enhance: support set lindera dict build dir and download url in yaml (#43541)

relate: https://github.com/milvus-io/milvus/issues/43120

---------

Signed-off-by: aoiasd <zhicheng.yue@zilliz.com>

configs/milvus.yaml

@@ -1409,3 +1409,5 @@ function:
           enable: true # Whether to enable TEI rerank service
         vllm:
           enable: true # Whether to enable vllm rerank service
+  analyzer:
+    local_resource_path: /var/lib/milvus/analyzer

internal/core/thirdparty/tantivy/tantivy-binding/Cargo.lock

@@ -932,6 +932,16 @@ dependencies = [
  "num",
 ]
 
+[[package]]
+name = "fs2"
+version = "0.4.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9564fc758e15025b46aa6643b1b77d047d1a56a1aea6e01002ac0c7026876213"
+dependencies = [
+ "libc",
+ "winapi",
+]
+
 [[package]]
 name = "fs4"
 version = "0.6.6"
@@ -3853,6 +3863,7 @@ dependencies = [
  "env_logger",
  "fancy-regex",
  "flate2",
+ "fs2",
  "futures",
  "icu_segmenter",
  "jieba-rs",

internal/core/thirdparty/tantivy/tantivy-binding/Cargo.toml

@@ -26,6 +26,7 @@ overflow-checks = false
 tantivy = { git = "https://github.com/zilliztech/tantivy.git" }
 tantivy-5 = { package = "tantivy", git = "https://github.com/zilliztech/tantivy", tag = "0.21.1-fix4" }
 lindera = "0.42.4"
+fs2 = "0.4"
 lindera-dictionary = "0.42.4"
 futures = "0.3.21"
 libc = "0.2"

internal/core/thirdparty/tantivy/tantivy-binding/src/analyzer/analyzer.rs

@@ -78,9 +78,9 @@ impl AnalyzerBuilder<'_> {
         for filter in filters {
             if filter.is_string() {
                 let filter_name = filter.as_str().unwrap();
-                let costum = self.filters.remove(filter_name);
+                let customize = self.filters.remove(filter_name);
-                if !costum.is_none() {
+                if !customize.is_none() {
-                    builder = costum.unwrap().transform(builder);
+                    builder = customize.unwrap().transform(builder);
                     continue;
                 }
 

internal/core/thirdparty/tantivy/tantivy-binding/src/analyzer/dict/lindera/cc_cedict.rs

@@ -1,16 +1,13 @@
 use super::fetch;
 use crate::error::{Result, TantivyBindingError};
 use lindera_dictionary::dictionary_builder::cc_cedict::CcCedictBuilder;
-use tokio::runtime::Runtime;
 
 #[cfg(feature = "lindera-cc-cedict")]
 use lindera::dictionary::{load_dictionary_from_kind, DictionaryKind};
 
 #[cfg(not(feature = "lindera-cc-cedict"))]
-async fn download(params: &fetch::FetchParams) -> Result<()> {
+fn download(params: &fetch::FetchParams) -> Result<()> {
-    fetch::fetch(params, CcCedictBuilder::new())
+    fetch::fetch(params, CcCedictBuilder::new()).map_err(|e| {
-        .await
-        .map_err(|e| {
         TantivyBindingError::InternalError(format!(
             "fetch cc_cedict failed with error: {}",
             e.to_string()
@@ -39,8 +36,7 @@ pub fn load_cc_cedict(
         params.download_urls = download_url
     }
 
-    let rt = Runtime::new().unwrap();
+    download(&params)?;
-    rt.block_on(download(&params))?;
     fetch::load(&params)
 }
 

internal/core/thirdparty/tantivy/tantivy-binding/src/analyzer/dict/lindera/fetch.rs

@@ -1,28 +1,59 @@
+use std::env;
 use std::error::Error;
-use std::path::Path;
+use std::fs;
+use std::io::{self, Cursor, Read, Write};
+use std::path::{Path, PathBuf};
+use std::time::Instant;
+
+use flate2::read::GzDecoder;
+use fs2::FileExt;
+use tar::Archive;
 
 use lindera::dictionary::Dictionary;
 use lindera_dictionary::dictionary::character_definition::CharacterDefinition;
 use lindera_dictionary::dictionary::connection_cost_matrix::ConnectionCostMatrix;
 use lindera_dictionary::dictionary::prefix_dictionary::PrefixDictionary;
 use lindera_dictionary::dictionary::unknown_dictionary::UnknownDictionary;
+use lindera_dictionary::dictionary_builder::DictionaryBuilder;
 use log::{error, info, warn};
 use md5::Context;
 use rand::{rngs::SmallRng, seq::SliceRandom, SeedableRng};
 use reqwest::Client;
+use tokio::runtime::Runtime;
 use tokio::time::sleep;
 use tokio::time::Duration;
 
-use serde_json as json;
-use std::fs;
-use std::path::PathBuf;
-
 use super::common;
 use crate::error::TantivyBindingError;
-use lindera_dictionary::dictionary_builder::DictionaryBuilder;
 
 const MAX_ROUND: usize = 3;
 
+pub struct FileMutexGuard {
+    file: Option<fs::File>,
+    path: PathBuf,
+}
+
+impl FileMutexGuard {
+    fn build(path: PathBuf) -> io::Result<FileMutexGuard> {
+        let flock = fs::File::create(&path)?;
+        flock.lock_exclusive()?;
+        Ok(FileMutexGuard {
+            file: Some(flock),
+            path: path,
+        })
+    }
+}
+
+impl Drop for FileMutexGuard {
+    fn drop(&mut self) {
+        if let Some(file) = self.file.take() {
+            let _ = file.unlock();
+            drop(file); // drop file before remove file
+            let _ = std::fs::remove_file(&self.path);
+        }
+    }
+}
+
 pub struct FetchParams {
     pub lindera_dir: String,
 
@@ -138,35 +169,20 @@ async fn download_with_retry(
     Err("Failed to download a valid file from all sources".into())
 }
 
-/// Fetch the necessary assets and then build the dictionary using `builder`
+pub fn build(
-pub async fn fetch(
     params: &FetchParams,
     builder: impl DictionaryBuilder,
+    build_dir: &PathBuf,
+    input_dir: &PathBuf,
+    output_dir: &PathBuf,
+    tmp_dir: &PathBuf,
 ) -> Result<(), Box<dyn Error>> {
-    use std::env;
-    use std::fs::{rename, File};
-    use std::io::{self, Cursor, Read, Write};
-    use std::path::{Path, PathBuf};
-    use std::time::Instant;
-
-    use flate2::read::GzDecoder;
-    use tar::Archive;
-
     let start = Instant::now();
     info!(
-        "start fetch lindera dictionary name: {}\n",
+        "start donwload and build lindera dictionary. name: {} to {:?}\n",
-        params.file_name.as_str()
+        params.file_name.as_str(),
+        output_dir,
     );
-    let build_dir = PathBuf::from(params.lindera_dir.as_str());
-    std::fs::create_dir_all(&build_dir)?;
-
-    let input_dir = build_dir.join(params.input_dir.as_str());
-    let output_dir = build_dir.join(params.output_dir.as_str());
-
-    // Fast path where the data is already in cache
-    if output_dir.is_dir() {
-        return Ok(());
-    }
 
     // Source file path for build package
     let source_path_for_build = &build_dir.join(params.file_name.as_str());
@@ -179,33 +195,32 @@ pub async fn fetch(
         .user_agent(format!("Lindera/{}", env!("CARGO_PKG_VERSION")))
         .build()?;
 
-    let mut dest = File::create(tmp_path.as_path())?;
+    let mut dest = fs::File::create(tmp_path.as_path())?;
-    let content = download_with_retry(
+
+    let rt = Runtime::new().unwrap();
+    let content = rt.block_on(download_with_retry(
         &client,
         params.download_urls.iter().map(|s| s.as_str()).collect(),
         MAX_ROUND,
         params.md5_hash.as_str(),
-    )
+    ))?;
-    .await?;
 
     io::copy(&mut Cursor::new(content.as_slice()), &mut dest)?;
     dest.flush()?;
 
-    rename(tmp_path.clone(), source_path_for_build)?;
+    fs::rename(tmp_path.clone(), source_path_for_build)?;
 
-    // Decompress a tar.gz file
+    let tmp_extracted_path = tmp_dir.join(params.input_dir.as_str());
-    let tmp_extract_path = Path::new(&build_dir).join(format!("tmp-archive-{}", params.input_dir));
+    let _ = std::fs::remove_dir_all(&tmp_dir);
-    let tmp_extracted_path = tmp_extract_path.join(params.input_dir.as_str());
+    std::fs::create_dir_all(&tmp_dir)?;
-    let _ = std::fs::remove_dir_all(&tmp_extract_path);
-    std::fs::create_dir_all(&tmp_extract_path)?;
 
-    let mut tar_gz = File::open(source_path_for_build)?;
+    let mut tar_gz = fs::File::open(source_path_for_build)?;
     let mut buffer = Vec::new();
     tar_gz.read_to_end(&mut buffer)?;
     let cursor = Cursor::new(buffer);
     let decoder = GzDecoder::new(cursor);
     let mut archive = Archive::new(decoder);
-    archive.unpack(&tmp_extract_path)?;
+    archive.unpack(&tmp_dir)?;
 
     #[cfg(target_os = "windows")]
     {
@@ -229,10 +244,9 @@ pub async fn fetch(
     {
         // Empty the input directory first to avoid conflicts when renaming the directory later on Linux and macOS systems (which do not support overwriting directories).
         empty_directory(&input_dir)?;
-        rename(tmp_extracted_path, &input_dir)?;
+        fs::rename(tmp_extracted_path, &input_dir)?;
     }
 
-    let _ = std::fs::remove_dir_all(&tmp_extract_path);
     drop(dest);
     let _ = std::fs::remove_file(source_path_for_build);
 
@@ -265,19 +279,54 @@ pub async fn fetch(
         empty_directory(&output_dir)?;
 
         // Rename tmp_path to output_dir
-        rename(tmp_path, &output_dir)?;
+        fs::rename(tmp_path, &output_dir)?;
     }
-
-    let _ = std::fs::remove_dir_all(&input_dir);
-
     info!(
-        "finish fetch lindera dictionary name: {} duration: {} ms\n",
+        "finish donwload and build lindera dictionary. name: {} duration: {} ms\n",
         params.file_name.as_str(),
         start.elapsed().as_millis()
     );
     Ok(())
 }
 
+/// Fetch the necessary assets and then build the dictionary using `builder`
+pub fn fetch(params: &FetchParams, builder: impl DictionaryBuilder) -> Result<(), Box<dyn Error>> {
+    let build_dir = PathBuf::from(params.lindera_dir.as_str());
+    std::fs::create_dir_all(&build_dir)?;
+
+    let input_dir = build_dir.join(params.input_dir.as_str());
+    let output_dir = build_dir.join(params.output_dir.as_str());
+    let lock_path = build_dir.join(format!("lindera-{}.lock", params.file_name.as_str()));
+
+    // Skip create fs lock if already in cache
+    if output_dir.is_dir() {
+        return Ok(());
+    }
+
+    let _flock_guard = FileMutexGuard::build(lock_path)?;
+
+    // Fast path where the data is already in cache
+    if output_dir.is_dir() {
+        return Ok(());
+    }
+
+    // Decompress a tar.gz file
+    let tmp_dir = Path::new(&build_dir).join(format!("tmp-archive-{}", params.input_dir));
+
+    let build_result = build(
+        params,
+        builder,
+        &build_dir,
+        &input_dir,
+        &output_dir,
+        &tmp_dir,
+    );
+    let _ = std::fs::remove_dir_all(&tmp_dir);
+    let _ = std::fs::remove_dir_all(&input_dir);
+
+    build_result
+}
+
 pub fn load(params: &FetchParams) -> Result<lindera::dictionary::Dictionary, TantivyBindingError> {
     let dict_dir = PathBuf::from(params.lindera_dir.clone()).join(params.output_dir.clone());
     let da_data = fs::read(dict_dir.join(common::DA_DATA))?;
@@ -286,7 +335,7 @@ pub fn load(params: &FetchParams) -> Result<lindera::dictionary::Dictionary, Tan
     let words_data = fs::read(dict_dir.join(common::WORDS_DATA))?;
     let connection_data = fs::read(dict_dir.join(common::CONNECTION_DATA))?;
     let char_definition_data = fs::read(dict_dir.join(common::CHAR_DEFINITION_DATA))?;
-    let unkonwn_data = fs::read(dict_dir.join(common::UNKNOWN_DATA))?;
+    let unknown_data = fs::read(dict_dir.join(common::UNKNOWN_DATA))?;
 
     let dict = Dictionary {
         prefix_dictionary: PrefixDictionary::load(
@@ -305,7 +354,7 @@ pub fn load(params: &FetchParams) -> Result<lindera::dictionary::Dictionary, Tan
                 ))
             },
         )?,
-        unknown_dictionary: UnknownDictionary::load(unkonwn_data.as_slice()).map_err(|e| {
+        unknown_dictionary: UnknownDictionary::load(unknown_data.as_slice()).map_err(|e| {
             TantivyBindingError::InternalError(format!(
                 "lindera load unknown dictionary failed, err:{}",
                 e

internal/core/thirdparty/tantivy/tantivy-binding/src/analyzer/dict/lindera/ipadic.rs

@@ -1,16 +1,13 @@
 use super::fetch;
 use crate::error::{Result, TantivyBindingError};
 use lindera_dictionary::dictionary_builder::ipadic::IpadicBuilder;
-use tokio::runtime::Runtime;
 
 #[cfg(feature = "lindera-ipadic")]
 use lindera::dictionary::{load_dictionary_from_kind, DictionaryKind};
 
 #[cfg(not(feature = "lindera-ipadic"))]
-async fn download(params: &fetch::FetchParams) -> Result<()> {
+fn download(params: &fetch::FetchParams) -> Result<()> {
-    fetch::fetch(params, IpadicBuilder::new())
+    fetch::fetch(params, IpadicBuilder::new()).map_err(|e| {
-        .await
-        .map_err(|e| {
         TantivyBindingError::InternalError(format!(
             "fetch ipadic failed with error: {}",
             e.to_string()
@@ -40,8 +37,7 @@ pub fn load_ipadic(
         params.download_urls = download_url
     }
 
-    let rt = Runtime::new().unwrap();
+    download(&params)?;
-    rt.block_on(download(&params))?;
     fetch::load(&params)
 }
 

internal/core/thirdparty/tantivy/tantivy-binding/src/analyzer/dict/lindera/ipadic_neologd.rs

@@ -1,16 +1,12 @@
 use super::fetch;
 use crate::error::{Result, TantivyBindingError};
-use lindera_dictionary::dictionary_builder::ipadic_neologd::IpadicNeologdBuilder;
-use tokio::runtime::Runtime;
-
 #[cfg(feature = "lindera-ipadic-neologd")]
 use lindera::dictionary::{load_dictionary_from_kind, DictionaryKind};
+use lindera_dictionary::dictionary_builder::ipadic_neologd::IpadicNeologdBuilder;
 
 #[cfg(not(feature = "lindera-ipadic-neologd"))]
-async fn download(params: &fetch::FetchParams) -> Result<()> {
+fn download(params: &fetch::FetchParams) -> Result<()> {
-    fetch::fetch(params, IpadicNeologdBuilder::new())
+    fetch::fetch(params, IpadicNeologdBuilder::new()).map_err(|e| {
-        .await
-        .map_err(|e| {
         TantivyBindingError::InternalError(format!(
             "fetch ipadic-neologd failed with error: {}",
             e.to_string()
@@ -39,8 +35,7 @@ pub fn load_ipadic_neologd(
         params.download_urls = download_url
     }
 
-    let rt = Runtime::new().unwrap();
+    download(&params)?;
-    rt.block_on(download(&params))?;
     fetch::load(&params)
 }
 

internal/core/thirdparty/tantivy/tantivy-binding/src/analyzer/dict/lindera/ko_dic.rs

@@ -1,16 +1,13 @@
 use super::fetch;
 use crate::error::{Result, TantivyBindingError};
 use lindera_dictionary::dictionary_builder::ko_dic::KoDicBuilder;
-use tokio::runtime::Runtime;
 
 #[cfg(feature = "lindera-ko-dic")]
 use lindera::dictionary::{load_dictionary_from_kind, DictionaryKind};
 
 #[cfg(not(feature = "lindera-ko-dic"))]
-async fn download(params: &fetch::FetchParams) -> Result<()> {
+fn download(params: &fetch::FetchParams) -> Result<()> {
-    fetch::fetch(params, KoDicBuilder::new())
+    fetch::fetch(params, KoDicBuilder::new()).map_err(|e| {
-        .await
-        .map_err(|e| {
         TantivyBindingError::InternalError(format!(
             "fetch ko-dic failed with error: {}",
             e.to_string()
@@ -40,8 +37,7 @@ pub fn load_ko_dic(
         params.download_urls = download_url
     }
 
-    let rt = Runtime::new().unwrap();
+    download(&params)?;
-    rt.block_on(download(&params))?;
     fetch::load(&params)
 }
 

internal/core/thirdparty/tantivy/tantivy-binding/src/analyzer/dict/lindera/unidic.rs

@@ -1,16 +1,13 @@
 use super::fetch;
 use crate::error::{Result, TantivyBindingError};
 use lindera_dictionary::dictionary_builder::unidic::UnidicBuilder;
-use tokio::runtime::Runtime;
 
 #[cfg(feature = "lindera-unidic")]
 use lindera::dictionary::{load_dictionary_from_kind, DictionaryKind};
 
 #[cfg(not(feature = "lindera-unidic"))]
-async fn download(params: &fetch::FetchParams) -> Result<()> {
+fn download(params: &fetch::FetchParams) -> Result<()> {
-    fetch::fetch(params, UnidicBuilder::new())
+    fetch::fetch(params, UnidicBuilder::new()).map_err(|e| {
-        .await
-        .map_err(|e| {
         TantivyBindingError::InternalError(format!(
             "fetch unidic failed with error: {}",
             e.to_string()
@@ -39,8 +36,7 @@ pub fn load_unidic(
         params.download_urls = download_url
     }
 
-    let rt = Runtime::new().unwrap();
+    download(&params)?;
-    rt.block_on(download(&params))?;
     fetch::load(&params)
 }
 

internal/core/thirdparty/tantivy/tantivy-binding/src/analyzer/filter/util.rs

@@ -3,7 +3,7 @@ use serde_json as json;
 use super::stop_words;
 use crate::error::{Result, TantivyBindingError};
 
-pub(crate) fn get_string_list(value: &json::Value, label: &str) -> Result<Vec<String>> {
+pub fn get_string_list(value: &json::Value, label: &str) -> Result<Vec<String>> {
     if !value.is_array() {
         return Err(TantivyBindingError::InternalError(
             format!("{} should be array", label).to_string(),

internal/core/thirdparty/tantivy/tantivy-binding/src/analyzer/tokenizers/lindera_tokenizer.rs

@@ -16,6 +16,7 @@ use lindera::token_filter::korean_stop_tags::KoreanStopTagsTokenFilter;
 use lindera::token_filter::BoxTokenFilter as LTokenFilter;
 
 use crate::analyzer::dict::lindera::load_dictionary_from_kind;
+use crate::analyzer::filter::get_string_list;
 use crate::error::{Result, TantivyBindingError};
 use serde_json as json;
 
@@ -25,6 +26,8 @@ pub struct LinderaTokenStream<'a> {
 }
 
 const DICTKINDKEY: &str = "dict_kind";
+const DICTBUILDDIRKEY: &str = "dict_build_dir";
+const DICTDOWNLOADURLKEY: &str = "download_urls";
 const FILTERKEY: &str = "filter";
 
 impl<'a> TokenStream for LinderaTokenStream<'a> {
@@ -62,8 +65,12 @@ impl LinderaTokenizer {
     /// This function will create a new `LinderaTokenizer` with json parameters.
     pub fn from_json(params: &json::Map<String, json::Value>) -> Result<LinderaTokenizer> {
         let kind: DictionaryKind = fetch_lindera_kind(params)?;
-        let dictionary =
+
-            load_dictionary_from_kind(&kind, "/var/lib/milvus/dict/lindera".to_string(), vec![])?;
+        // for download dict online
+        let build_dir = fetch_dict_build_dir(params)?;
+        let download_urls = fetch_dict_download_urls(params)?;
+
+        let dictionary = load_dictionary_from_kind(&kind, build_dir, download_urls)?;
 
         let segmenter = Segmenter::new(Mode::Normal, dictionary, None);
         let mut tokenizer = LinderaTokenizer::from_segmenter(segmenter);
@@ -126,18 +133,34 @@ impl DictionaryKindParser for &str {
 fn fetch_lindera_kind(params: &json::Map<String, json::Value>) -> Result<DictionaryKind> {
     params
         .get(DICTKINDKEY)
-        .ok_or_else(|| {
+        .ok_or(TantivyBindingError::InvalidArgument(format!(
-            TantivyBindingError::InvalidArgument(format!("lindera tokenizer dict_kind must be set"))
+            "lindera tokenizer dict_kind must be set"
-        })?
+        )))?
         .as_str()
-        .ok_or_else(|| {
+        .ok_or(TantivyBindingError::InvalidArgument(format!(
-            TantivyBindingError::InvalidArgument(format!(
             "lindera tokenizer dict kind should be string"
-            ))
+        )))?
-        })?
         .into_dict_kind()
 }
 
+fn fetch_dict_build_dir(params: &json::Map<String, json::Value>) -> Result<String> {
+    params
+        .get(DICTBUILDDIRKEY)
+        .map_or(Ok("/var/lib/milvus/dict/lindera".to_string()), |v| {
+            v.as_str()
+                .ok_or(TantivyBindingError::InvalidArgument(format!(
+                    "dict build dir must be string"
+                )))
+                .map(|s| s.to_string())
+        })
+}
+
+fn fetch_dict_download_urls(params: &json::Map<String, json::Value>) -> Result<Vec<String>> {
+    params.get(DICTDOWNLOADURLKEY).map_or(Ok(vec![]), |v| {
+        get_string_list(v, "lindera dict download urls")
+    })
+}
+
 fn fetch_lindera_tags_from_params(
     params: &json::Map<String, json::Value>,
 ) -> Result<HashSet<String>> {
@@ -330,7 +353,6 @@ mod tests {
     use tantivy::tokenizer::Tokenizer;
 
     #[test]
-    #[cfg(feature = "lindera-ipadic")]
     fn test_lindera_tokenizer() {
         let params = r#"{
             "type": "lindera",
@@ -358,7 +380,6 @@ mod tests {
     }
 
     #[test]
-    #[cfg(feature = "lindera-cc-cedict")]
     fn test_lindera_tokenizer_cc() {
         let params = r#"{
             "type": "lindera",

internal/core/thirdparty/tantivy/tantivy-binding/src/analyzer/tokenizers/tokenizer.rs

@@ -89,7 +89,7 @@ pub fn get_builder_with_tokenizer(
             }
             _ => {
                 return Err(TantivyBindingError::InvalidArgument(format!(
-                    "costum tokenizer must set type"
+                    "customized tokenizer must set type"
                 )))
             }
         }

internal/core/thirdparty/tantivy/tantivy-binding/src/index_writer_v5/analyzer/analyzer.rs

@@ -82,9 +82,9 @@ impl AnalyzerBuilder<'_> {
         for filter in filters {
             if filter.is_string() {
                 let filter_name = filter.as_str().unwrap();
-                let costum = self.filters.remove(filter_name);
+                let customize = self.filters.remove(filter_name);
-                if !costum.is_none() {
+                if !customize.is_none() {
-                    builder = costum.unwrap().transform(builder);
+                    builder = customize.unwrap().transform(builder);
                     continue;
                 }
 

internal/core/thirdparty/tantivy/tantivy-binding/src/index_writer_v5/analyzer/tokenizers/tokenizer.rs

@@ -51,7 +51,7 @@ pub fn get_builder_with_tokenizer(params: &json::Value) -> Result<TextAnalyzerBu
             }
             _ => {
                 return Err(TantivyBindingError::InvalidArgument(
-                    "costum tokenizer must set type".to_string(),
+                    "customized tokenizer must set type".to_string(),
                 ))
             }
         }

internal/util/ctokenizer/c_tokenizer_factory.go

@@ -9,12 +9,22 @@ package ctokenizer
 import "C"
 
 import (
+	"encoding/json"
+	"fmt"
+	"path"
 	"unsafe"
 
 	"github.com/milvus-io/milvus/internal/util/tokenizerapi"
+	"github.com/milvus-io/milvus/pkg/v2/util/merr"
+	"github.com/milvus-io/milvus/pkg/v2/util/paramtable"
 )
 
 func NewTokenizer(param string) (tokenizerapi.Tokenizer, error) {
+	param, err := CheckAndFillParams(param)
+	if err != nil {
+		return nil, err
+	}
+
 	paramPtr := C.CString(param)
 	defer C.free(unsafe.Pointer(paramPtr))
 
@@ -27,7 +37,100 @@ func NewTokenizer(param string) (tokenizerapi.Tokenizer, error) {
 	return NewCTokenizer(ptr), nil
 }
 
+func CheckAndFillParams(params string) (string, error) {
+	if len(params) == 0 {
+		return "", nil
+	}
+
+	var paramMaps map[string]any
+	flag := false
+	err := json.Unmarshal([]byte(params), &paramMaps)
+	if err != nil {
+		return "", merr.WrapErrAsInputError(fmt.Errorf("unmarshal analyzer params failed with json error: %s", err.Error()))
+	}
+
+	tokenizer, ok := paramMaps["tokenizer"]
+	if !ok {
+		// skip check if no tokenizer params
+		return params, nil
+	}
+
+	switch value := tokenizer.(type) {
+	case string:
+		// return if use build-in tokenizer
+		return params, nil
+	case map[string]any:
+		flag, err = CheckAndFillTokenizerParams(value)
+		if err != nil {
+			return "", err
+		}
+	default:
+		return "", merr.WrapErrAsInputError(fmt.Errorf("analyzer params set tokenizer with unknown type"))
+	}
+
+	// remarshal json params if params map was changed.
+	if flag {
+		bytes, err := json.Marshal(paramMaps)
+		if err != nil {
+			return "", merr.WrapErrAsInputError(fmt.Errorf("marshal analyzer params failed with json error: %s", err.Error()))
+		}
+		return string(bytes), nil
+	}
+	return params, nil
+}
+
+// fill some milvus params to tokenizer params
+func CheckAndFillTokenizerParams(params map[string]any) (bool, error) {
+	v, ok := params["type"]
+	if !ok {
+		return false, merr.WrapErrAsInputError(fmt.Errorf("costom tokenizer must set type"))
+	}
+
+	tokenizerType, ok := v.(string)
+	if !ok {
+		return false, merr.WrapErrAsInputError(fmt.Errorf("costom tokenizer type must be string"))
+	}
+
+	switch tokenizerType {
+	case "lindera":
+		cfg := paramtable.Get()
+
+		if _, ok := params["dict_build_dir"]; ok {
+			return false, merr.WrapErrAsInputError(fmt.Errorf("costom tokenizer dict_build_dir was system params, should not be set"))
+		}
+		// build lindera to LocalResourcePath/lindera/dict_kind
+		params["dict_build_dir"] = path.Join(cfg.FunctionCfg.LocalResourcePath.GetValue(), "lindera")
+
+		v, ok := params["dict_kind"]
+		if !ok {
+			return false, merr.WrapErrAsInputError(fmt.Errorf("lindera tokenizer must set dict_kind"))
+		}
+		dictKind, ok := v.(string)
+		if !ok {
+			return false, merr.WrapErrAsInputError(fmt.Errorf("lindera tokenizer dict kind must be string"))
+		}
+		dictUrlsMap := cfg.FunctionCfg.LinderaDownloadUrls.GetValue()
+
+		if _, ok := params["download_urls"]; ok {
+			return false, merr.WrapErrAsInputError(fmt.Errorf("costom tokenizer download_urls was system params, should not be set"))
+		}
+
+		if value, ok := dictUrlsMap["."+dictKind]; ok {
+			// use download urls set in milvus yaml
+			params["download_urls"] = paramtable.ParseAsStings(value)
+		}
+		return true, nil
+	default:
+		return false, nil
+	}
+}
+
 func ValidateTokenizer(param string) error {
+	param, err := CheckAndFillParams(param)
+	if err != nil {
+		return err
+	}
+
 	paramPtr := C.CString(param)
 	defer C.free(unsafe.Pointer(paramPtr))
 

internal/util/ctokenizer/c_tokenizer_test.go

@@ -5,14 +5,19 @@ import (
 	"testing"
 
 	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/milvus-io/milvus/pkg/v2/util/paramtable"
 )
 
 func TestTokenizer(t *testing.T) {
+	paramtable.Init()
+
 	// default tokenizer.
 	{
 		m := "{\"tokenizer\": \"standard\"}"
 		tokenizer, err := NewTokenizer(m)
-		assert.NoError(t, err)
+		require.NoError(t, err)
 		defer tokenizer.Destroy()
 
 		tokenStream := tokenizer.NewTokenStream("football, basketball, pingpang")
@@ -26,7 +31,7 @@ func TestTokenizer(t *testing.T) {
 	{
 		m := "{\"tokenizer\": \"jieba\"}"
 		tokenizer, err := NewTokenizer(m)
-		assert.NoError(t, err)
+		require.NoError(t, err)
 		defer tokenizer.Destroy()
 
 		tokenStream := tokenizer.NewTokenStream("张华考上了北京大学；李萍进了中等技术学校；我在百货公司当售货员：我们都有光明的前途")
@@ -35,6 +40,20 @@ func TestTokenizer(t *testing.T) {
 			fmt.Println(tokenStream.Token())
 		}
 	}
+
+	// lindera tokenizer.
+	{
+		m := "{\"tokenizer\": {\"type\":\"lindera\", \"dict_kind\": \"ipadic\"}}"
+		tokenizer, err := NewTokenizer(m)
+		require.NoError(t, err)
+		defer tokenizer.Destroy()
+
+		tokenStream := tokenizer.NewTokenStream("東京スカイツリーの最寄り駅はとうきょうスカイツリー駅です")
+		defer tokenStream.Destroy()
+		for tokenStream.Advance() {
+			fmt.Println(tokenStream.Token())
+		}
+	}
 }
 
 func TestValidateTokenizer(t *testing.T) {
@@ -58,3 +77,71 @@ func TestValidateTokenizer(t *testing.T) {
 		assert.Error(t, err)
 	}
 }
+
+func TestCheckAndFillParams(t *testing.T) {
+	paramtable.Init()
+	paramtable.Get().SaveGroup(map[string]string{"function.analyzer.lindera.download_urls.ipadic": "/test/url"})
+
+	// normal case
+	{
+		m := "{\"tokenizer\": {\"type\":\"jieba\"}}"
+		_, err := CheckAndFillParams(m)
+		assert.NoError(t, err)
+	}
+
+	// fill lindera tokenizer download urls and dict local path
+	{
+		m := "{\"tokenizer\": {\"type\":\"lindera\", \"dict_kind\": \"ipadic\"}}"
+		_, err := CheckAndFillParams(m)
+		assert.NoError(t, err)
+	}
+
+	// error with wrong json
+	{
+		m := "{invalid json"
+		_, err := CheckAndFillParams(m)
+		assert.Error(t, err)
+	}
+
+	// skip if use default analyzer
+	{
+		m := "{}"
+		_, err := CheckAndFillParams(m)
+		assert.NoError(t, err)
+	}
+
+	// error tokenizer without type
+	{
+		m := "{\"tokenizer\": {\"dict_kind\": \"ipadic\"}}"
+		_, err := CheckAndFillParams(m)
+		assert.Error(t, err)
+	}
+
+	// error tokenizer type not string
+	{
+		m := "{\"tokenizer\": {\"type\": 1, \"dict_kind\": \"ipadic\"}}"
+		_, err := CheckAndFillParams(m)
+		assert.Error(t, err)
+	}
+
+	// error tokenizer params type
+	{
+		m := "{\"tokenizer\": 1}"
+		_, err := CheckAndFillParams(m)
+		assert.Error(t, err)
+	}
+
+	// error set dict_build_dir by user
+	{
+		m := "{\"tokenizer\": {\"type\": \"lindera\", \"dict_kind\": \"ipadic\", \"dict_build_dir\": \"/tmp/milvus\"}}"
+		_, err := CheckAndFillParams(m)
+		assert.Error(t, err)
+	}
+
+	// error lindera kind not set
+	{
+		m := "{\"tokenizer\": {\"type\": \"lindera\"}}"
+		_, err := CheckAndFillParams(m)
+		assert.Error(t, err)
+	}
+}

pkg/util/paramtable/function_param.go

@@ -23,6 +23,8 @@ import (
 type functionConfig struct {
 	TextEmbeddingProviders ParamGroup `refreshable:"true"`
 	RerankModelProviders   ParamGroup `refreshable:"true"`
+	LocalResourcePath      ParamItem  `refreshable:"true"`
+	LinderaDownloadUrls    ParamGroup `refreshable:"true"`
 }
 
 func (p *functionConfig) init(base *BaseTable) {
@@ -91,6 +93,20 @@ func (p *functionConfig) init(base *BaseTable) {
 		},
 	}
 	p.RerankModelProviders.Init(base.mgr)
+
+	p.LocalResourcePath = ParamItem{
+		Key:          "function.analyzer.local_resource_path",
+		Version:      "2.5.16",
+		Export:       true,
+		DefaultValue: "/var/lib/milvus/analyzer",
+	}
+	p.LocalResourcePath.Init(base.mgr)
+
+	p.LinderaDownloadUrls = ParamGroup{
+		KeyPrefix: "function.analyzer.lindera.download_urls",
+		Version:   "2.5.16",
+	}
+	p.LinderaDownloadUrls.Init(base.mgr)
 }
 
 const (

tests/go_client/testcases/query_test.go

@@ -1218,7 +1218,7 @@ func TestRunAnalyzer(t *testing.T) {
 
 	// run analyzer with invalid params
 	_, err = mc.RunAnalyzer(ctx, client.NewRunAnalyzerOption("text doc").WithAnalyzerParamsStr("invalid params}"))
-	common.CheckErr(t, err, false, "JsonError")
+	common.CheckErr(t, err, false, "json error")
 
 	// run analyzer with custom analyzer
 	tokens, err = mc.RunAnalyzer(ctx, client.NewRunAnalyzerOption("test doc").