diff --git a/Cargo.lock b/Cargo.lock
index ed47b24a64..0f93377962 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -9,13 +9,13 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "57a7559404a7f3573127aab53c08ce37a6c6a315c374a31070f3c91cd1b4a7fe"
 dependencies = [
  "bitflags 1.3.2",
- "bytes",
+ "bytes 1.6.0",
  "futures-core",
  "futures-sink",
  "log",
  "memchr",
- "pin-project-lite",
- "tokio",
+ "pin-project-lite 0.2.12",
+ "tokio 1.37.0",
  "tokio-util",
 ]
 
@@ -45,7 +45,7 @@ dependencies = [
  "actix-utils",
  "actix-web",
  "bitflags 2.4.1",
- "bytes",
+ "bytes 1.6.0",
  "derive_more",
  "futures-core",
  "http-range",
@@ -53,7 +53,7 @@ dependencies = [
  "mime",
  "mime_guess",
  "percent-encoding",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
  "v_htmlescape",
 ]
 
@@ -72,7 +72,7 @@ dependencies = [
  "base64 0.21.0",
  "bitflags 2.4.1",
  "brotli",
- "bytes",
+ "bytes 1.6.0",
  "bytestring",
  "derive_more",
  "encoding_rs",
@@ -87,11 +87,11 @@ dependencies = [
  "local-channel",
  "mime",
  "percent-encoding",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
  "rand 0.8.5",
  "sha1",
  "smallvec",
- "tokio",
+ "tokio 1.37.0",
  "tokio-util",
  "tracing",
  "zstd 0.13.0",
@@ -116,7 +116,7 @@ dependencies = [
  "actix-multipart-derive",
  "actix-utils",
  "actix-web",
- "bytes",
+ "bytes 1.6.0",
  "derive_more",
  "futures-core",
  "futures-util",
@@ -129,7 +129,7 @@ dependencies = [
  "serde_json",
  "serde_plain",
  "tempfile",
- "tokio",
+ "tokio 1.37.0",
 ]
 
 [[package]]
@@ -165,7 +165,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "7ea16c295198e958ef31930a6ef37d0fb64e9ca3b6116e6b93a8bdae96ee1000"
 dependencies = [
  "futures-core",
- "tokio",
+ "tokio 1.37.0",
 ]
 
 [[package]]
@@ -182,7 +182,7 @@ dependencies = [
  "mio",
  "num_cpus",
  "socket2 0.4.9",
- "tokio",
+ "tokio 1.37.0",
  "tracing",
 ]
 
@@ -194,7 +194,7 @@ checksum = "3b894941f818cfdc7ccc4b9e60fa7e53b5042a2e8567270f9147d5591893373a"
 dependencies = [
  "futures-core",
  "paste",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
 ]
 
 [[package]]
@@ -208,9 +208,9 @@ dependencies = [
  "actix-utils",
  "futures-core",
  "impl-more",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
  "rustls-pki-types",
- "tokio",
+ "tokio 1.37.0",
  "tokio-rustls 0.25.0",
  "tokio-util",
  "tracing",
@@ -223,7 +223,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "88a1dcdff1466e3c2488e1cb5c36a71822750ad43839937f85d2f4d9f8b705d8"
 dependencies = [
  "local-waker",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
 ]
 
 [[package]]
@@ -243,7 +243,7 @@ dependencies = [
  "actix-utils",
  "actix-web-codegen",
  "ahash",
- "bytes",
+ "bytes 1.6.0",
  "bytestring",
  "cfg-if",
  "cookie",
@@ -256,7 +256,7 @@ dependencies = [
  "log",
  "mime",
  "once_cell",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
  "regex",
  "serde",
  "serde_json",
@@ -288,7 +288,7 @@ dependencies = [
  "actix-web",
  "futures-core",
  "futures-util",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
 ]
 
 [[package]]
@@ -300,7 +300,7 @@ dependencies = [
  "actix-http",
  "actix-router",
  "actix-web",
- "bytes",
+ "bytes 1.6.0",
  "futures",
  "futures-util",
  "log",
@@ -482,7 +482,7 @@ dependencies = [
  "serde_json",
  "sparse",
  "thiserror",
- "tokio",
+ "tokio 1.37.0",
  "tonic",
  "tonic-build",
  "tracing",
@@ -520,6 +520,28 @@ version = "0.7.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "8da52d66c7071e2e3fa2a1e5c6d088fec47b593032b254f5e980de8ea54454d6"
 
+[[package]]
+name = "async-compression"
+version = "0.3.15"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "942c7cd7ae39e91bde4820d74132e9862e62c2f386c3aa90ccf55949f5bad63a"
+dependencies = [
+ "brotli",
+ "bytes 0.5.6",
+ "bzip2",
+ "flate2",
+ "futures-core",
+ "futures-io",
+ "memchr",
+ "pin-project-lite 0.2.12",
+ "tokio 0.2.25",
+ "tokio 0.3.7",
+ "tokio 1.37.0",
+ "xz2",
+ "zstd 0.11.2+zstd.1.5.2",
+ "zstd-safe 5.0.2+zstd.1.5.2",
+]
+
 [[package]]
 name = "async-stream"
 version = "0.3.3"
@@ -584,7 +606,7 @@ dependencies = [
  "async-trait",
  "axum-core",
  "bitflags 1.3.2",
- "bytes",
+ "bytes 1.6.0",
  "futures-util",
  "http 0.2.9",
  "http-body 0.4.5",
@@ -594,7 +616,7 @@ dependencies = [
  "memchr",
  "mime",
  "percent-encoding",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
  "rustversion",
  "serde",
  "sync_wrapper",
@@ -610,7 +632,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b2f958c80c248b34b9a877a643811be8dbca03ca5ba827f2b63baf3a81e5fc4e"
 dependencies = [
  "async-trait",
- "bytes",
+ "bytes 1.6.0",
  "futures-util",
  "http 0.2.9",
  "http-body 0.4.5",
@@ -797,6 +819,12 @@ version = "1.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "1fd0f2584146f6f2ef48085050886acf353beff7305ebd1ae69500e27c67f64b"
 
+[[package]]
+name = "bytes"
+version = "0.5.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0e4cec68f03f32e44924783795810fa50a7035d8c8ebe78580ad7e6c703fba38"
+
 [[package]]
 name = "bytes"
 version = "1.6.0"
@@ -809,7 +837,7 @@ version = "1.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f7f83e57d9154148e355404702e2694463241880b939570d7c97c014da7a69a1"
 dependencies = [
- "bytes",
+ "bytes 1.6.0",
 ]
 
 [[package]]
@@ -838,7 +866,7 @@ name = "cancel"
 version = "0.0.0"
 dependencies = [
  "thiserror",
- "tokio",
+ "tokio 1.37.0",
  "tokio-util",
 ]
 
@@ -1055,7 +1083,7 @@ dependencies = [
  "async-trait",
  "atomicwrites",
  "bitvec",
- "bytes",
+ "bytes 1.6.0",
  "cancel",
  "chrono",
  "collection",
@@ -1095,7 +1123,7 @@ dependencies = [
  "tempfile",
  "thiserror",
  "tinyvec",
- "tokio",
+ "tokio 1.37.0",
  "tokio-util",
  "tonic",
  "tracing",
@@ -1132,7 +1160,7 @@ dependencies = [
  "serde",
  "thiserror",
  "thread-priority",
- "tokio",
+ "tokio 1.37.0",
  "validator",
 ]
 
@@ -1199,7 +1227,7 @@ dependencies = [
  "serde",
  "serde_json",
  "thread_local",
- "tokio",
+ "tokio 1.37.0",
  "tokio-stream",
  "tonic",
  "tracing",
@@ -1521,6 +1549,18 @@ dependencies = [
  "parking_lot_core",
 ]
 
+[[package]]
+name = "dataset"
+version = "0.0.0"
+dependencies = [
+ "anyhow",
+ "async-compression",
+ "flate2",
+ "indicatif",
+ "project-root",
+ "reqwest",
+]
+
 [[package]]
 name = "debugid"
 version = "0.8.0"
@@ -1787,9 +1827,9 @@ checksum = "0ce7134b9999ecaf8bcd65542e436736ef32ddca1b3e06094cb6ec5755203b80"
 
 [[package]]
 name = "flate2"
-version = "1.0.28"
+version = "1.0.30"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "46303f565772937ffe1d394a4fac6f411c6013172fadde9dcdb1e147a086940e"
+checksum = "5f54427cfd1c7829e2a139fcefea601bf088ebca651d2bf53ebc600eac295dae"
 dependencies = [
  "crc32fast",
  "miniz_oxide 0.7.1",
@@ -1967,7 +2007,7 @@ dependencies = [
  "futures-sink",
  "futures-task",
  "memchr",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
  "pin-utils",
  "slab",
 ]
@@ -2126,7 +2166,7 @@ version = "0.3.26"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "81fe527a889e1532da5c525686d96d4c2e74cdd345badf8dfef9f6b39dd5f5e8"
 dependencies = [
- "bytes",
+ "bytes 1.6.0",
  "fnv",
  "futures-core",
  "futures-sink",
@@ -2134,7 +2174,7 @@ dependencies = [
  "http 0.2.9",
  "indexmap 2.2.6",
  "slab",
- "tokio",
+ "tokio 1.37.0",
  "tokio-util",
  "tracing",
 ]
@@ -2145,7 +2185,7 @@ version = "0.4.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "816ec7294445779408f36fe57bc5b7fc1cf59664059096c65f905c1c61f58069"
 dependencies = [
- "bytes",
+ "bytes 1.6.0",
  "fnv",
  "futures-core",
  "futures-sink",
@@ -2153,7 +2193,7 @@ dependencies = [
  "http 1.0.0",
  "indexmap 2.2.6",
  "slab",
- "tokio",
+ "tokio 1.37.0",
  "tokio-util",
  "tracing",
 ]
@@ -2266,7 +2306,7 @@ version = "0.2.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "bd6effc99afb63425aff9b05836f029929e345a6148a14b7ecd5ab67af944482"
 dependencies = [
- "bytes",
+ "bytes 1.6.0",
  "fnv",
  "itoa",
 ]
@@ -2277,7 +2317,7 @@ version = "1.0.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b32afd38673a8016f7c9ae69e5af41a58f81b1d31689040f2f1959594ce194ea"
 dependencies = [
- "bytes",
+ "bytes 1.6.0",
  "fnv",
  "itoa",
 ]
@@ -2288,9 +2328,9 @@ version = "0.4.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d5f38f16d184e36f2408a55281cd658ecbd3ca05cce6d6510a176eca393e26d1"
 dependencies = [
- "bytes",
+ "bytes 1.6.0",
  "http 0.2.9",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
 ]
 
 [[package]]
@@ -2299,7 +2339,7 @@ version = "1.0.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "1cac85db508abc24a2e48553ba12a996e87244a0395ce011e62b37158745d643"
 dependencies = [
- "bytes",
+ "bytes 1.6.0",
  "http 1.0.0",
 ]
 
@@ -2309,11 +2349,11 @@ version = "0.1.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0475f8b2ac86659c21b64320d5d653f9efe42acd2a4e560073ec61a155a34f1d"
 dependencies = [
- "bytes",
+ "bytes 1.6.0",
  "futures-core",
  "http 1.0.0",
  "http-body 1.0.0",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
 ]
 
 [[package]]
@@ -2356,7 +2396,7 @@ version = "0.14.26"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ab302d72a6f11a3b910431ff93aae7e773078c769f0a3ef15fb9ec692ed147d4"
 dependencies = [
- "bytes",
+ "bytes 1.6.0",
  "futures-channel",
  "futures-core",
  "futures-util",
@@ -2366,9 +2406,9 @@ dependencies = [
  "httparse",
  "httpdate",
  "itoa",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
  "socket2 0.4.9",
- "tokio",
+ "tokio 1.37.0",
  "tower-service",
  "tracing",
  "want",
@@ -2380,7 +2420,7 @@ version = "1.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "186548d73ac615b32a73aafe38fb4f56c0d340e110e5a200bcadbaf2e199263a"
 dependencies = [
- "bytes",
+ "bytes 1.6.0",
  "futures-channel",
  "futures-util",
  "h2 0.4.4",
@@ -2388,9 +2428,9 @@ dependencies = [
  "http-body 1.0.0",
  "httparse",
  "itoa",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
  "smallvec",
- "tokio",
+ "tokio 1.37.0",
  "want",
 ]
 
@@ -2406,7 +2446,7 @@ dependencies = [
  "hyper-util",
  "rustls 0.22.4",
  "rustls-pki-types",
- "tokio",
+ "tokio 1.37.0",
  "tokio-rustls 0.25.0",
  "tower-service",
 ]
@@ -2418,8 +2458,8 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "bbb958482e8c7be4bc3cf272a766a2b0bf1a6755e7a6ae777f017a31d11b13b1"
 dependencies = [
  "hyper 0.14.26",
- "pin-project-lite",
- "tokio",
+ "pin-project-lite 0.2.12",
+ "tokio 1.37.0",
  "tokio-io-timeout",
 ]
 
@@ -2429,15 +2469,15 @@ version = "0.1.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ca38ef113da30126bbff9cd1705f9273e15d45498615d138b0c20279ac7a76aa"
 dependencies = [
- "bytes",
+ "bytes 1.6.0",
  "futures-channel",
  "futures-util",
  "http 1.0.0",
  "http-body 1.0.0",
  "hyper 1.2.0",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
  "socket2 0.5.5",
- "tokio",
+ "tokio 1.37.0",
  "tower",
  "tower-service",
  "tracing",
@@ -3101,6 +3141,17 @@ dependencies = [
  "tracing-subscriber",
 ]
 
+[[package]]
+name = "lzma-sys"
+version = "0.1.20"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5fda04ab3764e6cde78b9974eec4f779acaba7c4e84b36eca3cf77c581b85d27"
+dependencies = [
+ "cc",
+ "libc",
+ "pkg-config",
+]
+
 [[package]]
 name = "macro_rules_attribute"
 version = "0.2.0"
@@ -3653,6 +3704,12 @@ dependencies = [
  "syn 1.0.107",
 ]
 
+[[package]]
+name = "pin-project-lite"
+version = "0.1.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "257b64915a082f7811703966789728173279bdebb956b143dbcd23f6f970a777"
+
 [[package]]
 name = "pin-project-lite"
 version = "0.2.12"
@@ -3818,6 +3875,12 @@ dependencies = [
  "hex",
 ]
 
+[[package]]
+name = "project-root"
+version = "0.2.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "8bccbff07d5ed689c4087d20d7307a52ab6141edeedf487c3876a55b86cf63df"
+
 [[package]]
 name = "prometheus"
 version = "0.13.4"
@@ -3858,7 +3921,7 @@ version = "0.11.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0b82eaa1d779e9a4bc1c3217db8ffbeabaae1dca241bf70183242128d48681cd"
 dependencies = [
- "bytes",
+ "bytes 1.6.0",
  "prost-derive 0.11.9",
 ]
 
@@ -3868,7 +3931,7 @@ version = "0.12.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "146c289cda302b98a28d40c8b3b90498d6e526dd24ac2ecea73e4e491685b94a"
 dependencies = [
- "bytes",
+ "bytes 1.6.0",
  "prost-derive 0.12.3",
 ]
 
@@ -3878,7 +3941,7 @@ version = "0.11.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "119533552c9a7ffacc21e099c24a0ac8bb19c2a2a3f363de84cd9b844feab270"
 dependencies = [
- "bytes",
+ "bytes 1.6.0",
  "heck 0.4.1",
  "itertools 0.10.5",
  "lazy_static",
@@ -3902,7 +3965,7 @@ version = "0.12.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c55e02e35260070b6f716a2423c2ff1c3bb1642ddca6f99e1f26d06268a0e2d2"
 dependencies = [
- "bytes",
+ "bytes 1.6.0",
  "heck 0.4.1",
  "itertools 0.10.5",
  "log",
@@ -4101,7 +4164,7 @@ dependencies = [
  "tempfile",
  "thiserror",
  "tikv-jemallocator",
- "tokio",
+ "tokio 1.37.0",
  "tonic",
  "tonic-reflection",
  "tower",
@@ -4364,7 +4427,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "566cafdd92868e0939d3fb961bd0dc25fcfaaed179291093b3d43e6b3150ea10"
 dependencies = [
  "base64 0.22.0",
- "bytes",
+ "bytes 1.6.0",
  "futures-channel",
  "futures-core",
  "futures-util",
@@ -4381,7 +4444,7 @@ dependencies = [
  "mime",
  "once_cell",
  "percent-encoding",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
  "rustls 0.22.4",
  "rustls-pemfile 2.1.2",
  "rustls-pki-types",
@@ -4389,7 +4452,7 @@ dependencies = [
  "serde_json",
  "serde_urlencoded",
  "sync_wrapper",
- "tokio",
+ "tokio 1.37.0",
  "tokio-rustls 0.25.0",
  "tokio-util",
  "tower-service",
@@ -4865,12 +4928,14 @@ dependencies = [
  "chrono",
  "common",
  "criterion",
+ "dataset",
  "fnv",
  "fs_extra",
  "generic-tests",
  "geo",
  "geohash",
  "indexmap 2.2.6",
+ "indicatif",
  "io",
  "io-uring",
  "is_sorted",
@@ -5213,15 +5278,21 @@ name = "sparse"
 version = "0.1.0"
 dependencies = [
  "common",
+ "criterion",
+ "dataset",
+ "indicatif",
  "io",
  "itertools 0.12.1",
  "memmap2 0.9.4",
  "memory",
  "ordered-float 4.2.0",
  "parking_lot",
+ "pprof",
  "rand 0.8.5",
  "schemars",
  "serde",
+ "serde_json",
+ "sparse",
  "tempfile",
  "validator",
 ]
@@ -5286,7 +5357,7 @@ dependencies = [
  "tar",
  "tempfile",
  "thiserror",
- "tokio",
+ "tokio 1.37.0",
  "tonic",
  "tracing",
  "url",
@@ -5602,6 +5673,26 @@ version = "0.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "cda74da7e1a664f795bb1f8a87ec406fb89a02522cf6e50620d016add6dbbf5c"
 
+[[package]]
+name = "tokio"
+version = "0.2.25"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6703a273949a90131b290be1fe7b039d0fc884aa1935860dfcbe056f28cd8092"
+dependencies = [
+ "bytes 0.5.6",
+ "pin-project-lite 0.1.12",
+]
+
+[[package]]
+name = "tokio"
+version = "0.3.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "46409491c9375a693ce7032101970a54f8a2010efb77e13f70788f0d84489e39"
+dependencies = [
+ "autocfg",
+ "pin-project-lite 0.2.12",
+]
+
 [[package]]
 name = "tokio"
 version = "1.37.0"
@@ -5609,12 +5700,12 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "1adbebffeca75fcfd058afa480fb6c0b81e165a0323f9c9d39c9697e37c46787"
 dependencies = [
  "backtrace",
- "bytes",
+ "bytes 1.6.0",
  "libc",
  "mio",
  "num_cpus",
  "parking_lot",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
  "signal-hook-registry",
  "socket2 0.5.5",
  "tokio-macros",
@@ -5628,8 +5719,8 @@ version = "1.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "30b74022ada614a1b4834de765f9bb43877f910cc8ce4be40e89042c9223a8bf"
 dependencies = [
- "pin-project-lite",
- "tokio",
+ "pin-project-lite 0.2.12",
+ "tokio 1.37.0",
 ]
 
 [[package]]
@@ -5650,7 +5741,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c28327cf380ac148141087fbfb9de9d7bd4e84ab5d2c28fbc911d753de8a7081"
 dependencies = [
  "rustls 0.21.11",
- "tokio",
+ "tokio 1.37.0",
 ]
 
 [[package]]
@@ -5661,7 +5752,7 @@ checksum = "775e0c0f0adb3a2f22a00c4745d728b479985fc15ee7ca6a2608388c5569860f"
 dependencies = [
  "rustls 0.22.4",
  "rustls-pki-types",
- "tokio",
+ "tokio 1.37.0",
 ]
 
 [[package]]
@@ -5671,8 +5762,8 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "d660770404473ccd7bc9f8b28494a811bc18542b915c0855c51e8f419d5223ce"
 dependencies = [
  "futures-core",
- "pin-project-lite",
- "tokio",
+ "pin-project-lite 0.2.12",
+ "tokio 1.37.0",
 ]
 
 [[package]]
@@ -5681,11 +5772,11 @@ version = "0.7.11"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9cf6b47b3771c49ac75ad09a6162f53ad4b8088b76ac60e8ec1455b31a189fe1"
 dependencies = [
- "bytes",
+ "bytes 1.6.0",
  "futures-core",
  "futures-sink",
- "pin-project-lite",
- "tokio",
+ "pin-project-lite 0.2.12",
+ "tokio 1.37.0",
 ]
 
 [[package]]
@@ -5740,7 +5831,7 @@ dependencies = [
  "async-trait",
  "axum",
  "base64 0.21.0",
- "bytes",
+ "bytes 1.6.0",
  "flate2",
  "futures-core",
  "futures-util",
@@ -5753,7 +5844,7 @@ dependencies = [
  "pin-project",
  "prost 0.11.9",
  "rustls-pemfile 1.0.3",
- "tokio",
+ "tokio 1.37.0",
  "tokio-rustls 0.24.1",
  "tokio-stream",
  "tower",
@@ -5783,7 +5874,7 @@ checksum = "0543d7092032041fbeac1f2c84304537553421a11a623c2301b12ef0264862c7"
 dependencies = [
  "prost 0.11.9",
  "prost-types 0.11.9",
- "tokio",
+ "tokio 1.37.0",
  "tokio-stream",
  "tonic",
 ]
@@ -5798,10 +5889,10 @@ dependencies = [
  "futures-util",
  "indexmap 1.9.2",
  "pin-project",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
  "rand 0.8.5",
  "slab",
- "tokio",
+ "tokio 1.37.0",
  "tokio-util",
  "tower-layer",
  "tower-service",
@@ -5827,7 +5918,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c3523ab5a71916ccf420eebdf5521fcef02141234bbc0b8a49f2fdc4544364ef"
 dependencies = [
  "log",
- "pin-project-lite",
+ "pin-project-lite 0.2.12",
  "tracing-attributes",
  "tracing-core",
 ]
@@ -6635,6 +6726,15 @@ dependencies = [
  "libc",
 ]
 
+[[package]]
+name = "xz2"
+version = "0.1.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "388c44dc09d76f1536602ead6d325eb532f5c122f17782bd57fb47baeeb767e2"
+dependencies = [
+ "lzma-sys",
+]
+
 [[package]]
 name = "yada"
 version = "0.5.0"
diff --git a/Cargo.toml b/Cargo.toml
index 804fa8e027..154ca19e1f 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -123,7 +123,9 @@ fnv = "1.0"
 futures = "0.3.30"
 futures-util = "0.3.30"
 indexmap = { version = "2", features = ["serde"] }
+indicatif = "0.17.8"
 parking_lot = { version = "0.12.2", features = ["deadlock_detection", "serde"] }
+pprof = { version = "0.12", features = ["flamegraph", "prost-codec"] }
 prost = "0.11.9"
 prost-wkt-types = "0.4.2"
 reqwest = { version = "0.12.4", default-features = false, features = ["http2", "stream", "rustls-tls", "blocking"] }
diff --git a/lib/collection/Cargo.toml b/lib/collection/Cargo.toml
index 29f5a761f1..d09d709ff0 100644
--- a/lib/collection/Cargo.toml
+++ b/lib/collection/Cargo.toml
@@ -21,7 +21,7 @@ collection = { path = ".", features = ["testing"] }
 common = { path = "../common/common", features = ["testing"] }
 
 [target.'cfg(not(target_os = "windows"))'.dev-dependencies]
-pprof = { version = "0.12", features = ["flamegraph", "prost-codec"] }
+pprof = { workspace = true }
 
 [dependencies]
 parking_lot = { workspace = true }
@@ -63,7 +63,7 @@ sparse = { path = "../sparse" }
 api = { path = "../api" }
 
 itertools = "0.12"
-indicatif = "0.17.8"
+indicatif = { workspace = true }
 chrono = { workspace = true }
 schemars = { workspace = true }
 tar = { workspace = true }
diff --git a/lib/common/dataset/Cargo.toml b/lib/common/dataset/Cargo.toml
new file mode 100644
index 0000000000..83b0d51a58
--- /dev/null
+++ b/lib/common/dataset/Cargo.toml
@@ -0,0 +1,18 @@
+[package]
+name = "dataset"
+version = "0.0.0"
+authors = [
+    "Andrey Vasnetsov <andrey@qdrant.com>",
+    "Qdrant Team <info@qdrant.tech>",
+]
+license = "Apache-2.0"
+edition = "2021"
+publish = false
+
+[dependencies]
+anyhow = "1.0.82"
+async-compression = {version="0.3.13", features=["all", "stream"]}
+flate2 = { version = "1.0.30" }
+indicatif = { workspace = true }
+project-root = "0.2.2"
+reqwest = { workspace = true }
diff --git a/lib/common/dataset/src/lib.rs b/lib/common/dataset/src/lib.rs
new file mode 100644
index 0000000000..a8a7c921f9
--- /dev/null
+++ b/lib/common/dataset/src/lib.rs
@@ -0,0 +1,108 @@
+use std::fs::{create_dir_all, File};
+use std::path::{Path, PathBuf};
+
+use anyhow::{anyhow, Context, Result};
+use flate2::read::GzDecoder;
+use indicatif::{ProgressBar, ProgressDrawTarget};
+
+pub enum Dataset {
+    // https://github.com/qdrant/sparse-vectors-experiments
+    SpladeWikiMovies,
+
+    // https://github.com/qdrant/sparse-vectors-benchmark
+    NeurIps2023Full,
+    NeurIps2023_1M,
+    NeurIps2023Small,
+    NeurIps2023Queries,
+}
+
+impl Dataset {
+    pub fn download(&self) -> Result<PathBuf> {
+        download_cached(&self.url())
+    }
+
+    fn url(&self) -> String {
+        const NEUR_IPS_2023_BASE: &str =
+            "https://storage.googleapis.com/ann-challenge-sparse-vectors/csr";
+        match self {
+            Dataset::SpladeWikiMovies => {
+                "https://storage.googleapis.com/dataset-sparse-vectors/sparse-vectors.jsonl.gz"
+                    .to_string()
+            }
+            Dataset::NeurIps2023Full => format!("{NEUR_IPS_2023_BASE}/base_full.csr.gz"),
+            Dataset::NeurIps2023_1M => format!("{NEUR_IPS_2023_BASE}/base_1M.csr.gz"),
+            Dataset::NeurIps2023Small => format!("{NEUR_IPS_2023_BASE}/base_small.csr.gz"),
+            Dataset::NeurIps2023Queries => format!("{NEUR_IPS_2023_BASE}/queries.dev.csr.gz"),
+        }
+    }
+}
+
+fn download_cached(url: &str) -> Result<PathBuf> {
+    // Filename without an ".gz" extension, e.g. "base_full.csr".
+    let basename = {
+        let path = Path::new(url);
+        match path.extension() {
+            Some(gz) if gz == "gz" => path.file_stem(),
+            _ => path.file_name(),
+        }
+        .ok_or_else(|| anyhow!("Failed to extract basename from {url}"))?
+    };
+
+    // Cache directory, e.g. "target/datasets".
+    let cache_dir = workspace_dir()
+        .join(std::env::var_os("CARGO_TARGET_DIR").unwrap_or_else(|| "target".into()))
+        .join("datasets");
+
+    // Cache file path, e.g. "target/datasets/base_full.csr".
+    let cache_path = cache_dir.join(basename);
+
+    if cache_path.exists() {
+        return Ok(cache_path);
+    }
+
+    eprintln!("Downloading {url} to {cache_path:?}...");
+
+    create_dir_all(cache_dir)?;
+
+    let resp = reqwest::blocking::get(url)?;
+    if !resp.status().is_success() {
+        anyhow::bail!("Failed to download {url}, status: {}", resp.status());
+    }
+    let total_size = resp.content_length();
+
+    // Download to a temporary file, e.g. "target/datasets/base_full.csr.tmp", to avoid
+    // incomplete files.
+    let mut tmp_fname = cache_path.clone().into_os_string();
+    tmp_fname.push(".tmp");
+
+    // Progress bar.
+    let pb = ProgressBar::with_draw_target(total_size, ProgressDrawTarget::stderr_with_hz(12));
+    pb.set_style(
+        indicatif::ProgressStyle::default_bar()
+            .template("{msg} {wide_bar} {bytes}/{total_bytes} (eta:{eta})")
+            .expect("failed to set style"),
+    );
+
+    // Download + decompress.
+    std::io::copy(
+        &mut GzDecoder::new(pb.wrap_read(resp)),
+        &mut File::create(&tmp_fname)?,
+    )?;
+
+    std::fs::rename(&tmp_fname, &cache_path)
+        .with_context(|| format!("Failed to rename {tmp_fname:?} to {cache_path:?}"))?;
+
+    Ok(cache_path)
+}
+
+fn workspace_dir() -> PathBuf {
+    let output = std::process::Command::new(env!("CARGO"))
+        .arg("locate-project")
+        .arg("--workspace")
+        .arg("--message-format=plain")
+        .output()
+        .unwrap()
+        .stdout;
+    let cargo_path = Path::new(std::str::from_utf8(&output).unwrap().trim());
+    cargo_path.parent().unwrap().to_path_buf()
+}
diff --git a/lib/segment/Cargo.toml b/lib/segment/Cargo.toml
index 775562cdac..114273e030 100644
--- a/lib/segment/Cargo.toml
+++ b/lib/segment/Cargo.toml
@@ -16,6 +16,8 @@ testing = []
 
 [dev-dependencies]
 criterion = "0.5"
+dataset = { path = "../common/dataset" }
+indicatif = { workspace = true }
 rmp-serde = "~1.3"
 rand_distr = "0.4.3"
 walkdir = "2.5.0"
@@ -24,7 +26,7 @@ segment = { path = ".", features = ["testing"] }
 proptest = "1.4.0"
 
 [target.'cfg(not(target_os = "windows"))'.dev-dependencies]
-pprof = { version = "0.12", features = ["flamegraph", "prost-codec"] }
+pprof = { workspace = true }
 
 [dependencies]
 bitpacking = "0.9.2"
diff --git a/lib/segment/benches/sparse_index_search.rs b/lib/segment/benches/sparse_index_search.rs
index 8adac31a14..23863cc405 100644
--- a/lib/segment/benches/sparse_index_search.rs
+++ b/lib/segment/benches/sparse_index_search.rs
@@ -6,10 +6,13 @@ use std::sync::Arc;
 
 use common::cpu::CpuPermit;
 use common::types::PointOffsetType;
-use criterion::{criterion_group, criterion_main, Criterion};
+use criterion::{criterion_group, criterion_main, BatchSize, Criterion};
+use dataset::Dataset;
+use indicatif::{ProgressBar, ProgressDrawTarget, ProgressStyle};
+use itertools::Itertools as _;
 use rand::rngs::StdRng;
 use rand::SeedableRng;
-use segment::fixtures::sparse_fixtures::fixture_sparse_index_ram;
+use segment::fixtures::sparse_fixtures::fixture_sparse_index_ram_from_iter;
 use segment::index::hnsw_index::num_rayon_threads;
 use segment::index::sparse_index::sparse_index_config::{SparseIndexConfig, SparseIndexType};
 use segment::index::sparse_index::sparse_vector_index::SparseVectorIndex;
@@ -17,29 +20,67 @@ use segment::index::{PayloadIndex, VectorIndex};
 use segment::types::PayloadSchemaType::Keyword;
 use segment::types::{Condition, FieldCondition, Filter, Payload};
 use serde_json::json;
-use sparse::common::sparse_vector_fixture::random_positive_sparse_vector;
+use sparse::common::sparse_vector::SparseVector;
+use sparse::common::sparse_vector_fixture::{random_positive_sparse_vector, random_sparse_vector};
 use sparse::index::inverted_index::inverted_index_mmap::InvertedIndexMmap;
+use sparse::index::loaders::Csr;
 use tempfile::Builder;
 
 const NUM_VECTORS: usize = 50_000;
 const MAX_SPARSE_DIM: usize = 30_000;
+const NUM_QUERIES: usize = 2048;
 const TOP: usize = 10;
 const FULL_SCAN_THRESHOLD: usize = 1; // low value to trigger index usage by default
 
 fn sparse_vector_index_search_benchmark(c: &mut Criterion) {
-    let mut group = c.benchmark_group("sparse-vector-search-group");
+    let mut rnd = StdRng::seed_from_u64(0);
+    let query_vectors = (0..NUM_QUERIES)
+        // Positive values to test pruning.
+        .map(|_| random_positive_sparse_vector(&mut rnd, MAX_SPARSE_DIM))
+        .collect::<Vec<_>>();
+
+    let mut rnd = StdRng::seed_from_u64(42);
+    let random_vectors = (0..NUM_VECTORS).map(|_| random_sparse_vector(&mut rnd, MAX_SPARSE_DIM));
+    sparse_vector_index_search_benchmark_impl(c, "random-50k", random_vectors, &query_vectors);
+
+    let dataset_vectors = Csr::open(Dataset::NeurIps2023_1M.download().unwrap()).unwrap();
+    let query_vectors = Csr::open(Dataset::NeurIps2023Queries.download().unwrap())
+        .unwrap()
+        .iter()
+        .map(|v| v.unwrap())
+        .collect_vec();
+    sparse_vector_index_search_benchmark_impl(
+        c,
+        "neurips2023-1M",
+        dataset_vectors.iter().map(|v| v.unwrap()),
+        &query_vectors,
+    );
+}
+
+fn sparse_vector_index_search_benchmark_impl(
+    c: &mut Criterion,
+    group: &str,
+    vectors: impl ExactSizeIterator<Item = SparseVector>,
+    query_vectors: &[SparseVector],
+) {
+    let mut group = c.benchmark_group(format!("sparse_vector_index_search/{}", group));
+    group.sample_size(10);
+
+    let vectors_len = vectors.len();
 
     let stopped = AtomicBool::new(false);
-    let mut rnd = StdRng::seed_from_u64(42);
 
     let data_dir = Builder::new().prefix("data_dir").tempdir().unwrap();
-    let sparse_vector_index = fixture_sparse_index_ram(
-        &mut rnd,
-        NUM_VECTORS,
-        MAX_SPARSE_DIM,
+
+    let sparse_vector_index = fixture_sparse_index_ram_from_iter(
+        progress("Indexing (1/3)", vectors_len).wrap_iter(vectors),
         FULL_SCAN_THRESHOLD,
         data_dir.path(),
         &stopped,
+        || {
+            let pb = progress("Indexing (2/3)", vectors_len);
+            move || pb.inc(1)
+        },
     );
 
     // adding payload on field
@@ -59,11 +100,7 @@ fn sparse_vector_index_search_benchmark(c: &mut Criterion) {
     }
     drop(payload_index);
 
-    // shared query vector (positive values to test pruning)
-    let vector = random_positive_sparse_vector(&mut rnd, MAX_SPARSE_DIM);
-    eprintln!("sparse_vector size = {:#?}", vector.values.len());
-    let sparse_vector = vector.clone();
-    let query_vector = vector.into();
+    let mut query_vector_it = query_vectors.iter().cycle();
 
     let permit_cpu_count = num_rayon_threads(0);
     let permit = Arc::new(CpuPermit::dummy(permit_cpu_count as u32));
@@ -82,31 +119,41 @@ fn sparse_vector_index_search_benchmark(c: &mut Criterion) {
             &stopped,
         )
         .unwrap();
+    let pb = progress("Indexing (3/3)", vectors_len);
     sparse_vector_index_mmap
-        .build_index(permit, &stopped)
+        .build_index_with_progress(permit, &stopped, || pb.inc(1))
         .unwrap();
-    assert_eq!(sparse_vector_index_mmap.indexed_vector_count(), NUM_VECTORS);
+    pb.finish_and_clear();
+    assert_eq!(sparse_vector_index_mmap.indexed_vector_count(), vectors_len);
 
     // intent: bench `search` without filter on mmap inverted index
     group.bench_function("mmap-inverted-index-search", |b| {
-        b.iter(|| {
-            let results = sparse_vector_index_mmap
-                .search(&[&query_vector], None, TOP, None, &Default::default())
-                .unwrap();
+        b.iter_batched(
+            || query_vector_it.next().unwrap().clone().into(),
+            |vec| {
+                let results = sparse_vector_index_mmap
+                    .search(&[&vec], None, TOP, None, &Default::default())
+                    .unwrap();
 
-            assert_eq!(results[0].len(), TOP);
-        })
+                assert_eq!(results[0].len(), TOP);
+            },
+            BatchSize::SmallInput,
+        )
     });
 
     // intent: bench `search` without filter
     group.bench_function("inverted-index-search", |b| {
-        b.iter(|| {
-            let results = sparse_vector_index
-                .search(&[&query_vector], None, TOP, None, &Default::default())
-                .unwrap();
+        b.iter_batched(
+            || query_vector_it.next().unwrap().clone().into(),
+            |vec| {
+                let results = sparse_vector_index
+                    .search(&[&vec], None, TOP, None, &Default::default())
+                    .unwrap();
 
-            assert_eq!(results[0].len(), TOP);
-        })
+                assert_eq!(results[0].len(), TOP);
+            },
+            BatchSize::SmallInput,
+        )
     });
 
     // filter by field
@@ -116,22 +163,28 @@ fn sparse_vector_index_search_benchmark(c: &mut Criterion) {
     )));
 
     // intent: bench plain search when the filtered payload key is not indexed
-    group.bench_function("inverted-index-filtered-plain", |b| {
-        b.iter(|| {
-            let mut prefiltered_points = None;
-            let results = sparse_vector_index
-                .search_plain(
-                    &sparse_vector,
-                    &filter,
-                    TOP,
-                    &mut prefiltered_points,
-                    &Default::default(),
-                )
-                .unwrap();
+    if vectors_len < 100_000 {
+        group.bench_function("inverted-index-filtered-plain", |b| {
+            b.iter_batched(
+                || query_vector_it.next().unwrap(),
+                |vec| {
+                    let mut prefiltered_points = None;
+                    let results = sparse_vector_index
+                        .search_plain(
+                            vec,
+                            &filter,
+                            TOP,
+                            &mut prefiltered_points,
+                            &Default::default(),
+                        )
+                        .unwrap();
 
-            assert_eq!(results.len(), TOP);
-        })
-    });
+                    assert_eq!(results.len(), TOP);
+                },
+                BatchSize::SmallInput,
+            )
+        });
+    }
 
     let mut payload_index = sparse_vector_index.payload_index.borrow_mut();
 
@@ -144,42 +197,58 @@ fn sparse_vector_index_search_benchmark(c: &mut Criterion) {
 
     // intent: bench `search` when the filtered payload key is indexed
     group.bench_function("inverted-index-filtered-payload-index", |b| {
-        b.iter(|| {
-            let results = sparse_vector_index
-                .search(
-                    &[&query_vector],
-                    Some(&filter),
-                    TOP,
-                    None,
-                    &Default::default(),
-                )
-                .unwrap();
+        b.iter_batched(
+            || query_vector_it.next().unwrap().clone().into(),
+            |vec| {
+                let results = sparse_vector_index
+                    .search(&[&vec], Some(&filter), TOP, None, &Default::default())
+                    .unwrap();
 
-            assert_eq!(results[0].len(), TOP);
-        })
+                assert_eq!(results[0].len(), TOP);
+            },
+            BatchSize::SmallInput,
+        );
     });
 
     // intent: bench plain search when the filtered payload key is indexed
-    group.bench_function("plain-filtered-payload-index", |b| {
-        b.iter(|| {
-            let mut prefiltered_points = None;
-            let results = sparse_vector_index
-                .search_plain(
-                    &sparse_vector,
-                    &filter,
-                    TOP,
-                    &mut prefiltered_points,
-                    &Default::default(),
-                )
-                .unwrap();
+    if vectors_len < 100_000 {
+        group.bench_function("plain-filtered-payload-index", |b| {
+            b.iter_batched(
+                || query_vector_it.next().unwrap(),
+                |vec| {
+                    let mut prefiltered_points = None;
+                    let results = sparse_vector_index
+                        .search_plain(
+                            vec,
+                            &filter,
+                            TOP,
+                            &mut prefiltered_points,
+                            &Default::default(),
+                        )
+                        .unwrap();
 
-            assert_eq!(results.len(), TOP);
-        })
-    });
+                    assert_eq!(results.len(), TOP);
+                },
+                BatchSize::SmallInput,
+            )
+        });
+    }
 
     group.finish();
 }
 
+fn progress(name: &str, len: usize) -> ProgressBar {
+    let pb =
+        ProgressBar::with_draw_target(Some(len as u64), ProgressDrawTarget::stderr_with_hz(12));
+    pb.set_style(
+        ProgressStyle::default_bar()
+            .template("{msg} {wide_bar} {pos}/{len} (eta:{eta})")
+            .unwrap(),
+    );
+    pb.set_message(name.to_owned());
+    pb
+}
+
 #[cfg(not(target_os = "windows"))]
 criterion_group! {
     name = benches;
diff --git a/lib/segment/src/fixtures/sparse_fixtures.rs b/lib/segment/src/fixtures/sparse_fixtures.rs
index 46d221b136..d0d9eb1faa 100644
--- a/lib/segment/src/fixtures/sparse_fixtures.rs
+++ b/lib/segment/src/fixtures/sparse_fixtures.rs
@@ -89,6 +89,24 @@ pub fn fixture_sparse_index_ram<R: Rng + ?Sized>(
     data_dir: &Path,
     stopped: &AtomicBool,
 ) -> SparseVectorIndex<InvertedIndexRam> {
+    fixture_sparse_index_ram_from_iter(
+        (0..num_vectors).map(|_| random_sparse_vector(rnd, max_dim)),
+        full_scan_threshold,
+        data_dir,
+        stopped,
+        || || (),
+    )
+}
+
+/// Prepares a sparse vector index with a given iterator of sparse vectors
+pub fn fixture_sparse_index_ram_from_iter<P: FnMut()>(
+    vectors: impl ExactSizeIterator<Item = SparseVector>,
+    full_scan_threshold: usize,
+    data_dir: &Path,
+    stopped: &AtomicBool,
+    progress: impl FnOnce() -> P,
+) -> SparseVectorIndex<InvertedIndexRam> {
+    let num_vectors = vectors.len();
     let mut sparse_vector_index = fixture_open_sparse_index(
         data_dir,
         num_vectors,
@@ -100,10 +118,9 @@ pub fn fixture_sparse_index_ram<R: Rng + ?Sized>(
     let mut borrowed_storage = sparse_vector_index.vector_storage.borrow_mut();
 
     // add points to storage
-    for idx in 0..num_vectors {
-        let vec = &random_sparse_vector(rnd, max_dim);
+    for (idx, vec) in vectors.enumerate() {
         borrowed_storage
-            .insert_vector(idx as PointOffsetType, vec.into())
+            .insert_vector(idx as PointOffsetType, (&vec).into())
             .unwrap();
     }
     drop(borrowed_storage);
@@ -114,7 +131,7 @@ pub fn fixture_sparse_index_ram<R: Rng + ?Sized>(
             .vector_storage
             .borrow()
             .available_vector_count(),
-        num_vectors
+        num_vectors,
     );
 
     // assert no points are indexed following open for RAM index
@@ -124,7 +141,10 @@ pub fn fixture_sparse_index_ram<R: Rng + ?Sized>(
     let permit = Arc::new(CpuPermit::dummy(permit_cpu_count as u32));
 
     // build index to refresh RAM index
-    sparse_vector_index.build_index(permit, stopped).unwrap();
+    let tick_progress = progress();
+    sparse_vector_index
+        .build_index_with_progress(permit, stopped, tick_progress)
+        .unwrap();
     assert_eq!(sparse_vector_index.indexed_vector_count(), num_vectors);
     sparse_vector_index
 }
diff --git a/lib/segment/src/index/hnsw_index/hnsw.rs b/lib/segment/src/index/hnsw_index/hnsw.rs
index 31edc91997..2931a563ab 100644
--- a/lib/segment/src/index/hnsw_index/hnsw.rs
+++ b/lib/segment/src/index/hnsw_index/hnsw.rs
@@ -665,7 +665,12 @@ impl<TGraphLinks: GraphLinks> VectorIndex for HNSWIndex<TGraphLinks> {
         }
     }
 
-    fn build_index(&mut self, permit: Arc<CpuPermit>, stopped: &AtomicBool) -> OperationResult<()> {
+    fn build_index_with_progress(
+        &mut self,
+        permit: Arc<CpuPermit>,
+        stopped: &AtomicBool,
+        _tick_progress: impl FnMut(),
+    ) -> OperationResult<()> {
         // Build main index graph
         let id_tracker = self.id_tracker.borrow();
         let vector_storage = self.vector_storage.borrow();
diff --git a/lib/segment/src/index/plain_payload_index.rs b/lib/segment/src/index/plain_payload_index.rs
index 31ee799ec7..65e22d8dfa 100644
--- a/lib/segment/src/index/plain_payload_index.rs
+++ b/lib/segment/src/index/plain_payload_index.rs
@@ -317,10 +317,11 @@ impl VectorIndex for PlainIndex {
         }
     }
 
-    fn build_index(
+    fn build_index_with_progress(
         &mut self,
         _permit: Arc<CpuPermit>,
         _stopped: &AtomicBool,
+        _tick_progress: impl FnMut(),
     ) -> OperationResult<()> {
         Ok(())
     }
diff --git a/lib/segment/src/index/sparse_index/sparse_vector_index.rs b/lib/segment/src/index/sparse_index/sparse_vector_index.rs
index f54681370e..53088386c6 100644
--- a/lib/segment/src/index/sparse_index/sparse_vector_index.rs
+++ b/lib/segment/src/index/sparse_index/sparse_vector_index.rs
@@ -71,6 +71,7 @@ impl<TInvertedIndex: InvertedIndex> SparseVectorIndex<TInvertedIndex> {
                 vector_storage.clone(),
                 path,
                 stopped,
+                || (),
             )?;
             (config, inverted_index, indices_tracker)
         } else if config_path.exists() {
@@ -114,6 +115,7 @@ impl<TInvertedIndex: InvertedIndex> SparseVectorIndex<TInvertedIndex> {
         vector_storage: Arc<AtomicRefCell<VectorStorageEnum>>,
         path: &Path,
         stopped: &AtomicBool,
+        mut tick_progress: impl FnMut(),
     ) -> OperationResult<(TInvertedIndex, IndicesTracker)> {
         let borrowed_vector_storage = vector_storage.borrow();
         let borrowed_id_tracker = id_tracker.borrow();
@@ -143,6 +145,7 @@ impl<TInvertedIndex: InvertedIndex> SparseVectorIndex<TInvertedIndex> {
                     ram_index_builder.add(id, vector);
                 }
             }
+            tick_progress();
         }
         Ok((
             TInvertedIndex::from_ram_index(ram_index_builder.build(), path)?,
@@ -440,16 +443,18 @@ impl<TInvertedIndex: InvertedIndex> VectorIndex for SparseVectorIndex<TInvertedI
         Ok(results)
     }
 
-    fn build_index(
+    fn build_index_with_progress(
         &mut self,
         _permit: Arc<CpuPermit>,
         stopped: &AtomicBool,
+        tick_progress: impl FnMut(),
     ) -> OperationResult<()> {
         let (inverted_index, indices_tracker) = Self::build_inverted_index(
             self.id_tracker.clone(),
             self.vector_storage.clone(),
             &self.path,
             stopped,
+            tick_progress,
         )?;
 
         self.inverted_index = inverted_index;
diff --git a/lib/segment/src/index/vector_index_base.rs b/lib/segment/src/index/vector_index_base.rs
index d240d56df0..3ead70bc28 100644
--- a/lib/segment/src/index/vector_index_base.rs
+++ b/lib/segment/src/index/vector_index_base.rs
@@ -30,7 +30,17 @@ pub trait VectorIndex {
     ) -> OperationResult<Vec<Vec<ScoredPointOffset>>>;
 
     /// Force internal index rebuild.
-    fn build_index(&mut self, permit: Arc<CpuPermit>, stopped: &AtomicBool) -> OperationResult<()>;
+    fn build_index(&mut self, permit: Arc<CpuPermit>, stopped: &AtomicBool) -> OperationResult<()> {
+        self.build_index_with_progress(permit, stopped, || ())
+    }
+
+    /// Force internal index rebuild.
+    fn build_index_with_progress(
+        &mut self,
+        permit: Arc<CpuPermit>,
+        stopped: &AtomicBool,
+        tick_progress: impl FnMut(),
+    ) -> OperationResult<()>;
 
     fn get_telemetry_data(&self, detail: TelemetryDetail) -> VectorIndexSearchesTelemetry;
 
@@ -91,13 +101,28 @@ impl VectorIndex for VectorIndexEnum {
         }
     }
 
-    fn build_index(&mut self, permit: Arc<CpuPermit>, stopped: &AtomicBool) -> OperationResult<()> {
+    fn build_index_with_progress(
+        &mut self,
+        permit: Arc<CpuPermit>,
+        stopped: &AtomicBool,
+        tick_progress: impl FnMut(),
+    ) -> OperationResult<()> {
         match self {
-            VectorIndexEnum::Plain(index) => index.build_index(permit, stopped),
-            VectorIndexEnum::HnswRam(index) => index.build_index(permit, stopped),
-            VectorIndexEnum::HnswMmap(index) => index.build_index(permit, stopped),
-            VectorIndexEnum::SparseRam(index) => index.build_index(permit, stopped),
-            VectorIndexEnum::SparseMmap(index) => index.build_index(permit, stopped),
+            VectorIndexEnum::Plain(index) => {
+                index.build_index_with_progress(permit, stopped, tick_progress)
+            }
+            VectorIndexEnum::HnswRam(index) => {
+                index.build_index_with_progress(permit, stopped, tick_progress)
+            }
+            VectorIndexEnum::HnswMmap(index) => {
+                index.build_index_with_progress(permit, stopped, tick_progress)
+            }
+            VectorIndexEnum::SparseRam(index) => {
+                index.build_index_with_progress(permit, stopped, tick_progress)
+            }
+            VectorIndexEnum::SparseMmap(index) => {
+                index.build_index_with_progress(permit, stopped, tick_progress)
+            }
         }
     }
 
diff --git a/lib/sparse/Cargo.toml b/lib/sparse/Cargo.toml
index 92799b8890..cc21e4c453 100644
--- a/lib/sparse/Cargo.toml
+++ b/lib/sparse/Cargo.toml
@@ -8,6 +8,9 @@ authors = [
 license = "Apache-2.0"
 edition = "2021"
 
+[features]
+testing = []
+
 [dependencies]
 common = { path = "../common/common" }
 io = { path = "../common/io" }
@@ -15,9 +18,23 @@ memory = { path = "../common/memory" }
 memmap2 = "0.9.4"
 schemars = { workspace = true }
 serde = { workspace = true }
+serde_json = { workspace = true }
 tempfile = "3.10.1"
 ordered-float = "4.2"
 rand = "0.8.5"
 validator = { workspace = true }
 itertools = "0.12.1"
 parking_lot = "0.12.2"
+
+[dev-dependencies]
+criterion = "0.5"
+dataset = { path = "../common/dataset" }
+indicatif = { workspace = true }
+sparse = { path = ".", features = ["testing"] }
+
+[target.'cfg(not(target_os = "windows"))'.dev-dependencies]
+pprof = { workspace = true }
+
+[[bench]]
+name = "search"
+harness = false
diff --git a/lib/sparse/benches/prof.rs b/lib/sparse/benches/prof.rs
new file mode 100644
index 0000000000..eeb8ab6065
--- /dev/null
+++ b/lib/sparse/benches/prof.rs
@@ -0,0 +1,89 @@
+use std::fs::File;
+use std::io::Write;
+use std::os::raw::c_int;
+use std::path::Path;
+
+use criterion::profiler::Profiler;
+use pprof::flamegraph::TextTruncateDirection;
+use pprof::protos::Message;
+use pprof::ProfilerGuard;
+
+/// Small custom profiler that can be used with Criterion to create a flamegraph for benchmarks.
+/// Also see [the Criterion documentation on this][custom-profiler].
+///
+/// ## Example on how to enable the custom profiler:
+///
+/// ```
+/// mod perf;
+/// use perf::FlamegraphProfiler;
+///
+/// fn fibonacci_profiled(criterion: &mut Criterion) {
+///     // Use the criterion struct as normal here.
+/// }
+///
+/// fn custom() -> Criterion {
+///     Criterion::default().with_profiler(FlamegraphProfiler::new())
+/// }
+///
+/// criterion_group! {
+///     name = benches;
+///     config = custom();
+///     targets = fibonacci_profiled
+/// }
+/// ```
+///
+/// The neat thing about this is that it will sample _only_ the benchmark, and not other stuff like
+/// the setup process.
+///
+/// Further, it will only kick in if `--profile-time <time>` is passed to the benchmark binary.
+/// A flamegraph will be created for each individual benchmark in its report directory under
+/// `profile/flamegraph.svg`.
+///
+/// [custom-profiler]: https://bheisler.github.io/criterion.rs/book/user_guide/profiling.html#implementing-in-process-profiling-hooks
+pub struct FlamegraphProfiler<'a> {
+    frequency: c_int,
+    active_profiler: Option<ProfilerGuard<'a>>,
+}
+
+impl<'a> FlamegraphProfiler<'a> {
+    #[allow(dead_code)]
+    pub fn new(frequency: c_int) -> Self {
+        FlamegraphProfiler {
+            frequency,
+            active_profiler: None,
+        }
+    }
+}
+
+impl<'a> Profiler for FlamegraphProfiler<'a> {
+    fn start_profiling(&mut self, _benchmark_id: &str, _benchmark_dir: &Path) {
+        self.active_profiler = Some(ProfilerGuard::new(self.frequency).unwrap());
+    }
+
+    fn stop_profiling(&mut self, _benchmark_id: &str, benchmark_dir: &Path) {
+        std::fs::create_dir_all(benchmark_dir).unwrap();
+        let pprof_path = benchmark_dir.join("profile.pb");
+        let flamegraph_path = benchmark_dir.join("flamegraph.svg");
+        eprintln!("\nflamegraph_path = {flamegraph_path:#?}");
+        let flamegraph_file = File::create(&flamegraph_path)
+            .expect("File system error while creating flamegraph.svg");
+        let mut options = pprof::flamegraph::Options::default();
+        options.hash = true;
+        options.image_width = Some(2500);
+        options.text_truncate_direction = TextTruncateDirection::Left;
+        options.font_size /= 3;
+        if let Some(profiler) = self.active_profiler.take() {
+            let report = profiler.report().build().unwrap();
+
+            let mut file = File::create(pprof_path).unwrap();
+            let profile = report.pprof().unwrap();
+            let mut content = Vec::new();
+            profile.encode(&mut content).unwrap();
+            file.write_all(&content).unwrap();
+
+            report
+                .flamegraph_with_options(flamegraph_file, &mut options)
+                .expect("Error writing flamegraph");
+        }
+    }
+}
diff --git a/lib/sparse/benches/search.rs b/lib/sparse/benches/search.rs
new file mode 100644
index 0000000000..05b7cb8444
--- /dev/null
+++ b/lib/sparse/benches/search.rs
@@ -0,0 +1,178 @@
+use std::io;
+use std::path::Path;
+use std::sync::atomic::AtomicBool;
+
+use common::types::PointOffsetType;
+use criterion::{criterion_group, criterion_main, Criterion};
+use dataset::Dataset;
+use indicatif::{ProgressBar, ProgressDrawTarget};
+use itertools::Itertools;
+use rand::rngs::StdRng;
+use rand::SeedableRng as _;
+use sparse::common::scores_memory_pool::ScoresMemoryPool;
+use sparse::common::sparse_vector::SparseVector;
+use sparse::common::sparse_vector_fixture::{random_positive_sparse_vector, random_sparse_vector};
+use sparse::index::inverted_index::inverted_index_ram::InvertedIndexRam;
+use sparse::index::inverted_index::inverted_index_ram_builder::InvertedIndexBuilder;
+use sparse::index::loaders::{self, Csr};
+use sparse::index::search_context::SearchContext;
+mod prof;
+
+const NUM_QUERIES: usize = 2048;
+const MAX_SPARSE_DIM: usize = 30_000;
+const TOP: usize = 10;
+
+pub fn bench_search(c: &mut Criterion) {
+    bench_uniform_random(c, "random-50k", 50_000);
+    bench_uniform_random(c, "random-500k", 500_000);
+
+    let query_vectors =
+        loaders::load_csr_vecs(Dataset::NeurIps2023Queries.download().unwrap()).unwrap();
+
+    let index_1m = load_csr_index(Dataset::NeurIps2023_1M.download().unwrap(), 1.0).unwrap();
+    run_bench(c, "neurips2023-1M", index_1m, query_vectors.clone());
+
+    let index_full = load_csr_index(Dataset::NeurIps2023Full.download().unwrap(), 0.25).unwrap();
+    run_bench(c, "neurips2023-full-25pct", index_full, query_vectors);
+
+    bench_movies(c);
+}
+
+fn bench_uniform_random(c: &mut Criterion, name: &str, num_vectors: usize) {
+    let mut rnd = StdRng::seed_from_u64(42);
+
+    let index = InvertedIndexBuilder::build_from_iterator((0..num_vectors).map(|idx| {
+        (
+            idx as PointOffsetType,
+            random_sparse_vector(&mut rnd, MAX_SPARSE_DIM).into_remapped(),
+        )
+    }));
+
+    let query_vectors = (0..NUM_QUERIES)
+        .map(|_| random_positive_sparse_vector(&mut rnd, MAX_SPARSE_DIM))
+        .collect::<Vec<_>>();
+
+    run_bench(c, name, index, query_vectors);
+}
+
+pub fn bench_movies(c: &mut Criterion) {
+    let mut iter =
+        loaders::JsonReader::open(Dataset::SpladeWikiMovies.download().unwrap()).unwrap();
+
+    // Use the first NUM_QUERIES vectors as queries, and the rest as index.
+
+    let query_vectors = (0..NUM_QUERIES)
+        .map(|_| iter.next().unwrap().unwrap())
+        .collect_vec();
+
+    let index = InvertedIndexBuilder::build_from_iterator(
+        iter.enumerate()
+            .map(|(idx, vec)| (idx as PointOffsetType, vec.unwrap().into_remapped())),
+    );
+
+    run_bench(c, "movies", index, query_vectors);
+}
+
+pub fn run_bench(
+    c: &mut Criterion,
+    name: &str,
+    index: InvertedIndexRam,
+    mut query_vectors: Vec<SparseVector>,
+) {
+    let pool = ScoresMemoryPool::new();
+    let stopped = AtomicBool::new(false);
+
+    let mut group = c.benchmark_group(format!("search/{}", name));
+
+    let mut it = query_vectors.iter().cycle();
+    group.bench_function("basic", |b| {
+        b.iter_batched(
+            || it.next().unwrap().clone().into_remapped(),
+            |vec| SearchContext::new(vec, TOP, &index, pool.get(), &stopped).search(&|_| true),
+            criterion::BatchSize::SmallInput,
+        )
+    });
+
+    let hottest_id = index
+        .postings
+        .iter()
+        .enumerate()
+        .map(|(i, p)| (i, p.elements.len()))
+        .max_by_key(|(_, len)| *len)
+        .unwrap()
+        .0 as u32;
+
+    let average_elements = index
+        .postings
+        .iter()
+        .map(|p| p.elements.len())
+        .sum::<usize>() as f64
+        / index.postings.len() as f64;
+
+    eprintln!(
+        "Hottest id: {hottest_id} (elements: {}), average elements: {average_elements}",
+        index.postings[hottest_id as usize].elements.len(),
+    );
+
+    for vec in &mut query_vectors {
+        vec.indices.truncate(4);
+        vec.values.truncate(4);
+        if let Err(idx) = vec.indices.binary_search(&hottest_id) {
+            if idx < vec.indices.len() {
+                vec.indices[idx] = hottest_id;
+                vec.values[idx] = 1.0;
+            } else {
+                vec.indices.push(hottest_id);
+                vec.values.push(1.0);
+            }
+        }
+    }
+
+    let mut it = query_vectors.iter().cycle();
+    group.bench_function("hottest", |b| {
+        b.iter(|| {
+            SearchContext::new(
+                it.next().unwrap().clone().into_remapped(),
+                TOP,
+                &index,
+                pool.get(),
+                &stopped,
+            )
+            .search(&|_| true)
+        })
+    });
+}
+
+fn load_csr_index(path: impl AsRef<Path>, ratio: f32) -> io::Result<InvertedIndexRam> {
+    let csr = Csr::open(path.as_ref())?;
+    let mut builder = InvertedIndexBuilder::new();
+    assert!(ratio > 0.0 && ratio <= 1.0);
+    let count = (csr.len() as f32 * ratio) as usize;
+    let bar =
+        ProgressBar::with_draw_target(Some(count as u64), ProgressDrawTarget::stderr_with_hz(12));
+    for (row, vec) in bar.wrap_iter(csr.iter().take(count).enumerate()) {
+        builder.add(
+            row as u32,
+            vec.map(|v| v.into_remapped())
+                .map_err(|e| io::Error::new(io::ErrorKind::InvalidData, e))?,
+        );
+    }
+    bar.finish_and_clear();
+    Ok(builder.build())
+}
+
+#[cfg(not(target_os = "windows"))]
+criterion_group! {
+    name = benches;
+    config = Criterion::default().with_profiler(prof::FlamegraphProfiler::new(100));
+    targets = bench_search,
+}
+
+#[cfg(target_os = "windows")]
+criterion_group! {
+    name = benches;
+    config = Criterion::default();
+    targets = bench_search,
+}
+
+criterion_main!(benches);
diff --git a/lib/sparse/src/common/sparse_vector.rs b/lib/sparse/src/common/sparse_vector.rs
index 22af4c8794..b0267f7233 100644
--- a/lib/sparse/src/common/sparse_vector.rs
+++ b/lib/sparse/src/common/sparse_vector.rs
@@ -207,6 +207,15 @@ impl SparseVector {
         debug_assert!(result.validate().is_ok());
         result
     }
+
+    /// Create [RemappedSparseVector] from this vector in a naive way. Only suitable for testing.
+    #[cfg(feature = "testing")]
+    pub fn into_remapped(self) -> RemappedSparseVector {
+        RemappedSparseVector {
+            indices: self.indices,
+            values: self.values,
+        }
+    }
 }
 
 impl TryFrom<Vec<(u32, f32)>> for RemappedSparseVector {
diff --git a/lib/sparse/src/index/inverted_index/inverted_index_ram_builder.rs b/lib/sparse/src/index/inverted_index/inverted_index_ram_builder.rs
index e2cde2b719..42c424bc6b 100644
--- a/lib/sparse/src/index/inverted_index/inverted_index_ram_builder.rs
+++ b/lib/sparse/src/index/inverted_index/inverted_index_ram_builder.rs
@@ -52,4 +52,15 @@ impl InvertedIndexBuilder {
             vector_count,
         }
     }
+
+    /// Creates an [InvertedIndexRam] from an iterator of (id, vector) pairs.
+    pub fn build_from_iterator(
+        iter: impl Iterator<Item = (PointOffsetType, RemappedSparseVector)>,
+    ) -> InvertedIndexRam {
+        let mut builder = InvertedIndexBuilder::new();
+        for (id, vector) in iter {
+            builder.add(id, vector);
+        }
+        builder.build()
+    }
 }
diff --git a/lib/sparse/src/index/loaders.rs b/lib/sparse/src/index/loaders.rs
new file mode 100644
index 0000000000..49f05c8a75
--- /dev/null
+++ b/lib/sparse/src/index/loaders.rs
@@ -0,0 +1,157 @@
+use std::collections::HashMap;
+use std::fs::File;
+use std::io::{self, BufRead as _, BufReader, Lines};
+use std::mem::size_of;
+use std::path::Path;
+
+use memmap2::Mmap;
+use memory::mmap_ops::{open_read_mmap, transmute_from_u8, transmute_from_u8_to_slice};
+use validator::ValidationErrors;
+
+use crate::common::sparse_vector::SparseVector;
+
+/// Compressed Sparse Row matrix, baked by memory-mapped file.
+///
+/// The layout of the memory-mapped file is as follows:
+///
+/// =======  ===========  ==========  ===================
+/// name     type         size        start
+/// =======  ===========  ==========  ===================
+/// nrow     u64          8           0
+/// ncol     u64          8           8
+/// nnz      u64          8           16
+/// indptr   u64[nrow+1]  8*(nrow+1)  24
+/// indices  u32[nnz]     4*nnz       24+8*(nrow+1)
+/// data     u32[nnz]     4*nnz       24+8*(nrow+1)+4*nnz
+pub struct Csr {
+    mmap: Mmap,
+
+    nrow: usize,
+    nnz: usize,
+    intptr: Vec<u64>,
+}
+
+const CSR_HEADER_SIZE: usize = size_of::<u64>() * 3;
+
+impl Csr {
+    pub fn open(path: impl AsRef<Path>) -> io::Result<Self> {
+        Self::from_mmap(open_read_mmap(path.as_ref())?)
+    }
+
+    #[inline]
+    #[allow(clippy::len_without_is_empty)]
+    pub fn len(&self) -> usize {
+        self.nrow
+    }
+
+    pub fn iter(&self) -> CsrIter<'_> {
+        CsrIter { csr: self, row: 0 }
+    }
+
+    fn from_mmap(mmap: Mmap) -> io::Result<Self> {
+        let (nrow, ncol, nnz) =
+            transmute_from_u8::<(u64, u64, u64)>(&mmap.as_ref()[..CSR_HEADER_SIZE]);
+        let (nrow, _ncol, nnz) = (*nrow as usize, *ncol as usize, *nnz as usize);
+
+        let indptr = Vec::from(transmute_from_u8_to_slice::<u64>(
+            &mmap.as_ref()[CSR_HEADER_SIZE..CSR_HEADER_SIZE + size_of::<u64>() * (nrow + 1)],
+        ));
+        if !indptr.windows(2).all(|w| w[0] <= w[1]) || indptr.last() != Some(&(nnz as u64)) {
+            return Err(io::Error::new(
+                io::ErrorKind::InvalidData,
+                "Invalid indptr array",
+            ));
+        }
+
+        Ok(Self {
+            mmap,
+            nrow,
+            nnz,
+            intptr: indptr,
+        })
+    }
+
+    #[inline]
+    unsafe fn vec(&self, row: usize) -> Result<SparseVector, ValidationErrors> {
+        let start = *self.intptr.get_unchecked(row) as usize;
+        let end = *self.intptr.get_unchecked(row + 1) as usize;
+
+        let mut pos = CSR_HEADER_SIZE + size_of::<u64>() * (self.nrow + 1);
+
+        let indices = transmute_from_u8_to_slice::<u32>(
+            self.mmap
+                .as_ref()
+                .get_unchecked(pos + size_of::<u32>() * start..pos + size_of::<u32>() * end),
+        );
+        pos += size_of::<u32>() * self.nnz;
+
+        let data = transmute_from_u8_to_slice::<f32>(
+            self.mmap
+                .as_ref()
+                .get_unchecked(pos + size_of::<f32>() * start..pos + size_of::<f32>() * end),
+        );
+
+        SparseVector::new(indices.to_vec(), data.to_vec())
+    }
+}
+
+/// Iterator over the rows of a CSR matrix.
+pub struct CsrIter<'a> {
+    csr: &'a Csr,
+    row: usize,
+}
+
+impl<'a> Iterator for CsrIter<'a> {
+    type Item = Result<SparseVector, ValidationErrors>;
+
+    fn next(&mut self) -> Option<Self::Item> {
+        (self.row < self.csr.nrow).then(|| {
+            let vec = unsafe { self.csr.vec(self.row) };
+            self.row += 1;
+            vec
+        })
+    }
+}
+
+impl<'a> ExactSizeIterator for CsrIter<'a> {
+    fn len(&self) -> usize {
+        self.csr.nrow - self.row
+    }
+}
+
+pub fn load_csr_vecs(path: impl AsRef<Path>) -> io::Result<Vec<SparseVector>> {
+    Csr::open(path)?
+        .iter()
+        .collect::<Result<Vec<_>, _>>()
+        .map_err(|e| io::Error::new(io::ErrorKind::InvalidData, e))
+}
+
+/// Stream of sparse vectors in JSON format.
+pub struct JsonReader(Lines<BufReader<File>>);
+
+impl JsonReader {
+    pub fn open(path: impl AsRef<Path>) -> io::Result<Self> {
+        Ok(JsonReader(BufReader::new(File::open(path)?).lines()))
+    }
+}
+
+impl Iterator for JsonReader {
+    type Item = Result<SparseVector, io::Error>;
+
+    fn next(&mut self) -> Option<Self::Item> {
+        self.0.next().map(|line| {
+            line.map_err(|e| io::Error::new(io::ErrorKind::InvalidData, e))
+                .and_then(|line| {
+                    let data: HashMap<String, f32> = serde_json::from_str(&line)?;
+                    SparseVector::new(
+                        data.keys()
+                            .map(|k| k.parse())
+                            .collect::<Result<Vec<_>, _>>()
+                            .map_err(|e| io::Error::new(io::ErrorKind::InvalidData, e))?,
+                        data.values().copied().collect(),
+                    )
+                    .map_err(|e| io::Error::new(io::ErrorKind::InvalidData, e))
+                })
+        })
+    }
+}
diff --git a/lib/sparse/src/index/mod.rs b/lib/sparse/src/index/mod.rs
index fbb7b48417..3b6b2110bc 100644
--- a/lib/sparse/src/index/mod.rs
+++ b/lib/sparse/src/index/mod.rs
@@ -1,5 +1,6 @@
 #![allow(dead_code)]
 
 pub mod inverted_index;
+pub mod loaders;
 pub mod posting_list;
 pub mod search_context;