diff --git a/operator/text/v0/markdown_splitter.go b/operator/text/v0/markdown_splitter.go
index 50522799..eb3801db 100644
--- a/operator/text/v0/markdown_splitter.go
+++ b/operator/text/v0/markdown_splitter.go
@@ -68,11 +68,8 @@ func (sp MarkdownTextSplitter) buildDocuments(rawRunes []rune) []MarkdownDocumen
 
 	documents := []MarkdownDocument{}
 
-	startPosition := 0
-	document, startPosition := sp.buildDocument(rawRunes, MarkdownDocument{}, startPosition)
-
-	documents = append(documents, document)
-
+	var startPosition int
+	var document MarkdownDocument
 	for startPosition < len(rawRunes) {
 		document, startPosition = sp.buildDocument(rawRunes, document, startPosition)
 
diff --git a/operator/text/v0/markdown_splitter_test.go b/operator/text/v0/markdown_splitter_test.go
new file mode 100644
index 00000000..6e08f6ea
--- /dev/null
+++ b/operator/text/v0/markdown_splitter_test.go
@@ -0,0 +1,74 @@
+package text
+
+import (
+	"testing"
+
+	"github.com/frankban/quicktest"
+	"github.com/tmc/langchaingo/textsplitter"
+)
+
+func Test_MarkdownSplitter(t *testing.T) {
+
+	c := quicktest.New(t)
+
+	testCases := []struct {
+		input     ChunkTextInput
+		outputLen int
+	}{
+		{
+			input: ChunkTextInput{
+				Text: `# asf65463
+	## 654654
+	fasdflj`,
+				Strategy: Strategy{
+					Setting: Setting{
+						ChunkMethod:  "Markdown",
+						ChunkSize:    800,
+						ChunkOverlap: 200,
+					},
+				},
+			},
+			outputLen: 1,
+		},
+		{
+			input: ChunkTextInput{
+				Text: `# 醫囑
+檢驗 : Urine routine(急) [尿液] [有蓋定量離心管(尿液收集管)] STAT 【註:Foley】
+=> **尿液檢查採樣來源為Foley，表示有裝置導尿管**
+
+# 護理
+病人2way尿管存，管路引流順暢，尿液呈淡黃色，管路固定於右大腿，無滑脫，續觀
+=> **尿管存，表示有裝置導尿管**
+
+# 個案泌尿道感染判定
+1. 有導尿管
+2. 病患無UTI感染症狀
+=> **判斷為:非泌尿道感染，僅無症狀菌尿症(not UTI; asymptomatic bacteuria only)**`,
+				Strategy: Strategy{
+					Setting: Setting{
+						ChunkMethod:  "Markdown",
+						ChunkSize:    800,
+						ChunkOverlap: 200,
+					},
+				},
+			},
+			outputLen: 3,
+		},
+	}
+	for _, testCase := range testCases {
+		c.Run("Test bug cases reported", func(c *quicktest.C) {
+			inputStruct := testCase.input
+			setting := inputStruct.Strategy.Setting
+			split := NewMarkdownTextSplitter(
+				textsplitter.WithChunkSize(setting.ChunkSize),
+				textsplitter.WithChunkOverlap(setting.ChunkOverlap),
+			)
+
+			chunks, err := split.SplitText(inputStruct.Text)
+
+			c.Assert(err, quicktest.IsNil)
+
+			c.Assert(len(chunks), quicktest.DeepEquals, testCase.outputLen)
+		})
+	}
+}