diff --git a/Cargo.lock b/Cargo.lock
index b9bd93825..2009d9658 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -752,6 +752,7 @@ dependencies = [
  "tokio-util",
  "tracing",
  "tracing-subscriber",
+ "unicode-segmentation",
  "uuid",
  "validator",
  "workspace-template",
diff --git a/services/appflowy-collaborate/Cargo.toml b/services/appflowy-collaborate/Cargo.toml
index 2e3151a8f..070547f36 100644
--- a/services/appflowy-collaborate/Cargo.toml
+++ b/services/appflowy-collaborate/Cargo.toml
@@ -87,6 +87,7 @@ lazy_static = "1.4.0"
 itertools = "0.12.0"
 validator = "0.16.1"
 rayon.workspace = true
+unicode-segmentation = "1.9.0"
 
 [dev-dependencies]
 rand = "0.8.5"
diff --git a/services/appflowy-collaborate/src/indexer/document_indexer.rs b/services/appflowy-collaborate/src/indexer/document_indexer.rs
index 7cd42abb3..6dd25c3a1 100644
--- a/services/appflowy-collaborate/src/indexer/document_indexer.rs
+++ b/services/appflowy-collaborate/src/indexer/document_indexer.rs
@@ -13,6 +13,7 @@ use collab_document::document::DocumentBody;
 use collab_document::error::DocumentError;
 use collab_entity::CollabType;
 use database_entity::dto::{AFCollabEmbeddingParams, AFCollabEmbeddings, EmbeddingContentType};
+use unicode_segmentation::UnicodeSegmentation;
 use uuid::Uuid;
 
 use crate::indexer::{DocumentDataExt, Indexer};
@@ -45,42 +46,12 @@ impl Indexer for DocumentIndexer {
     match result {
       Ok(document_data) => {
         let content = document_data.to_plain_text();
-        let mut result = Vec::with_capacity(1 + content.len() / Self::DOC_CONTENT_SPLIT);
-
-        let mut slice = content.as_str();
-        while slice.len() > Self::DOC_CONTENT_SPLIT {
-          // we should split document into multiple fragments
-          let (left, right) = slice.split_at(Self::DOC_CONTENT_SPLIT);
-          let param = AFCollabEmbeddingParams {
-            fragment_id: Uuid::new_v4().to_string(),
-            object_id: object_id.clone(),
-            collab_type: CollabType::Document,
-            content_type: EmbeddingContentType::PlainText,
-            content: left.to_string(),
-            embedding: None,
-          };
-          result.push(param);
-          slice = right;
-        }
-
-        let content = if slice.len() == content.len() {
-          content // we didn't slice the content
-        } else {
-          slice.to_string()
-        };
-        if !content.is_empty() {
-          let param = AFCollabEmbeddingParams {
-            fragment_id: object_id.clone(),
-            object_id: object_id.clone(),
-            collab_type: CollabType::Document,
-            content_type: EmbeddingContentType::PlainText,
-            content,
-            embedding: None,
-          };
-          result.push(param);
-        }
-
-        Ok(result)
+        create_embedding_params(
+          object_id,
+          content,
+          CollabType::Document,
+          Self::DOC_CONTENT_SPLIT,
+        )
       },
       Err(err) => {
         if matches!(err, DocumentError::NoRequiredData) {
@@ -141,3 +112,335 @@ impl Indexer for DocumentIndexer {
     }))
   }
 }
+#[inline]
+fn create_embedding_params(
+  object_id: String,
+  content: String,
+  collab_type: CollabType,
+  max_content_len: usize,
+) -> Result<Vec<AFCollabEmbeddingParams>, AppError> {
+  if content.is_empty() {
+    return Ok(vec![]);
+  }
+
+  // Helper function to create AFCollabEmbeddingParams
+  fn create_param(
+    fragment_id: String,
+    object_id: &str,
+    collab_type: &CollabType,
+    content: String,
+  ) -> AFCollabEmbeddingParams {
+    AFCollabEmbeddingParams {
+      fragment_id,
+      object_id: object_id.to_string(),
+      collab_type: collab_type.clone(),
+      content_type: EmbeddingContentType::PlainText,
+      content,
+      embedding: None,
+    }
+  }
+
+  if content.len() <= max_content_len {
+    // Content is short enough; return as a single fragment
+    let param = create_param(object_id.clone(), &object_id, &collab_type, content);
+    return Ok(vec![param]);
+  }
+
+  // Content is longer than max_content_len; need to split
+  let mut result = Vec::with_capacity(1 + content.len() / max_content_len);
+  let mut fragment = String::with_capacity(max_content_len);
+  let mut current_len = 0;
+
+  for grapheme in content.graphemes(true) {
+    let grapheme_len = grapheme.len();
+    if current_len + grapheme_len > max_content_len {
+      if !fragment.is_empty() {
+        // Move the fragment to avoid cloning
+        result.push(create_param(
+          Uuid::new_v4().to_string(),
+          &object_id,
+          &collab_type,
+          std::mem::take(&mut fragment),
+        ));
+      }
+      current_len = 0;
+
+3     // Check if the grapheme itself is longer than max_content_len
+      if grapheme_len > max_content_len {
+        // Push the grapheme as a fragment on its own
+        result.push(create_param(
+          Uuid::new_v4().to_string(),
+          &object_id,
+          &collab_type,
+          grapheme.to_string(),
+        ));
+        continue;
+      }
+    }
+    fragment.push_str(grapheme);
+    current_len += grapheme_len;
+  }
+
+  // Add the last fragment if it's not empty
+  if !fragment.is_empty() {
+    result.push(create_param(
+      object_id.clone(),
+      &object_id,
+      &collab_type,
+      fragment,
+    ));
+  }
+
+  Ok(result)
+}
+#[cfg(test)]
+mod tests {
+  use crate::indexer::document_indexer::create_embedding_params;
+  use collab_entity::CollabType;
+
+  #[test]
+  fn test_split_at_non_utf8() {
+    let object_id = "test_object".to_string();
+    let collab_type = CollabType::Document;
+    let max_content_len = 10; // Small number for testing
+
+    // Content with multibyte characters (emojis)
+    let content = "Hello 😃 World 🌍! This is a test 🚀.".to_string();
+
+    let params = create_embedding_params(
+      object_id.clone(),
+      content.clone(),
+      collab_type.clone(),
+      max_content_len,
+    )
+    .unwrap();
+
+    // Ensure that we didn't split in the middle of a multibyte character
+    for param in params {
+      assert!(param.content.is_char_boundary(0));
+      assert!(param.content.is_char_boundary(param.content.len()));
+    }
+  }
+
+  #[test]
+  fn test_exact_boundary_split() {
+    let object_id = "test_object".to_string();
+    let collab_type = CollabType::Document;
+    let max_content_len = 5; // Set to 5 for testing
+
+    // Content length is exactly a multiple of max_content_len
+    let content = "abcdefghij".to_string(); // 10 characters
+
+    let params = create_embedding_params(
+      object_id.clone(),
+      content.clone(),
+      collab_type.clone(),
+      max_content_len,
+    )
+    .unwrap();
+
+    assert_eq!(params.len(), 2);
+    assert_eq!(params[0].content, "abcde");
+    assert_eq!(params[1].content, "fghij");
+  }
+
+  #[test]
+  fn test_content_shorter_than_max_len() {
+    let object_id = "test_object".to_string();
+    let collab_type = CollabType::Document;
+    let max_content_len = 100;
+
+    let content = "Short content".to_string();
+
+    let params = create_embedding_params(
+      object_id.clone(),
+      content.clone(),
+      collab_type.clone(),
+      max_content_len,
+    )
+    .unwrap();
+
+    assert_eq!(params.len(), 1);
+    assert_eq!(params[0].content, content);
+  }
+
+  #[test]
+  fn test_empty_content() {
+    let object_id = "test_object".to_string();
+    let collab_type = CollabType::Document;
+    let max_content_len = 10;
+
+    let content = "".to_string();
+
+    let params = create_embedding_params(
+      object_id.clone(),
+      content.clone(),
+      collab_type.clone(),
+      max_content_len,
+    )
+    .unwrap();
+
+    assert_eq!(params.len(), 0);
+  }
+
+  #[test]
+  fn test_content_with_only_multibyte_characters() {
+    let object_id = "test_object".to_string();
+    let collab_type = CollabType::Document;
+    let max_content_len = 4; // Small number for testing
+
+    // Each emoji is 4 bytes in UTF-8
+    let content = "😀😃😄😁😆".to_string();
+
+    let params = create_embedding_params(
+      object_id.clone(),
+      content.clone(),
+      collab_type.clone(),
+      max_content_len,
+    )
+    .unwrap();
+
+    assert_eq!(params.len(), 5);
+    let expected_contents = vec!["😀", "😃", "😄", "😁", "😆"];
+    for (param, expected) in params.iter().zip(expected_contents.iter()) {
+      assert_eq!(param.content, *expected);
+    }
+  }
+
+  #[test]
+  fn test_split_with_combining_characters() {
+    let object_id = "test_object".to_string();
+    let collab_type = CollabType::Document;
+    let max_content_len = 5; // Small number for testing
+
+    // String with combining characters (e.g., letters with accents)
+    let content = "a\u{0301}e\u{0301}i\u{0301}o\u{0301}u\u{0301}".to_string(); // "áéíóú"
+
+    let params = create_embedding_params(
+      object_id.clone(),
+      content.clone(),
+      collab_type.clone(),
+      max_content_len,
+    )
+    .unwrap();
+
+    assert_eq!(params.len(), 5);
+    let expected_contents = vec!["á", "é", "í", "ó", "ú"];
+    for (param, expected) in params.iter().zip(expected_contents.iter()) {
+      assert_eq!(param.content, *expected);
+    }
+  }
+
+  #[test]
+  fn test_large_content() {
+    let object_id = "test_object".to_string();
+    let collab_type = CollabType::Document;
+    let max_content_len = 1000;
+
+    // Generate a large content string
+    let content = "a".repeat(5000); // 5000 characters
+
+    let params = create_embedding_params(
+      object_id.clone(),
+      content.clone(),
+      collab_type.clone(),
+      max_content_len,
+    )
+    .unwrap();
+
+    assert_eq!(params.len(), 5); // 5000 / 1000 = 5
+    for param in params {
+      assert_eq!(param.content.len(), 1000);
+    }
+  }
+  #[test]
+  fn test_non_ascii_characters() {
+    let object_id = "test_object".to_string();
+    let collab_type = CollabType::Document;
+    let max_content_len = 5;
+
+    // Non-ASCII characters: "áéíóú"
+    let content = "áéíóú".to_string();
+
+    let params = create_embedding_params(
+      object_id.clone(),
+      content.clone(),
+      collab_type.clone(),
+      max_content_len,
+    )
+    .unwrap();
+
+    // Content should be split into two fragments
+    assert_eq!(params.len(), 3);
+    assert_eq!(params[0].content, "áé");
+    assert_eq!(params[1].content, "íó");
+    assert_eq!(params[2].content, "ú");
+  }
+
+  #[test]
+  fn test_content_with_leading_and_trailing_whitespace() {
+    let object_id = "test_object".to_string();
+    let collab_type = CollabType::Document;
+    let max_content_len = 5;
+
+    let content = "  abcde  ".to_string();
+
+    let params = create_embedding_params(
+      object_id.clone(),
+      content.clone(),
+      collab_type.clone(),
+      max_content_len,
+    )
+    .unwrap();
+
+    // Content should include leading and trailing whitespace
+    assert_eq!(params.len(), 2);
+    assert_eq!(params[0].content, "  abc");
+    assert_eq!(params[1].content, "de  ");
+  }
+
+  #[test]
+  fn test_content_with_multiple_zero_width_joiners() {
+    let object_id = "test_object".to_string();
+    let collab_type = CollabType::Document;
+    let max_content_len = 10;
+
+    // Complex emoji sequence with multiple zero-width joiners
+    let content = "👩‍👩‍👧‍👧👨‍👨‍👦‍👦".to_string();
+
+    let params = create_embedding_params(
+      object_id.clone(),
+      content.clone(),
+      collab_type.clone(),
+      max_content_len,
+    )
+    .unwrap();
+
+    // Each complex emoji should be treated as a single grapheme
+    assert_eq!(params.len(), 2);
+    assert_eq!(params[0].content, "👩‍👩‍👧‍👧");
+    assert_eq!(params[1].content, "👨‍👨‍👦‍👦");
+  }
+
+  #[test]
+  fn test_content_with_long_combining_sequences() {
+    let object_id = "test_object".to_string();
+    let collab_type = CollabType::Document;
+    let max_content_len = 5;
+
+    // Character with multiple combining marks
+    let content = "a\u{0300}\u{0301}\u{0302}\u{0303}\u{0304}".to_string(); // a with multiple accents
+
+    let params = create_embedding_params(
+      object_id.clone(),
+      content.clone(),
+      collab_type.clone(),
+      max_content_len,
+    )
+    .unwrap();
+
+    // The entire combining sequence should be in one fragment
+    assert_eq!(params.len(), 1);
+    assert_eq!(params[0].content, content);
+  }
+}