InvalidColumnOptionSnafu {
+                    name: column_name.to_string(),
+                    msg: "expect INDEX after VECTOR keyword",
+                }
+            );
+
+            ensure!(
+                column_extensions.vector_index_options.is_none(),
+                InvalidColumnOptionSnafu {
+                    name: column_name.to_string(),
+                    msg: "duplicated VECTOR INDEX option",
+                }
+            );
+
+            // Check that column is a vector type
+            let column_type = get_unalias_type(column_type);
+            let data_type = sql_data_type_to_concrete_data_type(&column_type, column_extensions)?;
+            ensure!(
+                matches!(data_type, ConcreteDataType::Vector(_)),
+                InvalidColumnOptionSnafu {
+                    name: column_name.to_string(),
+                    msg: "VECTOR INDEX only supports Vector type columns",
+                }
+            );
+
+            let options = parser
+                .parse_options(Keyword::WITH)
+                .context(error::SyntaxSnafu)?
+                .into_iter()
+                .map(parse_option_string)
+                .collect::<Result<Vec<_>>>()?;
+
+            for (key, _) in options.iter() {
+                ensure!(
+                    validate_column_vector_index_create_option(key),
+                    InvalidColumnOptionSnafu {
+                        name: column_name.to_string(),
+                        msg: format!("invalid VECTOR INDEX option: {key}"),
+                    }
+                );
+            }
+
+            let options = OptionMap::new(options);
+            column_extensions.vector_index_options = Some(options);
+            is_index_declared |= true;
+        }
+
        Ok(is_index_declared)
    }

@@ -2714,7 +2771,8 @@ CREATE TABLE log (

    #[test]
    fn test_parse_column_extensions_vector() {
-        let sql = "VECTOR(128)";
+        // Test that vector options are parsed from data_type (no additional SQL needed)
+        let sql = "";
        let dialect = GenericDialect {};
        let mut tokenizer = Tokenizer::new(&dialect, sql);
        let tokens = tokenizer.tokenize().unwrap();
@@ -2734,7 +2792,8 @@ CREATE TABLE log (

    #[test]
    fn test_parse_column_extensions_vector_invalid() {
-        let sql = "VECTOR()";
+        // Test that vector with no dimension fails
+        let sql = "";
        let dialect = GenericDialect {};
        let mut tokenizer = Tokenizer::new(&dialect, sql);
        let tokens = tokenizer.tokenize().unwrap();
@@ -2912,4 +2971,174 @@ CREATE TABLE log (
                .unwrap();
        assert_eq!("SELECT '10 seconds'::INTERVAL", &stmts[0].to_string());
    }
+
+    #[test]
+    fn test_parse_create_table_vector_index_options() {
+        // Test basic vector index
+        let sql = r"
+CREATE TABLE vectors (
+    ts TIMESTAMP TIME INDEX,
+    vec VECTOR(128) VECTOR INDEX,
+)";
+        let result =
+            ParserContext::create_with_dialect(sql, &GreptimeDbDialect {}, ParseOptions::default())
+                .unwrap();
+
+        if let Statement::CreateTable(c) = &result[0] {
+            c.columns.iter().for_each(|col| {
+                if col.name().value == "vec" {
+                    assert!(
+                        col.extensions
+                            .vector_index_options
+                            .as_ref()
+                            .unwrap()
+                            .is_empty()
+                    );
+                }
+            });
+        } else {
+            panic!("should be create_table statement");
+        }
+
+        // Test vector index with options
+        let sql = r"
+CREATE TABLE vectors (
+    ts TIMESTAMP TIME INDEX,
+    vec VECTOR(128) VECTOR INDEX WITH (metric='cosine', connectivity='32', expansion_add='256', expansion_search='128')
+)";
+        let result =
+            ParserContext::create_with_dialect(sql, &GreptimeDbDialect {}, ParseOptions::default())
+                .unwrap();
+
+        if let Statement::CreateTable(c) = &result[0] {
+            c.columns.iter().for_each(|col| {
+                if col.name().value == "vec" {
+                    let options = col.extensions.vector_index_options.as_ref().unwrap();
+                    assert_eq!(options.len(), 4);
+                    assert_eq!(options.get("metric").unwrap(), "cosine");
+                    assert_eq!(options.get("connectivity").unwrap(), "32");
+                    assert_eq!(options.get("expansion_add").unwrap(), "256");
+                    assert_eq!(options.get("expansion_search").unwrap(), "128");
+                }
+            });
+        } else {
+            panic!("should be create_table statement");
+        }
+    }
+
+    #[test]
+    fn test_parse_create_table_vector_index_invalid_type() {
+        // Test vector index on non-vector type (should fail)
+        let sql = r"
+CREATE TABLE vectors (
+    ts TIMESTAMP TIME INDEX,
+    col INT VECTOR INDEX,
+)";
+        let result =
+            ParserContext::create_with_dialect(sql, &GreptimeDbDialect {}, ParseOptions::default());
+        assert!(result.is_err());
+        assert!(
+            result
+                .unwrap_err()
+                .to_string()
+                .contains("VECTOR INDEX only supports Vector type columns")
+        );
+    }
+
+    #[test]
+    fn test_parse_create_table_vector_index_duplicate() {
+        // Test duplicate vector index (should fail)
+        let sql = r"
+CREATE TABLE vectors (
+    ts TIMESTAMP TIME INDEX,
+    vec VECTOR(128) VECTOR INDEX VECTOR INDEX,
+)";
+        let result =
+            ParserContext::create_with_dialect(sql, &GreptimeDbDialect {}, ParseOptions::default());
+        assert!(result.is_err());
+        assert!(
+            result
+                .unwrap_err()
+                .to_string()
+                .contains("duplicated VECTOR INDEX option")
+        );
+    }
+
+    #[test]
+    fn test_parse_create_table_vector_index_invalid_option() {
+        // Test invalid option key (should fail)
+        let sql = r"
+CREATE TABLE vectors (
+    ts TIMESTAMP TIME INDEX,
+    vec VECTOR(128) VECTOR INDEX WITH (metric='l2sq', invalid_option='foo')
+)";
+        let result =
+            ParserContext::create_with_dialect(sql, &GreptimeDbDialect {}, ParseOptions::default());
+        assert!(result.is_err());
+        assert!(
+            result
+                .unwrap_err()
+                .to_string()
+                .contains("invalid VECTOR INDEX option")
+        );
+    }
+
+    #[test]
+    fn test_parse_column_extensions_vector_index() {
+        // Test vector index on vector type
+        {
+            let sql = "VECTOR INDEX WITH (metric = 'l2sq')";
+            let dialect = GenericDialect {};
+            let mut tokenizer = Tokenizer::new(&dialect, sql);
+            let tokens = tokenizer.tokenize().unwrap();
+            let mut parser = Parser::new(&dialect).with_tokens(tokens);
+            let name = Ident::new("vec_col");
+            let data_type =
+                DataType::Custom(vec![Ident::new("VECTOR")].into(), vec!["128".to_string()]);
+            // First, parse the vector type to set vector_options
+            let mut extensions = ColumnExtensions {
+                vector_options: Some(OptionMap::from([(
+                    VECTOR_OPT_DIM.to_string(),
+                    "128".to_string(),
+                )])),
+                ..Default::default()
+            };
+
+            let result = ParserContext::parse_column_extensions(
+                &mut parser,
+                &name,
+                &data_type,
+                &mut extensions,
+            );
+            assert!(result.is_ok());
+            assert!(extensions.vector_index_options.is_some());
+            let vi_options = extensions.vector_index_options.unwrap();
+            assert_eq!(vi_options.get("metric"), Some("l2sq"));
+        }
+
+        // Test vector index on non-vector type (should fail)
+        {
+            let sql = "VECTOR INDEX";
+            let dialect = GenericDialect {};
+            let mut tokenizer = Tokenizer::new(&dialect, sql);
+            let tokens = tokenizer.tokenize().unwrap();
+            let mut parser = Parser::new(&dialect).with_tokens(tokens);
+            let name = Ident::new("num_col");
+            let data_type = DataType::Int(None); // Non-vector type
+            let mut extensions = ColumnExtensions::default();
+            let result = ParserContext::parse_column_extensions(
+                &mut parser,
+                &name,
+                &data_type,
+                &mut extensions,
+            );
+            assert!(result.is_err());
+            assert!(
+                result
+                    .unwrap_err()
+                    .to_string()
+                    .contains("VECTOR INDEX only supports Vector type columns")
+            );
+        }
+    }
 }
--- a/src/sql/src/parsers/utils.rs
+++ b/src/sql/src/parsers/utils.rs
@@ -222,6 +222,29 @@ pub fn validate_column_skipping_index_create_option(key: &str) -> bool {
    .contains(&key)
 }

+/// Valid options for VECTOR INDEX:
+/// - engine: Vector index engine (usearch)
+/// - metric: Distance metric (l2sq, cosine, inner_product)
+/// - connectivity: HNSW M parameter
+/// - expansion_add: ef_construction parameter
+/// - expansion_search: ef_search parameter
+pub const COLUMN_VECTOR_INDEX_OPT_KEY_ENGINE: &str = "engine";
+pub const COLUMN_VECTOR_INDEX_OPT_KEY_METRIC: &str = "metric";
+pub const COLUMN_VECTOR_INDEX_OPT_KEY_CONNECTIVITY: &str = "connectivity";
+pub const COLUMN_VECTOR_INDEX_OPT_KEY_EXPANSION_ADD: &str = "expansion_add";
+pub const COLUMN_VECTOR_INDEX_OPT_KEY_EXPANSION_SEARCH: &str = "expansion_search";
+
+pub fn validate_column_vector_index_create_option(key: &str) -> bool {
+    [
+        COLUMN_VECTOR_INDEX_OPT_KEY_ENGINE,
+        COLUMN_VECTOR_INDEX_OPT_KEY_METRIC,
+        COLUMN_VECTOR_INDEX_OPT_KEY_CONNECTIVITY,
+        COLUMN_VECTOR_INDEX_OPT_KEY_EXPANSION_ADD,
+        COLUMN_VECTOR_INDEX_OPT_KEY_EXPANSION_SEARCH,
+    ]
+    .contains(&key)
+}
+
 /// Convert an [`IntervalMonthDayNano`] to a [`Duration`].
 #[cfg(feature = "enterprise")]
 pub fn convert_month_day_nano_to_duration(
--- a/src/sql/src/statements.rs
+++ b/src/sql/src/statements.rs
@@ -55,7 +55,7 @@ use crate::ast::{
 use crate::error::{
    self, ConvertToGrpcDataTypeSnafu, ConvertValueSnafu, Result,
    SerializeColumnDefaultConstraintSnafu, SetFulltextOptionSnafu, SetJsonStructureSettingsSnafu,
-    SetSkippingIndexOptionSnafu, SqlCommonSnafu,
+    SetSkippingIndexOptionSnafu, SetVectorIndexOptionSnafu, SqlCommonSnafu,
 };
 use crate::statements::create::{Column, ColumnExtensions};
 pub use crate::statements::option_map::OptionMap;
@@ -147,6 +147,12 @@ pub fn column_to_schema(
            .context(SetSkippingIndexOptionSnafu)?;
    }

+    if let Some(options) = column.extensions.build_vector_index_options()? {
+        column_schema = column_schema
+            .with_vector_index_options(&options)
+            .context(SetVectorIndexOptionSnafu)?;
+    }
+
    column_schema.set_inverted_index(column.extensions.inverted_index_options.is_some());

    if matches!(column.data_type(), SqlDataType::JSON) {
@@ -710,6 +716,7 @@ mod tests {
                skipping_index_options: None,
                inverted_index_options: None,
                json_datatype_options: None,
+                vector_index_options: None,
            },
        };

@@ -720,4 +727,82 @@ mod tests {
        assert_eq!(fulltext_options.analyzer, FulltextAnalyzer::English);
        assert!(fulltext_options.case_sensitive);
    }
+
+    #[test]
+    fn test_column_to_schema_with_vector_index() {
+        use datatypes::schema::{VectorDistanceMetric, VectorIndexEngineType};
+
+        // Test with custom metric and parameters
+        let column = Column {
+            column_def: ColumnDef {
+                name: "embedding".into(),
+                data_type: SqlDataType::Custom(
+                    vec![Ident::new(VECTOR_TYPE_NAME)].into(),
+                    vec!["128".to_string()],
+                ),
+                options: vec![],
+            },
+            extensions: ColumnExtensions {
+                fulltext_index_options: None,
+                vector_options: None,
+                skipping_index_options: None,
+                inverted_index_options: None,
+                json_datatype_options: None,
+                vector_index_options: Some(OptionMap::from([
+                    ("metric".to_string(), "cosine".to_string()),
+                    ("connectivity".to_string(), "32".to_string()),
+                    ("expansion_add".to_string(), "200".to_string()),
+                    ("expansion_search".to_string(), "100".to_string()),
+                ])),
+            },
+        };
+
+        let column_schema = column_to_schema(&column, "ts", None).unwrap();
+        assert_eq!("embedding", column_schema.name);
+        assert!(column_schema.is_vector_indexed());
+
+        let vector_options = column_schema.vector_index_options().unwrap().unwrap();
+        assert_eq!(vector_options.engine, VectorIndexEngineType::Usearch);
+        assert_eq!(vector_options.metric, VectorDistanceMetric::Cosine);
+        assert_eq!(vector_options.connectivity, 32);
+        assert_eq!(vector_options.expansion_add, 200);
+        assert_eq!(vector_options.expansion_search, 100);
+    }
+
+    #[test]
+    fn test_column_to_schema_with_vector_index_defaults() {
+        use datatypes::schema::{VectorDistanceMetric, VectorIndexEngineType};
+
+        // Test with default values (empty options map)
+        let column = Column {
+            column_def: ColumnDef {
+                name: "vec".into(),
+                data_type: SqlDataType::Custom(
+                    vec![Ident::new(VECTOR_TYPE_NAME)].into(),
+                    vec!["64".to_string()],
+                ),
+                options: vec![],
+            },
+            extensions: ColumnExtensions {
+                fulltext_index_options: None,
+                vector_options: None,
+                skipping_index_options: None,
+                inverted_index_options: None,
+                json_datatype_options: None,
+                vector_index_options: Some(OptionMap::default()),
+            },
+        };
+
+        let column_schema = column_to_schema(&column, "ts", None).unwrap();
+        assert_eq!("vec", column_schema.name);
+        assert!(column_schema.is_vector_indexed());
+
+        let vector_options = column_schema.vector_index_options().unwrap().unwrap();
+        // Verify defaults
+        assert_eq!(vector_options.engine, VectorIndexEngineType::Usearch);
+        assert_eq!(vector_options.metric, VectorDistanceMetric::L2sq);
+        assert_eq!(vector_options.connectivity, 16);
+        assert_eq!(vector_options.expansion_add, 128);
+        assert_eq!(vector_options.expansion_search, 64);
+    }
 }
--- a/src/sql/src/statements/create.rs
+++ b/src/sql/src/statements/create.rs
@@ -17,7 +17,10 @@ use std::fmt::{Display, Formatter};

 use common_catalog::consts::FILE_ENGINE;
 use datatypes::json::JsonStructureSettings;
-use datatypes::schema::{FulltextOptions, SkippingIndexOptions};
+use datatypes::schema::{
+    FulltextOptions, SkippingIndexOptions, VectorDistanceMetric, VectorIndexEngineType,
+    VectorIndexOptions,
+};
 use itertools::Itertools;
 use serde::Serialize;
 use snafu::ResultExt;
@@ -133,6 +136,8 @@ pub struct ColumnExtensions {
    ///
    /// Inverted index doesn't have options at present. There won't be any options in that map.
    pub inverted_index_options: Option<OptionMap>,
+    /// Vector index options for HNSW-based vector similarity search.
+    pub vector_index_options: Option<OptionMap>,
    pub json_datatype_options: Option<OptionMap>,
 }

@@ -208,6 +213,15 @@ impl Display for Column {
                write!(f, " INVERTED INDEX")?;
            }
        }
+
+        if let Some(vector_index_options) = &self.extensions.vector_index_options {
+            if !vector_index_options.is_empty() {
+                let options = vector_index_options.kv_pairs();
+                write!(f, " VECTOR INDEX WITH({})", format_list_comma!(options))?;
+            } else {
+                write!(f, " VECTOR INDEX")?;
+            }
+        }
        Ok(())
    }
 }
@@ -233,6 +247,89 @@ impl ColumnExtensions {
        ))
    }

+    pub fn build_vector_index_options(&self) -> Result<Option<VectorIndexOptions>> {
+        let Some(options) = self.vector_index_options.as_ref() else {
+            return Ok(None);
+        };
+
+        let options_map: HashMap<String, String> = options.clone().into_map();
+        let mut result = VectorIndexOptions::default();
+
+        if let Some(s) = options_map.get("engine") {
+            result.engine = s.parse::<VectorIndexEngineType>().map_err(|e| {
+                InvalidSqlSnafu {
+                    msg: format!("invalid VECTOR INDEX engine: {e}"),
+                }
+                .build()
+            })?;
+        }
+
+        if let Some(s) = options_map.get("metric") {
+            result.metric = s.parse::<VectorDistanceMetric>().map_err(|e| {
+                InvalidSqlSnafu {
+                    msg: format!("invalid VECTOR INDEX metric: {e}"),
+                }
+                .build()
+            })?;
+        }
+
+        if let Some(s) = options_map.get("connectivity") {
+            let value = s.parse::<u32>().map_err(|_| {
+                InvalidSqlSnafu {
+                    msg: format!(
+                        "invalid VECTOR INDEX connectivity: {s}, expected positive integer"
+                    ),
+                }
+                .build()
+            })?;
+            if !(2..=2048).contains(&value) {
+                return InvalidSqlSnafu {
+                    msg: "VECTOR INDEX connectivity must be in the range [2, 2048].".to_string(),
+                }
+                .fail();
+            }
+            result.connectivity = value;
+        }
+
+        if let Some(s) = options_map.get("expansion_add") {
+            let value = s.parse::<u32>().map_err(|_| {
+                InvalidSqlSnafu {
+                    msg: format!(
+                        "invalid VECTOR INDEX expansion_add: {s}, expected positive integer"
+                    ),
+                }
+                .build()
+            })?;
+            if value == 0 {
+                return InvalidSqlSnafu {
+                    msg: "VECTOR INDEX expansion_add must be greater than 0".to_string(),
+                }
+                .fail();
+            }
+            result.expansion_add = value;
+        }
+
+        if let Some(s) = options_map.get("expansion_search") {
+            let value = s.parse::<u32>().map_err(|_| {
+                InvalidSqlSnafu {
+                    msg: format!(
+                        "invalid VECTOR INDEX expansion_search: {s}, expected positive integer"
+                    ),
+                }
+                .build()
+            })?;
+            if value == 0 {
+                return InvalidSqlSnafu {
+                    msg: "VECTOR INDEX expansion_search must be greater than 0".to_string(),
+                }
+                .fail();
+            }
+            result.expansion_search = value;
+        }
+
+        Ok(Some(result))
+    }
+
    pub fn build_json_structure_settings(&self) -> Result<Option<JsonStructureSettings>> {
        let Some(options) = self.json_datatype_options.as_ref() else {
            return Ok(None);
@@ -893,4 +990,92 @@ AS SELECT number FROM numbers_input where number > 10"#,
            _ => unreachable!(),
        }
    }
+
+    #[test]
+    fn test_vector_index_options_validation() {
+        use super::{ColumnExtensions, OptionMap};
+
+        // Test zero connectivity should fail
+        let extensions = ColumnExtensions {
+            fulltext_index_options: None,
+            vector_options: None,
+            skipping_index_options: None,
+            inverted_index_options: None,
+            json_datatype_options: None,
+            vector_index_options: Some(OptionMap::from([(
+                "connectivity".to_string(),
+                "0".to_string(),
+            )])),
+        };
+        let result = extensions.build_vector_index_options();
+        assert!(result.is_err());
+        assert!(
+            result
+                .unwrap_err()
+                .to_string()
+                .contains("connectivity must be in the range [2, 2048]")
+        );
+
+        // Test zero expansion_add should fail
+        let extensions = ColumnExtensions {
+            fulltext_index_options: None,
+            vector_options: None,
+            skipping_index_options: None,
+            inverted_index_options: None,
+            json_datatype_options: None,
+            vector_index_options: Some(OptionMap::from([(
+                "expansion_add".to_string(),
+                "0".to_string(),
+            )])),
+        };
+        let result = extensions.build_vector_index_options();
+        assert!(result.is_err());
+        assert!(
+            result
+                .unwrap_err()
+                .to_string()
+                .contains("expansion_add must be greater than 0")
+        );
+
+        // Test zero expansion_search should fail
+        let extensions = ColumnExtensions {
+            fulltext_index_options: None,
+            vector_options: None,
+            skipping_index_options: None,
+            inverted_index_options: None,
+            json_datatype_options: None,
+            vector_index_options: Some(OptionMap::from([(
+                "expansion_search".to_string(),
+                "0".to_string(),
+            )])),
+        };
+        let result = extensions.build_vector_index_options();
+        assert!(result.is_err());
+        assert!(
+            result
+                .unwrap_err()
+                .to_string()
+                .contains("expansion_search must be greater than 0")
+        );
+
+        // Test valid values should succeed
+        let extensions = ColumnExtensions {
+            fulltext_index_options: None,
+            vector_options: None,
+            skipping_index_options: None,
+            inverted_index_options: None,
+            json_datatype_options: None,
+            vector_index_options: Some(OptionMap::from([
+                ("connectivity".to_string(), "32".to_string()),
+                ("expansion_add".to_string(), "200".to_string()),
+                ("expansion_search".to_string(), "100".to_string()),
+            ])),
+        };
+        let result = extensions.build_vector_index_options();
+        assert!(result.is_ok());
+        let options = result.unwrap().unwrap();
+        assert_eq!(options.connectivity, 32);
+        assert_eq!(options.expansion_add, 200);
+        assert_eq!(options.expansion_search, 100);
+    }
 }
--- a/src/store-api/src/storage.rs
+++ b/src/store-api/src/storage.rs
@@ -27,5 +27,8 @@ pub use datatypes::schema::{

 pub use self::descriptors::*;
 pub use self::file::{FileId, FileRef, FileRefsManifest, GcReport, IndexVersion, ParseIdError};
-pub use self::requests::{ScanRequest, TimeSeriesDistribution, TimeSeriesRowSelector};
+pub use self::requests::{
+    ScanRequest, TimeSeriesDistribution, TimeSeriesRowSelector, VectorDistanceMetric,
+    VectorIndexEngine, VectorIndexEngineType, VectorSearchMatches, VectorSearchRequest,
+};
 pub use self::types::{SequenceNumber, SequenceRange};
--- a/src/store-api/src/storage/requests.rs
+++ b/src/store-api/src/storage/requests.rs
@@ -14,11 +14,66 @@

 use std::fmt::{Display, Formatter};

+use common_error::ext::BoxedError;
 use common_recordbatch::OrderOption;
 use datafusion_expr::expr::Expr;
+// Re-export vector types from datatypes to avoid duplication
+pub use datatypes::schema::{VectorDistanceMetric, VectorIndexEngineType};
 use strum::Display;

-use crate::storage::SequenceNumber;
+use crate::storage::{ColumnId, SequenceNumber};
+
+/// A hint for KNN vector search.
+#[derive(Debug, Clone, PartialEq)]
+pub struct VectorSearchRequest {
+    /// Column ID of the vector column to search.
+    pub column_id: ColumnId,
+    /// The query vector to search for.
+    pub query_vector: Vec<f32>,
+    /// Number of nearest neighbors to return.
+    pub k: usize,
+    /// Distance metric to use (matches the index metric).
+    pub metric: VectorDistanceMetric,
+}
+
+/// Search results from vector index.
+#[derive(Debug, Clone, PartialEq)]
+pub struct VectorSearchMatches {
+    /// Keys (row offsets in the index).
+    pub keys: Vec<u64>,
+    /// Distances from the query vector.
+    pub distances: Vec<f32>,
+}
+
+/// Trait for vector index engines (HNSW implementations).
+///
+/// This trait defines the interface for pluggable vector index engines.
+/// Implementations (e.g., UsearchEngine) are provided by storage engines like mito2.
+pub trait VectorIndexEngine: Send + Sync {
+    /// Adds a vector with the given key.
+    fn add(&mut self, key: u64, vector: &[f32]) -> Result<(), BoxedError>;
+
+    /// Searches for k nearest neighbors.
+    fn search(&self, query: &[f32], k: usize) -> Result<VectorSearchMatches, BoxedError>;
+
+    /// Returns the serialized length.
+    fn serialized_length(&self) -> usize;
+
+    /// Serializes the index to a buffer.
+    fn save_to_buffer(&self, buffer: &mut [u8]) -> Result<(), BoxedError>;
+
+    /// Reserves capacity for vectors.
+    fn reserve(&mut self, capacity: usize) -> Result<(), BoxedError>;
+
+    /// Returns current size (number of vectors).
+    fn size(&self) -> usize;
+
+    /// Returns current capacity.
+    fn capacity(&self) -> usize;
+
+    /// Returns memory usage in bytes.
+    fn memory_usage(&self) -> usize;
+}

 /// A hint on how to select rows from a time-series.
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, Display)]
@@ -38,7 +93,7 @@ pub enum TimeSeriesDistribution {
    PerSeries,
 }

-#[derive(Default, Clone, Debug, PartialEq, Eq)]
+#[derive(Default, Clone, Debug, PartialEq)]
 pub struct ScanRequest {
    /// Indices of columns to read, `None` to read all columns. This indices is
    /// based on table schema.
@@ -66,6 +121,9 @@ pub struct ScanRequest {
    pub sst_min_sequence: Option<SequenceNumber>,
    /// Optional hint for the distribution of time-series data.
    pub distribution: Option<TimeSeriesDistribution>,
+    /// Optional hint for KNN vector search. When set, the scan should use
+    /// vector index to find the k nearest neighbors.
+    pub vector_search: Option<VectorSearchRequest>,
 }

 impl Display for ScanRequest {
@@ -138,6 +196,16 @@ impl Display for ScanRequest {
        if let Some(distribution) = &self.distribution {
            write!(f, "{}distribution: {}", delimiter.as_str(), distribution)?;
        }
+        if let Some(vector_search) = &self.vector_search {
+            write!(
+                f,
+                "{}vector_search: column_id={}, k={}, metric={}",
+                delimiter.as_str(),
+                vector_search.column_id,
+                vector_search.k,
+                vector_search.metric
+            )?;
+        }
        write!(f, " }}")
    }
 }
Author	SHA1	Message	Date
discord9	1afcddd5a9	chore: feature gate vector_index (#7428 ) Signed-off-by: discord9 <discord9@163.com>	2025-12-17 07:14:25 +00:00
shuiyisong	62808b887b	fix: using anonymous s3 access when ak and sk is not provided (#7425 ) * chore: allow s3 anon Signed-off-by: shuiyisong <xixing.sys@gmail.com> * chore: disable ec2 metadata Signed-off-by: shuiyisong <xixing.sys@gmail.com> --------- Signed-off-by: shuiyisong <xixing.sys@gmail.com>	2025-12-17 06:34:29 +00:00
discord9	04ddd40e00	chore: bump version to beta.3 (#7423 ) chore: bump to beta.3 Signed-off-by: discord9 <discord9@163.com>	2025-12-17 04:18:23 +00:00
liyang	b4f028be5f	chore: change etcd endpoints to array in the test scripts (#7419 ) chore: change etcd endpoint Signed-off-by: liyang <daviderli614@gmail.com>	2025-12-17 03:14:35 +00:00
Lei, HUANG	da964880f5	chore: expose symbols (#7417 ) * refactor/expose-symbols: ## Refactor `bulk/part.rs` to Simplify Mutation Handling - Removed the `mutations_to_record_batch` function and its associated helper functions, including `ArraysSorter`, `timestamp_array_to_iter`, and `binary_array_to_dictionary`, to simplify the mutation handling logic in `bulk/part.rs`. - Deleted related test functions `check_binary_array_to_dictionary` and `check_mutations_to_record_batches` from the test module, along with their associated test cases. Signed-off-by: Lei, HUANG <mrsatangel@gmail.com> * refactor/expose-symbols: ### Commit Message Refactor and Enhance Deduplication Logic - `flush.rs`: Refactored `maybe_dedup_one` function to accept `append_mode` and `merge_mode` as parameters instead of `RegionOptions`. This change enhances flexibility in deduplication logic. - `memtable/bulk.rs`: Made `BulkRangeIterBuilder` struct and its fields public to allow external access and modification, improving extensibility. - `sst.rs`: Corrected a typo in the schema documentation, changing `__prmary_key` to `__primary_key` for clarity and accuracy. Signed-off-by: Lei, HUANG <mrsatangel@gmail.com> --------- Signed-off-by: Lei, HUANG <mrsatangel@gmail.com>	2025-12-17 01:29:36 +00:00
dennis zhuang	a35a39f726	feat(vector_index): adds the foundational types and SQL parsing support for vector index (#7366 ) * feat: adds the foundational types and SQL parsing support for vector index Signed-off-by: Dennis Zhuang <killme2008@gmail.com> * refactor: by suggestions Signed-off-by: Dennis Zhuang <killme2008@gmail.com> * fix: ensure index option values must be greater than zero Signed-off-by: Dennis Zhuang <killme2008@gmail.com> * chore: validate connectivity strictly Signed-off-by: Dennis Zhuang <killme2008@gmail.com> * fix: compile error Signed-off-by: Dennis Zhuang <killme2008@gmail.com> * feat: disable SIMD for ci Signed-off-by: Dennis Zhuang <killme2008@gmail.com> --------- Signed-off-by: Dennis Zhuang <killme2008@gmail.com>	2025-12-16 22:45:36 +00:00